title
stringlengths 1
827
⌀ | uuid
stringlengths 36
36
| pmc_id
stringlengths 5
8
| search_term
stringclasses 18
values | text
stringlengths 0
8.42M
|
|---|---|---|---|---|
Obstetrician-gynecologist perceptions and utilization of prescription drug monitoring programs | 36756204-ac94-4554-a05c-14c267b0fe44 | 7793317 | Gynaecology[mh] | Introduction The US Centers for Disease Control and Prevention (CDC) issued guidelines in 2016 recommending that clinicians review their state Prescription Drug Monitoring Program (PDMP) data when initiating and/or continuing opioid therapies under certain clinical circumstances. PDMPs provide opioid and other controlled substance dispensing histories and other measures to clinicians for patients in their care. The American College of Obstetricians and Gynecologists (ACOG) and the American Society of Addiction Medicine (ASAM) jointly released a committee opinion to clarify recommendations for obstetrician-gynecologists (OB/GYN) that treat patients who are prescribed or may use opioids during pregnancy, medically or non-medically, following the release of the CDC guidelines. The ACOG-ASAM recommendations endorse OB/GYN usage of PDMPs as a primary prevention tool for opioid-related adverse events. As of mid-2020, most states now a) mandate that all controlled substance prescribers register with their state PDMP and b) require all or certain prescribers to check the PDMP when initiating controlled substance prescriptions, particularly for US Drug Enforcement Agency (DEA) Schedule II opioids. Physician use of PDMPs increases when administrative registration with the state is mandated, and prescribers reportedly comply with PDMP usage mandates. However, prescribers across multiple specialties report that stand-alone PDMP data is difficult to access and incorporate into their workflow. For OB/GYNs in particular, PDMPs are viewed as less effective, positive, or useful when compared to other primary care physicians. In this literature, OB/GYNs sample sizes are low and they have sometimes been categorized with “other” prescriber specialties making it difficult to understand their nuanced PDMP use and perceptions. One study in Washington Medicaid reported that OB/GYNs had the second lowest uptake in both PDMP registration and usage when compared with other physician specialties. Since OB/GYNs are the primary source of care for many women and comprise the majority of care during pregnancy; they are well-positioned to provide screening and intervention for opioid-related sequelae. The purpose of this study was to assess OB/GYN utilization and perceptions of their state PDMP as stratified by practice location in states with and without mandated PDMP query. Methods 2.1 Instrument development A workgroup consisting of an OB/GYN, a pharmacist, and health services researchers reviewed survey items from several publicly available state-level PDMP survey instruments. Survey items from previously published instruments were adapted for OB/GYNs to assess the perception of PDMP effectiveness, knowledge of PDMP functions, and self-reported use of PDMPs. The survey instrument was reviewed and approved by the ACOG District XII Committee on Health Care for Underserved Women prior to release and is available in Supplementary Materials. 2.2 Study design and protocol The study design was a cross-sectional survey. The research team partnered with ACOG leadership, who oversaw dissemination of the survey link and accompanying study description and explanation via email to a random sample of 5000 ACOG members with an active license to practice in the United States in May 2018. A reminder email was sent each week following the initial email invitation for a period of 6 weeks and the survey link remained active for a period of one week following the final reminder in July 2018. Survey responses were anonymous, but email read receipt data from the invitation were collected to calculate an adjusted response rate. Data were collected in Qualtrics (Qualtrics, Provo, Utah, USA). The University of Florida Institutional Review Board reviewed and approved this study. 2.3 Analysis Response frequencies were calculated for each item and all surveys with >1 item response were included in the analysis (n = 397). State regulatory environment was classified as “mandatory” or “voluntary” based on the legal requirements for PDMP query (as of July 2018) and the physicians primary practice location. Chi square analysis was used to compare differences in response distribution between respondents practicing in mandatory versus voluntary PDMP states. A priori significance was set at 0.05. Qualitative and free-text survey items were analyzed and coded for instances of similar thematic content by 3 reviewers, and, in instances of disagreement, our OB/GYN acted as a fourth and deciding vote. All analyses were conducted in Excel and SAS 9.4 (SAS Institute Inc., Cary, North Carolina, USA). Instrument development A workgroup consisting of an OB/GYN, a pharmacist, and health services researchers reviewed survey items from several publicly available state-level PDMP survey instruments. Survey items from previously published instruments were adapted for OB/GYNs to assess the perception of PDMP effectiveness, knowledge of PDMP functions, and self-reported use of PDMPs. The survey instrument was reviewed and approved by the ACOG District XII Committee on Health Care for Underserved Women prior to release and is available in Supplementary Materials. Study design and protocol The study design was a cross-sectional survey. The research team partnered with ACOG leadership, who oversaw dissemination of the survey link and accompanying study description and explanation via email to a random sample of 5000 ACOG members with an active license to practice in the United States in May 2018. A reminder email was sent each week following the initial email invitation for a period of 6 weeks and the survey link remained active for a period of one week following the final reminder in July 2018. Survey responses were anonymous, but email read receipt data from the invitation were collected to calculate an adjusted response rate. Data were collected in Qualtrics (Qualtrics, Provo, Utah, USA). The University of Florida Institutional Review Board reviewed and approved this study. Analysis Response frequencies were calculated for each item and all surveys with >1 item response were included in the analysis (n = 397). State regulatory environment was classified as “mandatory” or “voluntary” based on the legal requirements for PDMP query (as of July 2018) and the physicians primary practice location. Chi square analysis was used to compare differences in response distribution between respondents practicing in mandatory versus voluntary PDMP states. A priori significance was set at 0.05. Qualitative and free-text survey items were analyzed and coded for instances of similar thematic content by 3 reviewers, and, in instances of disagreement, our OB/GYN acted as a fourth and deciding vote. All analyses were conducted in Excel and SAS 9.4 (SAS Institute Inc., Cary, North Carolina, USA). Results A total of n = 1470 survey invitations were opened and read, resulting in an adjusted response rate of 27% (n = 397 surveys completed). About a third of respondents were in private practice settings, and few were still considered trainees (60.7% classified as Attending). Most respondents practiced in a mandatory PDMP state (80.6%), 9.6% practiced in voluntary PDMP states, and 9.8% did not indicate their practice location. The majority were currently registered with the PDMP (77.6%). To gauge OB/GYN familiarity and understanding of PDMP data, respondents were asked to identify what information is provided by the PDMP from a list of options. Approximately, 30% were unaware that the PDMP identifies the prescriber writing each prescription and nearly half of respondents were unaware that the PDMP identifies dispensing pharmacies. A summary of other respondent characteristics is shown in Table . Those practicing in mandatory versus voluntary states perceived the primary purpose of PDMPs differently (Table ) and the majority of respondents suspected that 0 to 10% of their patients misuse or abuse opioids (Fig. ). In free-text responses regarding the primary purpose of PDMPs, a majority of respondents that selected “other” purpose expressed frustration with PDMP usage and/or mandatory use laws (n = 14, Table ). Three content themes of PDMP purpose emerged from these free-text responses: 1. Increase in physician burden [sample response: “To burden physicians with police work”], 2. Skepticism of government involvement [sample response: “Government bull [expletive]”], and 3. Oversight of prescriber activity [sample response: “So that state government and legislators can say they are doing something about the “opioid crisis””]. Respondents report most frequently querying the PDMP for patients that are currently using or prescribed opioids, and when they treat patients suspected of drug abuse (Fig. ). Respondents most frequently report taking action as a result of using the PDMP by confirming prescription fills (31.3% in mandatory states; 23.7% in voluntary states), followed by speaking with patients about controlled substance use (27.8% mandatory states; 26.3% voluntary states). About 1 in 5 respondents indicated they confirmed doctor shopping behaviors as a result of querying the PDMP. No respondents reported referring patients to law enforcement (0%) and Child Protective Services referrals were also rare (1.9% in mandatory states; 0.0% in voluntary states; Table ). Overall, 53% of OB/GYNs agreed that “…mandating prescriber use of the PDMP was a good idea.” A greater proportion (58.3%) of respondents practicing in voluntary states agreed or strongly agreed with this statement (Fig. ). Discussion Our study is the largest to-date on OB/GYN perceptions and use of their state PDMPs, and is among the first to assess perception of opioid use among the patients in their care. These findings suggest that OB/GYN perceptions may be tied to experience with the PDMP as evidenced by a significantly different stated purpose of the PDMP when examined by practice legal environment. The skepticism expressed by many respondents regarding PDMP effectiveness as a primary prevention tool for several opioid-related sequelae is concerning, despite recommendations. The findings regarding PDMP utility as a primary prevention tool were documented in a separate report analyzing these same data. A recent survey of ACOG Fellows and Junior Fellows reported that most OB/GYN respondents continue to prescribe opioids for a variety of indications, but few reported adherence to opioid prescribing guidelines. In that ACOG survey, 81% of respondents also reported that they were unaware that the primary source of diverted opioids were prescriptions from friends and family members. 4.1 Clinical and research implications Many states have recently adopted legislation to restrict opioid prescribing and dispensing by limiting quantities of outpatient prescriptions of opioids for acute pain and several other states have similar legislation under consideration. Additionally, federal legislation has been proposed to limit new opioid prescriptions for acute pain conditions to a 7-day supply. These changes in the medico-legal landscape suggest that all prescribers, including OB/GYNs, will be checking PDMPs more frequently. Of particular importance for OB/GYN clinical practice, pregnancy may be the only time a woman with opioid use disorder or other forms of SUD engage in medical treatment, which suggests that OB/GYNs are optimally positioned for screenings and interventions. The delegate model, whereby a prescriber assigns responsibility for logging in and obtaining reports to another qualified health professional, for PDMP usage has been demonstrated to be more cost-effective than prescriber-initiated PDMP query and could reduce time and resource burden for OB/GYNs. As of 2020, all states (with the exception of Missouri, which is the only state that has not yet implemented a statewide PDMP) permit prescriber delegates to access the PDMP. After resolving workflow issues regarding PDMP access, however, there is evidence to suggest that physicians are uncertain about how and when to discuss information gleaned from PDMPs with their patients. This uncertainty may contribute to decreased perceptions of PDMP utility. 4.2 Strengths and limitations This study employed evidence-based practices for maximizing physician response rates, including the use of multiple, timely follow-up invitations, as well as delivery of the invitation via a trusted professional association (here, ACOG). Despite these efforts, the response rate to this survey is in line with typical response rates for web-based surveys to physicians that do not include financial incentives. An additional limitation is that we were reliant on self-reported measures of OB/GYN PDMP usage and were unable to compare these self-reports with patterns of actual PDMP use. Clinical and research implications Many states have recently adopted legislation to restrict opioid prescribing and dispensing by limiting quantities of outpatient prescriptions of opioids for acute pain and several other states have similar legislation under consideration. Additionally, federal legislation has been proposed to limit new opioid prescriptions for acute pain conditions to a 7-day supply. These changes in the medico-legal landscape suggest that all prescribers, including OB/GYNs, will be checking PDMPs more frequently. Of particular importance for OB/GYN clinical practice, pregnancy may be the only time a woman with opioid use disorder or other forms of SUD engage in medical treatment, which suggests that OB/GYNs are optimally positioned for screenings and interventions. The delegate model, whereby a prescriber assigns responsibility for logging in and obtaining reports to another qualified health professional, for PDMP usage has been demonstrated to be more cost-effective than prescriber-initiated PDMP query and could reduce time and resource burden for OB/GYNs. As of 2020, all states (with the exception of Missouri, which is the only state that has not yet implemented a statewide PDMP) permit prescriber delegates to access the PDMP. After resolving workflow issues regarding PDMP access, however, there is evidence to suggest that physicians are uncertain about how and when to discuss information gleaned from PDMPs with their patients. This uncertainty may contribute to decreased perceptions of PDMP utility. Strengths and limitations This study employed evidence-based practices for maximizing physician response rates, including the use of multiple, timely follow-up invitations, as well as delivery of the invitation via a trusted professional association (here, ACOG). Despite these efforts, the response rate to this survey is in line with typical response rates for web-based surveys to physicians that do not include financial incentives. An additional limitation is that we were reliant on self-reported measures of OB/GYN PDMP usage and were unable to compare these self-reports with patterns of actual PDMP use. Conclusions ACOG members are diverse in their perceptions regarding the utility and purpose of PDMPs; though, the majority agree that PDMPs are a primary prevention tool for drug abuse and diversion. However, a knowledge translation gap may still exist- as only a third of OB/GYNs report checking the PDMP for their patients with opioid prescriptions. Increased training is needed regarding clinical utility of PDMPs along with practical guidance for incorporating the PDMP into OB/GYN practice. The authors wish to thank the leadership of the American College of Obstetricians and Gynecologists (ACOG) for disseminating the survey to ACOG members and for providing feedback on the survey instrument. Additionally, the authors would like to thank the members of ACOG District XII for providing comment on the preliminary findings. Preliminary findings were presented at the 2019 ACOG Annual Clinical and Scientific Meeting in Nashville, Tennessee. Conceptualization: Amie Goodin, Chris Delcher, Joshua Brown, Dikea Roussos-Ross. Data curation: Jungjun Bae. Formal analysis: Amie Goodin. Methodology: Amie Goodin, Chris Delcher, Joshua Brown, Dikea Roussos-Ross. Project administration: Amie Goodin. Supervision: Amie Goodin, Chris Delcher, Dikea Roussos-Ross. Visualization: Jungjun Bae. Writing – original draft: Amie Goodin. Writing – review & editing: Amie Goodin, Jungjun Bae, Chris Delcher, Joshua Brown, Dikea Roussos-Ross. Supplemental Digital Content |
Uncovering metabolic dysregulation in schizophrenia and cannabis use disorder through untargeted plasma lipidomics | 0be97d51-cde9-4285-b524-5ee1ffb5eb58 | 11682106 | Biochemistry[mh] | Schizophrenia (SZ) is a chronic, disabling condition that typically manifests early in life, affecting up to 1% of the global population over a lifetime. The disease is associated with significant mortality and morbidity, leading to a reduced life expectancy of 10 to 15 years and no cure is available yet. Schizophrenia ranks as the seventh most costly illness worldwide, due to its early onset, frequent hospitalizations, the necessity for psychosocial support, and the associated loss of productivity . While various hypotheses have been proposed regarding the aetiopathogenesis of SZ, none have been definitively validated. Although the symptomatic onset usually occurs in late adolescence or early adulthood, the disorder is rooted in genetic and/or environmental factors that are present long before symptoms emerge . Cannabis is among the most widely used substances globally, with an estimated 228 million users aged 15 to 64. The risk of developing schizophrenia is significantly heightened with cannabis abuse, particularly when use begins at a younger age . Additionally, approximately 10% of cannabis users are estimated to develop cannabis use disorder (CUD) over their lifetime. Interestingly, nearly one-third of individuals diagnosed with schizophrenia (SZ) have also been reported to meet the criteria for CUD . However, the biological mechanisms that determine why some individuals develop schizophrenia while others experience only CUD, despite similar levels of cannabis exposure, remain unclear. Regarding this, many metabolomic studies have tried to find specific biomarkers of schizophrenia for its early detection and disease progress monitoring. Different metabolic abnormalities have been found in patients with schizophrenia, but most studies highlight a different content of various amino acids and lipids compared to healthy subjects – . Similarly, cannabis users show distinctive metabolic profiles when compared to subjects that do not consume cannabis. In this sense, abnormal content of various lipids, amino acids, sugars, small proteins and other metabolites have been found in the plasma of cannabis users, suggesting that cannabis influences various systems and metabolic pathways besides the endocannabinoid system , . Within the afore-mentioned families of metabolites lipids can be highlighted due to their abundance in the brain and the key role of their metabolism for the functioning of central nervous system – . The most widely used and best characterized biological matrices for clinical metabolomic studies are blood components, serum and plasma . As blood is in contact with all organs and tissues, it allows monitoring of overall homeostatic changes in the body, and its collection is minimally invasive . Nonetheless, plasma is usually preferred to serum in the search for potential biomarkers, as serum sampling presents a broader source of variability that can hinder comparisons between samples . In this context, schizophrenia and cannabis use may be reflected in the lipid composition through their association with neuroinflammation, oxidative stress and altered endocannabinoid signaling. Cannabis, in particular, disrupts endocannabinoid pathways , , which are lipid-derived signaling mechanisms that exacerbate or modify the symptoms of schizophrenia. By analyzing lipid profiles in the blood of patients with schizophrenia and cannabis use, we may gain insights into how cannabis interacts with pre-existing metabolic and neurochemical disturbances in schizophrenia. Having such a background as a starting point, in this work, an untargeted lipidomic analysis of plasma samples from patients with schizophrenia, cannabis use disorder or comorbidity of both disorders was carried out.
Annotation of untargeted non-polar metabolites According to the workflow and data filtering described in the next section, a final list of 119 potential compounds in all the analyzed samples was obtained. Final candidates were annotated based on the endogenous metabolites mass lists (Human Metabolome Database and LipidMaps structure database) and MS2 mass spectra. The potential name of each candidate, grouped by compound classes, together with their predicted molecular formula, predicted monoisotopic mass and its error, the match level (with mzCloud spectra or with in silico fragmentation), retention time and the identification confidence level are shown in Supplementary Table . Lipids, predominantly fatty acids, glycerolipids, glycerolphospholipids and sphingolipids, represented the most detected class among all the potential candidates of the final list. Leaving aside 6 fatty acids that were annotated at level 2a due to high MS2 spectra matching and lack of ambiguity with other potential candidates, most of the identified lipids were annotated at level 3. The exogenous compounds (i.e. non-endogenous metabolites) are identified as “Others” in Table S2 under the “Categories/Subclass” column. They represent 19 of the 77 compounds included among the 2a/2b and 3 identification confidence levels. All the metabolites at level 3 (51% of candidates) showed acceptable spectral matches, but this is not enough to assure the discrimination between positional isomers or equivalent structures (e.g. the position of a double bond in unsaturated fatty acids), and therefore, only general structures and compound classes could be assured. Nevertheless, the putative name of the probable candidate (those with the higher number of citations) in metabolites annotated at level 3 was maintained. In contrast, candidates annotated at level 4 (35%), showed uninformative MS2 spectra and big ambiguity between other potential candidates, and thus, it was not even possible to determine the functional group or the class of these compounds. Consecutively, neither name nor family name was given to these compounds and only molecular formula was maintained. The remaining candidates (13%) were annotated at level 2a or 2b due to unambiguous MS2 spectra matching with experimental MS2 available in mzCloud spectra or in silico fragmentation in Compound Discoverer, respectively. Within this group of compounds, three unsaturated fatty acids, two acyl carnitines, one N-acyl amine and one nicotinamide were identified. In addition to all the endogenous compounds mentioned above, some exogenous compounds and metabolites were identified at the same identification level, which deserve special attention. On the one hand, piperine (M003), tricyclazole (M012) and 2-hydroxybenzothiazol (M013) were identified, which have been detected in urine in other suspect analysis research works in the framework of human exposome assessment . More importantly, active substances of drugs and pharmaceuticals and related metabolites have been found as well. In this sense, we can highlight, nicotine (M001) and cotinine (M002), as tobacco consumption biomarkers ; aripiprazole (M009) and dehydroaripiprazole (M010), as antipsychotic treatment biomarkers , and 11-nor-9-carboxy-∆ 9 -THC (M015), as cannabis consumption biomarker . Aripiprazole (Supplementary Fig. a) and dehydroaripiprazole (Supplementary Fig. b) were found in DUAL and SZ subjects but not in controls (for both metabolites and groups, log 2 (FC) = 5.5–8.8) or CUD subjects (for both metabolites and groups, log 2 (FC) = 2.9–3.1). Moreover, CUD and DUAL subjects had higher levels of nicotine and cotinine in plasma compared to controls (for both metabolites and groups, log 2 (FC) = 4.7–5.7) and to SZ patients (for both metabolites and groups log 2 (FC) = 1.2–2.1) (Supplementary Fig. ). This finding suggests that there is a moderate tobacco consumption among CUD subjects (probably related to cannabis consumption as well) and DUAL subjects, and low or none tobacco consumption among SZ patients and controls. Higher 11-nor-9-carboxy-∆ 9 -THC content was found in CUD subjects compared to the rest of groups (log 2 (FC) = 3.1 compared to DUAL, and, log 2 (FC) = 5.4–6 compared to SZ or controls) (Supplementary Fig. ). These observations confirm that SZ and DUAL patients were under antipsychotic treatment, as well as cannabis consumption among CUD patients, and none consumption of any of these substances among controls. Although these expected findings were interesting to verify the consumption and treatment status of the subjects, these discriminative exogenous metabolites (together with the mentioned exogenous compounds belonging to exposome) directly related with the specific characteristic of the studied population were not taken into account for further analysis in order to go beyond the exposome analysis and study the alterations in lipidomic pathway. Metabolic clustering of SZ, CUD and DUAL patients and their controls Before analyzing the changes in the metabolic profiles of the three groups of patients, one of the first issues to be addressed was whether or not the recruited controls to each cohort (CUD, DUAL and SZ) were equivalent. To test this underlying hypothesis, control samples from different groups were compared by one way ANOVA using MetaboAnalyst (Supplementary Table ). This analysis showed that 12 metabolites were significantly different in the three groups of control samples. Consequently, they were eliminated from all the data (together with the exogenous metabolites mentioned above) in order to avoid biased results. The metabolic profiles of each group of patients (CUD, DUAL and SZ) were then contrasted with those of the control group. Multivariate analysis using OPLS-DA was run along univariate analysis using FC and FDR t-tests. The score plots, permutation analysis results and the variable importance in projection (VIP) scores from the three OPLS-DA models are shown in Fig. . In addition, volcano plots obtained from univariate analyses are shown in Fig. . The score plots of the OPLS-DA models (Fig. a, d and g), show that the variability in the contents of the annotated metabolites allows metabolic clustering between control subjects and either CUD, DUAL or SZ patients. The cumulative R2Y and Q2 values of the OPLS-DA models were respectively, 0.92 and 0.85 for CUD vs. control; 0.91 and 0.87 for DUAL vs. control; and, 0.96 and 0.89 for SZ vs. control. Overall, around 40% of the variance was explained with the first two principal components. These values suggests that the OPLS-DA models fit the data correctly (R2Y near to 1) and that they have good predictive ability (Q2 > 0.5). Moreover, in the three OPLS-DA models, permuted RY2 values were around 0.6 or below, and most permuted Q2 values were below 0 (Fig. b, e and h). Also, permuted RY2 and Q2 values were in all cases below cumulative RY2 and Q2 values. All this suggests that the model fitting is adequate, and that was unlikely to be built by chance. After checking the suitability of the OPLS-DA models, metabolites that allowed discrimination between patient groups and controls were taken into account (Fig. c, f and i). To ensure a higher reliability of these results, among the potential biomarkers elucidated by the OPLS-DA analyses, only metabolites that showed significant changes according to volcano plots (Fig. ) were further considered. Nonetheless, most of the metabolites that allowed discrimination of groups in the OPLS-DA models (those with the highest VIP) were significant according to volcano plots (FC > 2, p < 0.05), and vice versa. Compared to controls, CUD subjects (Fig. a) presented a significant increase in the levels of two metabolites identified at level 4 (M094 and M096), and a significant decrease in the content of two acylcarnitines (M007 and M008), two N-acyl amino acids (M016 and M023), one primary amide (M026), one acetate ester (M070) and other nine metabolites (M078, M081, M083, M085, M093, M095, M097, M099 and M103). DUAL (Fig. b) and SZ (Fig. c) patients, in contrast to CUD patients, presented a higher degree of metabolic impairments compared to controls. DUAL patients showed increased plasma levels of one oxo fatty acid (M019), four ceramides (M052, M053, M054 and M058), one sterol lipid (M062) and other four metabolites (M077, M094, M096 and M119); and decreased levels of two acylcarnitines (M007 and M008), one primary amide (M026), three glycerolphospholipids (M037, M040 and M043), one phenolic acid (M064) and other 11 metabolites (M078, M079, M081, M083, M085, M087, M093, M097, M103, M105 and M108). Similarly, SZ patients showed increased plasma levels of two ceramides (M044 and M053), one sterol lipid (M062) and other four metabolites (M090, M096, M107 and M115), along with decreased levels of two acylcarnitines (M007 and M008), one nicotinamide (M011), one N-acyl amino acid (M023), one primary amide (M026), seven glycerolphospholipids (M034, M035, M036, M038, M039, M040 and M043), one phenolic acid (M064), one dipeptide (M066), one acetate ester (M070) and other twelve metabolites (M078, M079, M081, M083, M085, M087, M095, M097, M099, M102, M103 and M108). These analyses showed that either CUD, DUAL or SZ patients had distinctive metabolic profiles compared to healthy subjects, and pointed out a few potential biomarkers, some of them shared by two patient groups or more. It should be noted though, that some metabolites that underwent significant changes were among the metabolites identified at level 4. Indeed, a few of them (i.e., M081, M083, M085, M094 and M096) showed some of the highest scores (or highest FCs); being key compounds to provide metabolic clustering between patient groups and controls. Thus, although being potentially relevant metabolites, these metabolites identified at level 4 were not taken into account for further data interpretation owing to the lack of feasibility in their identification, which could lead to erroneous interpretations. Plasma metabolic signatures in SZ, CUD and DUAL patients The metabolites differentially altered in the SZ, CUD and DUAL patients (Fig. , p values < 0.05, one-way ANOVA with Tukey’s post hoc test; see Supplementary Information for statistical data) were identified by inspection of the variable importance of projection (VIP) scores of the significant OPLS-DA models described above. One of the most relevant alterations is related with content of two acylcarnitines, L-octanoylcarnitine (Fig. a) and L-decanoylcarnitine (Fig. b), which are significantly decreased in the three groups of patients compared to control subjects. In addition, in SZ patients, the metabolite 1-methylnicotinamide (Fig. c) appears downregulated. The other metabolic impairments suffered by CUD patients, were the downregulations of the N-acyl amino acids (NAAAs) N-palmitoyl threonine (Fig. d) and N-palmitoyl serine (Fig. e), which was also downregulated in SZ and DUAL patients as well. In addition to the aforementioned changes, both SZ and DUAL patients showed a higher plasma content of a sterol lipid (7-dehydrodesmosterol) compared to control subjects (Fig. f).
According to the workflow and data filtering described in the next section, a final list of 119 potential compounds in all the analyzed samples was obtained. Final candidates were annotated based on the endogenous metabolites mass lists (Human Metabolome Database and LipidMaps structure database) and MS2 mass spectra. The potential name of each candidate, grouped by compound classes, together with their predicted molecular formula, predicted monoisotopic mass and its error, the match level (with mzCloud spectra or with in silico fragmentation), retention time and the identification confidence level are shown in Supplementary Table . Lipids, predominantly fatty acids, glycerolipids, glycerolphospholipids and sphingolipids, represented the most detected class among all the potential candidates of the final list. Leaving aside 6 fatty acids that were annotated at level 2a due to high MS2 spectra matching and lack of ambiguity with other potential candidates, most of the identified lipids were annotated at level 3. The exogenous compounds (i.e. non-endogenous metabolites) are identified as “Others” in Table S2 under the “Categories/Subclass” column. They represent 19 of the 77 compounds included among the 2a/2b and 3 identification confidence levels. All the metabolites at level 3 (51% of candidates) showed acceptable spectral matches, but this is not enough to assure the discrimination between positional isomers or equivalent structures (e.g. the position of a double bond in unsaturated fatty acids), and therefore, only general structures and compound classes could be assured. Nevertheless, the putative name of the probable candidate (those with the higher number of citations) in metabolites annotated at level 3 was maintained. In contrast, candidates annotated at level 4 (35%), showed uninformative MS2 spectra and big ambiguity between other potential candidates, and thus, it was not even possible to determine the functional group or the class of these compounds. Consecutively, neither name nor family name was given to these compounds and only molecular formula was maintained. The remaining candidates (13%) were annotated at level 2a or 2b due to unambiguous MS2 spectra matching with experimental MS2 available in mzCloud spectra or in silico fragmentation in Compound Discoverer, respectively. Within this group of compounds, three unsaturated fatty acids, two acyl carnitines, one N-acyl amine and one nicotinamide were identified. In addition to all the endogenous compounds mentioned above, some exogenous compounds and metabolites were identified at the same identification level, which deserve special attention. On the one hand, piperine (M003), tricyclazole (M012) and 2-hydroxybenzothiazol (M013) were identified, which have been detected in urine in other suspect analysis research works in the framework of human exposome assessment . More importantly, active substances of drugs and pharmaceuticals and related metabolites have been found as well. In this sense, we can highlight, nicotine (M001) and cotinine (M002), as tobacco consumption biomarkers ; aripiprazole (M009) and dehydroaripiprazole (M010), as antipsychotic treatment biomarkers , and 11-nor-9-carboxy-∆ 9 -THC (M015), as cannabis consumption biomarker . Aripiprazole (Supplementary Fig. a) and dehydroaripiprazole (Supplementary Fig. b) were found in DUAL and SZ subjects but not in controls (for both metabolites and groups, log 2 (FC) = 5.5–8.8) or CUD subjects (for both metabolites and groups, log 2 (FC) = 2.9–3.1). Moreover, CUD and DUAL subjects had higher levels of nicotine and cotinine in plasma compared to controls (for both metabolites and groups, log 2 (FC) = 4.7–5.7) and to SZ patients (for both metabolites and groups log 2 (FC) = 1.2–2.1) (Supplementary Fig. ). This finding suggests that there is a moderate tobacco consumption among CUD subjects (probably related to cannabis consumption as well) and DUAL subjects, and low or none tobacco consumption among SZ patients and controls. Higher 11-nor-9-carboxy-∆ 9 -THC content was found in CUD subjects compared to the rest of groups (log 2 (FC) = 3.1 compared to DUAL, and, log 2 (FC) = 5.4–6 compared to SZ or controls) (Supplementary Fig. ). These observations confirm that SZ and DUAL patients were under antipsychotic treatment, as well as cannabis consumption among CUD patients, and none consumption of any of these substances among controls. Although these expected findings were interesting to verify the consumption and treatment status of the subjects, these discriminative exogenous metabolites (together with the mentioned exogenous compounds belonging to exposome) directly related with the specific characteristic of the studied population were not taken into account for further analysis in order to go beyond the exposome analysis and study the alterations in lipidomic pathway.
Before analyzing the changes in the metabolic profiles of the three groups of patients, one of the first issues to be addressed was whether or not the recruited controls to each cohort (CUD, DUAL and SZ) were equivalent. To test this underlying hypothesis, control samples from different groups were compared by one way ANOVA using MetaboAnalyst (Supplementary Table ). This analysis showed that 12 metabolites were significantly different in the three groups of control samples. Consequently, they were eliminated from all the data (together with the exogenous metabolites mentioned above) in order to avoid biased results. The metabolic profiles of each group of patients (CUD, DUAL and SZ) were then contrasted with those of the control group. Multivariate analysis using OPLS-DA was run along univariate analysis using FC and FDR t-tests. The score plots, permutation analysis results and the variable importance in projection (VIP) scores from the three OPLS-DA models are shown in Fig. . In addition, volcano plots obtained from univariate analyses are shown in Fig. . The score plots of the OPLS-DA models (Fig. a, d and g), show that the variability in the contents of the annotated metabolites allows metabolic clustering between control subjects and either CUD, DUAL or SZ patients. The cumulative R2Y and Q2 values of the OPLS-DA models were respectively, 0.92 and 0.85 for CUD vs. control; 0.91 and 0.87 for DUAL vs. control; and, 0.96 and 0.89 for SZ vs. control. Overall, around 40% of the variance was explained with the first two principal components. These values suggests that the OPLS-DA models fit the data correctly (R2Y near to 1) and that they have good predictive ability (Q2 > 0.5). Moreover, in the three OPLS-DA models, permuted RY2 values were around 0.6 or below, and most permuted Q2 values were below 0 (Fig. b, e and h). Also, permuted RY2 and Q2 values were in all cases below cumulative RY2 and Q2 values. All this suggests that the model fitting is adequate, and that was unlikely to be built by chance. After checking the suitability of the OPLS-DA models, metabolites that allowed discrimination between patient groups and controls were taken into account (Fig. c, f and i). To ensure a higher reliability of these results, among the potential biomarkers elucidated by the OPLS-DA analyses, only metabolites that showed significant changes according to volcano plots (Fig. ) were further considered. Nonetheless, most of the metabolites that allowed discrimination of groups in the OPLS-DA models (those with the highest VIP) were significant according to volcano plots (FC > 2, p < 0.05), and vice versa. Compared to controls, CUD subjects (Fig. a) presented a significant increase in the levels of two metabolites identified at level 4 (M094 and M096), and a significant decrease in the content of two acylcarnitines (M007 and M008), two N-acyl amino acids (M016 and M023), one primary amide (M026), one acetate ester (M070) and other nine metabolites (M078, M081, M083, M085, M093, M095, M097, M099 and M103). DUAL (Fig. b) and SZ (Fig. c) patients, in contrast to CUD patients, presented a higher degree of metabolic impairments compared to controls. DUAL patients showed increased plasma levels of one oxo fatty acid (M019), four ceramides (M052, M053, M054 and M058), one sterol lipid (M062) and other four metabolites (M077, M094, M096 and M119); and decreased levels of two acylcarnitines (M007 and M008), one primary amide (M026), three glycerolphospholipids (M037, M040 and M043), one phenolic acid (M064) and other 11 metabolites (M078, M079, M081, M083, M085, M087, M093, M097, M103, M105 and M108). Similarly, SZ patients showed increased plasma levels of two ceramides (M044 and M053), one sterol lipid (M062) and other four metabolites (M090, M096, M107 and M115), along with decreased levels of two acylcarnitines (M007 and M008), one nicotinamide (M011), one N-acyl amino acid (M023), one primary amide (M026), seven glycerolphospholipids (M034, M035, M036, M038, M039, M040 and M043), one phenolic acid (M064), one dipeptide (M066), one acetate ester (M070) and other twelve metabolites (M078, M079, M081, M083, M085, M087, M095, M097, M099, M102, M103 and M108). These analyses showed that either CUD, DUAL or SZ patients had distinctive metabolic profiles compared to healthy subjects, and pointed out a few potential biomarkers, some of them shared by two patient groups or more. It should be noted though, that some metabolites that underwent significant changes were among the metabolites identified at level 4. Indeed, a few of them (i.e., M081, M083, M085, M094 and M096) showed some of the highest scores (or highest FCs); being key compounds to provide metabolic clustering between patient groups and controls. Thus, although being potentially relevant metabolites, these metabolites identified at level 4 were not taken into account for further data interpretation owing to the lack of feasibility in their identification, which could lead to erroneous interpretations.
The metabolites differentially altered in the SZ, CUD and DUAL patients (Fig. , p values < 0.05, one-way ANOVA with Tukey’s post hoc test; see Supplementary Information for statistical data) were identified by inspection of the variable importance of projection (VIP) scores of the significant OPLS-DA models described above. One of the most relevant alterations is related with content of two acylcarnitines, L-octanoylcarnitine (Fig. a) and L-decanoylcarnitine (Fig. b), which are significantly decreased in the three groups of patients compared to control subjects. In addition, in SZ patients, the metabolite 1-methylnicotinamide (Fig. c) appears downregulated. The other metabolic impairments suffered by CUD patients, were the downregulations of the N-acyl amino acids (NAAAs) N-palmitoyl threonine (Fig. d) and N-palmitoyl serine (Fig. e), which was also downregulated in SZ and DUAL patients as well. In addition to the aforementioned changes, both SZ and DUAL patients showed a higher plasma content of a sterol lipid (7-dehydrodesmosterol) compared to control subjects (Fig. f).
Many metabolomics studies have tried to find specific biomarkers of schizophrenia for its early detection and disease progress monitoring. Regarding this, mass spectrometry (MS) is the analytical technique of preference when it comes to clinical metabolomics of plasma and other biofluids , . Untargeted analysis allows to obtain a general metabolic profiling and eases the search of potential biomarkers, becoming the preferred approach for the preliminary steps in clinical metabolomic studies , . The results obtained from untargeted analyses have been further studied using both univariate and multivariate statistical analyses in order to seek for differences in the metabolic profiles of control subjects and patients from the different groups. In these sense, the relation between cannabis abuse and the development of schizophrenia from a metabolomic perspective is studied too, which could be useful for laying the groundwork for future metabolomic studies focused on the same concern. Among the observed changes, the upregulation of ceramides and the downregulation of various glycerolphospholipids paired by the two groups individuals suffering from schizophrenia are worthy of particular attention. Ceramides represent the central element in the metabolism of sphingolipids, ubiquitous constituents of eukaryotic cell membranes . Besides their structural role, sphingolipids are involved in cellular signaling, cell differentiation and proliferation, apoptosis and inflammation, among other functions , . In the central nervous system, sphingolipids are located in neuronal cell membranes, and impaired sphingolipid metabolism has been linked with brain dysfunction and the development of different psychological diseases. Overproduction of ceramides results to be toxic and accelerates apoptosis of cells . Additionally, increased content of different ceramides in brain tissues and blood has been related with neuroinflammation, altered synaptic transmission and accelerated apoptosis observed in various neurodegenerative and psychiatric disorders , . In this manner, compared to healthy subjects, increased levels of three ceramides and other three different ceramides have been quantified in postmortem white matter and plasma of patients with schizophrenia. These findings are in line with the observed upregulations of various ceramides in the plasma samples of patients of the SZ (M044, M053) and DUAL (M052-054, M058) groups in our study. In the case of schizophrenia, it has been proposed that the accumulation of ceramides may inhibit the action of excitatory amino acid transporter-2, leading to an accumulation of glutamate in the synaptic cleft, which results in excitotoxicity and hence, in neuronal damage . Glycerolphospholipids are the main constituents of neuronal and glial cell membranes, and are crucial to maintain the membrane’s structure and many membrane and cellular functions . Phosphatidylcholines (PCs) and phosphatidylethanolamines (PEs), comprise the most abundant classes of glycerolphospholipids in the brain. They are vulnerable to lipid peroxidation, and hence, the oxidative stress typically observed in schizophrenia has been related with an excessive breakdown of these membrane phospholipids, which leads to neuronal damage and brain dysfunction , , . Besides, lysophosphatidylcholines (LPCs) and lysophosphatidylethanolamines (LPEs) are less abundant, but stand out due to their involvement in membrane function, apoptosis, and their protective role during oxidative stress and inflammation . Therefore, a decreased content of these phospholipids may be also related with membrane dysfunction and a possible deficiency to face oxidative stress and neuroinflammation , . In line, with a suggested reduction of these key phospholipids in the brain, several studies have reported decreased content of various PCs, PEs, LPCs and LPEs in serum and plasma of patients with schizophrenia , , , , which is in accordance with our observations. Interestingly, these changes were mainly observed in SZ group subjects, where the content of three LPEs (M034, M035, M036), three PEs (M038, M038, M040) and one phosphatidylglycerol (PG, M043) were found to be significantly decreased compared to controls. Conversely, whereas subjects of the DUAL group only showed decreased levels of two PEs (M037, M040) and one PG (M043). Related to the excitotoxicity, oxidative stress and neuroinflammation that patients with schizophrenia usually show , , it is worth mentioning the observed decrease in 1-methylnicotinamide (M060) levels in plasma samples of the SZ group. 1-methylnicotinamide is a neuroprotective endogenous metabolite that reduces glutamate excitotoxicity , and can help to attenuate neuroinflammation, neuronal apoptosis and oxidative stress . Therefore, a deficiency of 1-methylnicotinamide can ease the apparition of neuronal damage in case of glutamate accumulation, neuroinflammation and/or oxidative stress. In agreement with our results, Fan and colleagues reported a decrease of 1-methylnicotinamide in the plasma of patients with schizophrenia compared to healthy controls . One of the most relevant alterations was related with the content of two acylcarnitines, L-octanoylcarnitine and L-decanoylcarnitine, which were significantly decreased in the three groups of patients compared to healthy subjects. In the brain, acylcarnitines play a role in neuroprotection, lipid synthesis, membrane stability, gene and protein modulation, mitochondrial function, antioxidant activity, and cholinergic neurotransmission . Altered acylcarnitine metabolism has been related with mitochondrial dysfunction, oxidative stress and inflammation – , processes that can lead to neuronal damage and that are usually found among schizophrenia patients , . In line with the observations found in this work, several studies have reported a decreased content of L-octanoylcarnitine and L-decanoylcarnitine in the plasma or serum of patients with schizophrenia, along with altered levels of other acylcarnitines , , . Interestingly, Maayah and colleagues reported a lower content of L-octanoylcarnitine and medium- and short-chain acylcarnitines, and higher levels of some long-chain acylcarnitines in the serum of rats treated with THC-containing cannabis extracts, and proposed these as potential biomarkers of behavioral changes related to cannabis use . Taking all this along with the findings of this work, we suggest that continued THC use could alter the regulation of acylcarnitine metabolism, exposing users to the psychological risks that these changes may entail. Along with acylcarnitines, the other metabolic impairments suffered by CUD patients, were the downregulations of the N-acyl amino acids (NAAAs) N-palmitoyl threonine (M016) and N-palmitoyl serine (M023), which were also downregulated in SZ patients. N-acyl amino acids (NAAAs) are an important family of endogenous signaling with relevant involvement in both physiological and/or pathological conditions , . These lipid mediators have been involved in energy homeostasis and neuroprotection . Moreover, N-acyl amino acids can modulate glutamatergic and GABAergic neurotransmission , key systems implicated in schizophrenia and addiction. NAAAs share one inactivation enzyme (fatty-acid amide hydrolase, FAAH) and various molecular targets with endocannabinoids and, thus, they have been considered part of the “endocannabinoidome” , . The biosynthesis and physiological functions of some NAAAs, such as N-acylglycines, are well known; but, in general, NAAAs have only been identified and have not yet been studied in detail , , . For instance, N-palmitoyl threonine (M048) was found to be upregulated in certain brain areas during induced acute inflammation in rats , and the stearoyl NAAA with the same amino acid, N-stearoyl threonine, was found to have neuroprotective effects . The neuroprotective action (through indirect action) of N-palmitoyl serine (Met_020) has been suggested ; and neuroprotective properties of other N-acyl serines have also been proposed , . Indeed, Wood et al., found elevated levels of some N-acylphosphatidylserines and N-acylserines (including N-palmitoyl serine) in postmortem frontal cortex of schizophrenia subjects , proposing them as potential biomarkers of endogenous neuroprotective mechanisms. Thus, the observed changes in plasma levels of the two NAAAs in subjects with CUD or SZ could imply a deficit in the neuroprotective mechanisms of these patients. However, due to the current lack of knowledge about the synthesis and biological implication of these NAAAs and the fact that fatty acids and amino acids are essential dietary elements , more research is needed to properly interpret such metabolic changes. The observed decrease in metabolite concentration in patients relative to controls may be indicative of underlying pathophysiological alterations associated with schizophrenia and cannabis use. Lipids and lipid-derived molecules play a pivotal role in neuroprotection, neurotransmitter modulation, and anti-inflammatory responses. A decrease of these metabolites may indicate an inability to maintain essential protective functions in the context of schizophrenia or cannabis use. Furthermore, diminished levels of key metabolites may also indicate mitochondrial dysfunction and energy deficiencies. Such changes are commonly observed in individuals with schizophrenia and are further exacerbated by the effects of cannabis on metabolic homeostasis . In addition to the aforementioned changes, both SZ and DUAL patients showed a higher plasma content of a sterol lipid (7-dehydrodesmosterol, M062) compared to healthy subjects. 7-dehydrodesmosterol is implied in the biosynthesis of cholesterol, neurosteroids and vitamin D3. A deficiency of either of these metabolites has been linked with an increased risk of psychosis , hence it is difficult to correlate the up-regulation of 7-dehydrodesmosterol with the pathological state of patients with SZ or DUAL. However, it is worth mentioning that some antipsychotics as well as cannabis use , may interact with lipid pathways. Aripiprazole (the antipsychotic mainly used by patients in the SZ and DUAL groups), inhibit the action of 7-dehydrocholesterol reductase, resulting in an increase of 7-dehydrocolesterol and 7-dehydrodesmosterol content , , . Therefore, the observed increase in 7-dehydrodesmosterol levels in both SZ and DUAL patients may be a consequence of aripiprazole treatment, and thus, an indirect biomarker of its consumption. The present work provides an additional evidence of the power of untargeted lipidomics for clinical studies. The LC-HMRS based analysis of plasma samples, followed by a proper treatment of chromatographic and mass spectrum data, allowed us to identify up to 119 unknown metabolites at different identification confidence level. Although many metabolites were identified merely at level 4 and obstructed the interpretation of some of the observed metabolic impairments, a considerable number of metabolites were identified between levels 2a and 3; which was enough to detect some key families of metabolites and other endogenous and exogenous compounds. Multivariate and univariate analyses were key to find altered metabolic pathways among CUD, DUAL and SZ groups. The OPLS-DA models allowed clear discrimination between either of the groups of patients and healthy subjects; and pointed out some potential biomarkers. In addition, univariate analyses showed significant alterations of the same discriminant metabolites elucidated by OPLS-DA, which confers a higher level of confidence in the results and conclusions obtained. Some of the potential biomarkers elucidated from both analyses were identified at level 4, and thus, the metabolic pathways and biological implications related to these impaired metabolites could not be analyzed. Nonetheless, metabolic alterations of key metabolites’ families such as acylcarnitines, ceramides and glycerolphospholipids were found, which were in accordance to the literature and led to elucidate some preliminary hypotheses around the relationship between cannabis abuse and the risk of psychosis. Overall, our findings highlight key metabolic disruptions, particularly the consistent reduction in acylcarnitines (L-octanoylcarnitine and L-decanoylcarnitine) across all patient groups, suggesting impaired fatty acid oxidation as a shared feature in both disorders. Additionally, the significant downregulation of N-acyl amino acids (N-palmitoyl threonine and N-palmitoyl serine) in CUD and schizophrenia suggests a disruption in lipid signaling pathways that may contribute to the pathophysiology of these conditions. The elevated levels of 7-dehydrodesmosterol in SZ and dual diagnosis patients point to altered cholesterol metabolism as a potential distinguishing metabolic signature. These results provide insight into the biological mechanisms linking cannabis abuse and schizophrenia, offering novel biomarker candidates that could improve diagnosis and therapeutic strategies. While our findings underscore the utility of metabolomics in identifying metabolic pathways implicated in neuropsychiatric disorders, further research with larger cohorts is needed to validate these biomarkers and clarify their roles in disease progression. In conclusion, lipidomic profiling holds promise for advancing our understanding of schizophrenia and CUD, potentially informing targeted interventions that address metabolic dysfunction in these conditions.
Subjects Subjects who met inclusion criteria for cannabis use disorder (CUD, n = 24), schizophrenia (SZ, n = 18), or dual diagnosis (DUAL, n = 12) (Table ) according to DSM-IV/DSM-IV-TR 70 were included in the study. The controls were voluntary donors who met the criteria for sex and age matching. The inclusion criteria for the controls were age 18–60, while the exclusion criteria included any use of cannabis or a diagnosis of a neuropsychiatric disease within the previous two years prior to blood extraction. All subjects were excluded if they met the criteria for a severe mental disorder other than schizophrenia or had a history of severe congenital, medical, or neurological illnesses. Moreover, a blood toxicological screening was performed in all subjects to determine the presence of antipsychotics as well as of THC . The demographic characteristics of the four subject groups used in the study as well as the antipsychotic treatment of all the cases are described in a previous work and in Supplementary Tables , and . All the participants gave written, witnessed, informed consent for the participation in the study, that was approved by the corresponding Human Research Ethics Committee (University Cruces Hospital, code CEIC E14/43). All methods in the study were carried out in accordance with the ethical principles of the Declaration of Helsinki and Good Clinical Practice. Plasma collection Approximately 20 mL of blood was collected via venipuncture into ACD solution A Vacutainer ® citrate blood collection tubes (Becton Dickinson & Company, Franklin Lakes, NJ, USA). The blood was drawn by nurses at the Drug Addiction Unit of the Uribe Mental Health Centre (Getxo, Spain), part of the Basque Health Service, and at the School of Medicine, University of the Basque Country (UPV/EHU) (Leioa, Spain). For plasma preparation, blood cells were separated from plasma by centrifugation at 1,000–2,000 × g for 10 min in a refrigerated centrifuge set to 4 ºC. The plasma supernatant was then carefully transferred into a clean polypropylene tube using a Pasteur pipette. Samples were stored at -70 ºC until analysis, with the cold chain strictly maintained throughout collection and handling to preserve the integrity of chemical compounds. Untargeted analysis of non-polar metabolites in plasma Plasma samples were processed in order to extract mainly lipophilic metabolites and were subsequently analyzed with an untargeted approach by means of liquid-chromatography coupled to tandem high resolution mass spectrometry (LC-MS/HRMS). A detailed description of the employed materials and methods for the extraction of non-polar metabolites, LCMS/HRMS analysis, and data handling of the raw chromatograms has been compiled in Supplementary Methods File. Briefly, a liquid-liquid extraction using methyl tert-butyl ether was carried out to extract non-polar metabolites from 0.5 mL aliquots of plasma samples. Organic phases were separated from the resulting biphasic mixtures, and were dried under nitrogen flow and reconstituted in xx mL of acetonitrile. Final extracts were analyzed along with instrumental blanks, procedural blanks and a pool of all studied samples as quality control (QC) sample in the LC-MS/HRMS system. Chromatographic separation of metabolites was carried out in reverse phase using a gradient chromatographic method, and targets were determined in both positive and negative ionization modes. The acquired raw chromatograms were filtered based on the adequacy of the chromatographic peaks, the mass accuracy and the quality of fragmentation spectra; and chromatograms of blanks and QC samples where used to correct potential analytical bias in sample processing and instrumental analysis. After data filtering, annotation of the final list of unknown metabolites was carried using freely available mass lists, and by contrasting the obtained fragmentation MS2 mass spectra with experimental fragmentation mass spectra libraries and in silico fragmentation. Identification confidence levels of the final candidates were done as proposed by Schymanski et al. (Schymanski et al., 2014). Statistical analysis To identify metabolites involved in altered metabolic pathways of the 104 individuals, the corrected areas determined in both sequences (i.e., positive and negative ionization modes) were merged in a spreadsheet treated in the same statistical analysis workflow using MetaboAnalyst 6.0 ( MetaboAnalyst , n.d.). Data was logarithmically transformed and autoscaled and potential outliers were identified based on Principal Component Analysis (PCA). Equivalence of control groups was checked using one way ANOVA with Fisher’s post hoc test (p-value ≤ 0.05). Univariate and multivariate analyses were combined in order to find metabolic alterations and potential biomarkers in CUD, DUAL and SZ patients compared to healthy subjects. For univariate analysis, FC and t-tests (with p-value FDR correction) were used, applying a cutoff value of ≥ 2 for FC and cutoff value of ≤ 0.05 for p-values. For multivariate analysis, OPLS-DA was used and the quality of the models was evaluated based on R2Ycum (goodness of fit) and cumulated Q2cum values. Additionally, OPLS-DA models were validated using permutation tests (1000 permutations). For comparisons between the levels of discriminating metabolites selected by the OPLS-DA models in SZ, CUD and DUAL groups, ANOVA with Tukey’s post-hoc test was used.
Subjects who met inclusion criteria for cannabis use disorder (CUD, n = 24), schizophrenia (SZ, n = 18), or dual diagnosis (DUAL, n = 12) (Table ) according to DSM-IV/DSM-IV-TR 70 were included in the study. The controls were voluntary donors who met the criteria for sex and age matching. The inclusion criteria for the controls were age 18–60, while the exclusion criteria included any use of cannabis or a diagnosis of a neuropsychiatric disease within the previous two years prior to blood extraction. All subjects were excluded if they met the criteria for a severe mental disorder other than schizophrenia or had a history of severe congenital, medical, or neurological illnesses. Moreover, a blood toxicological screening was performed in all subjects to determine the presence of antipsychotics as well as of THC . The demographic characteristics of the four subject groups used in the study as well as the antipsychotic treatment of all the cases are described in a previous work and in Supplementary Tables , and . All the participants gave written, witnessed, informed consent for the participation in the study, that was approved by the corresponding Human Research Ethics Committee (University Cruces Hospital, code CEIC E14/43). All methods in the study were carried out in accordance with the ethical principles of the Declaration of Helsinki and Good Clinical Practice.
Approximately 20 mL of blood was collected via venipuncture into ACD solution A Vacutainer ® citrate blood collection tubes (Becton Dickinson & Company, Franklin Lakes, NJ, USA). The blood was drawn by nurses at the Drug Addiction Unit of the Uribe Mental Health Centre (Getxo, Spain), part of the Basque Health Service, and at the School of Medicine, University of the Basque Country (UPV/EHU) (Leioa, Spain). For plasma preparation, blood cells were separated from plasma by centrifugation at 1,000–2,000 × g for 10 min in a refrigerated centrifuge set to 4 ºC. The plasma supernatant was then carefully transferred into a clean polypropylene tube using a Pasteur pipette. Samples were stored at -70 ºC until analysis, with the cold chain strictly maintained throughout collection and handling to preserve the integrity of chemical compounds.
Plasma samples were processed in order to extract mainly lipophilic metabolites and were subsequently analyzed with an untargeted approach by means of liquid-chromatography coupled to tandem high resolution mass spectrometry (LC-MS/HRMS). A detailed description of the employed materials and methods for the extraction of non-polar metabolites, LCMS/HRMS analysis, and data handling of the raw chromatograms has been compiled in Supplementary Methods File. Briefly, a liquid-liquid extraction using methyl tert-butyl ether was carried out to extract non-polar metabolites from 0.5 mL aliquots of plasma samples. Organic phases were separated from the resulting biphasic mixtures, and were dried under nitrogen flow and reconstituted in xx mL of acetonitrile. Final extracts were analyzed along with instrumental blanks, procedural blanks and a pool of all studied samples as quality control (QC) sample in the LC-MS/HRMS system. Chromatographic separation of metabolites was carried out in reverse phase using a gradient chromatographic method, and targets were determined in both positive and negative ionization modes. The acquired raw chromatograms were filtered based on the adequacy of the chromatographic peaks, the mass accuracy and the quality of fragmentation spectra; and chromatograms of blanks and QC samples where used to correct potential analytical bias in sample processing and instrumental analysis. After data filtering, annotation of the final list of unknown metabolites was carried using freely available mass lists, and by contrasting the obtained fragmentation MS2 mass spectra with experimental fragmentation mass spectra libraries and in silico fragmentation. Identification confidence levels of the final candidates were done as proposed by Schymanski et al. (Schymanski et al., 2014).
To identify metabolites involved in altered metabolic pathways of the 104 individuals, the corrected areas determined in both sequences (i.e., positive and negative ionization modes) were merged in a spreadsheet treated in the same statistical analysis workflow using MetaboAnalyst 6.0 ( MetaboAnalyst , n.d.). Data was logarithmically transformed and autoscaled and potential outliers were identified based on Principal Component Analysis (PCA). Equivalence of control groups was checked using one way ANOVA with Fisher’s post hoc test (p-value ≤ 0.05). Univariate and multivariate analyses were combined in order to find metabolic alterations and potential biomarkers in CUD, DUAL and SZ patients compared to healthy subjects. For univariate analysis, FC and t-tests (with p-value FDR correction) were used, applying a cutoff value of ≥ 2 for FC and cutoff value of ≤ 0.05 for p-values. For multivariate analysis, OPLS-DA was used and the quality of the models was evaluated based on R2Ycum (goodness of fit) and cumulated Q2cum values. Additionally, OPLS-DA models were validated using permutation tests (1000 permutations). For comparisons between the levels of discriminating metabolites selected by the OPLS-DA models in SZ, CUD and DUAL groups, ANOVA with Tukey’s post-hoc test was used.
Below is the link to the electronic supplementary material. Supplementary Material 1
|
Autonomous surgical planning of mandibular angle reduction based on anatomical landmarks and osteotomy plane detection | 859e802e-b5e3-4fbc-9334-8410344e9b67 | 11836450 | Dentistry[mh] | Mandibular angle reduction is a surgical procedure aimed at altering the contour of the mandibular angle, the posterior border where the lower jaw transitions into the ramus . This surgery, part of facial skeletal contouring, follows well-established clinical design steps to achieve a smoother facial profile by selectively removing the more prominent part of the mandibular angle, tailored to each patient’s unique facial characteristics . This procedure is intricate, and the surgical risk is high. A variety of important neurovascular vessels are adjacent around the surgical area, including the facial artery, maxillary artery, alveolar artery, mandibular nerve, inferior alveolar nerve (IAN), mental nerve, and facial nerve. Improper operation will cause complications such as neurovascular injury, masticatory dysfunction, and mandibular fracture . Traditional freehand operation , surgical guide , and surgical navigation technology are commonly used in clinical practice for mandibular angle osteotomy. In recent years, technologies such as augmented reality and surgical robotics , have also been used in maxillofacial surgery. However, regardless of the surgical approach described above, the establishment of a preoperative surgical plan and the measurement of the patient’s individual anatomical parameters are prerequisites for achieving a precise and quantitative mandibular osteotomy. The above process begins with image segmentation and 3D reconstruction of the mandible using the patient’s preoperative CT imaging data, and then the surgeon set the osteotomy plane by the clinical experience. Due to the individualization of the patient and difference of the surgeon’s clinical experience, the design of the surgical planning plan requires repeated adjustments and evaluations, resulting in a low efficiency of preoperative planning , . According to the literature , the average surgical planning time of a mandibular osteotomy for a surgeon is more than 40 min. Numerous methods have been employed for preoperative planning of mandibular osteotomies. Most of these studies rely on commercial medical image processing software. Ming Zhu et al. used Mimics software to visualize the soft and hard tissues in the patient’s CT and manually complete the osteotomy surgical planning, which was subsequently integrated into the surgical navigation system. Ye et al. initially conducted a semi-automatic segmentation of key maxillofacial tissues and then imported the results into Geomagic Freeform (3D Systems, USA) software for surgical planning and guide plate design. Similarly, He et al. used the Geomagic Freeform software to manually design the inferior horn osteotomy line and the surgical guide plate. Clinical trials were carried out to verify effectiveness. Zhang et al. used ProPlan CMF (Materialise, Belgium) software to segment and reconstruct the mandibular CT images and completed the surgical planning plan according to the predefined rules for mandibular angle osteotomy line generation, combined with manually selected mandibular anatomical landmark points. In the robot-assisted mandibular osteotomy, Lin et al. and Sun et al. both firstly generated the osteotomy line based on anatomical landmark points, and then generated the drilling path. While the manual or semi-automatic mandibular angle osteotomy planning methods mentioned above are viable and effective, certain limitations persist. Firstly, the segmentation and 3D reconstruction of the mandible and soft tissues is time-consuming. Various software is used in different steps, leading to a complex process. Additionally, the osteotomy plan cannot be automatically generated, and the results of the anatomical parameters cannot be automatically generated, resulting in longer preoperative preparation time. Therefore, automated surgical planning has gained prominence due to its high-efficiency. Gupta et al. proposed a knowledge-based algorithm to achieve automatic detection of anatomical marker points. The method generates a volume of interest based on the searched seed points and achieves automatic recognition of marker points by edge detection. Neelapu et al. proposed a marker point detection method based on edge detection for 3D cephalometric measurement of CBCT images and achieved automatic recognition of 20 marker points in the head. The aforementioned marker point automatic recognition methods rely on the initially identified region of interest, and any misidentification will lead to the failure of the subsequent marker point detection methods. Several advanced deep-learning methods have been proposed for locating landmarks on the mandible. PoseNet effectively identified 27 anatomical landmarks by segmenting the mandible from cone-beam computed tomography (CBCT) images using a deep learning mechanism. Additionally, Ahn et al. utilized a deep convolutional neural network (CNN) based on artificial intelligence for accurate landmark detection. The unique shape of the mandible is often overlooked in common deep-learning techniques, where landmark localization can be ambiguous or not explicitly defined. To address these issues, this paper introduces a fully automated method for detecting mandibular anatomical feature points and a model for optimizing the mandibular osteotomy plane based on key anatomical measurement parameters. Additionally, these algorithms are seamlessly integrated into an automated mandibular osteotomy surgical planning software system. The goal is to reduce the time required for surgical planning and enhance the efficiency of preoperative planning for mandibular angle osteotomy. To achieve full preoperative planning for mandibular angle osteotomy, this paper proposes an algorithm for automatic identification of bony markers at the mandibular angle as well as the automatic measurement of anatomical parameters. The overall algorithm workflow is illustrated in Fig. . The workflow of the integrated surgical planning system is as follows: The CT image data of the patient was segmented using the previously proposed deep learning-based automatic craniomaxillofacial segmentation algorithm . The mandible and inferior alveolar nerve were reconstructed using a face mapping algorithm. The mandibular anatomical landmark points and occlusal planes were automatically identified and calculated, which would be described in detail in Sect. . The identified marker points were checked again and manually fine-tuned. An optimization model was built to calculate the mandibular osteotomy plane, which would be described in detail in Sect. . The key anatomical parameters were measured and a surgical planning scheme was generated, which will be described in detail in Sect. . Automatic recognition of anatomical marker points The purpose of mandibular anatomical marker point identification is to facilitate the planning of the surgical osteotomy plane and the measurement of key anatomical parameters. These anatomical markers are particularly significant in defining the shape of the mandible, especially in the region near the mandibular angle. As shown in Fig. , the planning of mandibular angle osteotomy requires the identification of key anatomical landmark points on each of the left and right sides of the mandible. These include the condylar apex (Condylion, Co), the mandibular angle point (Gonion, Go), the subchin point (Meton, Me), the projection of the second molar point on the inferior margin of the mandible (point E), the intersection of the occlusal plane with the external mandibular oblique line (point), and the intersection of the occlusal plane with the ramus of the mandible (point O), as well as a characteristic plane: the occlusal plane (OP). Definitions for these points are provided in Table . To detect these features, the HOBB (Hierarchical Oriented Bounding Box)-based method is proposed. The HOBB is a tool in the field of computer collision detection, and the author discovered that the shape of the mandible resembles a half-circle with two upward supports. The HOBB calculates a surface that splits the volume of the three-dimensional object into two equal parts. Therefore, if the first-order bounding box of the mandible is calculated, due to its symmetry, the split surface will be close to the mid-surface; if the second-order bounding box is calculated, due to the anatomical features of the mandible (width of the mandibular body is close to the width of the mandibular ramus, and the length of the mandibular body is approximately twice the length of the mandibular ramus), the volume of each side of the mandibular ramus and mandibular body will be approximately equal, with the split surface positioned centrally between them. As a result, the second-order bounding box of the mandible roughly divides the mandible into the left and right mandibular ramus and the left and right mandibular body. Therefore, through repeated research, this paper proposes a HOBB-based method for automatic recognition of the anatomical structure of the mandible, which is as follows: First, obtain the overall mandible model’s bounding box [12pt]{minimal} $$\:OB{B}_{MB}$$ and its center [12pt]{minimal} $$\:{O}_{MB}$$ . Then, model the reconstructed three-dimensional mandible model with HOBB, and obtain four bounding boxes (as shown in Fig. b), referred to as [12pt]{minimal} $$\:OB{B}_{MAL}$$ (located in the left mandibular arch area), [12pt]{minimal} $$\:OB{B}_{MAR}$$ (located in the right mandibular arch area), [12pt]{minimal} $$\:OB{B}_{MBL}$$ (located in the left mandibular body area), and [12pt]{minimal} $$\:OB{B}_{MBR}$$ (located in the right mandibular body area). After obtaining the left and right mandibular bodies, the cross product of the normal vectors of the longest axes of their bounding boxes is computed to determine the approximate direction of the mandible’s center of gravity [12pt]{minimal} $$\:_{fin}}.$$ The detailed method is as follows: First, the reference direction of the mandible center of gravity is established using [12pt]{minimal} $$\:OB{B}_{MBL}$$ and [12pt]{minimal} $$\:OB{B}_{MBR}.$$ Its initial value can be represented as: 1 [12pt]{minimal} $$\:_{ini}}=_{MBL}}\:_{MBR}}$$ The long axis vectors of [12pt]{minimal} $$\:_{MBL}}$$ and [12pt]{minimal} $$\:_{MBR}}$$ are respectively denoted as [12pt]{minimal} $$\:\:OB{B}_{MBL}$$ and [12pt]{minimal} $$\:OB{B}_{MBR}$$ . The final reference direction of the center of gravity [12pt]{minimal} $$\:_{fin}}$$ is obtained through an iterative optimization method as detailed in Algorithm 1. Finally, the left direction [12pt]{minimal} $$\:$$ can be obtained by taking the cross product of the gravity direction and the forward-backward direction. Subsequently, after obtaining the gravity reference direction [12pt]{minimal} $$\:_{fin}}$$ , the farthest point from [12pt]{minimal} $$\:_{fin}}$$ in the mandibular reconstruction model is determined to be [12pt]{minimal} $$\:C{o}_{L}$$ , [12pt]{minimal} $$\:C{o}_{R}$$ . A local coordinate system is established with [12pt]{minimal} $$\:{O}_{MB}$$ as the origin, [12pt]{minimal} $$\:$$ as the x -axis, [12pt]{minimal} $$\:_{fin}}$$ as the y -axis, and [12pt]{minimal} $$\:$$ as the z -axis. Subsequently, cutting planes are generated at intervals of [12pt]{minimal} $$\:p$$ , perpendicular to the direction of [12pt]{minimal} $$\:$$ , as illustrated in Fig. a. The [12pt]{minimal} $$\:i$$ -th cutting plane is denoted as [12pt]{minimal} $$\:{M}_{i}^{Cut}$$ , where [12pt]{minimal} $$\:1\:i\:{N}_{cut}$$ . [12pt]{minimal} $$\:{M}_{i}^{Cut}$$ cuts both the left and right mandibular models, resulting in separate cutting contours, which are denoted as [12pt]{minimal} $$\:{L}_{i}^{Left}$$ and [12pt]{minimal} $$\:{L}_{i}^{Right}$$ , respectively. For [12pt]{minimal} $$\:{L}_{i}^{Left}$$ , the contour is sparsely sampled to obtain a series of contour points, denoted as [12pt]{minimal} $$\:{p}_{i,j}^{Left}=\{{X}_{i,j}^{Left},\:{Y}_{i,j}^{Left},{Z}_{i,j}^{Left}\}$$ , [12pt]{minimal} $$\:1\:j\:{N}_{Left}$$ . The highest point on each segmented contour, which corresponds to the point with the maximum [12pt]{minimal} $$\:-{Y}_{i,j}^{Left}$$ , is denoted as [12pt]{minimal} $$\:{p}_{i}^{L\_high}=\{{X}_{i}^{L\_high},\:{Y}_{i}^{L\_high},{Z}_{i}^{L\_high}\}$$ . The absolute value of the interpolation in the y -axis direction between the highest point obtained from the i-th cutting plane and the highest point obtained from the [12pt]{minimal} $$\:(i-1)$$ -th cutting plane is defined as [12pt]{minimal} $$\:{\:H}_{i}=|{Y}_{i}^{L\_high}-{Y}_{i-1}^{L\_high}|,$$ where the maximum [12pt]{minimal} $$\:{\:H}_{i}$$ as corresponding to the highest point on the segmented contour, which is the left mandibular ascending ramus point and is represented as [12pt]{minimal} $$\:{Q}_{p}^{Left}$$ . Similarly, the right mandibular ascending ramus point, denoted as [12pt]{minimal} $$\:{Q}_{p}^{Right}$$ , can be obtained in the same manner. The cutting plane corresponding to the mandibular ascending ramus point is denoted as [12pt]{minimal} $$\:{M}_{Q}$$ , and a plane [12pt]{minimal} $$\:{M}_{b}$$ at a distance of [12pt]{minimal} $$\:{d}_{cut}$$ along the direction of [12pt]{minimal} $$\:$$ from [12pt]{minimal} $$\:{M}_{Q}$$ is constructed, as illustrated in Fig. b. The highest points on the cutting contours obtained by cutting the left and right sides of the mandibular bone model with [12pt]{minimal} $$\:{M}_{b}$$ are denoted as [12pt]{minimal} $$\:{M}_{o}^{Left}$$ and [12pt]{minimal} $$\:{M}_{o}^{Right}$$ , which correspond to the left and right first molar points. Additionally, the highest point on the groove, corresponding to the incisor, is designated as [12pt]{minimal} $$\:{I}_{a}$$ . These three points, [12pt]{minimal} $$\:{M}_{o}^{Left}$$ , [12pt]{minimal} $$\:{M}_{o}^{Right}$$ , and [12pt]{minimal} $$\:{I}_{a}$$ , can define the occlusal plane OP, as shown in Fig. c. To filter out the cut regions corresponding to the cutting plane and the dental area, OP is used to segment the mandibular model, two largest connected regions are extracted from the segmented contours, denoted as [12pt]{minimal} $$\:{C}_{Left}$$ and [12pt]{minimal} $$\:{C}_{Right}$$ , respectively. This step is employed. Taking the left boundary contour as an example, it is sparsely sampled to obtain boundary points, which can be represented as [12pt]{minimal} $$\:{C}_{Left}=\{{q}_{1}^{C},\:,{q}_{{N}_{q}}^{C}\}$$ , where [12pt]{minimal} $$\:{N}_{q}\:$$ is the total number of contour points. The distance from each contour point to [12pt]{minimal} $$\:{O}_{MB}$$ is calculated and denoted as [12pt]{minimal} $$\:dis\{{q}_{i}^{C},{O}_{MB}\}$$ . Let [12pt]{minimal} $$\:{O}_{Left}$$ be the contour point corresponding to the maximum [12pt]{minimal} $$\:dis\{{q}_{i}^{C},{O}_{MB}\}$$ , and let [12pt]{minimal} $$\:{{A}_{2}}_{Left}$$ be the contour point corresponding to the minimum [12pt]{minimal} $$\:dis\{{q}_{i}^{C},{O}_{MB}\}$$ . Similarly, [12pt]{minimal} $$\:{O}_{Right}$$ and [12pt]{minimal} $$\:{A}_{2Right}$$ can be determined for the right boundary contour. To obtain the posterior point [12pt]{minimal} $$\:Go$$ of the mandibular angle, a helper line [12pt]{minimal} $$\:{l}_{MA}$$ must be constructed first. As shown in Fig. a, [12pt]{minimal} $$\:{l}_{MA}$$ is defined as the line of intersection between the mandibular bone edge [12pt]{minimal} $$\:{S}_{down}$$ and the posterior surface of the mandibular ramus [12pt]{minimal} $$\:{S}_{back}$$ . The calculation of the mandibular bone edge surface [12pt]{minimal} $$\:{S}_{down}$$ is as follows: Find the angle between the normal vector [12pt]{minimal} $$\:$$ for all points on the surface of the mandible and the angle between the normal vector [12pt]{minimal} $$\:_{i}}$$ of point [12pt]{minimal} $$\:i$$ and the initial gravity direction [12pt]{minimal} $$\:_{ini}}$$ . If the angle [12pt]{minimal} $$\:{\:}_{i}$$ is less than the predetermined threshold [12pt]{minimal} $$\:\:{T}_{NV}$$ , the point [12pt]{minimal} $$\:i$$ is a point on the lower surface [12pt]{minimal} $$\:\:{M}_{down}$$ , which is the colored part of the mandibular bone lower edge in Fig. a. After obtaining all lower surface points, a filtering operation is applied, and the largest connected region is selected. The OBB bounding box of the new lower surface, [12pt]{minimal} $$\:OB{B}_{{M}_{down}}$$ , is then calculated, and the gravity direction of the OBB bounding box is used as the new gravity direction. Then, iterate until the gravity direction no longer changes or reaches the specified iteration count limit. The lower surface of the OBB bounding box generated in the last iteration is the edge surface [12pt]{minimal} $$\:{S}_{down}$$ , and the detailed steps can be seen in Algorithm 1 (11–23). Meanwhile, [12pt]{minimal} $$\:{G}_{o}$$ is the closest point to [12pt]{minimal} $$\:{l}_{MA}$$ in the preoperative reconstructed three-dimensional model of the mandibular bone, which can be calculated as follows: 2 [12pt]{minimal} $$\:Go=Dis({m}_{i},{l}_{MA})\:$$ where [12pt]{minimal} $$\:Dis()$$ represents the distance function, [12pt]{minimal} $$\:{m}_{i}\:M$$ . Thus, the posterior points of the left and right mandibular angles, [12pt]{minimal} $$\:G{o}_{Left}$$ and [12pt]{minimal} $$\:G{o}_{Right}$$ , can be obtained. Finally, as depicted in Fig. b, the preliminary position of the second molar point is determined, and the projections of the left and right subcondylar regions of the mandible are obtained along the specified direction of [12pt]{minimal} $$\:_{fin}}$$ , yielding [12pt]{minimal} $$\:{E}_{Left}$$ and [12pt]{minimal} $$\:{E}_{Right}$$ , respectively. For the lowest point of the ramus [12pt]{minimal} $$\:{M}_{e}$$ , it can be obtained by calculating the symmetrical plane of the mandible and cutting the lowest point of the mandible model. In conclusion, all the anatomical landmark points of the left and right sides of the mandible can be obtained, which are useful for surgical planning and parameter measurement. Acquisition of the mandibular angle osteotomy plane: automatic planning After identifying all the anatomical points of the mandible, it is necessary to formulate a preoperative planning scheme for the mandibular angle osteotomy by considering the anatomical relationships and postoperative facial appearance. This paper presents an automatic planning for the mandibular angle osteotomy through the optimization model of the cutting plane planning for the procedure. The left cutting plane for the mandible, represented as [12pt]{minimal} $$\:{W}_{Left}$$ and the right cutting plane for the mandible, represented as [12pt]{minimal} $$\:{W}_{Right},$$ can be represented as: 3 [12pt]{minimal} $$\:{A}_{Left}x+{B}_{Left}y+{C}_{Left}z+{D}_{Left}=0$$ 4 [12pt]{minimal} $$\:{A}_{Right}x+{B}_{Right}y+{C}_{Right}z+{D}_{Right}=0$$ To reduce postoperative complications, it is necessary to protect the mandibular nerve during surgery by avoiding interference between the cutting plane and the mandibular nerve in the preoperative planning. Taking the left side of the mandible as an example, enough gap between the nerve canal model reconstructed before the operation and the left cutting plane should be ensured. The nerve canal model can be represented as a sequence of nodes, [12pt]{minimal} $$\:N=\{{N}_{1},\:,{N}_{t}\},$$ where t is the total number of nodes in the nerve canal model. The objective function to be optimized can be represented as: 5 [12pt]{minimal} $$\:_{i=1}^{t}Dis({N}_{i},{W}_{left})$$ where [12pt]{minimal} $$\:Dis({N}_{i},{W}_{left})$$ represents the distance between any node [12pt]{minimal} $$\:{N}_{i}\:$$ in the nerve canal model and the cutting plane [12pt]{minimal} $$\:\:{W}_{left}$$ . As shown in Fig. b. To ensure postoperative effectiveness, it is necessary to maximize the area of the mandibular bone [12pt]{minimal} $$\:M$$ ’s osteotomy plane. Therefore, another optimization objective can be expressed as [12pt]{minimal} $$\:m{ax}S=Area\{{W}_{Left},M\}$$ , where [12pt]{minimal} $$\:Area\{{W}_{Left},M\}\:$$ represents the area of the osteotomy contour obtained by cutting [12pt]{minimal} $$\:M$$ with the left cutting plane [12pt]{minimal} $$\:{W}_{Left}$$ . According to the anatomical structural relationships, the cutting plane must pass through both anatomical landmarks E and [12pt]{minimal} $$\:{A}_{2}$$ , which means it needs to simultaneously satisfy: 6 [12pt]{minimal} $$\:s.t.\{_{Left}{x}_{E}+{B}_{Left}{y}_{E}+{C}_{Left}{z}_{E}+{D}_{Left}=0\\\:{A}_{Left}{x}_{{A}_{2}}+{B}_{Left}{y}_{{A}_{2}}+{C}_{Left}{z}_{{A}_{2}}+{D}_{Left}=0.$$ Then, the left cutting plane [12pt]{minimal} $$\:{W}_{Left}$$ can be obtained through an iterative optimization process. The objective function and constraints for the mandibular right cutting plane can be formulated similarly, eventually resulting in the determination of the right cutting plane [12pt]{minimal} $$\:{W}_{Right}$$ , as shown in Fig. a. Based on clinical experience, it is imperative to ensure that the distance from the cutting planes to the neurovascular canal is not less than 5 mm. Therefore, once the cutting planes [12pt]{minimal} $$\:{W}_{Left}$$ and [12pt]{minimal} $$\:{W}_{Right}$$ are determined, the shortest distances from points on the neurovascular canal to both [12pt]{minimal} $$\:{W}_{Left}$$ and [12pt]{minimal} $$\:{W}_{Right}$$ are calculated. If the distances are greater than 5 mm, these cutting planes are accepted as the final ones. If the distances are less than 5 mm, [12pt]{minimal} $$\:{W}_{Left}$$ and [12pt]{minimal} $$\:{W}_{Right}$$ are translated in the opposite direction of the neurovascular canal by 5 mm, and the resulting [12pt]{minimal} $$\:{W}_{Left}$$ and [12pt]{minimal} $$\:{W}_{Right}$$ are considered as the final cutting planes. Due to the discrepancy between the automatically recognized anatomical landmarks and their true values, manual adjustment by the physician is necessary during the planning process. The system provides interactive adjustment capabilities for anatomical landmarks, and the cutting plane of the surgical plan will be updated in real-time during the adjustment process until it meets the clinical surgical requirements. Anatomical parameter measurement and evaluation methods To assess the effectiveness of the planning scheme, it is crucial to measure the key parameters of the mandible before and after the planning process. As shown in Fig. (c), the important parameters to be measured specifically include: [12pt]{minimal} $$\:{l}^{Co-Go}$$ , defined as the distance from the most top point of the condylar to the most bottom point of the mandibular angle. The left and right sides are represented as [12pt]{minimal} $$\:{l}_{Left}^{Co-Go}$$ and [12pt]{minimal} $$\:{l}_{Right}^{Co-Go}$$ , respectively. [12pt]{minimal} $$\:{l}^{Co-G{o}^{{\:}}}$$ , defined as the distance from the most top point of the condylar to the newly cut mandibular angle’s most bottom point, where [12pt]{minimal} $$\:G{o}^{{\:}}$$ represents the new mandibular angle’s most bottom point. The left and right sides are represented as [12pt]{minimal} $$\:\:{l}_{Left}^{Co-G{o}^{{\:}}}$$ and [12pt]{minimal} $$\:{l}_{Right}^{Co-G{o}^{{\:}}}$$ , respectively. [12pt]{minimal} $$\:{l}^{Go-Go}$$ , defined as the distance between the post-bottom points of the left and right mandibular angles. [12pt]{minimal} $$\:{l}^{G{o}^{{\:}}-G{o}^{{\:}}}$$ , defined as the distance between the newly cut mandibular angle’s most bottom points on the left and right sides. Mandibular angle angle [12pt]{minimal} $$\:\:(Co-Go-ME)$$ , defined as the angle formed by the lines connecting the most top point of the condylar, the most bottom point of the mandibular angle, and the submental point. The left and right sides are represented as [12pt]{minimal} $$\:\:(Co-Go-ME{)}_{Left}$$ and [12pt]{minimal} $$\:\:(Co-Go-ME{)}_{Right}$$ , respectively. New mandibular angle angle [12pt]{minimal} $$\:\:(Co-G{o}^{{\:}}-ME)$$ , defined as the angle formed by the lines connecting the most top point of the condylar, the newly cut mandibular angle’s most bottom point, and the submental point. The left and right sides are represented as [12pt]{minimal} $$\:\:(Co-G{o}^{{\:}}-ME{)}_{Left}$$ and [12pt]{minimal} $$\:\:(Co-G{o}^{{\:}}-ME{)}_{Right}$$ , respectively. By obtaining the above parameters and combining them with the doctor’s clinical experience, the postoperative patient’s mandibular bone morphology can be quantitatively evaluated and compared with the preoperative morphology. Additionally, the surgical planning scheme can be further adjusted based on the evaluation results. Software system integration The algorithms were finally integrated into a software for preoperative planning of mandibular angle osteotomy. The software was developed in a Python environment and utilized the following open-source libraries: VTK ( https://www.vtk.org ) for display, SimpleITK ( https://simpleitk.org/ ) for image processing, numpy ( https://numpy.org/ ) for numerical computation, and PySide2 ( https://pypi.org/project/PySide2/ ) for user interface. The software comprises of five functional modules: data import, image segmentation and reconstruction, automatic feature point recognition, surgical planning, measurement, and result export. The overall operating procedure of the system is outlined as follows: Imports patient’s cranial-facial CT image data. Then, the deep learning-based automatic segmentation module is utilized to achieve automatic segmentation of the mandibular bone region and the mandibular nerve canal, and manual adjustment is performed on the poorly segmented regions. Finally, a 3D model of the mandibular bone and mandibular nerve canal is established. Uses the landmark recognition algorithm to automatically recognize the key anatomical landmarks of the mandibular bone and automatically generates an initial osteotomy plan. Manually fine-tunes the recognized feature points, adjusts the planning plane accordingly, until the clinical requirements are met. Measures the important anatomical parameters before and after the planning and evaluates the surgical plan. Exports the data of the reconstructed model and planning plane and generates a surgical planning report. The purpose of mandibular anatomical marker point identification is to facilitate the planning of the surgical osteotomy plane and the measurement of key anatomical parameters. These anatomical markers are particularly significant in defining the shape of the mandible, especially in the region near the mandibular angle. As shown in Fig. , the planning of mandibular angle osteotomy requires the identification of key anatomical landmark points on each of the left and right sides of the mandible. These include the condylar apex (Condylion, Co), the mandibular angle point (Gonion, Go), the subchin point (Meton, Me), the projection of the second molar point on the inferior margin of the mandible (point E), the intersection of the occlusal plane with the external mandibular oblique line (point), and the intersection of the occlusal plane with the ramus of the mandible (point O), as well as a characteristic plane: the occlusal plane (OP). Definitions for these points are provided in Table . To detect these features, the HOBB (Hierarchical Oriented Bounding Box)-based method is proposed. The HOBB is a tool in the field of computer collision detection, and the author discovered that the shape of the mandible resembles a half-circle with two upward supports. The HOBB calculates a surface that splits the volume of the three-dimensional object into two equal parts. Therefore, if the first-order bounding box of the mandible is calculated, due to its symmetry, the split surface will be close to the mid-surface; if the second-order bounding box is calculated, due to the anatomical features of the mandible (width of the mandibular body is close to the width of the mandibular ramus, and the length of the mandibular body is approximately twice the length of the mandibular ramus), the volume of each side of the mandibular ramus and mandibular body will be approximately equal, with the split surface positioned centrally between them. As a result, the second-order bounding box of the mandible roughly divides the mandible into the left and right mandibular ramus and the left and right mandibular body. Therefore, through repeated research, this paper proposes a HOBB-based method for automatic recognition of the anatomical structure of the mandible, which is as follows: First, obtain the overall mandible model’s bounding box [12pt]{minimal} $$\:OB{B}_{MB}$$ and its center [12pt]{minimal} $$\:{O}_{MB}$$ . Then, model the reconstructed three-dimensional mandible model with HOBB, and obtain four bounding boxes (as shown in Fig. b), referred to as [12pt]{minimal} $$\:OB{B}_{MAL}$$ (located in the left mandibular arch area), [12pt]{minimal} $$\:OB{B}_{MAR}$$ (located in the right mandibular arch area), [12pt]{minimal} $$\:OB{B}_{MBL}$$ (located in the left mandibular body area), and [12pt]{minimal} $$\:OB{B}_{MBR}$$ (located in the right mandibular body area). After obtaining the left and right mandibular bodies, the cross product of the normal vectors of the longest axes of their bounding boxes is computed to determine the approximate direction of the mandible’s center of gravity [12pt]{minimal} $$\:_{fin}}.$$ The detailed method is as follows: First, the reference direction of the mandible center of gravity is established using [12pt]{minimal} $$\:OB{B}_{MBL}$$ and [12pt]{minimal} $$\:OB{B}_{MBR}.$$ Its initial value can be represented as: 1 [12pt]{minimal} $$\:_{ini}}=_{MBL}}\:_{MBR}}$$ The long axis vectors of [12pt]{minimal} $$\:_{MBL}}$$ and [12pt]{minimal} $$\:_{MBR}}$$ are respectively denoted as [12pt]{minimal} $$\:\:OB{B}_{MBL}$$ and [12pt]{minimal} $$\:OB{B}_{MBR}$$ . The final reference direction of the center of gravity [12pt]{minimal} $$\:_{fin}}$$ is obtained through an iterative optimization method as detailed in Algorithm 1. Finally, the left direction [12pt]{minimal} $$\:$$ can be obtained by taking the cross product of the gravity direction and the forward-backward direction. Subsequently, after obtaining the gravity reference direction [12pt]{minimal} $$\:_{fin}}$$ , the farthest point from [12pt]{minimal} $$\:_{fin}}$$ in the mandibular reconstruction model is determined to be [12pt]{minimal} $$\:C{o}_{L}$$ , [12pt]{minimal} $$\:C{o}_{R}$$ . A local coordinate system is established with [12pt]{minimal} $$\:{O}_{MB}$$ as the origin, [12pt]{minimal} $$\:$$ as the x -axis, [12pt]{minimal} $$\:_{fin}}$$ as the y -axis, and [12pt]{minimal} $$\:$$ as the z -axis. Subsequently, cutting planes are generated at intervals of [12pt]{minimal} $$\:p$$ , perpendicular to the direction of [12pt]{minimal} $$\:$$ , as illustrated in Fig. a. The [12pt]{minimal} $$\:i$$ -th cutting plane is denoted as [12pt]{minimal} $$\:{M}_{i}^{Cut}$$ , where [12pt]{minimal} $$\:1\:i\:{N}_{cut}$$ . [12pt]{minimal} $$\:{M}_{i}^{Cut}$$ cuts both the left and right mandibular models, resulting in separate cutting contours, which are denoted as [12pt]{minimal} $$\:{L}_{i}^{Left}$$ and [12pt]{minimal} $$\:{L}_{i}^{Right}$$ , respectively. For [12pt]{minimal} $$\:{L}_{i}^{Left}$$ , the contour is sparsely sampled to obtain a series of contour points, denoted as [12pt]{minimal} $$\:{p}_{i,j}^{Left}=\{{X}_{i,j}^{Left},\:{Y}_{i,j}^{Left},{Z}_{i,j}^{Left}\}$$ , [12pt]{minimal} $$\:1\:j\:{N}_{Left}$$ . The highest point on each segmented contour, which corresponds to the point with the maximum [12pt]{minimal} $$\:-{Y}_{i,j}^{Left}$$ , is denoted as [12pt]{minimal} $$\:{p}_{i}^{L\_high}=\{{X}_{i}^{L\_high},\:{Y}_{i}^{L\_high},{Z}_{i}^{L\_high}\}$$ . The absolute value of the interpolation in the y -axis direction between the highest point obtained from the i-th cutting plane and the highest point obtained from the [12pt]{minimal} $$\:(i-1)$$ -th cutting plane is defined as [12pt]{minimal} $$\:{\:H}_{i}=|{Y}_{i}^{L\_high}-{Y}_{i-1}^{L\_high}|,$$ where the maximum [12pt]{minimal} $$\:{\:H}_{i}$$ as corresponding to the highest point on the segmented contour, which is the left mandibular ascending ramus point and is represented as [12pt]{minimal} $$\:{Q}_{p}^{Left}$$ . Similarly, the right mandibular ascending ramus point, denoted as [12pt]{minimal} $$\:{Q}_{p}^{Right}$$ , can be obtained in the same manner. The cutting plane corresponding to the mandibular ascending ramus point is denoted as [12pt]{minimal} $$\:{M}_{Q}$$ , and a plane [12pt]{minimal} $$\:{M}_{b}$$ at a distance of [12pt]{minimal} $$\:{d}_{cut}$$ along the direction of [12pt]{minimal} $$\:$$ from [12pt]{minimal} $$\:{M}_{Q}$$ is constructed, as illustrated in Fig. b. The highest points on the cutting contours obtained by cutting the left and right sides of the mandibular bone model with [12pt]{minimal} $$\:{M}_{b}$$ are denoted as [12pt]{minimal} $$\:{M}_{o}^{Left}$$ and [12pt]{minimal} $$\:{M}_{o}^{Right}$$ , which correspond to the left and right first molar points. Additionally, the highest point on the groove, corresponding to the incisor, is designated as [12pt]{minimal} $$\:{I}_{a}$$ . These three points, [12pt]{minimal} $$\:{M}_{o}^{Left}$$ , [12pt]{minimal} $$\:{M}_{o}^{Right}$$ , and [12pt]{minimal} $$\:{I}_{a}$$ , can define the occlusal plane OP, as shown in Fig. c. To filter out the cut regions corresponding to the cutting plane and the dental area, OP is used to segment the mandibular model, two largest connected regions are extracted from the segmented contours, denoted as [12pt]{minimal} $$\:{C}_{Left}$$ and [12pt]{minimal} $$\:{C}_{Right}$$ , respectively. This step is employed. Taking the left boundary contour as an example, it is sparsely sampled to obtain boundary points, which can be represented as [12pt]{minimal} $$\:{C}_{Left}=\{{q}_{1}^{C},\:,{q}_{{N}_{q}}^{C}\}$$ , where [12pt]{minimal} $$\:{N}_{q}\:$$ is the total number of contour points. The distance from each contour point to [12pt]{minimal} $$\:{O}_{MB}$$ is calculated and denoted as [12pt]{minimal} $$\:dis\{{q}_{i}^{C},{O}_{MB}\}$$ . Let [12pt]{minimal} $$\:{O}_{Left}$$ be the contour point corresponding to the maximum [12pt]{minimal} $$\:dis\{{q}_{i}^{C},{O}_{MB}\}$$ , and let [12pt]{minimal} $$\:{{A}_{2}}_{Left}$$ be the contour point corresponding to the minimum [12pt]{minimal} $$\:dis\{{q}_{i}^{C},{O}_{MB}\}$$ . Similarly, [12pt]{minimal} $$\:{O}_{Right}$$ and [12pt]{minimal} $$\:{A}_{2Right}$$ can be determined for the right boundary contour. To obtain the posterior point [12pt]{minimal} $$\:Go$$ of the mandibular angle, a helper line [12pt]{minimal} $$\:{l}_{MA}$$ must be constructed first. As shown in Fig. a, [12pt]{minimal} $$\:{l}_{MA}$$ is defined as the line of intersection between the mandibular bone edge [12pt]{minimal} $$\:{S}_{down}$$ and the posterior surface of the mandibular ramus [12pt]{minimal} $$\:{S}_{back}$$ . The calculation of the mandibular bone edge surface [12pt]{minimal} $$\:{S}_{down}$$ is as follows: Find the angle between the normal vector [12pt]{minimal} $$\:$$ for all points on the surface of the mandible and the angle between the normal vector [12pt]{minimal} $$\:_{i}}$$ of point [12pt]{minimal} $$\:i$$ and the initial gravity direction [12pt]{minimal} $$\:_{ini}}$$ . If the angle [12pt]{minimal} $$\:{\:}_{i}$$ is less than the predetermined threshold [12pt]{minimal} $$\:\:{T}_{NV}$$ , the point [12pt]{minimal} $$\:i$$ is a point on the lower surface [12pt]{minimal} $$\:\:{M}_{down}$$ , which is the colored part of the mandibular bone lower edge in Fig. a. After obtaining all lower surface points, a filtering operation is applied, and the largest connected region is selected. The OBB bounding box of the new lower surface, [12pt]{minimal} $$\:OB{B}_{{M}_{down}}$$ , is then calculated, and the gravity direction of the OBB bounding box is used as the new gravity direction. Then, iterate until the gravity direction no longer changes or reaches the specified iteration count limit. The lower surface of the OBB bounding box generated in the last iteration is the edge surface [12pt]{minimal} $$\:{S}_{down}$$ , and the detailed steps can be seen in Algorithm 1 (11–23). Meanwhile, [12pt]{minimal} $$\:{G}_{o}$$ is the closest point to [12pt]{minimal} $$\:{l}_{MA}$$ in the preoperative reconstructed three-dimensional model of the mandibular bone, which can be calculated as follows: 2 [12pt]{minimal} $$\:Go=Dis({m}_{i},{l}_{MA})\:$$ where [12pt]{minimal} $$\:Dis()$$ represents the distance function, [12pt]{minimal} $$\:{m}_{i}\:M$$ . Thus, the posterior points of the left and right mandibular angles, [12pt]{minimal} $$\:G{o}_{Left}$$ and [12pt]{minimal} $$\:G{o}_{Right}$$ , can be obtained. Finally, as depicted in Fig. b, the preliminary position of the second molar point is determined, and the projections of the left and right subcondylar regions of the mandible are obtained along the specified direction of [12pt]{minimal} $$\:_{fin}}$$ , yielding [12pt]{minimal} $$\:{E}_{Left}$$ and [12pt]{minimal} $$\:{E}_{Right}$$ , respectively. For the lowest point of the ramus [12pt]{minimal} $$\:{M}_{e}$$ , it can be obtained by calculating the symmetrical plane of the mandible and cutting the lowest point of the mandible model. In conclusion, all the anatomical landmark points of the left and right sides of the mandible can be obtained, which are useful for surgical planning and parameter measurement. After identifying all the anatomical points of the mandible, it is necessary to formulate a preoperative planning scheme for the mandibular angle osteotomy by considering the anatomical relationships and postoperative facial appearance. This paper presents an automatic planning for the mandibular angle osteotomy through the optimization model of the cutting plane planning for the procedure. The left cutting plane for the mandible, represented as [12pt]{minimal} $$\:{W}_{Left}$$ and the right cutting plane for the mandible, represented as [12pt]{minimal} $$\:{W}_{Right},$$ can be represented as: 3 [12pt]{minimal} $$\:{A}_{Left}x+{B}_{Left}y+{C}_{Left}z+{D}_{Left}=0$$ 4 [12pt]{minimal} $$\:{A}_{Right}x+{B}_{Right}y+{C}_{Right}z+{D}_{Right}=0$$ To reduce postoperative complications, it is necessary to protect the mandibular nerve during surgery by avoiding interference between the cutting plane and the mandibular nerve in the preoperative planning. Taking the left side of the mandible as an example, enough gap between the nerve canal model reconstructed before the operation and the left cutting plane should be ensured. The nerve canal model can be represented as a sequence of nodes, [12pt]{minimal} $$\:N=\{{N}_{1},\:,{N}_{t}\},$$ where t is the total number of nodes in the nerve canal model. The objective function to be optimized can be represented as: 5 [12pt]{minimal} $$\:_{i=1}^{t}Dis({N}_{i},{W}_{left})$$ where [12pt]{minimal} $$\:Dis({N}_{i},{W}_{left})$$ represents the distance between any node [12pt]{minimal} $$\:{N}_{i}\:$$ in the nerve canal model and the cutting plane [12pt]{minimal} $$\:\:{W}_{left}$$ . As shown in Fig. b. To ensure postoperative effectiveness, it is necessary to maximize the area of the mandibular bone [12pt]{minimal} $$\:M$$ ’s osteotomy plane. Therefore, another optimization objective can be expressed as [12pt]{minimal} $$\:m{ax}S=Area\{{W}_{Left},M\}$$ , where [12pt]{minimal} $$\:Area\{{W}_{Left},M\}\:$$ represents the area of the osteotomy contour obtained by cutting [12pt]{minimal} $$\:M$$ with the left cutting plane [12pt]{minimal} $$\:{W}_{Left}$$ . According to the anatomical structural relationships, the cutting plane must pass through both anatomical landmarks E and [12pt]{minimal} $$\:{A}_{2}$$ , which means it needs to simultaneously satisfy: 6 [12pt]{minimal} $$\:s.t.\{_{Left}{x}_{E}+{B}_{Left}{y}_{E}+{C}_{Left}{z}_{E}+{D}_{Left}=0\\\:{A}_{Left}{x}_{{A}_{2}}+{B}_{Left}{y}_{{A}_{2}}+{C}_{Left}{z}_{{A}_{2}}+{D}_{Left}=0.$$ Then, the left cutting plane [12pt]{minimal} $$\:{W}_{Left}$$ can be obtained through an iterative optimization process. The objective function and constraints for the mandibular right cutting plane can be formulated similarly, eventually resulting in the determination of the right cutting plane [12pt]{minimal} $$\:{W}_{Right}$$ , as shown in Fig. a. Based on clinical experience, it is imperative to ensure that the distance from the cutting planes to the neurovascular canal is not less than 5 mm. Therefore, once the cutting planes [12pt]{minimal} $$\:{W}_{Left}$$ and [12pt]{minimal} $$\:{W}_{Right}$$ are determined, the shortest distances from points on the neurovascular canal to both [12pt]{minimal} $$\:{W}_{Left}$$ and [12pt]{minimal} $$\:{W}_{Right}$$ are calculated. If the distances are greater than 5 mm, these cutting planes are accepted as the final ones. If the distances are less than 5 mm, [12pt]{minimal} $$\:{W}_{Left}$$ and [12pt]{minimal} $$\:{W}_{Right}$$ are translated in the opposite direction of the neurovascular canal by 5 mm, and the resulting [12pt]{minimal} $$\:{W}_{Left}$$ and [12pt]{minimal} $$\:{W}_{Right}$$ are considered as the final cutting planes. Due to the discrepancy between the automatically recognized anatomical landmarks and their true values, manual adjustment by the physician is necessary during the planning process. The system provides interactive adjustment capabilities for anatomical landmarks, and the cutting plane of the surgical plan will be updated in real-time during the adjustment process until it meets the clinical surgical requirements. To assess the effectiveness of the planning scheme, it is crucial to measure the key parameters of the mandible before and after the planning process. As shown in Fig. (c), the important parameters to be measured specifically include: [12pt]{minimal} $$\:{l}^{Co-Go}$$ , defined as the distance from the most top point of the condylar to the most bottom point of the mandibular angle. The left and right sides are represented as [12pt]{minimal} $$\:{l}_{Left}^{Co-Go}$$ and [12pt]{minimal} $$\:{l}_{Right}^{Co-Go}$$ , respectively. [12pt]{minimal} $$\:{l}^{Co-G{o}^{{\:}}}$$ , defined as the distance from the most top point of the condylar to the newly cut mandibular angle’s most bottom point, where [12pt]{minimal} $$\:G{o}^{{\:}}$$ represents the new mandibular angle’s most bottom point. The left and right sides are represented as [12pt]{minimal} $$\:\:{l}_{Left}^{Co-G{o}^{{\:}}}$$ and [12pt]{minimal} $$\:{l}_{Right}^{Co-G{o}^{{\:}}}$$ , respectively. [12pt]{minimal} $$\:{l}^{Go-Go}$$ , defined as the distance between the post-bottom points of the left and right mandibular angles. [12pt]{minimal} $$\:{l}^{G{o}^{{\:}}-G{o}^{{\:}}}$$ , defined as the distance between the newly cut mandibular angle’s most bottom points on the left and right sides. Mandibular angle angle [12pt]{minimal} $$\:\:(Co-Go-ME)$$ , defined as the angle formed by the lines connecting the most top point of the condylar, the most bottom point of the mandibular angle, and the submental point. The left and right sides are represented as [12pt]{minimal} $$\:\:(Co-Go-ME{)}_{Left}$$ and [12pt]{minimal} $$\:\:(Co-Go-ME{)}_{Right}$$ , respectively. New mandibular angle angle [12pt]{minimal} $$\:\:(Co-G{o}^{{\:}}-ME)$$ , defined as the angle formed by the lines connecting the most top point of the condylar, the newly cut mandibular angle’s most bottom point, and the submental point. The left and right sides are represented as [12pt]{minimal} $$\:\:(Co-G{o}^{{\:}}-ME{)}_{Left}$$ and [12pt]{minimal} $$\:\:(Co-G{o}^{{\:}}-ME{)}_{Right}$$ , respectively. By obtaining the above parameters and combining them with the doctor’s clinical experience, the postoperative patient’s mandibular bone morphology can be quantitatively evaluated and compared with the preoperative morphology. Additionally, the surgical planning scheme can be further adjusted based on the evaluation results. The algorithms were finally integrated into a software for preoperative planning of mandibular angle osteotomy. The software was developed in a Python environment and utilized the following open-source libraries: VTK ( https://www.vtk.org ) for display, SimpleITK ( https://simpleitk.org/ ) for image processing, numpy ( https://numpy.org/ ) for numerical computation, and PySide2 ( https://pypi.org/project/PySide2/ ) for user interface. The software comprises of five functional modules: data import, image segmentation and reconstruction, automatic feature point recognition, surgical planning, measurement, and result export. The overall operating procedure of the system is outlined as follows: Imports patient’s cranial-facial CT image data. Then, the deep learning-based automatic segmentation module is utilized to achieve automatic segmentation of the mandibular bone region and the mandibular nerve canal, and manual adjustment is performed on the poorly segmented regions. Finally, a 3D model of the mandibular bone and mandibular nerve canal is established. Uses the landmark recognition algorithm to automatically recognize the key anatomical landmarks of the mandibular bone and automatically generates an initial osteotomy plan. Manually fine-tunes the recognized feature points, adjusts the planning plane accordingly, until the clinical requirements are met. Measures the important anatomical parameters before and after the planning and evaluates the surgical plan. Exports the data of the reconstructed model and planning plane and generates a surgical planning report. Experimental data With the approval of the Ethics Committee of the Ninth People’s Hospital, affiliated with Shanghai Jiao Tong University School of Medicine (Ethics Code: SH9H-2020-T267-3), preoperative CT images were collected from 100 patients who underwent mandibular angle resection. There were 32 male and 68 female patients, with ages ranging from 20 to 48 years and an average age of 35 years. The resolution of the CT images was 0.45 mm/pixel and the slice thickness was 1.25 mm. Our deep learning-based mandibular bone segmentation algorithm was trained on 100 CT data and manually annotated data collected from the same hospital. The algorithm achieved a Dice error of 0.926 for segmentation, which meets the accuracy requirements for clinical application. Evaluation of the accuracy of automatic identification of anatomical landmarks and measurement of anatomical structures This paper evaluates the accuracy of the system by assessing the recognition precision of anatomic landmarks. To this end, 100 mandibular angle resection cases were utilized and the anatomic landmarks were automatically recognized by the open surgical planning system. Meanwhile, experienced clinical doctors manually annotated the positions of all anatomic landmarks on the three-dimensional model of the mandible, serving as the ground truth for the recognition algorithm. A one-to-one correspondence between the automatically recognized landmarks and ground truth landmarks was established and root mean square (RMS) error was used as the accuracy evaluation indicator. The results of the accuracy experiment are shown in Fig. (a). The average RMS error of the proposed landmark recognition algorithm is 1.87 ± 0.33 mm, with a minimum error value of 0.63 mm and a maximum error value of 3.77 mm. In addition, the feasibility of the planning scheme was also evaluated as shown in Fig. (b), with the statistical results of the resection amount and the distance between the nerve canal and the planning plane for all cases on both sides. The distance between the nerve canal and the planning plane, along with controlling the size of the resection amount, is also a factor that needs to be considered in surgical planning, with a general recommendation of not exceeding 600mm 3 . The distance between the planning plane and the nerve canal is the most critical parameter in surgery, ensuring its safety, and is a key parameter for preoperative planning evaluation, generally requiring a minimum of 3 mm. Clinical Experimentation and evaluation The validity of the automatic measurement algorithm is verified in this study, as shown in Table , which summarizes the measurement results for the important parameters of the mandible across all cases. The average time for automatic surgical planning was 57 s, with the core algorithm taking 12.7 s to detect the anatomical landmarks and osteotomy plane. The remaining time was used for the surgeon’s operation in the software and final confirmation. In contrast, manual planning by an experienced surgeon typically takes over 20 min, based on a five-year analysis of 231 cases performed by 10 surgeons in our hospital’s Department of Plastic and Reconstructive Surgery and the Department of Oral Maxillofacial - Head & Neck Oncology. To validate the efficacy of the mandibular angle resection surgery planning system in clinical practice, three trials were conducted in collaboration with medical professionals. For each patient, a surgical plan was developed using the system, and the actual surgeries were performed by the medical team. 3D-printed templates were created based on the surgical plans to guide the procedure. These templates were designed to fit precisely onto the patient’s mandible and featured a groove computed from the osteotomy plane to guide the saw during the osteotomy. During the surgery, the patients underwent CT scans and the post-operative mandibular bone models were reconstructed. The results were then compared with the pre-operative planning model, and the average surface distance error was calculated. Figure a shows the pre-operative surgical plan for the patient, and the same patient’s CT results before and after the surgery were registered using iterative closest point method (Geomagic Control X, 3D System, USA) as shown in Fig. b, which demonstrated reasonable error analysis within the acceptable range. Figure c and d show the post-operative appearance of the patient, demonstrating that the expected outcome was achieved as the surgery was carried out according to the automatic planning algorithm. With the approval of the Ethics Committee of the Ninth People’s Hospital, affiliated with Shanghai Jiao Tong University School of Medicine (Ethics Code: SH9H-2020-T267-3), preoperative CT images were collected from 100 patients who underwent mandibular angle resection. There were 32 male and 68 female patients, with ages ranging from 20 to 48 years and an average age of 35 years. The resolution of the CT images was 0.45 mm/pixel and the slice thickness was 1.25 mm. Our deep learning-based mandibular bone segmentation algorithm was trained on 100 CT data and manually annotated data collected from the same hospital. The algorithm achieved a Dice error of 0.926 for segmentation, which meets the accuracy requirements for clinical application. This paper evaluates the accuracy of the system by assessing the recognition precision of anatomic landmarks. To this end, 100 mandibular angle resection cases were utilized and the anatomic landmarks were automatically recognized by the open surgical planning system. Meanwhile, experienced clinical doctors manually annotated the positions of all anatomic landmarks on the three-dimensional model of the mandible, serving as the ground truth for the recognition algorithm. A one-to-one correspondence between the automatically recognized landmarks and ground truth landmarks was established and root mean square (RMS) error was used as the accuracy evaluation indicator. The results of the accuracy experiment are shown in Fig. (a). The average RMS error of the proposed landmark recognition algorithm is 1.87 ± 0.33 mm, with a minimum error value of 0.63 mm and a maximum error value of 3.77 mm. In addition, the feasibility of the planning scheme was also evaluated as shown in Fig. (b), with the statistical results of the resection amount and the distance between the nerve canal and the planning plane for all cases on both sides. The distance between the nerve canal and the planning plane, along with controlling the size of the resection amount, is also a factor that needs to be considered in surgical planning, with a general recommendation of not exceeding 600mm 3 . The distance between the planning plane and the nerve canal is the most critical parameter in surgery, ensuring its safety, and is a key parameter for preoperative planning evaluation, generally requiring a minimum of 3 mm. The validity of the automatic measurement algorithm is verified in this study, as shown in Table , which summarizes the measurement results for the important parameters of the mandible across all cases. The average time for automatic surgical planning was 57 s, with the core algorithm taking 12.7 s to detect the anatomical landmarks and osteotomy plane. The remaining time was used for the surgeon’s operation in the software and final confirmation. In contrast, manual planning by an experienced surgeon typically takes over 20 min, based on a five-year analysis of 231 cases performed by 10 surgeons in our hospital’s Department of Plastic and Reconstructive Surgery and the Department of Oral Maxillofacial - Head & Neck Oncology. To validate the efficacy of the mandibular angle resection surgery planning system in clinical practice, three trials were conducted in collaboration with medical professionals. For each patient, a surgical plan was developed using the system, and the actual surgeries were performed by the medical team. 3D-printed templates were created based on the surgical plans to guide the procedure. These templates were designed to fit precisely onto the patient’s mandible and featured a groove computed from the osteotomy plane to guide the saw during the osteotomy. During the surgery, the patients underwent CT scans and the post-operative mandibular bone models were reconstructed. The results were then compared with the pre-operative planning model, and the average surface distance error was calculated. Figure a shows the pre-operative surgical plan for the patient, and the same patient’s CT results before and after the surgery were registered using iterative closest point method (Geomagic Control X, 3D System, USA) as shown in Fig. b, which demonstrated reasonable error analysis within the acceptable range. Figure c and d show the post-operative appearance of the patient, demonstrating that the expected outcome was achieved as the surgery was carried out according to the automatic planning algorithm. In order to achieve precise and reliable mandibular osteotomy, preoperative planning that is personalized to the patient is extremely necessary. Traditional manual – or semi-automatic , mandibular osteotomy planning methods, the results of which depend on the surgeon’s experience, are less efficient and prolong the preoperative preparation time. For maxillofacial surgical planning, recognition of anatomical landmark points is a crucial step, and the accuracy and robustness of this recognition has important implications for the reliability of the surgical plan, as well as being a prerequisite for achieving automatic surgical planning. In this paper, a mandibular osteotomy automatic surgical planning method based on automatic recognition of mandibular landmark points is presented. The proposed method includes all processes from medical image segmentation and reconstruction to automatic recognition of anatomical landmark points, to plan generation and measurement of key anatomical parameters, and all algorithms can be automatically implemented. The surgeon only needs to confirm the plan and make simple adjustments, effectively reducing the efficiency of preoperative planning. In recent years, some artificial intelligence (AI) based automatic surgical planning methods for mandibular angle osteotomy have been proposed. Our team , proposed an AI-based mandibular point cloud automatic segmentation algorithm to generate osteotomy planning schemes. Lang et al. implemented automatic recognition of landmark points in CBCT images of the craniofacial region based on 3D Mask R-CNN network. Liu et al. proposed a multi-node CNN network SkullEngine, which realizes automatic segmentation and landmark point detection of craniofacial CBCT images. Despite the results of deep learning-based marker point approach, large data sample sets for personalized patients have significant bias for different central treatment protocols, leading to the disadvantage that training results of various deep learning algorithms are difficult to interpret and have inconsistent matching for surgical design options for new patients . Our proposed method does not rely on large-scale training data to achieve the identification of all anatomical landmark points in personalized mandibular osteotomy with accuracy and robustness that meet clinical requirements. Using the way based on surgical principles, the precise identification of some anatomically clear sites allows the surgeon to design the surgical plan based on the above-mentioned clear anatomical points, which is more in line with the actual situation of clinical treatment and the surgeon’s thinking habits, thus making the algorithm more interpretable and robust. This paper presents a comprehensive evaluation of the accuracy of anatomical marker point identification and surgical planning. The results show that the average RMS error of the automatic recognition algorithm is 1.87 ± 0.33 mm, and the results of the surgical planning scheme can meet the clinical needs. The errors of the planning scheme mainly originated from the following aspects: (1) A deep learning-based automatic segmentation algorithm was employed for the segmentation of the mandibular region. Although this method demonstrated strong performance, errors were still introduced during the segmentation process. (2) Our point identification was based on the preoperative reconstructed mandibular model, and reconstruction errors were introduced during the reconstruction process, including those related to mesh resolution. Higher mesh resolution results in smaller positioning errors because the mandible’s shape is more accurately described by the mesh. (3) Mark identification errors. There is a mathematical relationship between our final obtained surgical planning plane and multiple marker points, which may lead to the accumulation of errors. There are some limitations to the method presented in this paper. First, only mandibular osteotomy in normal patients is considered. However, for special cases such as mandibular lesions, asymmetry, and tumor erosion, the accuracy and robustness of the method require further validation. Additionally, the method uses the bony structure of the mandible and does not incorporate teeth information, which could provide more data when detecting anatomical landmarks. For better clinical application, the results of this method still need to be confirmed by the surgeon twice, even after the automatic surgical planning is completed. In cases of unreasonable anatomical marker points or planning planes, manual fine-tuning is necessary. In the future, more clinical experimental studies will be conducted to further evaluate the comprehensive performance of the proposed method. |
Drug‐Induced Liver Injury Caused by Metamizole: Identification of a Characteristic Injury Pattern | eb8d5cc2-8648-4786-b979-ff8722049f77 | 11801327 | Surgical Procedures, Operative[mh] | Introduction Metamizole is a widely used analgesic agent in many regions around the world, although it has been banned from the market in a few of countries due to a high risk for agranulocytosis . In addition to this rare but life‐threatening adverse event, drug‐induced liver injury (DILI) caused by metamizole has recently gained increasing attention . After sporadic reporting on metamizole causing liver injury, two larger case series were published in Germany reporting on a total of 55 cases with metamizole‐induced liver injury . These publications have raised awareness among clinicians and led to an official Healthcare Professional Communication Letter by the German Federal Institute for Drugs and Medical Devices warning of DILI caused by metamizole. However, despite the increasing knowledge on metamizole DILI, causality assessment can be a challenge. Since metamizole is usually taken together with other medications in patients with pain, acute infections, or undergoing medical procedures, identification of the causative agent in the case of this simultaneous multiple drug intake can pose difficulties. For a variety of medications, typical signature patterns have been proposed that can help to identify the causative medication in the case of polypharmacy and therefore aid to avoid discontinuation of innocent bystander agents . We therefore aimed to further evaluate metamizole DILI cases with the goal of identify characteristic patterns that can help differential diagnosis and causality assessment in the future. In addition, we investigated the predictive value of baseline parameters as well as the time‐dependent dynamic changes of liver parameters in metamizole DILI cases with the aim of rapidly identifying patients at high risk.
Methods Patients with acute liver injury (ALI) and suspicion of DILI, who were referred to the LMU Klinikum Munich, Germany, were enrolled in our single‐centre prospective study on patients with acute liver injury and suspected drug‐related causes, which has been described in more detail elsewhere . The data from patients that were enrolled between July 2012 and December 2023 and declared metamizole intake were analysed retrospectively. The procedures were in accordance with the Helsinki Declaration of 1975, as revised in 2013, and the study protocol was approved by the local ethics committee (Project number 55‐13). Written informed consent was obtained from all subjects. ALI was defined according to international consensus criteria : (a) Alanine aminotransferase (ALT) ≥ 5 × upper limit of normal (ULN), (b) alkaline phosphatase (ALP) ≥ 2 × ULN or (c) ALT ≥ 3 × ULN and total bilirubin (TBIL) ≥ 2 × ULN. The type of liver injury was classified using the R ‐ratio values (ALT/ULN)/(ALP/ULN), with R ≥ 5 defining a hepatocellular, R ≤ 2 a cholestatic, and 2 < R < 5 a mixed‐type injury . Severity of liver injury was graded according to the DILI severity index proposed by the international consensus criteria . According to these guidelines, mild DILI (severity category 1) is defined as elevation of ALT and/or ALP without elevation of TBIL; moderate DILI (severity category 2) as elevation of ALT and/or ALP with TBIL elevation of ≥ 2 × ULN; severe DILI (severity category 3) as additional coagulopathy, ascites, and/or encephalopathy or other organ failure associated with DILI; and fatal DILI (severity category 4) as death or transplantation due to DILI . Hy's law was defined as TBIL > 2 × ULN and ALT > 3 × ULN at the time of onset , while jaundice was defined as peak TBIL > 2 mg/dL and coagulopathy as international normalised ratio (INR) ≥ 1.5 at any stage of liver injury. A fatal adverse outcome was defined as death or liver transplantation (LT). Acute liver failure (ALF) was defined according to international criteria: (1) severe ALI characterised by an at least twofold elevation of transaminases, (2) the absence of pre‐existing liver disease, i.e., a disease duration of a maximum of 28 weeks, (3) coagulopathy with an INR ≥ 1.5 in the absence of oral anticoagulants, and (4) hepatic encephalopathy . A thorough hepatological work‐up was performed, including virology testing, liver ultrasound, magnetic resonance imaging (MRI), or computed tomography scan, as well as testing for autoantibodies and metabolic and hereditary liver diseases. The diagnosis of DILI and causality assessment of the causative agent was based on clinical, laboratory, and histopathological findings; the exclusion of alternative causes for liver injury; the Roussel Uclaf Causality Assessment Method (RUCAM) score , assessment by the supervising physician; and upon long‐term follow‐up. Forty‐nine of the 80 cases in total were additionally evaluated in a structured expert opinion causality case assessment process. This process included a detailed case presentation and evaluation by at least three independent reviewers with high expertise in the DILI field . The calculation of the RUCAM was performed for each medication with a potential for an association with DILI, meaning that an individual RUCAM score was given for each agent implicated. The total RUCAM score is the sum of points given in seven categories, which are comprised of (1) time to onset, (2) course of liver disease, (3) risk factors, (4) potential DILI by concomitant drugs, (5) exclusion of non‐drug causes of liver injury, (6) previous information on the hepatotoxicity of the drug, and (7) response to rechallenge . The individual points per category range from −3 to +3, and the total achievable score for each adjudicated medication and case ranges from −9 to +14. Likelihood of DILI is rated by the total score as follows: ≤ 0 indicates that the drug is “excluded” as a cause, while 1–2 means that the drug is an “unlikely,” 3–5 a “possible,” 6–8 a “probable,” and > 8 a “highly probable” cause of DILI . Autoimmune hepatitis (AIH) was diagnosed based on laboratory findings, histopathological features, the revised and simplified AIH scores as well as on response to corticosteroids and evaluation during long‐term follow‐up. Biliary obstruction was diagnosed with ultrasound or MRI. Whenever a liver biopsy was performed, histopathological reports were extracted from the patients' medical records in a standardised manner according to a pre‐defined catalogue. Remission was defined as resolution of liver test abnormalities, while chronicity was defined as an elevation of ALT, AST and/or TBIL at least 6 months after liver injury onset. Data are presented as median and range for continuous and as numbers and percentages for categorical variables. Categorical variables were compared using the Chi‐square test, while non‐parametric tests were applied for continuous variables. If continuous variables were compared between more than two groups, the Kruskal–Wallis test was used, while the Mann–Whitney U test was applied for the comparison of variables between two groups only. In case the Kruskal–Wallis test showed a significant difference between the metamizole DILI and two or more control groups, the differences to each comparator group alone were then analysed by the Mann–Whitney U test in a second step. p ≤ 0.05 was considered statistically significant. For the evaluation of the influence of baseline parameters on outcome, univariate and multivariate logistic regression, the latter with a stepwise backward elimination, were performed. Only variables with p ‐values < 0.1 in the univariate analysis were included in the multivariate analysis. Significant correlation ( r ≥ 0.8) between variables was excluded. A receiver‐operating characteristics (ROC) curve was utilised to identify the concordance statistic ( c ‐statistic) of baseline values associated with the diagnosis of the characteristic metamizole DILI pattern and with a fatal adverse outcome in metamizole DILI. Statistical analyses were performed using SPSS (IBM, Armonk, New York, USA, version 29).
Results 3.1 Prevalence of Metamizole DILI Between 2012 and 2023, 511 patients were included in our prospective study on ALI with a potential drug‐related cause. After thorough work‐up, 324 patients were diagnosed with DILI. Metamizole was associated with liver injury in 61 of those 324 DILI cases, translating into a prevalence of 18.8%. In addition, we identified nine patients with DILI due to an alternative drug despite concomitant metamizole treatment and 10 patients who were initially suspected to have suffered from DILI due to metamizole alone or due to metamizole and concomitant drugs but who were later diagnosed with AIH. 3.2 Clinical Characteristics The clinical characteristics of patients with metamizole DILI in comparison to control cases (comprised of non‐metamizole DILI [ n = 9] and AIH [ n = 10] patients with concomitant metamizole intake) can be reviewed in Table . There were no significant differences regarding age, body mass index (BMI), or sex between the three groups. The majority of metamizole DILI patients were female (68.9%) and middle‐aged (median: 44 years). The median daily dosage of metamizole was significantly lower in the metamizole DILI group when compared to the non‐metamizole DILI group (1000 mg vs. 2000 mg, p = 0.011) but comparable to the AIH patients (both 1000 mg, non‐significant [ns]; Table ). The latency from the start of metamizole intake until the onset of liver injury was comparable between all three groups (41 days [range: 2–674 days] in patients with metamizole DILI vs. 38 days [1–515] and 52 days [3–113] in the control groups, respectively; p = 0.827). Thirty‐eight patients with metamizole DILI (62.2%) had discontinued metamizole before the appearance of liver injury, while this was the case in six non‐metamizole DILI (66.6%) and five AIH cases (55.5%). Latency from the end of drug intake until DILI recognition was comparable in metamizole DILI and the control groups as well (Table ). In the metamizole DILI group, the median RUCAM for metamizole was 6 (range 2–11), and 98.4% had a RUCAM of 3 or higher, indicating that metamizole DILI was at least a possible cause for liver injury. In the metamizole DILI cohort, only two patients did not have any concomitant medications, while most patients (67.2%) had been taking other remedies at the same time as metamizole. In addition, 29.5% were under treatment with co‐medications; however, with no compatible time to onset. Most patients concomitantly used non‐steroidal anti‐inflammatory drugs (NSAID, n = 27, 44.3%) or antimicrobial drugs ( n = 17, 27.9%, data not shown). Concomitant medication with compatible timing to onset was used by every patient in the control groups, and the RUCAM scores for the concomitant medications were comparable between all three groups (Table ). Interestingly, significant differences regarding some baseline characteristics were mostly observed between metamizole DILI and non‐metamizole DILI patients, while the clinical features were widely comparable between metamizole DILI and AIH patients. As such, the median RUCAM for metamizole was significantly higher in metamizole DILI than in non‐metamizole DILI cases (6 [range: 2–11] vs. 4 [range: 2–6], p = 0.009), but similar in the AIH patients (6 [range: 2–11] vs. 6 [range: 0–9], p = 0.218; Table ). Moreover, patients with metamizole DILI showed higher elevation of ALT and aspartate aminotransferase (AST) when compared to non‐metamizole DILI patients (ALT: 36.5 × ULN vs. 6.0 × ULN, p = 0.007; AST: 26.9 × ULN vs. 4.4 × ULN, p = 0.009), while ALT and AST levels were comparable to those observed in AIH patients (ALT: 36.5 × ULN vs. 26.3 × ULN, p = 0.259; AST: 26.9 × ULN vs. 17.4 × ULN, p = 0.523; Table ). TBIL was also significantly higher in metamizole DILI patients when compared to the non‐metamizole DILI controls (8.2 × ULN vs. 0.9 × ULN, p = 0.003), yet a relevant increase of ALP was neither observed in metamizole DILI nor in control cases (Table ). Most patients with metamizole DILI and AIH presented with a hepatocellular type of injury (83.6% and 90.0%), while only 55.6% of the non‐metamizole‐DILI cases did ( p = 0.037). Accordingly, the R ‐ratio at onset was significantly higher in patients with metamizole DILI when compared to non‐metamizole DILI controls (20.5 vs. 5.1, p = 0.007). Immunoglobulin G (IgG) levels were significantly higher in AIH patients when compared to metamizole DILI and non‐metamizole DILI cases (16.6 g/dL vs. 12.9 g/dL and 9.7 g/dL, respectively, p = 0.017 and p = 0.009); however all three groups showed comparable rates of antinuclear antibodies (ANA) positivity (72.1% vs. 55.6% vs. 70.0%, p = 0.834; Table ). The DILI severity index, as well as the proportions of patients with Hy's law positivity, coagulopathy, and ALF, were similar in metamizole DILI and AIH patients but higher than in the non‐metamizole DILI group (Table ). Significantly more patients with metamizole DILI developed jaundice throughout the liver injury episode when compared to the non‐metamizole DILI controls (86.9% vs. 44.4%; Table ). Despite the higher severity in metamizole DILI patients, comparable remission rates were observed in all three groups (77.0% vs. 66.7% and 70.0%, p = 0.525). Time to remission, however, was significantly shorter in metamizole DILI patients than in AIH patients (10 vs. 20 weeks, p = 0.021), while there was a tendency towards a longer time to remission when compared to non‐metamizole DILI cases (10 vs. 3 weeks), although this difference was not statistically significant ( p = 0.125, Table ). Throughout the whole cohort of metamizole DILI cases, 18 of the 61 cases scored a RUCAM below 6 (29.5%), meaning that metamizole was an “unlikely” or “possible” cause of liver injury, while 43 patients (70.5%) scored 6 or higher, indicating metamizole was a “probable” or “highly probable” cause of DILI. However, apart from a higher INR at onset and peak levels in patients with a RUCAM of 3–5, no differences were observed between metamizole DILI patients with RUCAM < 6 or ≥ 6 regarding liver parameters, neither at baseline nor at peak levels, nor regarding the evolution of liver parameters over time (Table ). 3.3 Recurrence of Liver Injury Upon Rechallenge—A Common Phenomenon in Metamizole DILI Patients Nine patients from the metamizole DILI group (14.8%) presented with another episode of DILI after re‐challenge with metamizole alone ( n = 5, 8.2%) or metamizole and concomitant medication ( n = 4, 6.6%; Table ). One patient, for instance, presented at our centre with abdominal pain for 7 days and jaundice for 5 days. Laboratory sampling at presentation showed marked elevation of ALT (34.4 × ULN), AST (41.2 × ULN) and TBIL (9.1 × ULN) as well as coagulopathy (INR 1.9). The patient reported the intake of ibuprofen and paracetamol, which she had started 23 days and stopped 7 days before presentation, as well as metamizole and amoxicillin/clavulanic acid, which she had started 23 days and stopped 17 days before presentation. The hepatological work‐up excluded viral hepatitis, biliary obstruction, and metabolic as well as hereditary liver diseases. When the medical history was re‐evaluated closely, it became apparent that the patient had suffered from ALF 14 years earlier. Initial liver function tests back then had been similar, with highly elevated ALT, AST, and TBIL (40.9 × ULN, 33.7 × ULN and 16.1 × ULN) as well as coagulopathy (INR = 3.5). Previously, alternative diagnoses had been excluded as well. In addition, liver biopsy had shown subacute liver dystrophy with mixed cell type inflammation and ductular proliferation but no fibrosis nor elevation of iron or copper and was therefore rated as most likely DILI. During that episode, liver injury was associated with cyproterone acetate; however, at thorough re‐interrogation, the patient reported intake of metamizole in the weeks before that episode as well. In addition, she had used cyproterone acetate after this DILI episode again without recurrence of liver injury, the negative rechallenge therefore excludes causality with this agent. Rapidly after both liver injury incidents, liver function tests completely normalised. However, the patient once again presented with ALI 22 months after the second episode. Due to suffering from a traumatic brain injury following a bike accident, she had received amoxicillin/clavulanic acid, ibuprofen, and once again metamizole. Four weeks later, a laboratory control showed a marked elevation of ALT and AST (62.9 × ULN and 57.1 × ULN) as well as increasing levels of TBIL (1.9 × ULN) and INR (1.5). Amoxicillin/clavulanic acid and metamizole had been started 29 days and stopped 22 days before that presentation; ibuprofen had been started 29 days before and had been taken intermittently as needed until the day of presentation. Liver biopsy was re‐performed, showing eosinophilic hepatitis with signs of interface hepatitis as well as single‐cell necrosis, and was rated as most likely being associated with DILI. After onset of this last episode, AST, and ALT rapidly decreased, while TBIL peaked 12 days after presentation. Fifty‐eight days later, liver parameters had completely normalised and did not show another increase in the long‐term follow‐up. The patient reported that she intermittently was taking ibuprofen again after the third episode; metamizole and amoxicillin/clavulanic acid were never used again. Due to the spontaneous and rapid normalisation and the lack of signs for chronic liver disease in the repeatedly performed liver biopsy, AIH could be excluded in this case, leaving DILI as the only possible diagnosis. The only drug that was taken at all three time points was metamizole, strongly suggesting the causal relationship in this case. 3.4 The Characteristic Metamizole DILI Signature A characteristic clinical pattern was noticed in a large proportion of patients with metamizole DILI, which was also apparent in the case presented above: High elevation of AST and ALT at the time of DILI recognition with a rapid decline shortly after onset and a relatively high initial TBIL value, which however, continues to increase after DILI recognition (Figure ). As demonstrated in Figure , non‐metamizole DILI cases rather show a secondary increase of ALT and AST and no secondary TBIL peak. Forty‐three of the 61 metamizole DILI patients (70.5%) fit this clinical picture. When the patients with the characteristic metamizole DILI signature pattern were compared to a control group comprised of metamizole DILI cases without this clinical picture ( n = 18), patients with non‐metamizole DILI despite metamizole intake ( n = 9) and AIH patients with prior metamizole intake ( n = 10), significantly higher AST, ALT, TBIL, and INR levels, as well as MELD scores were seen at onset and peak values in the patients with the characteristic metamizole DILI pattern (Table ). The differences regarding liver parameter elevation were most pronounced between patients with the characteristic metamizole DILI pattern when compared to other metamizole DILI or non‐metamizole DILI cases, while baseline and peak liver parameters were not significantly different when compared to AIH patients. One of the only two features that was significantly higher in patients with the characteristic metamizole DILI pattern when compared to each control group alone, including AIH patients, was the R ‐ratio at peak ALT or ALP levels (29.6 vs. 17.6 vs. 12.5 and 14.1 in patients with the characteristic metamizole DILI pattern, other metamizole DILI, non‐metamizole DILI, and AIH cases, respectively, p = 0.025; Table ). In addition, the time‐based change of TBIL from day 1 to day 3 expressed as fold ULN (ΔTBIL3) was the only other factor with a significant difference when compared to each control group alone (1.8 × ULN vs. 0.0 × ULN, 0.0 × ULN, and −0.7 × ULN, p < 0.001; Table ). Even when all metamizole cases were compared to controls, ΔTBIL3 remained statistically different (+1.1 × ULN vs. −0.1; p = 0.022, data not shown). Figure demonstrates the ROC curve and c ‐statistics for baseline TBIL, MELD, ALT, AST, and INR regarding the diagnosis of metamizole DILI with the signature pattern in comparison to non‐metamizole DILI and AIH cases despite metamizole intake: The highest c ‐statistics were observed for ALT and TBIL at onset (0.74, respectively). For both baseline parameters, high positive predictive values (PPV) were observed (ALT: 87.2% at a cut‐off of 29 × ULN; TBIL: 83.3% at a cut‐off of 5.2 × ULN), while the highest sensitivity was observed for INR at a cut‐off of 1.2 (88.4%; Figure ). The time‐dependent dynamic changes of liver parameters, in particular ΔTBIL3, showed an even better discriminatory power with a c ‐statistic of 0.84. At a cut‐off of +0.1 × ULN above baseline 3 days after DILI recognition, metamizole DILI could be distinguished from controls with a remarkable PPV of 96.6% as well as high sensitivity and specificity (86.7% and 85.7%, respectively; Figure ). Regarding the clinical characteristics at baseline, there was no feature that showed a significant difference between patients with the metamizole DILI signature pattern and each control group alone (Table ). The average daily dosage of metamizole, however, was higher in the non‐metamizole DILI cases, while the RUCAM was lower in this group when compared to the patients with the characteristic metamizole DILI injury pattern (Table ). When the clinical characteristics were compared between patients with the characteristic metamizole DILI signature pattern and other metamizole DILI cases, the only parameter that was significantly different was IgG (13.7 g/dL vs. 10.8 g/dL, p = 0.047; Table ). Regarding severity and outcome, patients with the characteristic metamizole DILI injury pattern showed a higher DILI severity index than other metamizole and non‐metamizole DILI cases (3 vs. 2 and 1, p < 0.001, respectively) but not when compared to AIH patients (Table ). In addition, patients with the characteristic metamizole DILI signature pattern were more likely to be Hy's law positive upon presentation, develop jaundice or coagulopathy, and to progress to ALF. However, despite higher severity, remission rates did not differ significantly between the four groups (Table ). 3.5 Histological Analysis Forty‐five of the 61 patients (73.7%) with metamizole DILI received liver biopsy sampling. All of the patients presented with inflammatory infiltrates, which were predominately of a mixed cell type or comprised of lymphocytes and plasma cells (Table ). Inflammatory infiltrates were mainly localised in the portal and/or intralobular areas. 44.4% presented with interface hepatitis, while plasma cells were detected in 48.9%. In addition, eosinophilic cells were observed in 68.9%. Cholestasis was seen in 42.2%, which was mainly canalicular and hepatocellular. Necrosis was highly present (84.4%), with 28.2% showing severe confluent necrosis (Table ). Representative histological slides showing eosinophilic hepatitis, interface hepatitis, and extensive necrosis can be reviewed in Figure . 3.6 Parameters Influencing Outcome A relevant proportion of patients with metamizole DILI ( n = 16, 26.2%) developed ALF, and 13.1% ( n = 8) died or underwent LT. Patients with ALF showed higher AST, ALT, TBIL, INR, and MELD at onset and peak levels (Table ). While remission rates were lower in patients with ALF, time to remission was comparable in those patients who went into remission (Table ). In addition, the decline of AST and ALT until day 7 after DILI recognition on day 1 was significantly higher in patients with metamizole‐induced ALF (−22.7 × ULN vs. −2.1 × ULN, p < 0.001, and −18.1 × ULN vs. 2.4 × ULN, p = 0.002), while TBIL showed a significantly higher increase at day 3 (2.3 × ULN vs. 0.5 × ULN, p = 0.014; Table ). Hy's law, as a conventional measure for prediction of ALF, was positive in 43 metamizole DILI patients. Out of those 43 patients, 16 developed ALF (37.2%), and all patients with ALF also were Hy's law positive, translating into a sensitivity of 100.0% and specificity of 37.2% for Hy's law predicting ALF in our metamizole DILI cohort (Table ). Thus, due to low specificity, Hy's law was a suboptimal prediction marker for the development of ALF in our cohort. We therefore aimed to identify better prediction markers for adverse outcomes in our cohort of patients with metamizole DILI. On univariate analysis, ALT, AST, TBIL, INR, and MELD at onset were significantly associated with fatal adverse outcomes (as defined by LT or death; Table ). Age, sex, daily dosage of metamizole, or latency until onset of DILI did not have a significant effect on outcome; neither did corticosteroid treatment (Table ). In the multivariate logistic regression analysis, only AST (odds ratio [OR]: 1.07, 95% confidence interval [CI]: 1.02–1.13, p = 0.009) and INR at onset (OR: 3.42, 95% CI: 1.47–7.93, p = 0.004, Table ) remained significantly associated with a fatal adverse outcome. The c ‐statistic for fatal adverse outcomes for INR, MELD, AST, TBIL, and ALT at onset were 0.85 (95% CI: 0.70–1.01, p < 0.001), 0.84 (95% CI: 0.70–0.99, p < 0.001), 0.83 (95% CI: 0.67–0.98, p < 0.001), 0.80 (95% CI: 0.67–0.94, p < 0.001), and 0.74 (95% CI: 0.55–0.92, p = 0.015; Figure ). The optimal cut‐off for INR was determined at ≥ 2.1, at which a sensitivity of 75.0% and a specificity of 95.6% for a fatal adverse outcome were observed, while TBIL with a cut‐off of 10 × ULN showed the highest sensitivity (87.5%) at still reasonable specificity (71.1%; Figure ). INR in particular showed an extraordinary discriminatory power with both high PPV (75.0%) and negative predictive values (NPV; 95.8%). In addition, while all of the baseline laboratory parameters showed high NPV above 90%, with 97.2% the highest NPV was observed for TBIL at a cut‐off of 10 × ULN. The time‐dependent dynamic changes of transaminases following DILI recognition performed even better in predicting fatal adverse outcomes: The c ‐statistics for ΔAST from day 1 to 3, as well as ΔALT from day 1 to 3 and from day 1 to day 7, were all > 0.9 ( p < 0.001; Figure ). Sensitivity and specificity for a fatal adverse outcome were 100.0% and 89.7%, respectively, for a decline of AST of at least 13 × ULN by day 3 or a decline of ALT by at least 15 × ULN at day 3 and 37 × ULN at day 7 (Figure ). With regards to PPV and NPV, ΔAST between day 1 and 7 at a cut‐off of −52 × ULN had the highest discriminatory power with a PPV of 100.0% and NPV of 89.5% (Figure ).
Prevalence of Metamizole DILI Between 2012 and 2023, 511 patients were included in our prospective study on ALI with a potential drug‐related cause. After thorough work‐up, 324 patients were diagnosed with DILI. Metamizole was associated with liver injury in 61 of those 324 DILI cases, translating into a prevalence of 18.8%. In addition, we identified nine patients with DILI due to an alternative drug despite concomitant metamizole treatment and 10 patients who were initially suspected to have suffered from DILI due to metamizole alone or due to metamizole and concomitant drugs but who were later diagnosed with AIH.
Clinical Characteristics The clinical characteristics of patients with metamizole DILI in comparison to control cases (comprised of non‐metamizole DILI [ n = 9] and AIH [ n = 10] patients with concomitant metamizole intake) can be reviewed in Table . There were no significant differences regarding age, body mass index (BMI), or sex between the three groups. The majority of metamizole DILI patients were female (68.9%) and middle‐aged (median: 44 years). The median daily dosage of metamizole was significantly lower in the metamizole DILI group when compared to the non‐metamizole DILI group (1000 mg vs. 2000 mg, p = 0.011) but comparable to the AIH patients (both 1000 mg, non‐significant [ns]; Table ). The latency from the start of metamizole intake until the onset of liver injury was comparable between all three groups (41 days [range: 2–674 days] in patients with metamizole DILI vs. 38 days [1–515] and 52 days [3–113] in the control groups, respectively; p = 0.827). Thirty‐eight patients with metamizole DILI (62.2%) had discontinued metamizole before the appearance of liver injury, while this was the case in six non‐metamizole DILI (66.6%) and five AIH cases (55.5%). Latency from the end of drug intake until DILI recognition was comparable in metamizole DILI and the control groups as well (Table ). In the metamizole DILI group, the median RUCAM for metamizole was 6 (range 2–11), and 98.4% had a RUCAM of 3 or higher, indicating that metamizole DILI was at least a possible cause for liver injury. In the metamizole DILI cohort, only two patients did not have any concomitant medications, while most patients (67.2%) had been taking other remedies at the same time as metamizole. In addition, 29.5% were under treatment with co‐medications; however, with no compatible time to onset. Most patients concomitantly used non‐steroidal anti‐inflammatory drugs (NSAID, n = 27, 44.3%) or antimicrobial drugs ( n = 17, 27.9%, data not shown). Concomitant medication with compatible timing to onset was used by every patient in the control groups, and the RUCAM scores for the concomitant medications were comparable between all three groups (Table ). Interestingly, significant differences regarding some baseline characteristics were mostly observed between metamizole DILI and non‐metamizole DILI patients, while the clinical features were widely comparable between metamizole DILI and AIH patients. As such, the median RUCAM for metamizole was significantly higher in metamizole DILI than in non‐metamizole DILI cases (6 [range: 2–11] vs. 4 [range: 2–6], p = 0.009), but similar in the AIH patients (6 [range: 2–11] vs. 6 [range: 0–9], p = 0.218; Table ). Moreover, patients with metamizole DILI showed higher elevation of ALT and aspartate aminotransferase (AST) when compared to non‐metamizole DILI patients (ALT: 36.5 × ULN vs. 6.0 × ULN, p = 0.007; AST: 26.9 × ULN vs. 4.4 × ULN, p = 0.009), while ALT and AST levels were comparable to those observed in AIH patients (ALT: 36.5 × ULN vs. 26.3 × ULN, p = 0.259; AST: 26.9 × ULN vs. 17.4 × ULN, p = 0.523; Table ). TBIL was also significantly higher in metamizole DILI patients when compared to the non‐metamizole DILI controls (8.2 × ULN vs. 0.9 × ULN, p = 0.003), yet a relevant increase of ALP was neither observed in metamizole DILI nor in control cases (Table ). Most patients with metamizole DILI and AIH presented with a hepatocellular type of injury (83.6% and 90.0%), while only 55.6% of the non‐metamizole‐DILI cases did ( p = 0.037). Accordingly, the R ‐ratio at onset was significantly higher in patients with metamizole DILI when compared to non‐metamizole DILI controls (20.5 vs. 5.1, p = 0.007). Immunoglobulin G (IgG) levels were significantly higher in AIH patients when compared to metamizole DILI and non‐metamizole DILI cases (16.6 g/dL vs. 12.9 g/dL and 9.7 g/dL, respectively, p = 0.017 and p = 0.009); however all three groups showed comparable rates of antinuclear antibodies (ANA) positivity (72.1% vs. 55.6% vs. 70.0%, p = 0.834; Table ). The DILI severity index, as well as the proportions of patients with Hy's law positivity, coagulopathy, and ALF, were similar in metamizole DILI and AIH patients but higher than in the non‐metamizole DILI group (Table ). Significantly more patients with metamizole DILI developed jaundice throughout the liver injury episode when compared to the non‐metamizole DILI controls (86.9% vs. 44.4%; Table ). Despite the higher severity in metamizole DILI patients, comparable remission rates were observed in all three groups (77.0% vs. 66.7% and 70.0%, p = 0.525). Time to remission, however, was significantly shorter in metamizole DILI patients than in AIH patients (10 vs. 20 weeks, p = 0.021), while there was a tendency towards a longer time to remission when compared to non‐metamizole DILI cases (10 vs. 3 weeks), although this difference was not statistically significant ( p = 0.125, Table ). Throughout the whole cohort of metamizole DILI cases, 18 of the 61 cases scored a RUCAM below 6 (29.5%), meaning that metamizole was an “unlikely” or “possible” cause of liver injury, while 43 patients (70.5%) scored 6 or higher, indicating metamizole was a “probable” or “highly probable” cause of DILI. However, apart from a higher INR at onset and peak levels in patients with a RUCAM of 3–5, no differences were observed between metamizole DILI patients with RUCAM < 6 or ≥ 6 regarding liver parameters, neither at baseline nor at peak levels, nor regarding the evolution of liver parameters over time (Table ).
Recurrence of Liver Injury Upon Rechallenge—A Common Phenomenon in Metamizole DILI Patients Nine patients from the metamizole DILI group (14.8%) presented with another episode of DILI after re‐challenge with metamizole alone ( n = 5, 8.2%) or metamizole and concomitant medication ( n = 4, 6.6%; Table ). One patient, for instance, presented at our centre with abdominal pain for 7 days and jaundice for 5 days. Laboratory sampling at presentation showed marked elevation of ALT (34.4 × ULN), AST (41.2 × ULN) and TBIL (9.1 × ULN) as well as coagulopathy (INR 1.9). The patient reported the intake of ibuprofen and paracetamol, which she had started 23 days and stopped 7 days before presentation, as well as metamizole and amoxicillin/clavulanic acid, which she had started 23 days and stopped 17 days before presentation. The hepatological work‐up excluded viral hepatitis, biliary obstruction, and metabolic as well as hereditary liver diseases. When the medical history was re‐evaluated closely, it became apparent that the patient had suffered from ALF 14 years earlier. Initial liver function tests back then had been similar, with highly elevated ALT, AST, and TBIL (40.9 × ULN, 33.7 × ULN and 16.1 × ULN) as well as coagulopathy (INR = 3.5). Previously, alternative diagnoses had been excluded as well. In addition, liver biopsy had shown subacute liver dystrophy with mixed cell type inflammation and ductular proliferation but no fibrosis nor elevation of iron or copper and was therefore rated as most likely DILI. During that episode, liver injury was associated with cyproterone acetate; however, at thorough re‐interrogation, the patient reported intake of metamizole in the weeks before that episode as well. In addition, she had used cyproterone acetate after this DILI episode again without recurrence of liver injury, the negative rechallenge therefore excludes causality with this agent. Rapidly after both liver injury incidents, liver function tests completely normalised. However, the patient once again presented with ALI 22 months after the second episode. Due to suffering from a traumatic brain injury following a bike accident, she had received amoxicillin/clavulanic acid, ibuprofen, and once again metamizole. Four weeks later, a laboratory control showed a marked elevation of ALT and AST (62.9 × ULN and 57.1 × ULN) as well as increasing levels of TBIL (1.9 × ULN) and INR (1.5). Amoxicillin/clavulanic acid and metamizole had been started 29 days and stopped 22 days before that presentation; ibuprofen had been started 29 days before and had been taken intermittently as needed until the day of presentation. Liver biopsy was re‐performed, showing eosinophilic hepatitis with signs of interface hepatitis as well as single‐cell necrosis, and was rated as most likely being associated with DILI. After onset of this last episode, AST, and ALT rapidly decreased, while TBIL peaked 12 days after presentation. Fifty‐eight days later, liver parameters had completely normalised and did not show another increase in the long‐term follow‐up. The patient reported that she intermittently was taking ibuprofen again after the third episode; metamizole and amoxicillin/clavulanic acid were never used again. Due to the spontaneous and rapid normalisation and the lack of signs for chronic liver disease in the repeatedly performed liver biopsy, AIH could be excluded in this case, leaving DILI as the only possible diagnosis. The only drug that was taken at all three time points was metamizole, strongly suggesting the causal relationship in this case.
The Characteristic Metamizole DILI Signature A characteristic clinical pattern was noticed in a large proportion of patients with metamizole DILI, which was also apparent in the case presented above: High elevation of AST and ALT at the time of DILI recognition with a rapid decline shortly after onset and a relatively high initial TBIL value, which however, continues to increase after DILI recognition (Figure ). As demonstrated in Figure , non‐metamizole DILI cases rather show a secondary increase of ALT and AST and no secondary TBIL peak. Forty‐three of the 61 metamizole DILI patients (70.5%) fit this clinical picture. When the patients with the characteristic metamizole DILI signature pattern were compared to a control group comprised of metamizole DILI cases without this clinical picture ( n = 18), patients with non‐metamizole DILI despite metamizole intake ( n = 9) and AIH patients with prior metamizole intake ( n = 10), significantly higher AST, ALT, TBIL, and INR levels, as well as MELD scores were seen at onset and peak values in the patients with the characteristic metamizole DILI pattern (Table ). The differences regarding liver parameter elevation were most pronounced between patients with the characteristic metamizole DILI pattern when compared to other metamizole DILI or non‐metamizole DILI cases, while baseline and peak liver parameters were not significantly different when compared to AIH patients. One of the only two features that was significantly higher in patients with the characteristic metamizole DILI pattern when compared to each control group alone, including AIH patients, was the R ‐ratio at peak ALT or ALP levels (29.6 vs. 17.6 vs. 12.5 and 14.1 in patients with the characteristic metamizole DILI pattern, other metamizole DILI, non‐metamizole DILI, and AIH cases, respectively, p = 0.025; Table ). In addition, the time‐based change of TBIL from day 1 to day 3 expressed as fold ULN (ΔTBIL3) was the only other factor with a significant difference when compared to each control group alone (1.8 × ULN vs. 0.0 × ULN, 0.0 × ULN, and −0.7 × ULN, p < 0.001; Table ). Even when all metamizole cases were compared to controls, ΔTBIL3 remained statistically different (+1.1 × ULN vs. −0.1; p = 0.022, data not shown). Figure demonstrates the ROC curve and c ‐statistics for baseline TBIL, MELD, ALT, AST, and INR regarding the diagnosis of metamizole DILI with the signature pattern in comparison to non‐metamizole DILI and AIH cases despite metamizole intake: The highest c ‐statistics were observed for ALT and TBIL at onset (0.74, respectively). For both baseline parameters, high positive predictive values (PPV) were observed (ALT: 87.2% at a cut‐off of 29 × ULN; TBIL: 83.3% at a cut‐off of 5.2 × ULN), while the highest sensitivity was observed for INR at a cut‐off of 1.2 (88.4%; Figure ). The time‐dependent dynamic changes of liver parameters, in particular ΔTBIL3, showed an even better discriminatory power with a c ‐statistic of 0.84. At a cut‐off of +0.1 × ULN above baseline 3 days after DILI recognition, metamizole DILI could be distinguished from controls with a remarkable PPV of 96.6% as well as high sensitivity and specificity (86.7% and 85.7%, respectively; Figure ). Regarding the clinical characteristics at baseline, there was no feature that showed a significant difference between patients with the metamizole DILI signature pattern and each control group alone (Table ). The average daily dosage of metamizole, however, was higher in the non‐metamizole DILI cases, while the RUCAM was lower in this group when compared to the patients with the characteristic metamizole DILI injury pattern (Table ). When the clinical characteristics were compared between patients with the characteristic metamizole DILI signature pattern and other metamizole DILI cases, the only parameter that was significantly different was IgG (13.7 g/dL vs. 10.8 g/dL, p = 0.047; Table ). Regarding severity and outcome, patients with the characteristic metamizole DILI injury pattern showed a higher DILI severity index than other metamizole and non‐metamizole DILI cases (3 vs. 2 and 1, p < 0.001, respectively) but not when compared to AIH patients (Table ). In addition, patients with the characteristic metamizole DILI signature pattern were more likely to be Hy's law positive upon presentation, develop jaundice or coagulopathy, and to progress to ALF. However, despite higher severity, remission rates did not differ significantly between the four groups (Table ).
Histological Analysis Forty‐five of the 61 patients (73.7%) with metamizole DILI received liver biopsy sampling. All of the patients presented with inflammatory infiltrates, which were predominately of a mixed cell type or comprised of lymphocytes and plasma cells (Table ). Inflammatory infiltrates were mainly localised in the portal and/or intralobular areas. 44.4% presented with interface hepatitis, while plasma cells were detected in 48.9%. In addition, eosinophilic cells were observed in 68.9%. Cholestasis was seen in 42.2%, which was mainly canalicular and hepatocellular. Necrosis was highly present (84.4%), with 28.2% showing severe confluent necrosis (Table ). Representative histological slides showing eosinophilic hepatitis, interface hepatitis, and extensive necrosis can be reviewed in Figure .
Parameters Influencing Outcome A relevant proportion of patients with metamizole DILI ( n = 16, 26.2%) developed ALF, and 13.1% ( n = 8) died or underwent LT. Patients with ALF showed higher AST, ALT, TBIL, INR, and MELD at onset and peak levels (Table ). While remission rates were lower in patients with ALF, time to remission was comparable in those patients who went into remission (Table ). In addition, the decline of AST and ALT until day 7 after DILI recognition on day 1 was significantly higher in patients with metamizole‐induced ALF (−22.7 × ULN vs. −2.1 × ULN, p < 0.001, and −18.1 × ULN vs. 2.4 × ULN, p = 0.002), while TBIL showed a significantly higher increase at day 3 (2.3 × ULN vs. 0.5 × ULN, p = 0.014; Table ). Hy's law, as a conventional measure for prediction of ALF, was positive in 43 metamizole DILI patients. Out of those 43 patients, 16 developed ALF (37.2%), and all patients with ALF also were Hy's law positive, translating into a sensitivity of 100.0% and specificity of 37.2% for Hy's law predicting ALF in our metamizole DILI cohort (Table ). Thus, due to low specificity, Hy's law was a suboptimal prediction marker for the development of ALF in our cohort. We therefore aimed to identify better prediction markers for adverse outcomes in our cohort of patients with metamizole DILI. On univariate analysis, ALT, AST, TBIL, INR, and MELD at onset were significantly associated with fatal adverse outcomes (as defined by LT or death; Table ). Age, sex, daily dosage of metamizole, or latency until onset of DILI did not have a significant effect on outcome; neither did corticosteroid treatment (Table ). In the multivariate logistic regression analysis, only AST (odds ratio [OR]: 1.07, 95% confidence interval [CI]: 1.02–1.13, p = 0.009) and INR at onset (OR: 3.42, 95% CI: 1.47–7.93, p = 0.004, Table ) remained significantly associated with a fatal adverse outcome. The c ‐statistic for fatal adverse outcomes for INR, MELD, AST, TBIL, and ALT at onset were 0.85 (95% CI: 0.70–1.01, p < 0.001), 0.84 (95% CI: 0.70–0.99, p < 0.001), 0.83 (95% CI: 0.67–0.98, p < 0.001), 0.80 (95% CI: 0.67–0.94, p < 0.001), and 0.74 (95% CI: 0.55–0.92, p = 0.015; Figure ). The optimal cut‐off for INR was determined at ≥ 2.1, at which a sensitivity of 75.0% and a specificity of 95.6% for a fatal adverse outcome were observed, while TBIL with a cut‐off of 10 × ULN showed the highest sensitivity (87.5%) at still reasonable specificity (71.1%; Figure ). INR in particular showed an extraordinary discriminatory power with both high PPV (75.0%) and negative predictive values (NPV; 95.8%). In addition, while all of the baseline laboratory parameters showed high NPV above 90%, with 97.2% the highest NPV was observed for TBIL at a cut‐off of 10 × ULN. The time‐dependent dynamic changes of transaminases following DILI recognition performed even better in predicting fatal adverse outcomes: The c ‐statistics for ΔAST from day 1 to 3, as well as ΔALT from day 1 to 3 and from day 1 to day 7, were all > 0.9 ( p < 0.001; Figure ). Sensitivity and specificity for a fatal adverse outcome were 100.0% and 89.7%, respectively, for a decline of AST of at least 13 × ULN by day 3 or a decline of ALT by at least 15 × ULN at day 3 and 37 × ULN at day 7 (Figure ). With regards to PPV and NPV, ΔAST between day 1 and 7 at a cut‐off of −52 × ULN had the highest discriminatory power with a PPV of 100.0% and NPV of 89.5% (Figure ).
Discussion Following multiple reports on DILI due to metamizole, the hepatotoxic potential of metamizole is now well established . This is underlined by our finding that metamizole was associated with liver injury in 18.8% of 324 DILI patients included in our prospective study on ALI with a potential drug‐related cause. However, polypharmacy can hinder the identification of the causative agent. Metamizole as an analgesic and antipyretic agent is seldomly used alone, as it was demonstrated in a study on agranulocytosis induced by metamizole . In fact, in our cohort only two of the 61 patients with metamizole DILI had been taking no medication other than metamizole, while the majority of patients from our cohort had been using co‐medications with a compatible timing to onset of liver injury. The concomitant medication was mainly comprised of NSAIDs and antimicrobial substances, two drug classes that have been described as most commonly associated with DILI . Thus, due to those confounding co‐medications, metamizole might be underrecognized as a likely agent causing DILI. This can then lead to the erroneous discontinuation of innocent bystander medication and/or unintentional re‐exposure with metamizole. Alarmingly, re‐exposure to a medication that has previously caused DILI can lead to another DILI episode and in some cases also result in ALF and death . In line with this, we found that nine patients from our cohort (14.8%) had an involuntary positive re‐challenge with metamizole leading to a secondary (and even tertiary) episode of DILI. We observed that these re‐exposures occurred when patients and their physicians had related DILI to the co‐medication rather than to metamizole, which led to the misperception of metamizole being a safe drug in the respective patients. Overall, the phenotype of liver injury caused by metamizole which we observed in the current study, was comparable to previous data . The majority of patients presented with a hepatocellular injury pattern with marked elevation of AST, ALT, and TBIL as well as high rates of positive ANA, low ALP values, as well as strong inflammatory activity and extensive necrosis on histological analysis. Interestingly, the clinical features of metamizole DILI and AIH patients were quite similar, while relevant differences were rather observed in comparison to patients with non‐metamizole DILI, in particular regarding the level of liver parameter elevation, which was generally lower in patients with non‐metamizole DILI. In addition to a high rate of ANA positivity, a relatively high proportion of metamizole DILI patients presented with AMA (25%). The prevalence of ANA is commonly associated with AIH, while AMA is mostly correlated with primary biliary cholangitis; however, as we have described earlier, relevant proportions of DILI patients can present with both ANA and AMA positivity . Moreover, an unexpectedly high proportion of DILI patients had fibrosis (38%), which, however, was mostly mild to mild–moderate. Only one patient presented with severe fibrosis; however, this patient was transplanted due to severe DILI, and no other typical features for AIH were observed; most importantly, IgG was normal, and interface hepatitis was absent in the histological work‐up, which strongly indicates that DILI and not AIH was the cause of liver injury in this case. Due to the lack of specific features, differential diagnosis between DILI and AIH largely depends on clinical follow‐up . All of our cases were evaluated by RUCAM to complement thorough assessments of causality of liver injury performed by expert consensus. In addition, cases were followed up long‐term (median follow‐up of the metamizole cases: 29 [1–470] weeks) with no evidence of spontaneous recurrence of liver injury, which is one of the most accurate methods to distinguish DILI from AIH. Thus, despite relatively high rates of ANA positivity and fibrosis, we can provide a convincing differential diagnosis between DILI and AIH in our cohort. Interestingly, we identified a characteristic phenotype in 43 of the 61 patients with metamizole DILI. This typical injury pattern included a hepatocellular type of injury with marked elevation of transaminases at the time of DILI onset (AST: 36.9 × ULN, ALT:43.6 × ULN), which generally peaked at the time of DILI recognition, and a delayed increase of bilirubin with a median peak 7 days after DILI onset. Patients with the metamizole DILI signature pattern could best be distinguished from controls by baseline ALT at a cut‐off of 29 × ULN or baseline TBIL at a cut‐off of 5.2 × ULN with a PPV of 87% and 83%, respectively. Moreover, we observed that the increase of TBIL in the first 3–7 days after DILI recognition was a crucial parameter when it comes to identifying metamizole DILI cases. As such, the time‐based change of TBIL from day 1 to day 3 was one of the only parameters that was significantly different in patients with the characteristic metamizole DILI when compared to each control group alone. Strikingly, an increase of TBIL above baseline values by at least 0.1 × ULN in the first 3 days after liver injury recognition could distinguish metamizole DILI from controls with a high discriminatory power (PPV: 97%, sensitivity 87%, specificity 86%). Furthermore, patients with the characteristic metamizole DILI pattern showed a relatively long latency from the start of drug intake until the time of DILI recognition of 48 days. This is in line with the previous observations on metamizole DILI showing average latencies of 4–6 weeks and latencies as long as 35 weeks . In addition, 43% of patients with the characteristic metamizole DILI had discontinued the drug before DILI onset with a median time from drug discontinuation until the appearance of liver injury of 17 days. This time interval is longer than the latencies usually described for DILI, which are generally expected to be no longer than 15 days after drug withdrawal in case of earlier discontinuation . However, longer latencies have also been observed for other agents, e.g., minocycline or nitrofurantoin, a fact that has been considered by the revised electronic causality assessment method (RECAM) recently proposed by Hayashi et al. . Since 30% of all metamizole cases from our cohort only had a RUCAM below 6, meaning metamizole was an “unlikely” or “possible” cause for liver injury , we evaluated whether RUCAM had an influence on the severity of liver parameter elevation or the evolution of liver parameters. However, no significant differences could be observed regarding baseline or peak liver parameters nor the time‐dependent dynamic changes of serum transaminases or TBIL. Thus, our results indicate that metamizole DILI patients with RUCAM ≥ 6 or < 6 were comparable, arguing against the RUCAM score being a reliable discriminating tool in the assessment of metamizole DILI. It is well known that the RUCAM has major limitations, in particular in patients with polymedication or in severe cases . For instance, a higher score is given if the latency between drug initiation and DILI onset is between 5 and 90 days or ≤ 15 days after drug discontinuation. However, as described above, we observed longer latencies in some of our cases, in particular in patients that had discontinued metamizole before recognition of DILI. This might be due to a prolonged asymptomatic period until liver injury becomes apparent in metamizole DILI cases. In addition, simultaneous intake of other agents with DILI potential will lead to a lower RUCAM score for both metamizole and the concomitant medication. As we have pointed out, the majority of our patients were treated with a concomitant medication with compatible timing to the onset of liver injury, artificially leading to a lower RUCAM score for metamizole, which, however, does not reflect a lower probability of metamizole being the causative agent. Moreover, a relatively large proportion of patients presented with ALF, which, due to an increase of liver function tests after the onset of liver injury can lead to a lower score for the item “course of reaction,” however, once again, it does not lower the probability of metamizole causing DILI. Apart from the severity of liver enzyme increase, the characteristic metamizole DILI signature included Hy's law positivity in the majority of cases, coagulopathy in 72% of the patients, and with 37% also a high rate of ALF. Strinkingly despite the high severity in the subgroup with the metamizole DILI signature pattern, recovery rates remained high at 74%, which were also comparable to the recovery rates observed in larger registry studies on DILI cases in general . In addition, despite the higher severity and the more prominent increases in liver enzymes, no significant differences regarding the rates of recovery were observed between patients with the metamizole DILI signature pattern and control cases. Our findings therefore indicate that the characteristic metamizole DILI signature pattern is characterised by high severity at the time of DILI onset but not by higher mortality rates or lower remission rates. With high peak serum transaminase and rapid reversal of these changes after drug discontinuation, the biochemical profile of metamizole DILI, which was identified by us, has some similarities with acetaminophen overdose . Nevertheless, the timing to onset of liver injury after drug exposure differs significantly between those agents. In addition, while similarly to acetaminophen overdose, necrosis was highly present in our patients with metamizole DILI, histopathological findings also showed substantial differences when compared to those commonly observed in acetaminophen‐induced liver injury . The characteristic metamizole DILI signature pattern can potentially help clinicians in the future when confronted with a patient with suspected DILI and intake of metamizole among other agents. We propose that the intake of metamizole together with a marked elevation of transaminases at the time of liver injury detection, a secondary peak of bilirubin, and a reported latency from the start of the drug intake of approximately 7–8 weeks should prompt the diagnosis of a metamizole‐induced liver injury. In this regard, our data suggests that patients with baseline ALT of ≥ 29 × ULN or TBIL of ≥ 5.2 × ULN, as well as an increase of TBIL of ≥ 0.1 × ULN during the 3 days after onset of liver injury, are more prone to suffering from metamizole DILI than an alternative cause of liver injury or DILI caused by alternative agents. While these cut‐offs were established for the metamizole DILI subgroup with the typical injury pattern, there is the possibility that the distinct subtypes of metamizole DILI presented here might merely reflect that liver injury became apparent at a different time point in the evolution of DILI. For a further evaluation and validation of the signature pattern identified by us, the clinical features of metamizole DILI should be evaluated in prospective DILI cohorts and then be compared to larger control groups in the future. Since DILI is one of the leading causes of ALF , and metamizole is a relevant cause of DILI as demonstrated by our data, prediction of a fatal adverse outcome defined by death or LT in metamizole DILI has major clinical implications. We therefore aimed to further identify parameters or scores that can help to predict which patients with metamizole DILI are at high risk for a fatal adverse outcome even before the typical metamizole pattern becomes apparent. In this regard, Hy's law had a high sensitivity of 100% but a low specificity of only 37% for developing ALF in our cohort, which is comparable to previous data . On multivariate analysis, the only baseline parameters significantly associated with fatal adverse outcomes as defined by death or LT were baseline INR and AST values. INR in particular showed the highest discriminatory power, with a NPV of 96% and a PPV of 75% at a cut‐off of ≥ 2.1. Strinkingly when the time‐dependent dynamic changes of liver parameters after DILI recognition were evaluated, the predictive power regarding a fatal adverse outcome could be drastically improved: A fatal adverse outcome could best be predicted within 3 days after clinical presentation if a decline of AST of at least 13 × ULN or ALT of at least 15 × ULN was observed (sensitivity 100%, specificity 90%). Moreover, a decline of AST by ≥ 52 × ULN 7 days after DILI recognition was highly associated with a fatal adverse outcome (PPV 100%, NPV 90%). Thus, in the case of a marked coagulopathy with an INR of 2.1 or higher at the time of DILI recognition or a decline of AST of at least 13 × ULN within the first 3 days or 52 × ULN during the first week after DILI onset, patients are at a particular high risk of a fatal adverse outcome and should rapidly be transferred to a transplant centre since the risk for death or LT is high. Our study has limitations. The control groups, for instance, to which the patients with the characteristic metamizole DILI pattern were compared to, were limited in sample size and were rather heterogeneous. In addition, due to the lack of a general gold standard for DILI diagnosis and causality assessment in patients with multiple drug intake, one might argue that causality assessment and, in particular, differentiation from AIH cannot be conducted in a standardised manner. While the latter limitation is inherently unavoidable in clinical studies on DILI, the strengths of our study are the large sample size of patients with metamizole DILI, the thorough clinical work‐up, as well as the long‐term follow‐up enabling a detailed causality assessment. In addition, a relatively large proportion of our patients presented with recurrence of liver injury upon re‐exposure towards metamizole, which is a strong indicator of this drug being the underlying cause of liver injury . Moreover, the comparison to patients with non‐metamizole DILI or alternative causes for liver injury despite metamizole intake provides the possibility to diagnose or exclude metamizole as the causative agent based on the clinical picture. Our findings should stimulate further prospective studies evaluating the accuracy and efficacy of the proposed metamizole signature and outcome markers. In summary, we have identified a characteristic metamizole DILI phenotype comprised of high transaminases at the time of DILI onset, a secondary increase of TBIL, and significant proportions developing ALF but still high remission rates. Indicators for a fatal adverse outcome are high AST and especially INR at the time of DILI recognition, as well as a drastic decline of AST in the first 3–7 days after hospitalisation, which should lead to rapid liver transplant evaluation in order to avoid death from ALF.
Conceptualization: S.W., F.E. and A.L.G.; methodology: S.W. and F.E.; software: S.W. and F.E.; investigation: S.W., F.E., J.A., D.S., N.D., J.N. and C.M.L.; formal analysis: S.W., F.E. and J.N.; validation: S.W., F.E. and A.L.G.; resources: S.W., J.A., C.M.L. and A.L.G.; data curation: S.W. and F.E.; writing – original draft: S.W.; writing – review and editing: S.W., F.E., J.N., C.M.L. and A.L.G.; visualisation: S.W., F.E. and J.N.; supervision: S.W. and A.L.G.; project administration: S.W. and A.L.G.; funding acquisition: S.W. and A.L.G. All authors approved the final version of the manuscript.
The study protocol conforms to the ethical guidelines of the Declaration of Helsinki and was approved by the ethics committee of the Faculty of Medicine, LMU Munich (Project Number 55‐13).
Written informed consent was obtained from each participant.
The authors declare no conflicts of interest.
Table S1 Table S2 Table S3 Table S4
|
Microbial Proxies
for Anoxic Microsites Vary with
Management and Partially Explain Soil Carbon Concentration | 87e63ea6-ce30-4b2e-9b13-deefe8e461e1 | 11223465 | Microbiology[mh] | Introduction Restoring soil organic
carbon (C) content in cropland soils promises
to support climate resilient agriculture , and help mitigate
climate change. − Changes in soil organic C content (g m –2 ) are often approximated by changes in soil organic C concentration
. However, precise management practices that increase soil C concentration
and thus content remain largely unknown, partially because our understanding
of soil C cycling is incomplete. To increase soil organic C content,
soil C inputs must exceed outputs. Photosynthetically derived C is
the primary input to soil organic C, while soil microbial respiration,
the oxidation of soil C to carbon dioxide (CO 2 ), serves
as the primary output. While producing more plant biomass is an obvious
strategy to increase soil C inputs, microbial activity stimulated
by organic C additions may offset potential gains in soil C content. − Thus, management solutions to increase soil C content must also
consider constraints on soil microbial respiration. Soil carbon
protection mechanisms constrain soil microbial respiration.
Soil C binds to minerals, rendering C less available for microbial
respiration in a process termed mineral protection. Physical protection occurs when microorganisms are separated
from substrate within soil micropores. , Warmer temperatures
and moderate moisture contents increase microbial activity, while
cooler temperatures and immoderate moisture contents can slow the
turnover of organic C (i.e., climate protection). − More recently, anoxic
protection , the soil C protection conferred by anoxic microsites,
has been identified as an important and under-characterized soil C
protection mechanism. − Anoxic microsites, zones of oxygen depletion
in otherwise well-aerated
(i.e., upland) soils, form when soil oxygen supply from the atmosphere
– mediated by soil texture, moisture, and structure –
is slower than microbial demand for oxygen. Anoxic microsites represent
soil redox gradients and thus serve as
habitat space for a variety of anaerobes. Depending on the precise
terminal electron acceptor utilized for anaerobic respiration, anaerobic
respiration of soil organic C is often slower than aerobic respiration
of soil organic C. , , Recently, it has been shown that CO 2 production from
anoxic microsites can be up to 90% slower than CO 2 production
from equally sized oxic habitat space. Consequently, anoxic microsites have been associated with enhanced
soil C content , and residence times, while the aeration of anoxic microsites is associated
with increased CO 2 fluxes. Common agricultural management practices regularly alter oxygen
supply and demand in soils. Tillage increases oxygen supply by disrupting
soil structure and amplifying porosity, , while organic
matter (OM) amendments enhance microbial oxygen demand by providing
bioavailable C to microbes. However, the
influence of these management practices on anoxic microsites and,
subsequently, anoxic protection and storage of soil C remains unknown.
As a result, our ability to design biogeochemically informed management
strategies that increase soil C content is limited. Here, we
sought to determine the response of anoxic microsites
to management practices (e.g., agricultural use, tillage practices,
and OM amendments) and determine the degree to which anoxic microsites
impact soil C concentration across these soils. To achieve this goal,
we used a novel approach for detecting anoxic microsites based on
the abundance of anaerobic functional genes at several long-term agricultural
research experiments across the United States. We hypothesized that
(i) no-till and OM-amended soils would have greater evidence of anoxic
microsites than their conventional till and unamended counterparts
and (ii) anoxic microsites partially drive soil organic C concentration
in agricultural soils.
Methods 2.1 Site Description We collected soils
from four agricultural experiment stations representing a range of
climates and soil properties across the continental U.S. . All experiment
stations were part of Soil Health Institute’s North American
Project to Evaluate Soil Health Measurements (NAPESHM). Corn was planted at all sites, and soils were
sampled 3–4 weeks after planting, before V4 growth stage fertilizer
application, in the spring or summer of 2021. Annual climate averages
were obtained from the NAPESHM data set and are listed in . Antecedent precipitation
for date of sampling was obtained from the nearest public weather
stations. − Three of the four sites were long-term tillage trials:
Carrington
Research Extension Center (Carrington, ND), Sand Mountain Research
and Extension Center (Crossville, AL), and Wooster Ohio Agricultural
Research Development Center (Wooster, OH). For more detail on the
establishment of each long-term tillage trial, see refs. − For all three tillage trials, each unique treatment was maintained
in triplicate and sampled once from each replicate plot. Additionally,
we sampled uncultivated areas bordering agricultural fields in triplicate.
In Carrington, ND, the uncultivated area was a bordering grassland
overlain by a single row deciduous windbreak; in Crossville, AL, the
understory of nearby mature oak and/or oak forest; and in Wooster,
OH, grassland. The uncultivated areas were sampled once along each
edge of the cultivated area. We also sampled no-till and conventional
till plots amended with OM as manure at the site in Carrington, ND;
the unamended plots at Carrington received no additional N. Our fourth site, University of Missouri Grace Greenley Farm of
the Greenley Research Center (Novelty, MO), was a recently established,
replicated cover crop study in a no-till, terrace-tile system. For
an in-depth description of the site, please see ref . We sampled at three landscape
positions (channel, footslope, and shoulder) in three replicate terrace-tile
treatments, only from the no cover crop treatment. 2.2 Sampling Procedure Within each plot,
we collected both bulk and intact soils. Bulk soil was collected by
inserting two 2.5 cm diameter push probes to 15 cm depth. We focused
our analysis to the top 15 cm, as this depth is most likely to be
affected by tillage. Contents of both push probes were combined and
homogenized in a clean (wiped 3× with EtOH) tub with a sterile
scoop. Some of the homogenized soil was transferred to a 15 mL Nalgene
cryogenic vial and placed in a cooler with ice packs to preserve it
for DNA analyses. Intact soil cores were collected by inserting two
preweighed, 18 cm long, 5 cm diameter, sharpened PVC sections to a
15 cm depth. Cores were excavated, tightly capped, and stored in a
sealed plastic bag to maintain the original moisture content. Cryogenic vials were frozen within 12 h of sample collection, sent
on dry ice overnight to Stanford University, and received and stored
at −80 °C until DNA extraction. Remaining bulk samples
and intact soil cores were sent in a cooler with ice packs to Stanford
University and arrived within 7 days of sample collection. 2.3 General Soil Properties The volume
of each intact soil core was estimated by multiplying the mean of
four height measurements by the core surface area. Cores were weighed,
extruded, air-dried, and sieved to 2 mm to determine dry weight, plant
biomass (>2 mm) and coarse fraction (>2 mm). Dry bulk density
was
calculated with an assumed particle density of 2.65 g cm –3 . Soil clay content was determined using the hydrometer method. Air-dried soils were ground in a ball mill (SPEX
8000D Mixer Mill) and analyzed for organic C concentration and nitrogen
concentration using an elemental analyzer (NC Technologies, ECS 8020,
Milan, Italy). Inorganic carbon content was assumed to be negligible
as pH W of all samples was less than 7 (data not shown). We performed acid ammonium oxalate extractions
to approximate the short-range order (SRO) mineral content. In a 50
mL tube, 0.5 g of ground soil was shaken in the dark for 2 h with
30 mL of 0.2 M ammonium oxalate (pH = 3) according to Loeppert and
Inskeep (1996). Extracts were filtered
to 0.2 μm and analyzed for Fe, Al, and Mn by ICP-OES (iCAP 6300
Duo View, Thermo Scientific). Soil specific surface area (SSA)
was determined by the BET-N 2 method on hydrogen-peroxide-treated
soils (to remove organic
matter). Approximately 90–98% of soil organic C was removed
during this treatment as confirmed by elemental analyzer (NC Technologies,
ECS 8020, Milan, Italy, data not shown). Samples were analyzed on
a NovaTouch LX 2 surface area analyzer (Quantachrome Corporation,
Boynton Beach, FL). Samples were degassed at 105 °C under a vacuum
for 12 h prior to analysis. Surface area was calculated via the BET
equation from at least six data points over a relative vapor pressure
( P / P 0 ) range of 0.05
to 0.30. Finally, we obtained percentage of water stable aggregate
data
from the NAPESHM data set and recently published literature. , , − For cases in
which data were unavailable for a precise plot, we used water stable
aggregate data from similarly managed plots. See the Supporting Information for more detail. 2.4 DNA Extraction and Nucleic Acid Quantification DNA was extracted from approximately 0.25 g of field moist soil
using the DNeasy PowerSoil Pro Kit (QIAGEN), according to the manufacturer's
protocol. Extractions were performed in singlet, and extracts were
stored at −20 °C until droplet digital PCR (ddPCR) analysis. Using droplet digital PCR (ddPCR), we quantified genes directly
responsible for facilitating anaerobic metabolisms ( nirK and nirS – denitrification; dsrAB – sulfate reduction; mcrA – methanogenesis)
or genes unique to known anaerobes ( gltA – Geobacter , iron reduction). Unlike 16S rRNA sequencing,
which provides the relative abundance of specific taxonomic groups,
ddPCR allowed us to directly quantify organisms based on oxygen-sensitive
microbial functions. Furthermore, ddPCR is particularly well suited
for the detection and quantification of low-abundance targets and,
thus, allows for the detection and quantification of organisms that
would otherwise be missed in 16S rRNA sequencing or even traditional
quantitative PCR. Droplet digital PCR reaction conditions were
optimized prior to
the final analysis. Primer references and concentrations, cycling
conditions, and template DNA dilutions are listed in Supplementary Table S1 . For all targets, PCR reactions were
performed in 25 μL volumes, containing: 9 μL of nuclease
free water, 12.5 μL of EvaGreen SuperMix, 1 μL of forward
and reverse primers, 0.5 μL of bovine serum albumin (2.5 μg
μL –1 ), and 1 μL of template DNA. PCR
reactions were mixed, and 20 μL was emulsified using the Bio-Rad
QX200 Droplet Generator (Bio-Rad, Hercules, CA). The generated droplet
mixture (38 μL) was carefully transferred to a 96-well plate
and sealed. For each primer set, all samples from a single site were
run on the same plate, including a no-template control (nuclease free
water) and a positive control. PCR was performed on a Bio-Rad C1000
Touch Thermal Cycler. For all reactions, PCR protocols began with
a single initialization step of 5 min at 95 °C followed by 45
cycles of 1 min at 95 °C (denaturation) and annealing and extension
steps optimized for each target gene. All reactions ended with a 5
min hold at 4 °C, a 5 min hold at 90 °C (enzyme deactivation),
and infinite hold at 4 °C until droplet reading within 16 h.
Droplets were read on a Bio-Rad QX200 Droplet Reader and analyzed
using
Quantasoft software (Bio-Rad). In cases in which droplet separation
was suboptimal, threshold fluorescence was set manually based on the
fluorescence of the no-template control. Samples with fewer than 10 000
droplets were omitted from the data set. Quantasoft calculated target
copies were normalized to copies per gram of dry soil to enable comparisons
between management practices. 2.5 Statistical Analyses All statistical
analyses were performed using R version 4.2.2. We used a series of parametric and nonparametric tests
to compare the effect of tillage on anaerobe abundance. One high statistical
outlier of anaerobe abundance (Wooster, OH, uncultivated) was removed
from the data set. Data normality was confirmed using the Shapiro-Wilk
test, and the data were transformed to meet the assumptions of normality
where possible. For normal data, the Bartlett test was used to confirm
the equality of variances. The Levene’s test was used to confirm
equality of variances for non-normal data. We tested for differences
between conventional till, no-till, and uncultivated soils using (i)
ANOVA, when comparing data that were normally distributed with equal
variances, (ii) Welch’s ANOVA, for normally distributed data
with unequal variances, and (iii) Kruskal–Wallis test for non-normal
data. Posthoc mean separation was performed using Tukey’s Honest
Significant Difference, Games-Howell, and Dunn tests, respectively.
We tested if manure-amended soils had significantly greater anaerobe
abundances using one-tailed t tests (equal variances)
and Welch’s t tests (unequal variances). To examine the relationship between anaerobe copies (per g of soil)
and soil C concentration, we performed simple linear regression on
the overall data set ( n = 47) and within each site.
We also examined the relationship between anaerobe abundance and soil
C concentration using linear mixed effects modeling with the site
included as a random effect. The simple linear regression of anaerobe
abundance and soil C concentration for Crossville, AL, was performed
twice, both with and without a high outlier for anaerobe abundance
( Supplementary Figure S1 ). We performed
variance partitioning analysis to determine the unique
contribution of various soil C protection mechanisms (mineral, climate,
physical, and anoxic) to soil C concentration. The variables within
each grouping were finalized by eliminating multicollinearity by sequentially
deleting predictors with a VIF > 5 from the group. Our sample size
for each variance partitioning analysis ( n = 39–47)
agrees with others who have employed similar methods. , Variance partitioning analysis was done using the vegan package in R. Negative results for
the variance partitioning are reported as zeroes as negative values
signify lack of shared explained variance. To ensure that our analyses were not driven solely by high
denitrifier
abundance, we calculated a secondary, normalized proxy for anaerobe
abundance. We performed a principal component analysis on all genes
across the data set and found that the first principal component represented
an axis of anaerobe abundance ( Supplementary Figure S2 ). We repeated all analyses involving anaerobe abundance
with the principal component representation of anaerobe abundance
and report these secondary results in the Supplement and throughout.
Site Description We collected soils
from four agricultural experiment stations representing a range of
climates and soil properties across the continental U.S. . All experiment
stations were part of Soil Health Institute’s North American
Project to Evaluate Soil Health Measurements (NAPESHM). Corn was planted at all sites, and soils were
sampled 3–4 weeks after planting, before V4 growth stage fertilizer
application, in the spring or summer of 2021. Annual climate averages
were obtained from the NAPESHM data set and are listed in . Antecedent precipitation
for date of sampling was obtained from the nearest public weather
stations. − Three of the four sites were long-term tillage trials:
Carrington
Research Extension Center (Carrington, ND), Sand Mountain Research
and Extension Center (Crossville, AL), and Wooster Ohio Agricultural
Research Development Center (Wooster, OH). For more detail on the
establishment of each long-term tillage trial, see refs. − For all three tillage trials, each unique treatment was maintained
in triplicate and sampled once from each replicate plot. Additionally,
we sampled uncultivated areas bordering agricultural fields in triplicate.
In Carrington, ND, the uncultivated area was a bordering grassland
overlain by a single row deciduous windbreak; in Crossville, AL, the
understory of nearby mature oak and/or oak forest; and in Wooster,
OH, grassland. The uncultivated areas were sampled once along each
edge of the cultivated area. We also sampled no-till and conventional
till plots amended with OM as manure at the site in Carrington, ND;
the unamended plots at Carrington received no additional N. Our fourth site, University of Missouri Grace Greenley Farm of
the Greenley Research Center (Novelty, MO), was a recently established,
replicated cover crop study in a no-till, terrace-tile system. For
an in-depth description of the site, please see ref . We sampled at three landscape
positions (channel, footslope, and shoulder) in three replicate terrace-tile
treatments, only from the no cover crop treatment.
Sampling Procedure Within each plot,
we collected both bulk and intact soils. Bulk soil was collected by
inserting two 2.5 cm diameter push probes to 15 cm depth. We focused
our analysis to the top 15 cm, as this depth is most likely to be
affected by tillage. Contents of both push probes were combined and
homogenized in a clean (wiped 3× with EtOH) tub with a sterile
scoop. Some of the homogenized soil was transferred to a 15 mL Nalgene
cryogenic vial and placed in a cooler with ice packs to preserve it
for DNA analyses. Intact soil cores were collected by inserting two
preweighed, 18 cm long, 5 cm diameter, sharpened PVC sections to a
15 cm depth. Cores were excavated, tightly capped, and stored in a
sealed plastic bag to maintain the original moisture content. Cryogenic vials were frozen within 12 h of sample collection, sent
on dry ice overnight to Stanford University, and received and stored
at −80 °C until DNA extraction. Remaining bulk samples
and intact soil cores were sent in a cooler with ice packs to Stanford
University and arrived within 7 days of sample collection.
General Soil Properties The volume
of each intact soil core was estimated by multiplying the mean of
four height measurements by the core surface area. Cores were weighed,
extruded, air-dried, and sieved to 2 mm to determine dry weight, plant
biomass (>2 mm) and coarse fraction (>2 mm). Dry bulk density
was
calculated with an assumed particle density of 2.65 g cm –3 . Soil clay content was determined using the hydrometer method. Air-dried soils were ground in a ball mill (SPEX
8000D Mixer Mill) and analyzed for organic C concentration and nitrogen
concentration using an elemental analyzer (NC Technologies, ECS 8020,
Milan, Italy). Inorganic carbon content was assumed to be negligible
as pH W of all samples was less than 7 (data not shown). We performed acid ammonium oxalate extractions
to approximate the short-range order (SRO) mineral content. In a 50
mL tube, 0.5 g of ground soil was shaken in the dark for 2 h with
30 mL of 0.2 M ammonium oxalate (pH = 3) according to Loeppert and
Inskeep (1996). Extracts were filtered
to 0.2 μm and analyzed for Fe, Al, and Mn by ICP-OES (iCAP 6300
Duo View, Thermo Scientific). Soil specific surface area (SSA)
was determined by the BET-N 2 method on hydrogen-peroxide-treated
soils (to remove organic
matter). Approximately 90–98% of soil organic C was removed
during this treatment as confirmed by elemental analyzer (NC Technologies,
ECS 8020, Milan, Italy, data not shown). Samples were analyzed on
a NovaTouch LX 2 surface area analyzer (Quantachrome Corporation,
Boynton Beach, FL). Samples were degassed at 105 °C under a vacuum
for 12 h prior to analysis. Surface area was calculated via the BET
equation from at least six data points over a relative vapor pressure
( P / P 0 ) range of 0.05
to 0.30. Finally, we obtained percentage of water stable aggregate
data
from the NAPESHM data set and recently published literature. , , − For cases in
which data were unavailable for a precise plot, we used water stable
aggregate data from similarly managed plots. See the Supporting Information for more detail.
DNA Extraction and Nucleic Acid Quantification DNA was extracted from approximately 0.25 g of field moist soil
using the DNeasy PowerSoil Pro Kit (QIAGEN), according to the manufacturer's
protocol. Extractions were performed in singlet, and extracts were
stored at −20 °C until droplet digital PCR (ddPCR) analysis. Using droplet digital PCR (ddPCR), we quantified genes directly
responsible for facilitating anaerobic metabolisms ( nirK and nirS – denitrification; dsrAB – sulfate reduction; mcrA – methanogenesis)
or genes unique to known anaerobes ( gltA – Geobacter , iron reduction). Unlike 16S rRNA sequencing,
which provides the relative abundance of specific taxonomic groups,
ddPCR allowed us to directly quantify organisms based on oxygen-sensitive
microbial functions. Furthermore, ddPCR is particularly well suited
for the detection and quantification of low-abundance targets and,
thus, allows for the detection and quantification of organisms that
would otherwise be missed in 16S rRNA sequencing or even traditional
quantitative PCR. Droplet digital PCR reaction conditions were
optimized prior to
the final analysis. Primer references and concentrations, cycling
conditions, and template DNA dilutions are listed in Supplementary Table S1 . For all targets, PCR reactions were
performed in 25 μL volumes, containing: 9 μL of nuclease
free water, 12.5 μL of EvaGreen SuperMix, 1 μL of forward
and reverse primers, 0.5 μL of bovine serum albumin (2.5 μg
μL –1 ), and 1 μL of template DNA. PCR
reactions were mixed, and 20 μL was emulsified using the Bio-Rad
QX200 Droplet Generator (Bio-Rad, Hercules, CA). The generated droplet
mixture (38 μL) was carefully transferred to a 96-well plate
and sealed. For each primer set, all samples from a single site were
run on the same plate, including a no-template control (nuclease free
water) and a positive control. PCR was performed on a Bio-Rad C1000
Touch Thermal Cycler. For all reactions, PCR protocols began with
a single initialization step of 5 min at 95 °C followed by 45
cycles of 1 min at 95 °C (denaturation) and annealing and extension
steps optimized for each target gene. All reactions ended with a 5
min hold at 4 °C, a 5 min hold at 90 °C (enzyme deactivation),
and infinite hold at 4 °C until droplet reading within 16 h.
Droplets were read on a Bio-Rad QX200 Droplet Reader and analyzed
using
Quantasoft software (Bio-Rad). In cases in which droplet separation
was suboptimal, threshold fluorescence was set manually based on the
fluorescence of the no-template control. Samples with fewer than 10 000
droplets were omitted from the data set. Quantasoft calculated target
copies were normalized to copies per gram of dry soil to enable comparisons
between management practices.
Statistical Analyses All statistical
analyses were performed using R version 4.2.2. We used a series of parametric and nonparametric tests
to compare the effect of tillage on anaerobe abundance. One high statistical
outlier of anaerobe abundance (Wooster, OH, uncultivated) was removed
from the data set. Data normality was confirmed using the Shapiro-Wilk
test, and the data were transformed to meet the assumptions of normality
where possible. For normal data, the Bartlett test was used to confirm
the equality of variances. The Levene’s test was used to confirm
equality of variances for non-normal data. We tested for differences
between conventional till, no-till, and uncultivated soils using (i)
ANOVA, when comparing data that were normally distributed with equal
variances, (ii) Welch’s ANOVA, for normally distributed data
with unequal variances, and (iii) Kruskal–Wallis test for non-normal
data. Posthoc mean separation was performed using Tukey’s Honest
Significant Difference, Games-Howell, and Dunn tests, respectively.
We tested if manure-amended soils had significantly greater anaerobe
abundances using one-tailed t tests (equal variances)
and Welch’s t tests (unequal variances). To examine the relationship between anaerobe copies (per g of soil)
and soil C concentration, we performed simple linear regression on
the overall data set ( n = 47) and within each site.
We also examined the relationship between anaerobe abundance and soil
C concentration using linear mixed effects modeling with the site
included as a random effect. The simple linear regression of anaerobe
abundance and soil C concentration for Crossville, AL, was performed
twice, both with and without a high outlier for anaerobe abundance
( Supplementary Figure S1 ). We performed
variance partitioning analysis to determine the unique
contribution of various soil C protection mechanisms (mineral, climate,
physical, and anoxic) to soil C concentration. The variables within
each grouping were finalized by eliminating multicollinearity by sequentially
deleting predictors with a VIF > 5 from the group. Our sample size
for each variance partitioning analysis ( n = 39–47)
agrees with others who have employed similar methods. , Variance partitioning analysis was done using the vegan package in R. Negative results for
the variance partitioning are reported as zeroes as negative values
signify lack of shared explained variance. To ensure that our analyses were not driven solely by high
denitrifier
abundance, we calculated a secondary, normalized proxy for anaerobe
abundance. We performed a principal component analysis on all genes
across the data set and found that the first principal component represented
an axis of anaerobe abundance ( Supplementary Figure S2 ). We repeated all analyses involving anaerobe abundance
with the principal component representation of anaerobe abundance
and report these secondary results in the Supplement and throughout.
Results and Discussion 3.1 Anaerobe DNA: An Informative, Space-Time Integrated
Proxy for Anoxic Microsites We used DNA copies of anaerobic
functional genes over a confined volume (2 × 2.5 cm 2 diameter, 15 cm deep cores) as a proxy for anaerobe habitat space
and, hence, anoxic volume . Alternative indicators for anoxic microsites, such
as Fe(II) content,
anaerobe RNA, and CH 4 and N 2 O fluxes, reflect
anaerobic processes within a given volume at the time of sampling. However, we required a quantitative indicator for anoxic microsites
that was impervious to short-term weather conditions as anoxic microsites
may respond to changes in temperature and moisture. , DNA is relatively stable, remaining in soils for months to years. , Thus, bulk-sampled DNA represents a volume-averaged and near-past
time-integrated estimate of metabolic capacity within soils. We use the sum of the absolute abundances of anaerobe
target genes (referred to as “anaerobe abundance”) as
our proxy for anaerobe habitat space and thus anoxic microsites. We
focused our analysis on absolute anaerobe abundance
rather than abundance normalized to 16S rRNA copies
(i.e., relative abundance) for a variety of reasons. First, prokaryotes
can have multiple 16S rRNA copies per genome; thus, normalizing to 16S rRNA copies is typically not recommended. Additionally, there are multiple scenarios in which a higher relative
abundance of anaerobes would not reflect a higher absolute abundance
of anoxic microsites and vice versa. For example, microbial activity
in soils is known to be concentrated in biological “hotspots”,
regions of intense microbial activity that may serve as habitat space for both anaerobes and aerobes.
In a soil with comparatively more or a greater volume of hotspots,
the total abundance of both anaerobes and aerobes may increase, while
the relative abundance of anaerobes would remain unchanged. Thus, as we use it here, anaerobe abundance does not reveal the
relative proportion of aerobic vs anaerobic habitat space, nor can
it confirm actual function of anaerobic metabolisms. Furthermore,
the extraction of anaerobe DNA from bulk samples provides no information
about the relative distribution of anaerobes throughout the sampled
volume. Instead, the presence of anaerobe DNA represents two possibilities:
(i) sustained capacity for anaerobic respiration, possibly over a
relatively small volume, within the domain sampled, or (ii) short
bursts of capacity for anaerobic respiration, possibly over a relatively
large volume, within the domain sampled. We posit that soils with
more DNA copies of anaerobic genes, and thus a greater capacity for
anaerobic respiration, generally supported more anoxia over the soil
spacetime continuum within the domain sampled (i.e., the top 15 cm
of soil). 3.2 Anaerobe DNA Is Ubiquitous Target
genes associated with anaerobic metabolisms were found in every sample
of our data set, suggesting that anoxic microsites, or at least microorganisms
able to subsist under variable redox conditions, are present across
a broad diversity of upland, agricultural soils. Genes associated
with the most energetically favorable anaerobic metabolisms (e.g.,
denitrification, nirK and nirS )
were greatest in abundance, and genes associated with the least energetically
favorable metabolisms (e.g., methanogenesis, mcrA ) were least abundant . The relative distribution of the various genes suggests
that there are gradients in electron acceptors throughout the soil
space. Our work builds upon the work of others who have found methanogen
or other anaerobe DNA in upland soils. , − 3.3 Anaerobe Abundance Responds to Management Anaerobe abundance was largely similar between no-till (NT) and
conventional till (CT) soils. Only at the Wooster, OH site was anaerobe
abundance greater in NT vs CT soils ( , Supplementary Table S2 ), and marginally significantly greater when using the secondary
indicator for anaerobes ( p = 0.052, Supplementary Table S2 , Supplementary Figure S3 ). Typically, changes in topsoil C dynamics are observed
within 10–20 years of management implementation. The “youngest” long-term tillage
trial was established 34 years prior to sampling; thus, legacy effects
are unlikely to explain the difference, or lack thereof, between CT,
NT, and/or uncultivated soils. However, given obvious or potential
differences in site climate, DNA extraction efficiency, and/or management
legacy, we avoid explicit comparisons between sites and instead examine
treatment effects within sites for each of the three long-term tillage
trials. Tillage is well-known to enhance soil aeration, and previous work has shown that physical soil
disturbances, like tillage, can enhance soil oxygen supply and increase
microbial respiration. Thus, the lack
of effect in Carrington and Crossville is surprising and suggests
that enhanced oxygen supply post-tillage may be ephemeral. If we presume
anoxic microsites reform a short time after tillage, there may be
few discernible differences in DNA-based anaerobe abundance between
conventional till and no-till soils. However, the functional impact
of short periods of aeration on soil C degradation may be pronounced. − Alternatively, the lack of a tillage effect may suggest that limitations
on anoxic microsite formation within cropland soils are determined
more so by oxygen demand than they are by oxygen supply. We posit
that in soils with higher oxygen demand, such as those with greater
C inputs, the effect of the oxygen supply on anoxic microsites may
be more evident. The effect of manure amendments at the Carrington,
ND site on anaerobe
abundance further emphasized the role of bioavailable C and biological
oxygen demand in establishing anaerobe habitat space. No-till manure
amended soils had greater anaerobe abundance than NT unamended soils
( , Supplementary Table S3 , and Supplementary Figure S4 ). Although manure amendments may directly
add anaerobe DNA to soil, , it is unlikely that
the anaerobe DNA from manure application would overshadow anaerobe
abundance from the native soil community. In batch incubations, Ho
et al. (2015) found that 10% w/w manure amendments increased mcrA abundance in soils by stimulating soil-borne methanogens;
however, methanogens from the manure itself had little to no effect
on mcrA abundance. Thus, we interpret greater anaerobe
abundance in no-till manure amended soils to reflect greater soil
anaerobe habitat space in OM-amended soils. Nitrogen deposition has been shown to increase,
decrease, and have
no effect on our target gene abundances; − whereas numerous studies have
associated excess OM or high C concentrations with anoxic microsites.
Brewer et al. (2018) found that soils receiving OM amendments sustained
the highest rates of methanogenesis, compared with chemically fertilized
CT and NT soils. Kravchenko et al. (2017)
showed that OM residues, such as plant detritus, served as hotspots
for denitrification, an anaerobic metabolism, within upland soils. Therefore, we surmise that changes in anaerobe
abundance with OM amendments are largely driven by the addition of
organic C rather than an increase in N availability. Interestingly,
manure amendments had no effect on anaerobe abundance in CT soils
. Given that
tillage alone did not affect anaerobe abundance at Carrington , our results illustrate
the coupled effects of oxygen supply and demand; increased anoxic
microsites resulted only from the combination of increased oxygen
demand (through manure addition) and limited oxygen supply (through
NT management). Uncultivated soils represent a unique scenario
in which there are
presumably greater OM inputs due to continuous living and extensive
vegetation and minimal soil disturbance compared to croplands. Across
all three tillage experiments, nearby uncultivated areas had significantly
(or marginally significantly) greater anaerobe abundance than cultivated
(CT and NT) soils ( , Supplementary Figure S3 , Supplementary Table S2 ). Our results demonstrate
that the expansion of cropland (i.e., extensification) may decrease
the extent of anoxic microsites. By logical extension, our results
also suggest that implementing cropping practices that emulate the
uncultivated systems could enhance anoxic microsite formation. Specifically,
these findings reaffirm the joint effect of enhanced OM inputs and
minimal soil disturbance on anoxic microsite formation. Finally,
there were no apparent differences in anaerobe abundance
between landscape positions at Novelty, MO ( Supplementary Figure S5 ). Thus, the samples were not differentiated by landscape
position in any subsequent analyses. We posit that the lack of differences
observed across the hillslope is due to the recent establishment of
the site. 3.4 Microbial Proxies for Anoxic Microsites Partially
Explain Soil C Content Across the entire data set and within
each site except Novelty, MO, there was a clear and positive relationship
between anaerobe abundance and soil organic C . We hypothesize that the lack of relationship
at the Novelty, MO site is a result of a recently established field
site (5 years prior to sampling). Even when accounting for site as
a random effect and using our secondary metric for anaerobe abundance,
the relationship between soil C and anaerobe abundance remained significant
( Supplementary Table S4 and Supplementary Figure S6 ). The correlation between
anaerobes and soil C illustrates that anoxic microsites may contribute
— at least in part — to soil C protection. Of course,
correlation between soil C concentration and anaerobe abundance does
not imply a causal relationship (see ). To account for other possible predictors
of soil C, we performed variance partitioning analysis for all sites
and cultivated lands. Variance partitioning analysis allowed us to
account for collinearity between regressors by reporting the combined
and unique variances explained by groups of predictor variables. We
used soil C concentration as our response variable and groups of soil
and climate properties representing various soil C protection mechanisms
as our predictor variables. Variance partitioning analysis revealed that anoxic
microsites
uniquely contributed to soil C preservation in cropland soils. Consistent
with global studies, the intersection of mineralogy and climate explained
the greatest variance in soil C concentration across the entire data
set as well as within cropland soils . Across all
sites, anaerobe abundance explained 44% of the variance in soil C
concentration ( a, sum of values in “Anoxic” oval), but none of this
variance could be discerned from other soil C preservation factors.
However, within cropland soils, anaerobe abundance explained 5% of
the unique variance in soil C concentration–the largest unique
variance explained by a management-responsive soil C protection mechanism
( b). Using
our secondary proxy for anaerobe abundance amplified these patterns,
with our secondary metric for anaerobes explaining more than 13% of
the unique variance in soil C content in cropland soils ( Supplementary Figure S7 ). 3.5 Study Limitations and Implications for Management
of Cropland Soils This study imperfectly examines soil C
dynamics within agricultural systems. We sampled only a handful of
sites and treatments and, within each site, a single time point. Additionally,
our analysis focused solely on topsoils (0–15 cm), though it
is well-known that management practices, such as tillage and OM amendments,
can also alter subsoil C dynamics. − Rather than measuring
long-term biomass inputs, we assumed that within a single site, cultivated
soils had similar biomass inputs and uncultivated soils would have
greater biomass inputs than the cultivated soils. We did not directly
quantify mineral and/or particulate organic C, and we investigated
only a handful of soil C processes, neglecting to examine microbial
vs plant-derived C or carbon use efficiency – though the utility
of these measures is sometimes questioned. , In order to gain a more wholistic understanding of anoxic microsites
and soil C response to management, weather events, climate change,
and various soil properties, additional studies examining a broader
range of sites, soil depths, management practices, and soil processes
will be required. Although our results suggest that anoxic microsites
contribute to soil C preservation within agricultural systems, the
directionality of this relationship remains ambiguous. As indicated
by this study among others, , organic C can stimulate anoxic
microsites and, thus, anaerobe abundance. Therefore, the simple correlation
between anaerobe abundance (independent variable) and soil C concentration
(dependent variable) in cannot be interpreted as proof that anoxic
microsites unidirectionally drive changes in soil C concentration.
More likely, a positive feedback between soil C and anoxic microsites
exists. Soils with high organic C availability may stimulate anoxic
microsite formation by stimulating aerobe growth and ultimately anoxic
conditions (anaerobic habitat space), and the limited oxygen conditions
within those anoxic microsites may enhance soil C preservation. Although
our variance partitioning analysis and previous studies that link anoxic microsites
to enhanced soil C preservation , , suggest that anoxic microsites at least partially drive soil C concentration,
future studies must closely examine the directionality of the relationship
between anoxic microsites, anaerobe abundance, and soil C concentration. Finally, our understanding of the relationships between anoxic
microsites, soil C, and greenhouse dynamics is still evolving. Anoxic
microsites can host denitrification, reductive dissolution of Fe-oxides,
and methanogenesis, potentially undermining climate benefits afforded
by enhanced soil C retention. , − Whether anoxic microsites will confer soil C increases and net climate
benefits globally remains to be seen and must remain an urgent topic
of study. For this reason, we refrain from recommending management
practices to leverage anoxic microsites for net climate benefit. Despite its limitations, this study critically advances our understanding
of anoxic microsites and soil biogeochemistry more broadly. We found
anaerobic genes in every soil sampled, challenging the paradigm that
well-drained soils are entirely oxic. Additionally, our results demonstrate that anoxic microsites likely
respond to management and at least partially contribute to soil C
protection. Perhaps most critically, our work evokes questions about
the relative importance of various soil C protection mechanisms in
managed ecosystems. Mineral protection capacity is immutable, with
available mineral surface area largely dictating how much C can associate
with minerals. Furthermore, in our study,
proxies for mineral protection (i.e., soil specific surface area and
short-range order mineral content) explained less unique variance
in soil C than anoxic protection. Physical protection of soil C, such
as the occlusion of C within aggregates, presumably can be manipulated
through tillage practices. , However, our proxies
for physical protection (i.e., tillage and water stable aggregates)
explained only 2% of the unique variance in soil C. Climate explained
the most unique variance in the cropland data set (13%), but like
mineral protection, climate cannot be controlled and is expected to
become more erratic with climate change. Meanwhile, anoxic microsites
are seemingly responsive to management, and our proxy for anoxic microsites
explained 41% of the variance in soil C concentration in cropland
soils; even after excluding all possible multicollinearities with
proxies for other soil C protection mechanisms, anoxic microsites
uniquely explained 5% of the variance in soil C concentration in cropland
soils. Nearly all global plans for mitigating climate change
depend on
the restoration of soil organic C. , , Yet, global increases in demand for cropland and
declines in soil organic C stocks are projected over the coming decades. , Thus, it is essential to understand the mechanisms of soil C protection,
and all avenues for soil C accumulation must be explored. Continuing
to define anoxic microsites, their contribution to soil C content,
and the response of anoxic microsites to management remain urgent
topics of research.
Anaerobe DNA: An Informative, Space-Time Integrated
Proxy for Anoxic Microsites We used DNA copies of anaerobic
functional genes over a confined volume (2 × 2.5 cm 2 diameter, 15 cm deep cores) as a proxy for anaerobe habitat space
and, hence, anoxic volume . Alternative indicators for anoxic microsites, such
as Fe(II) content,
anaerobe RNA, and CH 4 and N 2 O fluxes, reflect
anaerobic processes within a given volume at the time of sampling. However, we required a quantitative indicator for anoxic microsites
that was impervious to short-term weather conditions as anoxic microsites
may respond to changes in temperature and moisture. , DNA is relatively stable, remaining in soils for months to years. , Thus, bulk-sampled DNA represents a volume-averaged and near-past
time-integrated estimate of metabolic capacity within soils. We use the sum of the absolute abundances of anaerobe
target genes (referred to as “anaerobe abundance”) as
our proxy for anaerobe habitat space and thus anoxic microsites. We
focused our analysis on absolute anaerobe abundance
rather than abundance normalized to 16S rRNA copies
(i.e., relative abundance) for a variety of reasons. First, prokaryotes
can have multiple 16S rRNA copies per genome; thus, normalizing to 16S rRNA copies is typically not recommended. Additionally, there are multiple scenarios in which a higher relative
abundance of anaerobes would not reflect a higher absolute abundance
of anoxic microsites and vice versa. For example, microbial activity
in soils is known to be concentrated in biological “hotspots”,
regions of intense microbial activity that may serve as habitat space for both anaerobes and aerobes.
In a soil with comparatively more or a greater volume of hotspots,
the total abundance of both anaerobes and aerobes may increase, while
the relative abundance of anaerobes would remain unchanged. Thus, as we use it here, anaerobe abundance does not reveal the
relative proportion of aerobic vs anaerobic habitat space, nor can
it confirm actual function of anaerobic metabolisms. Furthermore,
the extraction of anaerobe DNA from bulk samples provides no information
about the relative distribution of anaerobes throughout the sampled
volume. Instead, the presence of anaerobe DNA represents two possibilities:
(i) sustained capacity for anaerobic respiration, possibly over a
relatively small volume, within the domain sampled, or (ii) short
bursts of capacity for anaerobic respiration, possibly over a relatively
large volume, within the domain sampled. We posit that soils with
more DNA copies of anaerobic genes, and thus a greater capacity for
anaerobic respiration, generally supported more anoxia over the soil
spacetime continuum within the domain sampled (i.e., the top 15 cm
of soil).
Anaerobe DNA Is Ubiquitous Target
genes associated with anaerobic metabolisms were found in every sample
of our data set, suggesting that anoxic microsites, or at least microorganisms
able to subsist under variable redox conditions, are present across
a broad diversity of upland, agricultural soils. Genes associated
with the most energetically favorable anaerobic metabolisms (e.g.,
denitrification, nirK and nirS )
were greatest in abundance, and genes associated with the least energetically
favorable metabolisms (e.g., methanogenesis, mcrA ) were least abundant . The relative distribution of the various genes suggests
that there are gradients in electron acceptors throughout the soil
space. Our work builds upon the work of others who have found methanogen
or other anaerobe DNA in upland soils. , −
Anaerobe Abundance Responds to Management Anaerobe abundance was largely similar between no-till (NT) and
conventional till (CT) soils. Only at the Wooster, OH site was anaerobe
abundance greater in NT vs CT soils ( , Supplementary Table S2 ), and marginally significantly greater when using the secondary
indicator for anaerobes ( p = 0.052, Supplementary Table S2 , Supplementary Figure S3 ). Typically, changes in topsoil C dynamics are observed
within 10–20 years of management implementation. The “youngest” long-term tillage
trial was established 34 years prior to sampling; thus, legacy effects
are unlikely to explain the difference, or lack thereof, between CT,
NT, and/or uncultivated soils. However, given obvious or potential
differences in site climate, DNA extraction efficiency, and/or management
legacy, we avoid explicit comparisons between sites and instead examine
treatment effects within sites for each of the three long-term tillage
trials. Tillage is well-known to enhance soil aeration, and previous work has shown that physical soil
disturbances, like tillage, can enhance soil oxygen supply and increase
microbial respiration. Thus, the lack
of effect in Carrington and Crossville is surprising and suggests
that enhanced oxygen supply post-tillage may be ephemeral. If we presume
anoxic microsites reform a short time after tillage, there may be
few discernible differences in DNA-based anaerobe abundance between
conventional till and no-till soils. However, the functional impact
of short periods of aeration on soil C degradation may be pronounced. − Alternatively, the lack of a tillage effect may suggest that limitations
on anoxic microsite formation within cropland soils are determined
more so by oxygen demand than they are by oxygen supply. We posit
that in soils with higher oxygen demand, such as those with greater
C inputs, the effect of the oxygen supply on anoxic microsites may
be more evident. The effect of manure amendments at the Carrington,
ND site on anaerobe
abundance further emphasized the role of bioavailable C and biological
oxygen demand in establishing anaerobe habitat space. No-till manure
amended soils had greater anaerobe abundance than NT unamended soils
( , Supplementary Table S3 , and Supplementary Figure S4 ). Although manure amendments may directly
add anaerobe DNA to soil, , it is unlikely that
the anaerobe DNA from manure application would overshadow anaerobe
abundance from the native soil community. In batch incubations, Ho
et al. (2015) found that 10% w/w manure amendments increased mcrA abundance in soils by stimulating soil-borne methanogens;
however, methanogens from the manure itself had little to no effect
on mcrA abundance. Thus, we interpret greater anaerobe
abundance in no-till manure amended soils to reflect greater soil
anaerobe habitat space in OM-amended soils. Nitrogen deposition has been shown to increase,
decrease, and have
no effect on our target gene abundances; − whereas numerous studies have
associated excess OM or high C concentrations with anoxic microsites.
Brewer et al. (2018) found that soils receiving OM amendments sustained
the highest rates of methanogenesis, compared with chemically fertilized
CT and NT soils. Kravchenko et al. (2017)
showed that OM residues, such as plant detritus, served as hotspots
for denitrification, an anaerobic metabolism, within upland soils. Therefore, we surmise that changes in anaerobe
abundance with OM amendments are largely driven by the addition of
organic C rather than an increase in N availability. Interestingly,
manure amendments had no effect on anaerobe abundance in CT soils
. Given that
tillage alone did not affect anaerobe abundance at Carrington , our results illustrate
the coupled effects of oxygen supply and demand; increased anoxic
microsites resulted only from the combination of increased oxygen
demand (through manure addition) and limited oxygen supply (through
NT management). Uncultivated soils represent a unique scenario
in which there are
presumably greater OM inputs due to continuous living and extensive
vegetation and minimal soil disturbance compared to croplands. Across
all three tillage experiments, nearby uncultivated areas had significantly
(or marginally significantly) greater anaerobe abundance than cultivated
(CT and NT) soils ( , Supplementary Figure S3 , Supplementary Table S2 ). Our results demonstrate
that the expansion of cropland (i.e., extensification) may decrease
the extent of anoxic microsites. By logical extension, our results
also suggest that implementing cropping practices that emulate the
uncultivated systems could enhance anoxic microsite formation. Specifically,
these findings reaffirm the joint effect of enhanced OM inputs and
minimal soil disturbance on anoxic microsite formation. Finally,
there were no apparent differences in anaerobe abundance
between landscape positions at Novelty, MO ( Supplementary Figure S5 ). Thus, the samples were not differentiated by landscape
position in any subsequent analyses. We posit that the lack of differences
observed across the hillslope is due to the recent establishment of
the site.
Microbial Proxies for Anoxic Microsites Partially
Explain Soil C Content Across the entire data set and within
each site except Novelty, MO, there was a clear and positive relationship
between anaerobe abundance and soil organic C . We hypothesize that the lack of relationship
at the Novelty, MO site is a result of a recently established field
site (5 years prior to sampling). Even when accounting for site as
a random effect and using our secondary metric for anaerobe abundance,
the relationship between soil C and anaerobe abundance remained significant
( Supplementary Table S4 and Supplementary Figure S6 ). The correlation between
anaerobes and soil C illustrates that anoxic microsites may contribute
— at least in part — to soil C protection. Of course,
correlation between soil C concentration and anaerobe abundance does
not imply a causal relationship (see ). To account for other possible predictors
of soil C, we performed variance partitioning analysis for all sites
and cultivated lands. Variance partitioning analysis allowed us to
account for collinearity between regressors by reporting the combined
and unique variances explained by groups of predictor variables. We
used soil C concentration as our response variable and groups of soil
and climate properties representing various soil C protection mechanisms
as our predictor variables. Variance partitioning analysis revealed that anoxic
microsites
uniquely contributed to soil C preservation in cropland soils. Consistent
with global studies, the intersection of mineralogy and climate explained
the greatest variance in soil C concentration across the entire data
set as well as within cropland soils . Across all
sites, anaerobe abundance explained 44% of the variance in soil C
concentration ( a, sum of values in “Anoxic” oval), but none of this
variance could be discerned from other soil C preservation factors.
However, within cropland soils, anaerobe abundance explained 5% of
the unique variance in soil C concentration–the largest unique
variance explained by a management-responsive soil C protection mechanism
( b). Using
our secondary proxy for anaerobe abundance amplified these patterns,
with our secondary metric for anaerobes explaining more than 13% of
the unique variance in soil C content in cropland soils ( Supplementary Figure S7 ).
Study Limitations and Implications for Management
of Cropland Soils This study imperfectly examines soil C
dynamics within agricultural systems. We sampled only a handful of
sites and treatments and, within each site, a single time point. Additionally,
our analysis focused solely on topsoils (0–15 cm), though it
is well-known that management practices, such as tillage and OM amendments,
can also alter subsoil C dynamics. − Rather than measuring
long-term biomass inputs, we assumed that within a single site, cultivated
soils had similar biomass inputs and uncultivated soils would have
greater biomass inputs than the cultivated soils. We did not directly
quantify mineral and/or particulate organic C, and we investigated
only a handful of soil C processes, neglecting to examine microbial
vs plant-derived C or carbon use efficiency – though the utility
of these measures is sometimes questioned. , In order to gain a more wholistic understanding of anoxic microsites
and soil C response to management, weather events, climate change,
and various soil properties, additional studies examining a broader
range of sites, soil depths, management practices, and soil processes
will be required. Although our results suggest that anoxic microsites
contribute to soil C preservation within agricultural systems, the
directionality of this relationship remains ambiguous. As indicated
by this study among others, , organic C can stimulate anoxic
microsites and, thus, anaerobe abundance. Therefore, the simple correlation
between anaerobe abundance (independent variable) and soil C concentration
(dependent variable) in cannot be interpreted as proof that anoxic
microsites unidirectionally drive changes in soil C concentration.
More likely, a positive feedback between soil C and anoxic microsites
exists. Soils with high organic C availability may stimulate anoxic
microsite formation by stimulating aerobe growth and ultimately anoxic
conditions (anaerobic habitat space), and the limited oxygen conditions
within those anoxic microsites may enhance soil C preservation. Although
our variance partitioning analysis and previous studies that link anoxic microsites
to enhanced soil C preservation , , suggest that anoxic microsites at least partially drive soil C concentration,
future studies must closely examine the directionality of the relationship
between anoxic microsites, anaerobe abundance, and soil C concentration. Finally, our understanding of the relationships between anoxic
microsites, soil C, and greenhouse dynamics is still evolving. Anoxic
microsites can host denitrification, reductive dissolution of Fe-oxides,
and methanogenesis, potentially undermining climate benefits afforded
by enhanced soil C retention. , − Whether anoxic microsites will confer soil C increases and net climate
benefits globally remains to be seen and must remain an urgent topic
of study. For this reason, we refrain from recommending management
practices to leverage anoxic microsites for net climate benefit. Despite its limitations, this study critically advances our understanding
of anoxic microsites and soil biogeochemistry more broadly. We found
anaerobic genes in every soil sampled, challenging the paradigm that
well-drained soils are entirely oxic. Additionally, our results demonstrate that anoxic microsites likely
respond to management and at least partially contribute to soil C
protection. Perhaps most critically, our work evokes questions about
the relative importance of various soil C protection mechanisms in
managed ecosystems. Mineral protection capacity is immutable, with
available mineral surface area largely dictating how much C can associate
with minerals. Furthermore, in our study,
proxies for mineral protection (i.e., soil specific surface area and
short-range order mineral content) explained less unique variance
in soil C than anoxic protection. Physical protection of soil C, such
as the occlusion of C within aggregates, presumably can be manipulated
through tillage practices. , However, our proxies
for physical protection (i.e., tillage and water stable aggregates)
explained only 2% of the unique variance in soil C. Climate explained
the most unique variance in the cropland data set (13%), but like
mineral protection, climate cannot be controlled and is expected to
become more erratic with climate change. Meanwhile, anoxic microsites
are seemingly responsive to management, and our proxy for anoxic microsites
explained 41% of the variance in soil C concentration in cropland
soils; even after excluding all possible multicollinearities with
proxies for other soil C protection mechanisms, anoxic microsites
uniquely explained 5% of the variance in soil C concentration in cropland
soils. Nearly all global plans for mitigating climate change
depend on
the restoration of soil organic C. , , Yet, global increases in demand for cropland and
declines in soil organic C stocks are projected over the coming decades. , Thus, it is essential to understand the mechanisms of soil C protection,
and all avenues for soil C accumulation must be explored. Continuing
to define anoxic microsites, their contribution to soil C content,
and the response of anoxic microsites to management remain urgent
topics of research.
|
Reported Incidence of Infections Caused by Pathogens Transmitted Commonly Through Food: Impact of Increased Use of Culture-Independent Diagnostic Tests — Foodborne Diseases Active Surveillance Network, 1996–2023 | d197325c-56c9-41a4-9452-c053fb981c78 | 11221634 | Microbiology[mh] | Increased use of culture-independent diagnostic tests (CIDTs) affects observed trends in foodborne infection incidence.
During 2023, the incidence of eight domestically acquired infections transmitted commonly through food either increased or remained stable compared with 2016–2018, the baseline used to track progress toward disease reduction goals. Incidence of CIDT-diagnosed infection also increased during 2023.
CIDTs allow for diagnosis of infections that previously would have been undetected; recent increases in incidence appear to be driven by increased CIDT use. Continued surveillance is needed to monitor the impact of changing diagnostic practices on disease trends. Targeted prevention efforts are needed to reduce disease incidence.
Reducing foodborne disease incidence is a public health priority. This report summarizes preliminary 2023 Foodborne Diseases Active Surveillance Network (FoodNet) data and highlights efforts to increase the representativeness of FoodNet. During 2023, incidences of domestically acquired campylobacteriosis, Shiga toxin-producing Escherichia coli infection, yersiniosis, vibriosis, and cyclosporiasis increased, whereas those of listeriosis, salmonellosis, and shigellosis remained stable compared with incidences during 2016–2018, the baseline used for tracking progress towards federal disease reduction goals. During 2023, the incidence and percentage of infections diagnosed by culture-independent diagnostic tests (CIDTs) reported to FoodNet continued to increase, and the percentage of cases that yielded an isolate decreased, affecting observed trends in incidence. Because CIDTs allow for diagnosis of infections that previously would have gone undetected, lack of progress toward disease reduction goals might reflect changing diagnostic practices rather than an actual increase in incidence. Continued surveillance is needed to monitor the impact of changing diagnostic practices on disease trends, and targeted prevention efforts are needed to meet disease reduction goals. During 2023, FoodNet expanded its catchment area for the first time since 2004. This expansion improved the representativeness of the FoodNet catchment area, the ability of FoodNet to monitor trends in disease incidence, and the generalizability of FoodNet data.
Reducing the incidence of foodborne and enteric diseases is a public health priority. The Healthy People 2030 (HP2030) initiative established disease reduction goals for Campylobacter, Listeria , Salmonella, and Shiga toxin-producing Escherichia coli (STEC) infections . To evaluate progress toward HP2030 goals, CDC’s Foodborne Diseases Active Surveillance Network (FoodNet) monitors infections caused by eight pathogens transmitted commonly through food. This report summarizes preliminary 2023 surveillance data and describes changes in incidence compared with average annual incidence during 2016–2018, the reference period used by HP2030 .
Data Source FoodNet conducts active, population-based surveillance for laboratory-diagnosed Campylobacter, Cyclospora, Listeria, Salmonella, Shigella, STEC, Vibrio, and Yersinia infections and pediatric hemolytic uremic syndrome (HUS) at 10 U.S. sites; HUS is monitored because it can be a complication of STEC infection. FoodNet’s catchment area expanded during 2023 to include all of Colorado, and now represents 16% of the U.S. population (53.6 million persons); in 2023, the historic catchment area represented 15% of the U.S. population (51.0 million persons). Compared with the historic catchment area, the expansion increased representation for specific populations, including Hispanic or Latino ([Hispanic]; 8% increase), American Indian or Alaska Native (AI/AN; 8% increase), and Native Hawaiian or Pacific Islander (NH/PI; 6% increase) persons (FoodNet collects race and ethnicity as separate variables) as well as persons living in rural counties (10% increase). Laboratory Testing and Data Collection Bacterial infections were diagnosed by culture or culture-independent diagnostic tests (CIDTs). Cyclosporiasis was diagnosed by polymerase chain reaction or microscopy. Pediatric HUS surveillance is conducted through a network of nephrologists and infection preventionists and by hospital discharge data review. This report includes 2022 data on pediatric HUS cases, the most recent year for which data are available. This activity was reviewed by CDC, deemed not research, and conducted in accordance with applicable federal law and CDC policy. Statistical Methods Bayesian negative binomial models were implemented to estimate changes in incidence in the historic catchment area during 2023 compared with average annual incidence during 2016–2018 (overall, and for domestically acquired infections), using R statistical software (version 2.14.0; R Foundation). , Incidence in 2023 was considered substantially different from that during 2016–2018 if the 95% credible interval (CrI) for the incidence rate ratio (IRR) did not include 1.0. Cross-tabulations by demographic and other characteristics were also performed.
FoodNet conducts active, population-based surveillance for laboratory-diagnosed Campylobacter, Cyclospora, Listeria, Salmonella, Shigella, STEC, Vibrio, and Yersinia infections and pediatric hemolytic uremic syndrome (HUS) at 10 U.S. sites; HUS is monitored because it can be a complication of STEC infection. FoodNet’s catchment area expanded during 2023 to include all of Colorado, and now represents 16% of the U.S. population (53.6 million persons); in 2023, the historic catchment area represented 15% of the U.S. population (51.0 million persons). Compared with the historic catchment area, the expansion increased representation for specific populations, including Hispanic or Latino ([Hispanic]; 8% increase), American Indian or Alaska Native (AI/AN; 8% increase), and Native Hawaiian or Pacific Islander (NH/PI; 6% increase) persons (FoodNet collects race and ethnicity as separate variables) as well as persons living in rural counties (10% increase).
Bacterial infections were diagnosed by culture or culture-independent diagnostic tests (CIDTs). Cyclosporiasis was diagnosed by polymerase chain reaction or microscopy. Pediatric HUS surveillance is conducted through a network of nephrologists and infection preventionists and by hospital discharge data review. This report includes 2022 data on pediatric HUS cases, the most recent year for which data are available. This activity was reviewed by CDC, deemed not research, and conducted in accordance with applicable federal law and CDC policy.
Bayesian negative binomial models were implemented to estimate changes in incidence in the historic catchment area during 2023 compared with average annual incidence during 2016–2018 (overall, and for domestically acquired infections), using R statistical software (version 2.14.0; R Foundation). , Incidence in 2023 was considered substantially different from that during 2016–2018 if the 95% credible interval (CrI) for the incidence rate ratio (IRR) did not include 1.0. Cross-tabulations by demographic and other characteristics were also performed.
Incidence in 2023 Compared with Average Annual Incidence During 2016–2018 During 2023, FoodNet identified 29,607 infections, 7,234 hospitalizations, and 177 deaths overall (including domestically acquired and travel-associated infections) in the historic catchment area , compared with 31,492 infections, 7,588 hospitalizations, and 184 deaths in the expanded catchment area (Supplementary Table; https://stacks.cdc.gov/view/cdc/157822 ). In both the historic and expanded catchment areas, 15% of cases were associated with international travel. Overall, and for domestically acquired infections only, incidence of campylobacteriosis was highest, followed by salmonellosis and STEC infection. In the historic catchment area during 2023, incidences of domestically acquired campylobacteriosis, cyclosporiasis, STEC infection, vibriosis, and yersiniosis increased compared with those during 2016–2018, whereas listeriosis, salmonellosis, and shigellosis incidences remained stable. Generally, the overall percentage of infections attributable to specific Campylobacter, Shigella, Vibrio, and Yersinia species, Salmonella serotypes, and STEC serogroups was lower in 2023 than in all previous years . The overall incidence of infections for which the pathogen was not speciated, serotyped, or serogrouped increased substantially compared with incidence during 2016–2018 . During 2023, 78% of all bacterial infections were diagnosed by CIDTs in the historic catchment area, including 46% diagnosed using only CIDTs. The percentage of CIDT-diagnosed infections for which a reflex culture was attempted decreased from 71% during 2016–2018 to 68% during 2023. This decrease was largest for Yersinia , Vibrio , and STEC infections. For all illnesses except listeriosis , the percentage of reflex cultures that yielded an isolate (successful [or positive] reflex culture) was lower in 2023 than during previous years . This decrease in isolate availability has been associated with a decrease in serotyped, serogrouped, and speciated infections. For example, from 2016–2018 to 2023, the overall incidence of unspeciated infections increased substantially for Campylobacter , Shigella , Yersinia , and Vibrio ; the percentage of speciated infections declined from 33% to 26% for Campylobacter , from 65% to 41% for Shigella , from 49% to 23% for Yersinia , and from 61% to 34% for Vibrio . Although only culture-independent methods are used to diagnose cyclosporiasis, increases in CIDT-diagnosed cyclosporiasis and cyclosporiasis incidence mirror CIDT-driven increases in bacterial infection incidence. Salmonella Infections Of 8,454 total (i.e., both domestically acquired and travel-associated) Salmonella infections during 2023 in the historic catchment area, 83% yielded an isolate; 89% of isolates were fully serotyped. The incidence of nonserotyped infections increased substantially. The incidences of the most frequently reported serotypes, S. Enteritidis and S. Newport, remained stable during 2023 compared with those during 2016–2018, whereas the incidences of the next-most frequently reported serotypes, S. Typhimurium, S. Javiana, and S. I 4,[5],12:i:- decreased substantially. STEC Infections Of 3,351 total STEC infections in the historic catchment area during 2023, 57% yielded an isolate; 87% of isolates were fully serogrouped. The incidence of nonserogrouped infections increased substantially in 2023 compared with that during 2016–2018. During 2023, STEC O157 incidence decreased compared with incidence during 2016–2018, and non-O157 STEC incidence remained stable. Hemolytic Uremic Syndrome During 2022, FoodNet identified 61 cases of postdiarrheal HUS in persons aged <18 years, including 39 among children aged <5 years. The incidence of postdiarrheal HUS among persons aged <18 years (0.6 per 100,000 persons) and those aged <5 years (1.4 per 100,000) remained stable in 2022 compared with that during 2016–2018.
During 2023, FoodNet identified 29,607 infections, 7,234 hospitalizations, and 177 deaths overall (including domestically acquired and travel-associated infections) in the historic catchment area , compared with 31,492 infections, 7,588 hospitalizations, and 184 deaths in the expanded catchment area (Supplementary Table; https://stacks.cdc.gov/view/cdc/157822 ). In both the historic and expanded catchment areas, 15% of cases were associated with international travel. Overall, and for domestically acquired infections only, incidence of campylobacteriosis was highest, followed by salmonellosis and STEC infection. In the historic catchment area during 2023, incidences of domestically acquired campylobacteriosis, cyclosporiasis, STEC infection, vibriosis, and yersiniosis increased compared with those during 2016–2018, whereas listeriosis, salmonellosis, and shigellosis incidences remained stable. Generally, the overall percentage of infections attributable to specific Campylobacter, Shigella, Vibrio, and Yersinia species, Salmonella serotypes, and STEC serogroups was lower in 2023 than in all previous years . The overall incidence of infections for which the pathogen was not speciated, serotyped, or serogrouped increased substantially compared with incidence during 2016–2018 . During 2023, 78% of all bacterial infections were diagnosed by CIDTs in the historic catchment area, including 46% diagnosed using only CIDTs. The percentage of CIDT-diagnosed infections for which a reflex culture was attempted decreased from 71% during 2016–2018 to 68% during 2023. This decrease was largest for Yersinia , Vibrio , and STEC infections. For all illnesses except listeriosis , the percentage of reflex cultures that yielded an isolate (successful [or positive] reflex culture) was lower in 2023 than during previous years . This decrease in isolate availability has been associated with a decrease in serotyped, serogrouped, and speciated infections. For example, from 2016–2018 to 2023, the overall incidence of unspeciated infections increased substantially for Campylobacter , Shigella , Yersinia , and Vibrio ; the percentage of speciated infections declined from 33% to 26% for Campylobacter , from 65% to 41% for Shigella , from 49% to 23% for Yersinia , and from 61% to 34% for Vibrio . Although only culture-independent methods are used to diagnose cyclosporiasis, increases in CIDT-diagnosed cyclosporiasis and cyclosporiasis incidence mirror CIDT-driven increases in bacterial infection incidence.
Infections Of 8,454 total (i.e., both domestically acquired and travel-associated) Salmonella infections during 2023 in the historic catchment area, 83% yielded an isolate; 89% of isolates were fully serotyped. The incidence of nonserotyped infections increased substantially. The incidences of the most frequently reported serotypes, S. Enteritidis and S. Newport, remained stable during 2023 compared with those during 2016–2018, whereas the incidences of the next-most frequently reported serotypes, S. Typhimurium, S. Javiana, and S. I 4,[5],12:i:- decreased substantially.
Of 3,351 total STEC infections in the historic catchment area during 2023, 57% yielded an isolate; 87% of isolates were fully serogrouped. The incidence of nonserogrouped infections increased substantially in 2023 compared with that during 2016–2018. During 2023, STEC O157 incidence decreased compared with incidence during 2016–2018, and non-O157 STEC incidence remained stable.
During 2022, FoodNet identified 61 cases of postdiarrheal HUS in persons aged <18 years, including 39 among children aged <5 years. The incidence of postdiarrheal HUS among persons aged <18 years (0.6 per 100,000 persons) and those aged <5 years (1.4 per 100,000) remained stable in 2022 compared with that during 2016–2018.
The current findings and previous FoodNet reports suggest a lack of progress toward foodborne disease reduction goals; however, this outcome might reflect changing diagnostic practices such as the increased use of CIDTs rather than an actual increase in disease incidence. Increased use of CIDTs facilitates prompt clinical diagnosis and treatment but also complicates the interpretation of surveillance data and trends because CIDT adoption has varied over time, among clinical labs, and by pathogen. In addition, although CIDTs are generally considered more sensitive than are culture-based methods, some have high false-positive rates for certain pathogens (e.g., Vibrio ) . Previous studies have indicated that increased CIDT use has resulted in the diagnosis of infections that previously would have gone undetected; increased use of CIDTs has been associated with marked increases in reported incidence . Increases in CIDT-diagnosed infections are also associated with decreased rates of reflex culture, thereby reducing the number of isolates available for subtyping, whole genome sequencing, and antimicrobial resistance characterization . The impact of this reduction differs by species, serotype, and serogroup. Because an isolate is required for speciation, serotyping, and serogrouping, reduced isolate availability might result in underdetection of illnesses attributable to specific Campylobacter, Shigella, Vibrio, and Yersinia species, Salmonella serotypes, and STEC serogroups. The substantial increase in the incidence of infections for which the pathogen was not speciated, serotyped, or serogrouped is likely an artifact of changing diagnostic practices (i.e., increased CIDT use), resulting in a reduced availability of isolates for speciation and typing. Continued reductions in isolate availability might hinder outbreak identification and response (e.g., whole genome sequencing–based cluster identification and source attribution), detection of emerging antimicrobial resistance, and tracking of trends in illnesses attributable to specific species, subtypes, serotypes, and resistant strains. Increasing successful reflex culture rates after a CIDT diagnosis is a public health priority, which requires focused efforts and resources at the federal, state, and local levels. FoodNet data are used to track trends in enteric illness, monitor progress toward disease reduction goals, and guide food safety policy , . Because FoodNet is a sentinel surveillance system representing 10 sites, national extrapolation relies on strong assumptions of representativeness. Although the sites included in the FoodNet catchment area were selected nonrandomly, past analyses suggest that FoodNet’s catchment area is broadly representative of the national population . Previously, the only notable difference between FoodNet’s historic catchment area and the national population identified by these studies was that Hispanic persons were underrepresented in the catchment area relative to national representation . Investigating enteric disease epidemiology for AI/AN and NH/PI persons using FoodNet data has also been complicated by the small size of these populations in the historic catchment area. By increasing representation for these specific populations in the FoodNet catchment area, FoodNet’s expansion has helped to partially alleviate these limitations and improve the generalizability of FoodNet data. Additional expansion might be needed as national and catchment area demographics change. Limitations The findings in this report are subject to at least three limitations. First, underreporting might affect case counts because ill persons must seek care and be tested for their illness to be recorded as a case. Second, although ill persons might meet the FoodNet criteria for hospitalization or death, the underlying reason for hospitalization or death might be unknown. Deaths that occurred >1 week after specimen collection among nonhospitalized persons or after discharge for hospitalized persons might not be recorded. Finally, domestically acquired cases might be overestimated because of the inclusion of persons with unknown travel status. Implications for Public Health Practice FoodNet’s surveillance efforts are critical for tracking foodborne and enteric illnesses in the United States. During 2023, FoodNet expanded its catchment area for the first time since 2004, and it now includes all of Colorado. This expansion improved the representativeness of the FoodNet catchment area, and the ability of FoodNet to monitor trends in disease incidence, including the impact of changing diagnostic practices and the generalizability of FoodNet data. Continued surveillance is needed to monitor the impact of changing diagnostic practices on disease trends and evaluate the efficacy of prevention efforts in reducing incidence.
The findings in this report are subject to at least three limitations. First, underreporting might affect case counts because ill persons must seek care and be tested for their illness to be recorded as a case. Second, although ill persons might meet the FoodNet criteria for hospitalization or death, the underlying reason for hospitalization or death might be unknown. Deaths that occurred >1 week after specimen collection among nonhospitalized persons or after discharge for hospitalized persons might not be recorded. Finally, domestically acquired cases might be overestimated because of the inclusion of persons with unknown travel status.
FoodNet’s surveillance efforts are critical for tracking foodborne and enteric illnesses in the United States. During 2023, FoodNet expanded its catchment area for the first time since 2004, and it now includes all of Colorado. This expansion improved the representativeness of the FoodNet catchment area, and the ability of FoodNet to monitor trends in disease incidence, including the impact of changing diagnostic practices and the generalizability of FoodNet data. Continued surveillance is needed to monitor the impact of changing diagnostic practices on disease trends and evaluate the efficacy of prevention efforts in reducing incidence.
|
Preparation and Reliability and Validity Test of the Questionnaire on the Maintenance of Intravenous Catheter in | bb3a51f2-c22d-45c5-81d4-8a48496dfd53 | 11789573 | Surgical Procedures, Operative[mh] | Introduction A central venous catheter (CVC) is a large‐bore CVC placed through the skin using sterile techniques in certain clinical situations. In adult patients, the three placement sites for CVCs are the internal jugular vein, femoral vein, and subclavian vein, with the catheter reaching the right atrium via the puncture site (Bleichmder ). CVCs have been widely used in the clinical treatment of critically ill patients and are an important route for monitoring the condition of critically ill patients, infusing fluids and blood, providing total parenteral nutrition and administering life‐saving drugs (Lafuente Cabrero et al. ). However, due to the fact that the patients' conditions are generally severe, invasive procedures are often performed, and their immune function is low, deep venous catheterization is associated with an increasing number of complications, including catheter‐related bloodstream infections, catheter occlusion, slippage, pneumothorax and air embolism (Wang, Sun, and Wang ). Of these, improper maintenance is the main cause of complications. Nurses monitor CVCs to prevent complications such as infection, pneumothorax, hematoma, bleeding or extravasation, so that corrective measures can be taken in a timely manner to improve medical care (Sun et al. ). The knowledge, attitude, and practices (KAP) of CVC care among intensive care unit (ICU) nurses have a significant impact on the occurrence of CVC complications. To date, no specific questionnaires have been found to assess the KAP of ICU nurses regarding CVC maintenance, making it difficult to standardise and improve the regulation of CVC maintenance by ICU nurses. Therefore, this study developed an ICU Nurse Central Venous Catheter Maintenance Knowledge‐Attitude‐Practice Questionnaire to provide a reliable theoretical basis for evaluating ICU nurses' knowledge, attitudes and behaviours towards CVC maintenance.
Materials and Methods 2.1 Research Object 2.1.1 Consulting Experts by Mail Medical experts and psychological experts with intermediate technical titles and above engaged in critical care, clinical medicine, nursing management for more than 10 years; (1) Bachelor degree or above; (2) Interested and willing to participate in the study. The study selected 19 experts from east, west and central China. 2.1.2 Pre‐Survey Subjects In July 2020, 20 ICU nurses from REDACTED were selected by convenience sampling to conduct a pre‐survey on questionnaires. The eligibility criteria are: (1) Have a nurse practicing qualification certificate issued by the Ministry of Health within the valid registration period; (2) Years of front‐line nursing work in ICU > 1 year; (3) Able to correctly understand the questionnaire content; (4) Informed consent and voluntary participation in this survey. Exclusion criteria: training, rotation, practice nurses; Nurses who are out of work due to marriage, illness, childbirth, etc. 2.1.3 Formal Survey Subjects were selected by convenience sampling method. From July to September 2020, on‐site questionnaire survey was conducted among ICU nurses in REDACTED. The inclusion and exclusion criteria are the same. 2.2 Research Methods 2.2.1 Preliminary Questionnaire Construction A multidisciplinary team consisting of two critical care clinical experts, one critical care research expert, one intravenous therapy expert, one psychological measurement expert, one statistician and three graduate students was established. Based on the theory of knowledge, belief and action (Li and Liu ) and referring to domestic and foreign research and related guidelines (Gorski et al. ; Chinese Nursing Association intravenous Infusion Therapy Professional Committee ; Estrada‐Orozco et al. ; National Health and Family Planning Commission ) as well as expert group discussions, the initial questionnaire compiled included 14 items in the knowledge dimension, 8 items in the attitude dimension, and 26 items in the behaviour dimension. Using the Delphi method, two rounds of expert correspondence were conducted from July to September 2020. According to the concentration degree of expert opinions, coordination degree of expert opinions and other indicators, combined with expert suggestions, the initial items of the questionnaire were screened and modified to form a pre‐test version of the questionnaire. The questionnaire contains 55 items in total, including 16 items in the knowledge dimension, 9 items in the attitude dimension and 30 items in the behaviour dimension. Five points are scored for all correct answers to knowledge items, 1 point is scored for all correct answers to multiple choice items, and 0 points is scored for all wrong answers. All attitude and behaviour items are scored on the Likert five‐point scoring scale (Likert ): strongly agree, agree, not necessarily, disagree and strongly disagree five kinds of answers, which are recorded as 5, 4, 3, 2 and 1 points respectively. 2.2.2 Pre‐Survey The pre‐test version of questionnaire was used to conduct a questionnaire survey on 20 nurses who met the inclusion criteria. The questionnaire was collected on site with an effective recovery rate of 100%. After the preliminary survey, no nurses put forward new demands, and the questionnaire content was not modified. So the initial version of the questionnaire consisted of three dimensions and 55 items for clinical testing. 2.2.3 Formal Investigation An on‐site questionnaire survey was conducted on 360 nurses who met the inclusion criteria using the self‐prepared general data questionnaire of nurses and the CVC maintenance questionnaire of ICU nurses (clinical test version), and 334 valid questionnaires were collected. 2.2.4 Item Analysis Items are screened based on data from formal surveys. The following two methods are mainly adopted in this study: (1) critical ratio (CR): also known as extreme value method, that is, the scores of all subjects are sorted from the largest to the smallest. If the score is in the top 27%, that is, the sample of the high group, while the score is in the bottom 27%, that is, the sample of the low group. Through the t ‐test analysis of two independent samples, if the CR value of the item is > 3 or the difference is statistically significant ( p < 0.05), the item will be retained; otherwise, it will be deleted (Xiaoyong ). (2) Total correlation analysis: The correlation coefficient between each item and the total score was calculated. If the correlation coefficient was not significant ( p > 0.05) or r < 0.4, the item was deleted (Xiaoyong ). Based on the project analysis results, a questionnaire on the maintenance of CVC for ICU nurses was formed (preliminary version). 2.2.5 Reliability and Validity Analysis The content validity of this study was obtained by letter consultation with experts. The validity of questionnaire structure was tested by exploratory factor analysis. The reliability is tested by internal consistency reliability and retest reliability. 2.3 Quality Control After unified training, the quality control investigators introduced the research purpose and precautions to the nurses with unified guidance. After completing the questionnaire, they checked the filling status on the spot and corrected and supplemented the errors and omissions in time. 2.4 Ethical Principles This study follows the principles of voluntariness and confidentiality to avoid any harm to the participants. The study was reviewed and approved by the Ethics Committee of REDACTED (Approval number: REDACTED). 2.5 Statistical Methods Excel 2013 was used for data entry and sorting. SPSS 26.0 and Amos24.0 were used for statistical analysis of the data. The measurement data were described by mean ± standard deviation, and the counting data were described by frequency and percentage. The difference was statistically significant when p < 0.05. The correlation between items and total score and the critical ratio method were used to screen and analyse the questionnaire items (Xiaoyong ). Factor analysis was used to test the validity of the questionnaire. The validity analysis included content validity and structure validity. Expert evaluation method was used for Content Validity. Item of Content Validity Index (I‐CVI): For each item, the number of experts with a rating of 3 or 4 divided by the total number of experts participating in the evaluation is the I‐CVI. Scale of Content Validity Index (S‐CVI): The number of times of 3 or 4 ratings divided by the number of evaluations. Exploratory factor analysis and confirmatory factor analysis were used for structural validity. Cronbach α coefficient and broken half reliability were used to describe the validity of the questionnaire. The validity of item level content and item level content were calculated by using expert evaluation method. Exploratory factor analysis was used to test the validity of the structure. If KMO ≥ 0.8, p < 0.05 indicates that exploratory factor analysis is suitable. Principal component analysis is used to extract common factors with eigenvalues > 1, and maximum variance orthogonal spin method is used to delete items with factor loads < 0.4 (Zhang and Dong ). Confirmatory factor analysis was used to test the fit degree of each dimension and item of the questionnaire. Cronbach's α and broken half reliability were used to evaluate the reliability of the questionnaire (Wu ).
Research Object 2.1.1 Consulting Experts by Mail Medical experts and psychological experts with intermediate technical titles and above engaged in critical care, clinical medicine, nursing management for more than 10 years; (1) Bachelor degree or above; (2) Interested and willing to participate in the study. The study selected 19 experts from east, west and central China. 2.1.2 Pre‐Survey Subjects In July 2020, 20 ICU nurses from REDACTED were selected by convenience sampling to conduct a pre‐survey on questionnaires. The eligibility criteria are: (1) Have a nurse practicing qualification certificate issued by the Ministry of Health within the valid registration period; (2) Years of front‐line nursing work in ICU > 1 year; (3) Able to correctly understand the questionnaire content; (4) Informed consent and voluntary participation in this survey. Exclusion criteria: training, rotation, practice nurses; Nurses who are out of work due to marriage, illness, childbirth, etc. 2.1.3 Formal Survey Subjects were selected by convenience sampling method. From July to September 2020, on‐site questionnaire survey was conducted among ICU nurses in REDACTED. The inclusion and exclusion criteria are the same.
Consulting Experts by Mail Medical experts and psychological experts with intermediate technical titles and above engaged in critical care, clinical medicine, nursing management for more than 10 years; (1) Bachelor degree or above; (2) Interested and willing to participate in the study. The study selected 19 experts from east, west and central China.
Pre‐Survey Subjects In July 2020, 20 ICU nurses from REDACTED were selected by convenience sampling to conduct a pre‐survey on questionnaires. The eligibility criteria are: (1) Have a nurse practicing qualification certificate issued by the Ministry of Health within the valid registration period; (2) Years of front‐line nursing work in ICU > 1 year; (3) Able to correctly understand the questionnaire content; (4) Informed consent and voluntary participation in this survey. Exclusion criteria: training, rotation, practice nurses; Nurses who are out of work due to marriage, illness, childbirth, etc.
Formal Survey Subjects were selected by convenience sampling method. From July to September 2020, on‐site questionnaire survey was conducted among ICU nurses in REDACTED. The inclusion and exclusion criteria are the same.
Research Methods 2.2.1 Preliminary Questionnaire Construction A multidisciplinary team consisting of two critical care clinical experts, one critical care research expert, one intravenous therapy expert, one psychological measurement expert, one statistician and three graduate students was established. Based on the theory of knowledge, belief and action (Li and Liu ) and referring to domestic and foreign research and related guidelines (Gorski et al. ; Chinese Nursing Association intravenous Infusion Therapy Professional Committee ; Estrada‐Orozco et al. ; National Health and Family Planning Commission ) as well as expert group discussions, the initial questionnaire compiled included 14 items in the knowledge dimension, 8 items in the attitude dimension, and 26 items in the behaviour dimension. Using the Delphi method, two rounds of expert correspondence were conducted from July to September 2020. According to the concentration degree of expert opinions, coordination degree of expert opinions and other indicators, combined with expert suggestions, the initial items of the questionnaire were screened and modified to form a pre‐test version of the questionnaire. The questionnaire contains 55 items in total, including 16 items in the knowledge dimension, 9 items in the attitude dimension and 30 items in the behaviour dimension. Five points are scored for all correct answers to knowledge items, 1 point is scored for all correct answers to multiple choice items, and 0 points is scored for all wrong answers. All attitude and behaviour items are scored on the Likert five‐point scoring scale (Likert ): strongly agree, agree, not necessarily, disagree and strongly disagree five kinds of answers, which are recorded as 5, 4, 3, 2 and 1 points respectively. 2.2.2 Pre‐Survey The pre‐test version of questionnaire was used to conduct a questionnaire survey on 20 nurses who met the inclusion criteria. The questionnaire was collected on site with an effective recovery rate of 100%. After the preliminary survey, no nurses put forward new demands, and the questionnaire content was not modified. So the initial version of the questionnaire consisted of three dimensions and 55 items for clinical testing. 2.2.3 Formal Investigation An on‐site questionnaire survey was conducted on 360 nurses who met the inclusion criteria using the self‐prepared general data questionnaire of nurses and the CVC maintenance questionnaire of ICU nurses (clinical test version), and 334 valid questionnaires were collected. 2.2.4 Item Analysis Items are screened based on data from formal surveys. The following two methods are mainly adopted in this study: (1) critical ratio (CR): also known as extreme value method, that is, the scores of all subjects are sorted from the largest to the smallest. If the score is in the top 27%, that is, the sample of the high group, while the score is in the bottom 27%, that is, the sample of the low group. Through the t ‐test analysis of two independent samples, if the CR value of the item is > 3 or the difference is statistically significant ( p < 0.05), the item will be retained; otherwise, it will be deleted (Xiaoyong ). (2) Total correlation analysis: The correlation coefficient between each item and the total score was calculated. If the correlation coefficient was not significant ( p > 0.05) or r < 0.4, the item was deleted (Xiaoyong ). Based on the project analysis results, a questionnaire on the maintenance of CVC for ICU nurses was formed (preliminary version). 2.2.5 Reliability and Validity Analysis The content validity of this study was obtained by letter consultation with experts. The validity of questionnaire structure was tested by exploratory factor analysis. The reliability is tested by internal consistency reliability and retest reliability.
Preliminary Questionnaire Construction A multidisciplinary team consisting of two critical care clinical experts, one critical care research expert, one intravenous therapy expert, one psychological measurement expert, one statistician and three graduate students was established. Based on the theory of knowledge, belief and action (Li and Liu ) and referring to domestic and foreign research and related guidelines (Gorski et al. ; Chinese Nursing Association intravenous Infusion Therapy Professional Committee ; Estrada‐Orozco et al. ; National Health and Family Planning Commission ) as well as expert group discussions, the initial questionnaire compiled included 14 items in the knowledge dimension, 8 items in the attitude dimension, and 26 items in the behaviour dimension. Using the Delphi method, two rounds of expert correspondence were conducted from July to September 2020. According to the concentration degree of expert opinions, coordination degree of expert opinions and other indicators, combined with expert suggestions, the initial items of the questionnaire were screened and modified to form a pre‐test version of the questionnaire. The questionnaire contains 55 items in total, including 16 items in the knowledge dimension, 9 items in the attitude dimension and 30 items in the behaviour dimension. Five points are scored for all correct answers to knowledge items, 1 point is scored for all correct answers to multiple choice items, and 0 points is scored for all wrong answers. All attitude and behaviour items are scored on the Likert five‐point scoring scale (Likert ): strongly agree, agree, not necessarily, disagree and strongly disagree five kinds of answers, which are recorded as 5, 4, 3, 2 and 1 points respectively.
Pre‐Survey The pre‐test version of questionnaire was used to conduct a questionnaire survey on 20 nurses who met the inclusion criteria. The questionnaire was collected on site with an effective recovery rate of 100%. After the preliminary survey, no nurses put forward new demands, and the questionnaire content was not modified. So the initial version of the questionnaire consisted of three dimensions and 55 items for clinical testing.
Formal Investigation An on‐site questionnaire survey was conducted on 360 nurses who met the inclusion criteria using the self‐prepared general data questionnaire of nurses and the CVC maintenance questionnaire of ICU nurses (clinical test version), and 334 valid questionnaires were collected.
Item Analysis Items are screened based on data from formal surveys. The following two methods are mainly adopted in this study: (1) critical ratio (CR): also known as extreme value method, that is, the scores of all subjects are sorted from the largest to the smallest. If the score is in the top 27%, that is, the sample of the high group, while the score is in the bottom 27%, that is, the sample of the low group. Through the t ‐test analysis of two independent samples, if the CR value of the item is > 3 or the difference is statistically significant ( p < 0.05), the item will be retained; otherwise, it will be deleted (Xiaoyong ). (2) Total correlation analysis: The correlation coefficient between each item and the total score was calculated. If the correlation coefficient was not significant ( p > 0.05) or r < 0.4, the item was deleted (Xiaoyong ). Based on the project analysis results, a questionnaire on the maintenance of CVC for ICU nurses was formed (preliminary version).
Reliability and Validity Analysis The content validity of this study was obtained by letter consultation with experts. The validity of questionnaire structure was tested by exploratory factor analysis. The reliability is tested by internal consistency reliability and retest reliability.
Quality Control After unified training, the quality control investigators introduced the research purpose and precautions to the nurses with unified guidance. After completing the questionnaire, they checked the filling status on the spot and corrected and supplemented the errors and omissions in time.
Ethical Principles This study follows the principles of voluntariness and confidentiality to avoid any harm to the participants. The study was reviewed and approved by the Ethics Committee of REDACTED (Approval number: REDACTED).
Statistical Methods Excel 2013 was used for data entry and sorting. SPSS 26.0 and Amos24.0 were used for statistical analysis of the data. The measurement data were described by mean ± standard deviation, and the counting data were described by frequency and percentage. The difference was statistically significant when p < 0.05. The correlation between items and total score and the critical ratio method were used to screen and analyse the questionnaire items (Xiaoyong ). Factor analysis was used to test the validity of the questionnaire. The validity analysis included content validity and structure validity. Expert evaluation method was used for Content Validity. Item of Content Validity Index (I‐CVI): For each item, the number of experts with a rating of 3 or 4 divided by the total number of experts participating in the evaluation is the I‐CVI. Scale of Content Validity Index (S‐CVI): The number of times of 3 or 4 ratings divided by the number of evaluations. Exploratory factor analysis and confirmatory factor analysis were used for structural validity. Cronbach α coefficient and broken half reliability were used to describe the validity of the questionnaire. The validity of item level content and item level content were calculated by using expert evaluation method. Exploratory factor analysis was used to test the validity of the structure. If KMO ≥ 0.8, p < 0.05 indicates that exploratory factor analysis is suitable. Principal component analysis is used to extract common factors with eigenvalues > 1, and maximum variance orthogonal spin method is used to delete items with factor loads < 0.4 (Zhang and Dong ). Confirmatory factor analysis was used to test the fit degree of each dimension and item of the questionnaire. Cronbach's α and broken half reliability were used to evaluate the reliability of the questionnaire (Wu ).
Results 3.1 Expert Correspondence Among the 19 experts who completed 2 rounds of Delphi expert correspondence, 4 were male and 15 were female; Age < 40 years old 5 people, 40 ~ 50 years old 11 people, > 50 years old 3 people; 10 students have bachelor degree, 5 master degree and 4 doctor degree. Professional titles are intermediate 6 people, associate senior 10 people, senior 3 people; The working years are 10 ~ 20 years 8 people, 21 ~ 30 years 9 people, 30 years or more 2 people; Specialist expertise includes critical care (7 people), intravenous care (4 people), nursing management (4 people), nursing research (2 people), clinical medicine (1 person), psychology (1 person). With the average score of importance < 3.50 and coefficient of variation > 0.25 as the deletion criteria (Xiaoyong ), entries were added, deleted and modified based on expert opinions. The recovery rates of the two rounds of expert correspondence questionnaires were 95.0% (19/20) and 100% (19/19), and the positive coefficient of experts were 95.00% and 100.00%, respectively. The Cr of the two rounds of expert consultation is 0.83 ~ 0.98, which can be considered that the experts who participated in this letter have high authority. After the first round of consultation, 12 experts made constructive written suggestions, and after the second round, 4 experts made written suggestions. The revised questionnaire developed after the first round of consultation modified the expression of 6 items, adding 2 items of knowledge dimension, 2 items of attitude dimension and 4 items of behaviour dimension. After the second round of consultation, one item in the attitude dimension was deleted, and the expression ways of the two items were further modified. The consensus questionnaire formed after the letter consultation contained 55 items. The Kendall coordination coefficient of the two rounds was statistically significant ( χ 2 test, p < 0.001, Table ), indicating a good degree of coordination of expert opinions, so the conclusion is credible. 3.2 Project Analysis 3.2.1 Correlation Coefficient Method Pearson correlation coefficient method was used to analyse the correlation between each item and the total score of the questionnaire. Item 2 item 3, 6, 8, 9, 11, 12, 13, 14, 15 in addition to knowledge, attitude, behaviour items 1, 2, 8, 11, 16, 18, 19, 20, 22, 24, 25, 27, 28, 29 ( p > 0.05), the rest of the entries correlation coefficient of 0.40 or higher. 3.2.2 The Critical Ratio Method Conducted independent sample t test to analyse the critical difference between the items in the high and low groups. The decision values of all items in the questionnaire ranged from 3.241 to 11.582 with statistical significance ( p < 0.001), and the results showed that no items were deleted, as shown in Table . 3.3 Reliability and Validity Test 3.3.1 Validity Analysis 3.3.1.1 Content Validity The results of evaluation by nine experts showed that CVI (I‐CVI) = 0.889 ~ 1.000, CVI(S‐CVI) = 0.974, and the content validity was good. 3.3.1.2 Structural Validity Based on the formal investigation data, factor analysis was performed, and the KMO value was 0.913, and Bartlett sphericity test reached a significant level (c2 = 5886.897, p < 0.001), which was suitable for exploratory factor analysis. In this paper, 31 items were analysed by principal component analysis and variance maximum rotation method. There were three common factors (F1–F3) with feature root λ > 1, and the corresponding total interpretation rate of variation was 65.656%, indicating that the questionnaire had good structural validity. The factor load and commonality of each item of the questionnaire are shown in Table . The fit index of the two‐factor structural equation model is shown in Table . The standardised path analysis is shown in Figure . 3.3.2 Reliability Analysis The Cronbach's α coefficient of the questionnaire was 0.843, and the Cronbach's α coefficient of each dimension was 0.754, 0.887, 0.940. The partial half reliability was 0.816, and the retest reliability was 0.813. It shows that the questionnaire has reliability and stability.
Expert Correspondence Among the 19 experts who completed 2 rounds of Delphi expert correspondence, 4 were male and 15 were female; Age < 40 years old 5 people, 40 ~ 50 years old 11 people, > 50 years old 3 people; 10 students have bachelor degree, 5 master degree and 4 doctor degree. Professional titles are intermediate 6 people, associate senior 10 people, senior 3 people; The working years are 10 ~ 20 years 8 people, 21 ~ 30 years 9 people, 30 years or more 2 people; Specialist expertise includes critical care (7 people), intravenous care (4 people), nursing management (4 people), nursing research (2 people), clinical medicine (1 person), psychology (1 person). With the average score of importance < 3.50 and coefficient of variation > 0.25 as the deletion criteria (Xiaoyong ), entries were added, deleted and modified based on expert opinions. The recovery rates of the two rounds of expert correspondence questionnaires were 95.0% (19/20) and 100% (19/19), and the positive coefficient of experts were 95.00% and 100.00%, respectively. The Cr of the two rounds of expert consultation is 0.83 ~ 0.98, which can be considered that the experts who participated in this letter have high authority. After the first round of consultation, 12 experts made constructive written suggestions, and after the second round, 4 experts made written suggestions. The revised questionnaire developed after the first round of consultation modified the expression of 6 items, adding 2 items of knowledge dimension, 2 items of attitude dimension and 4 items of behaviour dimension. After the second round of consultation, one item in the attitude dimension was deleted, and the expression ways of the two items were further modified. The consensus questionnaire formed after the letter consultation contained 55 items. The Kendall coordination coefficient of the two rounds was statistically significant ( χ 2 test, p < 0.001, Table ), indicating a good degree of coordination of expert opinions, so the conclusion is credible.
Project Analysis 3.2.1 Correlation Coefficient Method Pearson correlation coefficient method was used to analyse the correlation between each item and the total score of the questionnaire. Item 2 item 3, 6, 8, 9, 11, 12, 13, 14, 15 in addition to knowledge, attitude, behaviour items 1, 2, 8, 11, 16, 18, 19, 20, 22, 24, 25, 27, 28, 29 ( p > 0.05), the rest of the entries correlation coefficient of 0.40 or higher. 3.2.2 The Critical Ratio Method Conducted independent sample t test to analyse the critical difference between the items in the high and low groups. The decision values of all items in the questionnaire ranged from 3.241 to 11.582 with statistical significance ( p < 0.001), and the results showed that no items were deleted, as shown in Table .
Correlation Coefficient Method Pearson correlation coefficient method was used to analyse the correlation between each item and the total score of the questionnaire. Item 2 item 3, 6, 8, 9, 11, 12, 13, 14, 15 in addition to knowledge, attitude, behaviour items 1, 2, 8, 11, 16, 18, 19, 20, 22, 24, 25, 27, 28, 29 ( p > 0.05), the rest of the entries correlation coefficient of 0.40 or higher.
The Critical Ratio Method Conducted independent sample t test to analyse the critical difference between the items in the high and low groups. The decision values of all items in the questionnaire ranged from 3.241 to 11.582 with statistical significance ( p < 0.001), and the results showed that no items were deleted, as shown in Table .
Reliability and Validity Test 3.3.1 Validity Analysis 3.3.1.1 Content Validity The results of evaluation by nine experts showed that CVI (I‐CVI) = 0.889 ~ 1.000, CVI(S‐CVI) = 0.974, and the content validity was good. 3.3.1.2 Structural Validity Based on the formal investigation data, factor analysis was performed, and the KMO value was 0.913, and Bartlett sphericity test reached a significant level (c2 = 5886.897, p < 0.001), which was suitable for exploratory factor analysis. In this paper, 31 items were analysed by principal component analysis and variance maximum rotation method. There were three common factors (F1–F3) with feature root λ > 1, and the corresponding total interpretation rate of variation was 65.656%, indicating that the questionnaire had good structural validity. The factor load and commonality of each item of the questionnaire are shown in Table . The fit index of the two‐factor structural equation model is shown in Table . The standardised path analysis is shown in Figure . 3.3.2 Reliability Analysis The Cronbach's α coefficient of the questionnaire was 0.843, and the Cronbach's α coefficient of each dimension was 0.754, 0.887, 0.940. The partial half reliability was 0.816, and the retest reliability was 0.813. It shows that the questionnaire has reliability and stability.
Validity Analysis 3.3.1.1 Content Validity The results of evaluation by nine experts showed that CVI (I‐CVI) = 0.889 ~ 1.000, CVI(S‐CVI) = 0.974, and the content validity was good. 3.3.1.2 Structural Validity Based on the formal investigation data, factor analysis was performed, and the KMO value was 0.913, and Bartlett sphericity test reached a significant level (c2 = 5886.897, p < 0.001), which was suitable for exploratory factor analysis. In this paper, 31 items were analysed by principal component analysis and variance maximum rotation method. There were three common factors (F1–F3) with feature root λ > 1, and the corresponding total interpretation rate of variation was 65.656%, indicating that the questionnaire had good structural validity. The factor load and commonality of each item of the questionnaire are shown in Table . The fit index of the two‐factor structural equation model is shown in Table . The standardised path analysis is shown in Figure .
Content Validity The results of evaluation by nine experts showed that CVI (I‐CVI) = 0.889 ~ 1.000, CVI(S‐CVI) = 0.974, and the content validity was good.
Structural Validity Based on the formal investigation data, factor analysis was performed, and the KMO value was 0.913, and Bartlett sphericity test reached a significant level (c2 = 5886.897, p < 0.001), which was suitable for exploratory factor analysis. In this paper, 31 items were analysed by principal component analysis and variance maximum rotation method. There were three common factors (F1–F3) with feature root λ > 1, and the corresponding total interpretation rate of variation was 65.656%, indicating that the questionnaire had good structural validity. The factor load and commonality of each item of the questionnaire are shown in Table . The fit index of the two‐factor structural equation model is shown in Table . The standardised path analysis is shown in Figure .
Reliability Analysis The Cronbach's α coefficient of the questionnaire was 0.843, and the Cronbach's α coefficient of each dimension was 0.754, 0.887, 0.940. The partial half reliability was 0.816, and the retest reliability was 0.813. It shows that the questionnaire has reliability and stability.
Discussion 4.1 Significance of Questionnaire Preparation The standard catheter maintenance performed by nurses can reduce complications, extend the service life of the catheter and reduce the economic burden of patients, which is of great significance in the whole course of intravenous therapy. In order to improve the quality of CVC maintenance, it is necessary to have a scientific and quantitative measurement tool to evaluate the knowledge and practice of ICU nurses on CVC maintenance, so as to formulate targeted interventions to improve the quality of CVC maintenance and reduce the occurrence of CVC complications. At present, there are no evaluation tools to evaluate the knowledge and practice of ICU nurses' central venous pipeline maintenance at home and abroad. The existing universal knowledge and practice assessment tools and the standard questionnaire for intravenous therapy involve a wide range of contents (Yao ; Chen et al. ), which is difficult to reflect the specific knowledge, attitude and behaviour of CVC maintenance. Specific assessment tools such as peripherally inserted central catheter and the questionnaire providing catheter maintenance knowledge and practice are highly targeted (Zhang et al. ; Ren et al. ) and are not applicable to the assessment of catheter maintenance knowledge and practice. In view of this, it is of practical significance to construct a questionnaire on the knowledge and practice of ICU nurses' central venous pipeline maintenance and provide a targeted quantitative evaluation tool for clinical practice. 4.2 The Questionnaire Is Applicable In this study, knowledge, belief and practice were selected as the theoretical framework to construct the dimension of the questionnaire, and the questionnaire items were constructed in combination with relevant literature and expert consensus on Clinical venous catheter maintenance and Operation (Chinese Nursing Association intravenous Infusion Therapy Professional Committee ). After group discussion, the item pool was established by brainstorming method. Delphi expert correspondence method was used to select items and form the initial questionnaire. In this study, medical experts and psychological experts from intravenous nursing, critical care, clinical medicine and nursing management were selected, and the initial entries were modified and screened through two rounds of Delphi expert correspondence, and professional opinions on the questionnaire were put forward from multiple angles and directions. The data were collected by investigating ICU nurses, and the reliability of the questionnaire was analysed by various statistical methods. The higher the score of the questionnaire, the higher the knowledge of CVC maintenance, the better the attitude and the more normative the behaviour. This questionnaire met nine aspects of CVC maintenance, including evaluation, tube flushing, tube sealing, dressing replacement, catheter fixation, infusion joint, catheter removal, and infection prevention and quantitatively assessed the knowledge, belief and practice of ICU nurses on CVC maintenance. The language of the questionnaire is easy to understand, and it usually takes less than 10 min to fill in, which is easy to promote and apply. 4.3 The Questionnaire Has Good Reliability and Validity The preparation process strictly follows the standardisation of questionnaire preparation. The questionnaire in this study has good stability. Principal component analysis and variance maximum rotation method showed that there were three common factors with eigenroot λ > 1, and the total interpretation rate of corresponding variation was 65.656%, indicating that the questionnaire structure was valid. The quantitative questionnaire for the maintenance of CVC in ICU nurses has high reliability, validity and operability, and can be used as a quantitative evaluation tool for the maintenance of CVC in ICU nurses. However, this study only decided to delete items from the perspective of statistical analysis, without considering the integrity of catheter maintenance. In addition, the ICU nurses involved were all from Anhui province, which has a certain region, and the sample size and cross‐provincial multi‐center studies need to be increased in the future, so as to improve the questionnaire.
Significance of Questionnaire Preparation The standard catheter maintenance performed by nurses can reduce complications, extend the service life of the catheter and reduce the economic burden of patients, which is of great significance in the whole course of intravenous therapy. In order to improve the quality of CVC maintenance, it is necessary to have a scientific and quantitative measurement tool to evaluate the knowledge and practice of ICU nurses on CVC maintenance, so as to formulate targeted interventions to improve the quality of CVC maintenance and reduce the occurrence of CVC complications. At present, there are no evaluation tools to evaluate the knowledge and practice of ICU nurses' central venous pipeline maintenance at home and abroad. The existing universal knowledge and practice assessment tools and the standard questionnaire for intravenous therapy involve a wide range of contents (Yao ; Chen et al. ), which is difficult to reflect the specific knowledge, attitude and behaviour of CVC maintenance. Specific assessment tools such as peripherally inserted central catheter and the questionnaire providing catheter maintenance knowledge and practice are highly targeted (Zhang et al. ; Ren et al. ) and are not applicable to the assessment of catheter maintenance knowledge and practice. In view of this, it is of practical significance to construct a questionnaire on the knowledge and practice of ICU nurses' central venous pipeline maintenance and provide a targeted quantitative evaluation tool for clinical practice.
The Questionnaire Is Applicable In this study, knowledge, belief and practice were selected as the theoretical framework to construct the dimension of the questionnaire, and the questionnaire items were constructed in combination with relevant literature and expert consensus on Clinical venous catheter maintenance and Operation (Chinese Nursing Association intravenous Infusion Therapy Professional Committee ). After group discussion, the item pool was established by brainstorming method. Delphi expert correspondence method was used to select items and form the initial questionnaire. In this study, medical experts and psychological experts from intravenous nursing, critical care, clinical medicine and nursing management were selected, and the initial entries were modified and screened through two rounds of Delphi expert correspondence, and professional opinions on the questionnaire were put forward from multiple angles and directions. The data were collected by investigating ICU nurses, and the reliability of the questionnaire was analysed by various statistical methods. The higher the score of the questionnaire, the higher the knowledge of CVC maintenance, the better the attitude and the more normative the behaviour. This questionnaire met nine aspects of CVC maintenance, including evaluation, tube flushing, tube sealing, dressing replacement, catheter fixation, infusion joint, catheter removal, and infection prevention and quantitatively assessed the knowledge, belief and practice of ICU nurses on CVC maintenance. The language of the questionnaire is easy to understand, and it usually takes less than 10 min to fill in, which is easy to promote and apply.
The Questionnaire Has Good Reliability and Validity The preparation process strictly follows the standardisation of questionnaire preparation. The questionnaire in this study has good stability. Principal component analysis and variance maximum rotation method showed that there were three common factors with eigenroot λ > 1, and the total interpretation rate of corresponding variation was 65.656%, indicating that the questionnaire structure was valid. The quantitative questionnaire for the maintenance of CVC in ICU nurses has high reliability, validity and operability, and can be used as a quantitative evaluation tool for the maintenance of CVC in ICU nurses. However, this study only decided to delete items from the perspective of statistical analysis, without considering the integrity of catheter maintenance. In addition, the ICU nurses involved were all from Anhui province, which has a certain region, and the sample size and cross‐provincial multi‐center studies need to be increased in the future, so as to improve the questionnaire.
This paper was approved by the Ethics Committee of the First Affiliated Hospital of Anhui Medical University (Approval number: Kuai‐Lun Review of the First Affiliated Hospital of An Medical University‐P2020‐17‐12), and we were in accordance with the 1975 Helsinki declaration and its later amendments.
The authors have nothing to report.
The authors declare no conflicts of interest.
|
Guidelines of the Polish Society of Pediatric Endocrinology and Diabetology and Pediatric Section of Diabetes Poland on insulin therapy using hybrid closed-loop systems in children and adolescents with diabetes in Poland | 6f458a88-f29c-403c-82fa-55b01ed0fa15 | 11538916 | Internal Medicine[mh] | Type 1 diabetes (T1D) is characterised by an absolute insulin deficiency. Unfortunately, despite the use of modern insulin therapies, insulin pumps, and continuous glucose monitoring (CGM) systems, a significant number of individuals living with T1D still fail to achieve their therapeutic goals. Moreover, T1D continues to drastically reduce life expectancy—by nearly 12 years in Poland compared to those without diabetes . This ongoing challenge highlights the need for new methods that enable optimal metabolic control while also ensuring a good quality of life for people with diabetes. The broader adoption of new diabetes technologies, including hybrid closed loop (HCL) systems, by facilitating better glycaemic control, offers hope for reclaiming the years of life lost due to diabetes. Automated insulin delivery (AID) systems use real-time CGM (rtCGM) data integrated into a control algorithm that automatically adjusts insulin delivery rates via Continuous Subcutaneous Insulin Infusion (CSII, i.e. an insulin pump). Currently available HCL systems can automatically reduce or suspend basal insulin infusion when glucose levels drop, and increase insulin delivery when glucose levels rise, aiming to maintain a preset target glucose level . AID systems significantly improve metabolic control, both in terms of reducing HbA1c and increasing time spent within the recommended glycaemic range, as well as in reducing hypoglycaemia. Reports indicate that the average time in the range 70–180 mg/dl (time in range – TIR) for individuals using various HCL systems is around 76%, compared to 45–60% for those using other forms of therapy combined with continuous glucose monitoring . Additionally, improvements have been demonstrated in time in the tight range of 70–140 mg/dl (time in tight range – TITR), especially during nighttime hours . The therapeutic efficacy of AID systems, understood as achieving time within the recommended range, has been demonstrated across a broad population of individuals with T1D in all age groups, both in randomised controlled trials and in real-world evidence studies. These results have proven stable and sustained over the years . The use of HCL has also not been associated with an increase in users’ body weight . Achieving glycaemic targets still requires the user to program mealtime boluses; however, even an inaccurate estimation of the carbohydrate content in a meal allows TIR to be maintained at the recommended level . A particularly important aspect of using HCL systems is the significant reduction of therapeutic decisions made by a person with diabetes. The algorithm handles them on its own leading to a significant improvement in a life quality. It also reduces the burden of living with diabetes. A decrease in anxiety related to hypoglycaemia is also highlighted. Improving nighttime glycaemia leads to better sleep quality, not only for the person with diabetes but also for their family . The therapeutic effectiveness of HCL systems is also evaluated through health economics analyses. Numerous studies have demonstrated that these systems are cost-effective interventions for individuals with T1D, particularly compared to a range of other non-HCL methods. A recently published study confirmed that HCL systems are cost-effective in every scenario despite higher incremental costs . This is because HCL systems generate sufficient gains in quality-adjusted life years (QALY) to outweigh the additional costs, bringing them below the willingness-to-pay thresholds of healthcare payers in various countries. In the conducted analyses, the QALY benefits were primarily attributed to reducing the incidence of diabetes-related complications (such as diabetic ketoacidosis, severe hypoglycaemic episodes, and micro- and macrovascular complications) . Additionally, in some cases, the reduction in anxiety related to hypoglycaemia – not only for the person with diabetes but also for their caregivers – contributed to these benefits. In this context, the impact of HCL systems on the quality of life of caregivers – who are integral partners in therapy – particularly in alleviating the mental stress and burdens associated with caregiving, is invaluable. Compelling scientific evidence has led several scientific societies worldwide, including Diabetes Poland (PTD), to develop recommendations that clearly identify HCL systems as the therapy of choice for all individuals with T1D, particularly those with suboptimal glycaemic control . For many patients who have been unable to achieve acceptable glycaemic control over the years due to various reasons, the use of HCL systems may be the only way to improve metabolic control of diabetes and reduce the risk of complications. In light of this, we present the recommendations of the Polish Society of Paediatric Endocrinology and Diabetology (PTEiDD) and the Paediatric Section of Diabetes Poland (PTD) regarding the use of closed loop systems in children and adolescents, focusing on HCL systems commercially available in Poland (as of August 2024). 1.1. Available systems The number of commercially available AID systems worldwide is growing. These systems operate on the principle of HCL, meaning that all of them still require the user to administer mealtime boluses . The following systems are used: MiniMed™ 780G, CamAPS® FX with mylife™ YpsoPump®, Tandem t:slim™ X2 Control IQ™, and systems utilising tubeless pumps (so-called “patch” pumps: Omnipod® 5, Tandem Mobi®, DBLG1™ System Diabeloop, Medtrum TouchCare® Nano System) . In Poland, two commercial HCL systems are available: the MiniMed™ 780G system from Medtronic and the CamAPS® FX, which connects the mylife™ YpsoPump® with the Dexcom G6® CGM system (Dexcom G7® is currently planned). These systems are discussed in detail in . Other configurations of these systems may be available in different countries . In HCL systems, instead of the typical distinction between basal insulin (the basic insulin dose) and boluses, there is rather a distinction between insulin delivery initiated by the algorithm and insulin delivery initiated by the user. Other examples of commercial HCL systems available in the European Union are as follows: the Tandem Control IQ™ (based on the Tandem™ t:slim™ X2 pump and Dexcom G6® or G7® CGM or FreeStyle Libre® 2 Plus) is registered for individuals aged 6 years and older. The target glucose value of the algorithm can be set every 30 minutes within a range of 110 to 160 mg/dl (6.1 to 8.9 mmol/l); the system has additional options such as a sleep mode (which increases the aggressiveness of the algorithm) and 6 exercise modes; SmartAdjust (based on the tubeless Omnipod® 5 pump and Dexcom G6® CGM), which is registered for individuals aged 2 years and older; the target glucose value can be set in 60-minute intervals (with values of 110, 120, 130, 140, 150 mg/dl [6.1, 6.7, 7.2, 7.8, or 8.3 mmol/l]). The user can use the SmartBolus calculator, which utilises CGM data and trends in its calculations. Some users also utilise “Artificial Pancreas System – Do it Yourself” (APS-DIY) systems, which they create themselves based on commercial devices and algorithms often made available from the Internet for free (open source). Since these algorithms do not have registration, their use remains the responsibility of the person with diabetes and/or their caregivers. Regardless of the HCL system being used, including APS-DIY, all individuals should receive the same support from the diabetes care team in managing their diabetes therapy. 1.2. Automatic mode setting Each HCL system, in addition to the mandatory manual mode settings (including basal infusion and the bolus calculator, see ), requires the input of parameters necessary for functioning in automatic mode . In the MiniMed™ 780G system, before starting operation in automatic mode, in addition to setting the target blood glucose level, it is usually recommended to adjust the duration of the active insulin time . For the CamAPS® FX system, the initial activation of automatic mode requires the patient’s data to be entered: body weight and daily insulin dose. Additionally, the algorithm uses the entered insulin-to-carbohydrate ratios. If the user wishes to administer an additional bolus manually, they can use the previously entered correction factor and target value in the bolus calculator. In the pump settings, 2 identical basal rates must always be entered (in case the automatic mode is turned off). Access to the CamAPS® FX application is granted after completing training developed by the application creators, confirmed by a physician . The number of commercially available AID systems worldwide is growing. These systems operate on the principle of HCL, meaning that all of them still require the user to administer mealtime boluses . The following systems are used: MiniMed™ 780G, CamAPS® FX with mylife™ YpsoPump®, Tandem t:slim™ X2 Control IQ™, and systems utilising tubeless pumps (so-called “patch” pumps: Omnipod® 5, Tandem Mobi®, DBLG1™ System Diabeloop, Medtrum TouchCare® Nano System) . In Poland, two commercial HCL systems are available: the MiniMed™ 780G system from Medtronic and the CamAPS® FX, which connects the mylife™ YpsoPump® with the Dexcom G6® CGM system (Dexcom G7® is currently planned). These systems are discussed in detail in . Other configurations of these systems may be available in different countries . In HCL systems, instead of the typical distinction between basal insulin (the basic insulin dose) and boluses, there is rather a distinction between insulin delivery initiated by the algorithm and insulin delivery initiated by the user. Other examples of commercial HCL systems available in the European Union are as follows: the Tandem Control IQ™ (based on the Tandem™ t:slim™ X2 pump and Dexcom G6® or G7® CGM or FreeStyle Libre® 2 Plus) is registered for individuals aged 6 years and older. The target glucose value of the algorithm can be set every 30 minutes within a range of 110 to 160 mg/dl (6.1 to 8.9 mmol/l); the system has additional options such as a sleep mode (which increases the aggressiveness of the algorithm) and 6 exercise modes; SmartAdjust (based on the tubeless Omnipod® 5 pump and Dexcom G6® CGM), which is registered for individuals aged 2 years and older; the target glucose value can be set in 60-minute intervals (with values of 110, 120, 130, 140, 150 mg/dl [6.1, 6.7, 7.2, 7.8, or 8.3 mmol/l]). The user can use the SmartBolus calculator, which utilises CGM data and trends in its calculations. Some users also utilise “Artificial Pancreas System – Do it Yourself” (APS-DIY) systems, which they create themselves based on commercial devices and algorithms often made available from the Internet for free (open source). Since these algorithms do not have registration, their use remains the responsibility of the person with diabetes and/or their caregivers. Regardless of the HCL system being used, including APS-DIY, all individuals should receive the same support from the diabetes care team in managing their diabetes therapy. Each HCL system, in addition to the mandatory manual mode settings (including basal infusion and the bolus calculator, see ), requires the input of parameters necessary for functioning in automatic mode . In the MiniMed™ 780G system, before starting operation in automatic mode, in addition to setting the target blood glucose level, it is usually recommended to adjust the duration of the active insulin time . For the CamAPS® FX system, the initial activation of automatic mode requires the patient’s data to be entered: body weight and daily insulin dose. Additionally, the algorithm uses the entered insulin-to-carbohydrate ratios. If the user wishes to administer an additional bolus manually, they can use the previously entered correction factor and target value in the bolus calculator. In the pump settings, 2 identical basal rates must always be entered (in case the automatic mode is turned off). Access to the CamAPS® FX application is granted after completing training developed by the application creators, confirmed by a physician . 2.1. Indications HCL systems are recommended for all children and adolescents with diabetes requiring intensive insulin therapy, provided that the following conditions are met: they accept the principles of this treatment method; they have no contraindications for its use; the technical requirements for using the specific HCL system are met. Every person with T1D has the right to use the HCL system from diagnosis. The earliest possible initiation of HCL therapy is justified and recommended for every child with T1D. 2.2. Contraindications Contraindications for HCL therapy are consistent with the contraindications for therapy using personal insulin pumps according to PTD : Certain psychological disorders in the person with diabetes and/or parents/caregivers that, in the opinion of the attending physician, prevent the safe use of a personal insulin pump. Certain eating disorders that, in the opinion of the attending physician, prevent the safe use of a personal insulin pump. Intellectual disabilities in parents/caregivers of children under 13 years old that prevent understanding the principles of intensive insulin therapy and pump operation or dependence on alcohol and psychoactive substances. For children aged 13-16 years, the decision to use HCL should be made individually based on the child’s ability to operate the insulin pump and manage therapy under increased supervision from the diabetes therapeutic team and a family assistant. Unjustified absences from medical appointments (attending only one appointment per year or missing appointments) at the diabetes clinic. Non-compliance or misunderstanding of the principles of intensive basal-bolus insulin therapy (lack of proper self-monitoring of blood glucose, lack of ketone body monitoring in situations of prolonged hyperglycaemia). More than one episode of severe ketoacidosis within a year during the use of HCL. Non-compliance with personal hygiene standards. 2.3. Use of hybrid closed loop systems beyond registered age According to the PTEiDD’s position ( https://pteidd.pl/index_subpage.php/?n=3 ), in the case of children younger than the age for which a given system is registered, the attending physician makes the decision to initiate its use individually. This decision is based on the current available knowledge and in accordance with the principles of evidence-based medicine, after obtaining the consent of the parents/caregivers. 2.4. Temporary discontinuation of hybrid closed loop therapy Problems with using sensors or infusion sets (e.g. severe skin reactions) – requiring the cessation of using the pump or CGM system that is part of the HCL system. Malfunction of HCL system components (insulin pump, CGM system, smartphone in the case of the CamAPS® FX application). Period of complete remission not requiring insulin administration or partial remission requiring only basal insulin administration. At the patient’s request (e.g. during holidays, sports competitions). Period of additional illness during which insulin requirements change rapidly, and the automatic mode modifies insulin delivery too slowly. HCL systems are recommended for all children and adolescents with diabetes requiring intensive insulin therapy, provided that the following conditions are met: they accept the principles of this treatment method; they have no contraindications for its use; the technical requirements for using the specific HCL system are met. Every person with T1D has the right to use the HCL system from diagnosis. The earliest possible initiation of HCL therapy is justified and recommended for every child with T1D. Contraindications for HCL therapy are consistent with the contraindications for therapy using personal insulin pumps according to PTD : Certain psychological disorders in the person with diabetes and/or parents/caregivers that, in the opinion of the attending physician, prevent the safe use of a personal insulin pump. Certain eating disorders that, in the opinion of the attending physician, prevent the safe use of a personal insulin pump. Intellectual disabilities in parents/caregivers of children under 13 years old that prevent understanding the principles of intensive insulin therapy and pump operation or dependence on alcohol and psychoactive substances. For children aged 13-16 years, the decision to use HCL should be made individually based on the child’s ability to operate the insulin pump and manage therapy under increased supervision from the diabetes therapeutic team and a family assistant. Unjustified absences from medical appointments (attending only one appointment per year or missing appointments) at the diabetes clinic. Non-compliance or misunderstanding of the principles of intensive basal-bolus insulin therapy (lack of proper self-monitoring of blood glucose, lack of ketone body monitoring in situations of prolonged hyperglycaemia). More than one episode of severe ketoacidosis within a year during the use of HCL. Non-compliance with personal hygiene standards. According to the PTEiDD’s position ( https://pteidd.pl/index_subpage.php/?n=3 ), in the case of children younger than the age for which a given system is registered, the attending physician makes the decision to initiate its use individually. This decision is based on the current available knowledge and in accordance with the principles of evidence-based medicine, after obtaining the consent of the parents/caregivers. Problems with using sensors or infusion sets (e.g. severe skin reactions) – requiring the cessation of using the pump or CGM system that is part of the HCL system. Malfunction of HCL system components (insulin pump, CGM system, smartphone in the case of the CamAPS® FX application). Period of complete remission not requiring insulin administration or partial remission requiring only basal insulin administration. At the patient’s request (e.g. during holidays, sports competitions). Period of additional illness during which insulin requirements change rapidly, and the automatic mode modifies insulin delivery too slowly. The scope of training depends on the duration of diabetes (e.g. newly diagnosed, reintroduction of therapy after a period of complete remission), the type of insulin therapy previously used, the type of insulin pump, and the glucose monitoring method . Complete training on the use of the HCL system includes the following points: general information on available systems and maintenance costs, including the principles of reimbursement in Poland, with personalised selection of HCL by the person with diabetes and their caregivers, as well as members of the therapeutic team; basics of insulin therapy in a basal-bolus system; nutritional principles in diabetes, including recognising and evaluating food quality and calculating the number of carbohydrates in a meal, at least at a basic level; technical operation of HCL system components: the insulin pump and the CGM system; familiarity with all system functions: manual and automatic modes, “basic” and “automatic mode” bolus calculators, evaluation and interpretation of active insulin, automatic and manual corrections, using applications that enable connection with the treating centre, therapy partners, smartwatch, contact with the helpline; management during physical activity, hypoglycaemia, hyperglycaemia, ketonaemia/ketonuria, acute infection, gastroenteritis, and system malfunction; analysis and interpretation of reports in the patient’s application/computer application; modification of system settings. Training should conclude with a positive result on a knowledge and skills test confirmed by a member of the therapeutic team. For individuals with diabetes who have daily insulin requirements below the threshold for activating the automatic insulin delivery function (e.g. small children, partial remission period), dilution of insulin may be considered. Ideally, the dilution should be performed using a diluent specifically designed for insulin (which is difficult to get in Poland) or with a 0.9% NaCl solution. The degree of insulin dilution depends on the individual’s daily insulin dose. Parents/caregivers must undergo appropriate training in insulin dilution techniques. Additionally, it is necessary to place information about the degree of insulin dilution on the pump . The principles for initiating therapy typically vary depending on the specific HCL system used. However, all HCL systems require the setting of essential parameters for manual mode operation, including basal infusion and bolus calculations . The timing for activating automatic mode depends on the clinical situation of individual’s with diabetes. For most individuals who have previously used insulin pumps and regularly utilised CGM systems, the automatic mode can be activated as soon as the system requirements are met. For children beginning therapy with a personal insulin pump (e.g. those newly diagnosed with diabetes or previously treated with multiple daily injections, MDI), it is advisable to remain in manual mode for an extended period. This duration should be tailored to the person to allow them to develop the necessary skills to manage therapy in manual mode (i.e. in the case of the MiniMed™ 780G, using the predictive low-glucose suspend [PLGS] algorithm). For individuals with highly variable daily insulin requirements using the MiniMed™ 780G system, the period of manual mode use may also need to be extended. 4.1. MiniMed™ 780G System Before initiating the automatic SmartGuard™ mode, 2 parameters must be set: target glucose level and insulin activity time. The system also utilises insulin-to-carbohydrate ratio (ICR) settings. Automatic mode can be activated after 48 hours of using manual mode, starting from midnight after the pump is connected. The recommended optimal settings are a target glucose level of 100 mg/dl and an insulin activity time of 2 hours . Evidence from studies and clinical practice suggests that these settings should be used from the start of therapy in most individuals because they allow for the highest percentage of time spent within range and a tight time in range without increasing the risk of hypoglycaemia. More conservative settings (target glucose of 110 or 120 mg/dl, insulin activity time of more than 2 hours) may be advisable at the beginning of therapy, particularly when glucose levels should not be lowered too rapidly. This approach is especially relevant for individuals with hypoglycaemia unawareness, fear of hypoglycaemia, advanced retinopathy, or very young children. The temporary target (set at 150 mg/dl) is used during physical activity or in situations with an increased risk of hypoglycaemia (e.g. gastrointestinal infections with diarrhoea/vomiting). 4.2. CamAPS® FX The first activation of the CamAPS® FX automatic mode requires the individual’s data to be entered, including body weight and average total daily insulin dose. Any significant change in body weight must be updated in the application’s settings. Automatic mode is activated as soon as CGM data are available. The algorithm’s default target glucose value is 104 mg/dl. A higher target glucose value can be set for individuals with hypoglycaemia unawareness, fear of hypoglycaemia, advanced retinopathy, or very young children. Different target values can be set for specific time intervals during the day. The Boost function can be used during an increased insulin requirement (e.g. during infections, menstruation, stress, or increased food intake). The Ease-off function is used when there is a reduced insulin requirement or an increased risk of hypoglycaemia. It is particularly useful during physical activity. Both the Boost and Ease-off modes can be scheduled in advance to start at a specific time. 4.3. Mealtime boluses In traditional pump therapy, the doses for mealtime boluses are calculated using a bolus calculator . The settings for the bolus calculator parameters (except active insulin time) vary depending on the time of day, which is why they are programmed in time intervals. Typically, the ICR is lowest in the morning, i.e. during breakfast/the first meal of the day. In HCL systems, the algorithm calculates the bolus dose based on the entered carbohydrate amount and the user-set ICR and target glucose values in automatic mode. Additionally, it is based on the analysis of data from previous days (particularly considering hypoglycaemic episodes), the current glucose level (SG), the trend in glucose changes, the amount of active insulin, and the insulin sensitivity factor (ISF) calculated by the algorithm. In the MiniMed 780G system, the safe meal bolus feature is designed to maintain SG above 50 mg/dl for the next 4 hours. If SG is expected to drop below 80 mg/dl within 2 hours, the algorithm will reduce the dose by 25% or more (to the highest safe dose). In CSII therapy, the effectiveness of the mealtime dose depends on the accuracy of calculating the number of grams of carbohydrates in the meal and the correct setting of the ICR. Less precise estimation of the amount of carbohydrates based on the size of the meal leads to poorer metabolic control of diabetes. In HCL systems, the ability to automatically administer insulin based on changing glucose levels allows for slightly less rigorous calculation of the number of carbohydrates consumed. For individuals who have difficulty counting carbohydrates, a portion size system (i.e. estimated portions, known as a fixed meal), such as “large meal” – 60 g, “standard meal” – 40 g, “small meal” – 20 g of carbohydrates, can be used. The number of grams of carbohydrates in a portion size system (large-standard-small meal) depends on the child’s age and the average daily calorie intake. It has been shown that more accurate calculations of carbohydrate intake lead to better metabolic control of diabetes . For older individuals, it is suggested not to exceed 60 grams of carbohydrates per meal. Meals containing more carbohydrates may require individualised insulin dose adjustments. Furthermore, it has been demonstrated that in HCL systems, not administering a mealtime bolus for meals containing up to 20 g of carbohydrates does not significantly affect TIR. This can be useful for individuals who cannot administer boluses independently and when their temporary caregivers (e.g. teachers, preschool staff) refuse to administer them. In such cases, small meals without boluses can be planned at school/preschool (for small children, the allowable carbohydrate portions not requiring a bolus may be smaller, proportional to body weight). Not only the amount of carbohydrates but also the protein and fat content modulate the glycaemic response. In HCL systems, protein-fat meals generate an additional insulin dose that is administered automatically. However, high-fat foods can cause a prolonged rise in glucose levels beyond 2 hours after insulin administration, which is not compensated for by automatic insulin delivery. In such situations, to increase the insulin dose for this type of meal: in the MiniMed™ 780G system, you can administer an additional mealtime bolus. To do this, enter approximately 20–30% of the carbohydrates entered before the meal 1.5 hours after the initial mealtime bolus. It is also acceptable to enter them into the system immediately after the meal; in the CamAPS® FX system, you can activate the Boost function at the beginning of the meal or increase the number of grams of carbohydrates by at least 30% and mark the meal as slowly absorbed. 4.3.1. Timing of bolus administration Mealtime boluses should be administered approximately 15 minutes before the meal . If the meal contains high-glycaemic index foods, it may be advisable to administer the bolus even earlier or optionally, to “increase the bolus dose”. The timing of bolus administration before consuming carbohydrates is particularly important because it helps prevent a sudden postprandial glucose spike. Following an initial rise in glucose levels, the system automatically increases the algorithm-modulated insulin dose, significantly increasing the amount of active insulin. The accumulation of insulin from a delayed mealtime bolus, increased basal rate, and auto-corrections may result in hypoglycaemia. If hypoglycaemia occurs before a meal, it is recommended to treat the hypoglycaemia first, and after glucose values stabilise, administer the mealtime bolus and begin the meal. 4.3.2. Missed or delayed mealtime bolus If a meal bolus is skipped or delayed (not administered before the meal), it is suggested that half of the amount of carbohydrates consumed be entered and the suggested bolus be administered within 30-60 minutes of starting the meal. If the meal bolus is delayed by more than 60 minutes after the start of the meal, it is recommended not to administer the meal bolus, because the system will automatically begin correcting blood glucose by increasing the automatic insulin supply. Alternatively, a correction dose initiated by the user can be administered by entering the current glucose level and zero carbohydrates (if blood glucose is within the recommended range, insulin administration is not advised) . 4.4. Correction dose of insulin The MiniMed 780G system automatically calculates and administers correction boluses. Additionally, a user-initiated correction dose can be given by entering the blood glucose value from a glucose meter into the system. The CamAPS® FX system automatically calculates and administers a correction dose of insulin. Additionally, after manually entering glucose levels in the bolus calculator (and zero grams of carbohydrates), the system calculates an additional correction dose. Administering a correction dose by entering fictional (i.e. not really consumed) amounts of carbohydrates („fake carbs”) into the HCL system may affect the system’s ability to respond appropriately to the situation, potentially reducing overall system performance and increasing glycaemic variability. Before initiating the automatic SmartGuard™ mode, 2 parameters must be set: target glucose level and insulin activity time. The system also utilises insulin-to-carbohydrate ratio (ICR) settings. Automatic mode can be activated after 48 hours of using manual mode, starting from midnight after the pump is connected. The recommended optimal settings are a target glucose level of 100 mg/dl and an insulin activity time of 2 hours . Evidence from studies and clinical practice suggests that these settings should be used from the start of therapy in most individuals because they allow for the highest percentage of time spent within range and a tight time in range without increasing the risk of hypoglycaemia. More conservative settings (target glucose of 110 or 120 mg/dl, insulin activity time of more than 2 hours) may be advisable at the beginning of therapy, particularly when glucose levels should not be lowered too rapidly. This approach is especially relevant for individuals with hypoglycaemia unawareness, fear of hypoglycaemia, advanced retinopathy, or very young children. The temporary target (set at 150 mg/dl) is used during physical activity or in situations with an increased risk of hypoglycaemia (e.g. gastrointestinal infections with diarrhoea/vomiting). The first activation of the CamAPS® FX automatic mode requires the individual’s data to be entered, including body weight and average total daily insulin dose. Any significant change in body weight must be updated in the application’s settings. Automatic mode is activated as soon as CGM data are available. The algorithm’s default target glucose value is 104 mg/dl. A higher target glucose value can be set for individuals with hypoglycaemia unawareness, fear of hypoglycaemia, advanced retinopathy, or very young children. Different target values can be set for specific time intervals during the day. The Boost function can be used during an increased insulin requirement (e.g. during infections, menstruation, stress, or increased food intake). The Ease-off function is used when there is a reduced insulin requirement or an increased risk of hypoglycaemia. It is particularly useful during physical activity. Both the Boost and Ease-off modes can be scheduled in advance to start at a specific time. In traditional pump therapy, the doses for mealtime boluses are calculated using a bolus calculator . The settings for the bolus calculator parameters (except active insulin time) vary depending on the time of day, which is why they are programmed in time intervals. Typically, the ICR is lowest in the morning, i.e. during breakfast/the first meal of the day. In HCL systems, the algorithm calculates the bolus dose based on the entered carbohydrate amount and the user-set ICR and target glucose values in automatic mode. Additionally, it is based on the analysis of data from previous days (particularly considering hypoglycaemic episodes), the current glucose level (SG), the trend in glucose changes, the amount of active insulin, and the insulin sensitivity factor (ISF) calculated by the algorithm. In the MiniMed 780G system, the safe meal bolus feature is designed to maintain SG above 50 mg/dl for the next 4 hours. If SG is expected to drop below 80 mg/dl within 2 hours, the algorithm will reduce the dose by 25% or more (to the highest safe dose). In CSII therapy, the effectiveness of the mealtime dose depends on the accuracy of calculating the number of grams of carbohydrates in the meal and the correct setting of the ICR. Less precise estimation of the amount of carbohydrates based on the size of the meal leads to poorer metabolic control of diabetes. In HCL systems, the ability to automatically administer insulin based on changing glucose levels allows for slightly less rigorous calculation of the number of carbohydrates consumed. For individuals who have difficulty counting carbohydrates, a portion size system (i.e. estimated portions, known as a fixed meal), such as “large meal” – 60 g, “standard meal” – 40 g, “small meal” – 20 g of carbohydrates, can be used. The number of grams of carbohydrates in a portion size system (large-standard-small meal) depends on the child’s age and the average daily calorie intake. It has been shown that more accurate calculations of carbohydrate intake lead to better metabolic control of diabetes . For older individuals, it is suggested not to exceed 60 grams of carbohydrates per meal. Meals containing more carbohydrates may require individualised insulin dose adjustments. Furthermore, it has been demonstrated that in HCL systems, not administering a mealtime bolus for meals containing up to 20 g of carbohydrates does not significantly affect TIR. This can be useful for individuals who cannot administer boluses independently and when their temporary caregivers (e.g. teachers, preschool staff) refuse to administer them. In such cases, small meals without boluses can be planned at school/preschool (for small children, the allowable carbohydrate portions not requiring a bolus may be smaller, proportional to body weight). Not only the amount of carbohydrates but also the protein and fat content modulate the glycaemic response. In HCL systems, protein-fat meals generate an additional insulin dose that is administered automatically. However, high-fat foods can cause a prolonged rise in glucose levels beyond 2 hours after insulin administration, which is not compensated for by automatic insulin delivery. In such situations, to increase the insulin dose for this type of meal: in the MiniMed™ 780G system, you can administer an additional mealtime bolus. To do this, enter approximately 20–30% of the carbohydrates entered before the meal 1.5 hours after the initial mealtime bolus. It is also acceptable to enter them into the system immediately after the meal; in the CamAPS® FX system, you can activate the Boost function at the beginning of the meal or increase the number of grams of carbohydrates by at least 30% and mark the meal as slowly absorbed. 4.3.1. Timing of bolus administration Mealtime boluses should be administered approximately 15 minutes before the meal . If the meal contains high-glycaemic index foods, it may be advisable to administer the bolus even earlier or optionally, to “increase the bolus dose”. The timing of bolus administration before consuming carbohydrates is particularly important because it helps prevent a sudden postprandial glucose spike. Following an initial rise in glucose levels, the system automatically increases the algorithm-modulated insulin dose, significantly increasing the amount of active insulin. The accumulation of insulin from a delayed mealtime bolus, increased basal rate, and auto-corrections may result in hypoglycaemia. If hypoglycaemia occurs before a meal, it is recommended to treat the hypoglycaemia first, and after glucose values stabilise, administer the mealtime bolus and begin the meal. 4.3.2. Missed or delayed mealtime bolus If a meal bolus is skipped or delayed (not administered before the meal), it is suggested that half of the amount of carbohydrates consumed be entered and the suggested bolus be administered within 30-60 minutes of starting the meal. If the meal bolus is delayed by more than 60 minutes after the start of the meal, it is recommended not to administer the meal bolus, because the system will automatically begin correcting blood glucose by increasing the automatic insulin supply. Alternatively, a correction dose initiated by the user can be administered by entering the current glucose level and zero carbohydrates (if blood glucose is within the recommended range, insulin administration is not advised) . Mealtime boluses should be administered approximately 15 minutes before the meal . If the meal contains high-glycaemic index foods, it may be advisable to administer the bolus even earlier or optionally, to “increase the bolus dose”. The timing of bolus administration before consuming carbohydrates is particularly important because it helps prevent a sudden postprandial glucose spike. Following an initial rise in glucose levels, the system automatically increases the algorithm-modulated insulin dose, significantly increasing the amount of active insulin. The accumulation of insulin from a delayed mealtime bolus, increased basal rate, and auto-corrections may result in hypoglycaemia. If hypoglycaemia occurs before a meal, it is recommended to treat the hypoglycaemia first, and after glucose values stabilise, administer the mealtime bolus and begin the meal. If a meal bolus is skipped or delayed (not administered before the meal), it is suggested that half of the amount of carbohydrates consumed be entered and the suggested bolus be administered within 30-60 minutes of starting the meal. If the meal bolus is delayed by more than 60 minutes after the start of the meal, it is recommended not to administer the meal bolus, because the system will automatically begin correcting blood glucose by increasing the automatic insulin supply. Alternatively, a correction dose initiated by the user can be administered by entering the current glucose level and zero carbohydrates (if blood glucose is within the recommended range, insulin administration is not advised) . The MiniMed 780G system automatically calculates and administers correction boluses. Additionally, a user-initiated correction dose can be given by entering the blood glucose value from a glucose meter into the system. The CamAPS® FX system automatically calculates and administers a correction dose of insulin. Additionally, after manually entering glucose levels in the bolus calculator (and zero grams of carbohydrates), the system calculates an additional correction dose. Administering a correction dose by entering fictional (i.e. not really consumed) amounts of carbohydrates („fake carbs”) into the HCL system may affect the system’s ability to respond appropriately to the situation, potentially reducing overall system performance and increasing glycaemic variability. In the case of persistent postprandial hyperglycaemia, one should do the following : Ensure that the mealtime bolus is administered at the appropriate time before the meal and that the actual number of grams of carbohydrates consumed is accurately entered. Decrease the ICR. Review the composition of the meal (e.g., an increased proportion of fats and proteins). If the TIR (time in range) is too low, one should: Check the settings of the automatic mode parameters: MiniMed™ 780G: Lower the target glucose and/or shorten the active insulin time, and check if the auto-correction function is enabled. CamAPS® FX: Lower the target glucose and ensure the current body weight is correctly entered. Assess postprandial glucose as mentioned above. If the TBR (Time Below Range) is too high: During nighttime: MiniMed™ 780G: Increase the target glucose value or extend the active insulin time, and sometimes activate the temporary target for a few hours at night. CamAPS® FX: Increase the target glucose. After meals: Review the ICR and assess adherence to mealtime bolus administration guidelines (e.g. boluses given after meals). When using HCL during physical activity, there is an option to activate functions that modify insulin dosing, known as the “sports mode”: In the MiniMed™ 780G system, this function is called the Temporary Target; it raises the target glucose level for the algorithm to 150 mg/dl and additionally turns off the administration of auto-correction boluses. The CamAPS® FX application has an “Ease-off” function, which temporarily raises the target glucose level and more quickly suspends insulin delivery when a drop in glucose below the target level is anticipated. This function can be programmed to activate temporarily, which can be useful for children and adolescents who may forget or are unable to activate this function. It is also possible to periodically increase the target glucose level, for example, to 150 mg/dl. Guidelines for managing physical activity: “Sports Mode”: Use this mode during physical activity lasting longer than 30 minutes; it is essential during prolonged or all-day physical activities. Activate 90–120 minutes before the planned physical activity. In the case of an increased risk of hypoglycaemia after physical activity, maintain it for an additional 30–120 minutes after the activity. Reduction of mealtime bolus: A bolus administered within 2 hours before physical activity requires a reduction in the insulin dose by 25% (in some individuals, this reduction may need to be greater, even up to 30–50%) by adjusting the number of carbohydrates entered into the “bolus” function. The degree of reduction depends on the current glucose level, the type, and the duration of the planned physical activity. In many situations, modifying the bolus requires decisions based solely on the individual’s own experiences. Management of hypoglycaemia risk before and during physical activity: Carbohydrate consumption should be limited to 5–10 minutes before starting the activity – carbohydrates should not be entered into the system. Before unplanned physical activity, it is often necessary to consume an additional portion of carbohydrates. Consuming carbohydrates earlier than 20 minutes before physical activity may cause the system to increase insulin delivery, which can result in hypoglycaemia. Before starting physical activity, checking glucose levels, the amount of active insulin, and the dose delivered as basal are essential to prevent hypoglycaemia. This is especially important for unplanned physical activity. For sports that require disconnection of the pump, insulin delivery must be stopped during the pump disconnection. Otherwise, the algorithm accounts for the insulin that was not delivered to the person’s body. In the case of persistent hyperglycaemia > 250 mg/dl (CGM data), the following steps are recommended: Verify the result by measuring blood glucose with a glucose meter. If hyperglycaemia persists for more than 2 hours without a known cause, it is recommended that the infusion set be replaced. Administer a user-initiated correction bolus in automatic mode. If there is no response within 90 minutes, the next correction should be administered using an insulin pen/syringe. This correction is more effective because, in the case of infusion set occlusion and multiple pump correction attempts, the amount of active insulin registered by the system may be falsely elevated, preventing the administration of an additional bolus. If a correction dose of insulin is administered using an insulin pen, the automatic mode should be turned off for 2 to 4 hours to prevent the system from administering additional corrections while the insulin from the pen is still active. For hyperglycaemia lasting > 4 hours, monitoring blood or urine ketone levels is necessary. If blood glucose exceeds 350-400 mg/dl, it is advisable to administer the first correction dose using a pen/syringe and replace the infusion set and insulin immediately. A particularly dangerous situation indicating the presence of ketoacidosis is when blood glucose exceeds 250–300 mg/dl (usually for longer than 4 hours), and the blood concentration of beta-hydroxybutyric acid (BOHB) is above 3 mmol/l (or ketones in urine are high [≥ +++]). In such a situation, the individual requires immediate medical staff attention. When using an HCL system, hypoglycaemia is usually caused by an overly large meal bolus, greater-than-expected physical activity, or an illness accompanied by vomiting/diarrhoea/malabsorption. In the event of impending hypoglycaemia, the system will suspend insulin delivery. Therefore, half the usual dose of carbohydrates required in standard therapy is often sufficient to treat hypoglycaemia . Treat hypoglycaemia episodes with 0.15 g of carbohydrates per kg of body weight (maximum 8 g), except in cases of hypoglycaemia induced by physical activity or significant overestimation of the carbohydrate content/meal bolus. It is important not to administer an additional portion of carbohydrates before 15–20 minutes have passed because this could lead to hyperglycaemia and, consequently, exacerbate glycaemic fluctuations. In the MiniMed™ 780G system, the amount of carbohydrates administered for hypoglycaemia treatment is not entered into the system. In the CamAPS® FX system, the carbohydrate amount should be entered with the note „hypoglycaemia treatment”. Situations where insulin requirements suddenly deviate significantly from “patterns” of previous days can be challenging for HCL systems. Currently available commercial HCL systems administer insulin based on algorithms that partially rely on insulin doses from preceding days, which may prevent appropriate automatic insulin adjustments during acute infections. These systems do not account for ketone levels in the body, which also has significant practical implications. The basic principles of management during acute illnesses do not differ from those applicable to all children treated with insulin . Even during infections, efforts should be made to maintain blood glucose levels within the 70-180 mg/dl range and keep ketone levels normal or minimal. 9.1. Management during acute infections accompanied by hyperglycaemia For blood glucose levels >250 mg/dl persisting for more than 4 hours or in the presence of vomiting or ketonaemia, switch to manual mode and follow the procedures described in the “Management of Hyperglycaemia” section and in accordance with the PTD recommendations . 9.2. Management during acute illnesses accompanied by reduced insulin dose (e.g. gastroenteritis) 9.2.1. MiniMed™ 780G System In cases of reduced insulin doses and declining glucose levels, the MiniMed™ 780G system reduces or stops the automatic delivery of basal insulin. To minimise the risk of hypoglycaemia, the “Temporary Target” option can be used, keeping in mind that it is set for a specific period, up to a maximum of 24 hours, and may need to be repeated afterwards. Alternatively, a temporary target of 120 mg/dl or 110 mg/dl can be selected if a lower value was previously chosen. With low glucose levels, even after entering meal carbohydrates, the system may suggest a dose of zero units; it is important to confirm this dose because it informs the algorithm that a meal has been consumed, preparing it to act more aggressively if glucose levels rise. When “starvation ketones” appear, it may be advisable to enter the amount of consumed carbohydrates (and administer the meal bolus) approximately 10 minutes after starting the meal – when an increase in sensor glucose (SG) begins to be visible; however, this suggestion requires further verification in subsequent studies. If low glucose levels persist or hypoglycaemia recurs, switching to manual mode and temporarily using the predictive low-glucose suspend (PLGS) algorithm may be more beneficial. In this case, it is also possible to periodically reduce the basal rate. 9.2.2. CamAPS® FX System The algorithm attempts to reduce insulin delivery by calculating it for the next 2.5–4 hours. If this is insufficient and low glucose levels persist, the personal target glucose level can be increased for part or all of the day. The “Ease-off” function can also help prevent glucose levels from dropping further. For blood glucose levels >250 mg/dl persisting for more than 4 hours or in the presence of vomiting or ketonaemia, switch to manual mode and follow the procedures described in the “Management of Hyperglycaemia” section and in accordance with the PTD recommendations . 9.2.1. MiniMed™ 780G System In cases of reduced insulin doses and declining glucose levels, the MiniMed™ 780G system reduces or stops the automatic delivery of basal insulin. To minimise the risk of hypoglycaemia, the “Temporary Target” option can be used, keeping in mind that it is set for a specific period, up to a maximum of 24 hours, and may need to be repeated afterwards. Alternatively, a temporary target of 120 mg/dl or 110 mg/dl can be selected if a lower value was previously chosen. With low glucose levels, even after entering meal carbohydrates, the system may suggest a dose of zero units; it is important to confirm this dose because it informs the algorithm that a meal has been consumed, preparing it to act more aggressively if glucose levels rise. When “starvation ketones” appear, it may be advisable to enter the amount of consumed carbohydrates (and administer the meal bolus) approximately 10 minutes after starting the meal – when an increase in sensor glucose (SG) begins to be visible; however, this suggestion requires further verification in subsequent studies. If low glucose levels persist or hypoglycaemia recurs, switching to manual mode and temporarily using the predictive low-glucose suspend (PLGS) algorithm may be more beneficial. In this case, it is also possible to periodically reduce the basal rate. 9.2.2. CamAPS® FX System The algorithm attempts to reduce insulin delivery by calculating it for the next 2.5–4 hours. If this is insufficient and low glucose levels persist, the personal target glucose level can be increased for part or all of the day. The “Ease-off” function can also help prevent glucose levels from dropping further. In cases of reduced insulin doses and declining glucose levels, the MiniMed™ 780G system reduces or stops the automatic delivery of basal insulin. To minimise the risk of hypoglycaemia, the “Temporary Target” option can be used, keeping in mind that it is set for a specific period, up to a maximum of 24 hours, and may need to be repeated afterwards. Alternatively, a temporary target of 120 mg/dl or 110 mg/dl can be selected if a lower value was previously chosen. With low glucose levels, even after entering meal carbohydrates, the system may suggest a dose of zero units; it is important to confirm this dose because it informs the algorithm that a meal has been consumed, preparing it to act more aggressively if glucose levels rise. When “starvation ketones” appear, it may be advisable to enter the amount of consumed carbohydrates (and administer the meal bolus) approximately 10 minutes after starting the meal – when an increase in sensor glucose (SG) begins to be visible; however, this suggestion requires further verification in subsequent studies. If low glucose levels persist or hypoglycaemia recurs, switching to manual mode and temporarily using the predictive low-glucose suspend (PLGS) algorithm may be more beneficial. In this case, it is also possible to periodically reduce the basal rate. The algorithm attempts to reduce insulin delivery by calculating it for the next 2.5–4 hours. If this is insufficient and low glucose levels persist, the personal target glucose level can be increased for part or all of the day. The “Ease-off” function can also help prevent glucose levels from dropping further. Currently, there is no evidence to suggest that HCL systems can be safely used during the perioperative period. However, some paediatric diabetes centres, based on their own experience, continue to use these systems in full functionality during certain minor procedures performed under anaesthesia . Based on the available literature (mainly expert opinions ) it can be suggested that when using HCL systems during minor procedures, it may be beneficial to activate the Temporary Target function (in the MiniMed™ 780G system) or set a higher personal target glucose level, or use the “Ease-off” function in the CamAPS® FX system. Maintaining insulin delivery through a personal insulin pump during procedures under general anaesthesia is possible only if the anaesthesiology team accepts this method and if close interdisciplinary collaboration with a diabetologist is possible within the facility. For major procedures planned for more than 2 hours that require skipping more than one meal, it is necessary to switch to intravenous insulin therapy and follow the guidelines for “major procedures” as described in the PTD recommendations . All emergency procedures should be performed according to the principles applicable to “major procedures” . shows the management during radiological examinations . During radiological examinations performed under sedation (e.g. MRI and CT in small and uncooperative children), insulin therapy should be managed according to the guidelines for minor surgical procedures, considering the remarks for procedures in individuals using a pump. Based on their own experiences, some paediatric diabetes centres continue to use HCL systems in full functionality during short-term sedation. However, there is currently no scientific evidence to confirm their complete safety in the perioperative/sedation period. The use of HCL is recommended during hospitalisation in non-diabetes units, provided it is not contraindicated due to the clinical condition or the needs related to the treatment or diagnostics being performed. In the case of children and younger adolescents, parental or guardian supervision over the system’s operation is necessary.. The issues associated with steroid therapy in the treatment of T1D are primarily related to significant changes in insulin sensitivity and insulin requirements, both when starting and when suddenly stopping their use. The management should be tailored to the individual, considering the steroid dosage, the planned duration of treatment, any plans for prolonged tapering of steroids, and the HCL system in use. For short-term treatment with very high doses of steroids, it may be necessary to switch to manual mode in the Medtronic MiniMed™ 780G and CamAPS® FX systems. In the case of CamAPS® FX, consideration may be given to using the “Boost” function and/or setting lower target glucose values. For individuals on long-term steroid therapy or those who plan to taper the dose gradually, it is recommended that HCL be continued under close supervision by the diabetes care team. The system may require modification of settings. After discontinuing steroids, to reduce the risk of hypoglycaemia, it may be helpful to temporarily set higher target values for 24–48 hours . During pregnancy and postpartum, insulin requirements change rapidly and significantly. It is important to note that lower target glucose values (63–140 mg/dl) are recommended during pregnancy, necessitating frequent adjustments to therapy. Therefore, constant and frequent adjustments of the HCL system settings could help achieve better metabolic control of diabetes . Currently, only the CamAPS® FX system is officially registered for use during pregnancy. During pregnancy, setting the target glucose level to below 100 mg/dl is recommended. The MiniMed™ 780G system is not yet registered for use during pregnancy. However, studies have shown that this system does not worsen obstetric outcomes and improves metabolic control at night . Differences between CGM readings and glucometer measurements are especially noticeable during episodes of hypoglycaemia or when there are rapid and significant fluctuations in glucose levels. Certain medications can noticeably impact the accuracy of CGM readings (e.g. acetaminophen and hydroxyurea). The body’s hydration level can affect the accuracy of CGM readings. It is essential to stay vigilant and cross-check CGM readings with glucometer measurements if the individual experiences symptoms that do not align with the CGM readings. Inflammatory skin reactions at the site of the sensor or infusion set may occur, which can be due to skin irritation, an allergic reaction to the device material, the adhesive, or the products used before device application. There can also be hypertrophy or hypotrophy of the subcutaneous tissue at the infusion set site. 16.1. Guidelines for conducting a visit for an individual treated with an hybrid closed loop system The use of HCL systems is based on data from continuous glucose monitoring (CGM). The success of the therapy can be measured by the improvement in CGM parameters, such as time in range (TIR), time below range (TBR), coefficient of variation (CV), and time in tight range (TITR). A diabetes management visit must be based on data from HCL system reports, allowing for proper interpretation of results and modification of insulin therapy settings. The HCL report, typically covering 14 days, provides insight into insulin therapy settings and CGM parameter outcomes 16.2. Report analysis and patient visit The analysis of the HCL report and the information discussed/verified during the visit should include the elements presented in . The use of HCL systems is based on data from continuous glucose monitoring (CGM). The success of the therapy can be measured by the improvement in CGM parameters, such as time in range (TIR), time below range (TBR), coefficient of variation (CV), and time in tight range (TITR). A diabetes management visit must be based on data from HCL system reports, allowing for proper interpretation of results and modification of insulin therapy settings. The HCL report, typically covering 14 days, provides insight into insulin therapy settings and CGM parameter outcomes The analysis of the HCL report and the information discussed/verified during the visit should include the elements presented in . Currently, HCL systems represent the most advantageous therapeutic option for individuals with diabetes who require intensive insulin therapy. They not only improve metabolic control of diabetes but also enhance the quality of life for children and adolescents with diabetes, as well as their families. For this reason, HCL systems should be the first choice regarding insulin therapy methods within the paediatric population. HCL systems can be implemented at any stage of diabetes management, whether in patients with newly diagnosed T1D or those with a longer disease duration, regardless of the previously used insulin therapy method. For many patients who have been unable to achieve good metabolic control using traditional intensive insulin therapy methods, applying HCL systems may be the only way to improve metabolic control and reduce the risk of acute and chronic complications of the disease. |
Comparative Outcomes of Surgical Techniques for Pilonidal Sinus: A Turkish Retrospective Study | 2e55c339-b85f-407f-af82-108fe588100b | 11871746 | Surgical Procedures, Operative[mh] | Pilonidal sinus is an acute or chronic infectious disease that occurs in the intergluteal cleft region, with an incidence rate of 26 in 100 000 . It is more commonly seen in young men aged 15 to 30 years who have obesity, are hairy, and have a sedentary lifestyle . This disease is characterized by subcutaneous epithelialized sinuses and inflamed pouches containing hair and foreign matter that form in the sacrococcygeal region . Although sometimes asymptomatic, it usually presents with purulent discharge and abscess from the sinus tracts. Its cause and appropriate treatment were first described by Mayo in 1833 . The diagnosis of this disease is made based on the specific history and physical examination by detecting characteristic midline pits in the gluteal cleft region . According to the most widely accepted hypothesis, the presence of dermopathy, keratin plugs, hair remnants in the affected area, and related foreign body reactions are important . Recently, it has been emphasized that the collagen type I/III ratio and total collagen amount in the midline sacrococcygeal region being lower than in the lateral skin region may be important in the formation of pilonidal sinus disease . This disease negatively affects patients’ quality of life and social life. Although it has been a long time since its definition, the most optimal treatment method is still a matter of debate. Several surgical methods are used in its treatment. Patients who require surgery due to chronic pilonidal disease can undergo excision and primary midline closure (PMC), excision with secondary healing, or excision with marsupialization. Flap techniques such as primary excision with Limber flap (LF) or Karydakis flap (KF) reconstruction are recommended for more complex and recurrent diseases . These known classical surgical methods require general anesthesia or spinal anesthesia. Especially considering that most pilonidal sinus disease is encountered in school-age children and young adults, these surgical techniques cause children to not be able to attend school for a long time after surgery and restrict social life in adults. At the same time, these classical methods also bring about large scars and cosmetic concerns. All these factors have led to the search for alternative treatment methods that are more minimally invasive and less disruptive to social life, for the treatment of pilonidal sinus disease. For this purpose, crystallized phenol and platelet-rich plasma applications have been used, especially in pediatric patients . Recently, the development of surgical technology has led to the emergence of minimally invasive techniques in pilonidal sinus disease treatment, such as laser pilonidotomy (LP) . Therefore, this retrospective study from a single center in Turkey aimed to compare outcomes following PMC, KF, LF, and LP in adult patients with pilonidal sinus disease.
Study Design This study was approved by the relevant clinical ethics committee (decision no: KA23/107). Informed consent forms were obtained from all patients. A retrospective analysis was conducted on the data of 476 patients aged 18 years and older who underwent surgery for pilonidal sinus disease at our clinic between January 2011 and September 2022. Fifty-four patients whose data could not be accessed were excluded from the study. The surgeries were performed by 4 different surgeons. The diagnosis of the patients was made based on the characteristic anamnesis and the observation of characteristic midline pits in the gluteal cleft on physical examination. Patients were divided into 4 groups according to the surgical technique: PMC (n=228), LF (n=82), KF (n=53), and LP (n=59). A flow chart of the study is shown in . Demographic data of patients (age, sex, comorbidities, primary or recurrent disease, preoperative abscess history, and follow-up period), surgical findings (operation time, presence of drain, hospital stay, postoperative complications, and postoperative recurrence) and, quality of life (QOL) indicators (pain-free toilet sitting time, pain-free walking time, time to return to work, and patient satisfaction levels) were recorded. All patients included in the study were contacted by telephone to assess their satisfaction using a Likert scale. Surgical Technique PMC, KF, and LF were performed under spinal anesthesia with the classical technique previously mentioned in the literature . Prior to anesthesia induction, all patients were administered intravenous cefazolin sodium (1 g). PMC Techinique After spinal anesthesia, with the patient in the jack-knife position, the skin was cleaned with antiseptic solution, and methylene blue dye was administered through the pits. The sinus was removed together with the healthy tissue, using an elliptical incision to include the stained diseased tissues and pits. A drain was used in some patients, according to the surgeon’s preference. Subcutaneous tissues were approximated with 2/0 and 3/0 vicryl sutures. The skin was closed with absorbable or non-absorbable sutures . LF technique The sinus tissue was excised down to the presacral fascia through a rhomboid incision, including the sinus pits and tissues stained with methylene blue. Then, a flap was prepared from the gluteal muscle and fixed to the presacral fascia. A drain was placed in the cavity in most patients. Then, the subcutaneous tissue was approximated with 2/0 and 3/0 vicryl sutures. The skin was closed with non-absorbable 2/0 monofilament sutures . KF Technique Sinus tissue was excised with an asymmetric and elliptical incision, including the sinus pits. Then, flap tissue was prepared, with a depth of 1 cm and a thickness of 2 cm. Then, the flap tissue was placed in the cavity to shift the midline by 2 cm. Depending on the surgeon’s preference, a drain was placed in the cavity in some of the patients, and the subcutaneous tissues were approximated with 2/0 or 3/0 vicryl sutures. The skin was closed with absorbable 3-0 sutures . LP Technique In the LP group, the procedures were performed as described by Dessily et al, using the NeoV 1470 diode laser machine device . LP was performed under local anesthesia with the patient in the jack-knife position. After cleansing and brushing the skin with povidone iodine antiseptic solution after shaving, the sinus pits were excised and expanded with a scalpel. Then, the hairs were removed from the sinus with a mosquito clamp, and the sinus was cleaned with a curette. The direction, length, and width of the sinus were confirmed with the help of a stylet, to use the appropriate laser probe. Before starting the procedure, saline solution was injected around the pit and under the skin to prevent burning of the surrounding tissues. A radial diode laser probe at a wavelength of 1470 nm was then used . Laser energy was 10 Watts. Fiber delivered energy homogeneously and continuously at 3600. As the probe was withdrawn at a rate of approximately 1 mm/s, the sinus shrunk and closed. If the channel did not close after the first pull, a second pull was made. The patient was discharged on the same day of surgery. The patient was recommended to use analgesics (preferably paracetamol) if needed. In the postoperative period, no special care was required other than covering the pits with compresses after taking a shower. Statistical Analysis Statistical analysis was performed retrospectively. Descriptive statistics are reported using numbers and percentages for categorical variables and median (minimum–maximum) for numerical variables, depending on the data distribution. The normal distribution of the data was evaluated using the Shapiro-Wilk test. Relationships between numerical measurements were investigated using Pearson or Spearman correlation coefficients, depending on the data distribution. To compare numerical measurements based on sociodemographic characteristics and research groups, the Kruskal-Wallis test was used for independent groups with more than 2 groups, in accordance with the data distribution. For comparisons of proportions or investigations based on research groups, the chi-square or Fisher exact test was used. A significance level of P <0.05 was considered statistically significant. The power of statistical tests was determined as 0.80, and effect size was determined as 0.5 in G power analysis.
This study was approved by the relevant clinical ethics committee (decision no: KA23/107). Informed consent forms were obtained from all patients. A retrospective analysis was conducted on the data of 476 patients aged 18 years and older who underwent surgery for pilonidal sinus disease at our clinic between January 2011 and September 2022. Fifty-four patients whose data could not be accessed were excluded from the study. The surgeries were performed by 4 different surgeons. The diagnosis of the patients was made based on the characteristic anamnesis and the observation of characteristic midline pits in the gluteal cleft on physical examination. Patients were divided into 4 groups according to the surgical technique: PMC (n=228), LF (n=82), KF (n=53), and LP (n=59). A flow chart of the study is shown in . Demographic data of patients (age, sex, comorbidities, primary or recurrent disease, preoperative abscess history, and follow-up period), surgical findings (operation time, presence of drain, hospital stay, postoperative complications, and postoperative recurrence) and, quality of life (QOL) indicators (pain-free toilet sitting time, pain-free walking time, time to return to work, and patient satisfaction levels) were recorded. All patients included in the study were contacted by telephone to assess their satisfaction using a Likert scale.
PMC, KF, and LF were performed under spinal anesthesia with the classical technique previously mentioned in the literature . Prior to anesthesia induction, all patients were administered intravenous cefazolin sodium (1 g).
After spinal anesthesia, with the patient in the jack-knife position, the skin was cleaned with antiseptic solution, and methylene blue dye was administered through the pits. The sinus was removed together with the healthy tissue, using an elliptical incision to include the stained diseased tissues and pits. A drain was used in some patients, according to the surgeon’s preference. Subcutaneous tissues were approximated with 2/0 and 3/0 vicryl sutures. The skin was closed with absorbable or non-absorbable sutures .
The sinus tissue was excised down to the presacral fascia through a rhomboid incision, including the sinus pits and tissues stained with methylene blue. Then, a flap was prepared from the gluteal muscle and fixed to the presacral fascia. A drain was placed in the cavity in most patients. Then, the subcutaneous tissue was approximated with 2/0 and 3/0 vicryl sutures. The skin was closed with non-absorbable 2/0 monofilament sutures .
Sinus tissue was excised with an asymmetric and elliptical incision, including the sinus pits. Then, flap tissue was prepared, with a depth of 1 cm and a thickness of 2 cm. Then, the flap tissue was placed in the cavity to shift the midline by 2 cm. Depending on the surgeon’s preference, a drain was placed in the cavity in some of the patients, and the subcutaneous tissues were approximated with 2/0 or 3/0 vicryl sutures. The skin was closed with absorbable 3-0 sutures .
In the LP group, the procedures were performed as described by Dessily et al, using the NeoV 1470 diode laser machine device . LP was performed under local anesthesia with the patient in the jack-knife position. After cleansing and brushing the skin with povidone iodine antiseptic solution after shaving, the sinus pits were excised and expanded with a scalpel. Then, the hairs were removed from the sinus with a mosquito clamp, and the sinus was cleaned with a curette. The direction, length, and width of the sinus were confirmed with the help of a stylet, to use the appropriate laser probe. Before starting the procedure, saline solution was injected around the pit and under the skin to prevent burning of the surrounding tissues. A radial diode laser probe at a wavelength of 1470 nm was then used . Laser energy was 10 Watts. Fiber delivered energy homogeneously and continuously at 3600. As the probe was withdrawn at a rate of approximately 1 mm/s, the sinus shrunk and closed. If the channel did not close after the first pull, a second pull was made. The patient was discharged on the same day of surgery. The patient was recommended to use analgesics (preferably paracetamol) if needed. In the postoperative period, no special care was required other than covering the pits with compresses after taking a shower.
Statistical analysis was performed retrospectively. Descriptive statistics are reported using numbers and percentages for categorical variables and median (minimum–maximum) for numerical variables, depending on the data distribution. The normal distribution of the data was evaluated using the Shapiro-Wilk test. Relationships between numerical measurements were investigated using Pearson or Spearman correlation coefficients, depending on the data distribution. To compare numerical measurements based on sociodemographic characteristics and research groups, the Kruskal-Wallis test was used for independent groups with more than 2 groups, in accordance with the data distribution. For comparisons of proportions or investigations based on research groups, the chi-square or Fisher exact test was used. A significance level of P <0.05 was considered statistically significant. The power of statistical tests was determined as 0.80, and effect size was determined as 0.5 in G power analysis.
Patient Characteristics The median age of the patients was 23 years. Of the patients, 75.9% were male (n=335) and 20.6% were female (n=87). There were no significant differences between the groups in terms of demographic characteristics such as age and sex. Seventy-five (17.8%) of the patients were recurrent cases. There was no significant difference between the groups in terms of preoperative primary or recurrent cases ( P= 0.782). There was no significant difference between the groups in terms of the presence of preoperative abscess drainage history and comorbidity (P =0.480 and P =0.174, respectively). The follow-up period of the LP group was significantly shorter than that of all other groups (mean 14 months; P <0.001). Patient characteristics are summarized in . Surgical Outcomes When the surgical findings were evaluated, the operation time was significantly shorter in the LP group than in all other groups ( P <0.001; mean 35 min). While no drains were used in the LP group, drains were used in most patients who underwent LF (92.7%; P <0.001). In evaluating postoperative complications, the PMC group had a significantly higher incidence of seroma and wound dehiscence ( P =0.006 and P <0.001, respectively). Although the incidence of wound infection was higher in the PMC group, this difference was not statistically significant. Necrosis did not develop in any group. Hospital stay in the LP group was significantly shorter than that of all other groups (mean, 8 h; P <0.001). Mean hospital stay was 24 h in the PMC and KF groups and 48 h in the LF group. The highest recurrence rate was observed in the PMC group during the follow-up period (14.5%; P =0.017). In contrast, recurrence occurred earlier in the LP group than in the other groups ( P =0.016). The surgical findings are summarized in . QOL Outcomes When the QOL of patients was evaluated, those who underwent LP started pain-free walking and pain-free toilet sitting statistically significantly earlier ( P <0.001 and P <0.001, respectively). Additionally, patients in the LP group returned to work earlier ( P <0.001). Patients in the LP group started pain-free toilet sitting on average 2 days after surgery, pain-free walking on average 3 days after surgery, and returned to work on average 7 days after surgery. A Likert scale was used to determine patient satisfaction levels. According to this scale, the percentage of patients dissatisfied with the surgery was statistically significantly higher in the PMC group ( P <0.001). The findings related to QOL are summarized in .
The median age of the patients was 23 years. Of the patients, 75.9% were male (n=335) and 20.6% were female (n=87). There were no significant differences between the groups in terms of demographic characteristics such as age and sex. Seventy-five (17.8%) of the patients were recurrent cases. There was no significant difference between the groups in terms of preoperative primary or recurrent cases ( P= 0.782). There was no significant difference between the groups in terms of the presence of preoperative abscess drainage history and comorbidity (P =0.480 and P =0.174, respectively). The follow-up period of the LP group was significantly shorter than that of all other groups (mean 14 months; P <0.001). Patient characteristics are summarized in .
When the surgical findings were evaluated, the operation time was significantly shorter in the LP group than in all other groups ( P <0.001; mean 35 min). While no drains were used in the LP group, drains were used in most patients who underwent LF (92.7%; P <0.001). In evaluating postoperative complications, the PMC group had a significantly higher incidence of seroma and wound dehiscence ( P =0.006 and P <0.001, respectively). Although the incidence of wound infection was higher in the PMC group, this difference was not statistically significant. Necrosis did not develop in any group. Hospital stay in the LP group was significantly shorter than that of all other groups (mean, 8 h; P <0.001). Mean hospital stay was 24 h in the PMC and KF groups and 48 h in the LF group. The highest recurrence rate was observed in the PMC group during the follow-up period (14.5%; P =0.017). In contrast, recurrence occurred earlier in the LP group than in the other groups ( P =0.016). The surgical findings are summarized in .
When the QOL of patients was evaluated, those who underwent LP started pain-free walking and pain-free toilet sitting statistically significantly earlier ( P <0.001 and P <0.001, respectively). Additionally, patients in the LP group returned to work earlier ( P <0.001). Patients in the LP group started pain-free toilet sitting on average 2 days after surgery, pain-free walking on average 3 days after surgery, and returned to work on average 7 days after surgery. A Likert scale was used to determine patient satisfaction levels. According to this scale, the percentage of patients dissatisfied with the surgery was statistically significantly higher in the PMC group ( P <0.001). The findings related to QOL are summarized in .
This study has some important implications. First, our study is the first to compare 4 different surgical techniques for pilonidal sinus surgery: PMC, KF, LF, and LP. Other important results include the following. The PMC group had the highest rate of postoperative seroma, wound dehiscence, and wound infection. The patients who underwent KF had the lowest rate of recurrence. Although postoperative morbidity was not low, the most satisfactory group in terms of QOL was the group of patients who underwent LP. Consistent with previous studies , our study found that young men are more likely to have pilonidal sinus disease, and there were no significant differences between the groups in terms of demographic data. Given that we have been performing LP for approximately 4 years in our clinic, it is not surprising that the follow-up period was significantly shorter in this group. In recent years, the development of minimally invasive surgery and its lesser impact on patients’ QOL have led us to occasionally choose the LP method. In this study, the possibility of performing LP under local anesthesia without the need for a drain and the significantly shorter operation time may make this method preferred. In the LF group, drains were used significantly more than in the other groups, to prevent seroma formation due to the wider tissue loss. In this study, the rates of postoperative seroma, wound dehiscence, and wound infection were higher in the PMC group than in the other groups. Similar to the results of a study by Erkent et al, who compared PMC, LF, and KF methods, wound site complications were higher in the PMC group . Another retrospective study comparing PMC, LF, and KF methods also found a significantly higher incidence of seroma, wound site infection, and wound dehiscence in the PMC group . In contrast, Ekici et al compared the lay-open, PMC, LF, and KF methods and found no significant differences in terms of postoperative complications . Similarly, Destek et al compared the LF and KF methods and found no differences in terms of postoperative complications . In the meta-analysis conducted by Gavriilidis et al comparing LF and KF methods, no significant difference was observed in terms of postoperative morbidity . A meta-analysis conducted in China indicated that PMC had the worst outcomes in terms of postoperative infectious complications . In the meta-analysis of randomized studies conducted by Berthier et al, flap methods were superior to primary closure methods in terms of wound complications, including wound infection and dehiscence . There are few studies in the literature that compare LP with other primary closure or flap methods. Algazar et al observed no significant difference between LP and LF methods in a prospective non-randomized study in terms of postoperative wound complications . In a retrospective study comparing the LP and sinus lay open techniques, no significant difference was found in terms of wound complications . In a case series of 60 patients undergoing LP, Georgiou et al found a high seroma formation rate (21.6%) and a low wound infection rate (1.6%) . Porwal et al reported that the incidence of wound infection was 0.88%, and the incidence of seroma was 6.58% in an LP case series of 228 patients . In our study, the incidence of seroma and wound infection in patients who underwent LP was 3.4%. Unlike the few studies in the literature on the laser method, Taşkin et al compared 2 minimally invasive methods, laser and crystallized phenol methods . As a result of their study, although postoperative pain and hematoma/bleeding were significantly less in the laser group, there was no significant difference between the groups in terms of patient satisfaction and recurrence. One of the most pressing issues in the treatment of pilonidal sinus disease is recurrence. The recurrence rate varies depending on the surgical technique used and the follow-up period. In our study, consistent with the literature, the postoperative recurrence rate was significantly higher in the PMC group (14.5%; P =0.017). In the LP group, recurrence occurred earlier after surgery than in other groups (mean 7 months; P =0.016). Studies by Ekici et al , Erkent et al , Karaca et al , and Kartal et al reported recurrence rates of 9.1%, 10.8%, 9.2%, and 11.44%, respectively. Recent literature has shown that PMC has a higher recurrence rate than other surgical techniques. Stauffer et al conducted a meta-analysis of 11 700 patients and showed that the recurrence rate of PMC was 2.1% at 12 months, 7% at 24 months, and 21.9% at 60 months . Among the methods studied in the present study, KF had the lowest recurrence rate (3.8%; P =0.017). The aim of the KF method is to reduce hair collection and mechanical irritation by flattening the natal cleft with a lateral shift of the midline and to reduce the possibility of recurrence of the disease . When Karydakis introduced this method to the literature in 1992, he reported a recurrence rate of 1% and a complication rate of 8% . Different authors have reported different recurrence rates for the KF method, ranging from 2.16% (18), 2.5% (14), and 3.16% (17), to 7.5% (19). A meta-analysis investigating the recurrence of pilonidal sinus after surgery reported a recurrence rate of 1.5% at 12 months, 2.4% at 24 months, and 10.2% at 60 months in patients who underwent KF . In this meta-analysis, it was reported that flap methods (KF and LF) had a lower recurrence rate compared to other methods . Surgeons tend to prefer minimally invasive methods, and recently, the LP procedure for the treatment of pilonidal sinus disease has been shown to improve patients’ QOL . There are no randomized controlled studies in the literature evaluating the results and follow-up period of the LP procedure, due to it being a new technique. In a non-randomized prospective study comparing the results of LP and LF, the operative time and hospital stay were significantly shorter, and the time to return to work was earlier in LP , which is consistent with our study. These findings can be attributed to the fact that LP is performed under local anesthesia and does not require a large incision. Pappas et al also showed that patients who underwent LP returned to normal social and physical activities earlier . In a study by Abdelnaby et al, patients who underwent LP started pain-free walking in about 4 days and returned to work in 7 days . In our study, patients who underwent LP started pain-free toilet sitting 2 days after surgery and pain-free walking after an average of 3 days and returned to work after an average of 7 days. When evaluated on the Likert scale, the number of highly satisfied patients in the LP group was significantly higher than that in other groups (66.1%; P <0.001). However, despite having a shorter follow-up period, the LP group had the second-highest recurrence rate (6.8%) after the PMC group. Additionally, recurrence occurred earlier in patients who underwent LP (mean 7 months; P =0.016). Consistent with our study, an exploratory study conducted in India reported that 82.5% of patients were highly satisfied with the LP method . Although LP may seem advantageous in terms of returning to social lives, a meta-analysis has indicated that it is equally disadvantageous in terms of postoperative recurrence (recurrence rate of 1.9% at 12 months, 5.1% at 24 months, and 36.6% at 60 months) . However, as our study also shows, recurrences in the LP group and their early detection negatively affected the follow-up period. We believe that recurrences associated with LP can be reduced by completing the learning curve and selecting appropriate patients. Limitations This study has some limitations. First, our study had a retrospective design. Second, the sample size of this study was small. Third, the follow-up period was shorter in the LP group.
This study has some limitations. First, our study had a retrospective design. Second, the sample size of this study was small. Third, the follow-up period was shorter in the LP group.
PMC appears to be the most disadvantageous method due to its high postoperative complication and recurrence rates. The KF method, which has a lower recurrence rate, appears to be a safer technique. In selected patients, the LP method may be more advantageous in terms of QOL and patient satisfaction. However, further advanced studies with larger sample sizes and longer follow-up periods are needed.
|
Molecular diagnosis of SARS-CoV-2 in seminal fluid | d5517c34-6b2d-4b01-8dce-7de9a4beb84d | 8085093 | Pathology[mh] | Coronaviruses are a family of positive-sense single-stranded RNA viruses that cause infections in birds and mammals as well as humans, inducing respiratory, hepatic, neurological and gastrointestinal diseases . The novel coronavirus SARS-CoV-2 causes pneumonia, a severe acute respiratory disease (COVID-19). Structurally, SARS-CoV-2 is composed of several proteins: nucleocapsid (N), spike (S), membrane (M) and envelope (E). The spike protein is particularly important, as it enables the virus to enter and infect host cells and determines viral pathogenesis, host tropism, and disease . The use of accurate molecular tests has enabled the presence and development of this virus to be monitored. The gold standard for the diagnosis of SARS-CoV-2 infection is qualitative reverse transcription-polymerase chain reaction (qRT-PCR) testing of nasopharyngeal swabs . The usual SARS-CoV-2 gene targets are E, S, N1, N2, and RpRd. The RT-PCR cycle threshold ( C t ) value is an indicator of the number of viral copies, with lower C t values corresponding to higher viral copy numbers. A C t less than 40 is interpreted as positive for SARS-CoV-2 RNA . However, it must be pointed out that C t values are not standardized to enable quantification of the viral concentration. Recently, some authors observed that the N2 gene may be prone to false positive results. Particularly high C t values (> 40) have been detected in nasopharyngeal swabs using N2 as the RT-PCR target, suggesting either “very low” viral load or "false positive" results. Careful interpretation of the clinical relevance of this “very low” test result is currently needed . Although respiratory samples are the reference specimens, the virus has been found in numerous human samples, including urine, feces, cerebrospinal fluid, lacrimal fluid and blood . In several studies, bronchoalveolar lavage fluid (BLF) (93%), sputum (72%) pharyngeal swabs (32%), feces (29%), and blood (1%) samples have tested positive. None of the urine samples tested were positive . According to some authors, these different viral loads could be attributed to the sample type or timing, the stage of the disease and/or where the specimen was taken from, all factors that play an important role in RT-qPCR results . Negative test results do not necessarily rule out infection. False negatives can occur in preanalytical steps (poor specimen collection, inappropriate sampling), analytical steps (PCR inhibition, target mutation or low viral load in the sample), and postanalytical steps (transcription error) . In contrast, false positives may arise due to two problems associated with RT-qPCR: contamination and determination of the limit at which it may be affirmed that a sample with a low viral load is in fact positive . Notably, a positive molecular test indicates only the detection of viral RNA and may be unrelated to the presence of infectious virus . In addition to methodological aspects, it should be stressed that a different SARS-CoV-2 level may be the result of different tissue expression of the receptors with which the virus interacts—angiotensin-converting enzyme 2 (ACE2) and transmembrane protease serine 2 (TMPRSS2)—suggesting possible routes of infection other than respiratory droplets. For this reason, research efforts have to date focused on various objectives, including the study of the routes of viral transmission and the research and validation of diagnostic methods. The impact of SARS-CoV-2 on male reproduction has not yet been established. An important aspect for reproductive medicine is whether or not this virus is found in seminal fluid. While a number of recent literature studies have investigated the presence of SARS-CoV-2 in semen, only one reported positive results, in four acute and two recovering COVID-19 patients (19%) . However, this study may have several major methodological limitations . The identification of SARS-CoV-2 in different clinical samples using RT-PCR is not yet well established. For this reason, the aim of our study was to verify RT-PCR in semen samples, to establish if SARS-CoV-2 is truly found in semen and if this can be used for diagnosis. Patients Patients from the Infectious Disease Department of the Policlinico Umberto I Hospital—“Sapienza” University of Rome were asked to provide a semen sample for viral RNA determination. They comprised: Mild COVID-19 patients with a recent nasopharyngeal swab positive for SARS-CoV-2 (pt #1 and pt #2). Recovered COVID-19 patients with a recent nasopharyngeal swab negative for SARS-CoV-2 ( pt #3 and pt #4). Controls were recruited from healthy men attending the Laboratory of Seminology—Sperm Bank “Loredana Gandini'' who were performing semen analysis as a part of pre-conceptional screening and had had a negative nasopharyngeal swab for SARS-CoV-2. All patients provided their written informed consent before any study procedures were carried out. This study was approved by our institution’s Ethics Committee (Ref. 5971, protocol 0646/2020). Semen analysis Semen samples were collected by masturbation into a sterile plastic container. Given the patients’ medical condition, days of abstinence were not taken into consideration. All samples were allowed to liquefy at 37 °C for 60 min and were then assessed according to World Health Organization guidelines (2010). The following variables were assessed: ejaculate volume (ml), sperm concentration ( n × 10 6 /ml), total sperm number ( n × 10 6 /ejaculate), progressive motility and morphology (% abnormal forms). A sperm viability test was carried out to differentiate cell death from immotility by staining with eosin Y 0.5% in saline solution. Processing of semen samples To evaluate the possible presence of SARS-CoV-2 in different fractions of seminal fluid, each sample was processed as follows (Fig. ): 140 µl of whole seminal fluid was used to extract Viral RNA. An aliquot of seminal fluid was centrifuged at 3000 rpm for 10 min to separate seminal plasma from spermatozoa and other cellular elements. The pellet containing spermatozoa, leukocytes and epithelial cells was diluted with 0.5% saline (cell suspension). 140 µl of seminal plasma and 140 µl of cell suspension were used to extract viral RNA. Detection of SARS-CoV-2 RNA in semen Viral RNA from 140 µl of whole seminal fluid, and seminal plasma was extracted using QIAamp viral RNA kit (Qiagen) according to the manufacturer's instructions. Six µl of a heterologous amplification system (Internal Control-RealStar SARS-CoV2 RT PCR, Altona Diagnostics) was used as control for extraction procedure and RT-PCR inhibition. Total RNA extraction from cell suspension was performed using Norgen total RNA purification kit (Norgen Biotek Corporation) according to the manufacturer's instructions. Ten µl of extracted RNA was reverse-transcribed and simultaneously amplified using a real-time RT-PCR system (RealStar SARS-CoV2 RT PCR, Altona Diagnostics) targeting E and S viral genes. PRM1 and PRM2 mRNA, sperm-specific nuclear proteins, was used as the control for sperm RNA extraction, using RT-PCR (TaqMan™ Gene Expression Assay, Applied Biosystems). Control panel To assess whether viral RNA extraction was affected by the use of whole seminal fluid, a known titer of SARS-CoV-2 virus was added to semen from a healthy donor. The panel was prepared with serial dilutions of a SARS-CoV-2 isolate (named 2019-nCoV/Italy-INMI1) . The virus was collected by nasopharyngeal swab and cultured in Vero E6 cell lines grown in MEM containing 2% FBS. The dilutions ranged from 4 × 10 2 to 4 × 10 6 viral RNA copies/ml, corresponding to 0.1 and 1000 TCID50 (50% tissue culture infective dose)/ml. SARS-CoV-2 RNA was amplified by qRT-PCR and quantified based on a standard curve prepared through serial dilutions of EURM-019 single-stranded RNA (ssRNA) fragments of SARS-CoV-2 including different target genes ( https://crm.jrc.ec.europa.eu/p/EURM-019 ). Viral titers were determined by limiting dilution assay on Vero E6 cells and infectivity was expressed as TCID50/ml, calculated according to the Reed and Muench method. A blank containing only cell culture medium was included in the panel. Two known titer viral preparations from the panel were diluted 1:2 in seminal fluid and in 0.5% saline solution (Fig. ). The final concentrations of the tested samples were 2 × 10 4 and 2 × 10 6 copies/ml. Patients from the Infectious Disease Department of the Policlinico Umberto I Hospital—“Sapienza” University of Rome were asked to provide a semen sample for viral RNA determination. They comprised: Mild COVID-19 patients with a recent nasopharyngeal swab positive for SARS-CoV-2 (pt #1 and pt #2). Recovered COVID-19 patients with a recent nasopharyngeal swab negative for SARS-CoV-2 ( pt #3 and pt #4). Controls were recruited from healthy men attending the Laboratory of Seminology—Sperm Bank “Loredana Gandini'' who were performing semen analysis as a part of pre-conceptional screening and had had a negative nasopharyngeal swab for SARS-CoV-2. All patients provided their written informed consent before any study procedures were carried out. This study was approved by our institution’s Ethics Committee (Ref. 5971, protocol 0646/2020). Semen samples were collected by masturbation into a sterile plastic container. Given the patients’ medical condition, days of abstinence were not taken into consideration. All samples were allowed to liquefy at 37 °C for 60 min and were then assessed according to World Health Organization guidelines (2010). The following variables were assessed: ejaculate volume (ml), sperm concentration ( n × 10 6 /ml), total sperm number ( n × 10 6 /ejaculate), progressive motility and morphology (% abnormal forms). A sperm viability test was carried out to differentiate cell death from immotility by staining with eosin Y 0.5% in saline solution. To evaluate the possible presence of SARS-CoV-2 in different fractions of seminal fluid, each sample was processed as follows (Fig. ): 140 µl of whole seminal fluid was used to extract Viral RNA. An aliquot of seminal fluid was centrifuged at 3000 rpm for 10 min to separate seminal plasma from spermatozoa and other cellular elements. The pellet containing spermatozoa, leukocytes and epithelial cells was diluted with 0.5% saline (cell suspension). 140 µl of seminal plasma and 140 µl of cell suspension were used to extract viral RNA. Viral RNA from 140 µl of whole seminal fluid, and seminal plasma was extracted using QIAamp viral RNA kit (Qiagen) according to the manufacturer's instructions. Six µl of a heterologous amplification system (Internal Control-RealStar SARS-CoV2 RT PCR, Altona Diagnostics) was used as control for extraction procedure and RT-PCR inhibition. Total RNA extraction from cell suspension was performed using Norgen total RNA purification kit (Norgen Biotek Corporation) according to the manufacturer's instructions. Ten µl of extracted RNA was reverse-transcribed and simultaneously amplified using a real-time RT-PCR system (RealStar SARS-CoV2 RT PCR, Altona Diagnostics) targeting E and S viral genes. PRM1 and PRM2 mRNA, sperm-specific nuclear proteins, was used as the control for sperm RNA extraction, using RT-PCR (TaqMan™ Gene Expression Assay, Applied Biosystems). To assess whether viral RNA extraction was affected by the use of whole seminal fluid, a known titer of SARS-CoV-2 virus was added to semen from a healthy donor. The panel was prepared with serial dilutions of a SARS-CoV-2 isolate (named 2019-nCoV/Italy-INMI1) . The virus was collected by nasopharyngeal swab and cultured in Vero E6 cell lines grown in MEM containing 2% FBS. The dilutions ranged from 4 × 10 2 to 4 × 10 6 viral RNA copies/ml, corresponding to 0.1 and 1000 TCID50 (50% tissue culture infective dose)/ml. SARS-CoV-2 RNA was amplified by qRT-PCR and quantified based on a standard curve prepared through serial dilutions of EURM-019 single-stranded RNA (ssRNA) fragments of SARS-CoV-2 including different target genes ( https://crm.jrc.ec.europa.eu/p/EURM-019 ). Viral titers were determined by limiting dilution assay on Vero E6 cells and infectivity was expressed as TCID50/ml, calculated according to the Reed and Muench method. A blank containing only cell culture medium was included in the panel. Two known titer viral preparations from the panel were diluted 1:2 in seminal fluid and in 0.5% saline solution (Fig. ). The final concentrations of the tested samples were 2 × 10 4 and 2 × 10 6 copies/ml. Patient information is summarized in Fig. and Table . The semen sample after a positive nasopharyngeal test was obtained from pt#1 on the same day of the positive nasopharyngeal swab was performed, and from pt #2 within 48 h of the last positive swab. All tested semen samples (both COVID-19 patients and controls) were negative for SARS-CoV-2 RNA, regardless of the nasopharyngeal swab result. To investigate the presence of the virus in different fractions of seminal fluid, we also extracted RNA from whole samples, seminal plasma and post-centrifugation pellets containing only the corpuscular part of the seminal fluid, namely spermatozoa, germ cells, leukocytes and epithelial cells. We did not detect the virus in any of these fractions. Internal control was detected in all samples. Semen parameters of all recruited subjects are shown in Table . RNA extraction To verify the efficiency of RNA extraction from sperm we used PRM1 and PRM2 mRNA as the control. Protamines 1 or 2 are the most abundant and specific nuclear proteins in human sperm . PRM1 and PRM2 mRNA expression was found in all the semen samples (data not shown), demonstrating a good extraction capacity from this matrix. Control panel The assays were evaluated against a panel using negative control samples and 0.5% saline solution. We tested diluted controls infected with a known titer of SARS-CoV-2 virus. This was required to assess if any substances in the seminal fluid might interfere with viral RNA extraction, inducing false negatives or false positives. Two known titer viral preparations from a panel were diluted 1:2 in seminal fluid (from SARS COV2-negative patients) and in 0.5% saline solution. All samples obtained by diluting viral preparation from the panel tested positive for SARS-CoV-2, with no RT-PCR inhibition detected. The final virus concentrations were 2 × 10 4 and 2 × 10 6 copies/ml. Viral preparations diluted in saline and seminal fluid showed a similar Ct value to the initial viral concentration, as shown in Table . To verify the efficiency of RNA extraction from sperm we used PRM1 and PRM2 mRNA as the control. Protamines 1 or 2 are the most abundant and specific nuclear proteins in human sperm . PRM1 and PRM2 mRNA expression was found in all the semen samples (data not shown), demonstrating a good extraction capacity from this matrix. The assays were evaluated against a panel using negative control samples and 0.5% saline solution. We tested diluted controls infected with a known titer of SARS-CoV-2 virus. This was required to assess if any substances in the seminal fluid might interfere with viral RNA extraction, inducing false negatives or false positives. Two known titer viral preparations from a panel were diluted 1:2 in seminal fluid (from SARS COV2-negative patients) and in 0.5% saline solution. All samples obtained by diluting viral preparation from the panel tested positive for SARS-CoV-2, with no RT-PCR inhibition detected. The final virus concentrations were 2 × 10 4 and 2 × 10 6 copies/ml. Viral preparations diluted in saline and seminal fluid showed a similar Ct value to the initial viral concentration, as shown in Table . The outbreak of coronavirus disease (COVID-19) caused by SARS-CoV-2 has raised a number of concerns about public health, including sex-related mortality . Epidemiological studies suggested that males are more likely to test positive for COVID-19 . This has prompted questions about the possible repercussions of SARS-CoV-2 for the male reproductive system. SARS-CoV-2 enters cells by means of a viral receptor, angiotensin-converting enzyme 2 (ACE2), which is highly expressed in a wide range of human tissues. In the testis, ACE2 expression has been found on seminiferous duct cells, spermatogonia, and Leydig and Sertoli cells, confirming the potential risks to the reproductive system associated with SARS-CoV-2 infection . SARS-CoV-2 also requires transmembrane protease serine 2 (TMPRSS2) to enter cells. This proteolytic enzyme is involved in numerous physiological processes . TMPRSS2 cleaves and modifies spike protein, enabling the permanent fusion of the virus and host cell . It is highly expressed in the prostate epithelial cells, and its expression is regulated by androgens. The question thus arises: can SARS-CoV-2 reach the seminal fluid? Several authors have investigated the presence of SARS-CoV-2 in semen (Table ). They all conducted a search for viral RNA through RT-PCR, albeit screening for different genes. It must be stressed that of the 15 publications to date that have investigated this aspect, only 1 reported finding viral RNA in semen from both acute (26.7%) and recovering (8.7%) patients . Furthermore, SARS-CoV-2 has currently only been investigated in semen in 31 acute COVID-19 cases and relatively few recovering subjects, including the aforementioned study. Overall, only 4 acute and 3 recovered patients have been reported to have seminal fluid positive for viral RNA over a total of 341 subjects evaluated (Table ). Since most positive subjects came from the same study, the peculiar clinical conditions (disease severity) and methodological weaknesses of this paper have been discussed . Recently, another paper reported a SARS-CoV-2-positive seminal fluid in a caseload of 15 mild-asymptomatic subjects, however, the presented data are scant . However, further factors influencing the heterogeneity of these papers should also be recognized, including different ethnicities, slightly different definitions for acute cases, and huge differences in timing for the testing of recovering cases. On the hypothesis that the virus sheds into semen, all these factors could greatly affect both viral load and viral clearance in semen, and hence the chance of its detection. Consequently, although the presence of SARS-CoV-2 in semen cannot yet be completely excluded, the available data may be interpreted cautiously, but optimistically—especially given the absence of solid proof of its presence in the testes of non-severe COVID-19 cases . Gonzales et al. reviewed literature data on the presence of SARS-CoV-2 in semen. They found a very low risk in seminal fluid, and a negligible risk in recovered men . These results suggest that the likely absence of SARS-CoV-2 in seminal fluid may be influenced by biological or methodological factors. In relation to biological factors, we know that the testicles may be vulnerable to SARS-CoV-2 infection. However, given the concentration of SARS-CoV-2 receptors present in testicular tissue, why is the infection not clinically evident in the testes ? Studies based on single-cell RNA sequencing (sc RNAseq) in humans did not find any ACE2/TMPRSS2 co-expression in any type of testicular tissue . In theory, viruses could reach semen from the blood, as the blood–testis barrier does not seem to constitute an insurmountable obstacle to viruses in the presence of systemic or local inflammation . To date, few studies have investigated the presence of SARS-CoV-2 in blood. Bwire et al. reported a low (1.0%) detection of SARS‐CoV‐2 in blood samples . It could be that the virus only spreads to blood under certain circumstances, such as the acute phase or severe disease, and then to other organs such as the testis . Methodological factors are also important. While qRT-PCR assay, as discussed above, is the first-line screening method of choice for SARS-CoV-2 detection due to its high sensitivity and rapid detection , there is a real risk of false negative and false positive results . False negative results may be due to sample inhibitors, poor amplification efficiency, and reduced precision in low concentration samples. False positives could arise from contaminants or poor test specificity . In any case, the sensitivity and specificity of the RT-PCR methods used to detect SARS-CoV-2 in seminal fluid have not been evaluated . In our study, we investigated the presence of SARS-COV-2 in seminal fluid from four COVID-19 patients: two mild cases with a positive recent nasopharyngeal swab and two whose last swab was negative. We did not find SARS-CoV-2 RNA in any of these samples. Semen analyses from the positive patients showed some abnormalities; specifically, patient #1 was azoospermic and patient #2 asthenozoospermic. It should be stressed that these semen characteristics are likely to be due to their medical history: patient #1 had undergone chemotherapy for lymphoma, while the asthenozoospermia of patient #2 was probably caused by his clinical condition, its treatment, and prolonged abstinence. For a qualitative determination of the RT-PCR assays in semen we performed different attempts: (1) we verified the efficiency of RNA extraction from sperm and seminal plasma using PRM1 and PRM2 mRNA and a heterologous system, respectively, as control; (2) we tested samples obtained by diluting viral preparation from a panel tested positive for SARS-CoV-2, with no RT-PCR inhibition detected; (3) we investigated the presence of the virus in different fractions of seminal fluids, whole samples, seminal plasma and post-centrifugation pellets containing only the corpuscular part of the seminal fluid. We did not detect the virus in any of these fractions. Our study not only demonstrated the absence of SARS COV2 in the seminal fluid of patients in the acute phase with a positive nasopharyngeal swab and in recovered patients with a negative swab, but for the first time confirmed the feasibility of this test for the molecular diagnosis of SARS-CoV-2 in seminal fluid. This result is important in two ways. First, it confirms the literature data on the absence of the virus in seminal fluid in patients with mild COVID-19, and second, it verifies the molecular method used through various tests. This information is important for reproductive medicine, especially in assisted reproductive technology and sperm cryopreservation. The limitation of this method in relation to seminal fluid is that contamination could lead to a false positive. It should be stressed that semen collection is not sterile, and the sample could be contaminated with respiratory droplets or other body fluids from the patient, or by the patient’s hands. For this reason, any positive test result should be confirmed by repeating the test, alongside an evaluation of the patient’s symptoms and a thorough andrological history. In our opinion, the molecular diagnosis of SARS COV2 in seminal fluid could be a useful tool for specialists in reproductive medicine to evaluate the safety of sperm. |
A survey of perceptions of exposure to new technology in residents and practicing ophthalmologists | d4134a75-8ced-46ed-b4b4-a494993707b6 | 10976830 | Ophthalmology[mh] | Over the last several decades, advances in technology have allowed for the continued improvement and refinement of ophthalmic surgery . The plethora of new surgical technologies and procedures, although promising for the field, poses a challenge for residency training. The influx of new technologies into operating rooms requires surgical educators to provide trainee supervision and graduated independence across a wider range of techniques. Establishing a balance between teaching traditional techniques and exposure to new technologies requires a thoughtful approach as well as a comprehensive understanding of how the exposure to new technology during residency influences future practice patterns . Implementation of new technology in residency education has already become a significant consideration in cataract surgery , with integration of phacoemulsification and even femtosecond laser-assisted cataract surgery techniques into training . The reported challenges and benefits associated with incorporating such newer techniques only further reinforce the importance of considering exposure to other new technologies during resident surgical training. Although the Accreditation Council for Graduate Medical Education (ACGME) requires residents to perform a certain number of surgical procedures in predetermined areas, there are no guidelines for teaching new technologies or devices . This may contribute to the challenges in transitioning from residency training to independent surgical practice . In addition, the lack of specificity of the ACGME requirements provides flexibility in how each program incorporates new technology into surgical training. There may also be significant variation in the exposure residents get to novel technologies at different affiliate hospitals within each residency training program. Although potentially integral to the training of adaptable and innovative surgeons, the difference in exposure to new technologies within a residency training program and how that influences early practice has not yet been explored. While programs may examine overall resident satisfaction and case volume at different hospitals within each institution , these questionnaires rarely include questions on satisfaction with exposure to particular technologies or techniques. This study sought to examine resident and attending perceptions of exposure to new technology during residency and how exposure influences practice patterns and adoption rates in practice. Study enrollment and data collection This study was conducted in accordance with the tenets stated in the Declaration of Helsinki. Prior to beginning the study, approval was obtained for all protocols from the University of California, San Francisco Institutional Review Board. Informed consent was obtained by all participants prior to agreeing to complete the survey. In order to elicit the perception on exposure to residency during training and in practice, a survey study was conducted. With the aim of accessing a large diverse cohort of academic and private community physicians, the authors partnered with Market Scope (Des Peres, MO) to distribute the survey to attending physicians. In order to source US ophthalmologists, Market Scope invited US ophthalmologists who self-registered and were individually verified on market-scope.com ( n = 723) to participate in the survey, eliciting 69 total unique responses (9.54% response rate). Data was collected between December 5th 2022 and June 15th 2023. United States ophthalmology residents in all years of training and practicing ophthalmologists in the United States who completed training were also included. In order to source US ophthalmology residents, program directors and coordinators from all ACGME ophthalmology residency programs were emailed with the request to forward the survey to all their residents. Ophthalmology residents were then sent a link to the survey to participate from their program directors and/or program coordinators. Only 17 total unique responses were obtained, therefore YoungMD Connect (Bryn Mawr Communications, Conshohocken, PA) was recruited to help increase resident response numbers by reaching out to residents through a variety of methods (QR code at academic meetings, social media, email blasts), with an additional 46 resident responses amounting to a total of 63 unique responses. The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Survey answers were de-identified prior to any data sharing or analysis, and data was protected through the Market Scope authorized data management system. Incomplete questionnaires were not registered or included in this study. Survey and analysis A survey was developed to assess exposure to various technologies in residency and in practice. Questions were focused on, but not limited to, soliciting feedback on practice or future practice size and structure, training received in residency, Minimally Invasive Glaucoma Surgery (MIGS) and premium intraocular lens (IOL) volumes, and current or planned use of various technologies. (Supplemental Fig. ) Questions included year of practice, residency program (for geographic differentiation), practice setting (e.g. private practice, corporate practice, public hospital, academic center, etc.), intent to pursue fellowship, specialty, exposure to particular devices/procedures (Femtosecond Laser assisted Cataract Surgery (FLACs), digital surgical planning software, image management software/ Picture Archiving and Communication Systems (PACS), heads up microscope display, dry eye procedure device, premium IOLS, MIGS devices, presbyopia drops, sustained release drug options, etc.), perception of exposure to industry partnerships and newer technology during residency, and planned and current practice patterns for the above technologies during attending practice. (Fig. ) Questions were primarily formatted with structured answer options in multiple choice and check relevant answer format. There was no existing peer-reviewed framework, therefore the questions were created for the purpose of this study and validated across a cohort of physicians and Marketscope team with extensive experience in developing user-friendly surveys. A summary analysis of the reported data was performed without any additional extrapolation to the overall market. This study was conducted in accordance with the tenets stated in the Declaration of Helsinki. Prior to beginning the study, approval was obtained for all protocols from the University of California, San Francisco Institutional Review Board. Informed consent was obtained by all participants prior to agreeing to complete the survey. In order to elicit the perception on exposure to residency during training and in practice, a survey study was conducted. With the aim of accessing a large diverse cohort of academic and private community physicians, the authors partnered with Market Scope (Des Peres, MO) to distribute the survey to attending physicians. In order to source US ophthalmologists, Market Scope invited US ophthalmologists who self-registered and were individually verified on market-scope.com ( n = 723) to participate in the survey, eliciting 69 total unique responses (9.54% response rate). Data was collected between December 5th 2022 and June 15th 2023. United States ophthalmology residents in all years of training and practicing ophthalmologists in the United States who completed training were also included. In order to source US ophthalmology residents, program directors and coordinators from all ACGME ophthalmology residency programs were emailed with the request to forward the survey to all their residents. Ophthalmology residents were then sent a link to the survey to participate from their program directors and/or program coordinators. Only 17 total unique responses were obtained, therefore YoungMD Connect (Bryn Mawr Communications, Conshohocken, PA) was recruited to help increase resident response numbers by reaching out to residents through a variety of methods (QR code at academic meetings, social media, email blasts), with an additional 46 resident responses amounting to a total of 63 unique responses. The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Survey answers were de-identified prior to any data sharing or analysis, and data was protected through the Market Scope authorized data management system. Incomplete questionnaires were not registered or included in this study. A survey was developed to assess exposure to various technologies in residency and in practice. Questions were focused on, but not limited to, soliciting feedback on practice or future practice size and structure, training received in residency, Minimally Invasive Glaucoma Surgery (MIGS) and premium intraocular lens (IOL) volumes, and current or planned use of various technologies. (Supplemental Fig. ) Questions included year of practice, residency program (for geographic differentiation), practice setting (e.g. private practice, corporate practice, public hospital, academic center, etc.), intent to pursue fellowship, specialty, exposure to particular devices/procedures (Femtosecond Laser assisted Cataract Surgery (FLACs), digital surgical planning software, image management software/ Picture Archiving and Communication Systems (PACS), heads up microscope display, dry eye procedure device, premium IOLS, MIGS devices, presbyopia drops, sustained release drug options, etc.), perception of exposure to industry partnerships and newer technology during residency, and planned and current practice patterns for the above technologies during attending practice. (Fig. ) Questions were primarily formatted with structured answer options in multiple choice and check relevant answer format. There was no existing peer-reviewed framework, therefore the questions were created for the purpose of this study and validated across a cohort of physicians and Marketscope team with extensive experience in developing user-friendly surveys. A summary analysis of the reported data was performed without any additional extrapolation to the overall market. The survey was completed by 63 physician residents in a US ophthalmology program (across 42 unique programs, comprising 12.4% of the 509 US-based ophthalmology residents) and 69 physicians (9.5% of total contacted) beyond residency training between December 2022 and June 2023 (who graduated from 52 unique programs). A 36.2% ( n = 25) majority of the practicing ophthalmologists completed residency in the 1990s. The vast majority reported being part of a group practice (76.8%, n = 53) (Table ). 27% ( n = 17) of resident respondents were on track to complete residency in 2026, 23.8% ( n = 15) in 2025, 25.4% ( n = 16) in 2024, 15.9% ( n = 10) in 2023, and 6.3% ( n = 4) in 2022. Of the practicing ophthalmologists, 81.2% ( n = 56) were currently offering premium IOLs, and 68.1% ( n = 47) were offering MIGS. When asked about exposure in residency, 26.1% ( n = 18) reported being trained on premium IOLs in and 17.4% ( n = 12) reported being trained on MIGS. When asked about the level of discussion/training participants received in their residency program on diversity of brands and manufacturers available for product selection, 56.5% ( n = 39) percent of practicing ophthalmologists responded that this was “minimally discussed but not emphasized” or “not discussed at all.” When asked about the level of discussion/training participants received in their residency program on newly developed products on the market (premium IOLs, MIGS, etc.), 65.2% ( n = 45) responded that this was “minimally discussed but not emphasized” or “not discussed at all.” Practicing ophthalmologists were asked to rate their perception of exposure to newer surgical and therapeutic treatments and technologies in their own residency as compared to other programs. A total of 40.6% ( n = 28) of respondents reported more exposure than other programs, 50.7% ( n = 35) same exposure, and 8.7% ( n = 6) less exposure when compared to other programs. The majority (55.1%, n = 38) of respondents reported that exposure to new technologies during residency did influence types of technologies employed during practice, whereas 21.7% ( n = 15) reported that exposure did not, and 23.2% ( n = 16) were uncertain. Written commentary was elicited on whether respondents believed experience with technology in residency influenced the types of technology used or quantity of certain technologies in practice (Question 11 a.1, Supplementary Fig. ). Given the low sample size, thematic analysis was performed. Those who reported minimal exposure during residency reported that less exposure made them less willing to try new technology and made them inappropriately think there were not more than one or two options for technology companies to work with. Such exemplary comments included “Less exposure makes you less willing to try new tech”, “You use what you get experience with in residency. If you do not have exposure to newer technologies, you will not feel comfortable using them.” “One brands grip on the VA system made my inappropriately think there were no other options. It was stifling.” Those who reported same exposure and more exposure during residency when compared to other training programs provided commentary of a similar sentiment. Themes were extracted from individual comments. Respondents reported that exposure to new technologies and pharmaceuticals during residency helped define future habits and practice patterns and made them feel more adaptable and comfortable with a wide variety of techniques (e.g. various MIGS platforms and different IOLS). They also reported that greater exposure made them more open to test out new phacoemulsification machines and equipment and made them more incentivized to familiarize themselves with new cutting edge technologies. Finally, they reported that greater exposure helped them to be more adaptable surgeons with greater confidence in adopting a wide range of surgical techniques. Exemplary comments included; “Using multiple types of technology and wetlabs with industry has allowed me to be comfortable using them when in practice”, “Using them in residency allowed familiarity with it and allowed me to be less inhibited about trying other technologies”, “During residency, we were exposed to the latest phacoemulsification technology and were able to use different phaco platforms. This allowed me to decide which specific phaco platform that I wanted to use in private practice” and “We were privileged to use newer instruments and the latest surgical techniques. This experience taught me to stay ahead with these and advance patient care.” Among resident respondents, 87.3% ( n = 55) planned on pursuing a fellowship, and almost 39.7% ( n = 25) planned to eventually join a private practice. 63.5% ( n = 40) of respondents planned to treat cataracts, 46% ( n = 29) refractive surgery, 38.1% ( n = 24) retina, 38.1% ( n = 24) cornea/external disease, 36.5% ( n = 23) glaucoma, and 19% ( n = 12) uveitis. When asked about their perception of exposure to new technology in their own residency program versus other programs, 52.4% ( n = 33) of residents believed that they received the same amount of exposure to newer surgical and therapeutic treatments and technologies in residency, as compared to other programs. 17.5% ( n = 11) thought they received less exposure, 14.3% ( n = 9) thought they received more exposure, and 15.9% ( n = 10) percent were uncertain. When asked about partnerships with industry in terms of training and collaboration, 34.9% ( n = 22) reported good or very good exposure and 22.2% ( n = 14) reported poor or very poor exposure, with the rest reporting average exposure. A total of 46% ( n = 29) reported good or very good training and availability of newer technologies and 12.7% ( n = 8) reported poor or very poor exposure, with the rest reporting average exposure. 82.5% ( n = 52) percent had been trained on MIGS, and 69.8% ( n = 44) were trained on premium IOLs. When asked about the level of discussion/training participants received in their residency program on diversity of brands and manufacturers available for product selection, 55.6% ( n = 35) responded that this “minimally discussed but not emphasized” or “not discussed at all.” When asked about the level of discussion/training participants received in their residency program on newly developed products on the market (premium IOLs, MIGS, etc.), 47.6% ( n = 30) responded that this was “minimally discussed but not emphasized” or “not discussed at all.” Only 4.8% ( n = 3) reported prioritized discussion/training on diversity of brands and manufacturers available for product selection and 6.3% ( n = 4) on newly developed products on the market (premium IOLS, MIGs, etc.). Responses regarding specific new technologies practicing physicians and current residents have had access to in residency are demonstrated in Fig. . Compared to practicing attendings, a greater percent of resident respondents reported training on or access to premium IOLS, MIGS, imaging managing software/PACS, sustained release drug options/implants, dry eye procedures, FLACS, presbyopia drops, and heads up microscope display (Fig. ). Resident reported plans to offer certain technologies demonstrated a similar trend to the technologies residents reported training in during residency (Fig. ). Overall, the vast majority resident physicians reported that they enjoyed being trained on newer technology and exposure made them feel more prepared for future changes in the field (95.2%, n = 60). They also reported that having industry partnerships in residency enhances education and training (90.5%, n = 57), and that they were more likely to seek out employment opportunities that value advanced technology because of exposure during residency (81%, n = 51). ( Fig. ) Only 12.7% ( n = 8) strongly agreed that they would prefer to focus on the standard procedures and technology they are most likely to use in practice and to increase their comfort level ( Fig. ). This study explores reported exposure to new technology in residency and perceptions of how this exposure affects and/or will affect practice patterns among current ophthalmology residents and practicing ophthalmologists in the United States. In their responses, practicing clinicians emphasized the importance of exposure to innovative technologies in residency and the majority believed exposure to new technology in residency influenced their current practice. Despite this, over half responded that discussion or training on newly developed products on the market (premium IOLS, MIGS, etc.) was minimally discussed or not discussed at all. Similarly, despite increased reported exposure to new technologies such as MIGS, premium IOLS, sustained release drug options, over half of current residents still reported that diversity of brands and manufacturers available for product selection and newly developed products on the market were minimally discussed or not discussed at all. Combined, these findings suggest that residency programs are not perceived by trainees and graduates to have adapted to incorporate more comprehensive training newer technologies. This survey also demonstrates that residents are seeking greater training on newer technology as it makes them feel more prepared for what comes next, enhances education and training, and encourages them to further seek out advanced technology. This sentiment is supported by the corresponding responses from practicing ophthalmologists. Over half of practicing clinicians reporting that exposure and experience with technology in residency influenced the types of technology and quantity used in practice. The written commentary elicited further sheds light on how early exposure in residency impacts the adoption of new approaches in practicing. Those with increased exposure during residency expressed a greater willingness to try new technologies and exhibit adaptability in adopting a wide range of surgical techniques. On the other hand, respondents with minimal exposure during residency felt less comfortable exploring new technologies, potentially hindering adoption of advances that improve patient care. Developing a structured approach to incorporating newer technologies/devices during residency training may contribute to training more confident, adaptable surgeons. This also may have the benefit of teaching residents to critically consider new technologies and adopt promising ones into their future clinical practice. The challenges of incorporating new technologies into residency training has certainly been demonstrated in ophthalmology , and is present throughout all surgical fields . Furthermore, both residents and programs hiring new graduates have perceived gaps in readiness of residency graduates for independent practice . With changing regulations and innovations, ophthalmology residents have reported feeling unprepared in business operations, finance, practice management, coding, advocacy . Despite the importance placed on emerging technologies as a mechanism to support mentoring, precepting, and proctoring to improve transition from residency to independent practice, iterations in ACGME accreditation have been limited in their ability to foster innovation . In light of this, ACGME focus groups have recognized that early experiences with new technology may serve as the basis for further exploration of innovative approaches . Given the challenges in striking a balance between training residents in traditional techniques and exposing them to new technologies, it is important to consider ways to incorporate exposure to new technologies without detracting from the standard ACGME case guidelines for traditional surgeries. For example, there has been concern that incorporating femtosecond laser-assisted cataract surgery into residency training may decrease manual proficiency in other parts of surgery performed by the laser, such as corneal incisions and capsulorhexis . Virtual reality simulators may prove beneficial to teach advanced new techniques within the bounds of academic residency training . Research blocks may also be utilized to develop rotations specifically focused on new technologies and techniques . Structured industry partnerships may be another solution to supplement resident education without detracting from proficiency of key surgical steps. In our survey, most respondents reported that industry partnerships in residency enhance their education and training, indicating that such collaborations can play a vital role in providing access to the latest technologies and advancements in the field. Therefore, in instances where residencies feel limited in how much exposure they can provide within the structure of a three-year program, industry partnerships may be able to fill the gaps with wet labs, didactic training sessions, and master surgeon videos to teach new techniques. In addition, greater breadth of industry exposure (rather than usage of just one or two companies’ devices), may increase physician resident confidence in critically evaluating which products they would like to incorporate into their practice to optimize patient care. Similarly, in addition to academic meetings, exposure to industry largely facilitates continued learning and training after residency, therefore establishing these relationships earlier on in training may have implications for continued training throughout practice. This study must be considered in the context of its limitations. The small sample size of resident respondents (63 unique responses, 12.4% of US residents) may limit the generalizability of the findings. Additionally, the response rate from practicing ophthalmologists (69 unique responses) could be further improved to obtain more robust and representative results. The authors were unable to extract a view rate or participation rate, therefore the response rate is only representative of the completion rate of the online survey. This study likely suffers from response bias, as residents and practicing clinicians more interested in considering exposure to innovation and new technologies may have been more likely to complete the survey. Similarly, the study may be limited by hindsight and reporting bias, and only represents resident and practicing ophthalmologists perceived exposure during residency rather than actual exposure. In addition, given that this is a survey study, the authors may only provide reported response data, without making associations or conclusions about how exposure to new technology in residency directly affects physicians’ future practice patterns. This study also did not inquire about specific brand or manufacturer usage and exposure during residency. However, it is possible that exposure to particular brands or manufacturers during residency may influence practice patterns after residency just as it does general perception and willingness to incorporate newer technologies. In addition, further qualitative studies with more extensive commentary from the cohort of participants who reported being uncertain about the effect of exposure to newer technologies in residency may be revealing. Further research with larger sample sizes and longitudinal follow-up could provide more comprehensive data to support the conclusions and recommendations of this study. In addition, future studies would benefit from implementing a structured program to optimize exposure to new technologies, incorporating both residency program teaching and industry partnerships and prospectively measuring the impact on future practice patterns with comparisons to programs without structured innovation programs. In conclusion, this paper draws attention to the challenges and opportunities posed by the integration of new technologies into ophthalmology residency training. It underscores the need for standardization in training pathways, ensuring a comprehensive understanding of resident exposure to new technologies, and its potential impact on their future practice patterns. The findings highlight the importance of incorporating exposure to new technologies during residency to enhance residents’ adaptability, innovation, and confidence in adopting cutting-edge surgical techniques. The insights provided by this study can guide program directors and educators in designing effective residency training programs that prepare ophthalmologists for the evolving landscape of ophthalmic surgery. Below is the link to the electronic supplementary material. Supplementary Material 1 |
Comparison of two different contrast sensitivity devices in young adults with normal visual acuity with or without refractive surgery | 0e439ce2-bcb3-43d9-b054-0a3cb046e30e | 9334259 | Ophthalmology[mh] | Visual acuity is a commonly used method to check visual power, usually using the Snellen chart. It uses targets of very high contrast and measures spatial-resolving ability. However, Snellen acuity is a measure of only visual quantity and provides limited information about functional vision . On the other hand, contrast sensitivity quantifies the lightness or darkness needed to identify a target against its background . Measuring contrast sensitivity is just as important as visual acuity as it reflects the quality of vision and in many cases declines earlier, while visual acuity remains normal (6/6 or better) . Contrast sensitivity plays a role in many aspects of vision, specifically motion detection, visual field, pattern recognition, dark adaptation, and visual acuity , , and affects patients’ daily lives. Many conditions, including age , myopia , higher-order aberrations and neurologic degeneration including parkinsonism affect contrast sensitivity, just before any change in visual acuity is detected. Contrast sensitivity changes even in the early stages of cataract , age-related macular degeneration , open angle glaucoma , without necessarily affecting the visual acuity from such an early stage. Therefore, contrast sensitivity tests (CSTs) may help clinicians to understand a patient’s impairment and complaints in those with normal visual acuity , . Laser in situ keratomileusis (LASIK) and laser-assisted sub-epithelial keratectomy (LASEK) have gained widespread popularity as a method to correct refractive errors. Although refractive surgery can reduce refractive errors and improve uncorrected visual acuity, previous studies have shown that it can degrade the quality of vision, resulting in reports of reduced night vision clarity and contrast sensitivity yet some studies argue that this decline recovers within 3 or 6 months postoperative . Therefore, contrast sensitivity plays an important role in describing the visual function of patients post refractive surgery. However, only a few previous studies have focused on investigating the long-term course of contrast sensitivity after refractive surgery , and to our knowledge, there has been no previous study to compare two devices in patients post refractive surgery. The first CST devices used wall charts such as the Pelli-Robson chart, which measures contrast sensitivity using a single large letter size (20/60 optotype), with contrast varying across groups of letters . However, certain problems are inherent in paper charts. First, the charts fade over time, making the results less accurate. The lighting environment also affects the results. The chart must be illuminated, but without calibrating the room, ambient light from windows and neighboring exam lanes can alter the results of the test . These limitations led to the introduction of built-in charts, including the OPTEC-6500 (Stereo Optical, Chicago, IL), in which lighting, distance, and glare are standardized. These are manual CST devices that use pattern fringe stimuli that are presented manually and whose orientation must be identified manually. The CGT-2000 (Takagi Seiko Co., Ltd., Nagano-Ken, Japan) is an “automated” CST device, where lighting, distance, and glare are standardized as in other built-in CST devices but differs in that it is fully automatic. The presentation of the stimulus lasts for a certain period (0.2, 0.4, 0.8 or 1.6 s according to the examiner’s preference), as in the stimuli of automatic visual field perimetry devices, and its measurement is carried out using an automatic threshold recognition strategy. The automated CST device is a patient-driven, standardized test that can easily maintain examination conditions under control and eliminate technician bias . If they show comparable repeatability and reliability, automated CST devices can be useful instruments to measure contrast sensitivity quickly and easily for patients. In this study, we assessed the repeatability and correlation of a manual CST device (OPTEC-6500) and an automated CST device (CGT-2000) in healthy young adults with normal visual acuity with or without a history of myopic refractive surgery (LASIK or LASEK).
Study population This was a retrospective study of patients aged 20 to 39 years who underwent contrast sensitivity tests using both manual and automated CSTs from June 19 to July 24, 2021 at Severance Hospital, Yonsei University College of Medicine (n = 57). Subjects with the following were excluded: (i) those who received less than two examinations using either device (n = 8); (ii) those with corrected distance visual acuity under 20/20 (n = 2); (iii) those with a history of ocular surgery other than myopic LASIK or LASEK (n = 3); and (iv) contact lens wearers (n = 3) (Fig. ). All participants responded to a survey about their history of systemic and ocular disease, history of ocular surgery, age, sex, and subjective visual disturbance during the day and night. The ocular problems of the participants were confirmed through slit-lamp examination, and only participants with no anterior surface problems were enrolled. This study was approved by the Institutional Review Board of Yonsei University Health System (IRB no. 1-2021-0052), and informed consent was waived due to the retrospective nature of the study. All methods were performed in accordance with the relevant guidelines and regulations. Measurements Two CST devices, a manual CST (OPTEC-6500, Stereo Optical) and an automated CST (CGT-2000, Takagi Seiko Co., Ltd.), were used in this study. The manual CST was used in mesopic and photopic conditions with spatial frequencies of 1.5, 3, 6, 12, and 18 cycles per degree of visual angle (cpd) without glare. Monocular contrast sensitivity was measured at far (3 m) distance under photopic conditions at 85 cd/m 2 and mesopic conditions at 3 cd/m 2 with optimum refractive correction and a natural pupil, according to the manufacturer’s recommendation . The automated CST was also used for mesopic and photopic conditions with six target sizes (6.3°, 4.0°, 2.5°, 1.6°, 1.0°, and 0.64°) and 14 contrast levels (0.0071–0.64) without glare. Monocular contrast sensitivity was measured at far (5 m) distance under photopic at 100 cd/m 2 and mesopic at 5 cd/m 2 with optimum refractive correction and a natural pupil, according to the manufacturer’s recommendation . Two measurements were taken for each platform. Mesopic contrast sensitivity was tested first, and photopic contrast sensitivity was assessed after an interval of 30 min. The right eye was tested first in both mesopic and photopic contrast sensitivity. The CST device used first were randomly allocated. 48% (n = 20) of patients with history of refractive surgery and 50% (n = 20) of patients without history of refractive surgery were tested with manual CST first. Both CST were done on the same day with an interval of at least 1 h between the two sessions. The CST was repeated on another day with the same sequence. All psychophysical tests of this study were performed in the same testing laboratory where standardized lighting conditions were ensured by blocking daylight. The time taken to complete an examination was measured for both manual and automated CST. For manual CST, a stopwatch was started when the patient was shown the first row of targets and was stopped when the patient finished reading the last row of targets. For automated CST, a stopwatch was started as the start button was pressed and was stopped when the examination ended with a beep sound. After the examination, time taken for mesopic CST and photopic CST was added to calculate the time taken for each device. The time taken to explain the test method or register patient data to the device was not included in the examination time. For consistency of data, all examinations were performed by a single physician. Statistical analysis Descriptive statistics were used to characterize baseline characteristics and comorbidities. Continuous variables are expressed as mean ± standard deviation (SD), and categorical variables are reported as frequency (percentage). To examine the test–retest reliability of each method and the correlations between the two methods, the intraclass correlation coefficient (ICC) was calculated. The ICC of two CST measurements was calculated under absolute-agreement, 2-way random-effects model to calculate the interrater reliability . In concordance with a widely used guideline for reporting ICC by Koo et al. , ICC values less than 0.5, between 0.5 and 0.75, between 0.75 and 0.9, and greater than 0.90 were interpreted as poor, moderate, good, and excellent reliability, respectively, based on the 95% confidence interval of the ICC estimate. Area under the log contrast sensitivity formula (AULCSF) was calculated for each examination using the methods described previously . The difference between AULCSF in mesopic and photopic conditions was used to compare the quality of vision in different settings as in previous studies , . The Bonferroni method was applied to approve the statistical significance. All tests were two-tailed, with P < 0.05 considered statistically significant. Statistical analyses were conducted using SPSS (version 23.0; IBM, Armonk, NY) and MedCalc Statistical Software version 14.8.1 (MedCalc, Ostend, Belgium).
This was a retrospective study of patients aged 20 to 39 years who underwent contrast sensitivity tests using both manual and automated CSTs from June 19 to July 24, 2021 at Severance Hospital, Yonsei University College of Medicine (n = 57). Subjects with the following were excluded: (i) those who received less than two examinations using either device (n = 8); (ii) those with corrected distance visual acuity under 20/20 (n = 2); (iii) those with a history of ocular surgery other than myopic LASIK or LASEK (n = 3); and (iv) contact lens wearers (n = 3) (Fig. ). All participants responded to a survey about their history of systemic and ocular disease, history of ocular surgery, age, sex, and subjective visual disturbance during the day and night. The ocular problems of the participants were confirmed through slit-lamp examination, and only participants with no anterior surface problems were enrolled. This study was approved by the Institutional Review Board of Yonsei University Health System (IRB no. 1-2021-0052), and informed consent was waived due to the retrospective nature of the study. All methods were performed in accordance with the relevant guidelines and regulations.
Two CST devices, a manual CST (OPTEC-6500, Stereo Optical) and an automated CST (CGT-2000, Takagi Seiko Co., Ltd.), were used in this study. The manual CST was used in mesopic and photopic conditions with spatial frequencies of 1.5, 3, 6, 12, and 18 cycles per degree of visual angle (cpd) without glare. Monocular contrast sensitivity was measured at far (3 m) distance under photopic conditions at 85 cd/m 2 and mesopic conditions at 3 cd/m 2 with optimum refractive correction and a natural pupil, according to the manufacturer’s recommendation . The automated CST was also used for mesopic and photopic conditions with six target sizes (6.3°, 4.0°, 2.5°, 1.6°, 1.0°, and 0.64°) and 14 contrast levels (0.0071–0.64) without glare. Monocular contrast sensitivity was measured at far (5 m) distance under photopic at 100 cd/m 2 and mesopic at 5 cd/m 2 with optimum refractive correction and a natural pupil, according to the manufacturer’s recommendation . Two measurements were taken for each platform. Mesopic contrast sensitivity was tested first, and photopic contrast sensitivity was assessed after an interval of 30 min. The right eye was tested first in both mesopic and photopic contrast sensitivity. The CST device used first were randomly allocated. 48% (n = 20) of patients with history of refractive surgery and 50% (n = 20) of patients without history of refractive surgery were tested with manual CST first. Both CST were done on the same day with an interval of at least 1 h between the two sessions. The CST was repeated on another day with the same sequence. All psychophysical tests of this study were performed in the same testing laboratory where standardized lighting conditions were ensured by blocking daylight. The time taken to complete an examination was measured for both manual and automated CST. For manual CST, a stopwatch was started when the patient was shown the first row of targets and was stopped when the patient finished reading the last row of targets. For automated CST, a stopwatch was started as the start button was pressed and was stopped when the examination ended with a beep sound. After the examination, time taken for mesopic CST and photopic CST was added to calculate the time taken for each device. The time taken to explain the test method or register patient data to the device was not included in the examination time. For consistency of data, all examinations were performed by a single physician.
Descriptive statistics were used to characterize baseline characteristics and comorbidities. Continuous variables are expressed as mean ± standard deviation (SD), and categorical variables are reported as frequency (percentage). To examine the test–retest reliability of each method and the correlations between the two methods, the intraclass correlation coefficient (ICC) was calculated. The ICC of two CST measurements was calculated under absolute-agreement, 2-way random-effects model to calculate the interrater reliability . In concordance with a widely used guideline for reporting ICC by Koo et al. , ICC values less than 0.5, between 0.5 and 0.75, between 0.75 and 0.9, and greater than 0.90 were interpreted as poor, moderate, good, and excellent reliability, respectively, based on the 95% confidence interval of the ICC estimate. Area under the log contrast sensitivity formula (AULCSF) was calculated for each examination using the methods described previously . The difference between AULCSF in mesopic and photopic conditions was used to compare the quality of vision in different settings as in previous studies , . The Bonferroni method was applied to approve the statistical significance. All tests were two-tailed, with P < 0.05 considered statistically significant. Statistical analyses were conducted using SPSS (version 23.0; IBM, Armonk, NY) and MedCalc Statistical Software version 14.8.1 (MedCalc, Ostend, Belgium).
A total of 41 patients were analyzed in this study; 51.2% were female, and the mean age was 26.6 ± 3.7 years. In total, 164 measurements of 82 eyes (20 eyes with prior LASIK, 20 eyes with prior LASEK, and 42 eyes without a history of ocular surgery) were performed. The baseline characteristics of patients with or without a history of refractive surgery are presented in Table . Those who received refractive surgery were older (28.1 ± 4.4 years vs. 25.3 ± 2.2 years), were more frequently female (60% vs. 43%) and reported more decreased subjective night vision (55% vs. 38%, P = 0.03) compared with those who did not receive refractive surgery. Time to complete examinations The mean time taken to complete one full examination (including four steps; both right and left eyes in mesopic and photopic conditions) was significantly shorter in automated CST examinations compared with manual CST examinations (396.4 ± 20.4 s vs. 286.8 ± 2.3 s, respectively; P < 0.001) (Table ). Examination time was significantly shorter in the second examination than in the first examination using the manual CST (P = 0.01); however, the first examinations performed with the automated CST were even quicker than the second examinations taken with the manual CST (P < 0.001). Test repeatability The test–retest ICC analysis of both CST devices revealed that the manual CST showed good repeatability (i.e., ICC value of 0.75–0.90 based on the 95% confidence interval ) in both photopic and mesopic conditions, while the automated CST showed moderate repeatability (i.e., ICC value of 0.50–0.90 based on the 95% confidence interval ) in both mesopic and photopic conditions (Appendix in the Supplementary material). The ICC of the inter-test analysis was moderate in both photopic and mesopic conditions (Appendix in the Supplementary material). Ceiling and floor effects Appendix and in the Supplementary material show the proportion of eyes with a maximum or minimum score in examinations using the manual and automated CSTs according to the history of refractive surgery. The ceiling effect (i.e., obtaining the maximum score on the chart) and floor effect (i.e., obtaining the minimum score on the chart) were more prominent with the manual CST than with the automated CST in both patients with and without refractive surgery. There was also a floor effect in eyes with and without refractive surgery at the highest spatial frequency in automated CST examinations. Meanwhile, the floor effect in manual CST examinations occurred at both the highest and second-highest spatial frequencies in both patients with and without refractive surgery. Correlation with history of refractive surgery There was no significant difference between the AULCSF of patients who did or did not undergo refractive surgery in either examination (Table ). However, patients who underwent refractive surgery showed significantly larger AULCSF difference (calculated as photopic AULCSF − mesopic AULCSF) in automated CST examinations compared with patients without a history of ocular surgery (AULCSF difference 0.415 vs. 0.323 in patients with and without refractive surgery, respectively; P < 0.001). However, there was no significant difference between patients with and without refractive surgery in manual CST examinations (AULCSF difference 0.189 vs. 0.217 in patients with and without refractive surgery, P = 0.22) (Fig. ). Correlation with decreased subjective night vision Patients who reported decreased subjective night vision showed significantly larger AULCSF differences in automated CST examinations compared with those who did not report decreased subjective night vision (AULCSF difference 0.397 vs. 0.343 in patients who did and did not report decreased subjective night vision, respectively; P = 0.02). However, there was no significant difference between patients who did and did not report decreased subjective night vision in manual CST examinations (AULCSF difference 0.200 vs. 0.206 in patients who did and did not report decreased subjective night vision, respectively, P = 0.76) (Fig. ).
The mean time taken to complete one full examination (including four steps; both right and left eyes in mesopic and photopic conditions) was significantly shorter in automated CST examinations compared with manual CST examinations (396.4 ± 20.4 s vs. 286.8 ± 2.3 s, respectively; P < 0.001) (Table ). Examination time was significantly shorter in the second examination than in the first examination using the manual CST (P = 0.01); however, the first examinations performed with the automated CST were even quicker than the second examinations taken with the manual CST (P < 0.001).
The test–retest ICC analysis of both CST devices revealed that the manual CST showed good repeatability (i.e., ICC value of 0.75–0.90 based on the 95% confidence interval ) in both photopic and mesopic conditions, while the automated CST showed moderate repeatability (i.e., ICC value of 0.50–0.90 based on the 95% confidence interval ) in both mesopic and photopic conditions (Appendix in the Supplementary material). The ICC of the inter-test analysis was moderate in both photopic and mesopic conditions (Appendix in the Supplementary material).
Appendix and in the Supplementary material show the proportion of eyes with a maximum or minimum score in examinations using the manual and automated CSTs according to the history of refractive surgery. The ceiling effect (i.e., obtaining the maximum score on the chart) and floor effect (i.e., obtaining the minimum score on the chart) were more prominent with the manual CST than with the automated CST in both patients with and without refractive surgery. There was also a floor effect in eyes with and without refractive surgery at the highest spatial frequency in automated CST examinations. Meanwhile, the floor effect in manual CST examinations occurred at both the highest and second-highest spatial frequencies in both patients with and without refractive surgery.
There was no significant difference between the AULCSF of patients who did or did not undergo refractive surgery in either examination (Table ). However, patients who underwent refractive surgery showed significantly larger AULCSF difference (calculated as photopic AULCSF − mesopic AULCSF) in automated CST examinations compared with patients without a history of ocular surgery (AULCSF difference 0.415 vs. 0.323 in patients with and without refractive surgery, respectively; P < 0.001). However, there was no significant difference between patients with and without refractive surgery in manual CST examinations (AULCSF difference 0.189 vs. 0.217 in patients with and without refractive surgery, P = 0.22) (Fig. ).
Patients who reported decreased subjective night vision showed significantly larger AULCSF differences in automated CST examinations compared with those who did not report decreased subjective night vision (AULCSF difference 0.397 vs. 0.343 in patients who did and did not report decreased subjective night vision, respectively; P = 0.02). However, there was no significant difference between patients who did and did not report decreased subjective night vision in manual CST examinations (AULCSF difference 0.200 vs. 0.206 in patients who did and did not report decreased subjective night vision, respectively, P = 0.76) (Fig. ).
The present study revealed the following findings. First, in the ICC analysis, the manual CST performed better than the automated CST, but both showed moderate to good test repeatability and fair to good inter-test correlation. In addition, the ceiling and floor effects were both more prominent with the manual CST than with the automated CST in both patients with and without refractive surgery. These results show that the automated CST shows comparable repeatability and reliability compared to manual CST devices. Second, the mean time taken to complete one full examination was significantly shorter when using the automated CST compared with the manual CST. Contrast sensitivity relates the visibility of a spatial pattern to both its size and contrast and is therefore a more comprehensive assessment of visual function than visual acuity but can be more time-consuming . Therefore, efficiency has become an important part of CST as well as precision, and there have been many approaches reduce time to detect contrast sensitivity changes , . In this study, the mean time taken to complete one full examination was significantly shorter when using the automated CST compared with the manual CST. Unlike the manual CST, in which patients are given unlimited patient response time, the automated CST requires a timed response, which not only reduce the overall time taken for each test, but also reduce the bias due to exposure time to visual stimuli. Human vision requires certain contrast, size, and exposure time for an object to be detected , . Therefore, the time of exposure to a stimulus is a critical variable in assessing the patient’s contrast sensitivity, while the manual CST lose examining value because the time of the motor response is not controlled . Finally, patients who underwent refractive surgery showed significantly larger AULCSF differences in automated CST examinations compared with patients without a history of ocular surgery, while there was no significant difference in manual CST examinations. Patients who reported decreased subjective night vision showed significantly larger AULCSF differences in automated CST examinations compared with those who did not report decreased subjective night vision, while there was no significant difference in manual CST examinations. These results imply that compared with manual CST, the automated CST correlated well with subjective night vision decrease and had higher sensitivity to history of refractive surgery. This superiority is probably due to improvements in the methodology of the automated CST. First, previous CST devices made use of vertical linear gratings that were biased for with-the-rule astigmatism and horizontal coma. Astigmatism and higher-order aberrations (coma, trefoil, and tetrafoil) cause lines to appear darker (higher contrast) in one angular orientation than in the orthogonal orientation. In contrast, the automated CST device made use of bull’s-eye sine-wave gratings, which is a rotationally symmetric target. For two patients with the same magnitude of astigmatism but at different orientations, the target will appear the same, and no advantage or disadvantage will happen based on the orientation of the appearance of the target . Second, the automated CST allows a blanking period between the presentation of each stimulus, allowing for a more accurate test. Third, while the manual CST have a high probability of correct guessing with only 3 answer choices, and is more prone to false positives, the automated CST system reduced false positivity by randomly presenting the lowest contrast target . Most previous studies performed CSTs using only the manual CST vision testing system, with a background luminance of 3 cd/m 2 for mesopic conditions and 85 cd/m 2 for photopic conditions, while automated CST was performed under a background luminance of 5 cd/m 2 for mesopic and 100 cd/m 2 for photopic conditions, which may have resulted in more prominent differences between photopic and mesopic conditions. In addition, the manual CST is based on spatial frequencies of 1.5, 3, 6, 12, and 18 cpd, while the automated CST is conducted under visual angles of 6.3°, 4°, 2.5°, 1.6°, 1°, and 0.64°, deviating the test results to lower spatial frequency, equal to a larger visual angle . To our knowledge, no previous studies have compared the reproducibility and correlation of the manual and automated CSTs. While the manual CST is typically criticized for being prone to guessing and showing low repeatability under mesopic conditions , the automated CST is typically criticized for poor repeatability and ceiling effects in normal individuals . In this study, both showed moderate to good test repeatability and moderate inter-test correlation. In addition, compared with a manual CST, the automated CST showed 33.8% and 20.2% decreases in time in the first and second examinations, respectively. The ceiling and floor effects were also smaller with the automated CST. These findings suggest that an automated CST can be a good alternative to manual CST in young adults with normal visual acuity. The current literature is unclear regarding the long-term compromise of contrast sensitivity after refractive surgery . Some demonstrate only a temporary decrease in contrast sensitivity within a few months after refractive surgery , , while others demonstrate a persistent decrease in contrast sensitivity, especially in mesopic conditions and in intermediate to low cycles per degree , , . Increased ocular higher-order aberrations and the oblate shape of the cornea following refractive surgery are suggested to be the reason for compromised contrast sensitivity , , . The present study reveals the long-term outcome in contrast sensitivity of patients who underwent refractive surgery at least one year earlier (mean time 5.5 ± 3.5 [range 1–12] years). Eyes with prior LASIK or LASEK had similar contrast sensitivity to normal eyes in both photopic and mesopic conditions (all P > 0.05). However, the AULCSF difference was significantly higher in eyes with prior refractive surgery in automated CST examinations, while no significant difference was seen in manual CST examinations. In addition, patients who reported decreased subjective night vision showed significantly larger AULCSF differences in automated CST examinations compared with those who did not report decreased subjective night vision, while there was no significant difference in manual CST examinations. In this study, the automated CST showed higher sensitivity to the presence of a history of refractive surgery and had better correlations with subjective night vision decrease compared with manual CST. This study has several limitations. First, this retrospective study with a relatively small sample size is prone to higher variability and selection bias. Second, the patients enrolled in this study underwent refractive surgery from different hospitals, and the modality of refractive surgeries performed on each patient will vary. Finally, the effects of presbyopia were not determined. However, the age at onset of 40 years for presbyopia was widely used for Asian populations in previous epidemiologic studies , , and only subjects aged < 40 years were analyzed in this study. Another study reported the overall prevalence of functional presbyopia to be only 9.07% in Asian subjects aged 35–44 years . Despite these limitations, this study presents the first comparison of two contrast sensitivity devices that are widely used in the real world. Compared with a manual CST, the automated CST took less time to complete and correlated well with subjective night vision decrease in patients after refractive surgery. These findings suggest that an automated CST can be a good alternative to a manual CST, with limited repeatability but higher sensitivity to the presence of a history of refractive surgery and with better correlations with subjective night vision decrease.
Supplementary Information.
|
Ammonia leakage can underpin nitrogen-sharing among soil microorganisms | 4619292c-17fe-4769-930c-5a27e73799b6 | 11440039 | Microbiology[mh] | The vast array of soil microorganisms and their interactions, the “soil microbial community”, confer a myriad of functions in soils. These functions underpin the contribution of soil community to biogeochemical cycles . The latter contribution is mediated through microbial processing of nutrients needed by the plant, suppression of plant pathogens, enhanced stress tolerance (reviewed by ), and by modulation of plant hormone signalling . There is therefore a requirement to better understand soil microbial communities and the microbial interactions within, so to preserve and possibly increase crop/ soil fertility. Several studies have shown that abiotic environmental inputs influence soil and plant microbiome assembly and the topic had been reviewed . At the community level, nutrient availability has been shown to strongly influence community diversity and it is suggested that high nutrient availability promotes the proliferation of few competitive species and low nutrient availability promotes diversity . External nutrient supply, as well as micro-scale nutrient gradients, are therefore important factors in shifting the balance of metabolic microbial interactions, as shown in experiments with model systems . The environmental influences on community composition are likely mediated through direct effects on specific inter-microbial interactions, however, specific interactions among soil microorganisms are often not well-characterised. Generally speaking, interactions between microorganisms can take many forms from competitive to cooperative and can be mediated by a range of different mechanisms . Among these, metabolic interactions through cross-feeding and auxotrophy are shown to be wide-spread in many different environments, including the soil . Auxotrophic interactions occur when one organism loses ability to synthesize a growth-essential compound, and instead receives this from another organism . Among specific soil microorganisms, previous work has highlighted mutual growth promotion between the fungus Mortierella elongata and bacterium Burkholderia BT03, mediated in-part by fungal organic acids . An auxotrophic interaction has also previously been identified, via vitamin B1 (thiamine), between the bacterium Bacillus subtilis and the fungus Serendipit indica . This specific interaction is notable because S. indica is an important soil fungi believed to promote plant growth and stress tolerance across a range of plant species . The inability of S. indica to produce thiamine means that its beneficial activity towards plants is dependent on nutrient input from other soil organisms such as B. subtilis or the plants themselves. In addition to its inability to produce thiamine, S. indica has been shown previously to have impaired growth in media with nitrate as the only N-source, and lacks genes coding for nitrate transporters and, nitrate and nitrite reductases . Considering that nitrate is a major nitrogen source in global soils and agricultural practices , this raises a question about why fungi such as S. indica would lose their ability to assimilate such an abundant nitrogen source and how they sustain themselves in nitrate-dominant environments. One possible answer to these questions could be that alternative nitrogen sources, that are easier to utilise, readily co-existing with nitrate, in nitrate-rich environments due to activities of nitrate utilising bacteria. Here, we have explored this hypothesis using the B. subtilis – S. indica pair as a model bacteria – fungi interaction system. We reconfirmed the incapability of S. indica to use nitrate as a sole nitrogen source and demonstrated that this incapacity is ameliorated by the presence of B. subtilis . We showed that this effect is mainly due to ammonia, which we found to be released into the media by B. subtilis . This finding is supported by a mathematical model, showing that ammonia can be readily leaked by cells due to its pH dependent equilibrium with ammonium and the high permeability of the latter through cell membranes. Thus, our results show that inevitable leakage of ammonia can act as a nitrogen sharing mechanism among soil microorganisms. This highlights the possible importance of incidental leakage of membrane-permeable metabolites in the development of auxotrophic interactions and the maintenance of soil microbial community stability.
Growth media Here we use modified ATS media : 70 μ M H 3 BO 3 , 14 μ M MnCl 2 , 0.5 μ M CuSO 4 , 1 μ M ZnSO4, 0.2 μ M NaMoO 4 , 10 μ M NaCl, 0.01 μ M CoCl 2 , 2.5 mM KPO 4 (2.3 mM KH 2 PO 4 and 0.2 mM K 2 HPO 4 for pH 5.8), 3 mM MgSO 4 , 3 mM CaCl 2 , 50 μ M Fe-EDTA, 100 mM glucose, 500 nM thiamine. N-sources were 10 mM KNO 3 , 5 mM (NH 4 ) 2 SO 4 or 5 mM glutamine. Where needed 2% agarose was added. Modified BG11+ media composition was as follows 1.5 g/L NaNo 3 , 0.04 g/L K 2 HPO 4 ·3H 2 O, 0.02 g/L Na 2 CO 3 , 6 mg/L FeCl 3 ·6H 2 O, 3 mg/L H 3 BO 3 , 2 mg/L MnCl 2 ·4H 2 O, 0.39 mg/L Na 2 MoO 4 ·2H 2 O, 0.22 mg/L ZnSO 4 ·7H 2 O, 0.02 mg/L biotin, 0.02 mg/L folic acid, 0.1 mg/L pyridoxinehydrochloride, 0.05 mg/L thiamine hydrochloride, 0.05 mg/L riboflavin, 0.05 mg/L nicotinic acid, 0.05 mg/L D-calcium pantothenate, 0.05 mg/L para-aminobenzoic acid, 0.001 mg/L cabalamin, 0.05 mg/L lipoic acid, and 1% glucose. Microbial preparation S. indica stocks (−80°C 500000 spores/mL in 0.02% tween 20 ) were germinated on CM-agar plates (20 g/L glucose, 6 g/L NaNO 3 , 2 g/L peptone, 1 g/L casein hydrolysate, 1 g/L yeast extract, 1.52 g/L KH 2 PO 4 , 502 mg/L MgSO 4 ·7H 2 O, 502 mg/L KCl, 6 mg/L MnCl 2 ·4H 2 O, 2.65 mg/L ZnSO 4 ·H 2 O, 1.5 mg/L H 3 BO 3 , 0.75 mg/L KI, 0.13 mg/L CuSO 4 ·5H 2 O, 2.4 ng/L Na 2 MO 4 ·2H 2 O, 15 g/L agar) and matured for 4 weeks. Agar plugs were transferred to fresh CM-agar plates and matured for minimum 8 weeks. Spores were harvested by washing with 0.02% tween 20 solution and adjusting to 500 000 spores/mL using a Fuchs-Rosenthal haemocytometer. All bacterial strains: Pseudomonas composti (two strains) and Allorhizobium rhizophilum from and B. subtilis strains NCIB3610, 168 and 168∆amtB were maintained at −80°C in 25% glycerol. 168∆amtB was generated by . Glycerol scrapes were streaked on LB-agar (10 g/L Peptone, 5 g/L Yeast extract, 10 g/L NaCl, 1.5% Agar) plates and incubated at 30°C overnight. Single colonies from these plates were inoculated into 10 mL LB-broth for overnight cultures. After overnight growth the cells were spun at 12000 x g for 1 min, washed twice with 100 mM NaCl and adjusted to the desired optical density (OD 600 = 0.5 unless otherwise stated) for use as inoculum. Plate inoculation of S.Indica and B. Subtilis experiments For plating, 2 mL of 2% agarose ATS media in 3.5 cm Petri dishes was used as described above and N-sources were added at the concentrations indicated. To inoculate S. indica, 5 μ L of S. indica spore suspension was placed centrally onto plates for mono-culture experiments, plates were sealed with parafilm and allowed to grow for 3 weeks in the dark in static 30°C incubator. For co-culture experiments, the S. indica inoculum was slightly offset to one side to allow for 2 cm separation between organisms. These plates were placed in static 30°C incubator for 2 days prior to B. subtilis inoculation. After 2 days plates were opened and allowed to dry for around 1 hr in sterile conditions. For co-culture plates, 1 μ L of B. subtilis inoculum or 1 μ L 100 mM NaCl was placed on the opposite side of the plate. Plates were wrapped in parafilm and kept in the 30°C static incubator until imaging. Preparation of B. Subtilis supernatant To create supernatant, 200 mL ATS media with 10 mM KNO 3 was added to sterile 500 mL conical flasks and 1 mL of B. subtilis inoculum (prepared as described above) was added and cultures were placed in 30°C with 170 rpm shaking. For initial experiments considering a range of B. subtilis OD 600 supernatants, cultures were sampled at approximately 24, 28, 32, and 52 hrs into the experiment. For later experiments, cultures were sampled at 28 hrs and the exact OD 600 measured. The cultures were centrifuged at x3200g for 10mins followed by vacuum filter sterilisation of the supernatant through Corning PES filters 0.22 μ m for use in later experiments and analysis. S. Indica liquid culture experiments S. indica liquid cultures were grown in 250 mL conical flasks filled to a final volume of 100 mL ATS with the indicated N-sources, B. subtilis supernatant and 100 μ L spore inoculum. These were placed in shaking 30°C incubator 170 rpm for 1 week before sampling. On sampling cultures were passed through Miracloth (Merck-Millipore, Burlington, MA, USA) to filter mycelia from the growth supernatant. Supernatant was collected for metabolite analysis and mycelia were scraped from the Miracloth surface and placed in 2 mL Eppendorf tubes to dry for weight measurements. Tubes and samples were dried at 70°C for 1–2 days and then returned to room temperature for around 24 hours before weighing to equilibrate to ambient humidity. Weights were measured using a Sartorius Secura 124–15 balance by measuring the tubes plus fungal material, then removing the fungal material and weighing only the tubes and calculating the difference. Cross species ammonium quantification Bacterial inocula were prepared as described above to a density of OD 600 = 1 in 0.9% saline. To initiate cultures, 100 μ L of inoculum was added to 100 mL BG11+ media (listed above) in a 250 ml conical flask. These cultures were incubated shaking at 30°C and sampled twice daily. OD measurements were taken regularly until OD > 0.5 was reached. Supernatant was collected by centrifugation and filter sterilisation. Ammonium was quantified by using the Supelco Spectroquant ammonium test kit and absorbance measured at 700 nm using CLARIOstar BMG Labtech plate reader. Fluorescence pH quantification Fluorescence standard curve was constructed by preparing KPO 4 buffer at four different pH values by combining different ratios of 1 M KH 2 PO 4 :K 2 HPO 4 as follows: pH 4, 100:0 (adjusted down to pH 4 with 0.1 M HCl), pH 5.8 91.5:8.5, pH 6.8 50.3:49.7, pH 8 6:94. Propidium iodide (PI), for normalisation, and the pH sensitive 2′,7’-Bis-(2-Carboxyethyl)-5-(and-6)-Carboxyfluorescein (BCECF) (Invitrogen, Waltham, MA, USA) were added to these at a final concentration of 100 μ M and 10 μ M respectively. Solutions in 3.5 cm Petri dishes were imaged with bright field illumination with 0.5 ms exposure, HcRED1 filter set 41 043 (Chroma) exposure (100 ms) and EGFP filter set 41 018 (Chroma) exposure (100 ms). Light source was pE-300white (CoolLED). Standard curve for pH quantification is performed . Nitrate, ammonium and amino acid quantification Quantification of free amino acids and ammonium, plus total protein content (amino acid content post acid hydrolysis) was conducted externally by Genaxxon (Ulm, Germany) using HPLC (high performance liquid chromatography) LC3000 with post-column ninhydrin derivitisation at 125°C. For detection of protein-derived amino acids, samples were hydrolysed, prior to HPLC separation, in 6 N HCl at 110°C for 60 hours. For detection of nitrate ions we used the Dionex (Sunnyvale, CA, USA) ICS-5000 + ion-chomotography system. We used an anionic column with KOH eluent. 2.5 μ L of sample (filtered through 0.22 μ m syringe filter) was injected and run at a flow rate of 0.38 mL/min for 30 minutes in a constant gradient as follows: 0 mins 1.5 mM KOH, 8 mins 1.5 mM KOH, 18 mins 15 mM KOH, 23 mins 24 mM KOH, 24 mins 60 mM KOH, 30 mins 60 mM KOH. Nitrate was detected at an elution time of around 12.8 minutes using a conductivity detector. Standard curve for quantification is performed . Mathematical modelling We used a modified version of the model developed by to consider the dynamics of ammonia and ammonium between the interior and exterior of a cell growing in a nitrate-only and given pH environment. We defined differential equations for the external (Equation ) and internal (Equation ) ammonium concentrations. (1) [12pt]{minimal}
dN{H}_4\_}/ dt= kN- diff- NH4\_} k1 (2) [12pt]{minimal}
dN{H}_4\_}/ dt= kN\_ in+ diff+N{H}_4\_} k1-N{H}_4\_ in kbm (3) [12pt]{minimal}
diff=P\_ VA(k\_} NH4\_}-k\_ in NH4\_ in) Parameters kN and kN_ in represent the external influx (from cells into the external environment) and internal influx (from nitrate reduction) of ammonium. Parameter diff represents diffusion into or out of the cell from the exterior environment and is given by Equation where P_V A is the rate of ammonia diffusion from the cell and k_in and k_ext are the internal and external dissociation constants respectively for the ammonium-ammonia equilibrium at the pH of those compartments. The rates k1 and kbm represent the rates of ammonium import into the cell and ammonium incorporation into biomass respectively. For the parameter kN_in , we inferred an amount of ammonium per hour that must be generated through nitrate reduction. We considered in the experiments, that there was a reduction of approximately 2.5 mM nitrate over the course of the 28 hrs growth . Thereby indicating at least 90 μ M per hour (2500/28) of internal ammonium production at the population level. We solved differential equations for change in external ammonium concentration (Equation ) and internal ammonium concentration (Equation ) until simulations reached steady state. We check steady state has been reached by verifying that the external ammonium concentration is not changing by more than 10 −5 within experimental timeframes. Endpoint measurements were then used to generate surface plots. Python code for this analysis is available at: https://github.com/lukeZrich/RichardsNsharing2024 . Plate reader assessment of bacterial growth ATS broth was used for plate reader growth experiments with the following modifications. KPO 4 buffer composition was altered to increase the pH by increasing the ratio of K 2 HPO 4 to KH 2 PO 4 such that the final media contained 1.3 mM KH 2 PO 4 and 1.2 mM K 2 HPO 4 for pH 6.8. To combat precipitation, we reduced the concentration of MgSO 4 and CaCl 2 by ten fold for 0.3 mM each. B. subtilis inocula were prepared as described above and 1 μ L used to inculate 200 μ L of media per well. Cultures were grown in a plate reader (CLARIOstar BMG Labtech) for 3 days at 30°C, 200 rpm and OD 600 measurements taken every 30 minutes. BLAST searching of nitrogen assimilation genes homology Protein gene accession as described in were used to perform protein–protein BLAST against the S. indica and S. vermifera genetic information stored in the NCBI (TaxID 1 109 443 and 109 899, respectively). Default settings were used to run Blastp protein–protein Blast.
Here we use modified ATS media : 70 μ M H 3 BO 3 , 14 μ M MnCl 2 , 0.5 μ M CuSO 4 , 1 μ M ZnSO4, 0.2 μ M NaMoO 4 , 10 μ M NaCl, 0.01 μ M CoCl 2 , 2.5 mM KPO 4 (2.3 mM KH 2 PO 4 and 0.2 mM K 2 HPO 4 for pH 5.8), 3 mM MgSO 4 , 3 mM CaCl 2 , 50 μ M Fe-EDTA, 100 mM glucose, 500 nM thiamine. N-sources were 10 mM KNO 3 , 5 mM (NH 4 ) 2 SO 4 or 5 mM glutamine. Where needed 2% agarose was added. Modified BG11+ media composition was as follows 1.5 g/L NaNo 3 , 0.04 g/L K 2 HPO 4 ·3H 2 O, 0.02 g/L Na 2 CO 3 , 6 mg/L FeCl 3 ·6H 2 O, 3 mg/L H 3 BO 3 , 2 mg/L MnCl 2 ·4H 2 O, 0.39 mg/L Na 2 MoO 4 ·2H 2 O, 0.22 mg/L ZnSO 4 ·7H 2 O, 0.02 mg/L biotin, 0.02 mg/L folic acid, 0.1 mg/L pyridoxinehydrochloride, 0.05 mg/L thiamine hydrochloride, 0.05 mg/L riboflavin, 0.05 mg/L nicotinic acid, 0.05 mg/L D-calcium pantothenate, 0.05 mg/L para-aminobenzoic acid, 0.001 mg/L cabalamin, 0.05 mg/L lipoic acid, and 1% glucose.
S. indica stocks (−80°C 500000 spores/mL in 0.02% tween 20 ) were germinated on CM-agar plates (20 g/L glucose, 6 g/L NaNO 3 , 2 g/L peptone, 1 g/L casein hydrolysate, 1 g/L yeast extract, 1.52 g/L KH 2 PO 4 , 502 mg/L MgSO 4 ·7H 2 O, 502 mg/L KCl, 6 mg/L MnCl 2 ·4H 2 O, 2.65 mg/L ZnSO 4 ·H 2 O, 1.5 mg/L H 3 BO 3 , 0.75 mg/L KI, 0.13 mg/L CuSO 4 ·5H 2 O, 2.4 ng/L Na 2 MO 4 ·2H 2 O, 15 g/L agar) and matured for 4 weeks. Agar plugs were transferred to fresh CM-agar plates and matured for minimum 8 weeks. Spores were harvested by washing with 0.02% tween 20 solution and adjusting to 500 000 spores/mL using a Fuchs-Rosenthal haemocytometer. All bacterial strains: Pseudomonas composti (two strains) and Allorhizobium rhizophilum from and B. subtilis strains NCIB3610, 168 and 168∆amtB were maintained at −80°C in 25% glycerol. 168∆amtB was generated by . Glycerol scrapes were streaked on LB-agar (10 g/L Peptone, 5 g/L Yeast extract, 10 g/L NaCl, 1.5% Agar) plates and incubated at 30°C overnight. Single colonies from these plates were inoculated into 10 mL LB-broth for overnight cultures. After overnight growth the cells were spun at 12000 x g for 1 min, washed twice with 100 mM NaCl and adjusted to the desired optical density (OD 600 = 0.5 unless otherwise stated) for use as inoculum.
For plating, 2 mL of 2% agarose ATS media in 3.5 cm Petri dishes was used as described above and N-sources were added at the concentrations indicated. To inoculate S. indica, 5 μ L of S. indica spore suspension was placed centrally onto plates for mono-culture experiments, plates were sealed with parafilm and allowed to grow for 3 weeks in the dark in static 30°C incubator. For co-culture experiments, the S. indica inoculum was slightly offset to one side to allow for 2 cm separation between organisms. These plates were placed in static 30°C incubator for 2 days prior to B. subtilis inoculation. After 2 days plates were opened and allowed to dry for around 1 hr in sterile conditions. For co-culture plates, 1 μ L of B. subtilis inoculum or 1 μ L 100 mM NaCl was placed on the opposite side of the plate. Plates were wrapped in parafilm and kept in the 30°C static incubator until imaging.
To create supernatant, 200 mL ATS media with 10 mM KNO 3 was added to sterile 500 mL conical flasks and 1 mL of B. subtilis inoculum (prepared as described above) was added and cultures were placed in 30°C with 170 rpm shaking. For initial experiments considering a range of B. subtilis OD 600 supernatants, cultures were sampled at approximately 24, 28, 32, and 52 hrs into the experiment. For later experiments, cultures were sampled at 28 hrs and the exact OD 600 measured. The cultures were centrifuged at x3200g for 10mins followed by vacuum filter sterilisation of the supernatant through Corning PES filters 0.22 μ m for use in later experiments and analysis.
S. indica liquid cultures were grown in 250 mL conical flasks filled to a final volume of 100 mL ATS with the indicated N-sources, B. subtilis supernatant and 100 μ L spore inoculum. These were placed in shaking 30°C incubator 170 rpm for 1 week before sampling. On sampling cultures were passed through Miracloth (Merck-Millipore, Burlington, MA, USA) to filter mycelia from the growth supernatant. Supernatant was collected for metabolite analysis and mycelia were scraped from the Miracloth surface and placed in 2 mL Eppendorf tubes to dry for weight measurements. Tubes and samples were dried at 70°C for 1–2 days and then returned to room temperature for around 24 hours before weighing to equilibrate to ambient humidity. Weights were measured using a Sartorius Secura 124–15 balance by measuring the tubes plus fungal material, then removing the fungal material and weighing only the tubes and calculating the difference.
Bacterial inocula were prepared as described above to a density of OD 600 = 1 in 0.9% saline. To initiate cultures, 100 μ L of inoculum was added to 100 mL BG11+ media (listed above) in a 250 ml conical flask. These cultures were incubated shaking at 30°C and sampled twice daily. OD measurements were taken regularly until OD > 0.5 was reached. Supernatant was collected by centrifugation and filter sterilisation. Ammonium was quantified by using the Supelco Spectroquant ammonium test kit and absorbance measured at 700 nm using CLARIOstar BMG Labtech plate reader.
Fluorescence standard curve was constructed by preparing KPO 4 buffer at four different pH values by combining different ratios of 1 M KH 2 PO 4 :K 2 HPO 4 as follows: pH 4, 100:0 (adjusted down to pH 4 with 0.1 M HCl), pH 5.8 91.5:8.5, pH 6.8 50.3:49.7, pH 8 6:94. Propidium iodide (PI), for normalisation, and the pH sensitive 2′,7’-Bis-(2-Carboxyethyl)-5-(and-6)-Carboxyfluorescein (BCECF) (Invitrogen, Waltham, MA, USA) were added to these at a final concentration of 100 μ M and 10 μ M respectively. Solutions in 3.5 cm Petri dishes were imaged with bright field illumination with 0.5 ms exposure, HcRED1 filter set 41 043 (Chroma) exposure (100 ms) and EGFP filter set 41 018 (Chroma) exposure (100 ms). Light source was pE-300white (CoolLED). Standard curve for pH quantification is performed .
Quantification of free amino acids and ammonium, plus total protein content (amino acid content post acid hydrolysis) was conducted externally by Genaxxon (Ulm, Germany) using HPLC (high performance liquid chromatography) LC3000 with post-column ninhydrin derivitisation at 125°C. For detection of protein-derived amino acids, samples were hydrolysed, prior to HPLC separation, in 6 N HCl at 110°C for 60 hours. For detection of nitrate ions we used the Dionex (Sunnyvale, CA, USA) ICS-5000 + ion-chomotography system. We used an anionic column with KOH eluent. 2.5 μ L of sample (filtered through 0.22 μ m syringe filter) was injected and run at a flow rate of 0.38 mL/min for 30 minutes in a constant gradient as follows: 0 mins 1.5 mM KOH, 8 mins 1.5 mM KOH, 18 mins 15 mM KOH, 23 mins 24 mM KOH, 24 mins 60 mM KOH, 30 mins 60 mM KOH. Nitrate was detected at an elution time of around 12.8 minutes using a conductivity detector. Standard curve for quantification is performed .
We used a modified version of the model developed by to consider the dynamics of ammonia and ammonium between the interior and exterior of a cell growing in a nitrate-only and given pH environment. We defined differential equations for the external (Equation ) and internal (Equation ) ammonium concentrations. (1) [12pt]{minimal}
dN{H}_4\_}/ dt= kN- diff- NH4\_} k1 (2) [12pt]{minimal}
dN{H}_4\_}/ dt= kN\_ in+ diff+N{H}_4\_} k1-N{H}_4\_ in kbm (3) [12pt]{minimal}
diff=P\_ VA(k\_} NH4\_}-k\_ in NH4\_ in) Parameters kN and kN_ in represent the external influx (from cells into the external environment) and internal influx (from nitrate reduction) of ammonium. Parameter diff represents diffusion into or out of the cell from the exterior environment and is given by Equation where P_V A is the rate of ammonia diffusion from the cell and k_in and k_ext are the internal and external dissociation constants respectively for the ammonium-ammonia equilibrium at the pH of those compartments. The rates k1 and kbm represent the rates of ammonium import into the cell and ammonium incorporation into biomass respectively. For the parameter kN_in , we inferred an amount of ammonium per hour that must be generated through nitrate reduction. We considered in the experiments, that there was a reduction of approximately 2.5 mM nitrate over the course of the 28 hrs growth . Thereby indicating at least 90 μ M per hour (2500/28) of internal ammonium production at the population level. We solved differential equations for change in external ammonium concentration (Equation ) and internal ammonium concentration (Equation ) until simulations reached steady state. We check steady state has been reached by verifying that the external ammonium concentration is not changing by more than 10 −5 within experimental timeframes. Endpoint measurements were then used to generate surface plots. Python code for this analysis is available at: https://github.com/lukeZrich/RichardsNsharing2024 .
ATS broth was used for plate reader growth experiments with the following modifications. KPO 4 buffer composition was altered to increase the pH by increasing the ratio of K 2 HPO 4 to KH 2 PO 4 such that the final media contained 1.3 mM KH 2 PO 4 and 1.2 mM K 2 HPO 4 for pH 6.8. To combat precipitation, we reduced the concentration of MgSO 4 and CaCl 2 by ten fold for 0.3 mM each. B. subtilis inocula were prepared as described above and 1 μ L used to inculate 200 μ L of media per well. Cultures were grown in a plate reader (CLARIOstar BMG Labtech) for 3 days at 30°C, 200 rpm and OD 600 measurements taken every 30 minutes.
Protein gene accession as described in were used to perform protein–protein BLAST against the S. indica and S. vermifera genetic information stored in the NCBI (TaxID 1 109 443 and 109 899, respectively). Default settings were used to run Blastp protein–protein Blast.
S. Indica lacks the ability to assimilate nitrate Both bacteria and fungi, generally, use the same metabolic pathways for the assimilation of nitrate into amino acids . Using the pBLAST tool , and genes from Bacillus subtilis as a reference, we searched both the S. indica and Serendipita vermifera NCBI sequence repositories for GDH (Glutamate dehydrogenase), GS (Glutamine synthase), and GOGAT (glutamate synthase) homologs, as well as for NTR (nitrate transporters), NR (nitrate reductase), and NiR (nitrite reductase) (see Methods ). For GDH , GS , and GOGAT we found likely homologs in both fungal species with a > 25% sequence identity match over > 89% of the query sequence length. For NTR , NR , and NiR possible homologs were found in the S. vermifera genome with > 26% sequence identity match over > 42% of the query sequence length but no hits were identified in S. indica for NTR and those identified for NiR and NR had lower identity and length than that quoted for S. vermifera . Additionally, the annotations for those hits do not match NR s and NiR s . To explore this further, we used the Ogatea angusta formerly Hansenula polymorpha yeast species’ nitrate assimilation genes OaYNT1 nitrate transporter, OaYNR1 nitrate reductase, and OaYNI1 nitrite reductase to search the S. indica and S. vermifera genomes for homologous genes. This represents an example of a well-characterised nitrate-assimilating fungus . Likely homologs were identified for all three genes in the S. vermifera genome but only the nitrate reductase ( OaYNR1 ) produced hits in the S. indica genome . These had low sequence similarity over only the latter portion of the query sequence where the NADH binding domain is located in other fungal NR s . These results, and in particular the lack of NTR s, suggest that S. indica will be unable to use nitrate as the sole nitrogen source. To experimentally test the capability of S. indica to use nitrate, we grew S. indica spores on ATS media containing different nitrogen sources: nitrate, ammonium, and glutamine (see Methods ). S. indica displayed extremely reduced growth on media containing nitrate when compared with ammonium and glutamine . S. Indica growth in nitrate media is significantly enhanced in the presence of B. Subtilis Given that S. indica can readily use ammonia and glutamine, we hypothesised that soil bacteria may be able to provide S. indica with these alternative nitrogen sources in a nitrate-dominated environment. To investigate this, S. indica was grown on ATS-nitrate plates with and without the addition of B. subtilis . B. subtilis improves the growth of S. indica on agar plates ( and ), leading to the hypothesis that B. subtilis is providing some exuded nitrogen compound capable of diffusing across the media and being taken up by S. indica . To further investigate nitrogen sharing between B. subtilis and S. indica , we assessed growth of S. indica in liquid culture in the presence of B. subtilis supernatant. Trial experiments indicated that, qualitatively, B. subtilis supernatant promoted the growth of S. indica in liquid culture . Nitrate and ammonium quantification of the B. subtilis supernatant samples used in these experiments shows consumption of nitrate as OD increases and a production of ammonium . At higher optical densities growth promotion was reduced, possibly because of a re-consumption of ammonium by B. subtilis at high bacterial cell densities . To re-confirm this growth-promotion result quantitatively, we designed experiments focusing on B. subtilis supernatant collected at an OD below 2 ( and Methods ). The B. subtilis supernatant addition greatly increased the dry mass of S. indica grown in liquid culture over 1 week , supporting the hypothesis that exudates from B. subtilis are used as a nitrogen source by S. indica . To further confirm this observation, we repeated the supernatant supplementation experiment and again found a significant growth promotion of S. indica with B. subtilis supernatant . We also quantified the nitrate in supernatants from these experiments (see Methods ) and were able to show a clear reduction in the amount of nitrate in B. subtilis supernatants which was not seen in the supernatants of S. indica . Thus, B. subtilis growth results in nitrate consumption and a release of a nitrogen source into supernatant that can be used by S. indica . S. Indica uses ammonia released by B. Subtilis to grow in the nitrate-only media To identify the compounds in the supernatant of nitrate-grown B. subtilis that S. indica could use as N-source, we analysed supernatant samples by HPLC. We could not detect any free amino acids in any of the supernatant samples other than those collected at OD greater than 4 . We also tested supernatant samples for the presence of protein-derived amino acids by subjecting them to acid hydrolysis prior to HPLC quantification. The sum of protein-derived amino acids was not significantly different between samples and generally the amounts detected were low . High amounts of ammonia and taurine were detected in all of these hydrolysis-treated samples, including media only controls, indicating that hydrolysis can result in production of these compounds from media components. The lack of any significant excreted or protein-derived amino acids in supernatant of nitrate-grown B. subtilis , led us to hypothesise that the effect of the supernatant on S. indica could be due to ammonium. It has been noted that ammonia (NH 3 ) exists in equilibrium with ammonium (NH + 4) and that the former can readily leak in and out of the cell due to its high permeability to the membrane ; making it an ideal candidate for such a cross-feeding interaction. If leakage of ammonia is an unavoidable process, this would also explain the observation of ammonium in S. indica only samples , as some of the nitrogen stored in spores would be lost through ammonia leakage during its utilisation. Leakage of ammonia would also complicate the assessment of ammonium usage by S. indica when grown in B. subtilis supernatant, as consumption and leakage would occur simultaneously. In order to correct for this, we subtracted the mean ammonium concentration observed in S. indica -only cultures . We subsequently calculate the change in ammonium concentration over the course of the experiment and compare these to the change in ammonium concentration in B. subtilis only cultures . We find that in cultures where S. indica is placed in media with B. subtilis supernatant there is clear consumption of ammonium. This effect is correlated with S. indica growth , supporting the conclusion that S. indica is using this leaked ammonium to support growth. To further verify that ammonia levels found in B. subtilis supernatant can promote S. indica growth to the levels seen, we supplemented S. indica with the same levels of ammonium found in B. subtilis supernatant. We found that a supplementation of ammonium, equivalent to 37.5 μM nitrogen can significantly enhance S. indica growth in nitrate media . Further confirming ammonia as mediator of the B. subtilis and S. indica interaction in these experiments, the lowest ammonium producing B. subtilis culture - one with the highest OD 600 - produced the smallest growth benefit to S. indica . These findings strongly support the notion that the growth promotion effect of B. subtilis supernatant on S. indica in the nitrate-media is due to ammonium transferred between these organisms. The leakage and sharing of ammonia is likely to be a universal phenomenon. To investigate this idea, we grew B. subtilis and three additional microbial strains ( Allorhizobium rhizophilum and two strains of Pseudomonas composti ) in growth media known to support their growth (BG11) with nitrate as the sole nitrogen source. We found that all bacterial strains produced micromolar amounts of ammonium . Mathematical modelling of ammonia assimilation highlights the role of environmental pH on ammonia leakage and microbial interactions To contextualise the above experimental results, we explored a simplified cellular growth model involving nitrate assimilation and ammonia leakage from cells, building on from a previously published model . We considered a given internal ammonium production rate, mimicking nitrate uptake and subsequent reduction . We explored the effect of key parameters: ammonium uptake rate, ammonium assimilation rate (i.e. biomass incorporation rate) and environmental pH level. We found that increasing uptake or assimilation rates reduced loss of ammonia to the media and vice versa ( &C), although a certain level of leakage is always present even at high uptake and assimilation rates ( &C). Additionally, the impact of lowering the uptake or assimilation rate increases with lower environmental pH ( &C). This suggests that organisms must maintain high ammonium uptake or assimilation to negate the effects of ammonia leakage, particularly under low pH conditions. It may also suggest a benefit for” ammonium-scavenging” organisms to reduce their external pH, inducing leakage in others. This is especially pertinent in light of the observation that S. indica can reduce its external local pH . Reduced ammonia uptake in B. Subtilis results in higher pH impact on its growth and increases its ability to support S. Indica growth in nitrate media One key consideration arising from our model is that the reduction of active uptake of ammonium by a given cell would increase its leakage of ammonia to the environment, and that this effect would be exacerbated under lower pH. To test this theoretical finding, we used an ammonium uptake mutant ( B. subtilis 168∆amtB) and assessed its ability to grow under different pH environments . These experiments were performed in liquid culture, i.e. a well-mixed homogenous environment, to mimic the situation we modelled . Growth of wild-type B. subtilis was not affected significantly by a pH change from 6.8 to 5.8, whereas the mutant strain, with reduced active uptake of ammonium, displays significantly lower growth under reduced pH . This supports the theoretical prediction that cells with reduced ammonium uptake will have a pH dependent impact on their ability to keep ammonium and therefore might be affected in their growth rate. We expect a B. subtilis ammonium uptake mutant to have higher ammonia leakage (at any pH level). To test this, we have used the wildtype and the mutant strains to initiate co-cultures with S. indica on agar plates - we could not use supernatant-based liquid culture experiments, as this B.subtilis strain is a tryptophan auxotroph and added amino acid in monoculture would interfere with the ammonium-based interaction. On agar media with only nitrate as nitrogen source (including no tryptophan), we found that both the wildtype and the mutant can enhance S. indica growth, but the mutant can do so to a higher degree . This finding supports the idea that the mutant with reduced ammonia uptake results in higher ammonia levels in its environment.
lacks the ability to assimilate nitrate Both bacteria and fungi, generally, use the same metabolic pathways for the assimilation of nitrate into amino acids . Using the pBLAST tool , and genes from Bacillus subtilis as a reference, we searched both the S. indica and Serendipita vermifera NCBI sequence repositories for GDH (Glutamate dehydrogenase), GS (Glutamine synthase), and GOGAT (glutamate synthase) homologs, as well as for NTR (nitrate transporters), NR (nitrate reductase), and NiR (nitrite reductase) (see Methods ). For GDH , GS , and GOGAT we found likely homologs in both fungal species with a > 25% sequence identity match over > 89% of the query sequence length. For NTR , NR , and NiR possible homologs were found in the S. vermifera genome with > 26% sequence identity match over > 42% of the query sequence length but no hits were identified in S. indica for NTR and those identified for NiR and NR had lower identity and length than that quoted for S. vermifera . Additionally, the annotations for those hits do not match NR s and NiR s . To explore this further, we used the Ogatea angusta formerly Hansenula polymorpha yeast species’ nitrate assimilation genes OaYNT1 nitrate transporter, OaYNR1 nitrate reductase, and OaYNI1 nitrite reductase to search the S. indica and S. vermifera genomes for homologous genes. This represents an example of a well-characterised nitrate-assimilating fungus . Likely homologs were identified for all three genes in the S. vermifera genome but only the nitrate reductase ( OaYNR1 ) produced hits in the S. indica genome . These had low sequence similarity over only the latter portion of the query sequence where the NADH binding domain is located in other fungal NR s . These results, and in particular the lack of NTR s, suggest that S. indica will be unable to use nitrate as the sole nitrogen source. To experimentally test the capability of S. indica to use nitrate, we grew S. indica spores on ATS media containing different nitrogen sources: nitrate, ammonium, and glutamine (see Methods ). S. indica displayed extremely reduced growth on media containing nitrate when compared with ammonium and glutamine .
Given that S. indica can readily use ammonia and glutamine, we hypothesised that soil bacteria may be able to provide S. indica with these alternative nitrogen sources in a nitrate-dominated environment. To investigate this, S. indica was grown on ATS-nitrate plates with and without the addition of B. subtilis . B. subtilis improves the growth of S. indica on agar plates ( and ), leading to the hypothesis that B. subtilis is providing some exuded nitrogen compound capable of diffusing across the media and being taken up by S. indica . To further investigate nitrogen sharing between B. subtilis and S. indica , we assessed growth of S. indica in liquid culture in the presence of B. subtilis supernatant. Trial experiments indicated that, qualitatively, B. subtilis supernatant promoted the growth of S. indica in liquid culture . Nitrate and ammonium quantification of the B. subtilis supernatant samples used in these experiments shows consumption of nitrate as OD increases and a production of ammonium . At higher optical densities growth promotion was reduced, possibly because of a re-consumption of ammonium by B. subtilis at high bacterial cell densities . To re-confirm this growth-promotion result quantitatively, we designed experiments focusing on B. subtilis supernatant collected at an OD below 2 ( and Methods ). The B. subtilis supernatant addition greatly increased the dry mass of S. indica grown in liquid culture over 1 week , supporting the hypothesis that exudates from B. subtilis are used as a nitrogen source by S. indica . To further confirm this observation, we repeated the supernatant supplementation experiment and again found a significant growth promotion of S. indica with B. subtilis supernatant . We also quantified the nitrate in supernatants from these experiments (see Methods ) and were able to show a clear reduction in the amount of nitrate in B. subtilis supernatants which was not seen in the supernatants of S. indica . Thus, B. subtilis growth results in nitrate consumption and a release of a nitrogen source into supernatant that can be used by S. indica .
uses ammonia released by B. Subtilis to grow in the nitrate-only media To identify the compounds in the supernatant of nitrate-grown B. subtilis that S. indica could use as N-source, we analysed supernatant samples by HPLC. We could not detect any free amino acids in any of the supernatant samples other than those collected at OD greater than 4 . We also tested supernatant samples for the presence of protein-derived amino acids by subjecting them to acid hydrolysis prior to HPLC quantification. The sum of protein-derived amino acids was not significantly different between samples and generally the amounts detected were low . High amounts of ammonia and taurine were detected in all of these hydrolysis-treated samples, including media only controls, indicating that hydrolysis can result in production of these compounds from media components. The lack of any significant excreted or protein-derived amino acids in supernatant of nitrate-grown B. subtilis , led us to hypothesise that the effect of the supernatant on S. indica could be due to ammonium. It has been noted that ammonia (NH 3 ) exists in equilibrium with ammonium (NH + 4) and that the former can readily leak in and out of the cell due to its high permeability to the membrane ; making it an ideal candidate for such a cross-feeding interaction. If leakage of ammonia is an unavoidable process, this would also explain the observation of ammonium in S. indica only samples , as some of the nitrogen stored in spores would be lost through ammonia leakage during its utilisation. Leakage of ammonia would also complicate the assessment of ammonium usage by S. indica when grown in B. subtilis supernatant, as consumption and leakage would occur simultaneously. In order to correct for this, we subtracted the mean ammonium concentration observed in S. indica -only cultures . We subsequently calculate the change in ammonium concentration over the course of the experiment and compare these to the change in ammonium concentration in B. subtilis only cultures . We find that in cultures where S. indica is placed in media with B. subtilis supernatant there is clear consumption of ammonium. This effect is correlated with S. indica growth , supporting the conclusion that S. indica is using this leaked ammonium to support growth. To further verify that ammonia levels found in B. subtilis supernatant can promote S. indica growth to the levels seen, we supplemented S. indica with the same levels of ammonium found in B. subtilis supernatant. We found that a supplementation of ammonium, equivalent to 37.5 μM nitrogen can significantly enhance S. indica growth in nitrate media . Further confirming ammonia as mediator of the B. subtilis and S. indica interaction in these experiments, the lowest ammonium producing B. subtilis culture - one with the highest OD 600 - produced the smallest growth benefit to S. indica . These findings strongly support the notion that the growth promotion effect of B. subtilis supernatant on S. indica in the nitrate-media is due to ammonium transferred between these organisms. The leakage and sharing of ammonia is likely to be a universal phenomenon. To investigate this idea, we grew B. subtilis and three additional microbial strains ( Allorhizobium rhizophilum and two strains of Pseudomonas composti ) in growth media known to support their growth (BG11) with nitrate as the sole nitrogen source. We found that all bacterial strains produced micromolar amounts of ammonium .
To contextualise the above experimental results, we explored a simplified cellular growth model involving nitrate assimilation and ammonia leakage from cells, building on from a previously published model . We considered a given internal ammonium production rate, mimicking nitrate uptake and subsequent reduction . We explored the effect of key parameters: ammonium uptake rate, ammonium assimilation rate (i.e. biomass incorporation rate) and environmental pH level. We found that increasing uptake or assimilation rates reduced loss of ammonia to the media and vice versa ( &C), although a certain level of leakage is always present even at high uptake and assimilation rates ( &C). Additionally, the impact of lowering the uptake or assimilation rate increases with lower environmental pH ( &C). This suggests that organisms must maintain high ammonium uptake or assimilation to negate the effects of ammonia leakage, particularly under low pH conditions. It may also suggest a benefit for” ammonium-scavenging” organisms to reduce their external pH, inducing leakage in others. This is especially pertinent in light of the observation that S. indica can reduce its external local pH .
B. Subtilis results in higher pH impact on its growth and increases its ability to support S. Indica growth in nitrate media One key consideration arising from our model is that the reduction of active uptake of ammonium by a given cell would increase its leakage of ammonia to the environment, and that this effect would be exacerbated under lower pH. To test this theoretical finding, we used an ammonium uptake mutant ( B. subtilis 168∆amtB) and assessed its ability to grow under different pH environments . These experiments were performed in liquid culture, i.e. a well-mixed homogenous environment, to mimic the situation we modelled . Growth of wild-type B. subtilis was not affected significantly by a pH change from 6.8 to 5.8, whereas the mutant strain, with reduced active uptake of ammonium, displays significantly lower growth under reduced pH . This supports the theoretical prediction that cells with reduced ammonium uptake will have a pH dependent impact on their ability to keep ammonium and therefore might be affected in their growth rate. We expect a B. subtilis ammonium uptake mutant to have higher ammonia leakage (at any pH level). To test this, we have used the wildtype and the mutant strains to initiate co-cultures with S. indica on agar plates - we could not use supernatant-based liquid culture experiments, as this B.subtilis strain is a tryptophan auxotroph and added amino acid in monoculture would interfere with the ammonium-based interaction. On agar media with only nitrate as nitrogen source (including no tryptophan), we found that both the wildtype and the mutant can enhance S. indica growth, but the mutant can do so to a higher degree . This finding supports the idea that the mutant with reduced ammonia uptake results in higher ammonia levels in its environment.
We have studied here the possibility of nitrogen sharing, among two common, plant-growth promoting microorganisms, S. indica and B. subtilis . This specific fungi-bacteria pair is previously shown to present a thiamine-mediated auxotrophic interaction , and the fungi S. indica was shown to be incapable of nitrate assimilation . We re-confirmed the latter proposition and showed that S. indica is indeed incapable of growth when nitrate is the sole nitrogen source. We found, however, that this incapacity is lifted, and growth significantly enhanced, in the presence of B. subtilis . We find that this effect is mediated through ammonia, which is leaked from B. subtilis . We find that these results can be rationalised by a mathematical model, incorporating known permeability of ammonia to cell membranes, active ammonium uptake and assimilation, and ammonia-ammonium equilibrium. Utilising this model, we predict that some level of ammonium leakage is inevitable for cells and increases under low environmental pH and reduced uptake and assimilation rates. Taken together, these results experimentally prove that nitrogen sharing among soil microorganisms is a feasible and specific interaction mechanism, and that ammonia-based interactions can be influenced by environmental pH around the microorganisms and their individual ammonia uptake and assimilation rates. The presented findings are relevant in our understanding of nitrogen dynamics in soils. Nitrate is a major component of global soils and applied fertilisers making it an abundant source of nitrogen for many microorganisms and plants. Assimilating nitrate is an energetically expensive process, necessitating both active uptake of nitrate and its reduction to ammonia . Ammonia, free amino acids and more complex organic nitrogen-containing compounds also exist in global soils . Microorganisms and plants have been demonstrated to have to ability to take up not only nitrate and ammonium but also free amino acids and more complex compounds including peptides and proteins . The presented finding that ammonia sharing can be possible for microorganisms that do not utilise nitrate suggest that this mechanism can allow such microorganisms to cut an energetic corner in their search for nitrogen and rely on the exuded/ leaked compounds provided by other members of a soil community. This mechanism could then lead to loss of genetic capacity to assimilate nitrate. The mechanism we describe here is likely not unique to ammonium and ammonia. Many organisms, including microorganisms, produce a range of volatile organic compounds that can diffuse through the gas phase . As such, these cannot be held by the producer and may act as a nutrient source for surrounding organisms. More specifically, any chemical species that exists in equilibrium between an ion (non-permeable to cell membranes) and a dissolved gas (highly permeable) may exhibit similar dynamics exactly like ammonia and ammonium. This is true for at least two other important growth elements, carbon and sulfur. Carbonate ions and sulfate ions exist in equilibria with dissolved gaseous carbon dioxide and sulfur dioxide respectively. The model presented here shows that ammonia leaked from cells will relate to their uptake and biomass incorporation rates. Subject to exact values of these parameters, we see a wide range of leaked ammonia concentrations are attainable in the vicinity of cells . Indeed, we found, testing 3 other bacteria, common in the soil, that they have all leaked ammonia into culture supernatant when grown in nitrate-media . The high membrane permeability of ammonia means that all organisms are ill-fated to leak nitrogen in this form. Furthermore, a broad spectrum of organisms are known to be incapable of growth on nitrate as the sole nitrogen source or are missing the required genetic machinery . This fact, coupled with our results, indicates that this mechanism of nitrogen sharing may be widespread in microbial communities and alludes to the potential prevalence of the loss of such machinery. We have also identified here a small amount of ammonium production in the S. indica cultures grown on nitrate media and without bacterial supernatant supplementation. This suggests that S. indica can use spore stored nitrogen sources, or recycling of amino acids, to achieve some growth in nitrate media and ammonia is leaked as a consequence. Deamination of nucleic and amino acids, in nitrogen free media, has been observed in response to nitrogen limitation in Escherichia coli and Neurospora crassa respectively. Considering that this process of spore-stored nitrogen recycling and ammonia loss would be happening in tandem with ammonia assimilation from B. subtilis supernatant, we conclude that growth of S. indica in media supplemented with B. subtilis supernatant results in net ammonia consumption. Ammonia dynamics can be influenced by environmental pH. Specifically, any nitrogen assimilating microbe will suffer at low pH and any impact of reduced assimilation will be exacerbated. Furthermore, any microbe relying on such leaked ammonia would benefit from lower pH in its local environment, provided sufficiently high ammonia uptake. Experimental results presented here align with these considerations, we found an ammonium transporter mutant of B. subtilis to be more susceptible to reduced growth in face of environmental pH reduction, and that S. indica promotes a low pH local environment upon growth. Additionally, B. subtilis may represent a particularly good” leaker” of ammonia; B. subtilis GDH may exclusively function to degrade, rather than produce, glutamate owing to a very low affinity for the ammonium ion . This would mean that in B. subtilis ammonia is only assimilated via GS . This would effectively reduce the assimilation rate described by our model and lead to higher rates of leakage into exterior media. These results also have implications in the context of soil. Low pH around Arabidopsis root systems has been shown to be vital in the suppression of plant immune responses by Pseudomonas . Beneficial microorganisms also have to suppress plant immune responses to colonise host tissue . Thus, it is possible that pH lowering is a strategy used by S. indica , or other fungi, initially to aid in host immune responses to allow tissue colonisation and has subsequently allowed S. indica to utilise low pH environments and high ammonium uptake as an ammonium-scavenging strategy.
Richards_etal_final_SI_wrae171 richards_etal_SupplementaryFigures_wrae171 richards_etal_DatafileS1_wrae171
|
Metabolomics Combined with Transcriptomics Analysis Reveals the Regulation of Flavonoids in the Leaf Color Change of | 5e5495cc-5f3e-4f51-87b2-cb97e5eb1d4b | 11678339 | Biochemistry[mh] | Acer truncatum (Sapindaceae) is one of the most important deciduous maple trees and is often used for its ornamental autumn leaf color in China . More and more studies are focusing on the mechanism of leaf discoloration in which the leaf color of trees often changes from green to yellow and then to red in autumn . Some studies found that the chloroplast ultrastructure of Forsythia suspensa (thunb.) vahl had a looser structure than green leaves, and V-myb avian myeloblastosis viral oncogene homolog ( MYB ), basic helix–loop–helix ( bHLH), NAM, ATAF and CUC ( NAC ) were associated with the leaf pigment compounds of Fraxinus angustifolia vahl . Some studies on maple leaf color changes found that anthocyanin synthase ( ANS ) and bronze-1 ( BZ1 ) in A. mandshuricum Maxim resulted in the accumulation of cyanidin 3-O-glucoside, which causes a significant reddening of the leaf blade . However, little effort has been made to elucidate the mechanism underlying the variation in color between red, yellow and green leaves of A. truncatum . In general, the color variation of leaves is more complex than that of corollas . Green leaves are mainly the result of the dominance of chlorophyll content among all pigments . By contrast, the formation of yellow leaves is mainly due to the gradual degradation of chlorophyll, resulting in the dominance of carotenoids in the leaves . However, flavonoid biosynthesis is the determining factor for the change of leaf color to a non-yellow color . Flavones, flavonols, flavanones, flavanonols, isoflavones, catechins, anthocyanins and proanthocyanidins belong to subgroups of flavonoids . Anthocyanins are particularly the main reason for the red color of the leaves, and this pigment also helps the plant resist various biotic and abiotic stresses . In line with the previous analyses, there are two main functional functions of naturally occurring anthocyanins : one is the resistance to external stresses and the other is using different color strategies to complete the life cycle . Consequently, the accurate identification of flavonoids and even anthocyanin compounds in maple is very important for the effective use of forest resources. Plant metabolomics is the qualitative and quantitative study of small-molecule metabolites in plants that helps researchers better understand patterns of metabolite synthesis and accumulation . Currently, research on plant metabolites is mainly concerned with crop improvement, assisted breeding, discovering biomarkers, the assessment of nutrients, and biotic and abiotic stress studies . Metabolomics for the analysis of flavonoid metabolites is commonly used to analyze the mechanism of plant color formation . For example, rice during yellowing is induced by the metabolism of flavones, flavonols, isoflavones, and anthocyanidins . Furthermore, a significant correlation was found between the accumulation of malvidin 3-O-glucoside and pelargonidin 3-O-glucoside and the change in leaf color from green to red in Padus virginiana . These results indicate that metabolomics is an important and effective method for analyzing the mechanisms of plant color formation. In order to comprehensively analyze the relationship between leaf color and flavonoids and anthocyanins, leaves from the same plant at three developmental stages, namely red, medium and green leaves, were selected as study materials. Metabolomics was used to analyze the changing pattern of flavonoid metabolites and key metabolites in the process of the leaf color change from red to green, and the mechanism of the leaf color difference was explored using transcriptomics analysis. Candidate genes and metabolic pathways for leaf color variation were further demonstrated. Our results provide a new perspective to understand the flavonoid metabolism of A. truncatum at different developmental stages, which is conducive to the utilization of its leaf resources.
2.1. Variations in Phenotypes and Analysis of Colour Parameters The leaves of A. truncatum showed colors at different stages of fall leaf coloration. With the development of the leaves, the leaves gradually changed from green to middle, and finally the mature leaves developed into red, as seen in A. During the change of leaf color from green to red, observations showed that the trend of L* and b* values followed the same patterns, with an initial increase and then a decrease, and the overall fluctuation range was large ( B). However, L* and b* were not significantly different between red and green leaves. Notably, a* showed a continuous upward trend, which was significantly different among all groups. 2.2. Analysis of Bioactive Flavonoids In the OPLS-DA model, the Q2 values for pairwise comparisons exceeded 0.82 and the Q2 values exceeded 0.9 in the two pairwise groups compared with green leaves ( A). The PCA showed that the composition of the metabolites for the leaf colors of the three different developmental phases differed considerably ( B). In the clustered heat map containing all samples, leaf metabolites in the red and middle phases had similar expression patterns. The metabolites of the green phase were distinctly different from the metabolites of the other phases and they were assigned between the two branches of the cluster ( C). The flavonoids were analyzed in A. truncatum ( and ). In the flavonoid category, o-methylated flavonoids, flavans, flavonoid glycosides, flavones, and biflavonoids and polyflavonoids were identified. Among the o-methylated flavonoids, hesperetin was identified. Among the flavonoid glycosides, cyanidin 3-glucoside, isoquercitrin, kaempferitrin, peonidin-3-glucoside, pelargonidin 3-sophoroside, myricitrin, quercitrin and astragalin were found. In the flavans class, epicatechin, catechin, naringenin, leucopelargonidin, -epigallocatechin, epigallocatechin gallate and eriodictyol were identified. Among the biflavonoids and polyflavonoids, procyanidin B2 was found. The last remaining six belonged to the flavans. 2.3. Analysis of Differentially Expressed Flavonoid Metabolites and KEGG Classification Pairwise comparisons were made between the three groups of materials with different periods of leaf color change ( A and ). A total of 134 significant differentially expressed metabolites (DEMs) were identified between red and middle leaves, of which 98 increased and 36 decreased. In addition, 243 significant DEMs were identified in the comparison between red and green leaves, of which 158 were increasing and 85 were decreasing. Meanwhile, 237 significant DEMs were identified in middle leaves compared with green leaves, of which 143 increased and 94 decreased. The KEGG classification showed that the significant DEMs of red and middle color were mainly involved in aminobenzoate degradation, the biosynthesis of phenylpropanoids, central carbon metabolism in cancer, tyrosine metabolism and styrene degradation pathways ( B). Red and green were mainly related to the biosynthesis of biosynthesis of phenylpropanoids, flavonoid biosynthesis, phenylpropanoid biosynthesis, phenylalanine metabolism and ABC transporters having the smallest p -value. The significant DEMs of middle and green were mainly enriched in the biosynthesis of phenylpropanoids, flavonoid biosynthesis, phenylpropanoid biosynthesis, ABC transporters and the biosynthesis of plant secondary metabolites. The significant DEMs of flavonoid biosynthesis in red vs. green and middle vs. green were subsequently analyzed ( C). In red compared with green, there were 13 flavonoid metabolites of which 5 were significantly up-regulated. In addition, the expression of seven flavonoid metabolites was increased in middle compared to green. Notably, most flavonoid metabolites, especially naringenin, chlorogenic acid, apigenin, taxifolin, dihydromyricetin, 4-coumaroylshikimate, leucopelargonidin, 5,7-dihydroxyflavone, dattelic acid and -gallocatechin, were expressed in both comparison groups. 2.4. Analysis of Transcriptome Results and Functional Annotation In the clustered heat map containing all samples, genes in the red and middle phases had similar expression patterns ( A). The genes of the green phase were distinctly different from the metabolites of the other phases and they were assigned between the two branches of the cluster. The differentially expressed genes (DEGs) of red, middle and green leaves in different developmental stages were analyzed ( B). A total of 2101 DEGs were obtained from the red vs. middle comparison, and the higher number of 1051 DEGs was upregulated. In addition, the highest number of DEGs was 9627 in red vs. green, among which 5369 genes showed upregulated expression, and 4258 genes showed downregulated expression. Simultaneously, the number of 7102 DEGs was obtained in middle vs. green, in which the genes with downregulated expression accounted for 41.98% of all DEGs. The gene ontology (GO) enrichment results showed that the DEGs of the red vs. middle group were significantly enriched in biological processes such as response to starvation, cellular response to phosphate starvation, response to extracellular stimulus, cellular response to starvation and cellular response to external stimulus ( C). The DEGs of red vs. green were enriched in plastid, chloroplast, structural constituent of ribosome, ribosome and thylakoid. The DEGs of the red vs. green group were enriched in chloroplast. In addition, 37 of these genes were annotated to the porphyrin metabolism pathway ( and ). The Kyoto encyclopedia of genes and genomes (KEGG) enrichment results showed that the red vs. middle group was annotated into 116 metabolic pathways ( D), where alanine, aspartate and glutamate metabolism, starch and sucrose metabolism, glycerolipid metabolism, ether lipid metabolism and steroid biosynthesis pathways were significantly enriched. The red vs. green were enriched in 130 metabolic pathways and the middle vs. green were enriched in 128 metabolic pathways. 2.5. Flavonoid Biosynthesis in Relation to Genes and Metabolites The nine-quadrant plot based on correlation analysis shows that metabolite and gene expression patterns are consistent in quadrants three and seven ( A). We found that kaempferin, proanthocyanidin B2, 5,7-dihydroxyflavone, apigenin, quercetin, epigallocatechin gallate and astragalin were identified as being related the six genes F3H , FLS , ANS , LAR , DFR, CHS and CYP75B1 . The epicatechin, naringenin, leucopelargonidin and UGT75C1 genes were identified as related. To further confirm the reliability of the RNA-Seq results, these eight candidate genes were selected for verification. The qRT-PCR analysis of the genes encoding these enzymes showed that all genes were significantly down-regulated ( B). A comprehensive analysis of enzyme activity and the corresponding gene expression pattern showed that the changing trend of enzyme activity of ANS , CHS , DFR , FLS and LAR was consistent with that of the gene expression pattern, while that of the other two enzymes was different from that of the gene expression pattern. 2.6. Analysis of Flavonoid and Anthocyanin Biosynthesis The combination analysis indicated that the expression of genes related to flavonoid synthesis in red leaves was higher than that in green leaves in the same developmental stage . Under the action of DFR , dihydroquercetin (DHQ), dihydrokempferol (DHK) and dihydromyricetin (DHM) were transformed into leucocyanidin, leucopelargonidin and eucodelphinidin. They were generated as anthocyanins under the action of ANS , so anthocyanins accumulate in red and middle leaves. With the development of leaf color, the CHS , F3H , DFR and ANS genes were continuously upregulated in the subsequent developmental stages. By contrast, LAR showed continuous downregulation, which led to the reduced conversion of leucocyanidin into catechin. Five relevant metabolites, namely cyanidin 3-O-β-D-sambubioside, cyanidin 3-O rutinoside, pelargonidin 3-O-3″,6″-O-dimalonylglucoside, delphinidin 3,7-di-O-β-D-glucoside and 3-O-β-D-sambubioside, showed differential changes in the process. In addition, compared with green leaves, cyanidin 3-O-beta-D-sambubioside and cyanidin 3-O rutinoside showed the most significant increases in red leaves during the two stages. 2.7. Analysis of Proteins Encoding Genes for Flavonoids and Anthocyanin Metabolites The predicted secondary structures of the eight proteins encoding genes showed that all the proteins consisted of four parts: an alpha helix, an extended chain, a beta turn, and a random coil . For another, these proteins had the higher proportion of random coil or α-helix, followed by an alpha helix and an extended chain, and the secondary structure of CYP75B1 had the lowest proportion of β-turn. Second, their encoded proteins’ tertiary structure includes alpha helices, extended chains, and random coils . The tertiary structure shows that the secondary structure further folds in a more regular manner, with similar structures formed by different proteins, indicating that their functions are different and further demonstrating the diversity of functions of its members. These genes encode a minimum of 265 amino acids and a maximum of 517 amino acids (CYP75B1) . Meanwhile, three identical structural domains with two identical motifs were found in F3H, FLS and ANS. The three proteins were predicted to belong to the protein family Plant 2OG-oxidoreductases (2ODOs). CHS possesses the active site of the enzyme chalcone/stilbene synthase. Furthermore, the protein encoded by CYP75B1 belongs to the cytochrome P450. In addition, the structural genes were usually controlled by transcription factors in leaf color-related biosynthetic pathways, such as MYB , bHLH , NAC and WRKY . Our study identified 36 MYBs , 126 bHLHs , 78 NACs and 41 WRKYs that were up-regulated in the group of red vs. green. These up-regulated transcription factors were positively correlated with anthocyanin metabolism in leaves. One NAC was annotated to the flavonoid synthesis pathway, while two MYBs were localized to the porphyrin metabolism pathway .
The leaves of A. truncatum showed colors at different stages of fall leaf coloration. With the development of the leaves, the leaves gradually changed from green to middle, and finally the mature leaves developed into red, as seen in A. During the change of leaf color from green to red, observations showed that the trend of L* and b* values followed the same patterns, with an initial increase and then a decrease, and the overall fluctuation range was large ( B). However, L* and b* were not significantly different between red and green leaves. Notably, a* showed a continuous upward trend, which was significantly different among all groups.
In the OPLS-DA model, the Q2 values for pairwise comparisons exceeded 0.82 and the Q2 values exceeded 0.9 in the two pairwise groups compared with green leaves ( A). The PCA showed that the composition of the metabolites for the leaf colors of the three different developmental phases differed considerably ( B). In the clustered heat map containing all samples, leaf metabolites in the red and middle phases had similar expression patterns. The metabolites of the green phase were distinctly different from the metabolites of the other phases and they were assigned between the two branches of the cluster ( C). The flavonoids were analyzed in A. truncatum ( and ). In the flavonoid category, o-methylated flavonoids, flavans, flavonoid glycosides, flavones, and biflavonoids and polyflavonoids were identified. Among the o-methylated flavonoids, hesperetin was identified. Among the flavonoid glycosides, cyanidin 3-glucoside, isoquercitrin, kaempferitrin, peonidin-3-glucoside, pelargonidin 3-sophoroside, myricitrin, quercitrin and astragalin were found. In the flavans class, epicatechin, catechin, naringenin, leucopelargonidin, -epigallocatechin, epigallocatechin gallate and eriodictyol were identified. Among the biflavonoids and polyflavonoids, procyanidin B2 was found. The last remaining six belonged to the flavans.
Pairwise comparisons were made between the three groups of materials with different periods of leaf color change ( A and ). A total of 134 significant differentially expressed metabolites (DEMs) were identified between red and middle leaves, of which 98 increased and 36 decreased. In addition, 243 significant DEMs were identified in the comparison between red and green leaves, of which 158 were increasing and 85 were decreasing. Meanwhile, 237 significant DEMs were identified in middle leaves compared with green leaves, of which 143 increased and 94 decreased. The KEGG classification showed that the significant DEMs of red and middle color were mainly involved in aminobenzoate degradation, the biosynthesis of phenylpropanoids, central carbon metabolism in cancer, tyrosine metabolism and styrene degradation pathways ( B). Red and green were mainly related to the biosynthesis of biosynthesis of phenylpropanoids, flavonoid biosynthesis, phenylpropanoid biosynthesis, phenylalanine metabolism and ABC transporters having the smallest p -value. The significant DEMs of middle and green were mainly enriched in the biosynthesis of phenylpropanoids, flavonoid biosynthesis, phenylpropanoid biosynthesis, ABC transporters and the biosynthesis of plant secondary metabolites. The significant DEMs of flavonoid biosynthesis in red vs. green and middle vs. green were subsequently analyzed ( C). In red compared with green, there were 13 flavonoid metabolites of which 5 were significantly up-regulated. In addition, the expression of seven flavonoid metabolites was increased in middle compared to green. Notably, most flavonoid metabolites, especially naringenin, chlorogenic acid, apigenin, taxifolin, dihydromyricetin, 4-coumaroylshikimate, leucopelargonidin, 5,7-dihydroxyflavone, dattelic acid and -gallocatechin, were expressed in both comparison groups.
In the clustered heat map containing all samples, genes in the red and middle phases had similar expression patterns ( A). The genes of the green phase were distinctly different from the metabolites of the other phases and they were assigned between the two branches of the cluster. The differentially expressed genes (DEGs) of red, middle and green leaves in different developmental stages were analyzed ( B). A total of 2101 DEGs were obtained from the red vs. middle comparison, and the higher number of 1051 DEGs was upregulated. In addition, the highest number of DEGs was 9627 in red vs. green, among which 5369 genes showed upregulated expression, and 4258 genes showed downregulated expression. Simultaneously, the number of 7102 DEGs was obtained in middle vs. green, in which the genes with downregulated expression accounted for 41.98% of all DEGs. The gene ontology (GO) enrichment results showed that the DEGs of the red vs. middle group were significantly enriched in biological processes such as response to starvation, cellular response to phosphate starvation, response to extracellular stimulus, cellular response to starvation and cellular response to external stimulus ( C). The DEGs of red vs. green were enriched in plastid, chloroplast, structural constituent of ribosome, ribosome and thylakoid. The DEGs of the red vs. green group were enriched in chloroplast. In addition, 37 of these genes were annotated to the porphyrin metabolism pathway ( and ). The Kyoto encyclopedia of genes and genomes (KEGG) enrichment results showed that the red vs. middle group was annotated into 116 metabolic pathways ( D), where alanine, aspartate and glutamate metabolism, starch and sucrose metabolism, glycerolipid metabolism, ether lipid metabolism and steroid biosynthesis pathways were significantly enriched. The red vs. green were enriched in 130 metabolic pathways and the middle vs. green were enriched in 128 metabolic pathways.
The nine-quadrant plot based on correlation analysis shows that metabolite and gene expression patterns are consistent in quadrants three and seven ( A). We found that kaempferin, proanthocyanidin B2, 5,7-dihydroxyflavone, apigenin, quercetin, epigallocatechin gallate and astragalin were identified as being related the six genes F3H , FLS , ANS , LAR , DFR, CHS and CYP75B1 . The epicatechin, naringenin, leucopelargonidin and UGT75C1 genes were identified as related. To further confirm the reliability of the RNA-Seq results, these eight candidate genes were selected for verification. The qRT-PCR analysis of the genes encoding these enzymes showed that all genes were significantly down-regulated ( B). A comprehensive analysis of enzyme activity and the corresponding gene expression pattern showed that the changing trend of enzyme activity of ANS , CHS , DFR , FLS and LAR was consistent with that of the gene expression pattern, while that of the other two enzymes was different from that of the gene expression pattern.
The combination analysis indicated that the expression of genes related to flavonoid synthesis in red leaves was higher than that in green leaves in the same developmental stage . Under the action of DFR , dihydroquercetin (DHQ), dihydrokempferol (DHK) and dihydromyricetin (DHM) were transformed into leucocyanidin, leucopelargonidin and eucodelphinidin. They were generated as anthocyanins under the action of ANS , so anthocyanins accumulate in red and middle leaves. With the development of leaf color, the CHS , F3H , DFR and ANS genes were continuously upregulated in the subsequent developmental stages. By contrast, LAR showed continuous downregulation, which led to the reduced conversion of leucocyanidin into catechin. Five relevant metabolites, namely cyanidin 3-O-β-D-sambubioside, cyanidin 3-O rutinoside, pelargonidin 3-O-3″,6″-O-dimalonylglucoside, delphinidin 3,7-di-O-β-D-glucoside and 3-O-β-D-sambubioside, showed differential changes in the process. In addition, compared with green leaves, cyanidin 3-O-beta-D-sambubioside and cyanidin 3-O rutinoside showed the most significant increases in red leaves during the two stages.
The predicted secondary structures of the eight proteins encoding genes showed that all the proteins consisted of four parts: an alpha helix, an extended chain, a beta turn, and a random coil . For another, these proteins had the higher proportion of random coil or α-helix, followed by an alpha helix and an extended chain, and the secondary structure of CYP75B1 had the lowest proportion of β-turn. Second, their encoded proteins’ tertiary structure includes alpha helices, extended chains, and random coils . The tertiary structure shows that the secondary structure further folds in a more regular manner, with similar structures formed by different proteins, indicating that their functions are different and further demonstrating the diversity of functions of its members. These genes encode a minimum of 265 amino acids and a maximum of 517 amino acids (CYP75B1) . Meanwhile, three identical structural domains with two identical motifs were found in F3H, FLS and ANS. The three proteins were predicted to belong to the protein family Plant 2OG-oxidoreductases (2ODOs). CHS possesses the active site of the enzyme chalcone/stilbene synthase. Furthermore, the protein encoded by CYP75B1 belongs to the cytochrome P450. In addition, the structural genes were usually controlled by transcription factors in leaf color-related biosynthetic pathways, such as MYB , bHLH , NAC and WRKY . Our study identified 36 MYBs , 126 bHLHs , 78 NACs and 41 WRKYs that were up-regulated in the group of red vs. green. These up-regulated transcription factors were positively correlated with anthocyanin metabolism in leaves. One NAC was annotated to the flavonoid synthesis pathway, while two MYBs were localized to the porphyrin metabolism pathway .
The regulation of plant leaf color is a complex process . Some results showed that anthocyanins were a major class of compounds that cause changes in leaf color in plants . The color of the leaves was in turn related to the content of the mostly studied phytochemicals, which include flavonoids and phenolic acids . Furthermore, fluctuations in plant leaf chemical levels could trigger leaves to take on different colors. The other results showed that leaf pigment content affects leaf color differences and its content was negatively correlated with L* . Furthermore, when the anthocyanin content of Lycium barbarum L. accumulated, it was reflected in the color by the process of turning from green to red and continuing to deepen . In this study, L* was assigned the highest value in the middle color, which indicated that the chlorophyll content might reach its lowest at this point. When the leaf color changed from green to red, the chlorophyll content had a tendency to decrease and then increase. It is known that the a* value is positively correlated with anthocyanin; when a* gradually increased, the leaf undergoes the process of anthocyanin accumulation. Therefore, we assumed that the content of both flavonoids and anthocyanins increase gradually in the process of coloration of the leaves to red. The molecular basis of flavonoid biosynthesis is increasingly important today . For instance, the metabolite analysis of Ficus carica L. detected 15 different flavonoid-related metabolites, which included the very significant accumulation of the colorless flavonoids procyanidin B1, luteolin-3′,7-di-O-glucoside, epicatechin and quercetin-3-O-rhamnoside in the mature purple peel . A total of 40 flavonoid metabolites were identified through the metabolite extraction and characterization of Cucumis melo . In these metabolites, flavonoids, flavanones, isoflavones and anthocyanins were the substances that mainly affected fruit color . The quantitative analysis revealed that the four varieties in question contained a combination of 125 distinct flavonoids in Actinidia arguta , with only delphinidin 3-O-glucoside, cyanidin O-octanoic acid and pelargonidin 3-O-β-D-glucoside being detected in the red and purple fruits . An aggregate of 23 flavonoid-related metabolites were detected in this article. These belong to o-methylated flavonoids, flavones, flavans, flavonoid glycosides and biflavonoids and polyflavonoids. However, in this study, only five anthocyanins had significant discrepancies among the different stages: cyanidin 3-O-β-D-sambubioside, cyanidin 3-O rutinoside, pelargonidin 3-O-3″,6″-O-dimalonylglucoside, delphinidin 3,7-di-O-β-D-glucoside and 3-O-beta-D-sambubioside. Interestingly, more anthocyanins associated with the leaf color difference were detected. Additionally, there were similarities and differences in the types of flavonoid metabolites between the above plants and A. truncatum . Both it and Actinidia arguta also contained delphinidin, which was not present in the other plants, suggesting differences in flavonoid composition between species. Phenylalanine, an upstream reaction of flavonoids and anthocyanins, was first converted to p-coumaroyl-CoA coenzyme a catalyzed by PAL , C4H and 4CL . In the presence of CHS and CHI , p-coumaroyl-CoA was converted to naringenin chalcone and then to naringenin. The expression of CHS was up-regulated in the process of the leaf color change, which was consistent with the qRT-PCR results. Meanwhile, CHS was the first key enzyme in flavonoid synthesis and its activity determines the formation of related metabolites . The expression of CHS in the red leaf stage was highly significant for the green leaf stage, suggesting that CHS may be involved in the process of leaf color changes. Then, naringenin was catalyzed by F3H to form DHK, which then continued to be catalyzed by F3H to form DHQ and DHM, respectively. Under the catalytic action of DFR and ANS , DHK, DHQ and DHM formed unstable anthocyanins such as cyanidin, pelargonidin and delphinidin, respectively. DHK, DHQ and DHM catalyzed the DFR to produce different ones in different plants and the ANS was necessary for the accumulation of different anthocyanins . The unstable anthocyanins produced eventually formed stable anthocyanins in the presence of UGT75C1 . To summarize, the five stabilities of anthocyanins that were described in detail in the previous paragraph were predominantly found in A. truncatum . 2ODDs are a family of proteins with both DIOX_N and 2OG-FeII_Oxy conserved structures . The related article pointed to them as the second-largest family of oxidative enzymes in plants, involved in various oxidative reactions . They are widely involved in secondary metabolic processes in plants, such as the biosynthesis of flavonoids, alkaloids and terpenoids . We expected that 2ODD could catalyze the conversion of naringenin into dihydrokaempferol, indicating that the enzyme was a typical F3H . This was demonstrated in the investigation that we have carried out. In our study, F3H, FLS and ANS belonged to this family. They possessed the same conserved structural domains DIOX_N and 2OG-FeII_Oxy described above, which could be involved in flavonoid formation and have an impact on anthocyanin biosynthesis. Correspondingly, we need to focus on the cytochrome P450 family. The P450s are thioredoxin proteins involved in the oxidative degradation of various compounds. P450 was imprinted in Scutellaria baicalensis for anthocyanin modification . The characterization and analysis of P450 from grapes revealed that its subfamily CYP75 is involved in anthocyanin production in a similar way . CYP75B1, which undoubtedly possesses the conserved structural domain of P450, affects the production of DHQ, DHK and DHM from naringenin. Increased gene activity of these members from different families triggers the accumulation of flavonoids and anthocyanins in red leaves.
4.1. Plant Materials and Sampling A. truncatum growing in the wild on Jilin Agricultural University campus (43°05′–45°15′ N, 124°18′–127°05′ E) was used. Based on the color pattern of the leaves, three different colored leaves of the same plant were selected as study material (red, middle and green leaves). Ten leaves were collected from each group and leaves with a similar location and color on the branches were selected. The test materials were divided into two parts: one half of the samples was quickly scanned for leaf color parameters, and the other half of the samples was quickly fixed with liquid nitrogen and later moved to storage at −80 °C. 4.2. Determination of Leaf Colour Parameters Leaf color was quantified in accordance with the International Commission on Illumination (CII) color standard. Luminosity (L*), a*, and b* values were obtained through the use of a CR30 colorimeter (CHNSpec, Hanghzhou China). After calibration using the colorimetric plate, five points were randomly selected and averaged on each blade, with the objective of avoiding the leaf edges and radial main veins . The exercise was repeated five times for each leaf color. The meanings of the three parameters are as follows. L* indicates the brightness of the color, where a positive number means whitish and a negative number means blackish. Additionally, a* indicates the red–green value, where a positive value means the color is redder and a negative value means it is greener. Finally, b* indicates the yellow–blue value, where positive values are yellowish and negative values are bluish. 4.3. Extraction of Total Metabolites First, the sample was weighed accurately in a 2 mL centrifuge tube and 600 µL of MeOH containing 2-Amino-3-(2-chloro-phenyl)-propionic acid (4 ppm) was added and vortexed for 30 s. Second, The sample was then placed in a tissue grinder and ground at 55 Hz for 60 s, followed by sonication at room temperature for 15 min. Finally, the sample was centrifuged at 12,000 rpm at 4 °C for 10 min on a H1850-R refrigerated centrifuge (Cence, Changsha, China), and the supernatant was taken through a 0.22 μm membrane and added into the detection vial for LC-MS detection . 4.4. Metabolomics Analysis The LC analysis was performed on a Vanquish UHPLC System (Thermo Fisher Scientific, Waltham, MA, USA). Chromatography was carried out with an ACQUITY UPLC ® HSS T3 (2.1 × 100 mm, 1.8 µm) (Waters, Milford, MA, USA). The column was maintained at 40 °C. The flow rate and injection volume were set at 0.3 mL/min and 2 μL, respectively. Mass spectrometric detection of metabolites was performed on Q Exactive (Thermo Fisher Scientific, Waltham, MA, USA) with an ESI ion source. Simultaneous MS1 and MS/MS (full MS-ddMS2 mode, data-dependent MS/MS) acquisition was used. The parameters were as follows: capillary temperature, 325 °C; MS1 range, m / z 100–1000; MS1 resolving power, 70,000 FWHM; number of data-dependent scans per cycle, 10; MS/MS resolving power, FWHM; normalized collision energy, 30 eV; dynamic exclusion time, automatic. The raw mass spectrometry downcomer files were converted into the mzXML file format using the MSConvert tool in the Proteowizard software package (v3.0.8789). Peak detection, peak filtering and peak alignment were carried out using the R XCMS software package (v3.12.0), and a list of substances for quantification was obtained. Substances with coefficients of variation smaller than 30% in the QC samples were then retained for subsequent analysis. The molecular weight of the metabolite was ascertained by means of the mass-to-charge ratio of the parent ion present in the primary mass spectrum. Furthermore, the molecular formula was deduced on the basis of the mass number deviation, as well as the information provided by the additional ions. Following this, the metabolite was matched with a previously existing record in a database, thereby achieving its preliminary identification. Meanwhile, the metabolites detected in the secondary spectrum were subjected to a process of secondary identification. This involved the matching of the metabolite data with the information contained in the database, including the fragment ions of each metabolite. 4.5. Total RNA Extraction and Transcriptome Sequencing The mRNA with polyA structure in the total RNA was enriched by Oligo(dT) magnetic beads. Subsequently, the RNA was subjected to ionic interruption, which fragmented it into fragments of approximately 300 base pairs in length. The initial cDNA strand was synthesized using RNA as a template with a 6-base random primer and reverse transcriptase. The second cDNA strand was then synthesized with the initial cDNA strand serving as the template. Following the construction of the library, the library fragments were amplified by PCR. The library size was 450 bp, and the total and effective concentrations were subsequently determined by an Agilent 2100 Bioanalyzer (Agilent, Santa Clara, CA, USA). Subsequently, the mixed libraries were uniformly diluted to 2 nM and denatured by alkaline denaturation, thereby forming single-stranded libraries. Following the extraction, purification and construction of the libraries, the libraries were subjected to paired-end sequencing using next-generation sequencing (NCBI project: PRJNA1196921). 4.6. Identification and Analysis of DEGs FPKM was used to normalize the raw gene expression calculation. In order to identify genes that were differentially expressed among the three different groups, we identified genes with |log 2 foldchang| > 1 and a p -value < 0.05 as DEGs. GO functional enrichment analysis and KEGG pathway analysis were performed on the confirmed DEGs. 4.7. qRT-PCR Analysis Primers specific to structural genes involved in flavonoid biosynthesis were designed for qRT-PCR analysis using Primer Premier 5 software . All samples were subjected to three replicates, as were three technical replicates. The internal control genes employed were actin and β-tubulin . The relative expression levels of the target genes were calculated employing the 2 −∆∆Ct methodology. Three experimental replicates were performed for each sample. 4.8. Protein Biology Analysis The secondary structure of the protein encoding the key enzyme gene for flavonoids and anthocyanin synthesis was predicted using the SOPMA online tool ( https://npsa.lyon.inserm.fr/cgi-bin/npsa_automat.pl?page=/NPSA/npsa_sopma.html , accessed on 4 October 2024) and modeled for the tertiary structure using the Phyre2 online tool ( http://www.sbg.bio.ic.ac.uk/phyre2/ , accessed on 4 October 2024) for tertiary structure modeling. The proteins were also functionally analyzed using the website InterPro ( https://www.ebi.ac.uk/interpro/ , accessed on 5 October 2024).
A. truncatum growing in the wild on Jilin Agricultural University campus (43°05′–45°15′ N, 124°18′–127°05′ E) was used. Based on the color pattern of the leaves, three different colored leaves of the same plant were selected as study material (red, middle and green leaves). Ten leaves were collected from each group and leaves with a similar location and color on the branches were selected. The test materials were divided into two parts: one half of the samples was quickly scanned for leaf color parameters, and the other half of the samples was quickly fixed with liquid nitrogen and later moved to storage at −80 °C.
Leaf color was quantified in accordance with the International Commission on Illumination (CII) color standard. Luminosity (L*), a*, and b* values were obtained through the use of a CR30 colorimeter (CHNSpec, Hanghzhou China). After calibration using the colorimetric plate, five points were randomly selected and averaged on each blade, with the objective of avoiding the leaf edges and radial main veins . The exercise was repeated five times for each leaf color. The meanings of the three parameters are as follows. L* indicates the brightness of the color, where a positive number means whitish and a negative number means blackish. Additionally, a* indicates the red–green value, where a positive value means the color is redder and a negative value means it is greener. Finally, b* indicates the yellow–blue value, where positive values are yellowish and negative values are bluish.
First, the sample was weighed accurately in a 2 mL centrifuge tube and 600 µL of MeOH containing 2-Amino-3-(2-chloro-phenyl)-propionic acid (4 ppm) was added and vortexed for 30 s. Second, The sample was then placed in a tissue grinder and ground at 55 Hz for 60 s, followed by sonication at room temperature for 15 min. Finally, the sample was centrifuged at 12,000 rpm at 4 °C for 10 min on a H1850-R refrigerated centrifuge (Cence, Changsha, China), and the supernatant was taken through a 0.22 μm membrane and added into the detection vial for LC-MS detection .
The LC analysis was performed on a Vanquish UHPLC System (Thermo Fisher Scientific, Waltham, MA, USA). Chromatography was carried out with an ACQUITY UPLC ® HSS T3 (2.1 × 100 mm, 1.8 µm) (Waters, Milford, MA, USA). The column was maintained at 40 °C. The flow rate and injection volume were set at 0.3 mL/min and 2 μL, respectively. Mass spectrometric detection of metabolites was performed on Q Exactive (Thermo Fisher Scientific, Waltham, MA, USA) with an ESI ion source. Simultaneous MS1 and MS/MS (full MS-ddMS2 mode, data-dependent MS/MS) acquisition was used. The parameters were as follows: capillary temperature, 325 °C; MS1 range, m / z 100–1000; MS1 resolving power, 70,000 FWHM; number of data-dependent scans per cycle, 10; MS/MS resolving power, FWHM; normalized collision energy, 30 eV; dynamic exclusion time, automatic. The raw mass spectrometry downcomer files were converted into the mzXML file format using the MSConvert tool in the Proteowizard software package (v3.0.8789). Peak detection, peak filtering and peak alignment were carried out using the R XCMS software package (v3.12.0), and a list of substances for quantification was obtained. Substances with coefficients of variation smaller than 30% in the QC samples were then retained for subsequent analysis. The molecular weight of the metabolite was ascertained by means of the mass-to-charge ratio of the parent ion present in the primary mass spectrum. Furthermore, the molecular formula was deduced on the basis of the mass number deviation, as well as the information provided by the additional ions. Following this, the metabolite was matched with a previously existing record in a database, thereby achieving its preliminary identification. Meanwhile, the metabolites detected in the secondary spectrum were subjected to a process of secondary identification. This involved the matching of the metabolite data with the information contained in the database, including the fragment ions of each metabolite.
The mRNA with polyA structure in the total RNA was enriched by Oligo(dT) magnetic beads. Subsequently, the RNA was subjected to ionic interruption, which fragmented it into fragments of approximately 300 base pairs in length. The initial cDNA strand was synthesized using RNA as a template with a 6-base random primer and reverse transcriptase. The second cDNA strand was then synthesized with the initial cDNA strand serving as the template. Following the construction of the library, the library fragments were amplified by PCR. The library size was 450 bp, and the total and effective concentrations were subsequently determined by an Agilent 2100 Bioanalyzer (Agilent, Santa Clara, CA, USA). Subsequently, the mixed libraries were uniformly diluted to 2 nM and denatured by alkaline denaturation, thereby forming single-stranded libraries. Following the extraction, purification and construction of the libraries, the libraries were subjected to paired-end sequencing using next-generation sequencing (NCBI project: PRJNA1196921).
FPKM was used to normalize the raw gene expression calculation. In order to identify genes that were differentially expressed among the three different groups, we identified genes with |log 2 foldchang| > 1 and a p -value < 0.05 as DEGs. GO functional enrichment analysis and KEGG pathway analysis were performed on the confirmed DEGs.
Primers specific to structural genes involved in flavonoid biosynthesis were designed for qRT-PCR analysis using Primer Premier 5 software . All samples were subjected to three replicates, as were three technical replicates. The internal control genes employed were actin and β-tubulin . The relative expression levels of the target genes were calculated employing the 2 −∆∆Ct methodology. Three experimental replicates were performed for each sample.
The secondary structure of the protein encoding the key enzyme gene for flavonoids and anthocyanin synthesis was predicted using the SOPMA online tool ( https://npsa.lyon.inserm.fr/cgi-bin/npsa_automat.pl?page=/NPSA/npsa_sopma.html , accessed on 4 October 2024) and modeled for the tertiary structure using the Phyre2 online tool ( http://www.sbg.bio.ic.ac.uk/phyre2/ , accessed on 4 October 2024) for tertiary structure modeling. The proteins were also functionally analyzed using the website InterPro ( https://www.ebi.ac.uk/interpro/ , accessed on 5 October 2024).
In this study, we focused on the diversity of flavonoid compounds in the red, middle and green leaves of A. truncatum to explore the molecular mechanisms of leaf color formation. The visual diversity of different leaf colors was first described through leaf color parameters, using a digital method. Moreover, a total of 23 different modified flavonoids were detected by metabolomics. In particular, cyanidin 3-O-β-D-sambubioside, cyanidin 3-O rutinoside, pelargonidin 3-O-3″,6″-O-dimalonylglucoside, delphinidin 3,7-di-O-β-D-glucoside and 3-O-β-D-sambubioside could be responsible for the differences between green and red leaves. Furthermore, RNA-seq analysis showed that the up-regulation of CHS , DFR and ANS expression led to an increase in the corresponding anthocyanin red color, which resulted in the reddening of the leaves. Additionally, UGT75C1 was correspondingly essential as a downstream gene for the synthesis of anthocyanin. By modeling the proteins encoded by these genes and analyzing the conserved structural domains, the results corroborated the reliability of the metabolomic and transcriptomic data. Overall, the study offers valuable insights into the flavonoid-related metabolite composition in A. truncatum, providing essential reference points for breeders seeking to enhance the pigmentation of this species.
|
Policy and care in tandem: structuring youth volunteerism for psychological benefits in pediatric palliative care | e2242c88-3310-4c11-8dd5-a3006aa28ec3 | 11907901 | Pediatrics[mh] | Terminally ill children (TIC), particularly those abandoned by their parents and residing in a pediatric palliative care center in southern China, are adversely impacted by delays in self-discovery. This issue is primarily due to enforced geographical isolation in remote urban areas, compounded by social isolation characterized by limited social interaction beyond their caregivers. The government recommends isolation to protect TIC’s health, especially during the COVID-19 and post-pandemic period. A policy recommendation suggests establishing a registration system for all personnel entering and existing, and avoiding unnecessary visits . This measure aims to safeguard their fragile health, as they are more vulnerable to secondary infections. For instance, one such risk is pulmonary tuberculosis, which can be transmitted through airborne droplets from visitors carrying Mycobacterium tuberculosis, posing serious harm to immunosuppressed TICs . However, an unintended consequence of protective isolation is psychological developmental delays, characterized by a limited understanding of themselves and the external world . In palliative care centers, managers and caregivers serve as the legal guardians and often assume parental roles for these children, primarily because many are orphans. They strive to meet the needs of TICs and avoid conflicts, given the children’s limited life expectancy. This situation leads to prevalent hyper-parenting behaviors among the caregiving staff, who, through overprotection and restricted autonomy, unintentionally limit the TIC’s ability to navigate and internalize norms of interpersonal relationships . Consequently, there is a tendency towards self-centeredness, as the focus is predominantly on safeguarding their physical health rather than fostering psychological development . The lack of opportunities for socialization in external environments further exacerbates delays in the self-discovery process, inhibiting the TIC from engaging in enriching interactions like playing with peers and exploring diverse settings . These constrained childhood experiences adversely affect children’s understanding of their identity, including talents, interests, and vision by limiting their exploration of the world . The impact of peer interaction in the self-discovery process Self-discovery, which includes identifying personal interests, character traits, and social roles, plays a crucial role in mental health, particularly during adolescence. This process aids in understanding social interactions and daily functioning, mainly including living habits and recreational activities . It is important for identity formation, a process of understanding one’s self within a social context, where group interactions play a facilitating role . Adolescents typically experience a transition period where their life focus shifts from family members to peers , moving from parent-dependent self-definition to a self-identity integrated within society . Peer interactions are critical in developing social-emotional competence, enabling adolescents to self-regulate and maintain relationships . Moreover, the quality of peer relationships significantly influences this process. Positive peer interactions can enhance mental health and psychological well-being, reducing issues like social anxiety and depression . However, interactions with at-risk peers can negatively affect the TIC’s identity formation . This emphasizes the need for careful selection and training in volunteerism. Volunteerism’s impact on pediatric palliative care Acknowledging the delays in the self-discovery process of TIC compared to their peers highlights their unique developmental challenges. Children raised in institutions often experience developmental delays and psychological deprivation, and positive peer interactions are a viable solution . In this context, China’s Young Pioneer Voluntary Teams (YPVTs) in pediatric palliative care centers emerge as a promising strategy to address these issues. The YPVTs are groups of adolescents who engage in voluntary activities as a form of social practice, which are under the control of the Ministry of Education . YPVTs typically volunteer during weekends and summer and winter holidays, ensuring consistent operation of these programs. Traditional voluntary activities in the health sector include outpatient medical guidance and assistance with self-service registration, with the goal of equipping adolescents with healthcare knowledge. However, these programs often fall short in imparting practical health knowledge, underscoring the additional purpose of volunteer intervention in pediatric palliative care. Moving to the impact on TIC, research by Haski-Leventhal et al. emphasized the essential role of youth volunteerism in addressing social exclusion and discord among TIC, highlighting that, in contrast to the service-oriented approach of adults, youth volunteering tends to be relationship-oriented . Unlike their adult counterparts, teenage volunteers within the YPVTs are less inclined towards excessive coddling, thereby enabling TIC to cultivate interpersonal skills and emotional intelligence through more naturalistic social interactions. Moreover, the proximity in age and interests between YPVTs and TIC facilitates a more authentic understanding and engagement . These volunteers bring diverse experiences, from understanding popular culture to sharing travel stories, enriching TIC’s learning and broadening their worldviews. This facilitates personal identity development and a deeper self-discovery process in TIC. The viewpoints of YPVTs bring to light new challenges, significantly influencing the development of interventions within pediatric care. Further exploring the volunteers’ perspective, YPVTs also experience benefits in terms of psychological well-being. Engaging in voluntary programs allows YPVTs to interact with people from diverse socio-economic backgrounds, which gives them a sense of community and being needed. Moreover, lower possibility of depression and higher life satisfaction and self-esteem is linked with youth volunteering . However, challenges exist in the implementation of YPVTs. Inconsistencies in volunteer selection and training can lead to unpreparedness and behaviors that are not conducive to the sensitive environment required for TIC. Inadequate psychological readiness may result in inappropriate interactions, impacting the TIC’s psychological well-being and hindering their growth. Additionally, volunteers may experience psychological distress, including anxiety, depression, and even post-traumatic stress disorder, when confronted with the suffering or death of the children they assist . This issue is exacerbated by the lack of comprehensive death education in China, emphasizing the need for rigorous selection and training processes . Another concern is the high turnover rate among volunteers, which can disrupt the continuity of care and reduce service quality . Increased investment in the selection and training processes is necessary to enhance the overall performance of organizations, as it directly impacts the quality of care provided . Research has investigated adult volunteer selection and training in palliative care and the effect of youth volunteering, but most of them are fragmented, indicating an absence of research focus on the impact of youth volunteering on TIC and a lack of systematic frameworks for adolescent volunteers working in pediatric palliative care. Studies like Niinomi demonstrate the effectiveness of training programs in enhancing volunteers’ confidence in providing palliative care through a series of five progressive lectures, but these methods are not directly transferable to adolescent volunteers due to differences in age and maturity. Training content for adolescent volunteers should be simpler and more relatable, emphasizing easy-to-grasp concepts in pediatric palliative care. Furthermore, the incorporation of adolescent volunteers’ perspectives, crucial for feedback collection, has been largely overlooked in policy-making . A formal feedback mechanism for regular volunteerism is essential for efficient improvement. Zeanah also highlighted the issue of attachment disruption in orphanage volunteering, where short-term volunteer relationships can negatively impact children’s psychological well-being, leading to poor social behavior and even serious psychiatric disorders. To address this gap, this paper advocates for youth volunteerism as a transformative element of pediatric palliative care. A comprehensive framework aims to achieve mutual psychological development in both YPVTs and TIC. The effectiveness of this framework will be assessed through two distinct psychological questionnaires, tailored respectively to the YPVTs and TIC, to understand the benefits each group derived from the volunteer program and guide future enhancement.
Self-discovery, which includes identifying personal interests, character traits, and social roles, plays a crucial role in mental health, particularly during adolescence. This process aids in understanding social interactions and daily functioning, mainly including living habits and recreational activities . It is important for identity formation, a process of understanding one’s self within a social context, where group interactions play a facilitating role . Adolescents typically experience a transition period where their life focus shifts from family members to peers , moving from parent-dependent self-definition to a self-identity integrated within society . Peer interactions are critical in developing social-emotional competence, enabling adolescents to self-regulate and maintain relationships . Moreover, the quality of peer relationships significantly influences this process. Positive peer interactions can enhance mental health and psychological well-being, reducing issues like social anxiety and depression . However, interactions with at-risk peers can negatively affect the TIC’s identity formation . This emphasizes the need for careful selection and training in volunteerism.
Acknowledging the delays in the self-discovery process of TIC compared to their peers highlights their unique developmental challenges. Children raised in institutions often experience developmental delays and psychological deprivation, and positive peer interactions are a viable solution . In this context, China’s Young Pioneer Voluntary Teams (YPVTs) in pediatric palliative care centers emerge as a promising strategy to address these issues. The YPVTs are groups of adolescents who engage in voluntary activities as a form of social practice, which are under the control of the Ministry of Education . YPVTs typically volunteer during weekends and summer and winter holidays, ensuring consistent operation of these programs. Traditional voluntary activities in the health sector include outpatient medical guidance and assistance with self-service registration, with the goal of equipping adolescents with healthcare knowledge. However, these programs often fall short in imparting practical health knowledge, underscoring the additional purpose of volunteer intervention in pediatric palliative care. Moving to the impact on TIC, research by Haski-Leventhal et al. emphasized the essential role of youth volunteerism in addressing social exclusion and discord among TIC, highlighting that, in contrast to the service-oriented approach of adults, youth volunteering tends to be relationship-oriented . Unlike their adult counterparts, teenage volunteers within the YPVTs are less inclined towards excessive coddling, thereby enabling TIC to cultivate interpersonal skills and emotional intelligence through more naturalistic social interactions. Moreover, the proximity in age and interests between YPVTs and TIC facilitates a more authentic understanding and engagement . These volunteers bring diverse experiences, from understanding popular culture to sharing travel stories, enriching TIC’s learning and broadening their worldviews. This facilitates personal identity development and a deeper self-discovery process in TIC. The viewpoints of YPVTs bring to light new challenges, significantly influencing the development of interventions within pediatric care. Further exploring the volunteers’ perspective, YPVTs also experience benefits in terms of psychological well-being. Engaging in voluntary programs allows YPVTs to interact with people from diverse socio-economic backgrounds, which gives them a sense of community and being needed. Moreover, lower possibility of depression and higher life satisfaction and self-esteem is linked with youth volunteering . However, challenges exist in the implementation of YPVTs. Inconsistencies in volunteer selection and training can lead to unpreparedness and behaviors that are not conducive to the sensitive environment required for TIC. Inadequate psychological readiness may result in inappropriate interactions, impacting the TIC’s psychological well-being and hindering their growth. Additionally, volunteers may experience psychological distress, including anxiety, depression, and even post-traumatic stress disorder, when confronted with the suffering or death of the children they assist . This issue is exacerbated by the lack of comprehensive death education in China, emphasizing the need for rigorous selection and training processes . Another concern is the high turnover rate among volunteers, which can disrupt the continuity of care and reduce service quality . Increased investment in the selection and training processes is necessary to enhance the overall performance of organizations, as it directly impacts the quality of care provided . Research has investigated adult volunteer selection and training in palliative care and the effect of youth volunteering, but most of them are fragmented, indicating an absence of research focus on the impact of youth volunteering on TIC and a lack of systematic frameworks for adolescent volunteers working in pediatric palliative care. Studies like Niinomi demonstrate the effectiveness of training programs in enhancing volunteers’ confidence in providing palliative care through a series of five progressive lectures, but these methods are not directly transferable to adolescent volunteers due to differences in age and maturity. Training content for adolescent volunteers should be simpler and more relatable, emphasizing easy-to-grasp concepts in pediatric palliative care. Furthermore, the incorporation of adolescent volunteers’ perspectives, crucial for feedback collection, has been largely overlooked in policy-making . A formal feedback mechanism for regular volunteerism is essential for efficient improvement. Zeanah also highlighted the issue of attachment disruption in orphanage volunteering, where short-term volunteer relationships can negatively impact children’s psychological well-being, leading to poor social behavior and even serious psychiatric disorders. To address this gap, this paper advocates for youth volunteerism as a transformative element of pediatric palliative care. A comprehensive framework aims to achieve mutual psychological development in both YPVTs and TIC. The effectiveness of this framework will be assessed through two distinct psychological questionnaires, tailored respectively to the YPVTs and TIC, to understand the benefits each group derived from the volunteer program and guide future enhancement.
A four-step framework to improve volunteerism in pediatric palliative care To address identified deficiencies and optimize the contributions of YPVTs, a four-step framework is proposed. It is grounded in three important virtues: psychological resilience, respectful demeanour, and a proclivity for enthusiastic sharing. Firstly, psychological resilience is crucial as it promotes mutual psychological well-being among TIC and YPVTs, allowing volunteers to navigate the complexities of rare diseases without undue distress . This resilience is vital in maintaining a supportive presence and safeguarding TIC from potential negative reactions. Secondly, a respectful attitude is imperative for building a foundation of mutual trust. Physical demeanor, such as maintaining eye contact, is an essential expression of this respect, which contributes to creating an environment conducive to equitable communication . Lastly, the embodiment of sharing enthusiasm is also important to facilitate interaction between TIC and YPVTs and further encourage TIC’s self-discovery process. The volunteers are required to show a willingness to openly share their interests, talents, and experiences, cultivating a rich exchange of perspectives. This enthusiastic sharing fosters a reciprocal interaction, encouraging TIC in their journey of self-discovery while also allowing volunteers to glean deeper insights into the lived experiences of the TIC. This framework (Fig. ) includes meticulous selection processes, comprehensive training program, and dedicated follow-up procedures. Each stage is designed to reinforce these core qualities, ensuring that volunteers are well-prepared. The goal is to minimize the risks of adverse impacts and establish a platform that facilitates the recognition and constructive utilization of the YPVT’s insights.
To address identified deficiencies and optimize the contributions of YPVTs, a four-step framework is proposed. It is grounded in three important virtues: psychological resilience, respectful demeanour, and a proclivity for enthusiastic sharing. Firstly, psychological resilience is crucial as it promotes mutual psychological well-being among TIC and YPVTs, allowing volunteers to navigate the complexities of rare diseases without undue distress . This resilience is vital in maintaining a supportive presence and safeguarding TIC from potential negative reactions. Secondly, a respectful attitude is imperative for building a foundation of mutual trust. Physical demeanor, such as maintaining eye contact, is an essential expression of this respect, which contributes to creating an environment conducive to equitable communication . Lastly, the embodiment of sharing enthusiasm is also important to facilitate interaction between TIC and YPVTs and further encourage TIC’s self-discovery process. The volunteers are required to show a willingness to openly share their interests, talents, and experiences, cultivating a rich exchange of perspectives. This enthusiastic sharing fosters a reciprocal interaction, encouraging TIC in their journey of self-discovery while also allowing volunteers to glean deeper insights into the lived experiences of the TIC. This framework (Fig. ) includes meticulous selection processes, comprehensive training program, and dedicated follow-up procedures. Each stage is designed to reinforce these core qualities, ensuring that volunteers are well-prepared. The goal is to minimize the risks of adverse impacts and establish a platform that facilitates the recognition and constructive utilization of the YPVT’s insights.
Selection process The selection process for YPVTs consists of two stages: a questionnaire and an interview. Initially, candidates complete a series of closed-ended questions covering personal background, interests, and talents, providing program managers with preliminary insights. Further, the process mainly evaluates volunteers’ qualities of psychological resilience and a proclivity for enthusiastic sharing by incorporating the General Self-Efficacy Scale (GSES) (Fig. ) and the Motivation to Volunteer Scale (MTVS) (Fig. ), both pivotal in assessing crucial traits for volunteering. The GSES is a 10-item questionnaire using a four-point Likert scale to measure self-efficacy, reflecting a volunteer’s commitment to caring for TIC and their ability to effectively navigate challenges while volunteering . Each item is rated as “Not at all true,” “Hardly true,” “Moderately true,” and “Exactly true.” “Not at all true” indicates no confidence in managing the described situation or task. “Hardly true” suggests minimal confidence, with occasional capability under limited conditions. “Moderately true” reflects a moderate belief in handling challenges, indicating capability in many situations, though not consistently. “Exactly true” signifies strong confidence, showing the ability to effectively dealing with challenges and tasks in most situations. Higher GSES scores indicate commendable psychological resilience among YPVTs. Considering the nature of volunteerism, special attention should be paid to voluntary motivation. The MTVS uses the same scoring method to assess the candidates’ voluntary commitment to the program . It categorizes motivation into voluntary and involuntary, reflecting their enthusiasm when meeting TIC. An evaluative approach will be applied to the MTVS by computing the average scores for both voluntary (e.g., “Volunteering makes me feel good about myself”) and involuntary motivation (e.g., “They told me to volunteer in the school”), which is to assess YPVTs' underlying motivation. Those who are self-motivated will be preferred. To achieve the accurate assessment, the GSES and MTVS must be precisely translated and culturally adapted for Chinese youth volunteers. This process involves multiple steps to guarantee accuracy and cultural appropriateness. The translation team is composed of at least two independent subgroups, including native speakers of both English and Chinese with bicultural backgrounds. This diversity allows culturally specific terms, such as “self-efficacy,” to be appropriately translated and adapted to the Chinese cultural context . Team members experienced with the target population help align terminology with the group’s understanding and needs. Professional child psychologists play a key role in refining the translations for clarity and comprehensibility. The independent teams then compare their translations and reach a consensus, followed by a separate team’s back-translation into the original language. Finally, pilot testing with a small sample identifies and resolves any ambiguities or misunderstandings, ensuring the final versions of the scales are accurate and contextually appropriate for the intended population . With the translation process confirming cultural relevance and clarity, the refined scales are now poised for effective deployment in evaluating candidates. When considering the two scales comprehensively, criteria for identifying “relatively lower scores” on the GSES will be established through using normative data and statistical benchmarks derived from the pilot tests. Previous research with a sample size of 9,578 students in China indicates a mean of 28.75, which is used as a benchmark to measure the self-efficacy level for the YPVTs . Scores below 28.74 may be classified as having a relatively lower self-efficacy level. However, candidates with strong motivation on the MTVS will still be eligible for admission, provided they undergo additional training to build their self-efficacy and better equip them to handle challenges effectively during their volunteer service. The scores from scales will not be the sole determinants for the eligibility of YPVTs to participate in volunteering, unless combining the results from further interviews. The interview phase builds upon the insights gained from these scales. During this stage, evaluators pay attention to how candidates respond and their level of enthusiasm, which are key indicators of their genuineness and suitability. Given that the candidates are adolescents, a trait-based selection method is preferred to minimize stress and potential subjectivity . The use of the GSES and MTVS in this phase ensures a standardized and objective approach to scoring the candidates’ performance. In addition to standardized scales, the selection process incorporates scenario-based interviews. These interviews utilize real-life cases from pediatric palliative care centers, featuring virtual avatars and narrative explanations to create realistic scenarios. The goal is to elicit natural responses from candidates, allowing them to demonstrate their problem-solving abilities in practical situations . Candidates who feel uncomfortable during the interview have the right to withdraw at any time. The materials used in these interviews will be reviewed by volunteer managers and child psychologists to mitigate potential psychological risks, underscoring commitment to safeguarding the mental well-being of all participants. Candidates who exhibit psychological resilience, intrinsic self-motivation, and adeptness in addressing scenario-based interviews will be prioritized for entry into the training process. Conversely, those with transmittable diseases, a history of misconduct, or insufficient time commitment will be excluded from the program. These exclusion criteria are designed to protect the psychological and physical health of TIC and to enhance the overall quality of the volunteer program. This rigorous selection acts as an effective way to protect both TIC’s and YPVTs’ physical and psychological health during the program. Training Upon the conscientious selection, a comprehensive training session is imperative to aptly equip YPVTs with the requisite competencies, especially the three qualities mentioned above, to navigate the complexities of their volunteering roles effectively. This training initiates with an orientation, aiming to imbue the YPVTs with a deeper understanding of the program’s objectives . It’s essential that YPVTs are furnished with foundational knowledge of palliative care paradigms, coupled with an insight into the specificities governing the experiences of the TIC, such as their daily routines and prevalent symptomatic manifestations. This pedagogical approach seeks to promote an atmosphere of familiarity and contextual sensitivity, enabling the YPVTs to better resonate with the challenges faced by the TIC as well as a psychological preparation for themselves . Elaboration on the responsibilities of YPVTs also constitutes a pivotal aspect of the training, ensuring alignment with the program’s ethos and objectives. The training incorporates lectures, workshops and exercises to foster competency and preparedness among the YPVTs. Key topics including “Cultivating Empathetic Perspectives” and “Nurturing Care in Interpersonal Engagements” will be emphasized. These topics are designed to reinforce the three main qualities mentioned in the selection section, and help YPVT to empathize effectively with TIOs . Feedback collection To further refine the program, a sophisticated feedback collection and evaluation framework is proposed, envisioning a synergistic collaboration between various pediatric palliative care centers. Throughout their volunteer engagements, YPVTs acquire invaluable insights into the obscured challenges and needs represented within the palliative environments through observation and communication . Moreover, the personal feelings of YPVTs should be collected to ensure they are not stressful after the interaction with TIC, protecting their psychological well-being . Feedback, reflective of these insights, will be systematically harvested through a variety of methods, including questionnaires, post-engagement interviews and discussions. Such feedback endeavours to drive internal enhancements, tailored specifically to improve the program and daily care provision. To uphold the integrity and authenticity of the feedback, the preliminary collection process will be led by professional’s adept in child psychology. This approach is strategic in circumventing potential biases or power imbalances inherent in adult-centric interpretations, as it acknowledges the propensity for adults to unconsciously position teenagers as their subordinates . Naturalistic settings characterized by peer companionship, where teenagers feel emboldened to express their perspectives, is also important, aiming to foster an environment conducive to open expression and reflection. Utilizing a holistic evaluative lens, results collated from group discussions and interviews will be synergistically analyzed alongside field observations, ensuring a consideration of all pertinent insights . Ultimately, this feedback and evaluation framework is instrumental in driving continuous refinement within this program. Effectiveness evaluation To assess the impact of the four-step framework on the psychological well-being of YPVTs and TIC, three different scales will be administered post-volunteering. For TIC, evaluations will be conducted by volunteer team managers in pediatric palliative care centers with caregiver assistance. The Identity Scale for Adolescents, suitable for individuals aged 13–18, will be employed. This 39-items self-report questionnaire uses a 4-point Likert rating scale 0 (never), 1 (rarely), 2 (sometimes) and 3 (often) to categorize adolescents into three personality types: positive, negative, and arrogant self-identity . It measures identity formation, with an expectation of a shift towards a positive self-identity, characterized by sociability and optimism, after interactions with YPVTs. Negative identity shows a lack of confidence and social skills and arrogant identity indicates egoistic and feeling superior which potentially results from hyper-parenting behaviors. In evaluating YPVTs, the GSES and MTVS will be reapplied to develop deeper into their motivations and the changes they experience through the program. This approach aims to assess personal traits such as dedication and professionalism among the youth volunteers, ensuring alignment with the framework’s objectives. By comparing the results obtained during the selection process with those gathered post-program, any increase in scores will indicate an improvement in self-efficacy, a desirable outcome of the volunteer experience. In addition, it is anticipated that responses on the MTVS will shift towards affirmations like “Volunteering makes me feel good about myself” and “It will help me in the future,” reflecting a positive transformation in their perception of volunteer work. Opposite responses would indicate the potential reasons for dropping out of the voluntary program, where more attention should be paid to retaining the youth volunteers. These two scales can be used to regularly measure and track the volunteers' self-efficacy and motivation. This ensures they are engaged and find meaning in their work, while also helping to identify suitable roles and responsibilities for future adaptation. For instance, volunteers who resonate with the statement “I can handle whatever comes my way” and score high in feeling good about themselves through volunteering could be encouraged to take on leadership roles within the volunteer group. They can also provide immediate feedback on the impact of their work. Moreover, this approach helps identify when volunteers might need additional support or a change in role to stay motivated. Such a strategy not only aids in their self-discovery process but also contributes to fostering a culture of continuous improvement within the volunteer program. In practical terms, these self-report scales will be utilized for self-evaluation by the young volunteers. They will rate their agreement on a Likert scale, ranging from 1 (“Not at all true”) to 4 (“Exactly true”), under condition of confidentiality to encourage honest responses. Post-volunteering, these scales will be re-administered to assess any changes in self-efficacy and motivations. This approach provides valuable insights into the program’s impact and the volunteers’ development. The longitudinal data obtained will help in understanding the volunteers’ personal traits, their dedication, and professionalism. This information is significant for informing the refinement of future volunteer training and the overall development of the program. Limitations One of the challenges in youth volunteerism is the high turnover rate. The first step to controlling the turnover rate is to improve the recruiting process to gain more qualified candidates since recruitment and retention are two interrelated processes . Moreover, peer groups positively strengthen the relationships within the volunteering groups, improve their commitment to the task, and help control the turnover rate . Additionally, the volunteers in palliative care centers, who bear the sadness of witnessing their clients suffering, tend to be overstressed psychologically, being a major reason for withdrawing from the volunteering program . Timely counselling and taking time off are common ways to release stress, indicating that professional psychological support and a proper volunteering schedule may help volunteers maintain a good status and retain them . In the feedback collection section, YPVTs should be interviewed about their motivation to join and reasons for leaving the program, which can comprehensively improve the framework from both TIC’s and YPVTs’ perspectives, striving for a sustainable program . There might be some potential flaws or weaknesses in the framework worth noting. In the selection section, the threshold for determining YPVTs’ eligibility to enroll in the program is unclear due to the limited data available. A significant issue in the framework evaluation section is the self-assessment bias inherent in these tools. Since all motivation items and statements are self-reported, responses could be influenced by the individual’s current mood or self-perception. This might lead to an inaccurate portrayal of one’s true capabilities or motivations. Furthermore, complex personal traits such as dedication and professionalism may not be fully captured through self-reporting, as such qualities are often better demonstrated through actions rather than introspection. Moreover, governmental regulations pose challenges in data collection from local palliative care centers in collaboration with welfare institutions, primarily due to concerns over data and privacy protection. To address these limitations, the questionnaire scores should be collected in the pilot experiments to determine a more reliable threshold by calculating the range of scores of youth volunteers who withdraw from the program. The incorporation of performance-based assessment with multiple raters is planned. This approach will include assessments from peers, supervisors, or community members, providing a more objective measure of the volunteer's traits and performance. Additionally, the process of synergistically analyzing feedback from various sources, including group discussions, interviews, and observations, though complex, is essential. Ensuring accurate reflection of all viewpoints in this data synthesis is significant for a comprehensive evaluation. Combining both qualitative feedback collection and quantitative methods, such as graded scales, offers a more balanced approach to evaluation.
The selection process for YPVTs consists of two stages: a questionnaire and an interview. Initially, candidates complete a series of closed-ended questions covering personal background, interests, and talents, providing program managers with preliminary insights. Further, the process mainly evaluates volunteers’ qualities of psychological resilience and a proclivity for enthusiastic sharing by incorporating the General Self-Efficacy Scale (GSES) (Fig. ) and the Motivation to Volunteer Scale (MTVS) (Fig. ), both pivotal in assessing crucial traits for volunteering. The GSES is a 10-item questionnaire using a four-point Likert scale to measure self-efficacy, reflecting a volunteer’s commitment to caring for TIC and their ability to effectively navigate challenges while volunteering . Each item is rated as “Not at all true,” “Hardly true,” “Moderately true,” and “Exactly true.” “Not at all true” indicates no confidence in managing the described situation or task. “Hardly true” suggests minimal confidence, with occasional capability under limited conditions. “Moderately true” reflects a moderate belief in handling challenges, indicating capability in many situations, though not consistently. “Exactly true” signifies strong confidence, showing the ability to effectively dealing with challenges and tasks in most situations. Higher GSES scores indicate commendable psychological resilience among YPVTs. Considering the nature of volunteerism, special attention should be paid to voluntary motivation. The MTVS uses the same scoring method to assess the candidates’ voluntary commitment to the program . It categorizes motivation into voluntary and involuntary, reflecting their enthusiasm when meeting TIC. An evaluative approach will be applied to the MTVS by computing the average scores for both voluntary (e.g., “Volunteering makes me feel good about myself”) and involuntary motivation (e.g., “They told me to volunteer in the school”), which is to assess YPVTs' underlying motivation. Those who are self-motivated will be preferred. To achieve the accurate assessment, the GSES and MTVS must be precisely translated and culturally adapted for Chinese youth volunteers. This process involves multiple steps to guarantee accuracy and cultural appropriateness. The translation team is composed of at least two independent subgroups, including native speakers of both English and Chinese with bicultural backgrounds. This diversity allows culturally specific terms, such as “self-efficacy,” to be appropriately translated and adapted to the Chinese cultural context . Team members experienced with the target population help align terminology with the group’s understanding and needs. Professional child psychologists play a key role in refining the translations for clarity and comprehensibility. The independent teams then compare their translations and reach a consensus, followed by a separate team’s back-translation into the original language. Finally, pilot testing with a small sample identifies and resolves any ambiguities or misunderstandings, ensuring the final versions of the scales are accurate and contextually appropriate for the intended population . With the translation process confirming cultural relevance and clarity, the refined scales are now poised for effective deployment in evaluating candidates. When considering the two scales comprehensively, criteria for identifying “relatively lower scores” on the GSES will be established through using normative data and statistical benchmarks derived from the pilot tests. Previous research with a sample size of 9,578 students in China indicates a mean of 28.75, which is used as a benchmark to measure the self-efficacy level for the YPVTs . Scores below 28.74 may be classified as having a relatively lower self-efficacy level. However, candidates with strong motivation on the MTVS will still be eligible for admission, provided they undergo additional training to build their self-efficacy and better equip them to handle challenges effectively during their volunteer service. The scores from scales will not be the sole determinants for the eligibility of YPVTs to participate in volunteering, unless combining the results from further interviews. The interview phase builds upon the insights gained from these scales. During this stage, evaluators pay attention to how candidates respond and their level of enthusiasm, which are key indicators of their genuineness and suitability. Given that the candidates are adolescents, a trait-based selection method is preferred to minimize stress and potential subjectivity . The use of the GSES and MTVS in this phase ensures a standardized and objective approach to scoring the candidates’ performance. In addition to standardized scales, the selection process incorporates scenario-based interviews. These interviews utilize real-life cases from pediatric palliative care centers, featuring virtual avatars and narrative explanations to create realistic scenarios. The goal is to elicit natural responses from candidates, allowing them to demonstrate their problem-solving abilities in practical situations . Candidates who feel uncomfortable during the interview have the right to withdraw at any time. The materials used in these interviews will be reviewed by volunteer managers and child psychologists to mitigate potential psychological risks, underscoring commitment to safeguarding the mental well-being of all participants. Candidates who exhibit psychological resilience, intrinsic self-motivation, and adeptness in addressing scenario-based interviews will be prioritized for entry into the training process. Conversely, those with transmittable diseases, a history of misconduct, or insufficient time commitment will be excluded from the program. These exclusion criteria are designed to protect the psychological and physical health of TIC and to enhance the overall quality of the volunteer program. This rigorous selection acts as an effective way to protect both TIC’s and YPVTs’ physical and psychological health during the program.
Upon the conscientious selection, a comprehensive training session is imperative to aptly equip YPVTs with the requisite competencies, especially the three qualities mentioned above, to navigate the complexities of their volunteering roles effectively. This training initiates with an orientation, aiming to imbue the YPVTs with a deeper understanding of the program’s objectives . It’s essential that YPVTs are furnished with foundational knowledge of palliative care paradigms, coupled with an insight into the specificities governing the experiences of the TIC, such as their daily routines and prevalent symptomatic manifestations. This pedagogical approach seeks to promote an atmosphere of familiarity and contextual sensitivity, enabling the YPVTs to better resonate with the challenges faced by the TIC as well as a psychological preparation for themselves . Elaboration on the responsibilities of YPVTs also constitutes a pivotal aspect of the training, ensuring alignment with the program’s ethos and objectives. The training incorporates lectures, workshops and exercises to foster competency and preparedness among the YPVTs. Key topics including “Cultivating Empathetic Perspectives” and “Nurturing Care in Interpersonal Engagements” will be emphasized. These topics are designed to reinforce the three main qualities mentioned in the selection section, and help YPVT to empathize effectively with TIOs .
To further refine the program, a sophisticated feedback collection and evaluation framework is proposed, envisioning a synergistic collaboration between various pediatric palliative care centers. Throughout their volunteer engagements, YPVTs acquire invaluable insights into the obscured challenges and needs represented within the palliative environments through observation and communication . Moreover, the personal feelings of YPVTs should be collected to ensure they are not stressful after the interaction with TIC, protecting their psychological well-being . Feedback, reflective of these insights, will be systematically harvested through a variety of methods, including questionnaires, post-engagement interviews and discussions. Such feedback endeavours to drive internal enhancements, tailored specifically to improve the program and daily care provision. To uphold the integrity and authenticity of the feedback, the preliminary collection process will be led by professional’s adept in child psychology. This approach is strategic in circumventing potential biases or power imbalances inherent in adult-centric interpretations, as it acknowledges the propensity for adults to unconsciously position teenagers as their subordinates . Naturalistic settings characterized by peer companionship, where teenagers feel emboldened to express their perspectives, is also important, aiming to foster an environment conducive to open expression and reflection. Utilizing a holistic evaluative lens, results collated from group discussions and interviews will be synergistically analyzed alongside field observations, ensuring a consideration of all pertinent insights . Ultimately, this feedback and evaluation framework is instrumental in driving continuous refinement within this program.
To assess the impact of the four-step framework on the psychological well-being of YPVTs and TIC, three different scales will be administered post-volunteering. For TIC, evaluations will be conducted by volunteer team managers in pediatric palliative care centers with caregiver assistance. The Identity Scale for Adolescents, suitable for individuals aged 13–18, will be employed. This 39-items self-report questionnaire uses a 4-point Likert rating scale 0 (never), 1 (rarely), 2 (sometimes) and 3 (often) to categorize adolescents into three personality types: positive, negative, and arrogant self-identity . It measures identity formation, with an expectation of a shift towards a positive self-identity, characterized by sociability and optimism, after interactions with YPVTs. Negative identity shows a lack of confidence and social skills and arrogant identity indicates egoistic and feeling superior which potentially results from hyper-parenting behaviors. In evaluating YPVTs, the GSES and MTVS will be reapplied to develop deeper into their motivations and the changes they experience through the program. This approach aims to assess personal traits such as dedication and professionalism among the youth volunteers, ensuring alignment with the framework’s objectives. By comparing the results obtained during the selection process with those gathered post-program, any increase in scores will indicate an improvement in self-efficacy, a desirable outcome of the volunteer experience. In addition, it is anticipated that responses on the MTVS will shift towards affirmations like “Volunteering makes me feel good about myself” and “It will help me in the future,” reflecting a positive transformation in their perception of volunteer work. Opposite responses would indicate the potential reasons for dropping out of the voluntary program, where more attention should be paid to retaining the youth volunteers. These two scales can be used to regularly measure and track the volunteers' self-efficacy and motivation. This ensures they are engaged and find meaning in their work, while also helping to identify suitable roles and responsibilities for future adaptation. For instance, volunteers who resonate with the statement “I can handle whatever comes my way” and score high in feeling good about themselves through volunteering could be encouraged to take on leadership roles within the volunteer group. They can also provide immediate feedback on the impact of their work. Moreover, this approach helps identify when volunteers might need additional support or a change in role to stay motivated. Such a strategy not only aids in their self-discovery process but also contributes to fostering a culture of continuous improvement within the volunteer program. In practical terms, these self-report scales will be utilized for self-evaluation by the young volunteers. They will rate their agreement on a Likert scale, ranging from 1 (“Not at all true”) to 4 (“Exactly true”), under condition of confidentiality to encourage honest responses. Post-volunteering, these scales will be re-administered to assess any changes in self-efficacy and motivations. This approach provides valuable insights into the program’s impact and the volunteers’ development. The longitudinal data obtained will help in understanding the volunteers’ personal traits, their dedication, and professionalism. This information is significant for informing the refinement of future volunteer training and the overall development of the program.
One of the challenges in youth volunteerism is the high turnover rate. The first step to controlling the turnover rate is to improve the recruiting process to gain more qualified candidates since recruitment and retention are two interrelated processes . Moreover, peer groups positively strengthen the relationships within the volunteering groups, improve their commitment to the task, and help control the turnover rate . Additionally, the volunteers in palliative care centers, who bear the sadness of witnessing their clients suffering, tend to be overstressed psychologically, being a major reason for withdrawing from the volunteering program . Timely counselling and taking time off are common ways to release stress, indicating that professional psychological support and a proper volunteering schedule may help volunteers maintain a good status and retain them . In the feedback collection section, YPVTs should be interviewed about their motivation to join and reasons for leaving the program, which can comprehensively improve the framework from both TIC’s and YPVTs’ perspectives, striving for a sustainable program . There might be some potential flaws or weaknesses in the framework worth noting. In the selection section, the threshold for determining YPVTs’ eligibility to enroll in the program is unclear due to the limited data available. A significant issue in the framework evaluation section is the self-assessment bias inherent in these tools. Since all motivation items and statements are self-reported, responses could be influenced by the individual’s current mood or self-perception. This might lead to an inaccurate portrayal of one’s true capabilities or motivations. Furthermore, complex personal traits such as dedication and professionalism may not be fully captured through self-reporting, as such qualities are often better demonstrated through actions rather than introspection. Moreover, governmental regulations pose challenges in data collection from local palliative care centers in collaboration with welfare institutions, primarily due to concerns over data and privacy protection. To address these limitations, the questionnaire scores should be collected in the pilot experiments to determine a more reliable threshold by calculating the range of scores of youth volunteers who withdraw from the program. The incorporation of performance-based assessment with multiple raters is planned. This approach will include assessments from peers, supervisors, or community members, providing a more objective measure of the volunteer's traits and performance. Additionally, the process of synergistically analyzing feedback from various sources, including group discussions, interviews, and observations, though complex, is essential. Ensuring accurate reflection of all viewpoints in this data synthesis is significant for a comprehensive evaluation. Combining both qualitative feedback collection and quantitative methods, such as graded scales, offers a more balanced approach to evaluation.
Due to the relatively recent introduction of pediatric palliative care in China, the application of youth volunteering as an effective intervention to promote the self-discovery process among TIC remains an area not yet extensively explored by researchers and policymakers. A further research direction involves conducting pilot experiments to assess the effectiveness of the proposed framework. This paper presents an innovative approach, advocating for the empowerment of private palliative care centers to address the psychological challenges among disadvantaged children.
|
Comparison of HercepTest™ mAb pharmDx (Dako Omnis, GE001) with Ventana PATHWAY anti-HER-2/neu (4B5) in breast cancer: correlation with | 27c7b30a-e5d9-4eea-99db-0e323fd059c7 | 9636083 | Anatomy[mh] | The human epidermal growth factor receptor 2 (HER2, also referred to as HER2/neu) is one of four members belonging to the epidermal growth factor receptor (EGFR) protein family. The HER2 protein is characterized by its tyrosine kinase activity and the HER2 oncogene controls cell proliferation and apoptosis . Initially described in 1985 by King et al. , HER2 overexpression has been demonstrated to play a major role in the onset, development, and progression of breast cancer (BC). About 15–20% of BC patients show HER2 amplification and/or HER2 over-expression, which are associated with increased tumor aggression and poor prognosis, although these patients are eligible for HER2-directed therapy . HER2-targeted monoclonal antibodies (mAbs) such as trastuzumab and/or pertuzumab (used as single or combined agents, with or without chemotherapy) are now the standard treatment for patients with HER2-positive advanced BC, acting to block the corresponding pathway(s) and provide improved overall survival rates . Beyond the use of these two drugs, novel therapies based on anti-HER2 antibody–drug conjugates (ADCs) have been developed. For example, trastuzumab-emtansin (T-DM1, Kadcyla®) was the first of its kind approved in 2013 by European Medicines Agency (EMA) for HER2 overexpressing and/or amplified advanced metastasized BC . New types of ADCs have recently been developed using trastuzumab linked to novel toxic agents (e.g., deruxtecan, a topoisomerase I inhibitor) (T-DXd, Enhertu®) and have shown efficacy in patients even after T-DM1 therapy failure . Interestingly, there is also evidence that T-DXd is effective in patients with BC exhibiting low levels of HER2 protein as determined by IHC (i.e., HER2 IHC 2 + /non-amplified or IHC 1 +) . Since, almost 40–50% of BC are classifiable as HER2-low , many more patients may benefit from this new type of HER2-targeted therapy (reviewed in ). Methods to screen for eligible BC patients who may benefit from HER2-targeted therapies currently include IHC demonstrating HER2 protein overexpression and in situ hybridization (ISH) to detect HER2 gene amplification. Other methods such as quantitative real-time PCR (transcript amplification) are not recommended for routine patient selection . Available IHC assays are well-established, robust, and inexpensive. While several different antibody clones have been successfully used in clinical trials (e.g., R60, 10H8, and CB11), the Agilent/Dako HercepTest™ pAb pharmDx (Autostainer Link) (HercepTest (poly)) and Ventana PATHWAY® anti-HER-2/neu (4B5) (PATHWAY 4B5) are currently the most widely used IHC assays . Many studies have analyzed the diagnostic value (i.e., sensitivity and specificity) of these two IHC assays for detecting HER2-positive BC by comparing IHC results to the HER2 gene amplification status determined by ISH assays . Accordingly, international guidelines for HER2 testing in BC focus on the correlation between IHC and ISH to reliably select those HER2-positive carcinomas most likely to respond to HER2-directed therapies. Due to the potential broader applicability of current anti-HER2-targeting drugs, the sensitivity of these assays is now of greater importance for selecting eligible patients . In this context, it has become necessary to evaluate the diagnostic utility of HER2 assays with respect to the detection of not only HER2-positive (IHC 3 + and/or amplified) but also HER2-low (IHC 2 + or IHC 1 + , non-amplified) BC cases. In this context, studies comparing the polyclonal HercepTest (poly) and the monoclonal PATHWAY 4B5 have revealed good concordance between the two methods for detection of HER2-positive BC . However, there is evidence that the HercepTest (poly) might be less sensitive in detecting HER2-low BC as compared to the PATHWAY 4B5 assay . Recently, a second-generation, CE-IVD-marked HercepTest™ mAb pharmDx (Dako Omnis) kit (HercepTest (mAb)) has become available in Europe and Canada. This new assay is run on the Dako Omnis staining platform using a monoclonal rabbit antibody (clone DG44) . Interestingly, according to the 2021 NordiQC data, laboratories applying the HercepTest (mAb) achieved highest overall pass rate (100%) . Herein, we report the results of an IHC concordance study comparing the HercepTest (mAb) run on the Dako Omnis platform and the PATHWAY 4B5 assay run on the Ventana BenchMark ULTRA using a BC cohort of 119 samples and assessing assay sensitivity and specificity with respect to amplification status and inter-assay and inter-observer variations.
Sample selection and study design ( Fig. ) The clinical performance of the new HercepTest (mAb) (Agilent Technologies, Santa Clara, CA. USA) was compared with the monoclonal PATHWAY 4B5 assay performance (Ventana) (Roche Tissue Diagnostics, Tucson, AZ. USA) using a selection of 120 BC samples. These commercially acquired tissue blocks were originally pre-tested for their HER2 status by the vendor using either Ventana or Leica antibodies and verified by Agilent Technologies applying HercepTest (mAb) and HercepTest (poly). The testing cohort composed of an equal number of cases representative of HER2 status 0, 1 + , 2 + , and 3 + , respectively ( n = 30/status). Within the IHC score 2 + group, 15 samples were selected to be HER2 amplified, while the remaining samples were non-amplified. Specimens were acquired by Agilent Technologies from Danish hospitals (with ethical permission) and external tissue vendors in USA and Canada (see Vendor list). The specimens were de-identified, and all were fixed in 4% neutral buffered formalin and paraffin-embedded compliant with ASCO/CAP guidelines. The BC specimens were enrolled in the study following assessment of tissue sections stained with H&E, HercepTest (mAb), and HercepTest (poly). A specimen was included if (1) invasive BC tissue and an adequate number of tumor cells (≥ 100) were present, (2) the tissue morphology was adequately preserved, and (3) there was an absence of processing artifacts that would negatively affect the assessment of the HER2 status. Each specimen entered the study with an enrollment IHC score based on HercepTest (mAb) and HercepTest (poly). FISH status for enrollment of amplified and non-amplified specimens was based on information previously provided by the commercial tissue vendor, if these data were available. One tumor sample had an inadequate amount of tumor tissue and was rejected by the observing pathologists; hence, a total of 119 BC specimens were used for this study. The final selection of samples consisted of 103 surgical resections and 16 biopsy specimens. Tumor types included 106 ductal (89.1%), 9 lobular (7.6%), and 4 mucinous (3.4%) carcinomas. IHC results were assigned to each of these 119 samples. HER2 FISH analysis revealed 114 evaluable samples out of 119 tested (see also Supplementary Data ); five of the BC samples produced non-evaluable FISH signals due to sub-optimal tissue pre-analytics despite repeat testing. Sample preparation Twelve tissue sections, 4–5 µm thick, cut from each of the selected specimens were mounted onto Epredia™ SuperFrost Plus™ Microscope Slides. On-slide controls containing HER2-positive (FFPE cell pellet from IHC 2 + cell line MDA-453) and negative (tonsil sample) cores were added to each slide. Mounted tissues were baked at 60 °C for 1 h. Two tissue sections (first and last) from each collected specimen were H&E stained. Immunohistochemistry HercepTest™ mAb pharmDx (Dako Omnis) (GE001) The IHC staining protocol using the HercepTest (mAb) was performed as described by the manufacturer . Freshly cut tissue was processed on the Dako Omnis platform (Agilent Technologies, Santa Clara, CA) together with kit control slides for every staining run, using an automated staining protocol validated for HER2 detection . PATHWAY® anti-HER-2/NEU, clone 4B5 (790–2991) IHC staining using the PATHWAY® HER-2/neu rabbit monoclonal antibody 4B5 was performed according to the recommendations of the manufacturer . Freshly cut tissue was processed on the Ventana BenchMark ULTRA (Ventana Medical Systems, Roche Diagnostics, Tucson, AZ) together with kit control slides for every staining run, using an automated staining protocol validated for HER2 detection . IHC scoring IHC staining for HER2 was independently evaluated by three trained pathologists (IN, MK, JR), followed by a consensus session for discordantly scored samples to define a consensus score for each case. IHC stains of the two assays were read after a 2-week wash-out period, and all the pathologists were blinded to the FISH results. In addition to a pre-study training provided by Dako/Agilent, all investigators had extensive experience in HER2 evaluation, having served over the past 20 years as readers in most of the trastuzumab, pertuzumab, and T-DM1 approval BC studies by Targos GmbH (Kassel, Germany) (for a review of studies screened by first-generation HercepTest (poly), see ). IHC scoring was performed according to the 2018 ASCO/CAP guidelines . Accordingly, cases with complete intense staining in ≤ 10% of tumor cells, as well as cases with intense and lateral or basolateral (“U-type”) staining, were included in the IHC 2 + category. For cases of IHC 1 + staining intensity (i.e., faint/barely perceptible membrane staining), the percentage of stained cells ≤ 10% or > 10% was recorded separately according to Ventana Instructions for Use (IFU) . Intensity scoring was performed by applying the magnification rule as published previously by our group . Fluorescence in situ hybridization assessment HER2 IQFISH pharmDx (K5731) Determination of HER2 gene amplification was analyzed using the HER2 IQFISH pharmDx kit according to the recommendations of the manufacturer . HER2 in situ hybridizations were evaluated by a pathologist (IN) using the updated 2018 ASCO/CAP guidelines. For final interpretation of the FISH data, newly defined ISH groups (1–5) were taken into consideration . Accordingly, group 1 (ratio ≥ 2.0 and gene count ≥ 4.0) and group 3 cases (ratio < 2.0 and gene count ≥ 6.0) with IHC 3 + or IHC 2 + were considered FISH positive. Statistical evaluations For comparison of datasets, chi-square test ( X 2 ) was used with p < 0.05 considered as statistically significant. Test performance was evaluated using FISH as a reference standard. Sensitivity and specificity were calculated as follows: [12pt]{minimal}
$$=100 }{\#+\#}$$ Estimated sensitivity = 100 % × # true positive events # true positive events + # false negative events [12pt]{minimal}
$$=100 }{\#+\#}$$ Estimated specificity = 100 % × # true negative events # false positive events + # true negative events Inter-rater reliability (IRR), defined as the ratio of the total number of agreements among raters and the total number of ratings, was calculated as follows: [12pt]{minimal}
$$[]=\#}{\#*\#} 100$$ IRR % = Total # of agreements Total # of ratings given by each rater ∗ # of raters × 100
Fig. ) The clinical performance of the new HercepTest (mAb) (Agilent Technologies, Santa Clara, CA. USA) was compared with the monoclonal PATHWAY 4B5 assay performance (Ventana) (Roche Tissue Diagnostics, Tucson, AZ. USA) using a selection of 120 BC samples. These commercially acquired tissue blocks were originally pre-tested for their HER2 status by the vendor using either Ventana or Leica antibodies and verified by Agilent Technologies applying HercepTest (mAb) and HercepTest (poly). The testing cohort composed of an equal number of cases representative of HER2 status 0, 1 + , 2 + , and 3 + , respectively ( n = 30/status). Within the IHC score 2 + group, 15 samples were selected to be HER2 amplified, while the remaining samples were non-amplified. Specimens were acquired by Agilent Technologies from Danish hospitals (with ethical permission) and external tissue vendors in USA and Canada (see Vendor list). The specimens were de-identified, and all were fixed in 4% neutral buffered formalin and paraffin-embedded compliant with ASCO/CAP guidelines. The BC specimens were enrolled in the study following assessment of tissue sections stained with H&E, HercepTest (mAb), and HercepTest (poly). A specimen was included if (1) invasive BC tissue and an adequate number of tumor cells (≥ 100) were present, (2) the tissue morphology was adequately preserved, and (3) there was an absence of processing artifacts that would negatively affect the assessment of the HER2 status. Each specimen entered the study with an enrollment IHC score based on HercepTest (mAb) and HercepTest (poly). FISH status for enrollment of amplified and non-amplified specimens was based on information previously provided by the commercial tissue vendor, if these data were available. One tumor sample had an inadequate amount of tumor tissue and was rejected by the observing pathologists; hence, a total of 119 BC specimens were used for this study. The final selection of samples consisted of 103 surgical resections and 16 biopsy specimens. Tumor types included 106 ductal (89.1%), 9 lobular (7.6%), and 4 mucinous (3.4%) carcinomas. IHC results were assigned to each of these 119 samples. HER2 FISH analysis revealed 114 evaluable samples out of 119 tested (see also Supplementary Data ); five of the BC samples produced non-evaluable FISH signals due to sub-optimal tissue pre-analytics despite repeat testing.
Twelve tissue sections, 4–5 µm thick, cut from each of the selected specimens were mounted onto Epredia™ SuperFrost Plus™ Microscope Slides. On-slide controls containing HER2-positive (FFPE cell pellet from IHC 2 + cell line MDA-453) and negative (tonsil sample) cores were added to each slide. Mounted tissues were baked at 60 °C for 1 h. Two tissue sections (first and last) from each collected specimen were H&E stained.
HercepTest™ mAb pharmDx (Dako Omnis) (GE001) The IHC staining protocol using the HercepTest (mAb) was performed as described by the manufacturer . Freshly cut tissue was processed on the Dako Omnis platform (Agilent Technologies, Santa Clara, CA) together with kit control slides for every staining run, using an automated staining protocol validated for HER2 detection . PATHWAY® anti-HER-2/NEU, clone 4B5 (790–2991) IHC staining using the PATHWAY® HER-2/neu rabbit monoclonal antibody 4B5 was performed according to the recommendations of the manufacturer . Freshly cut tissue was processed on the Ventana BenchMark ULTRA (Ventana Medical Systems, Roche Diagnostics, Tucson, AZ) together with kit control slides for every staining run, using an automated staining protocol validated for HER2 detection . IHC scoring IHC staining for HER2 was independently evaluated by three trained pathologists (IN, MK, JR), followed by a consensus session for discordantly scored samples to define a consensus score for each case. IHC stains of the two assays were read after a 2-week wash-out period, and all the pathologists were blinded to the FISH results. In addition to a pre-study training provided by Dako/Agilent, all investigators had extensive experience in HER2 evaluation, having served over the past 20 years as readers in most of the trastuzumab, pertuzumab, and T-DM1 approval BC studies by Targos GmbH (Kassel, Germany) (for a review of studies screened by first-generation HercepTest (poly), see ). IHC scoring was performed according to the 2018 ASCO/CAP guidelines . Accordingly, cases with complete intense staining in ≤ 10% of tumor cells, as well as cases with intense and lateral or basolateral (“U-type”) staining, were included in the IHC 2 + category. For cases of IHC 1 + staining intensity (i.e., faint/barely perceptible membrane staining), the percentage of stained cells ≤ 10% or > 10% was recorded separately according to Ventana Instructions for Use (IFU) . Intensity scoring was performed by applying the magnification rule as published previously by our group .
The IHC staining protocol using the HercepTest (mAb) was performed as described by the manufacturer . Freshly cut tissue was processed on the Dako Omnis platform (Agilent Technologies, Santa Clara, CA) together with kit control slides for every staining run, using an automated staining protocol validated for HER2 detection .
IHC staining using the PATHWAY® HER-2/neu rabbit monoclonal antibody 4B5 was performed according to the recommendations of the manufacturer . Freshly cut tissue was processed on the Ventana BenchMark ULTRA (Ventana Medical Systems, Roche Diagnostics, Tucson, AZ) together with kit control slides for every staining run, using an automated staining protocol validated for HER2 detection .
IHC staining for HER2 was independently evaluated by three trained pathologists (IN, MK, JR), followed by a consensus session for discordantly scored samples to define a consensus score for each case. IHC stains of the two assays were read after a 2-week wash-out period, and all the pathologists were blinded to the FISH results. In addition to a pre-study training provided by Dako/Agilent, all investigators had extensive experience in HER2 evaluation, having served over the past 20 years as readers in most of the trastuzumab, pertuzumab, and T-DM1 approval BC studies by Targos GmbH (Kassel, Germany) (for a review of studies screened by first-generation HercepTest (poly), see ). IHC scoring was performed according to the 2018 ASCO/CAP guidelines . Accordingly, cases with complete intense staining in ≤ 10% of tumor cells, as well as cases with intense and lateral or basolateral (“U-type”) staining, were included in the IHC 2 + category. For cases of IHC 1 + staining intensity (i.e., faint/barely perceptible membrane staining), the percentage of stained cells ≤ 10% or > 10% was recorded separately according to Ventana Instructions for Use (IFU) . Intensity scoring was performed by applying the magnification rule as published previously by our group .
in situ hybridization assessment HER2 IQFISH pharmDx (K5731) Determination of HER2 gene amplification was analyzed using the HER2 IQFISH pharmDx kit according to the recommendations of the manufacturer . HER2 in situ hybridizations were evaluated by a pathologist (IN) using the updated 2018 ASCO/CAP guidelines. For final interpretation of the FISH data, newly defined ISH groups (1–5) were taken into consideration . Accordingly, group 1 (ratio ≥ 2.0 and gene count ≥ 4.0) and group 3 cases (ratio < 2.0 and gene count ≥ 6.0) with IHC 3 + or IHC 2 + were considered FISH positive.
Determination of HER2 gene amplification was analyzed using the HER2 IQFISH pharmDx kit according to the recommendations of the manufacturer . HER2 in situ hybridizations were evaluated by a pathologist (IN) using the updated 2018 ASCO/CAP guidelines. For final interpretation of the FISH data, newly defined ISH groups (1–5) were taken into consideration . Accordingly, group 1 (ratio ≥ 2.0 and gene count ≥ 4.0) and group 3 cases (ratio < 2.0 and gene count ≥ 6.0) with IHC 3 + or IHC 2 + were considered FISH positive.
For comparison of datasets, chi-square test ( X 2 ) was used with p < 0.05 considered as statistically significant. Test performance was evaluated using FISH as a reference standard. Sensitivity and specificity were calculated as follows: [12pt]{minimal}
$$=100 }{\#+\#}$$ Estimated sensitivity = 100 % × # true positive events # true positive events + # false negative events [12pt]{minimal}
$$=100 }{\#+\#}$$ Estimated specificity = 100 % × # true negative events # false positive events + # true negative events Inter-rater reliability (IRR), defined as the ratio of the total number of agreements among raters and the total number of ratings, was calculated as follows: [12pt]{minimal}
$$[]=\#}{\#*\#} 100$$ IRR % = Total # of agreements Total # of ratings given by each rater ∗ # of raters × 100
Performance of HercepTest (mAb) and inter-rater agreement ( Fig. ) In HER2-expressing samples, each of the HER2 IHC assays produced specific membrane-bound staining that was easy to interpret at all intensities (weak to strong). Although non-specific background staining was not observed, a weak and only focally distributed staining of normal duct cells was detected with HercepTest (mAb) (Fig. , ). Furthermore, PATHWAY 4B5 staining was characterized by the occasional presence of diffuse and/or dot-like cytoplasmic staining in tumor and normal cells, as previously reported . Signal detection in normal duct samples was usually of low intensity (Fig. , ). Noteworthy, we did not observe relevant staining differences between the sample types, e.g., no higher frequency of edge artifacts in biopsies. Within the HercepTest (mAb) and the PATHWAY 4B5 assays, an overall inter-reader agreement of 84% (100/119) and of 89.1% (106/119) was observed. Study IRR was recorded as 89.4% and 92.7%, respectively. Discrepantly scored samples were re-evaluated by all three observers during a final review session and assigned consensus scores that were used for further analyses. It is noteworthy that most disagreements (68.8%) between pathologists’ scores were observed within the HER2-low range (later consented as IHC score 0 or 1 +), especially near the cut-off for HER2 ultra-low category exhibiting a HER2 score of 0 with incomplete and faint staining in ≤ 10% of tumor cells. This led to several challenging samples around the cut-off value (IHC 1 + versus IHC 0, according to ASCO/CAP 2018). HercepTest (mAb) and PATHWAY 4B5 — inter-assay concordance ( Table ) Based on the consented scores for both assays, complete concordance was reached in 83 of 119 tumors (69.7%). Thirty-six samples received discordant scores, including 26 resections (25.2%) and 10 biopsies (62.5%). Virtually, all these cases ( n = 35) showed higher scores with HercepTest (mAb), and in only one case (biopsy) was the staining recorded to be higher by PATHWAY 4B5. While 56 samples were evaluated as negative (IHC 0) for HER2 by PATHWAY 4B5, only 35 specimens were likewise identified by HercepTest (mAb). Thus, adjustments to discordant scores were mainly associated with the PATHWAY 4B5 negative group of IHC 0 and IHC 1 + (33 of 36). This led to a significantly different classification of BC samples by both assays. For instance, the group of HER2-low expressing samples (HER2 score 2 + or 1 + /not amplified) was found to be significantly larger for HercepTest (mAb) (35% versus 19%; p < 0.01). Only three of the discordant cases were observed in the PATHWAY 4B5 IHC 2 + and IHC 3 + group, with scores for two tumors being raised from IHC 2 + to 3 + , and one score downgraded from IHC 3 + to 2 + . Lastly, the concordance of both assays was found to be 83.7% (87/104 cases) for HER2-negative (IHC 0/1 +) versus HER2-positive (IHC 3 +). HercepTest (mAb) and PATHWAY 4B5 — correlation with FISH ( Fig. ) FISH data were obtained for 114 specimens, 42 of which showed HER2 amplification (Fig. ). All non-amplified cases ( n = 72) were identified as IHC negative (0/1 +) or equivocal (2 +) by both assays, i.e., no false positives were recorded, corresponding to 100% specificity. However, two false negatives were observed with the PATHWAY 4B5 assay in which two amplified (surgical) specimens showed an IHC 1 + score compared to a IHC 2 + score with the HercepTest (mAb), leading to a slightly lower sensitivity for PATHWAY 4B5 (95.2% versus 100%; Fig. ). Both cases were tested amplified according to the external vendor information as well as within this study. Although more IHC 2 + cases were identified by the HercepTest (mAb) as being not amplified (14 of 27) compared to PATHWAY 4B5 (3 of 15), all the amplified tumors were detected as positive (IHC 2 + or 3 +) when using the HercepTest (mAb) (see Figs. and ). A more detailed analysis of FISH data was conducted with respect to ISH groups according to ASCO/CAP 2018 guidance . Compared to PATHWAY 4B5, scores for 13 cases were increased to IHC 2 + when using HercepTest (mAb) ( n = 10 from IHC 1 + and n = 3 from IHC 0; see Table ). In two tumors, FISH revealed a HER2 ratio ≥ 2.0 and mean gene count per cell ≥ 4.0, corresponding to ISH group 1 (HER2 positive; see Fig. and Table ; sample nos. 86 and 116). In addition, four tumors with ratios ≤ 2.0 exhibited increased HER2 gene counts between ≥ 4 and < 6, corresponding to ISH group 4 (Table : samples 56, 78, 103, and 109). In these cases, HER2 amplification status should have been considered questionable and been reported as negative, with a comment about the uncertainty of a response to HER2-targeted drug therapy available at the time of guidance (i.e., 2018). HercepTest (mAb) and PATHWAY 4B5 — correlation with HER2-low status ( Fig. ) Since the development of novel HER2-directed drugs may benefit BC patients with low levels of HER2 expression (IHC 2 + /non-amplified and IHC 1 +) , the assay data were further analyzed with respect to their sensitivity and specificity for detecting HER2-low tumors (Fig. ). Out of 41 tumors (all non-amplified) that were determined to be completely negative by PATHWAY 4B5, only 19 cases (46.3%) showed no staining when using HercepTest (mAb), corresponding to the more strictly defined IHC 0 category using the Ventana score algorithm. In the remaining 22 cases, the HercepTest (mAb) stained at least some tumor cells, with approximately one-third of these cases belonging to the HER2-low group (7 × IHC 1 + , 1 × IHC 2 + non-amplified) and 14 cases in the HER2 “ultra-low” group with < 10% stained tumor cells (see also Fig. , marked in grey), thus highlighting the high sensitivity of the HercepTest (mAb) used in this study.
Fig. ) In HER2-expressing samples, each of the HER2 IHC assays produced specific membrane-bound staining that was easy to interpret at all intensities (weak to strong). Although non-specific background staining was not observed, a weak and only focally distributed staining of normal duct cells was detected with HercepTest (mAb) (Fig. , ). Furthermore, PATHWAY 4B5 staining was characterized by the occasional presence of diffuse and/or dot-like cytoplasmic staining in tumor and normal cells, as previously reported . Signal detection in normal duct samples was usually of low intensity (Fig. , ). Noteworthy, we did not observe relevant staining differences between the sample types, e.g., no higher frequency of edge artifacts in biopsies. Within the HercepTest (mAb) and the PATHWAY 4B5 assays, an overall inter-reader agreement of 84% (100/119) and of 89.1% (106/119) was observed. Study IRR was recorded as 89.4% and 92.7%, respectively. Discrepantly scored samples were re-evaluated by all three observers during a final review session and assigned consensus scores that were used for further analyses. It is noteworthy that most disagreements (68.8%) between pathologists’ scores were observed within the HER2-low range (later consented as IHC score 0 or 1 +), especially near the cut-off for HER2 ultra-low category exhibiting a HER2 score of 0 with incomplete and faint staining in ≤ 10% of tumor cells. This led to several challenging samples around the cut-off value (IHC 1 + versus IHC 0, according to ASCO/CAP 2018).
Table ) Based on the consented scores for both assays, complete concordance was reached in 83 of 119 tumors (69.7%). Thirty-six samples received discordant scores, including 26 resections (25.2%) and 10 biopsies (62.5%). Virtually, all these cases ( n = 35) showed higher scores with HercepTest (mAb), and in only one case (biopsy) was the staining recorded to be higher by PATHWAY 4B5. While 56 samples were evaluated as negative (IHC 0) for HER2 by PATHWAY 4B5, only 35 specimens were likewise identified by HercepTest (mAb). Thus, adjustments to discordant scores were mainly associated with the PATHWAY 4B5 negative group of IHC 0 and IHC 1 + (33 of 36). This led to a significantly different classification of BC samples by both assays. For instance, the group of HER2-low expressing samples (HER2 score 2 + or 1 + /not amplified) was found to be significantly larger for HercepTest (mAb) (35% versus 19%; p < 0.01). Only three of the discordant cases were observed in the PATHWAY 4B5 IHC 2 + and IHC 3 + group, with scores for two tumors being raised from IHC 2 + to 3 + , and one score downgraded from IHC 3 + to 2 + . Lastly, the concordance of both assays was found to be 83.7% (87/104 cases) for HER2-negative (IHC 0/1 +) versus HER2-positive (IHC 3 +).
Fig. ) FISH data were obtained for 114 specimens, 42 of which showed HER2 amplification (Fig. ). All non-amplified cases ( n = 72) were identified as IHC negative (0/1 +) or equivocal (2 +) by both assays, i.e., no false positives were recorded, corresponding to 100% specificity. However, two false negatives were observed with the PATHWAY 4B5 assay in which two amplified (surgical) specimens showed an IHC 1 + score compared to a IHC 2 + score with the HercepTest (mAb), leading to a slightly lower sensitivity for PATHWAY 4B5 (95.2% versus 100%; Fig. ). Both cases were tested amplified according to the external vendor information as well as within this study. Although more IHC 2 + cases were identified by the HercepTest (mAb) as being not amplified (14 of 27) compared to PATHWAY 4B5 (3 of 15), all the amplified tumors were detected as positive (IHC 2 + or 3 +) when using the HercepTest (mAb) (see Figs. and ). A more detailed analysis of FISH data was conducted with respect to ISH groups according to ASCO/CAP 2018 guidance . Compared to PATHWAY 4B5, scores for 13 cases were increased to IHC 2 + when using HercepTest (mAb) ( n = 10 from IHC 1 + and n = 3 from IHC 0; see Table ). In two tumors, FISH revealed a HER2 ratio ≥ 2.0 and mean gene count per cell ≥ 4.0, corresponding to ISH group 1 (HER2 positive; see Fig. and Table ; sample nos. 86 and 116). In addition, four tumors with ratios ≤ 2.0 exhibited increased HER2 gene counts between ≥ 4 and < 6, corresponding to ISH group 4 (Table : samples 56, 78, 103, and 109). In these cases, HER2 amplification status should have been considered questionable and been reported as negative, with a comment about the uncertainty of a response to HER2-targeted drug therapy available at the time of guidance (i.e., 2018).
Fig. ) Since the development of novel HER2-directed drugs may benefit BC patients with low levels of HER2 expression (IHC 2 + /non-amplified and IHC 1 +) , the assay data were further analyzed with respect to their sensitivity and specificity for detecting HER2-low tumors (Fig. ). Out of 41 tumors (all non-amplified) that were determined to be completely negative by PATHWAY 4B5, only 19 cases (46.3%) showed no staining when using HercepTest (mAb), corresponding to the more strictly defined IHC 0 category using the Ventana score algorithm. In the remaining 22 cases, the HercepTest (mAb) stained at least some tumor cells, with approximately one-third of these cases belonging to the HER2-low group (7 × IHC 1 + , 1 × IHC 2 + non-amplified) and 14 cases in the HER2 “ultra-low” group with < 10% stained tumor cells (see also Fig. , marked in grey), thus highlighting the high sensitivity of the HercepTest (mAb) used in this study.
Accurate assessment of HER2 status is of utmost importance for patient selection and the determination of those eligible for HER2-directed therapy. Test kits approved by the FDA have been introduced to minimize HER2 testing variability and are now recommended for use by ASCO/CAP . To the best of our knowledge, this is the first study comparing the technical and diagnostic performance of the new HercepTest (mAb) with the well-established Ventana PATHWAY 4B5 test kit. The original manual HercepTest (poly) was approved in 1998 by the FDA for assessing the eligibility of BC patients to receive trastuzumab antibody therapy. Recent reports, however, demonstrated higher specificity and sensitivity for alternative HER2 IHC assays compared to HercepTest (poly) for Autostainer . To meet these challenges, a new HercepTest (mAb) pharmDx kit was developed that uses a licensed monoclonal antibody (clone DG44) produced with a patented process by Epitomics Inc. (an Abcam company). IHC detection of HER2 using the new HercepTest (mAb) pharmDx kit is performed on the fully automated Dako Omnis staining device. The PATHWAY 4B5 was also run on an automated staining system (i.e., the Ventana Benchmark ULTRA); however, HercepTest (mAb) performed on Dako Omnis platform, using the newly invented “dynamic gap staining technology” (reviewed in ), was observed to provide slightly more consistent staining as indicated by lower numbers of repetitions and fewer staining artifacts (e.g., patchy staining, edge effects, and air bubbles; see Fig. ). In general, IHC staining of the HercepTest (mAb) assay was characterized by distinct and sharp detection of HER2, with low to no background and/or non-specific signal detection (see also Fig. ). Dot-like cytoplasmic staining, with or without basal membrane staining as outlined by Ventana IFU for PATHWAY 4B5, was not observed with HercepTest (mAb) in our sample series. However, a partial, mostly weak staining of normal epithelium could be seen in some samples, but was associated neither with the HER2 protein level of tumor tissue nor with false positive immunoreactions, i.e., IHC 3 + and FISH negatives as described in some cases previously for polyclonal HercepTest (poly) . Instead, the comparison with FISH data demonstrated the opposite, with 100% concordance between HercepTest (mAb) and amplification status for cases scored as 0, 1 + , or 3 + by HercepTest (mAb). Two false negatives were observed using the PATHWAY 4B5 assay (1 + by PATHWAY 4B5 IHC but FISH positive), resulting in a reduced sensitivity. In this context, it should be noted that previous studies frequently reported a certain number of tumors without HER2 protein overexpression (IHC score 0 or 1 +) but being HER2 gene amplified . While commonly considered as cases with DNA-uncoupled synthesis of HER2 protein, it might be of interest to confirm IHC score with the apparently more sensitive HercepTest (mAb). Notably, the HercepTest (mAb) assay generated a significantly higher rate of equivocal cases (30 versus 15; see Table ). Based on the 114 cases with available FISH data, 51.8% (14/27) were non-amplified by FISH compared to 20% (3/15) by PATHWAY 4B5. It may be argued then that HercepTest (mAb) could result in increased costs and delayed diagnosis due to an increased rate of reflex FISH testing in clinical practice. However, in light of the updated and more focused ASCO/CAP HER2 testing guideline that defines five diagnostic ISH groups , an in-depth analysis of FISH data revealed an additional four tumors identified by HercepTest (mAb) as IHC 2 + had increased gene counts (between ≥ 4.0 and < 6.0) and a ratio < 2.0, due to polysomy in three cases. These tumors correspond to ASCO/CAP ISH group 4 and would have been classified as negative by PATHWAY 4B5 testing. The clinical implication for these patients is still not clear. The prevalence of this group in different studies varies from 1.9% to 14.2% and has been classified as “equivocal” in ASCO/CAP 2013 guidelines. Since 2018 , these cases should be reported as HER2-negative with an associated comment describing the uncertainty about patient response to HER2-directed therapies available at the time of guidance. Meanwhile, novel HER2-directed drugs such as T-DXd have been developed using a new generation of ADCs . In contrast to the first approved ADC (T-DM1) for which therapy effectiveness is still dependent on the demonstration of HER2-positive tumors (IHC3 + and/or ISH amplified), T-DXd was beneficial even after Kadcyla therapy . Interestingly, there is increasing evidence that patients with HER2-low BC (HER2 2 + /non-amplified or IHC 1 + , according to ASCO/CAP 2018) also benefit from T-DXd. These new developments in HER2-targeted BC therapy have implications for both testing and the definition of HER2 sensitivity and specificity . Our data demonstrate a higher detection rate of HER2 amplified breast carcinomas (100% versus 95%) by the HercepTest (mAb) compared to the Ventana PATHWAY 4B5 assay. In addition, the number of HER2-low expressing samples was markedly increased by using HercepTest (mAb) (35% versus 19%). In the upcoming era of HER2-targeted therapies administered to HER2-low BC patients , both observations would significantly increase the number of patients eligible to HER2-directed therapies. These promising results have already raised much interest in the scientific community focusing on the assessment of HER2-low BC in future clinical diagnostics . Recent clinical trials using the HER2-directed antibody–drug conjugate T-DXd have already included patients exhibiting either IHC 1 + or 2 + /HER2 non-amplified in their HER2-low group (e.g., DB02 [NCT03523585]) or very low (“ultra-low”) HER2-expressing cohorts (HER2 IHC 0 < 1 + , weak staining in less than 10% of tumor cells, e.g., DB06 [NCT04494425]) eligible for therapy. Most recently, a large T-DXd phase III trial (DB-04) turned out to be effective in HER2-low metastatic breast cancer . Thus far, these studies are based on expression analysis using the PATHWAY 4B5 antibody clone. As demonstrated in this study, the increased sensitivity of HercepTest (mAb) may allow inclusion of more patients in clinical trials, specifically by enrolling patients with HER2-low and ultra-low expression and allow the investigation of clinical response rate and outcome in these cohorts. Another implication of testing HER2-low category of patients in this study is the accuracy of HER2 interpretation within this tumor group. Inter-rater variability was mostly restricted to the discordant assessment of HER2 0/1 + cases near the cut-off value. Future HER2 scoring will need to include more training for the HER2-low category of patients, and ASCO/CAP may refine their guidelines appropriately. Recently, French GEFPICS group published the first national recommendation for HER2 status evaluation in breast cancer with emphasis on the HER2-low concept underlining the need for harmonized testing guidelines . Finally, we regard the results of this study as representative for all HER2 scores including the recently delineated HER2 low category as the carefully pre-selected case series representing the entire range of IHC scores and amplification levels, including different ISH subgroups . Accordingly, about 35% of cases were HER2-positive (IHC 3 + or IHC 2 + amplified) belonging to ISH group 1 ( n = 40) and group 3 ( n = 2). In the remaining non-amplified cases, the accuracy of assays was determined both with respect to detection of HER2-negative versus HER2-positive and considering HER2-low (IHC 1 + and IHC 2 + /non-amplified) and HER2-ultra-low (IHC 0 < 1 +) tumors. Therefore, our comparative study of HercepTest (mAb) with PATHWAY 4B5 addresses the main challenges that may arise during HER2 testing in BC, particularly with consideration to the emerging anti-HER2-directed drugs and patients with lower HER2 expression. However, determining the predictive value of new HercepTest (mAb) clinical trials using this new assay is of crucial importance since more sensitive tests may not necessarily be the best predictors of response to targeted therapy. In conclusion, while both IHC assays are highly suitable for the detection of HER2 protein in BC samples, fewer assay-related failures (e.g., staining artifacts) were observed using HercepTest (mAb) Dako Omnis. The data demonstrated that HercepTest (mAb) exhibited both high specificity (100%) and high sensitivity (100%), which could be critical in patient selection for new HER2-targeting treatment options. Future studies will demonstrate whether this new assay has the capacity to provide better patient stratification, leading to better patient response rates and clinical outcomes.
Below is the link to the electronic supplementary material. Supplementary file1 (XLSX 19 KB)
|
The ill physician who self-discloses: What do patients think? | 0219eb3b-8703-4537-b7e6-27a9c6a978ee | 9718545 | Family Medicine[mh] | Physician self-disclosure is often used intentionally, to establish rapport, foster trust and reciprocity, express empathy, provide hope and reassurance or enhance the ability to make credible recommendations. However, there may also be negative consequences, including taking away time from the patient’s visit, changing the focus of the visit, burdening the patient with the doctor’s problems, and even role reversal . Revealing one’s health status is a unique type of self-disclosure. It is often unavoidable, colours the physician-patient relationship, and may have significant ramifications for the physician. Physicians who experience a severe illness face the complex decision of whether to inform their patients about their health status. Sharing such personal, sensitive information with patients may have a powerful impact on the physician, the patient, and the nature of their relationship . While physician self-disclosure on various issues has been addressed in the literature, very little is known about patients’ reactions to such revelations. Specifically, patients’ reactions to learning that their doctor is ill have never been studied systematically. In September 2017, one of the authors (BK) underwent septal myectomy for advanced hypertrophic cardiomyopathy. He was absent from clinical work for six weeks. Upon return, colleagues, friends and family gave a range of divergent opinions on whether it was correct to openly and honestly report to patients about what had transpired. In the absence of any evidence-based recommendations, he decided to inform every patient who came for a visit during the period after his return and to ask for their reactions to being told. This paper is an analysis of their written responses. This study aimed to learn about patients’ emotions and reactions to their family physician’s sharing with them about having a major illness.
Setting and population The study was carried out in a family practice office in a suburb of Jerusalem between November 12 and December 3, 2017. The town’s population numbers about 18,000, is middle class and has a mixture of religious and secular Jewish residents. The participants received care at the only local branch of the Meuhedet Health Services HMO, where the author (BK) and one other family physician were the primary care staff. There had been no advance notice to patients that the physician would be absent for an extended period. During the physician’s absence, secretaries told patients who came to the clinic that he was in the United States for a prolonged family visit. Study design We carried out a questionnaire study of both closed and open questions in a sample of patients. The two authors drafted the questionnaire independently and then together. It was piloted with eight men and women of various ages and then finalised. It included demographic data, followed by nine closed and three open questions about reactions to and opinions regarding learning of the physician’s illness and his sharing this information with patients. We included the open questions to explore an additional, more personal dimension of the observed phenomenon. In these questions, we asked participants to freely describe their thoughts and feelings after being informed that the physician had been ill, and how they felt about the fact that the doctor himself had informed them about his illness. A fourth open question added mid-survey, asked patients what they would think or feel if the doctor’s illness had not improved. This report focuses principally on the study’s open questions. Data collection The study began on the day that the physician returned to work. The physician told each patient who mentioned the absence: ‘I was out of the country because I had to have heart surgery for an unusual, genetic condition that could not be repaired in Israel. The operation was successful, and now I am feeling well.’ If the patient did not raise the issue, this sentence was prefaced with: ‘Did you know that I was not at work for a very long time?’. The topic was presented in an optimistic tone, and patient questions were answered straightforwardly. The focus of the discussion then switched quickly to the patient’s reason for visitation. At the end of the consultation, the physician asked the patient if s/he would be willing to fill out an anonymous questionnaire regarding the doctor’s severe illness. The questionnaire was then given to the patient by the physician himself, to be filled out in the waiting room and returned to the secretary. IRB approval was not requested, as the data collected were anonymous and did not include information directly relating to the health or medical treatment of the patients. The questionnaire was distributed to 200 consecutive patients 18 years of age or older who attended the clinic during the study period. The last 98 patients (103–200) received the questionnaire with the fourth open question. Data analysis – open questions Anonymous data extraction and compilation of the open questions were performed by one of the authors (NG). Content analysis of the first 20 questionnaires was then performed and discussed by both authors to elicit primary themes and formulate categories. The next phase entailed a dynamic and ongoing process of analysing the entire body of questionnaires, identifying additional themes. This involved constructing tables separately for each question that included: the questionnaire number with demographic data; 'significance units’ – word for word patient responses (e.g. ‘I was both sad and happy’); 'coding’ (e.g. using the method of marker words and sentences); author’s comments (e.g. search specifically for expressions of ambivalence in a specific age group); 'categories’ (e.g. emotions); 'themes’ (e.g. mixed feelings or ambivalence). All respondents’ data was organised and analysed according to gender, age and marital status. An ongoing dialogue between the researchers facilitated mutual exposure to their thoughts and feelings .
The study was carried out in a family practice office in a suburb of Jerusalem between November 12 and December 3, 2017. The town’s population numbers about 18,000, is middle class and has a mixture of religious and secular Jewish residents. The participants received care at the only local branch of the Meuhedet Health Services HMO, where the author (BK) and one other family physician were the primary care staff. There had been no advance notice to patients that the physician would be absent for an extended period. During the physician’s absence, secretaries told patients who came to the clinic that he was in the United States for a prolonged family visit.
We carried out a questionnaire study of both closed and open questions in a sample of patients. The two authors drafted the questionnaire independently and then together. It was piloted with eight men and women of various ages and then finalised. It included demographic data, followed by nine closed and three open questions about reactions to and opinions regarding learning of the physician’s illness and his sharing this information with patients. We included the open questions to explore an additional, more personal dimension of the observed phenomenon. In these questions, we asked participants to freely describe their thoughts and feelings after being informed that the physician had been ill, and how they felt about the fact that the doctor himself had informed them about his illness. A fourth open question added mid-survey, asked patients what they would think or feel if the doctor’s illness had not improved. This report focuses principally on the study’s open questions.
The study began on the day that the physician returned to work. The physician told each patient who mentioned the absence: ‘I was out of the country because I had to have heart surgery for an unusual, genetic condition that could not be repaired in Israel. The operation was successful, and now I am feeling well.’ If the patient did not raise the issue, this sentence was prefaced with: ‘Did you know that I was not at work for a very long time?’. The topic was presented in an optimistic tone, and patient questions were answered straightforwardly. The focus of the discussion then switched quickly to the patient’s reason for visitation. At the end of the consultation, the physician asked the patient if s/he would be willing to fill out an anonymous questionnaire regarding the doctor’s severe illness. The questionnaire was then given to the patient by the physician himself, to be filled out in the waiting room and returned to the secretary. IRB approval was not requested, as the data collected were anonymous and did not include information directly relating to the health or medical treatment of the patients. The questionnaire was distributed to 200 consecutive patients 18 years of age or older who attended the clinic during the study period. The last 98 patients (103–200) received the questionnaire with the fourth open question.
Anonymous data extraction and compilation of the open questions were performed by one of the authors (NG). Content analysis of the first 20 questionnaires was then performed and discussed by both authors to elicit primary themes and formulate categories. The next phase entailed a dynamic and ongoing process of analysing the entire body of questionnaires, identifying additional themes. This involved constructing tables separately for each question that included: the questionnaire number with demographic data; 'significance units’ – word for word patient responses (e.g. ‘I was both sad and happy’); 'coding’ (e.g. using the method of marker words and sentences); author’s comments (e.g. search specifically for expressions of ambivalence in a specific age group); 'categories’ (e.g. emotions); 'themes’ (e.g. mixed feelings or ambivalence). All respondents’ data was organised and analysed according to gender, age and marital status. An ongoing dialogue between the researchers facilitated mutual exposure to their thoughts and feelings .
A total of 200 patients visited the physician during the study period, 165 (82.5%) returned the questionnaires. Patient characteristics are presented in . Brief summary of the closed questions responses 20% of the patients felt that after being absent due to a major illness, a physician should tell all patients the reason, 55% that he should only tell those who inquire, 7% that he should give a different reason to those who ask, and 18% had no opinion or gave a different response. In total, 82% were very pleased that the physician shared the information with them, 12% were somewhat pleased, 6% responded that 'it doesn’t matter’, and none were displeased. Content analysis of the open questions’ responses The themes that were formulated stemmed from the content analysis. They will be presented with a description and a limited selection of representative citations. No significant differences in replies according to the demographic variables were identified. Thoughts and feelings about learning of the physician’s illness In this section, we received many emotional responses from the patients. Identification with the physician was another major theme. A small number of responses were matter-of-fact and non-emotional. Themes Emotional reactions There was a wide range of emotional responses. Sadness, concern and pain were the most common. Following these in order of frequency were expressions of surprise and even shock at learning that the doctor had been ill, often combined with concern or worry about his well-being and health. There were also moving responses that expressed happiness and relief that everything had gone smoothly and that the physician had returned to work as usual, several with compliments to this particular doctor. Many statements combined several emotions, such as sadness together with happiness about the recovery. On the one hand, I was very sorry to hear about this, and on the other, was happy that he is feeling well. (Q132) Physician as a human being A widespread reason for feeling surprised was the very prevalent perception that doctors, by the nature of their profession, do not or should not become ill themselves. I was a bit surprised because we’re talking about a doctor, and sometimes it seems that doctors don’t get sick because they know how to take care of themselves. (Q131) The realisation that the physician is a human being like everyone else and that there is a common human denominator was philosophically expressed by some. I'm not the only one with pain and feelings or poor health. We all have inherited genetic problems. We should know that the doctor is no different from us. He suffers pain and fears just like I do. Sometime he needs a doctor to take care of him. (Q36) Thoughts and feelings about the physician’s self-disclosure In this section, we asked the respondents to examine their thoughts and feelings that arose, not from learning of the doctor’s illness but from a unique and unconventional experience: that the person giving the notification of the health condition was the physician himself who had become ill. This subject elicited a significant number and wide range of responses. We note that most patients favoured their physician’s sharing information with them about having a serious illness – matching the findings in the quantitative analysis. Themes Feelings towards the physician and reactions to the revelation Empathy The experience of shared solidarity and human fragility was expressed in many responses. I identify with the difficulties that he went through, and just as I feel that he worries about me, so I worry and think about him. (Q44) Surprise Surprise was a widespread reaction among the respondents, who were divided between those who simply gave a one-word response – ‘surprise;’ and those who expanded on their surprise that the doctor had opened up to them about his disease. I felt surprised but it gave me a good feeling and a sense of closeness, in that if he tells me about this and shares something so personal with me, this reveals humanity, openness, frankness. (Q26) Appreciation In general, we have the impression that the status of openness and exposure was much appreciated. Most of the expressions of appreciation were brief (‘brave,’ ‘congratulations to him’), but some wrote more explicitly about the directness, courage and sincerity and how this affected them. I deeply respect the doctor’s decision to tell his patients about his illness… I appreciate him even more because of this. Such a step requires great courage and openness. (Q157) Pride Another common and notable reaction that was experienced as a result of this unexpected and unique encounter with the physician was pride, feeling worthy of mutual trust and important, in that the physician had chosen to share this very personal information with him. I'm proud that he was willing to open his private self and share with me. (Q36) I was happy. This allowed me to feel that the doctor trusts me enough as a human being to tell me what he had gone through, just as I trust him. (Q192) Criticism Reasons for criticism included a feeling that the physician should not share such information with patients. … he doesn’t have to publicise this to every patient. This is a personal matter. (Q136) I was surprised but mainly felt that this wasn’t so much my business. (Q180) As well as quite the opposite, he should have informed patients of the illness in advance. I would have wanted to hear about this before he travelled for the surgery, so that at least I could have wished him a full recovery… (Q159) Compassion and concern Beyond empathy, several patients felt compassion towards their now-vulnerable doctor. Sadness and concern for the doctor’s well-being were also often noted. However, very few respondents expressed pity or worry about his ability to function. The experience of actually seeing the physician facing them and appearing perfectly healthy undoubtedly affected their responses, minimising such reactions. Laconic, unemotional reactions We were impressed that one-word responses, clearly expressing a lack of emotion or thought, such as ‘normal’ or ‘regular’ were very uncommon, as were statements showing that the patient was not particularly moved. For example, 'nice’ was expressed by two very young participants. b. Feelings during the encounter Comfort/discomfort Many participants spoke of feeling comfortable with the fact that the doctor had shared personal medical information with them. I felt very comfortable, nice on his part and warm. (Q15) However, a few explicitly expressed discomfort, embarrassment or mixed feelings. We must learn from this that there were also reservations about the directness and openness of the doctor. Initially, I thought it was a bit weird that he was sharing this with me but ultimately I appreciated his sincerity and sharing. His sharing caused me to feel more at ease. (Q3) I felt embarrassed, unsure that this was any of my business. I appreciate privacy. (Q103) For some patients, discomfort or ambivalence was due to their respect for the physician’s right to privacy. I hope that he understood, that I didn’t ask about his absence out of respect for his privacy, and not out of indifference. (Q33) c. Impact on the physician-patient relationship Value of personal experience The value of sharing personal experiences was expressed multiple times as a feature that adds value to the physician. Patients perceived that a physicians’ age and experience, including the personal experience of illness, add to their ability to understand, advise, empathise and inspire trust. This strengthens the relationship and creates a feeling of trust between physician and patient…Only in this way does the patient feel more secure and involved and less uncertain. (Q8) The physician’s ability to honestly share with his patients not only his professional knowledge and experience, but also his personal experience… It strengthens and inspires confidence! …. A doctor who knows first-hand what a disease is and the feelings that accompany it can be more empathetic and sensitive to patients. (Q125) Engenders closeness Needless to say, several respondents expressed the feeling that sharing personal information with them heightened their feeling of closeness with the physician. This can draw the patient closer to the doctor and improve the communication between them. (Q130) Unique connection with family physician Finally, we found several noteworthy responses that distinguished Family Medicine from other specialties. The longstanding, multifaceted relationship with patients that develops with family doctors seemed to nurture a deep, bidirectional connection that may not be present with other specialities. It’s important for me to know, not because I'm curious but rather because I feel an important connection specifically with my family doctor, beyond what I feel for other kinds of doctors. (Q152) The relationship between a family doctor and a patient are generally over many years and so it’s important to me… that the doctor connects to me beyond tests and medications…that the doctor shares his feelings and his illness. ( Q44) Hypothetical reactions if the physician had not recovered We included a hypothetical question to understand whether the patients could share their thoughts and emotions about ‘what would happen if?.’ Most participants who chose to answer this question (76% of the subsample) selected one-word emotional responses, usually a non-quantifiable noun (e.g. sadness, worry, compassion, sorrow), and avoided verbal inflexion in the first person. This lack of content stood out, particularly in the questionnaires of participants who answered the other questions in considerable detail. It probably reflects significant patient discomfort in acknowledging or dealing with this scenario.
20% of the patients felt that after being absent due to a major illness, a physician should tell all patients the reason, 55% that he should only tell those who inquire, 7% that he should give a different reason to those who ask, and 18% had no opinion or gave a different response. In total, 82% were very pleased that the physician shared the information with them, 12% were somewhat pleased, 6% responded that 'it doesn’t matter’, and none were displeased.
The themes that were formulated stemmed from the content analysis. They will be presented with a description and a limited selection of representative citations. No significant differences in replies according to the demographic variables were identified.
In this section, we received many emotional responses from the patients. Identification with the physician was another major theme. A small number of responses were matter-of-fact and non-emotional.
Emotional reactions There was a wide range of emotional responses. Sadness, concern and pain were the most common. Following these in order of frequency were expressions of surprise and even shock at learning that the doctor had been ill, often combined with concern or worry about his well-being and health. There were also moving responses that expressed happiness and relief that everything had gone smoothly and that the physician had returned to work as usual, several with compliments to this particular doctor. Many statements combined several emotions, such as sadness together with happiness about the recovery. On the one hand, I was very sorry to hear about this, and on the other, was happy that he is feeling well. (Q132) Physician as a human being A widespread reason for feeling surprised was the very prevalent perception that doctors, by the nature of their profession, do not or should not become ill themselves. I was a bit surprised because we’re talking about a doctor, and sometimes it seems that doctors don’t get sick because they know how to take care of themselves. (Q131) The realisation that the physician is a human being like everyone else and that there is a common human denominator was philosophically expressed by some. I'm not the only one with pain and feelings or poor health. We all have inherited genetic problems. We should know that the doctor is no different from us. He suffers pain and fears just like I do. Sometime he needs a doctor to take care of him. (Q36)
There was a wide range of emotional responses. Sadness, concern and pain were the most common. Following these in order of frequency were expressions of surprise and even shock at learning that the doctor had been ill, often combined with concern or worry about his well-being and health. There were also moving responses that expressed happiness and relief that everything had gone smoothly and that the physician had returned to work as usual, several with compliments to this particular doctor. Many statements combined several emotions, such as sadness together with happiness about the recovery. On the one hand, I was very sorry to hear about this, and on the other, was happy that he is feeling well. (Q132)
A widespread reason for feeling surprised was the very prevalent perception that doctors, by the nature of their profession, do not or should not become ill themselves. I was a bit surprised because we’re talking about a doctor, and sometimes it seems that doctors don’t get sick because they know how to take care of themselves. (Q131) The realisation that the physician is a human being like everyone else and that there is a common human denominator was philosophically expressed by some. I'm not the only one with pain and feelings or poor health. We all have inherited genetic problems. We should know that the doctor is no different from us. He suffers pain and fears just like I do. Sometime he needs a doctor to take care of him. (Q36)
In this section, we asked the respondents to examine their thoughts and feelings that arose, not from learning of the doctor’s illness but from a unique and unconventional experience: that the person giving the notification of the health condition was the physician himself who had become ill. This subject elicited a significant number and wide range of responses. We note that most patients favoured their physician’s sharing information with them about having a serious illness – matching the findings in the quantitative analysis.
Feelings towards the physician and reactions to the revelation Empathy The experience of shared solidarity and human fragility was expressed in many responses. I identify with the difficulties that he went through, and just as I feel that he worries about me, so I worry and think about him. (Q44) Surprise Surprise was a widespread reaction among the respondents, who were divided between those who simply gave a one-word response – ‘surprise;’ and those who expanded on their surprise that the doctor had opened up to them about his disease. I felt surprised but it gave me a good feeling and a sense of closeness, in that if he tells me about this and shares something so personal with me, this reveals humanity, openness, frankness. (Q26) Appreciation In general, we have the impression that the status of openness and exposure was much appreciated. Most of the expressions of appreciation were brief (‘brave,’ ‘congratulations to him’), but some wrote more explicitly about the directness, courage and sincerity and how this affected them. I deeply respect the doctor’s decision to tell his patients about his illness… I appreciate him even more because of this. Such a step requires great courage and openness. (Q157) Pride Another common and notable reaction that was experienced as a result of this unexpected and unique encounter with the physician was pride, feeling worthy of mutual trust and important, in that the physician had chosen to share this very personal information with him. I'm proud that he was willing to open his private self and share with me. (Q36) I was happy. This allowed me to feel that the doctor trusts me enough as a human being to tell me what he had gone through, just as I trust him. (Q192) Criticism Reasons for criticism included a feeling that the physician should not share such information with patients. … he doesn’t have to publicise this to every patient. This is a personal matter. (Q136) I was surprised but mainly felt that this wasn’t so much my business. (Q180) As well as quite the opposite, he should have informed patients of the illness in advance. I would have wanted to hear about this before he travelled for the surgery, so that at least I could have wished him a full recovery… (Q159) Compassion and concern Beyond empathy, several patients felt compassion towards their now-vulnerable doctor. Sadness and concern for the doctor’s well-being were also often noted. However, very few respondents expressed pity or worry about his ability to function. The experience of actually seeing the physician facing them and appearing perfectly healthy undoubtedly affected their responses, minimising such reactions. Laconic, unemotional reactions We were impressed that one-word responses, clearly expressing a lack of emotion or thought, such as ‘normal’ or ‘regular’ were very uncommon, as were statements showing that the patient was not particularly moved. For example, 'nice’ was expressed by two very young participants. b. Feelings during the encounter Comfort/discomfort Many participants spoke of feeling comfortable with the fact that the doctor had shared personal medical information with them. I felt very comfortable, nice on his part and warm. (Q15) However, a few explicitly expressed discomfort, embarrassment or mixed feelings. We must learn from this that there were also reservations about the directness and openness of the doctor. Initially, I thought it was a bit weird that he was sharing this with me but ultimately I appreciated his sincerity and sharing. His sharing caused me to feel more at ease. (Q3) I felt embarrassed, unsure that this was any of my business. I appreciate privacy. (Q103) For some patients, discomfort or ambivalence was due to their respect for the physician’s right to privacy. I hope that he understood, that I didn’t ask about his absence out of respect for his privacy, and not out of indifference. (Q33) c. Impact on the physician-patient relationship Value of personal experience The value of sharing personal experiences was expressed multiple times as a feature that adds value to the physician. Patients perceived that a physicians’ age and experience, including the personal experience of illness, add to their ability to understand, advise, empathise and inspire trust. This strengthens the relationship and creates a feeling of trust between physician and patient…Only in this way does the patient feel more secure and involved and less uncertain. (Q8) The physician’s ability to honestly share with his patients not only his professional knowledge and experience, but also his personal experience… It strengthens and inspires confidence! …. A doctor who knows first-hand what a disease is and the feelings that accompany it can be more empathetic and sensitive to patients. (Q125) Engenders closeness Needless to say, several respondents expressed the feeling that sharing personal information with them heightened their feeling of closeness with the physician. This can draw the patient closer to the doctor and improve the communication between them. (Q130) Unique connection with family physician Finally, we found several noteworthy responses that distinguished Family Medicine from other specialties. The longstanding, multifaceted relationship with patients that develops with family doctors seemed to nurture a deep, bidirectional connection that may not be present with other specialities. It’s important for me to know, not because I'm curious but rather because I feel an important connection specifically with my family doctor, beyond what I feel for other kinds of doctors. (Q152) The relationship between a family doctor and a patient are generally over many years and so it’s important to me… that the doctor connects to me beyond tests and medications…that the doctor shares his feelings and his illness. ( Q44)
The experience of shared solidarity and human fragility was expressed in many responses. I identify with the difficulties that he went through, and just as I feel that he worries about me, so I worry and think about him. (Q44)
Surprise was a widespread reaction among the respondents, who were divided between those who simply gave a one-word response – ‘surprise;’ and those who expanded on their surprise that the doctor had opened up to them about his disease. I felt surprised but it gave me a good feeling and a sense of closeness, in that if he tells me about this and shares something so personal with me, this reveals humanity, openness, frankness. (Q26)
In general, we have the impression that the status of openness and exposure was much appreciated. Most of the expressions of appreciation were brief (‘brave,’ ‘congratulations to him’), but some wrote more explicitly about the directness, courage and sincerity and how this affected them. I deeply respect the doctor’s decision to tell his patients about his illness… I appreciate him even more because of this. Such a step requires great courage and openness. (Q157)
Another common and notable reaction that was experienced as a result of this unexpected and unique encounter with the physician was pride, feeling worthy of mutual trust and important, in that the physician had chosen to share this very personal information with him. I'm proud that he was willing to open his private self and share with me. (Q36) I was happy. This allowed me to feel that the doctor trusts me enough as a human being to tell me what he had gone through, just as I trust him. (Q192)
Reasons for criticism included a feeling that the physician should not share such information with patients. … he doesn’t have to publicise this to every patient. This is a personal matter. (Q136) I was surprised but mainly felt that this wasn’t so much my business. (Q180) As well as quite the opposite, he should have informed patients of the illness in advance. I would have wanted to hear about this before he travelled for the surgery, so that at least I could have wished him a full recovery… (Q159)
Beyond empathy, several patients felt compassion towards their now-vulnerable doctor. Sadness and concern for the doctor’s well-being were also often noted. However, very few respondents expressed pity or worry about his ability to function. The experience of actually seeing the physician facing them and appearing perfectly healthy undoubtedly affected their responses, minimising such reactions.
We were impressed that one-word responses, clearly expressing a lack of emotion or thought, such as ‘normal’ or ‘regular’ were very uncommon, as were statements showing that the patient was not particularly moved. For example, 'nice’ was expressed by two very young participants. b. Feelings during the encounter
Many participants spoke of feeling comfortable with the fact that the doctor had shared personal medical information with them. I felt very comfortable, nice on his part and warm. (Q15) However, a few explicitly expressed discomfort, embarrassment or mixed feelings. We must learn from this that there were also reservations about the directness and openness of the doctor. Initially, I thought it was a bit weird that he was sharing this with me but ultimately I appreciated his sincerity and sharing. His sharing caused me to feel more at ease. (Q3) I felt embarrassed, unsure that this was any of my business. I appreciate privacy. (Q103) For some patients, discomfort or ambivalence was due to their respect for the physician’s right to privacy. I hope that he understood, that I didn’t ask about his absence out of respect for his privacy, and not out of indifference. (Q33) c. Impact on the physician-patient relationship
The value of sharing personal experiences was expressed multiple times as a feature that adds value to the physician. Patients perceived that a physicians’ age and experience, including the personal experience of illness, add to their ability to understand, advise, empathise and inspire trust. This strengthens the relationship and creates a feeling of trust between physician and patient…Only in this way does the patient feel more secure and involved and less uncertain. (Q8) The physician’s ability to honestly share with his patients not only his professional knowledge and experience, but also his personal experience… It strengthens and inspires confidence! …. A doctor who knows first-hand what a disease is and the feelings that accompany it can be more empathetic and sensitive to patients. (Q125)
Needless to say, several respondents expressed the feeling that sharing personal information with them heightened their feeling of closeness with the physician. This can draw the patient closer to the doctor and improve the communication between them. (Q130)
Finally, we found several noteworthy responses that distinguished Family Medicine from other specialties. The longstanding, multifaceted relationship with patients that develops with family doctors seemed to nurture a deep, bidirectional connection that may not be present with other specialities. It’s important for me to know, not because I'm curious but rather because I feel an important connection specifically with my family doctor, beyond what I feel for other kinds of doctors. (Q152) The relationship between a family doctor and a patient are generally over many years and so it’s important to me… that the doctor connects to me beyond tests and medications…that the doctor shares his feelings and his illness. ( Q44)
We included a hypothetical question to understand whether the patients could share their thoughts and emotions about ‘what would happen if?.’ Most participants who chose to answer this question (76% of the subsample) selected one-word emotional responses, usually a non-quantifiable noun (e.g. sadness, worry, compassion, sorrow), and avoided verbal inflexion in the first person. This lack of content stood out, particularly in the questionnaires of participants who answered the other questions in considerable detail. It probably reflects significant patient discomfort in acknowledging or dealing with this scenario.
Main findings The significant finding of this study was that these encounters elicited and demonstrated a wide range of moving responses related to sharing an emotionally laden human experience with the family physician. Patients expressed sadness, identification and an enhanced sense of closeness. Noteworthy were the feelings of pride that the physician had chosen to confide in them and having the opportunity to show compassion towards the person who is usually the source of compassion. Nearly all patients were very pleased that they had been informed about the illness and were particularly touched by experiencing the face-to-face revelation by the physician himself. On the other hand, while no patients replied that they were displeased by the disclosure, very few did express discomfort about being in this situation. While most patients in this population agreed with telling only those who asked, a significant minority felt that all patients should be informed. Intuition would have it, and the physician himself ultimately felt it appropriate to share this intimate information only with veteran patients or those who inquired. Personal experience as a patient deepens the practitioner’s understanding of illness and its ramifications, often fostering empathy and the ability to be a mentor and guide. Practitioners may exploit it as an educational and motivational tool . From the patient’s perspective, such experience may enhance the physician’s value, inspiring confidence that cannot be obtained in formal training and is less likely to be felt with younger, less-weathered doctors. The value of family medicine as a unique speciality was explicitly and implicitly expressed throughout this collection of moving responses. The characteristically long-term, multifaceted relationships which develop undoubtedly have rich consequences. These responses exposed the deep and perhaps unrecognised emotional bonds that may form between patients and their family physicians. Strengths and limitations This paper is unique in that it gives patients’ views on physician self-disclosure about being ill, utilising a systematic and uniform methodology. It provides information not accessible from previous surveys or interviews with the physicians themselves, or from researchers’ analyses of visits to doctors. Numerous factors most certainly influenced patients’ reactions: the actual diagnosis, the physician’s specialty, the closeness of the relationship, the length of time under the physician’s care, the setting in a small community, and the physician’s own personality and style of communication. The results are therefore not generalisable to other populations. The study’s open questions provided particular strengths regarding the depth of information it provided about this specific and unique phenomenon. However, there were also limitations. Conducting interviews – a more optimal way of obtaining data – was impossible for lack of time. Additional identifying characteristics, such as time cared for by the physician, might have given greater insight into determinants of the patients’ responses. The primary source of bias in this study was the fact that the study was carried out by the same physician who cared for the patients being surveyed. Many patients with a longstanding relationship might not have wanted to upset a close relationship. This may have resulted in overly positive results. Possible patient concerns about lack of anonymity, especially with open questions, may have added to this problem. Nevertheless, the patients were promised and given absolute confidentiality and no identification in the data analysis. Another potentially significant source of bias was the 18% nonresponse rate. Many of these patients, specifically may have been critical of the doctor’s sharing information. Comparison with existing literature A study in which 50 physicians with HIV and other major illnesses were interviewed illuminates the feelings, struggles and dilemmas of these men and women, transformed into the role of physician-patient. Many of these physicians elected to self-disclose because of the discomfort in concealing or the conviction that their experience with illness and treatment could benefit their patients. Others decided against being open, feeling that disclosure was irrelevant or could threaten the relationship’s stability, lead to stigmatisation or harm them professionally . Essays by physician–patients caring for patients with the same diagnosis have poignantly described both the tangible benefits and the potential dangers of sharing their stories with patients . The current study is the first (to the best of our knowledge) that specifically and systematically examines the patients’ reactions to learning that their physician has a major illness and to self-disclosure about the illness by the physician. While to some extent confirming the impressions and beliefs of physicians and researchers in previous reports, this patient-oriented study reveals a rich world of patient experience and emotion that has not been previously described. Discovering that the healer himself is a vulnerable, even mortal human being was an eye-opening revelation for many respondents. This possibility had not occurred to them and did not fit into the accepted paradigms of an omniscient professional who prevents and cures illness. Cognitive coping with questions about possible death is naturally threatening, especially when it concerns those in whom we entrust our health. While much has been written about the therapist’s vulnerability in the mental health professions , these patients’ remarks illuminate just how poignant these issues are regarding the medical profession as well. The potentially negative consequences to the patient of physician self-disclosure, posited by several authors , were reflected in the comments of only a tiny minority in this sample. To some extent, this may be explained by the brief and reassuring manner in which the news was shared but it may also demonstrate that such concerns have been overstated. Perceived or actual risk of self-disclosure to the impaired physician’s career has been the subject of previous works, most notably regarding mental illness and also conditions such as substance abuse, HIV and advanced malignancy . To what degree the fear of disclosure is based on reality from the patient’s vantage point is not well known. In the current study, patients did not have to deal with the mixed feelings and dilemmas that would have arisen if the doctor had been perceived as posing a threat, unable to fully function, or having a more questionable future. It remains unanswered whether, under a different constellation of circumstances, their emotional responses would have been qualitatively different or whether they would have considered leaving the doctor. Implications for clinical practice and research The study adds to current knowledge about when to share this powerful information with patients. It warrants further investigation and discussion, especially in the education of future doctors.
The significant finding of this study was that these encounters elicited and demonstrated a wide range of moving responses related to sharing an emotionally laden human experience with the family physician. Patients expressed sadness, identification and an enhanced sense of closeness. Noteworthy were the feelings of pride that the physician had chosen to confide in them and having the opportunity to show compassion towards the person who is usually the source of compassion. Nearly all patients were very pleased that they had been informed about the illness and were particularly touched by experiencing the face-to-face revelation by the physician himself. On the other hand, while no patients replied that they were displeased by the disclosure, very few did express discomfort about being in this situation. While most patients in this population agreed with telling only those who asked, a significant minority felt that all patients should be informed. Intuition would have it, and the physician himself ultimately felt it appropriate to share this intimate information only with veteran patients or those who inquired. Personal experience as a patient deepens the practitioner’s understanding of illness and its ramifications, often fostering empathy and the ability to be a mentor and guide. Practitioners may exploit it as an educational and motivational tool . From the patient’s perspective, such experience may enhance the physician’s value, inspiring confidence that cannot be obtained in formal training and is less likely to be felt with younger, less-weathered doctors. The value of family medicine as a unique speciality was explicitly and implicitly expressed throughout this collection of moving responses. The characteristically long-term, multifaceted relationships which develop undoubtedly have rich consequences. These responses exposed the deep and perhaps unrecognised emotional bonds that may form between patients and their family physicians.
This paper is unique in that it gives patients’ views on physician self-disclosure about being ill, utilising a systematic and uniform methodology. It provides information not accessible from previous surveys or interviews with the physicians themselves, or from researchers’ analyses of visits to doctors. Numerous factors most certainly influenced patients’ reactions: the actual diagnosis, the physician’s specialty, the closeness of the relationship, the length of time under the physician’s care, the setting in a small community, and the physician’s own personality and style of communication. The results are therefore not generalisable to other populations. The study’s open questions provided particular strengths regarding the depth of information it provided about this specific and unique phenomenon. However, there were also limitations. Conducting interviews – a more optimal way of obtaining data – was impossible for lack of time. Additional identifying characteristics, such as time cared for by the physician, might have given greater insight into determinants of the patients’ responses. The primary source of bias in this study was the fact that the study was carried out by the same physician who cared for the patients being surveyed. Many patients with a longstanding relationship might not have wanted to upset a close relationship. This may have resulted in overly positive results. Possible patient concerns about lack of anonymity, especially with open questions, may have added to this problem. Nevertheless, the patients were promised and given absolute confidentiality and no identification in the data analysis. Another potentially significant source of bias was the 18% nonresponse rate. Many of these patients, specifically may have been critical of the doctor’s sharing information.
A study in which 50 physicians with HIV and other major illnesses were interviewed illuminates the feelings, struggles and dilemmas of these men and women, transformed into the role of physician-patient. Many of these physicians elected to self-disclose because of the discomfort in concealing or the conviction that their experience with illness and treatment could benefit their patients. Others decided against being open, feeling that disclosure was irrelevant or could threaten the relationship’s stability, lead to stigmatisation or harm them professionally . Essays by physician–patients caring for patients with the same diagnosis have poignantly described both the tangible benefits and the potential dangers of sharing their stories with patients . The current study is the first (to the best of our knowledge) that specifically and systematically examines the patients’ reactions to learning that their physician has a major illness and to self-disclosure about the illness by the physician. While to some extent confirming the impressions and beliefs of physicians and researchers in previous reports, this patient-oriented study reveals a rich world of patient experience and emotion that has not been previously described. Discovering that the healer himself is a vulnerable, even mortal human being was an eye-opening revelation for many respondents. This possibility had not occurred to them and did not fit into the accepted paradigms of an omniscient professional who prevents and cures illness. Cognitive coping with questions about possible death is naturally threatening, especially when it concerns those in whom we entrust our health. While much has been written about the therapist’s vulnerability in the mental health professions , these patients’ remarks illuminate just how poignant these issues are regarding the medical profession as well. The potentially negative consequences to the patient of physician self-disclosure, posited by several authors , were reflected in the comments of only a tiny minority in this sample. To some extent, this may be explained by the brief and reassuring manner in which the news was shared but it may also demonstrate that such concerns have been overstated. Perceived or actual risk of self-disclosure to the impaired physician’s career has been the subject of previous works, most notably regarding mental illness and also conditions such as substance abuse, HIV and advanced malignancy . To what degree the fear of disclosure is based on reality from the patient’s vantage point is not well known. In the current study, patients did not have to deal with the mixed feelings and dilemmas that would have arisen if the doctor had been perceived as posing a threat, unable to fully function, or having a more questionable future. It remains unanswered whether, under a different constellation of circumstances, their emotional responses would have been qualitatively different or whether they would have considered leaving the doctor.
The study adds to current knowledge about when to share this powerful information with patients. It warrants further investigation and discussion, especially in the education of future doctors.
Physician self-disclosure of major illness to patients can potentially reveal the physician’s humanity, encourage empathy on the part of patients, and strengthen the physician-patient relationship.
|
Effect of steroid eyedrops after trabeculectomy in glaucoma patients:
a keratograph analysis | f100732c-5b61-4daf-af9e-33fdd57d1566 | 11826601 | Surgical Procedures, Operative[mh] | Glaucoma is a leading cause of irreversible blindness and visual impairment . The disease is characterized by progressive optic nerve changes that may lead to loss of visual function and decrease in vision-related quality of life (QoL) . Epidemiological studies estimate that approximately 60.5 million people worldwide have glaucoma, and it is predicted that the number will increase to 79.6 million by 2020, mostly because of the rapidly aging population . Different parameters are associated with loss of QoL in these patients, including the use of eyedrop medications or the need for numerous surgical procedures in their eyes . The use of therapies that lower ocular pressure has already been reported to been associated with ocular surface disease (OSD) . Some patients may have irritation, burning, ocular dryness, lacrimation, foreign-body sensation, red eye, or blurred vision. It is known that both the active principle of ocular hypotensive eye drops and the preservative used, usually benzalkonium chloride (BAK), can cause and/or aggravate changes in the ocular surface . Careful observation is needed particularly for eyes that are treated with multiple eye drops, especially in older patients or those who have additional eye problems . A previous reported that moderate or severe OSD affected 38% of patients who re ceived a single topical therapy, 54% of those who received two topical therapies, and 71% of those who received three or more topical therapies . Previous studies have also investigated the relationship between OSD and surgical procedures in patients with glaucoma . One of the most concerning effects of the subclinical inflammation caused by glaucoma medication is the failure of filtration surgery, which is frequently the last alternative in the treatment of glaucoma . Baudouin et al. and Johnson et al. demonstrated that the duration and number of glaucoma medications used by the patient directly affect filtration surgery. In addition, several studies have demonstrated that the preoperative hypercellularity of chronic inflammatory cells (including fibroblasts, macrophages, and lymphocytes) of the trabecular meshwork is greatly reduced in patients who have undergone successful surgeries . Therefore, there is solid ground to assume that the presence of the chronic inflammatory response associated with preservative toxicity, specifically BAK toxicity, may cause adverse surgical outcomes as a result of the fibrosis of the bleb, which in turn indicates a strong positive correlation between glaucoma medication and surgery failure . Furthermore, corticosteroids have been reported to decrease symptoms of ocular irritation and corneal fluorescein staining in cases of OSD . In a retrospective clinical series by Marsh and Pflugfelder, topical administration of a 1% solution of nonpreserved methylprednisolone, given three to four times daily for two weeks, to patients with Sjögren’s syndrome keratoconjunctivitis sicca (KCS) provided moderate to complete symptom relief in all patients . This therapy was even effective for patients with severe KCS who demonstrated no improvement with maximum aqueous enhancement therapies . In a prospective, randomized clinical trial by Sainz de la Maza et al., topical treatment of dry eye patients with nonpreserved methylprednisolone and punctual plugs significantly decreased the severity of ocular irritation symptoms and corneal fluorescein staining as compared with the group receiving punctual occlusion alone . Until now, elevated intraocular pressure (IOP) has been considered the major known risk factor for glaucoma progression . IOP can be lowered using different methods, including topical medications, laser procedures, or incisional surgery . Among incisional surgeries, trabeculectomy is the most frequently used, and it is effective for decreasing IOP . However, the surgery can lead to some side effects, including OSD . Thus, the purpose of this study was to investigate the use of preoperative steroid eyedrops for OSD in glaucoma patients undergoing trabeculectomy using subjective (e.g., Ocular Surface Disease Index [OSDI]) and objective (e.g., keratograph and clinical analysis) parameters. This interventional study adhered to the tenets of the Declaration of Helsinki and was approved by the Institutional Review Board of the Federal University of São Paulo. In addition, all participants provided written informed consent. We included all patients who had an indication for trabeculectomy in the next few months. Trabeculectomy was performed by different surgeons in a standardized manner. The trabeculectomy technique was the same for all patients and consisted of topical or peribulbar anesthesia, fornix-based dissection of conjunctiva and tenon, application of mitomycin C in the subtenonian space for three minutes, confection of the scleral flap, trabeculectomy (with punch instrument), and application of flow-control sutures (Nylon 10-0) in the borders of the scleral flap, based on the surgeon’s intraoperative impression of flow control. All patients were treated with loteprednol etabonate ophthalmic suspension 0.5% four times a day for 1 week before trabeculectomy. Data from the baseline (day of surgery) and the follow-up visit (2 weeks after surgery) were included in the analysis. Study participants A total of 31 patients with open-angle glaucoma were included in the study. Only patients with glaucoma who had an indication for trabeculectomy from March 2018 until December 2018 were included. Glaucoma was defined as the presence of repeatable (≥three consecutive) abnormal standard automatic perimetry (SAP) test results on the 24-2 program of the visual field (Humphrey Field Analyzer; Carl Zeiss Meditec, Inc) or progressive glaucomatous optic disk changes noted on masked examination of stereo photographs, regardless of the results of the SAP testing. Abnormal SAP was defined as the presence of a pattern standard deviation index outside the 95% confidence limits or glaucoma hemifield test results outside the reference range. The exclusion criteria were (1) systemic diseases affecting the ocular surface, (2) any acute disease affecting the ocular surface (e.g., acute conjunctivitis), (3) use of contact lenses, (4) previous ocular surgery or trauma, and (5) history of OSD prior to starting hypotensive agents or a history of chronic BAK exposure. Demographic and socioeconomic parameters To avoid bias in our main results, we evaluated patients’ socioeconomic and clinical parameters. All par ticipants completed a questionnaire in which were asked to provide information on the following items: gender (female yes/no), ethnicity (black yes/no), marital status (married yes/no), and educational level (at least high school degree yes/no). These variables were added because they could affect the patient’s perception of QoL. To evaluate possible morbidities, the presence or history of several diseases was investigated, such as high blood pressure, diabetes mellitus, arthritis, heart disease, stroke, depression, cancer, and asthma. The comorbidity index was calculated from the sum of some scores given to each item. Patient visual acuity (VA) and number of topical medications were also collected from all patients. VA was measured using the Early Treatment Diabetic Retinopathy Study, and logMAR calculations were also included in the analysis. Ocular surface disease index All patients answered two questionnaires: a general epidemiological questionnaire and a questionnaire called the Ocular Surface Disease Index (OSDI) . The questionnaire was validated in Brazil by Prigol et al. and is a 12-item scale divided into three categories with the aim of assessing symptoms related to dry eye disease and their effect on vision . The first part is associated visual function (questions 1 to 5), the second part with ocular symptoms (questions 6 to 9), and the third with environmental triggers (questions 10 to 12). Each item is scored on a scale ranging from 0 to 4 according to the frequency of the symptoms: 0 indicates symptoms none of the time; 1, some of the time; 2, half of the time; 3, most of the time; and 4, all the time. The total OSDI score is calculated on the basis of the following formula: OSDI = [(sum of the score for all the questions answered) × 100]/[(total number of questions answered) × 4]. The total score ranges from 0 to 100, with higher scores indicating worse OSD. Clinical evaluation All patients underwent a detailed ophthalmic examination including best corrected VA, slit-lamp examination, IOP (Goldmann), and fundoscopy. To evaluate OSD, we used tear breakup time (BUT), bulbar redness (BR), and the presence/absence of keratitis. BUT was classified as (1) less than 5 seconds, (2) between 5 and 10 seconds, and (3) greater than 10 seconds. BR was scored from 0 to 4 according to the Institute for Eye Research-Brien Holden Vision Institute scales using comparative photos in which 0 indicated an absence of BR; 1, very slight BR; 2, slight BR; 3, moderate BR; and 4, severe BR. Keratitis was evaluated by staining the cornea cell surface with fluorescein eyedrops and classified accor ding to the absence or presence of kereatitis (slight, moderate, or severe). The ophthalmologic examination was performed last to avoid any influence of fluorescein on the stability of the tear film or the ocular surface. Keratograph analysis The Keratograph 5M (Oculus, Wetzlar, Germany) is a noninvasive imaging device that uses infrared light and has automated features that do not require topical anesthesia, fluorescein staining, white light, or manual timing . It was used to objectively analyze the ocular surface features by quantifying tear meniscus height (TMH), BR by keratograph (BR-K), noninvasive tear BUT (NIKBUT), and meibomian glands (meibography). Tear meniscus height was analyzed using Oculus TMH tool images, which were graded perpendicular to the lid margin at the central point relative to the pupil center and measured in millimeters. NIKBUT was measured by using infrared light video from the NIKBUT tool, which measures time in seconds until the first breakup of tears (NIKBUT FIRST), as well as using a graph that shows the location of the first break. BR-K was assessed and scored automatically by the keratograph using a photo of the anterior biomicroscopy. The light scan detects vessels in the conjunctiva and evaluates the degree of redness. The keratography scale of BR was scored from 0 to 4 according to the Institute for Eye Research-Brien Holden Vision Institute scales using comparative photos of BR, with 0 indicating no BR; 1, very slight BR; 2, slight BR; 3, moderate BR; and 4, severe BR. For the meibomian evaluation, we used a meibography tool to generate IR images of the tarsal conjunctiva. The upper and lower eyelid were everted, and manual grading of the meibomian gland images was performed using a meiboscale (degrees from 0 to 4): 0 eyes with total meibomian integrity, 1 to an area of meibomian loss less than 25%, 2 to an area of loss from 25% to 50%, 3 to an area of loss from 51% to 75%, and 4 to an area of loss more than 75% of the total area. Statistical analysis The descriptive analysis included the mean and standard deviation for variables with a normal distribution, whereas variables that were not distributed normally were presented as the median. We used the skewness-kurtosis test to confirm normality. The t test was used for multiple comparisons between pre-and postoperative measurements, and for non-normal variables, the corresponding nonparametric test (Wilcoxon rank test) was performed. Percentages were used to describe categorical values and achieve better comparators between the two groups. All statistical analyses were performed using the available software Stata version 13 (StataCorp LP, College Station, TX). The alpha level (type I error) was set at 0.05. A total of 31 patients with open-angle glaucoma were included in the study. Only patients with glaucoma who had an indication for trabeculectomy from March 2018 until December 2018 were included. Glaucoma was defined as the presence of repeatable (≥three consecutive) abnormal standard automatic perimetry (SAP) test results on the 24-2 program of the visual field (Humphrey Field Analyzer; Carl Zeiss Meditec, Inc) or progressive glaucomatous optic disk changes noted on masked examination of stereo photographs, regardless of the results of the SAP testing. Abnormal SAP was defined as the presence of a pattern standard deviation index outside the 95% confidence limits or glaucoma hemifield test results outside the reference range. The exclusion criteria were (1) systemic diseases affecting the ocular surface, (2) any acute disease affecting the ocular surface (e.g., acute conjunctivitis), (3) use of contact lenses, (4) previous ocular surgery or trauma, and (5) history of OSD prior to starting hypotensive agents or a history of chronic BAK exposure. To avoid bias in our main results, we evaluated patients’ socioeconomic and clinical parameters. All par ticipants completed a questionnaire in which were asked to provide information on the following items: gender (female yes/no), ethnicity (black yes/no), marital status (married yes/no), and educational level (at least high school degree yes/no). These variables were added because they could affect the patient’s perception of QoL. To evaluate possible morbidities, the presence or history of several diseases was investigated, such as high blood pressure, diabetes mellitus, arthritis, heart disease, stroke, depression, cancer, and asthma. The comorbidity index was calculated from the sum of some scores given to each item. Patient visual acuity (VA) and number of topical medications were also collected from all patients. VA was measured using the Early Treatment Diabetic Retinopathy Study, and logMAR calculations were also included in the analysis. All patients answered two questionnaires: a general epidemiological questionnaire and a questionnaire called the Ocular Surface Disease Index (OSDI) . The questionnaire was validated in Brazil by Prigol et al. and is a 12-item scale divided into three categories with the aim of assessing symptoms related to dry eye disease and their effect on vision . The first part is associated visual function (questions 1 to 5), the second part with ocular symptoms (questions 6 to 9), and the third with environmental triggers (questions 10 to 12). Each item is scored on a scale ranging from 0 to 4 according to the frequency of the symptoms: 0 indicates symptoms none of the time; 1, some of the time; 2, half of the time; 3, most of the time; and 4, all the time. The total OSDI score is calculated on the basis of the following formula: OSDI = [(sum of the score for all the questions answered) × 100]/[(total number of questions answered) × 4]. The total score ranges from 0 to 100, with higher scores indicating worse OSD. All patients underwent a detailed ophthalmic examination including best corrected VA, slit-lamp examination, IOP (Goldmann), and fundoscopy. To evaluate OSD, we used tear breakup time (BUT), bulbar redness (BR), and the presence/absence of keratitis. BUT was classified as (1) less than 5 seconds, (2) between 5 and 10 seconds, and (3) greater than 10 seconds. BR was scored from 0 to 4 according to the Institute for Eye Research-Brien Holden Vision Institute scales using comparative photos in which 0 indicated an absence of BR; 1, very slight BR; 2, slight BR; 3, moderate BR; and 4, severe BR. Keratitis was evaluated by staining the cornea cell surface with fluorescein eyedrops and classified accor ding to the absence or presence of kereatitis (slight, moderate, or severe). The ophthalmologic examination was performed last to avoid any influence of fluorescein on the stability of the tear film or the ocular surface. The Keratograph 5M (Oculus, Wetzlar, Germany) is a noninvasive imaging device that uses infrared light and has automated features that do not require topical anesthesia, fluorescein staining, white light, or manual timing . It was used to objectively analyze the ocular surface features by quantifying tear meniscus height (TMH), BR by keratograph (BR-K), noninvasive tear BUT (NIKBUT), and meibomian glands (meibography). Tear meniscus height was analyzed using Oculus TMH tool images, which were graded perpendicular to the lid margin at the central point relative to the pupil center and measured in millimeters. NIKBUT was measured by using infrared light video from the NIKBUT tool, which measures time in seconds until the first breakup of tears (NIKBUT FIRST), as well as using a graph that shows the location of the first break. BR-K was assessed and scored automatically by the keratograph using a photo of the anterior biomicroscopy. The light scan detects vessels in the conjunctiva and evaluates the degree of redness. The keratography scale of BR was scored from 0 to 4 according to the Institute for Eye Research-Brien Holden Vision Institute scales using comparative photos of BR, with 0 indicating no BR; 1, very slight BR; 2, slight BR; 3, moderate BR; and 4, severe BR. For the meibomian evaluation, we used a meibography tool to generate IR images of the tarsal conjunctiva. The upper and lower eyelid were everted, and manual grading of the meibomian gland images was performed using a meiboscale (degrees from 0 to 4): 0 eyes with total meibomian integrity, 1 to an area of meibomian loss less than 25%, 2 to an area of loss from 25% to 50%, 3 to an area of loss from 51% to 75%, and 4 to an area of loss more than 75% of the total area. The descriptive analysis included the mean and standard deviation for variables with a normal distribution, whereas variables that were not distributed normally were presented as the median. We used the skewness-kurtosis test to confirm normality. The t test was used for multiple comparisons between pre-and postoperative measurements, and for non-normal variables, the corresponding nonparametric test (Wilcoxon rank test) was performed. Percentages were used to describe categorical values and achieve better comparators between the two groups. All statistical analyses were performed using the available software Stata version 13 (StataCorp LP, College Station, TX). The alpha level (type I error) was set at 0.05. We included 31 eyes of 31 patients with glaucoma. The mean age of the glaucoma patients was 69.90 ± 10.77 years. Of the sample, 18 (58.06%) were female pa tients and 10 (32.26%) were of Caucasian ancestry. The average VA was 0.40 ± 0.34 logMAR. The average comorbidity index and number of oral medications were 1.10 ± 0.91 and 1.32 ± 0.48, respectively. It is important to point that, for the first month after the surgery, only topical antibiotic eyedrops (monofloxacin 0.5%) four times a day and steroid eyedrops (prednisolone acetate 1%) were prescribed for all subjects. summarizes the demographic and clinical findings of our study. For the subjective analysis, the overall OSDI prevalence rate was 27.20 ± 17.56 units. The mean value for the questions associated with visual function (questions 1 to 5) was 7.42 ± 5.37 units. The mean value for the questions associated with ocular symptoms (questions 6 to 9) was 2.68 ± 2.41 units, and the mean value for the questions related to environmental triggers (questions 10 to 12) was 2.97 ± 3.87 units. summarizes the overall results from the OSDI. The clinical evaluations showed that 58.00% of patients had an absence of BR before surgery and 48.39% of patients had slight BR after surgery ( p =0.056). Of the patients, 42.00% patients had slight keratitis before surgery, and 39.00% also had slight keratitis after the surgery (p=0.787). A total of 51.60% of patients had BUT between 5 and 10 seconds before surgery, and 48.39% of patients had BUT between 5 and 10 seconds after the surgery (p=0.537). For the keratograph analysis (TMH, BR-K, NIKBUT, meibography quantification for the upper and lower eyelid), the only two parameters that were significantly different before and after trabeculectomy was the BR-K and the average of the NIKBUT . After trabeculectomy, patients presented with more conjunctival hyperemia compared with before surgery (1.42 ± 0.36 and 1.68 ± 0.48, respectively, p=0.013; ). After trabeculectomy, patients presented shorter NIKBUT compared with before the surgery (16.22 ± 2.37 and 14.98 ± 3.1, respectively, p=0.041; ). There was no significant difference in TMH or quantification of meibography before and after treatment (p>0.05 for all comparisons). In this study, we found that although the use of loteprednol etabonate ophthalmic suspension 0.5% may improve OSD in glaucoma patients, our patients presented with more conjunctival hyperemia as measured by keratograph analysis. Previous studies have already addressed this issue in glaucoma patients and reported an improvement in OSD after administration of loteprednol etabonate. However, because these patients had recently undergone trabeculectomy surgery, they might present with worse OSD. To our knowledge, this preliminary study is the first to attempt to compare the same patient before and after trabeculectomy using keratograph analysis. The presence of conjunctival hyperemia measured by clinical evaluation has been discussed extensively in the past years. For example, the Glaucoma Adherence and Persistency Study (GAPS) evaluated 300 open-angle glaucoma patients without previous treatment (surgical or clinical) in the past 6 months using a data interview. The study showed that hyperemia was the most common side effect of topical eyedrops, and it was responsible for the stopping or switching of medication in 63% of patients, especially in those using prostaglandin. Park et al. examined the common side effects of topical an tiglaucomatous medications as well as the factors affecting compliance with glaucoma treatment and found that conjunctival injection, stinging sensation, and blurred vision were the most frequent uncomfortable side effects. Using keratograph technology, Pérez Bartolomé et al. compared the ocular redness from 211 eyes of 211 patients with open-angle glaucoma or ocular hypertension using topical medication with 51 eyes of healthy volunteers and found statistically significant results. In addition, Pérez et al. showed that higher redness scores were recorded in the medication group (p<0.01 for all scores). The present study found no difference in terms of keratitis before and after surgery using clinical assessment or keratograph analysis. After analyzing almost 400 eyes, Ono et al. found that the severity of OSD after trabeculectomy is related to its intensity before the surgery and also reported no statistically significance difference in keratitis or tear BUT before and after the procedure. Zhong et al. recently analyzed OSD after trabeculectomy using keratograph technology. Their study recruited 81 patients without previous OSD with an indication of trabeculectomy and followed them through 3 months postoperatively. Results showed worse tear BUT and fluo rescein stain in the first month, with partial recovery at the third month. However, this final recovery was still worse than the baseline. Many studies have proven that it is not just glaucoma eye drops that contribute to OSD, but also that the intraoperatively use of mitomycin C is a factor . However, this is the first study that has attempted not only to analyze the effect of preoperative steroid eye drops on OSD but also to objectively quantify these findings using keratograph technology. It is important to address some specific points of this study. Despite being the first study to use keratographs to investigate OSD in glaucoma patients, this was a cross-sectional study that enrolled a small sample size. In addition, although the clinical examinations were performed by only one ophthalmologist and findings were classified based on a well-established scale, issues remain regarding the study’s subjectiveness. Another point that must be considered is the fact that patients might present with worse OSD because of recent trabeculectomy surgery, as any surgery can be considered a “trauma” for the eye and stimulate proinflammatory agents. In conclusion, although loteprednol etabonate ophthalmic suspension 0.5% has been associated with an improvement in OSD in glaucoma patients, our sample presented with more conjunctival hyperemia as measured by keratograph analysis. Because these patients recent underwent trabeculectomy surgery, they might present with worse OSD. |
Diabetes knowledge, health literacy and diabetes self-care among older adults living with diabetes in Alexandria, Egypt | fe13e169-fd28-40a9-8aa3-b75e370e3438 | 11481765 | Health Literacy[mh] | The study sheds light on health literacy as an important and relatively new health parameter that is not thoroughly studied among older adults in developing countries, particularly Egypt. Among older adults living with diabetes, knowledge of diabetes and health literacy were essential determinants of self-care behaviors. Educating people with diabetes and promoting health literacy can help seniors care for themselves and have better health outcomes.
Diabetes mellitus is a global public health problem that has been rising for the past few decades. In 2021, 537 million individuals between 20 and 79 years old had diabetes and it is predicted to reach 643 million by 2030 and 783 million by 2045. Among adults aged 75–79 years diabetes prevalence was estimated to be 24.0% in 2021 and predicted to rise to 24.7% in 2045. The aging of the world’s population will produce an increasing proportion of those with diabetes being over the age of 60 years . The U.S. Department of Health and Human Services, Office of Disease Prevention & Health Promotion defined health literacy (HL) as “the degree to which individuals have the ability to find, understand, and use information and services to inform health-related decisions and actions for themselves and others” . Personal, situational, and socio-environmental factors were found to be associated with HL. Personal factors include age, gender, education, income, occupation and health insurance. Situational factors include marital status, living situation and social support. Socio-environmental factors include culture and spoken local language . Low HL has been reported as a significant predictor of poor health outcomes. Patients with low HL tend to be frequent users of emergency medical services . Health Literacy was categorized into three levels; functional/basic literacy, which includes simple reading and writing skills and basic understanding of common diseases and health care systems; communicative literacy, which includes communication, social skills and critical literacy which includes enhanced cognitive skills that can be used to evaluate information and apply it to regulate situations which will affect the health of the individual and the entire community . Diabetes specific HL skills include understanding verbal or written directions, apprehending appointment details, comprehending educational brochures and reading instructions or labels on pill bottles and fully understanding informed consent documents . Patients with diabetes must know the signs and symptoms of hyperglycemia and hypoglycemia, how to correctly self-administer oral medications and insulin to achieve the best glycemic control. They must also know appropriate foot-care routine and how to monitor their blood glucose level and what to do if the levels are very high or very low . Age was found to be negatively associated with diabetes knowledge and self-care practices, with older adults having less knowledge about managing their condition and its complications . The Health and Retirement Study in the USA reported that impairments in self-care activities were more prevalent among older adults with diabetes compared to non-diabetic older adults . Diabetes is an important chronic condition that requires extensive knowledge, continuous education and self-care management. Patients who are involved in their own disease management have a much better probability of learning about the disease and tend to maintain good glycemic control and adhere to self-care activities that could prevent complications . Several studies reported that good diabetes knowledge had a strong positive association with better glycemic control and better HbA1c levels . Although diabetes self-management in Egypt was previously investigated . Our research aims to address a specific gap by concentrating on older adults and examining the relationship between health literacy, diabetes knowledge, and self-care activities. By focusing on this vulnerable and underexplored population, our study seeks to generate insights that can inform targeted interventions and enhance diabetes outcomes among older adults in Egypt.
Study design, setting, and study participants A cross-sectional study was conducted among 400 older adults with type 2 diabetes and extended over a three-month period between June and August 2021. Study participants were recruited from two outpatient clinics affiliated to health insurance organization in Alexandria, Egypt. These clinics were selected because they have the highest attendance rate of diabetic patients treated on an outpatient basis in Alexandria. We excluded older adults with any communication problems such as hearing impairment. Sample size was calculated based on the assumption that 50% of patients have insufficient information about diabetes. Using a margin error of 5%, and alpha error of 0.05, the minimum sample size required was 384 which was rounded to 400 participants. The sample size was calculated using Epi info7 software. Measures A questionnaire was used to assess health literacy, diabetes literacy and self-care activities in older adults living with diabetes. The questionnaire was composed of the following parts: Predesigned structured interview questionnaire to collect the following data : socio-demographic data (such as age, gender, marital status, education, income, occupation and living situation), personal habits such as practicing regular physical activities (participant was considered physically active if he or she achieved 30 min of exercise, 5 times per week), smoking status and diabetes history including duration of diabetes and current treatment. The Arabic version of All Aspects of Health Literacy Scale (AAHLS) The scale was originally developed by Chinn and McCarthy in 2013 as an effective measure of health literacy . It is composed of 14 items assessing reading skills and understanding health information (Functional literacy), communication with health professionals (Communicative literacy) and using health information and capability to take action for one’s health (Critical literacy). AAHLS is a 3-point Likert scale scored as follow: “rarely” (0), “sometimes” (1), and “often” (2). Only the functional literacy section is scored as “rarely” (2), “sometimes” (1) and “often” (0). Higher scores indicate better health literacy levels. The Arabic version of Short Test of Functional Health literacy in Adults (STOFHLA) The original version of STOFHLA was created by Baker et al. . The STOFHLA has two sections: reading section and numeracy section. Only the numeracy section was used in the present study. It consists of four questions that evaluate understanding of glucose monitoring, prescription labels and appointment notice. The original total score for the STOFHLA is 100 points, 70 points for the reading section and 30 points for the numeracy Sect. (7.5 points for each correct answer of the four questions). Higher scores in each section indicate better functional health literacy levels. The Arabic version of Summary of Diabetes Self-Care activities Scale (SDSCA) The original English version of SDSCA was developed by Toobert et al. in 2000 . It consists of eight questions that assess four main aspects of diabetes self-care: diet, exercise, blood-glucose testing and foot care. The mean number of days per week for all four aspects of diabetes self-care activities (diet, exercise, blood glucose testing and foot-care) was calculated. Scores range from 0 to 7 with higher scores suggesting better self-care activities. We classified the responses into two categories according to the total mean score of SDSCA after another study done in 2016, with scores (˂ 3) representing poor self-care activities and (≥ 3) representing good self-care activities . The Arabic version of the revised brief Diabetes Knowledge Test (DKT2) The DKT2 was developed by Fitzgerald et al. in 2016 to assess the general knowledge of diabetes and its self-care . DKT2 contains 23 questions and is divided into two parts that can be used separately. Part 1 was translated and validated in Arabic by Alhaiti et al. in 2016 . It contains 14 questions that test the general knowledge about diabetes, and it is suitable for both type 1 and type 2 diabetes on any treatment. Part 2 contains 9 questions, and it is designed for patients who are using insulin only. Part 2 was translated by the research team using forward backward translation method. The total questionnaire consists of multiple-choice questions and each right answer was given one point. The higher the score is, the higher level of diabetes literacy the participant has. The original questionnaire does not have cut off scores for categorization of diabetes knowledge. A percent score (percent score = total score of each participant / total number of questions x 100) was calculated for all the participants, both who answered part 1 (14 questions) and those who answered both parts 1 and 2 (23 questions), so that all of the participants could be categorized together. The total percent score was categorized according to a study done in 2016 as; low diabetes knowledge (≤ 59%), average diabetes knowledge (60–74%) and high diabetes knowledge (≥ 75%) . Anthropometric measures : Weight and height of each participant were measured to calculate the Body Mass Index (BMI). The weight in kilograms and the height in meters were obtained using a standard office scale and by the standard procedure. The body mass index was calculated using the following formula: BMI = kg/m 2 . Participants were categorized according to BMI into 3 categories according to the WHO classification for adults over 20 years old: Normal weight: 18.5–24.9 kg/m 2 , Overweight: 25–29.9 kg/m 2 and Obese: ≥ 30 kg/m 2 . Blood pressure (BP) measurement : The BP of each participant was carefully measured using a standard mercury device twice and the mean of the two measurements was taken. We classified blood pressure according to the American diabetes association (ADA) recommendations for target BP in older adults with diabetes into 3 categories as follows : controlled BP (< 140/90), uncontrolled BP (≥ 140/90) and isolated systolic hypertension (systolic BP ≥ 140 + diastolic BP < 90). Glycated hemoglobin (HbA1c in %) was obtained from participants’ records. Prior to data collection, informed consent was obtained from each participant, wherein they were thoroughly informed about the purpose of the study, the type of data being collected, and their right to withdraw from the study at any point without any consequences. All personal identifiers were removed from the data to maintain anonymity. The data were coded and stored in a secure, password-protected database accessible only to the research team. Additionally, the study protocols were reviewed and approved by the institutional review board (IRB) to ensure compliance with ethical guidelines. A pilot study ( n = 20) was conducted to check the accuracy, reliability and the time needed to complete the questionnaire, there were no corrections needed. Time needed to complete the questionnaire ranged from 30 to 40 min. Participants in the pilot study were not included in the study sample. The clinics were visited 4 times per week. The average number of interviewed older adults per day was from 5 to 8. Response rate was 92%. Statistical methods The data was managed and analyzed using statistical software SPSS version 25 (Armonk, NY: IBM Corp). The Kolmogorov-Smirnov test was used to assess if the data follow normal distribution. Our data was found to be not normally distributed. Qualitative data were described using frequency and percentage. Quantitative data were described using median and interquartile range. Spearman’s correlation coefficient was determined for linear correlation of two quantitative variables. Chi-squared (χ2) test was used to analyze the associations between qualitative data. Multivariate logistic regression was calculated to assess predictors of good diabetes self-care activities. All variables in the bivariate analysis with p value < 0.2 were entered into the model. The significance of the obtained results was judged at the 5% level.
A cross-sectional study was conducted among 400 older adults with type 2 diabetes and extended over a three-month period between June and August 2021. Study participants were recruited from two outpatient clinics affiliated to health insurance organization in Alexandria, Egypt. These clinics were selected because they have the highest attendance rate of diabetic patients treated on an outpatient basis in Alexandria. We excluded older adults with any communication problems such as hearing impairment. Sample size was calculated based on the assumption that 50% of patients have insufficient information about diabetes. Using a margin error of 5%, and alpha error of 0.05, the minimum sample size required was 384 which was rounded to 400 participants. The sample size was calculated using Epi info7 software.
A questionnaire was used to assess health literacy, diabetes literacy and self-care activities in older adults living with diabetes. The questionnaire was composed of the following parts: Predesigned structured interview questionnaire to collect the following data : socio-demographic data (such as age, gender, marital status, education, income, occupation and living situation), personal habits such as practicing regular physical activities (participant was considered physically active if he or she achieved 30 min of exercise, 5 times per week), smoking status and diabetes history including duration of diabetes and current treatment. The Arabic version of All Aspects of Health Literacy Scale (AAHLS) The scale was originally developed by Chinn and McCarthy in 2013 as an effective measure of health literacy . It is composed of 14 items assessing reading skills and understanding health information (Functional literacy), communication with health professionals (Communicative literacy) and using health information and capability to take action for one’s health (Critical literacy). AAHLS is a 3-point Likert scale scored as follow: “rarely” (0), “sometimes” (1), and “often” (2). Only the functional literacy section is scored as “rarely” (2), “sometimes” (1) and “often” (0). Higher scores indicate better health literacy levels. The Arabic version of Short Test of Functional Health literacy in Adults (STOFHLA) The original version of STOFHLA was created by Baker et al. . The STOFHLA has two sections: reading section and numeracy section. Only the numeracy section was used in the present study. It consists of four questions that evaluate understanding of glucose monitoring, prescription labels and appointment notice. The original total score for the STOFHLA is 100 points, 70 points for the reading section and 30 points for the numeracy Sect. (7.5 points for each correct answer of the four questions). Higher scores in each section indicate better functional health literacy levels. The Arabic version of Summary of Diabetes Self-Care activities Scale (SDSCA) The original English version of SDSCA was developed by Toobert et al. in 2000 . It consists of eight questions that assess four main aspects of diabetes self-care: diet, exercise, blood-glucose testing and foot care. The mean number of days per week for all four aspects of diabetes self-care activities (diet, exercise, blood glucose testing and foot-care) was calculated. Scores range from 0 to 7 with higher scores suggesting better self-care activities. We classified the responses into two categories according to the total mean score of SDSCA after another study done in 2016, with scores (˂ 3) representing poor self-care activities and (≥ 3) representing good self-care activities . The Arabic version of the revised brief Diabetes Knowledge Test (DKT2) The DKT2 was developed by Fitzgerald et al. in 2016 to assess the general knowledge of diabetes and its self-care . DKT2 contains 23 questions and is divided into two parts that can be used separately. Part 1 was translated and validated in Arabic by Alhaiti et al. in 2016 . It contains 14 questions that test the general knowledge about diabetes, and it is suitable for both type 1 and type 2 diabetes on any treatment. Part 2 contains 9 questions, and it is designed for patients who are using insulin only. Part 2 was translated by the research team using forward backward translation method. The total questionnaire consists of multiple-choice questions and each right answer was given one point. The higher the score is, the higher level of diabetes literacy the participant has. The original questionnaire does not have cut off scores for categorization of diabetes knowledge. A percent score (percent score = total score of each participant / total number of questions x 100) was calculated for all the participants, both who answered part 1 (14 questions) and those who answered both parts 1 and 2 (23 questions), so that all of the participants could be categorized together. The total percent score was categorized according to a study done in 2016 as; low diabetes knowledge (≤ 59%), average diabetes knowledge (60–74%) and high diabetes knowledge (≥ 75%) . Anthropometric measures : Weight and height of each participant were measured to calculate the Body Mass Index (BMI). The weight in kilograms and the height in meters were obtained using a standard office scale and by the standard procedure. The body mass index was calculated using the following formula: BMI = kg/m 2 . Participants were categorized according to BMI into 3 categories according to the WHO classification for adults over 20 years old: Normal weight: 18.5–24.9 kg/m 2 , Overweight: 25–29.9 kg/m 2 and Obese: ≥ 30 kg/m 2 . Blood pressure (BP) measurement : The BP of each participant was carefully measured using a standard mercury device twice and the mean of the two measurements was taken. We classified blood pressure according to the American diabetes association (ADA) recommendations for target BP in older adults with diabetes into 3 categories as follows : controlled BP (< 140/90), uncontrolled BP (≥ 140/90) and isolated systolic hypertension (systolic BP ≥ 140 + diastolic BP < 90). Glycated hemoglobin (HbA1c in %) was obtained from participants’ records. Prior to data collection, informed consent was obtained from each participant, wherein they were thoroughly informed about the purpose of the study, the type of data being collected, and their right to withdraw from the study at any point without any consequences. All personal identifiers were removed from the data to maintain anonymity. The data were coded and stored in a secure, password-protected database accessible only to the research team. Additionally, the study protocols were reviewed and approved by the institutional review board (IRB) to ensure compliance with ethical guidelines. A pilot study ( n = 20) was conducted to check the accuracy, reliability and the time needed to complete the questionnaire, there were no corrections needed. Time needed to complete the questionnaire ranged from 30 to 40 min. Participants in the pilot study were not included in the study sample. The clinics were visited 4 times per week. The average number of interviewed older adults per day was from 5 to 8. Response rate was 92%.
The data was managed and analyzed using statistical software SPSS version 25 (Armonk, NY: IBM Corp). The Kolmogorov-Smirnov test was used to assess if the data follow normal distribution. Our data was found to be not normally distributed. Qualitative data were described using frequency and percentage. Quantitative data were described using median and interquartile range. Spearman’s correlation coefficient was determined for linear correlation of two quantitative variables. Chi-squared (χ2) test was used to analyze the associations between qualitative data. Multivariate logistic regression was calculated to assess predictors of good diabetes self-care activities. All variables in the bivariate analysis with p value < 0.2 were entered into the model. The significance of the obtained results was judged at the 5% level.
The study sample included 400 older adults living with diabetes; the majority were males (56.2%). The age of the participants ranged between 60 and 87 years with a mean of 65.75 ± 5.15 years and 77% of them were in the youngest age group (60–69 years). More than half of the study sample were married (61.2%) and nearly two thirds were below university level (62.7%). 43% (43.2%) of the participants reported that they did not have enough income or were in debt. Most of them (88.5%) were not working (housewives or retired). Regarding living situation, only 16.2% were living alone (Table ). The table also shows that only 22% of the study sample were practicing regular physical activities and only (9.3%) were current smokers. Nearly half (47.5%) had diabetes for 10 years or more, 65.5% were managed with oral hypoglycemic drugs and (16.0%) were managed with both oral hypoglycemic drugs and insulin. Regarding BMI, more than half of (51.2%) were overweight. According to ADA recommendations for target BP in older adults with diabetes, 23.8% had uncontrolled blood pressure and 13.2% had isolated systolic hypertension. Moreover, most of the sample had uncontrolled diabetes as measured by HbA1c (79.4%) (Table ). The total score of the AAHLS ranged from 4 to 24 and the median score was 12.0. The score of part 1 of DKT2 ranged from 2 to 12 with a median score of 5.0 while the score of part 2 ranged from 3 to 7 with a median score of 4.0. The Total score of correct answers calculated by percentage ranged from 14.2 to 85.7% with a median percent score of 43.4%. The total score of STOFHLA range was 0–30 with a median score of 22.5 (Table ). The highest correct answers were in understanding clinic appointment details (88.5%), understanding prescription instructions (75.2%) followed by understanding instructions on medication label (61.3%). The lowest correct answers were in understanding normal blood sugar ranges (32.3%) (Fig. ). The median score of SDSCA was 2 (Table ). Three quarters of the study participants (73.5%) had poor self-care activities (< 3) (Fig. ) The total DKT2 score was calculated through percentage of correct answers for each participant, 77.8% had low diabetes knowledge (≤ 59%) and only 6% had high diabetes knowledge (≥ 75%) (Fig. ). Association between health literacy and diabetes self-care activities was examined. A significant positive correlation was found between AAHLS and STOFHLA ( r = 0.768), SDSCA ( r = 0.338), DKT2 part 1 ( r = 0.336) and DKT2 part 2 ( r = 0.447). SDSCA was also found to have a significant positive correlation with both DKT2 part 1( r = 0.379), part 2 ( r = 0.254) and STOFHLA ( r = 0.338). Both AAHLS and SDSCA had significant negative correlation with HbA1c ( r = -0.156 and -0.560 respectively) (Table ). Education, income level and practicing physical activities were found to have a statistically significant association with good self-care activities. For education, participants who had a university degree were found to have the higher odds of good self-care activities (OR = 1.941, 95%CI; 1.591–2.368, p < 0.001). As for income levels, participants who reported having enough income had higher self-care activities (OR = 1.941, 95%CI; 1.591–2.368, p < 0.001). Those who practiced regular physical activities were also found to have higher self-care activities (OR = 1.941, 95%CI; 1.591–2.368, p < 0.001) (Table ). However, results of multiple logistic regression analysis shows that health literacy and diabetes knowledge were found to be the only significant predictors of good self-care activities ((aOR = 1.132; 95% CI:1.062–1.207, p < 0.001 and aOR = 1.313; 95% CI: 1.178–1.464, p < 0.001; respectively) (Table ).
The present study demonstrated that good self-care activities were independently associated with health literacy and diabetes knowledge. Moreover, diabetes self-care activities have a significant association with glycemic control. It was found that 73.5% of older adults with diabetes had poor self-care activities as measured by Summary of diabetes self-care activities (SDSCA) scale. Similar results (72.8%) were reported by Harikrishna et al. in a community based cross-sectional study among older adults with diabetes in India. On the other hand, a study conducted in Ethiopia reported that only 49.1% of patients had poor self-care activities. However, this study was conducted among a younger age group (40–60 years old) . Education was significantly associated with good self-care activities (SDSCA) in this study; those with university degrees had the highest self-care activities score. Most of the available literature suggests a strong significant association between level of education and HL . Similarly another study in Egypt found that better diabetes knowledge was associated with higher level of education . The consistency of these results is not surprising as knowledge is gained through education and adults with low education have been reported to have lower self-efficacy levels . Furthermore, studies in Ethiopia and Iran reported that lower educational level was a significant predictor of poor self-care activities . Highly educated individuals have been found to have good decision-making abilities and self-efficacy specifically in context of diabetes self-care management. These results highlight the value of education in the promotion and attainment of HL and better self-care practices and good health outcomes in all individuals, especially the older adults. Income level showed a statistically significant association with good self-care activities, with participants reporting sufficient income having higher self-care activity scores. Guo et al. , in their systematic review, also noted that health literacy was influenced by income level. Comparable findings were reported by Yılmazel & Cici in Turkey and Xie et al. in China. Additionally, a study in Iran found a significant positive association between monthly income and diabetes self-care activities. People with higher income are more likely to be educated and mostly have better access to health information and health-care services through private channels. Furthermore, in our culture, people with poor income usually have more pressing priorities than focusing on their health. Diabetes self-care management also requires a certain level of income to maintain a healthy lifestyle, obtain a home glucose monitoring device and regularly follow-up with health-care professionals in absence of an adequate health insurance system. Therefore, it is imperative to assess income-related inequality in both HL and diabetes literacy, because rectifying this inequality is a potential opportunity to improve health outcomes in the population. Physical activity had a statistically significant association with self-care activities in the present study. Older adults who practiced regular physical activity had higher scores on the SDSCA. A study in Brazil also found a statistically significant association between physical activity and better diabetes knowledge and positive attitude to self-care activities in older adults living with diabetes . Vicente et al. also reported a significant positive association between physical activity and some domains of self-care activities such as exercise and foot-care among older adults living with diabetes in Brazil. Patients who have good HL presumably know the significant value of physical activity in managing their condition and its impact on overall health status. Understanding the impact of physical activity on glycemic control in patients with diabetes is a crucial aspect of diabetes knowledge and self-care activities . In the current study, HbA1c was found to have a significant strong negative association with self-care activities in older adults living with diabetes. Similar to the results in this study; a cross-sectional study in a university hospital in Korea reported that diabetes self-care activities had a significant negative correlation with both HbA1c and fasting blood sugar . On the other hand, a study in Qatar reported no significant relationship between self-care activities and HbA1c levels . This result was surprising because most of the available studies reported a significant relationship between them. Good self-care activities, among other factors, were found to cause better glycemic control and diabetes outcomes in several studies . HbA1c was also found to have a significant negative correlation with AAHLS, STOFHLA and DKT2. Similar to our study, studies in Turkey and Brazil reported that HL had a statistically significant negative association with HbA1c levels in older adults living with diabetes . Surprisingly, a cross-sectional study in Qatar, with a mean age of 50.7 years, reported that there was no significant association between diabetes knowledge and glycemic control . In this context, low HL and diabetes knowledge prove to be vital variables that can explain the prevalence of poor glycemic control in patients with T2D. Several studies have highlighted the strong association between HL and diabetes management . Sufficient HL was reported to be a necessity for the effective utilization of the tools of diabetes self-care management . In the current study, a statistically significant correlation was found between all measures of HL and diabetes knowledge and self-care activities. Also, results of multiple logistic regression analysis shows that health literacy and diabetes knowledge were found to be the only significant predictors of good self-care activities. Moreover, Forghani et al. in Iran and Zhao in China reported a significant link between HL and diabetes self-care activities. Moreover, a cross-sectional study in Malaysia among older adults living with diabetes, reported a significant positive association between diabetes knowledge and self-care activities . Several cross-sectional studies have also reported a positive association between diabetes knowledge and self-care activities . All these results highlight the significance of diabetes knowledge in achieving good self-care practice and eventually better glycemic control and disease outcome in older adults with diabetes. The significant association between health literacy, diabetes knowledge, and self-care activities highlights the critical role of education and literacy in managing chronic conditions like diabetes. While our study was conducted in a specific population within Alexandria, the underlying principles regarding the impact of health literacy and diabetes knowledge on self-care are likely applicable to older adults’ population in Egypt, and also, in regions with similar sociodemographic profiles. However, caution should be exercised when generalizing these results to different cultural or healthcare contexts, where variations in healthcare access, education systems, and social support networks may influence the outcomes. Study limitations One of the limitations of this study is related to the cross-sectional research design, which can’t verify the cause-effect relationships between the variables. Moreover, this study used subjective self-rated health questions which could be subject to various social implications. Also, the actual estimate of health literacy levels in the studied sample could not be assessed or compared to other studies because all aspects of health literacy scale (AAHLS) have no cut-off score or universal total score. Conclusion and recommendations Health literacy and diabetes knowledge were identified as predictors of good self-care activities in older adults with diabetes, and all three; health literacy, diabetes knowledge, and self-care activities were significantly associated with diabetes outcomes as measured by HbA1c. Based on these findings several recommendations can be made. Firstly, enhancing health literacy programs for older adults with type 2 diabetes is crucial. These programs should be tailored, particularly for individuals with lower educational backgrounds, to improve their understanding and practice of diabetes management, medication adherence, and lifestyle modifications. Additionally, it is recommended to develop and offer personalized diabetes education sessions that cater to the individual needs of older adults. These sessions should consider the varying levels of health literacy and diabetes knowledge among patients to ensure effective learning and application of self-care practices. Furthermore, incorporating routine health literacy assessments into clinical practice can help identify those with low health literacy early, allowing healthcare providers to offer additional support and resources. By focusing on these areas, healthcare professionals can better support older adults in managing their diabetes, ultimately improving self-care activities and glycemic control.
One of the limitations of this study is related to the cross-sectional research design, which can’t verify the cause-effect relationships between the variables. Moreover, this study used subjective self-rated health questions which could be subject to various social implications. Also, the actual estimate of health literacy levels in the studied sample could not be assessed or compared to other studies because all aspects of health literacy scale (AAHLS) have no cut-off score or universal total score.
Health literacy and diabetes knowledge were identified as predictors of good self-care activities in older adults with diabetes, and all three; health literacy, diabetes knowledge, and self-care activities were significantly associated with diabetes outcomes as measured by HbA1c. Based on these findings several recommendations can be made. Firstly, enhancing health literacy programs for older adults with type 2 diabetes is crucial. These programs should be tailored, particularly for individuals with lower educational backgrounds, to improve their understanding and practice of diabetes management, medication adherence, and lifestyle modifications. Additionally, it is recommended to develop and offer personalized diabetes education sessions that cater to the individual needs of older adults. These sessions should consider the varying levels of health literacy and diabetes knowledge among patients to ensure effective learning and application of self-care practices. Furthermore, incorporating routine health literacy assessments into clinical practice can help identify those with low health literacy early, allowing healthcare providers to offer additional support and resources. By focusing on these areas, healthcare professionals can better support older adults in managing their diabetes, ultimately improving self-care activities and glycemic control.
|
Regression Analysis of Cancer Incidence Rates and Water Fluoride in the U.S.A. based on IACR / IARC (WHO) Data (1978-1992) | f871a159-f9e5-412d-96cc-b4217e512001 | 11735073 | Dentistry[mh] | Since water fluoridation for prevention of dental caries was initiated in several American and Canadian communities in 1945 ) , a large number of epidemiological studies have indicated the possibility of adverse effects, including increased risk of cancer development. From an evalution of over 50 epidemiological studies on cancer mortality or morbidity and fluoride levels in drinking water EG Knox ) in the U.K. and RN Hoover ) in the U.S.A. concluded no credible association. However there are methodological problems suggesting that re-appraisal might be necessary. For this purpose we have here employed data on registered cancers from the volumes entitles “Cancer Incidence in Five Continents” published by the International Agency for Research on Cancer (IARC)/World Hearth Organization (WHO) in 1987, 1992 and 1997 ) . The unpublished study by the National Cancer Institute, U.S. provided epidemiological evidence of relation between cancer incidence and water fluoridation ) in 1987. The findings provoked concern over the possibility that fluoride might have carcinogenic activity and prompted the NCI (National Cancer Institute), the EPA (Environmental Protection Agency), and the NIDR (National Institute for Dental Research) to nominate sodium fluoride for further study. The National Toxicology Program (NTP) ) of the U.S., Public Health Service (PHS), therefore conducted standard rat and mouse carcinogenicity studies on water fluoridation. Four cases of osteosarcoma were found among 261 male rats. In November, 1990, NTP concluded that there was “equivocal” evidence of carcinogenicity in fluoride. It also supplied a detailed description of the toxicology of fluoride not only in terms of osteosarcomas but also lesions in the oral mucosa, thyroid gland, skin and uterus. In February, 1991, PHS published a monograph entitled “Review of Fluoride: Benefits and Risks” ) , in which Hoover again negated completely any epidemiological relationship between water fluoridation and osteosarcoma and several other cancers on the basis of their registered SEER ) (Surveillance, Epidemiology and End Results) of cancer statistics (p.79-83, E-3, F-l). However, the results of the animal experiments by NTP prompted us to re-test the hypothesis of an epidemiological association between water fluoridation and cancer incidences in different sites of the human body. 1. The most important data sources were the three volumes of “Cancer Incidence in Five Continents” ) , giving the mean age-specific rates of registered cancer for the five year periods 1978-82, 1983-87 and 1988-92, and also the age-standardized rates relative to the world population (ASR world/100,000) for forty-five body sites (ICD, WHO, 1957). The data for three States and six cities of the U.S.A. were obtained from these publications (five other communities were excluded for the reasons given in Annex A). The nine communities in question, distributed widely over the U.S.A. included a total population of 21.8 million (males and females separately, mainly white). The three volumes supplied comparable surveillance data altogether for fifteen years. Data for water fluoridation were obtained from the “Fluoridation Census 1985” ) of the Public Health Service (PHS), U.S.A., giving the numbers of citizens receiving fluoridated water adjusted optimally to 1 ppm and using water with naturally occurring fluoride at levels of 0.7 ppm or higher. From these Census data the ratio of citizens receiving the above-defined fluoridated water to the total population of the community was calculated (Fluoridation Index, FD) (See Annex. B) , ) . For cases of lip cancer and melanoma of the skin the percentage of possible sunshine (SS) cited from the “Weather Almanac” ) was considered as an additional variable which may be influential in these sites. 2. From the standpoint of stochastics (R.A.Fisher), to maintain the homogeneity of the obtained data a rejection test was performed before analysis and rejected outlying values at the level of P < 0.05. Furthermore, according to the principle of the biostatistical assay (bioassay), the independent variables x (FD and SS) to be applied for regression analysis were transformed to their logarithm. The incidence rates of cancer y following the Poisson distribution were normalized by logarithmic transformation. These two logalithmic transformation confirms the linearity of the regression curve between log y and log x. The cancer incidence ratios FD at the level of 100%/1% (CIR · 100) were defined as the magnitude when the exposure of the inhabitants to fluoride increased hundred times from 1 to 100%. 1. The results of regression analysis The results of our regression analysis on the geometric means for three five-years’ ASRs (1978-82, 1983-87, 1988-92) are shown in , with 1) year where necessary, 2) regression coefficient b, 3) probability P for significance of b, 4) cancer incidence ratio CIR · 100, separately for males and females. When an analysis of a particular five-years’ ASR showed some deviation from the total mean, these data are also tabulated. Significant regression coefficients were obtained for about two-thirds of thirty six sites (positive in 63.9% and negative in 11.1%). I. Cancer of digestive organs ( -1, -1) In males, cancer of the digestive organs (oral cavity, pharynx, esophagus, colon, rectum, liver, gallbladder, and pancreas) were associated positively with the fluoride index (FD), the cancer incidence ratio (CIR · 100) ranging from 1.3 to 2.1. In females, cancers of most of these organs correlated with the FD only in 1990. For the incidence rates of cancer of the salivary gland, stomach and small intestine significant associations with fluoride intake were rare or never found in males or females. II. Cancer of the respiratory organs ( -II, ) In males, cancer of the bronchus and lung showed a significant correlation with FD. III. Cancer of the urinary organs ( -III, ) In females, cancers of the kidney and urinary bladder were associated with FD, whereas in males this was only the case for cancer of the urinary bladder. IV. Cancer of the sexual and hormonal organs ( -IV, ) The incidence rates for prostate and ovarial cancers were correlated with FD negatively in the former in 1978-82 and positively in the latter in 1988-92. In females, thyroid cancer was correlated negatively with FD. V. Bone cancer ( -V) The incidence rate of bone cancer as the mean of three five- years ASRs was significantly correlated with FD only in males, with CIR · 100 of 1.22, whereas in 1978-82 it showed a high CIR · 100 of 2.53. (a detailed analysis of the age specificity of the bone cancer in young boys will be reported in the future). VI. Brain tumors ( -VI) Brain tumors were correlated with FD significantly in 1978-82 in males, and in 1983-87 in males and females. VII. Lymphoma, multiple myeloma and leukaemia ( -VII. -2) Only in females, Hodgkin’s diseases and Non-Hodgkin lymphoma and multiple myeloma were associated with FD. Monocytic leukaemia was associated with FD significantly in males. VIII. Lip cancer and melanoma of the skin ( -VIII, ) The incidence rates for lip cancer were analyzed in relation to FD and SS. Significant negative associations were observed with FD in males, and with FD and SS in females. In case of melanoma of the skin, negative associations were evident with FD in both sexes. No association with SS was evident in males and females. IX. Cancer of all sites ( -IX) In males total cancers were associated significantly with FD, with CIR · 100 of 1.23. In females the CIR · 100 of FD was 1.19, but this was sub-significant. 2. Distribution of cancer incidence rates (CIR · 100) In both sexes CIR · 100 values ranged mainly from 1.20 to 1.90 (as a mean 1.50), but highest 3.00 in males and 5.00 in females. The results of our regression analysis on the geometric means for three five-years’ ASRs (1978-82, 1983-87, 1988-92) are shown in , with 1) year where necessary, 2) regression coefficient b, 3) probability P for significance of b, 4) cancer incidence ratio CIR · 100, separately for males and females. When an analysis of a particular five-years’ ASR showed some deviation from the total mean, these data are also tabulated. Significant regression coefficients were obtained for about two-thirds of thirty six sites (positive in 63.9% and negative in 11.1%). I. Cancer of digestive organs ( -1, -1) In males, cancer of the digestive organs (oral cavity, pharynx, esophagus, colon, rectum, liver, gallbladder, and pancreas) were associated positively with the fluoride index (FD), the cancer incidence ratio (CIR · 100) ranging from 1.3 to 2.1. In females, cancers of most of these organs correlated with the FD only in 1990. For the incidence rates of cancer of the salivary gland, stomach and small intestine significant associations with fluoride intake were rare or never found in males or females. II. Cancer of the respiratory organs ( -II, ) In males, cancer of the bronchus and lung showed a significant correlation with FD. III. Cancer of the urinary organs ( -III, ) In females, cancers of the kidney and urinary bladder were associated with FD, whereas in males this was only the case for cancer of the urinary bladder. IV. Cancer of the sexual and hormonal organs ( -IV, ) The incidence rates for prostate and ovarial cancers were correlated with FD negatively in the former in 1978-82 and positively in the latter in 1988-92. In females, thyroid cancer was correlated negatively with FD. V. Bone cancer ( -V) The incidence rate of bone cancer as the mean of three five- years ASRs was significantly correlated with FD only in males, with CIR · 100 of 1.22, whereas in 1978-82 it showed a high CIR · 100 of 2.53. (a detailed analysis of the age specificity of the bone cancer in young boys will be reported in the future). VI. Brain tumors ( -VI) Brain tumors were correlated with FD significantly in 1978-82 in males, and in 1983-87 in males and females. VII. Lymphoma, multiple myeloma and leukaemia ( -VII. -2) Only in females, Hodgkin’s diseases and Non-Hodgkin lymphoma and multiple myeloma were associated with FD. Monocytic leukaemia was associated with FD significantly in males. VIII. Lip cancer and melanoma of the skin ( -VIII, ) The incidence rates for lip cancer were analyzed in relation to FD and SS. Significant negative associations were observed with FD in males, and with FD and SS in females. In case of melanoma of the skin, negative associations were evident with FD in both sexes. No association with SS was evident in males and females. IX. Cancer of all sites ( -IX) In males total cancers were associated significantly with FD, with CIR · 100 of 1.23. In females the CIR · 100 of FD was 1.19, but this was sub-significant. -1, -1) In males, cancer of the digestive organs (oral cavity, pharynx, esophagus, colon, rectum, liver, gallbladder, and pancreas) were associated positively with the fluoride index (FD), the cancer incidence ratio (CIR · 100) ranging from 1.3 to 2.1. In females, cancers of most of these organs correlated with the FD only in 1990. For the incidence rates of cancer of the salivary gland, stomach and small intestine significant associations with fluoride intake were rare or never found in males or females. -II, ) In males, cancer of the bronchus and lung showed a significant correlation with FD. -III, ) In females, cancers of the kidney and urinary bladder were associated with FD, whereas in males this was only the case for cancer of the urinary bladder. -IV, ) The incidence rates for prostate and ovarial cancers were correlated with FD negatively in the former in 1978-82 and positively in the latter in 1988-92. In females, thyroid cancer was correlated negatively with FD. -V) The incidence rate of bone cancer as the mean of three five- years ASRs was significantly correlated with FD only in males, with CIR · 100 of 1.22, whereas in 1978-82 it showed a high CIR · 100 of 2.53. (a detailed analysis of the age specificity of the bone cancer in young boys will be reported in the future). -VI) Brain tumors were correlated with FD significantly in 1978-82 in males, and in 1983-87 in males and females. -VII. -2) Only in females, Hodgkin’s diseases and Non-Hodgkin lymphoma and multiple myeloma were associated with FD. Monocytic leukaemia was associated with FD significantly in males. -VIII, ) The incidence rates for lip cancer were analyzed in relation to FD and SS. Significant negative associations were observed with FD in males, and with FD and SS in females. In case of melanoma of the skin, negative associations were evident with FD in both sexes. No association with SS was evident in males and females. -IX) In males total cancers were associated significantly with FD, with CIR · 100 of 1.23. In females the CIR · 100 of FD was 1.19, but this was sub-significant. In both sexes CIR · 100 values ranged mainly from 1.20 to 1.90 (as a mean 1.50), but highest 3.00 in males and 5.00 in females. Fluoride as the strongest electronegative element reacts with other elements or chemicals to form a wide variety of inorganic and organic compounds in vivo. The present broad analysis of the association between cancer incidence rates in different organs and tissues of the body and the water fluoridation index (FD) revealed a characteristic spectrum of association. The water fluoridation initiated in the U.S.A. and Canada in 1945 has been promoted widely throughout the world under the recommendation of the WHO since 1969. Because of this situation we have now obtained over a half century of information on the human pathophysiology of fluoride. 1. The Principal Structure of the Spectrum of Cancer Sites Correlated with Water Fluoridation The dimension of the spectrum was reduced to 36 from original 45 after rejection of 9 not informative sites of neoplasia marked as “unspecified” or “others”. With regard to the association with fluoride, 23 were significantly positive (63.9%), 9 not significant (25.0%) and 4 significantly negative (11.1%). The data thus indicate a complexity of the action mechanisms of fluoride in the body. shows a list of the principal part of risk factors for the most common types of cancers cited from JR Bertino ) and partly from Harrison’s Principles of Medicine ) . None of the most common causes of cancer appears to demonstrate such a super-multidimensional and triphasic spectrum as fluoride. 2. An Approach to the Water Fluoridation as a Genetic Cause of Cancers From the epistemogical viewpoint, epidemiology may be principally a phenomenological approach, which can identify associations among several phenomena, but not directly assess causal genesis. However, if it satisfies conditions of consistency, strength, specificity, temporal relationships and coherence of the association in terms of physiology, conclusions can be drawn. The following is our attempt to evaluate our results in the light of the five criteria which were standardized in the epidemiological research on “Smoking and Health”, U.S.A., in 1964 , ) . 1) The consistency of the association : Significances were detected for total average of the mean values in 1977-82, 1983-87 and 1988-92, whereas the tongue, oropharynx, nose and sinuses, prostate, brain and nerves, and monocytic leukaemia were significant only in one or two of the particular five-year periods in males. In females significant association was observed in about a half of the total averages and for particular five-year means. Though WHO/IARC supplies a registered cancer incidence data for many countries in five continents, most do not provide fluoridation census data. Therefore our analysis was here limited to data for the U.S.A.. However, we found the SEER cancer data in Appendix E of the “Review of fluoride” (1991) ) unanalysed. Our analysis detected fluoride-associations for cancers in 11 among 26 sites (42.3%) in males and 6 among 29 (20.7%) in female, which coincided essentially with our results of analysis by IACR (WHO)/IACR (unpublished). As a sample of biological inconsistency the completely inverse response regarding hip fracture of menopausal females will be discussed in the following paper. 2) The strength of the association : Significancy was confirmed as a result of dose-response relationship with registered cancer incidence rates (ASR) at thirty six sites (ICD) to FD, covering 21.8 million inhabitants during fifteen years 1978-92, which was supplied by the IACA (WHO) and IARC ) . The calculation of FD is basing on the Fluoridation Census 1985 (CDC, U.S.A) ) . About a half were significant at the level of P < 0.01. The factor for excess cancer incidence rates in the inhabitants under the assumption that FD increased from 1% to 100%, was estimated to be about 1.5 . Under the condition that the water fluoridation as an artificial public nuisance does not usually exceed 1 ppm, the total daily intake of fluoride may not be more than a few milligram per day. The level of the strength of the fluoride-association seems moderately high, given the wide distribution in the body, reaching 3-5 in CIR · 100 for some sites. Principally it must be noted that the power of the strength of the association required must be evaluated relative to the necessity of fluoride for prevention of teeth caries, which has yet to be confirmed. 3) The specificity of the association : Twenty three of thirty six cancer sites (63.9%) were associated positively with FD. Such a broad spectrum association has never been observed for any particular known carcinogen, but it may be reasonable for fluoride, because of its strong electronegative nature. A number for points must be noted here. ① Lung cancer associated with FD is in accordance with exposure to the higher concentration of fluoride in the pulmonary vein blood which is absorbed directly through the stomach wall rapidly as hydrogen fluoride (HF) ) as states later on stomach cancer. ② Cancers in the oral cavity, bones and multiple myeloma associated with fluoride are in accordance with the higher magnitude of accumulation of fluoride in these sites - ) . ③ Cancers of the colon and rectum, urinary bladder and gallbladder associated with FD are in accordance with the extended presence of fluoride within the colorectal mass, urine and bile. ④ Brain tumors associated with FD are in accordance with infusion of fluoride facilitated by inactivation of the blood-brain-barrier (BBB), under the action of aluminum-fluoride (AlF 3 ) or sodium fluoride (NaF) ) . ⑤ Negative association of thyroid cancer with FD is in accordance with hypofunction of thyroid gland following to water fluoridation ) . ⑥ Non-association of stomach cancer with FD might be the result of the death of precancerous cells by the toxic action of hydrogen fluoride (HF) produced from sodium-fluoride (NaF) ingested and hydrogen chloride secreted in the gastric juice ) . ⑦ A negative association of prostate cancer and a positive association of ovarian cancer with FD are in coincidence with the decrease of circulating testosterone and increase of circulating gonadtropin observed in workers exposed to a F-containing compound (cryolite) for 10-25 years, though this has not yet been confirmed for areas of water fluoridation ) . ⑧ Lip cancer and melanoma of the skin may be provoked primarily by ultraviolet ray. As a possible mechanism of a negative association with fluoridation index (FD), a hypothesis on production of some toxic substance (for example HF) from fluoride under ultraviolet ray may be proposed. However this must be confirmed experimentally. ⑨ Hodgkin’s disease, Non-Hodgkin lymphoma and multiple myeloma associated with fluoride may be summarized as diseases of the T-cell system. Monocytes are a precursors for macrophages. Thus there appears to be an association between fluoride and malignant change in the T-cell system ) . An overview of the above-cited specifity of fluoride-associated cancers connected with the pathophysiology of fluoride in human body may indicate little room for any other influential co-factors. 4) The temporal relationship of the association : The first volume of registered cancer data by WHO/IARC was published in 1987, whereas, the situation of water fluoridation in the U.S.A. was only stationary after the publication of the Fluoridation Census 1985. Therefore analysis of the temporal relationship of the association between cancer incidence rates and FD was not possible in our analysis. However, when a significant regression has been confirmed for the range from zero to full percentage of variation of x (here FD), the dose may substitute for ‘the time’ as an alternative. As in our research all sites of cancer were confirmed by regression analysis, the criteria for a ‘Temporal’ relation of the association may be evaluated as fulfilled. 5) The coherence of the association: ① Clastogenicity in cultured cells i) Ogura ) (1995) reported that fluoride ion is clastogenic at 5-10 ppm and lethally injurious at 20 ppm using cultured human diploid cell (IMR-90). He also says that clastogenicity appeared interlinked with growth depression and G1 arrest on continuous treatment at the level of 10 ppm. These values suggest the possibility of cancer in oral cavity, as the concentration of fluoride in plaque fluid reaches some times to 30 ppm after water fluoridation and 30, 100 or 200 ppm after topical application of fluoride. ii) Mihashi and Tsutsui ) (1996) reported that cultured rat vertebral body-derived cells (RVBd) treated with NaF at 10∼40 ppm showed the reduction of growth and/or survival in a dose-dependent manner. Significant increases in the frequencies of chromosome aberrations were induced in a dose- and treatment time-dependent fashion with 10 and 40 ppm (0.5-2.0 mM) NaF for 24 and 48 hours. They concluded that NaF is genetoxic and carcinogenetic to rat vertebra. It is well established that fluoride accumulates in bones up to several hundred or thousand ppm during aging. ② Carcinogenicity in animal experiments Reacting to concern over the possibility of carcinogenic activity obtained from the epidemiological analysis by NCI et al, U.S.A. (1987), the NTP conducted experiment in male rats which provided the answer in 1990 of “yes” ) (equivocal) for carcinogenecity, as stated in the introduction. This is important as it may compensate for the lack of an analysis of “the temporal relationship of the association” in our present study. ③ Invalidity of the negation of osteosarcoma in young males by Hoover (NCI, U.S.A.)-Dose-response relationship is not always valid in the discussion of a causal genesis Hoover’s principal logic for negation of a causal relationship between fluoride and osteosarcoma in younger males was clarified in the last replay to the request of the Subcommittee in his Appendix F (F-2). He stated “For osteosarcomas among males, increases were seen for those under age 20 in both the “fluoridated” and “non-fluoridated” areas, although more prominently in the “fluoridated” counties ( , F-6). In the following page (F-3) he stated again “The ratios for osteosarcomas are lowest in the longest duration categories, probably citing Table 5 and 6 (F-7). These statements suggest that Hoover did not appreciate the principle of childhood cancer, which results in an elevated plateau of incidence rates at specified ages determined by children’s physiology during their development, which was numerically expressed in the regression analysis to the magnitude of FD as will be discussed in our next paper. But this phenomenon was already indicated in Miller’s paper ) (G1-3), published as an NCI Monograph in 1981 as “death rates for children with bone cancer rose as a function of (increasing gradient of) their stature”. Quite a surprise to say, following Hoover’s last statement in his Appendix F-7, we should find to Appendix G-l ∼ 3 entitled “Osteosarcoma” which introduced Miller’s paper in 3 pages, without the writer’s name. Moreover, in the list of members of the Subcommittee we find the name Robert W. Miller as Chair of the Fluoride Benefit Workgroup of the Ad-Hoc Subcommittee. ④ Reproductive toxicity and genotoxicity “Review on Fluoride : Benefits and Risks” (PHS, U.S.A., 1991) ) recommended the conduct of studies on the reproductive toxicity of fluoride and further to investigate whether or not fluoride is genotoxic (p.91). It should be a great surprise that water fluoridation has been practiced for about a half century without confirmation of the safety in terms of reproductive toxicity and genotoxicity. Takahashi (1983) ) confirmed the significant association between Down syndrome births in young mothers which Erickson et al ) (1976) had not identified in their paper because of a less effective statistical method for hypothesis testing. Further Takahashi ) (1998) clarified the methodological mistake of Erickson’s negation ) (1980) of the hypothesis on the fluoride-associated Down Syndrome births on the basis of new data, albeit of lower quality, from 44 cities. 3. Conclusion The result of epidemiologic study were here evaluated regarding causal significance of indicated associations between cancer incidence rates and water fluoridation in the light of five criteria, citing information relevant to this problem. Though the U.S.A. is the country where water fluoridation has been practiced systematically since 1945 and the Censes data on water fluoridation are available, our analysis may not be necessarily be specified to the U.S.A. Finally, we must conclude that the consistency of the fluoride-associations for cancers and temporal relationship are not yet adequately confirmed because of its sociological conditions. The strength may be limited by general toxicicity, whereas the specificity and coherence are relatively well established. We would like to ask for the cooperation of researchers throughout the world to further assess fluoride as a genetic cause of cancers from the standpoint of epidemiology and also in animal experiments, so as to strengthen the power of five criteria and stop the application of fluoride for prevention of teeth caries if this indeed presents as a risk factor for cancer. The dimension of the spectrum was reduced to 36 from original 45 after rejection of 9 not informative sites of neoplasia marked as “unspecified” or “others”. With regard to the association with fluoride, 23 were significantly positive (63.9%), 9 not significant (25.0%) and 4 significantly negative (11.1%). The data thus indicate a complexity of the action mechanisms of fluoride in the body. shows a list of the principal part of risk factors for the most common types of cancers cited from JR Bertino ) and partly from Harrison’s Principles of Medicine ) . None of the most common causes of cancer appears to demonstrate such a super-multidimensional and triphasic spectrum as fluoride. From the epistemogical viewpoint, epidemiology may be principally a phenomenological approach, which can identify associations among several phenomena, but not directly assess causal genesis. However, if it satisfies conditions of consistency, strength, specificity, temporal relationships and coherence of the association in terms of physiology, conclusions can be drawn. The following is our attempt to evaluate our results in the light of the five criteria which were standardized in the epidemiological research on “Smoking and Health”, U.S.A., in 1964 , ) . 1) The consistency of the association : Significances were detected for total average of the mean values in 1977-82, 1983-87 and 1988-92, whereas the tongue, oropharynx, nose and sinuses, prostate, brain and nerves, and monocytic leukaemia were significant only in one or two of the particular five-year periods in males. In females significant association was observed in about a half of the total averages and for particular five-year means. Though WHO/IARC supplies a registered cancer incidence data for many countries in five continents, most do not provide fluoridation census data. Therefore our analysis was here limited to data for the U.S.A.. However, we found the SEER cancer data in Appendix E of the “Review of fluoride” (1991) ) unanalysed. Our analysis detected fluoride-associations for cancers in 11 among 26 sites (42.3%) in males and 6 among 29 (20.7%) in female, which coincided essentially with our results of analysis by IACR (WHO)/IACR (unpublished). As a sample of biological inconsistency the completely inverse response regarding hip fracture of menopausal females will be discussed in the following paper. 2) The strength of the association : Significancy was confirmed as a result of dose-response relationship with registered cancer incidence rates (ASR) at thirty six sites (ICD) to FD, covering 21.8 million inhabitants during fifteen years 1978-92, which was supplied by the IACA (WHO) and IARC ) . The calculation of FD is basing on the Fluoridation Census 1985 (CDC, U.S.A) ) . About a half were significant at the level of P < 0.01. The factor for excess cancer incidence rates in the inhabitants under the assumption that FD increased from 1% to 100%, was estimated to be about 1.5 . Under the condition that the water fluoridation as an artificial public nuisance does not usually exceed 1 ppm, the total daily intake of fluoride may not be more than a few milligram per day. The level of the strength of the fluoride-association seems moderately high, given the wide distribution in the body, reaching 3-5 in CIR · 100 for some sites. Principally it must be noted that the power of the strength of the association required must be evaluated relative to the necessity of fluoride for prevention of teeth caries, which has yet to be confirmed. 3) The specificity of the association : Twenty three of thirty six cancer sites (63.9%) were associated positively with FD. Such a broad spectrum association has never been observed for any particular known carcinogen, but it may be reasonable for fluoride, because of its strong electronegative nature. A number for points must be noted here. ① Lung cancer associated with FD is in accordance with exposure to the higher concentration of fluoride in the pulmonary vein blood which is absorbed directly through the stomach wall rapidly as hydrogen fluoride (HF) ) as states later on stomach cancer. ② Cancers in the oral cavity, bones and multiple myeloma associated with fluoride are in accordance with the higher magnitude of accumulation of fluoride in these sites - ) . ③ Cancers of the colon and rectum, urinary bladder and gallbladder associated with FD are in accordance with the extended presence of fluoride within the colorectal mass, urine and bile. ④ Brain tumors associated with FD are in accordance with infusion of fluoride facilitated by inactivation of the blood-brain-barrier (BBB), under the action of aluminum-fluoride (AlF 3 ) or sodium fluoride (NaF) ) . ⑤ Negative association of thyroid cancer with FD is in accordance with hypofunction of thyroid gland following to water fluoridation ) . ⑥ Non-association of stomach cancer with FD might be the result of the death of precancerous cells by the toxic action of hydrogen fluoride (HF) produced from sodium-fluoride (NaF) ingested and hydrogen chloride secreted in the gastric juice ) . ⑦ A negative association of prostate cancer and a positive association of ovarian cancer with FD are in coincidence with the decrease of circulating testosterone and increase of circulating gonadtropin observed in workers exposed to a F-containing compound (cryolite) for 10-25 years, though this has not yet been confirmed for areas of water fluoridation ) . ⑧ Lip cancer and melanoma of the skin may be provoked primarily by ultraviolet ray. As a possible mechanism of a negative association with fluoridation index (FD), a hypothesis on production of some toxic substance (for example HF) from fluoride under ultraviolet ray may be proposed. However this must be confirmed experimentally. ⑨ Hodgkin’s disease, Non-Hodgkin lymphoma and multiple myeloma associated with fluoride may be summarized as diseases of the T-cell system. Monocytes are a precursors for macrophages. Thus there appears to be an association between fluoride and malignant change in the T-cell system ) . An overview of the above-cited specifity of fluoride-associated cancers connected with the pathophysiology of fluoride in human body may indicate little room for any other influential co-factors. 4) The temporal relationship of the association : The first volume of registered cancer data by WHO/IARC was published in 1987, whereas, the situation of water fluoridation in the U.S.A. was only stationary after the publication of the Fluoridation Census 1985. Therefore analysis of the temporal relationship of the association between cancer incidence rates and FD was not possible in our analysis. However, when a significant regression has been confirmed for the range from zero to full percentage of variation of x (here FD), the dose may substitute for ‘the time’ as an alternative. As in our research all sites of cancer were confirmed by regression analysis, the criteria for a ‘Temporal’ relation of the association may be evaluated as fulfilled. 5) The coherence of the association: ① Clastogenicity in cultured cells i) Ogura ) (1995) reported that fluoride ion is clastogenic at 5-10 ppm and lethally injurious at 20 ppm using cultured human diploid cell (IMR-90). He also says that clastogenicity appeared interlinked with growth depression and G1 arrest on continuous treatment at the level of 10 ppm. These values suggest the possibility of cancer in oral cavity, as the concentration of fluoride in plaque fluid reaches some times to 30 ppm after water fluoridation and 30, 100 or 200 ppm after topical application of fluoride. ii) Mihashi and Tsutsui ) (1996) reported that cultured rat vertebral body-derived cells (RVBd) treated with NaF at 10∼40 ppm showed the reduction of growth and/or survival in a dose-dependent manner. Significant increases in the frequencies of chromosome aberrations were induced in a dose- and treatment time-dependent fashion with 10 and 40 ppm (0.5-2.0 mM) NaF for 24 and 48 hours. They concluded that NaF is genetoxic and carcinogenetic to rat vertebra. It is well established that fluoride accumulates in bones up to several hundred or thousand ppm during aging. ② Carcinogenicity in animal experiments Reacting to concern over the possibility of carcinogenic activity obtained from the epidemiological analysis by NCI et al, U.S.A. (1987), the NTP conducted experiment in male rats which provided the answer in 1990 of “yes” ) (equivocal) for carcinogenecity, as stated in the introduction. This is important as it may compensate for the lack of an analysis of “the temporal relationship of the association” in our present study. ③ Invalidity of the negation of osteosarcoma in young males by Hoover (NCI, U.S.A.)-Dose-response relationship is not always valid in the discussion of a causal genesis Hoover’s principal logic for negation of a causal relationship between fluoride and osteosarcoma in younger males was clarified in the last replay to the request of the Subcommittee in his Appendix F (F-2). He stated “For osteosarcomas among males, increases were seen for those under age 20 in both the “fluoridated” and “non-fluoridated” areas, although more prominently in the “fluoridated” counties ( , F-6). In the following page (F-3) he stated again “The ratios for osteosarcomas are lowest in the longest duration categories, probably citing Table 5 and 6 (F-7). These statements suggest that Hoover did not appreciate the principle of childhood cancer, which results in an elevated plateau of incidence rates at specified ages determined by children’s physiology during their development, which was numerically expressed in the regression analysis to the magnitude of FD as will be discussed in our next paper. But this phenomenon was already indicated in Miller’s paper ) (G1-3), published as an NCI Monograph in 1981 as “death rates for children with bone cancer rose as a function of (increasing gradient of) their stature”. Quite a surprise to say, following Hoover’s last statement in his Appendix F-7, we should find to Appendix G-l ∼ 3 entitled “Osteosarcoma” which introduced Miller’s paper in 3 pages, without the writer’s name. Moreover, in the list of members of the Subcommittee we find the name Robert W. Miller as Chair of the Fluoride Benefit Workgroup of the Ad-Hoc Subcommittee. ④ Reproductive toxicity and genotoxicity “Review on Fluoride : Benefits and Risks” (PHS, U.S.A., 1991) ) recommended the conduct of studies on the reproductive toxicity of fluoride and further to investigate whether or not fluoride is genotoxic (p.91). It should be a great surprise that water fluoridation has been practiced for about a half century without confirmation of the safety in terms of reproductive toxicity and genotoxicity. Takahashi (1983) ) confirmed the significant association between Down syndrome births in young mothers which Erickson et al ) (1976) had not identified in their paper because of a less effective statistical method for hypothesis testing. Further Takahashi ) (1998) clarified the methodological mistake of Erickson’s negation ) (1980) of the hypothesis on the fluoride-associated Down Syndrome births on the basis of new data, albeit of lower quality, from 44 cities. The result of epidemiologic study were here evaluated regarding causal significance of indicated associations between cancer incidence rates and water fluoridation in the light of five criteria, citing information relevant to this problem. Though the U.S.A. is the country where water fluoridation has been practiced systematically since 1945 and the Censes data on water fluoridation are available, our analysis may not be necessarily be specified to the U.S.A. Finally, we must conclude that the consistency of the fluoride-associations for cancers and temporal relationship are not yet adequately confirmed because of its sociological conditions. The strength may be limited by general toxicicity, whereas the specificity and coherence are relatively well established. We would like to ask for the cooperation of researchers throughout the world to further assess fluoride as a genetic cause of cancers from the standpoint of epidemiology and also in animal experiments, so as to strengthen the power of five criteria and stop the application of fluoride for prevention of teeth caries if this indeed presents as a risk factor for cancer. |
Unsaturated Fatty Acids Are Decreased in Aβ Plaques in Alzheimer's Disease | d401ce97-b7c9-4b30-84a9-c5c092ac3cdc | 11742699 | Biochemistry[mh] | Introduction Alzheimer's disease (AD) is a neurodegenerative disorder, hallmarked by the deposition of Aβ peptides in the brain. Aβ aggregates into β‐sheet‐rich oligomers and fibrils (Kirschner et al. ; Michaels et al. ). These aggregated forms of Aβ continually accumulate in the intercellular space and drive the development of Aβ plaques in the brain (Rozemuller et al. ; Ikeda et al. ; Thal et al. ; Röhr et al. ). Aβ's interactions with surrounding tissues are complex and actively researched. Aβ, in its various aggregated forms, engages in dynamic interactions with adjacent cells, extracellular matrix components, and other proteins, shaping the microenvironment and influencing the progressive pathogenesis of AD (Hardy and Higgins ). Aβ's interaction with the lipid membranes of adjacent cells has been indicated as a potential mediator of toxicity (Arispe, Pollard, and Rojas ; Markesbery ). The amphipathic nature of Aβ allows it to integrate into lipid bilayers, disrupting their structural integrity. This perturbation, in turn, influences ion channel conductance and cellular homeostasis, contributing to synaptic dysfunction and neuronal damage (Arispe, Rojas, and Pollard ; Lin, Bhatia, and Lal ; Fantini et al. ; Ciudad et al. ). Furthermore, Aβ, especially in its aggregated form, has the capacity to directly induce oxidative stress in lipid membranes (Behl et al. ; Butterfield and Sultana ; Butterfield, Swomley, and Sultana ). The proposed mechanism involves the interaction of Aβ with metal ions and the generation of reactive oxygen species (ROS) (Smith, Cappai, and Barnham ; Butterfield et al. ). Additionally, Aβ may induce oxidative stress indirectly, by inducing mitochondrial dysfunction or by interfering with the antioxidant defense system (Reddy and Beal ; Butterfield, Swomley, and Sultana ). The abundance of highly oxidizable unsaturated fatty acids (UFAs) in the brain makes it particularly vulnerable to oxidative stress (Prasad et al. ). Oxidants abstract weak allylic hydrogens from UFAs, creating alkyl radicals. By addition of molecular oxygen, a peroxyl radical is formed, which can, in turn, abstract another allylic hydrogen from a nearby UFA. Thus, an FA hydroperoxide and another alkyl radical are produced. The latter can start the reaction cycle anew (Girotti ). This process, called lipid peroxidation, can propagate through lipid membranes, producing instable FA hydroperoxides. The FA hydroperoxides are cleaved, resulting in shorter, less unsaturated FAs and byproducts, which are detrimental for neuronal functions (Reiter ; Markesbery and Lovell ; Di Paolo and Kim ). The composition of FAs is known to be altered in the AD brain compared to the healthy control (HC) brain (Corrigan et al. ; Prasad et al. ; Fraser, Tayler, and Love ). This is particularly relevant because FAs play a central role in maintaining the fluidity and structural integrity of cell membranes (Di Paolo and Kim ). Notably, Nakada et al. revealed significantly decreased levels of UFAs in phospholipids in the temporal lobe (TL) of AD brains, compared to HC brains (Nakada, Kwee, and Ellis ). Likewise, Söderberg et al. reported decreased levels of UFAs in the major phospholipids phosphatidylethanolamine (PE) and phosphatidylcholine (PC) in multiple AD brain regions (Söderberg et al. ). In both studies, the reduction in UFAs was paralleled by increased levels of shorter saturated FAs (SFAs). The evidence of decreased UFAs in AD tissue has led to the hypothesis that Aβ‐mediated oxidative stress causes UFA degradation in Aβ plaques (Markesbery ). Garcia‐Alloza et al. observed increased levels of ROS in Aβ plaques of a transgenic AD mouse model using in vivo multi‐photon microscopy (Garcia‐Alloza et al. ). Likewise, Xie et al. reported markers of oxidative stress in neurons close to plaques and observed the resulting death of those neurons (Xie et al. ). This aligns with reports that neuronal damage follows plaque development in AD (Sheng et al. ; Xie et al. ; Kiskis et al. ). The cleavage byproducts of UFA hydroperoxides, such as 4‐hydroxy‐2‐nonenal (HNE), have been found to co‐localize with Aβ plaques (Ellis et al. ; Furman et al. ). However, recent advances have significantly expanded our understanding of the lipid composition, spatially resolved in Aβ plaques. Several studies have investigated the lipid composition in plaques (Blank and Hopf ). Notably, Kaya et al. reported the accumulation of saturated PC and concurrent depletion of unsaturated PC in plaques a transgenic AD mouse model. Similarly, Michno et al. employed multimodal imaging techniques to reveal both the decrease in unsaturated phosphoserines (PS) and PE, along with Aβ aggregation‐dependent lipid alterations in plaques, demonstrating the distinct accumulation of ceramides in cored plaques and phosphatidylinositols (PI) in diffuse plaques (Michno et al. , ). While these studies provide valuable insights, investigations into the lipidome of Aβ plaques in human AD cases have been limited. Notably, Panchal et al. employed mass spectrometry to analyze laser microdissected plaques and surrounding tissue, reporting elevated levels of saturated ceramides (Cer) and cholesterol within plaques (Panchal et al. , ). Most recently, mass spectrometry imaging has been instrumental in elucidating the lipid composition of plaques (Michno et al. ). Coupled with machine learning, it uncovered significant heterogeneity in lipid profiles between amyloid‐positive individuals with and without cognitive impairment (Enzlein et al. ). Furthermore, Huang et al. revealed the accumulation of lysophospholipids and ceramides (Cer) around Aβ plaques in advanced Braak stage AD brain tissue. This suggests spatially heterogeneous lipid changes, similar to those observed in transgenic mouse models (Huang et al. ). A significant limitation for the use of LMD extracted plaques is the necessity to perform immunohistochemical (IHC) staining prior to LMD. IHC requires the incubation of tissue sections in solvents as well as the application of reagents for fixation and blocking (Westermark, Johnson, and Westermark ). Depending on the exact IHC protocol, lipids may be dislodged from the samples during the staining procedure. Most recently, our group demonstrated the significant loss of soluble proteins in plaques during anti‐Aβ IHC (Müller et al. ). In this study, we used Fourier transform infrared (FTIR) imaging to investigate the degree of lipid unsaturation in Aβ plaques. This label‐free technique spatially resolves the distribution of chemical groups, such as the alkenes in UFAs, prior to the invasive IHC staining. We further apply label‐free quantum cascade laser infrared (QCL‐IR) imaging to detect and extract Aβ plaques in fresh frozen AD brain tissue sections. This novel method generates highly precise and chemically native plaque samples in sufficient quantities for mass spectrometric analysis (Müller et al. ). This allowed us to fully harbor the main advantage of mass spectrometry to detect and quantify a wide range of lipid species with high accuracy and reproducibility (Schwudke et al. ; Blank and Hopf ). The result is a comprehensive insight into the native plaque lipidome that allowed us to uncover a systematic depletion of long UFAs in plaques.
Methods 2.1 Case Selection Human brain tissue for this non‐pre‐registered study was obtained post‐mortem from the Netherlands Brain Bank (NBB), Netherlands Institute for Neuroscience, Amsterdam in accordance with strict ethical guidelines. The Institutional Review Board and Medical Ethical Board from Vrije University Medical Center Amsterdam approved the procedures of the NBB. All materials have been collected from donors for or from whom a written informed consent for brain autopsy and the use of the material and clinical information for research purposes had been obtained by NBB. The neuropathological diagnosis was performed using standardized procedures, with AD cases ( n = 8) meeting the National Institute on Aging‐Alzheimer's Association (NIA‐AA) criteria for AD (Montine et al. ). No formal sample size calculations were performed. The number of samples was granted from the NBB upon request. HC cases ( n = 8) were chosen based on the lack or low presence of AD pathology, indicated by their ABC scores, see Table , and no reported cognitive decline during life. To reduce confounding variables, the AD and HC groups were matched by age and sex. A detailed overview of individual patient information can be found in Table . 2.2 Brain Tissue Preparation Tissue samples from AD and HC were exclusively obtained from the TL, as this region is associated with Aβ plaque pathology in AD. Specifically, the samples were taken from the superior TL (STL) and middle TL (MTL), with details of the individual regions provided in Table . Tissue sections (10 μm) were thaw‐mounted on 1.4 μm Leica PET frame slides for vibrational imaging and subsequent anti‐Aβ IF or LMD and FIA‐MS. To preserve Aβ plaques and lipid changes in AD, samples were stored at −80°C until experimentation. 2.3 FTIR Imaging FTIR microspectroscopy was performed in transmission mode with a Cary 670 spectrometer coupled to a Cary 620 microscope (Agilent Technologies) according to established protocols (Röhr et al. ). The microscope features a nitrogen‐cooled focal plane array (FPA) detector with 128 × 128 elements and a 15× (0.62 NA) objective. The instrument can achieve a nominal pixel size of 1.1 μm at 5× optical magnification. The 128 × 128‐element data acquisition yields a field of view (FOV) of around 141 × 141 μm 2 . Interferograms were acquired with 128 scans, resulting in a spectral range of 3700–948 cm −1 at a spectral sampling interval of 1.9 cm −1 , using Blackman‐Harris‐4‐term apodization, power phase correction, and zero‐filling factor 2 for Fourier transform. Background correction was carried out by measuring the clean area of each CaF 2 slide (512 scans) and subtracting it from the sample measurements. The software Resolutions Pro 5.3 (Agilent Technologies) was used to facilitate image acquisition. Throughout the measurements, the instruments and sample cavity were continuously purged with dry air to reduce atmospheric water vapor contribution and keep the samples in a conserving dry state. 2.4 FTIR Spectral Data Analysis A lipid consists of a head group and FAs. The FAs are methylene (CH 2 ) chains ending in a methyl group (CH 3 ). UFAs contain alkene groups with double‐bonded carbon atoms (H‐C=C‐H). The =C‐H stretching vibration of these alkenes produces a distinct vibrational band 3012 cm −1 in infrared (IR) absorbance spectra (Socrates ; Dreissig et al. ). The intensity of the alkene band can be used to assess the degree of unsaturation of FAs (Yoshida and Yoshida ). The FAs are bound to the head groups by ester groups that feature a characteristic C=O stretching vibration band at 1738 cm −1 (Socrates ). The ester band serves as proxy for total lipid content (Robinson et al. ). Here, the absorbance ratio A 3012 /A 1738 is used to calculate the degree of lipid unsaturation. All calculations were performed in Python 3.10, utilizing NumPy 1.26.3, Pandas 1.5.3, and scikit‐learn 1.4.2. Significance levels were assessed using Welch's t‐test without adjustments for multiple comparisons, as only two independent groups were compared. For visualization, Statannotations 0.6 (Charlier ) was used. Confidence levels were determined based on p ‐values: < 0.05 , < 0.01 , < 0.001 , and < 0.0001 . Before analysis, FTIR spectra were thoroughly prepared to ensure data quality and reliability, as described previously (Röhr et al. ). In brief, to avoid statistical distortions caused by outlier spectra, spectra with high noise level (signal‐to‐noise‐ration SNR < 100) and high scattering were excluded from analysis. The remaining spectra were then corrected for Mie scattering using extended multiplicative signal correction (EMSC). This correction method effectively reduces the effects of Mie scattering, which can complicate spectral interpretation (Konevskikh, Lukacs, and Kohler ; Solheim et al. ). An average spectrum of a plaque core has been created by manually selecting and averaging fitting pixel‐spectra. This was done for illustration spectral differences. All plots were created with matplotlib 3.8.4 and seaborn 0.11.2. 2.5 Immunofluorescence ( IF ) Staining Anti‐Aβ IHC was performed using a multi‐label direct IF approach, staining samples with two anti‐Aβ probes with varying affinities for different Aβ polymorphs; see Figure . The goal was to distinguish Aβ fibrils from oligomeric forms without using any reagents that could alter the structure of Aβ. The samples were fixed in 4% paraformaldehyde (PFA) for 30 min and then washed with phosphate‐buffered saline (PBS). Samples were incubated with 20 μM Thioflavin T (ThT, Thermo Fisher) in 2% mouse serum (Novus Biologicals) in PBS for 45 min, washed in PBS, and then incubated overnight at 4°C with 4G8, labeled with Spark‐YG 570 (BioLegend) (1:1000) in normal antibody diluent (DAKO). Following additional washing, samples were mounted with EverBrite TrueBlack Hardset mounting medium (Biotium) and cover‐slipped. The stained sample was then imaged with an Olympus VS120 slide scanner and the UPlanSApo 20× 0.75 NA (numerical aperture) objective (Olympus). The rigorous protocol visualized Aβ using fluorescent labels ThT and 4G8. 2.6 Multimodal Image Registration To link spectral data with IF images, custom software (written in Matlab 2019a) was used to determine an affine 2D transformation, as described previously (Röhr et al. ). The user‐provided reference coordinates were used to transform the IF image into the vibrational image's coordinate system. The overlay's quality was visually confirmed while accounting for tissue morphology. To identify spectra corresponding to Aβ plaques, binary masks were generated from IF image cutouts using Otsu's method, delineating plaques, and surrounding regions. The resulting plaque spectrum was calculated as the arithmetic mean of all pixel‐spectra within the plaque mask, and the surrounding spectrum was derived from the pixel‐spectra within a ring‐shaped mask that surrounded the plaque; see Figure . 2.7 QCL ‐ IR Imaging QCL‐IR was used to generate microscopic spectral images of the tissue sections intended for the label‐free detection and extraction of plaques via LMD for FIA‐MS. The Spero‐QT 340 microspectrometer (Daylight Solutions, San Diego, USA) and Chemical Vision software version 3.2 were used for this purpose. This label‐free method employs infrared spectra, which contain useful chemical information in the form of vibrational bands (Goormaghtigh et al. ; Röhr et al. ). The spectral range of the instrument is 1800 to 948 cm −1 , with a spectral resolution of 2 cm −1 . A 4× objective (0.3 NA) was used to project a 2 × 2 mm 2 FOV onto a 480 × 480 pixel microbolometer FPA detector, yielding an image with 4.25 × 4.25 μm 2 pixel size. Prior to spectral measurements, tissue samples were thawed in a dry air‐filled container before being placed in the Spero‐QT cavity. Continuous purging with dry air was used throughout the measurements. 2.8 Aβ Plaque Detection Using a Neural Network In our previous work (Müller et al. ), we detailed the label‐free detection of plaques in QCL‐IR images using a deep convolutional neural network (CNN). In summary, we used a comparative segmenting network (CompSegNet) trained for Aβ plaque detection (Schuhmacher, Schörner, and Küpper ). The network's architecture is based on an extended U‐Net model with input dimensions of 64 × 64 × 427 pixels (Ronneberger, Fischer, and Brox ). QCL‐IR images are cropped to 64 × 64 pixels and then passed through convolution and pooling layers, which gradually reduce the image dimensions to an 8 × 8 × 512 matrix. Transposed convolutions, combined with skip‐connections, are then applied, resulting in the generation of a 64 × 64 × 1 output layer called the activation map. To facilitate whole‐slide image (WSI) segmentation, a tile‐based strategy was used, with a fixed window size of 64 × 64 pixels. Overlapping tiles of 16 pixels were created to ensure information continuity across adjacent regions, and the CNN evaluated each tile individually. Finally, the outputs from each tile were reassembled using the maximum value for overlapping regions, yielding a comprehensive WSI activation map that identifies the plaques; see Figure . 2.9 Plaque Extraction via Laser Microdissection ( LMD ) LMD is used to extract plaques from brain tissue sections, guided by the WSI activation maps, as described previously (Müller et al. ). In brief, by binarizing the activation maps with a predetermined threshold of 0.9 within a range of 0 to 1, “plaque masks” were created. Morphological operations were performed sequentially, excluding objects smaller than 100 μm 2 , dilation by 15 μm, hole filling, erosion by 10 μm, and exclusion of objects smaller than 300 μm 2 , based on eccentricity and solidity criteria; see Figure . The morphological processing and coordinate transformation were done in Matlab (version 2019a), resulting in plaque‐like shapes with a 5 μm margin to account for tissue loss during subsequent procedures. This results in plaques that are approximately spherical in shape, with areas exceeding 300 μm 2 . Following manual exclusion of damaged tissue regions, the sample was laser microdissected with a LMD microscope (PALM MicroBeam; Zeiss, Jena, Germany). A two‐dimensional Helmert transformation, which used reference points in each microscopy image, facilitated the plaque shape coordinates to the microscopy coordinate system. The PALM Robo software (Zeiss, version 4.6) was then used to import the coordinates and perform precise plaque extraction with the instrument's 20× objective. Extracted tissue samples were collected in 50 mM SDS, briefly sonified, covered with argon gas, and stored at −80°C until FIA‐MS analysis. In general, several thousand shapes were combined to reach a total area of 50 million μm 2 per sample. This procedure was applied to plaques, their surrounding regions, as well as control tissue. 2.10 Sample Preparation for FIA ‐ MS Lipid extraction was performed according to the procedure described by Bligh and Dyer (Bligh and Dyer ). A sample amount of 50 μg wet weight was subjected to lipid extraction in the presence of non‐naturally occurring internal standards. The following lipid species were added as internal standards: CE 17:0, CE 22:0, FC[D7], Cer 18:1;O2/14:0, Cer 18:1;O2[D7]/18:0, HexCer 18:1;O2/12:0, HexCer 18:1;O2[D5]/18:0, DG 14:0/14:0, DG 20:0/20:0, FC[D7], LPC 13:0, LPC 19:0, LPE 13:0, LPE 18:1[D7], PC 14:0/14:0, PC 22:0/22:0, PE 14:0/14:0, PE 20:0/20:0 (di‐phytanoyl), PI 18:1[D7]/15:0, PS 14:0/14:0, PS 20:0/20:0 (di‐phytanoyl), SM 18:1;O2/12:0, SM 18:1;O2/18:1[D9], TG 17:0/17:0/17:0, and TG 19:0/19:0/19:0. After extraction, a volume of 800 μL of the chloroform phase (total volume of 1 mL) was recovered by a pipetting robot and vacuum dried. The dried extracts were dissolved in 500 μL chloroform/methanol/2‐propanol (1:2:4 v/v/v) with 7.5 mM ammonium formate. 2.11 FIA ‐ MS Lipidomics The analysis of lipids was performed by direct FIA using a triple quadrupole mass spectrometer (FIA‐MS/MS) and a high‐resolution hybrid quadrupole‐Orbitrap mass spectrometer (FIA‐FTMS). FIA‐MS/MS was performed in positive ion mode using the analytical setup and strategy described previously (Liebisch et al. ). A Waters Acquity UPLC (Milford, Massachusetts, USA) delivered methanol/chloroform = 3/1 (v/v) with 7.5 mM ammonium acetate at an initial flow rate of 50 μL/min for 0.2 min, followed by 10 μL/min for 5.7 min and a wash at 500 μL/min for 1.9 min. The LC flow was coupled to a Waters Xevo TQ‐S micro equipped with an electrospray ionization source operated in positive mode. A fragment ion of m/z 184 was used for lysophosphatidylcholines (LPC) (Liebisch et al. ). The following neutral losses were applied: PE and lysophosphatidylethanolamine (LPE) 141, phosphatidylserine (PS) 185, and phosphatidylinositol (PI) 277 (Matyash et al. ). Sphingosine‐based Cer and hexosylceramides (HexCer) were analyzed using a fragment ion of m/z 264 (Liebisch et al. ). For FIA‐MS/MS glycerophospholipid species, annotation is based on the assumption of even numbered carbon chains only. A detailed description of the FIA‐FTMS method was published recently (Höring et al. , ). QExactive Orbitrap (Thermo Fisher Scientific, Bremen, Germany) equipped with a heated electrospray ionization source was used with the following settings: spray voltage of 3.5 kV, S‐lens RF level 50, capillary temperature of 250°C, aux gas heater temperature of 100°C, and settings of 15 for sheath gas and 5 for aux gas. Chloroform/methanol/2‐propanol (1:2:4 v/v/v) was delivered at an initial flow rate of 100 μL/min until 0.25 min followed by 10 μL/min for 2.5 min and a wash out with 300 μL/min for 0.5 min. PC, PC ether (PC O) and sphingomyelins (SM) were analyzed in negative ion mode m/z 520–960 as [M+HCOO]‐ at a target resolution of 140 000 (at m/z 200). The quantification was performed by multiplication of the spiked IS amount with analyte‐to‐IS ratio. Lipid species were annotated according to the latest proposal for shorthand notation of lipid structures that are derived from mass spectrometry (Liebisch et al. ). 2.12 Lipidomics Data Analysis Lipidomic analysis by mass spectrometry yields molar concentrations of lipid species for each sample. Here, each sample was normalized for the total molar lipid content to compare their lipid composition. Lysophospholipids were excluded from analysis because they contain only one FA and can therefore not be compared with the other lipid species investigated here. For the principal component analysis (PCA), standardized data with a mean of 0 and a standard deviation of 1 were utilized to ensure equal weighting of all features. The PCA was performed based on the correlation matrix, employing Singular Value Decomposition (SVD). All calculations were done in python 3.10, using numpy 1.26.3, pandas 1.5.3, and scikit‐learn 1.4.2. Significance levels were calculated using Welch's t‐test and visualized using Statannotations 0.6 (Charlier ). The confidence levels were determined according to p ‐values < 0.05 , < 0.01 , < 0.001 , and < 0.0001 . All plots were created with matplotlib 3.8.4 and seaborn 0.11.2.
Case Selection Human brain tissue for this non‐pre‐registered study was obtained post‐mortem from the Netherlands Brain Bank (NBB), Netherlands Institute for Neuroscience, Amsterdam in accordance with strict ethical guidelines. The Institutional Review Board and Medical Ethical Board from Vrije University Medical Center Amsterdam approved the procedures of the NBB. All materials have been collected from donors for or from whom a written informed consent for brain autopsy and the use of the material and clinical information for research purposes had been obtained by NBB. The neuropathological diagnosis was performed using standardized procedures, with AD cases ( n = 8) meeting the National Institute on Aging‐Alzheimer's Association (NIA‐AA) criteria for AD (Montine et al. ). No formal sample size calculations were performed. The number of samples was granted from the NBB upon request. HC cases ( n = 8) were chosen based on the lack or low presence of AD pathology, indicated by their ABC scores, see Table , and no reported cognitive decline during life. To reduce confounding variables, the AD and HC groups were matched by age and sex. A detailed overview of individual patient information can be found in Table .
Brain Tissue Preparation Tissue samples from AD and HC were exclusively obtained from the TL, as this region is associated with Aβ plaque pathology in AD. Specifically, the samples were taken from the superior TL (STL) and middle TL (MTL), with details of the individual regions provided in Table . Tissue sections (10 μm) were thaw‐mounted on 1.4 μm Leica PET frame slides for vibrational imaging and subsequent anti‐Aβ IF or LMD and FIA‐MS. To preserve Aβ plaques and lipid changes in AD, samples were stored at −80°C until experimentation.
FTIR Imaging FTIR microspectroscopy was performed in transmission mode with a Cary 670 spectrometer coupled to a Cary 620 microscope (Agilent Technologies) according to established protocols (Röhr et al. ). The microscope features a nitrogen‐cooled focal plane array (FPA) detector with 128 × 128 elements and a 15× (0.62 NA) objective. The instrument can achieve a nominal pixel size of 1.1 μm at 5× optical magnification. The 128 × 128‐element data acquisition yields a field of view (FOV) of around 141 × 141 μm 2 . Interferograms were acquired with 128 scans, resulting in a spectral range of 3700–948 cm −1 at a spectral sampling interval of 1.9 cm −1 , using Blackman‐Harris‐4‐term apodization, power phase correction, and zero‐filling factor 2 for Fourier transform. Background correction was carried out by measuring the clean area of each CaF 2 slide (512 scans) and subtracting it from the sample measurements. The software Resolutions Pro 5.3 (Agilent Technologies) was used to facilitate image acquisition. Throughout the measurements, the instruments and sample cavity were continuously purged with dry air to reduce atmospheric water vapor contribution and keep the samples in a conserving dry state.
FTIR Spectral Data Analysis A lipid consists of a head group and FAs. The FAs are methylene (CH 2 ) chains ending in a methyl group (CH 3 ). UFAs contain alkene groups with double‐bonded carbon atoms (H‐C=C‐H). The =C‐H stretching vibration of these alkenes produces a distinct vibrational band 3012 cm −1 in infrared (IR) absorbance spectra (Socrates ; Dreissig et al. ). The intensity of the alkene band can be used to assess the degree of unsaturation of FAs (Yoshida and Yoshida ). The FAs are bound to the head groups by ester groups that feature a characteristic C=O stretching vibration band at 1738 cm −1 (Socrates ). The ester band serves as proxy for total lipid content (Robinson et al. ). Here, the absorbance ratio A 3012 /A 1738 is used to calculate the degree of lipid unsaturation. All calculations were performed in Python 3.10, utilizing NumPy 1.26.3, Pandas 1.5.3, and scikit‐learn 1.4.2. Significance levels were assessed using Welch's t‐test without adjustments for multiple comparisons, as only two independent groups were compared. For visualization, Statannotations 0.6 (Charlier ) was used. Confidence levels were determined based on p ‐values: < 0.05 , < 0.01 , < 0.001 , and < 0.0001 . Before analysis, FTIR spectra were thoroughly prepared to ensure data quality and reliability, as described previously (Röhr et al. ). In brief, to avoid statistical distortions caused by outlier spectra, spectra with high noise level (signal‐to‐noise‐ration SNR < 100) and high scattering were excluded from analysis. The remaining spectra were then corrected for Mie scattering using extended multiplicative signal correction (EMSC). This correction method effectively reduces the effects of Mie scattering, which can complicate spectral interpretation (Konevskikh, Lukacs, and Kohler ; Solheim et al. ). An average spectrum of a plaque core has been created by manually selecting and averaging fitting pixel‐spectra. This was done for illustration spectral differences. All plots were created with matplotlib 3.8.4 and seaborn 0.11.2.
Immunofluorescence ( IF ) Staining Anti‐Aβ IHC was performed using a multi‐label direct IF approach, staining samples with two anti‐Aβ probes with varying affinities for different Aβ polymorphs; see Figure . The goal was to distinguish Aβ fibrils from oligomeric forms without using any reagents that could alter the structure of Aβ. The samples were fixed in 4% paraformaldehyde (PFA) for 30 min and then washed with phosphate‐buffered saline (PBS). Samples were incubated with 20 μM Thioflavin T (ThT, Thermo Fisher) in 2% mouse serum (Novus Biologicals) in PBS for 45 min, washed in PBS, and then incubated overnight at 4°C with 4G8, labeled with Spark‐YG 570 (BioLegend) (1:1000) in normal antibody diluent (DAKO). Following additional washing, samples were mounted with EverBrite TrueBlack Hardset mounting medium (Biotium) and cover‐slipped. The stained sample was then imaged with an Olympus VS120 slide scanner and the UPlanSApo 20× 0.75 NA (numerical aperture) objective (Olympus). The rigorous protocol visualized Aβ using fluorescent labels ThT and 4G8.
Multimodal Image Registration To link spectral data with IF images, custom software (written in Matlab 2019a) was used to determine an affine 2D transformation, as described previously (Röhr et al. ). The user‐provided reference coordinates were used to transform the IF image into the vibrational image's coordinate system. The overlay's quality was visually confirmed while accounting for tissue morphology. To identify spectra corresponding to Aβ plaques, binary masks were generated from IF image cutouts using Otsu's method, delineating plaques, and surrounding regions. The resulting plaque spectrum was calculated as the arithmetic mean of all pixel‐spectra within the plaque mask, and the surrounding spectrum was derived from the pixel‐spectra within a ring‐shaped mask that surrounded the plaque; see Figure .
QCL ‐ IR Imaging QCL‐IR was used to generate microscopic spectral images of the tissue sections intended for the label‐free detection and extraction of plaques via LMD for FIA‐MS. The Spero‐QT 340 microspectrometer (Daylight Solutions, San Diego, USA) and Chemical Vision software version 3.2 were used for this purpose. This label‐free method employs infrared spectra, which contain useful chemical information in the form of vibrational bands (Goormaghtigh et al. ; Röhr et al. ). The spectral range of the instrument is 1800 to 948 cm −1 , with a spectral resolution of 2 cm −1 . A 4× objective (0.3 NA) was used to project a 2 × 2 mm 2 FOV onto a 480 × 480 pixel microbolometer FPA detector, yielding an image with 4.25 × 4.25 μm 2 pixel size. Prior to spectral measurements, tissue samples were thawed in a dry air‐filled container before being placed in the Spero‐QT cavity. Continuous purging with dry air was used throughout the measurements.
Aβ Plaque Detection Using a Neural Network In our previous work (Müller et al. ), we detailed the label‐free detection of plaques in QCL‐IR images using a deep convolutional neural network (CNN). In summary, we used a comparative segmenting network (CompSegNet) trained for Aβ plaque detection (Schuhmacher, Schörner, and Küpper ). The network's architecture is based on an extended U‐Net model with input dimensions of 64 × 64 × 427 pixels (Ronneberger, Fischer, and Brox ). QCL‐IR images are cropped to 64 × 64 pixels and then passed through convolution and pooling layers, which gradually reduce the image dimensions to an 8 × 8 × 512 matrix. Transposed convolutions, combined with skip‐connections, are then applied, resulting in the generation of a 64 × 64 × 1 output layer called the activation map. To facilitate whole‐slide image (WSI) segmentation, a tile‐based strategy was used, with a fixed window size of 64 × 64 pixels. Overlapping tiles of 16 pixels were created to ensure information continuity across adjacent regions, and the CNN evaluated each tile individually. Finally, the outputs from each tile were reassembled using the maximum value for overlapping regions, yielding a comprehensive WSI activation map that identifies the plaques; see Figure .
Plaque Extraction via Laser Microdissection ( LMD ) LMD is used to extract plaques from brain tissue sections, guided by the WSI activation maps, as described previously (Müller et al. ). In brief, by binarizing the activation maps with a predetermined threshold of 0.9 within a range of 0 to 1, “plaque masks” were created. Morphological operations were performed sequentially, excluding objects smaller than 100 μm 2 , dilation by 15 μm, hole filling, erosion by 10 μm, and exclusion of objects smaller than 300 μm 2 , based on eccentricity and solidity criteria; see Figure . The morphological processing and coordinate transformation were done in Matlab (version 2019a), resulting in plaque‐like shapes with a 5 μm margin to account for tissue loss during subsequent procedures. This results in plaques that are approximately spherical in shape, with areas exceeding 300 μm 2 . Following manual exclusion of damaged tissue regions, the sample was laser microdissected with a LMD microscope (PALM MicroBeam; Zeiss, Jena, Germany). A two‐dimensional Helmert transformation, which used reference points in each microscopy image, facilitated the plaque shape coordinates to the microscopy coordinate system. The PALM Robo software (Zeiss, version 4.6) was then used to import the coordinates and perform precise plaque extraction with the instrument's 20× objective. Extracted tissue samples were collected in 50 mM SDS, briefly sonified, covered with argon gas, and stored at −80°C until FIA‐MS analysis. In general, several thousand shapes were combined to reach a total area of 50 million μm 2 per sample. This procedure was applied to plaques, their surrounding regions, as well as control tissue.
Sample Preparation for FIA ‐ MS Lipid extraction was performed according to the procedure described by Bligh and Dyer (Bligh and Dyer ). A sample amount of 50 μg wet weight was subjected to lipid extraction in the presence of non‐naturally occurring internal standards. The following lipid species were added as internal standards: CE 17:0, CE 22:0, FC[D7], Cer 18:1;O2/14:0, Cer 18:1;O2[D7]/18:0, HexCer 18:1;O2/12:0, HexCer 18:1;O2[D5]/18:0, DG 14:0/14:0, DG 20:0/20:0, FC[D7], LPC 13:0, LPC 19:0, LPE 13:0, LPE 18:1[D7], PC 14:0/14:0, PC 22:0/22:0, PE 14:0/14:0, PE 20:0/20:0 (di‐phytanoyl), PI 18:1[D7]/15:0, PS 14:0/14:0, PS 20:0/20:0 (di‐phytanoyl), SM 18:1;O2/12:0, SM 18:1;O2/18:1[D9], TG 17:0/17:0/17:0, and TG 19:0/19:0/19:0. After extraction, a volume of 800 μL of the chloroform phase (total volume of 1 mL) was recovered by a pipetting robot and vacuum dried. The dried extracts were dissolved in 500 μL chloroform/methanol/2‐propanol (1:2:4 v/v/v) with 7.5 mM ammonium formate.
FIA ‐ MS Lipidomics The analysis of lipids was performed by direct FIA using a triple quadrupole mass spectrometer (FIA‐MS/MS) and a high‐resolution hybrid quadrupole‐Orbitrap mass spectrometer (FIA‐FTMS). FIA‐MS/MS was performed in positive ion mode using the analytical setup and strategy described previously (Liebisch et al. ). A Waters Acquity UPLC (Milford, Massachusetts, USA) delivered methanol/chloroform = 3/1 (v/v) with 7.5 mM ammonium acetate at an initial flow rate of 50 μL/min for 0.2 min, followed by 10 μL/min for 5.7 min and a wash at 500 μL/min for 1.9 min. The LC flow was coupled to a Waters Xevo TQ‐S micro equipped with an electrospray ionization source operated in positive mode. A fragment ion of m/z 184 was used for lysophosphatidylcholines (LPC) (Liebisch et al. ). The following neutral losses were applied: PE and lysophosphatidylethanolamine (LPE) 141, phosphatidylserine (PS) 185, and phosphatidylinositol (PI) 277 (Matyash et al. ). Sphingosine‐based Cer and hexosylceramides (HexCer) were analyzed using a fragment ion of m/z 264 (Liebisch et al. ). For FIA‐MS/MS glycerophospholipid species, annotation is based on the assumption of even numbered carbon chains only. A detailed description of the FIA‐FTMS method was published recently (Höring et al. , ). QExactive Orbitrap (Thermo Fisher Scientific, Bremen, Germany) equipped with a heated electrospray ionization source was used with the following settings: spray voltage of 3.5 kV, S‐lens RF level 50, capillary temperature of 250°C, aux gas heater temperature of 100°C, and settings of 15 for sheath gas and 5 for aux gas. Chloroform/methanol/2‐propanol (1:2:4 v/v/v) was delivered at an initial flow rate of 100 μL/min until 0.25 min followed by 10 μL/min for 2.5 min and a wash out with 300 μL/min for 0.5 min. PC, PC ether (PC O) and sphingomyelins (SM) were analyzed in negative ion mode m/z 520–960 as [M+HCOO]‐ at a target resolution of 140 000 (at m/z 200). The quantification was performed by multiplication of the spiked IS amount with analyte‐to‐IS ratio. Lipid species were annotated according to the latest proposal for shorthand notation of lipid structures that are derived from mass spectrometry (Liebisch et al. ).
Lipidomics Data Analysis Lipidomic analysis by mass spectrometry yields molar concentrations of lipid species for each sample. Here, each sample was normalized for the total molar lipid content to compare their lipid composition. Lysophospholipids were excluded from analysis because they contain only one FA and can therefore not be compared with the other lipid species investigated here. For the principal component analysis (PCA), standardized data with a mean of 0 and a standard deviation of 1 were utilized to ensure equal weighting of all features. The PCA was performed based on the correlation matrix, employing Singular Value Decomposition (SVD). All calculations were done in python 3.10, using numpy 1.26.3, pandas 1.5.3, and scikit‐learn 1.4.2. Significance levels were calculated using Welch's t‐test and visualized using Statannotations 0.6 (Charlier ). The confidence levels were determined according to p ‐values < 0.05 , < 0.01 , < 0.001 , and < 0.0001 . All plots were created with matplotlib 3.8.4 and seaborn 0.11.2.
Results 3.1 Lipids Are Less Unsaturated in Plaques Post‐mortem brain tissue sections from the TL of eight AD cases and eight HC cases were analyzed in this study. FTIR was performed on the selected gray matter regions within the tissue sections, followed by IF staining against Aβ. Figure illustrates the decrease in lipid unsaturation within Aβ plaques, based upon the ratio between alkenes and ester groups. Figure displays an exemplary brain tissue section from the TL of an AD case. IF against Aβ visualizes the distribution and microscopic details of plaques in the tissue section. By combining FTIR with subsequent IF, plaques were localized in FTIR images with micrometer precision, as described previously (Röhr et al. ). This facilitated the extraction and subsequent analysis of FTIR spectra from plaque core, corona, and its surrounding, as shown in Figure . The reduced =C‐H stretching band of alkenes at 3012 cm −1 indicates a decrease of alkenes in plaques. Figure displays that the ratio of the alkene band and the ester band at 1738 cm −1 is significantly decreased in plaques, compared to their surrounding tissue and gray matter of HC cases. This indicates lower levels of lipid unsaturation in plaques. To further elucidate the specific lipidomic alterations associated with Aβ plaques, FIA‐MS was employed for in‐depth lipid profiling. Plaques were identified and localized in unstained tissue sections using fast, label‐free QCL‐IR imaging, circumventing potential artifacts introduced by staining methods. This non‐invasive approach, coupled with machine learning‐based plaque detection, enabled the precise microdissection of plaques and their surrounding tissue from tissue sections via LMD (Müller et al. ). This was done with brain tissue from all eight AD cases and gray matter from all HC cases. Figure shows the detailed lipid unsaturation profiles from FIA‐MS analysis. Figure illustrates the composition of lipid unsaturation in all FIA‐MS samples ( n = 24). Saturated lipids that contain only SFAs with no double bonds make up 11% ± 2% of the total lipid content. The residual lipids contain between one and seven double bonds. The most abundant species are monosaturated lipids that make up 36% ± 4%, whereas lipids with three double bonds are least abundant and make up 1.3% ± 0.4% of the lipid composition; see Figure . The differences in lipid unsaturation in plaques, their surroundings, and gray matter from HC cases are most significant in saturated lipids. Consequently, the percentage of UFAs also differ significantly between the tissue groups, as shown in Figure . HC tissue displays a broad range of UFA contents across the HC cases ( n = 8), whereas plaques and their surrounding tissue from the AD cases ( n = 8) feature tighter distributions. Presumably, this is the case because the extraction of plaques and their surroundings was guided by QCL‐IR imaging, whereas the HC tissue was collected from across the entire gray matter area. Plaques display the lowest UFA content (88% ± 1%), significantly lower than their surroundings (91% ± 1%). Figure further illustrates the differences between plaques and their surroundings. Saturated lipids are significantly increased in plaques by 1.6% ± 0.5%. Unsaturated lipids with between one and four double bonds are decreased in plaques. Lipids with two double bonds display the strongest decrease by −1.8% ± 0.9%. Lipids with more than five double bonds remain broadly unchanged in plaques. 3.2 Fatty Acids Are Shorter in Plaques The proposed cleavage of UFAs, described above, results in a length reduction of FAs. Figure presents the FA length distribution differences between plaques, their surroundings, and HC tissue. Figure displays the sum of acyl chain distribution in all our FIA‐MS samples ( n = 24) given as the sum number of C atoms contained in both FAs in a lipid. The shortest detected FA pair yields a sum of 30 C atoms and is most likely a pair of a myristic acid (14:0) and a palmitic (16:0) or palmitoleic acid (16:1) and constitutes 0.85% ± 0.17% of the lipid composition; see Figure . The longest FA pair we detected contains 44 C atoms and is most likely a combination of two C22 (22:x) FAs, because longer FAs are usually not detected in gray matter (Nakada, Kwee, and Ellis ; Söderberg et al. ; Fraser, Tayler, and Love ). The FA pairs with between 32 and 42 C atoms can contain various FA combinations and are correspondingly more abundant. The FA length distributions in plaques, their surroundings, and HC tissue display significant differences in the short FA pairs between 30 and 34 C atoms. Correspondingly, the average total FA length differs between the tissue groups, as shown in Figure . HC tissue shows a broader range of FA length, compared to plaques and their surrounding tissue. This is similar to the lipid unsaturation shown above and presumably also due to the tissue collection procedure. The average FA length in plaques (36.76 ± 0.11 C) is significantly lower than in their surroundings (37.02 ± 0.07 C). Figure details the differences in FA lengths between plaques and their surroundings. Short FA pairs (30–34 C) are significantly increased in plaques. FA pairs with 32 C atoms see the strongest increase by 1.9% ± 0.6%. All long FA pairs (36–44 C) are decreased in plaques, except C40 with a deviation much larger than the difference of 0.5% ± 1.8%. 3.3 PC Dominates the Lipidome Changes in Plaques An analysis of the lipid species composition reveals that PC contributes strongest to the lipidome changes in plaques. Figure displays the top nine contributors, five of which are PC species, three are PE species, and one is a Cer species. The strongest contributor is PC 32:0 that contributes 1.8% ± 0.6% more to the lipidome of plaques than to their surroundings. Figures and analyze PC separately and show that the reduction in lipid unsaturation and FA length is prominent in PC. A PCA reveals that the plaque lipidome differs systematically from the surrounding tissue in all AD cases ( n = 8). This can be seen by the respective cluster (blue = plaque, gray = surrounding) in Figure . Figure displays the respective component loadings. The cluster are linearly separated by a separation line, determined by a support vector machine (SVM). Figure presents a correlation matrix of the short and saturated PC species 30:0, 32:0, and 34:0 with all PC species in plaques. It is apparent that the short and saturated PC species broadly correlate with other short and (mono)saturated PC species (red hues, left side), whereas they anti‐correlate with long and unsaturated PC species (blue hues, right side). Correlations of all lipid species are shown in Figure .
Lipids Are Less Unsaturated in Plaques Post‐mortem brain tissue sections from the TL of eight AD cases and eight HC cases were analyzed in this study. FTIR was performed on the selected gray matter regions within the tissue sections, followed by IF staining against Aβ. Figure illustrates the decrease in lipid unsaturation within Aβ plaques, based upon the ratio between alkenes and ester groups. Figure displays an exemplary brain tissue section from the TL of an AD case. IF against Aβ visualizes the distribution and microscopic details of plaques in the tissue section. By combining FTIR with subsequent IF, plaques were localized in FTIR images with micrometer precision, as described previously (Röhr et al. ). This facilitated the extraction and subsequent analysis of FTIR spectra from plaque core, corona, and its surrounding, as shown in Figure . The reduced =C‐H stretching band of alkenes at 3012 cm −1 indicates a decrease of alkenes in plaques. Figure displays that the ratio of the alkene band and the ester band at 1738 cm −1 is significantly decreased in plaques, compared to their surrounding tissue and gray matter of HC cases. This indicates lower levels of lipid unsaturation in plaques. To further elucidate the specific lipidomic alterations associated with Aβ plaques, FIA‐MS was employed for in‐depth lipid profiling. Plaques were identified and localized in unstained tissue sections using fast, label‐free QCL‐IR imaging, circumventing potential artifacts introduced by staining methods. This non‐invasive approach, coupled with machine learning‐based plaque detection, enabled the precise microdissection of plaques and their surrounding tissue from tissue sections via LMD (Müller et al. ). This was done with brain tissue from all eight AD cases and gray matter from all HC cases. Figure shows the detailed lipid unsaturation profiles from FIA‐MS analysis. Figure illustrates the composition of lipid unsaturation in all FIA‐MS samples ( n = 24). Saturated lipids that contain only SFAs with no double bonds make up 11% ± 2% of the total lipid content. The residual lipids contain between one and seven double bonds. The most abundant species are monosaturated lipids that make up 36% ± 4%, whereas lipids with three double bonds are least abundant and make up 1.3% ± 0.4% of the lipid composition; see Figure . The differences in lipid unsaturation in plaques, their surroundings, and gray matter from HC cases are most significant in saturated lipids. Consequently, the percentage of UFAs also differ significantly between the tissue groups, as shown in Figure . HC tissue displays a broad range of UFA contents across the HC cases ( n = 8), whereas plaques and their surrounding tissue from the AD cases ( n = 8) feature tighter distributions. Presumably, this is the case because the extraction of plaques and their surroundings was guided by QCL‐IR imaging, whereas the HC tissue was collected from across the entire gray matter area. Plaques display the lowest UFA content (88% ± 1%), significantly lower than their surroundings (91% ± 1%). Figure further illustrates the differences between plaques and their surroundings. Saturated lipids are significantly increased in plaques by 1.6% ± 0.5%. Unsaturated lipids with between one and four double bonds are decreased in plaques. Lipids with two double bonds display the strongest decrease by −1.8% ± 0.9%. Lipids with more than five double bonds remain broadly unchanged in plaques.
Fatty Acids Are Shorter in Plaques The proposed cleavage of UFAs, described above, results in a length reduction of FAs. Figure presents the FA length distribution differences between plaques, their surroundings, and HC tissue. Figure displays the sum of acyl chain distribution in all our FIA‐MS samples ( n = 24) given as the sum number of C atoms contained in both FAs in a lipid. The shortest detected FA pair yields a sum of 30 C atoms and is most likely a pair of a myristic acid (14:0) and a palmitic (16:0) or palmitoleic acid (16:1) and constitutes 0.85% ± 0.17% of the lipid composition; see Figure . The longest FA pair we detected contains 44 C atoms and is most likely a combination of two C22 (22:x) FAs, because longer FAs are usually not detected in gray matter (Nakada, Kwee, and Ellis ; Söderberg et al. ; Fraser, Tayler, and Love ). The FA pairs with between 32 and 42 C atoms can contain various FA combinations and are correspondingly more abundant. The FA length distributions in plaques, their surroundings, and HC tissue display significant differences in the short FA pairs between 30 and 34 C atoms. Correspondingly, the average total FA length differs between the tissue groups, as shown in Figure . HC tissue shows a broader range of FA length, compared to plaques and their surrounding tissue. This is similar to the lipid unsaturation shown above and presumably also due to the tissue collection procedure. The average FA length in plaques (36.76 ± 0.11 C) is significantly lower than in their surroundings (37.02 ± 0.07 C). Figure details the differences in FA lengths between plaques and their surroundings. Short FA pairs (30–34 C) are significantly increased in plaques. FA pairs with 32 C atoms see the strongest increase by 1.9% ± 0.6%. All long FA pairs (36–44 C) are decreased in plaques, except C40 with a deviation much larger than the difference of 0.5% ± 1.8%.
PC Dominates the Lipidome Changes in Plaques An analysis of the lipid species composition reveals that PC contributes strongest to the lipidome changes in plaques. Figure displays the top nine contributors, five of which are PC species, three are PE species, and one is a Cer species. The strongest contributor is PC 32:0 that contributes 1.8% ± 0.6% more to the lipidome of plaques than to their surroundings. Figures and analyze PC separately and show that the reduction in lipid unsaturation and FA length is prominent in PC. A PCA reveals that the plaque lipidome differs systematically from the surrounding tissue in all AD cases ( n = 8). This can be seen by the respective cluster (blue = plaque, gray = surrounding) in Figure . Figure displays the respective component loadings. The cluster are linearly separated by a separation line, determined by a support vector machine (SVM). Figure presents a correlation matrix of the short and saturated PC species 30:0, 32:0, and 34:0 with all PC species in plaques. It is apparent that the short and saturated PC species broadly correlate with other short and (mono)saturated PC species (red hues, left side), whereas they anti‐correlate with long and unsaturated PC species (blue hues, right side). Correlations of all lipid species are shown in Figure .
Discussion By employing FTIR, we were able to spatially resolve and quantify the unsaturation levels of lipids in Aβ plaques. Our findings indicate a significant reduction in the =C‐H stretching band at 3012 cm −1 of alkenes, a marker of unsaturated FAs, within plaques compared to surrounding tissue and HC brain regions. This approach provided a direct measurement of lipid unsaturation, circumventing the limitations associated with traditional oxidative stress markers. Many studies that have previously investigated oxidative stress relied on specific molecular markers to trace oxidative stress. However, common markers like 4‐hydroxylnonenal (HNE) and malondialdehyde (MDA) have been criticized for their lack of specificity, sensitivity, and reproducibility (Sultana, Perluigi, and Butterfield ). Further molecular details of the lipidomic alterations in plaques were elucidated using FIA‐MS, which confirmed that the fraction of unsaturated lipids is decreased in plaques. Specifically, we observed a decrease in long‐chain, unsaturated lipid species, compensated by a significant increase in short‐chain, saturated lipids in plaques. Comparing our results to previous studies, the overall depletion of PUFAs in AD brains has been documented previously (Söderberg et al. ; Corrigan et al. ; Prasad et al. ; Fraser, Tayler, and Love ). Martín et al. reported reduced levels of PUFAs and a 13% decrease in alkenes in lipid rafts extracted from the frontal brain cortex of AD cases (Martín et al. ). Our observation of decreased alkenes within plaques suggests that lipid degradation in plaques significantly contributes to the overall decrease of alkenes in AD brains. However, our study extends this understanding by providing evidence that lipid degradation is specifically localized within Aβ plaques. This localized depletion highlights the direct impact of Aβ aggregation on lipid composition, differentiating our work from studies assessing global brain lipid levels without spatial specificity. The spatially resolved lipid composition of plaques has only been investigated by a few studies so far. Panchal et al. reported increased levels of saturated Cer and cholesterol in plaques using FIA‐MS to compare plaques and surrounding tissue (Panchal et al. , ). Our FIA‐MS analysis complements these findings by demonstrating a significant decrease in long‐chain UFAs within plaques. This suggests that while certain lipid species accumulate, others are selectively depleted, likely due to oxidative processes. Kaya et al. applied matrix‐assisted laser desorption/ionization imaging mass spectrometry (MALDI‐IMS) to Aβ plaques in transgenic mice, which offers high‐resolution insights into the distribution of specific lipid species within complex tissues, providing a comprehensive understanding of the lipid environment surrounding plaques. Therefore, depletion of UFA‐rich phospholipids was observed (Kaya et al. ). This observation corroborates our findings, as we also noted a significant reduction in UFA‐rich phospholipids within plaques. CARS imaging provides exceptionally high spatial resolution, enabling Kiskis et al. to investigate the microstructure of lipid changes in plaques. However, it is constrained by the limited information content of Raman bands. Our study specifically identified PC as a major contributor to this decrease, with a shift toward short‐chain, saturated PC species. Benseny‐Cases et al. used FTIR and reported the decrease of the 3012 cm −1 band in plaques of transgenic mice, indicating reduced lipid unsaturation (Benseny‐Cases et al. ). Surowka et al. similarly observed increased lipid oxidation in amyloid deposits in transgenic mouse brains (Surowka et al. ). This is supported by observations from Head et al., who found oxidized Aβ in 98% of plaque cores, indicating that oxidative stress is a central feature of plaque pathology (Head et al. ). These findings align well with our FTIR results on human brain samples, which also indicated a marked decrease in lipid unsaturation within Aβ plaques. Despite significant findings, our study has several limitations. First, FTIR and FIA‐MS for lipid analysis have inherent constraints: FTIR spectral data can be affected by overlapping signals, and FIA‐MS, while sensitive, requires meticulous sample preparation and can be influenced by ion suppression and matrix effects. Additionally, our focus was on a limited number of lipid species, excluding glycolipids and minor lipids that could offer deeper insights into lipid alterations in AD (Blank and Hopf ). We measured total lipid mass, which leaves room for interpretation. However, our observed percentage of the shortest FA pair (30:x, made up of 14:0 and 16:x) at 0.85% ± 0.17% aligns well with literature values of approximately 0.8% ± 0.2% in human TL, validating our findings (Fraser, Tayler, and Love ). The observed lipid depletion, though statistically significant, had a small effect size, indicating subtle changes requiring highly sensitive detection methods. The spatial resolution of QCL‐IR imaging also poses a limitation. Although LMD allowed isolation and analysis of Aβ plaques, heterogeneity within plaques and their microenvironment could influence lipid composition and oxidative stress levels, potentially accounting for discrepancies with other studies (Gaudin et al. ). Plaque, surrounding, and control regions for FTIR analyses were selected based on IHC, while the areas chosen for LMD were unstained, given the label‐free nature of this approach. Control gray matter regions were manually selected based on QCL‐IR measurements, potentially accounting for the broader distribution observed within this subcohort. In contrast, plaque and surrounding areas in the QCL‐IR data were selected using machine learning. As demonstrated in our most recent publication, this label‐free detection of plaques via machine learning shows a high degree of concordance with established IHC methods (Müller et al. ). Nevertheless, certain discrepancies persist between the two techniques, which may have contributed to the variations discussed above. Additionally, using post‐mortem brain tissue introduces variability due to factors such as post‐mortem interval, preservation methods, and disease progression, potentially affecting in vivo accuracy. Our cross‐sectional study did not address temporal lipid changes throughout disease stages. Furthermore, other mechanisms such as mitochondrial dysfunction and neuroinflammation may also contribute to oxidative stress in the AD brain, adding further complexity to the observed lipid alterations. We did not investigate the functional causes and consequences of lipid depletion within Aβ plaques. While we observed a reduction in UFAs, its origin and its impact on neuronal function, membrane integrity, and synaptic activity remains unclear. Functional studies using cell culture or animal models could link our findings to specific pathological mechanisms and identify therapeutic targets. To summarize, the relationship between Aβ pathology and overall lipid degradation has been well‐documented (Söderberg et al. ; Reiter ; Martín et al. ). Our study adds to this body of evidence by demonstrating that lipid degradation is not only a generalized phenomenon in AD brains but also a localized event within Aβ plaques. Our study has provided significant insights into the lipidomic landscape of Aβ plaques in AD, specifically highlighting the depletion of UFAs. By employing advanced techniques such as IR imaging and mass spectrometry, we were able to directly investigate the lipid composition within plaques, bypassing the limitations associated with traditional oxidative stress markers. In conclusion, our study highlights the significance of lipid degradation within Aβ plaques, providing a more detailed understanding of the molecular changes occurring in AD. The localized depletion of UFAs within plaques underscores the critical role of lipid alterations in the pathogenesis of AD. Further research should include longitudinal studies and in vivo imaging techniques. This will be crucial for understanding the dynamic changes in lipid composition throughout AD progression and exploring potential therapeutic avenues targeting lipid metabolism.
Dominik Röhr: conceptualization, methodology, investigation, formal analysis, software, visualization, writing – original draft, writing – review and editing. Melina Helfrich: investigation, formal analysis, visualization, writing – review and editing. Marcus Höring: investigation, writing – review and editing, methodology. Frederik Großerüschkamp: project administration, methodology, writing – review and editing. Gerhard Liebisch: project administration, methodology, writing – review and editing. Klaus Gerwert: supervision, funding acquisition, writing – review and editing.
The Institutional Review Board and Medical Ethical Board from the Vrije University Medical Center Amsterdam approved the procedures of the NBB. This project is registered at the NBB under project number 1430 and carries the title: “From plaque to CSF – and back again? An (un)expected journey of Amyloid‐β”.
The brain samples were obtained from the Netherlands Brain Bank (NBB), Netherlands Institute for Neuroscience, Amsterdam. All materials have been collected from donors for or from whom a written informed consent for brain autopsy and the use of the material and clinical information for research purposes had been obtained by NBB.
The authors declare no conflicts of interest.
Data S1.
|
Infección por | 67a57fdb-eb63-4b7c-97c3-77e49548622b | 6166253 | Pathology[mh] | |
Epithelioid Hemangioendothelioma as a Dangerous, Easy to Miss, and Nearly Impossible to Clinically Diagnose Condition: Case Report | d2b2408d-495d-4b00-ac5d-75553e3be91c | 11391149 | Anatomy[mh] | Epithelioid hemangioendothelioma (EHE) is an extremely rare cancer, accounting for less than 1% of all vascular tumors, demonstrating features between those of hemangioma and angiosarcoma . Although first described by Dail and Liebow in 1975, the term EHE was only first proposed by Weiss and Enzinger in 1982 . These tumors can occur at any age, with 38 years being the median age at diagnosis . The most common presenting symptom is pain, along with less commonly reported symptoms such as cough, palpable mass, or fatigue. Nearly one-third of patients with EHE are asymptomatic and tumors are discovered incidentally . While likely endothelial in origin, EHE is extremely heterogeneous in presentation and prognosis, complicating diagnosis and clinical decisions . EHE can occur nearly anywhere in the body. Primary cutaneous EHE is rare and should prompt suspicion of metastatic disease, especially if multifocal in the skin . Owing to their rarity and similarities to other diagnoses, cutaneous EHE lesions are commonly misdiagnosed . Previous studies suggest the diagnosis of strictly cutaneous EHE incurs a 17% mortality rate at 3 years, highlighting its relatively aggressive nature . It is paramount for dermatologists and dermatopathologists to be aware of EHE and its defining characteristics to minimize the risk of missing this crucial diagnosis. We report a case of two periauricular lesions with dermal proliferation consistent with EHE, leading to the discovery of underlying metastatic EHE with pulmonary and hepatic involvement in a 24-year-old woman. The aim of presenting this case is to enhance understanding of EHE, an uncommon cancer that is not well studied. A 24-year-old woman presented to our dermatology clinic with a left posterior auricular papule and left preauricular papule present for 8 and 4 months, respectively . The patient had no significant medical or social history, including no tobacco or heavy alcohol use. Both lesions were painful and progressively enlarging. The patient denied any other symptoms. Shave biopsy was taken of both lesions. The histology of both lesions demonstrated cellular dermal proliferations of epithelioid cells with the eosinophilic cytoplasm arranged in cords within a myxohyaline stroma . Subtle vacuoles containing red blood cells were present within some of the cells . Histological and immunohistochemical findings were consistent with the diagnosis of EHE . Due to multifocal cutaneous disease, there was high clinical suspicion of metastatic disease. Our patient was referred to medical and surgical oncology for further evaluation, and computed tomography (CT) scans of the head, neck, chest, abdomen, and pelvis were performed. Innumerable bilateral pulmonary nodules, a 1.8-cm hypoattenuated hepatic lesion, and prominent bilateral axillary lymph nodes were noted, all consistent with metastatic disease. After seeking multiple opinions from oncology, our patient elected the watchful waiting approach. Serial CT scans every 3 months were recommended to monitor disease progression. The patient provided consent to publish information regarding her case, including photographs and relevant findings. Identifiable patient information has been appropriately masked or omitted to comply with ethical standards and patient privacy. Prior Reports of EHE Literature pertaining to EHE is limited with case reports and case series comprising the majority. This can largely be attributed to the low prevalence of EHE, reported as approximately 1 in 1 million . Sites of primary and metastatic involvement in EHE most commonly involve the liver, lung, and bone; however, the disease has been reported in nearly every part of the body. When cutaneous EHE is discovered, it typically represents metastatic disease rather than primary malignancy. The appearance, location, and characteristics of cutaneous EHE vary immensely from case to case, with no clear consensus available . The extreme heterogeneity of this disease complicates detection and diagnosis . Histopathology and immunohistochemistry are often crucial for diagnosis of cutaneous disease. Histologically, tumors typically show circumcised nodules with an overlying acanthotic epidermis. A mixture of pleomorphic spindle and epithelioid cells with sharply eosinophilic cytoplasm will be present, typically embedded in a myxoid or hyaline matrix . Cells typically stain positive for CD31, CD34, factor VIII-related antigen, α-smooth muscle actin, and cytokeratin . When unable to be clearly differentiated from other vascular tumors, the presence of the WWTR1-CAMTA1 translocation can aid the diagnosis of EHE . This translocation dysregulates the Hippo pathway, promoting cancer proliferation and survival . The prognosis of strictly cutaneous EHE is not readily available. In a small case series of 30 patients with cutaneous EHE, at 36 months follow-up, 21% of cases had metastatic disease, 13% had local recurrence, and 17% had died from the disease . In all cases of EHE irrespective of site, 1-year overall survival is 90% with a 5-year overall survival of 73% . Given the low prevalence of EHE, no randomized clinical trials exist regarding the optimal treatment strategy . Patients with cutaneous EHE should receive additional imaging to evaluate for metastatic disease. When no metastatic disease is found, the treatment is surgical resection . A variety of treatments such as cytotoxic chemotherapy, immunotherapy, targeted therapies, and organ transplantation have been used for metastatic disease . With reports of spontaneous disease regression , watchful waiting can also be proposed as a reasonable course following EHE diagnosis, especially if the nature of the disease is not yet understood or the risks of treatment outweigh benefits. Conclusion The heterogeneity of EHE is also demonstrated in its variable course; EHE can be unpredictable, at times being indolent and at other times very aggressive . Given the uncertain course of the disease, joint decision-making between the patient and physician is necessary. Active surveillance includes monitoring progression, and the decision to treat with radiation or surgery often follows once the nature of the tumor is better understood . Systemic treatments have been recorded, but not enough data are currently available to determine a standard approach . Regardless of the course of management, close follow-up for local recurrence and metastatic disease is essential. Future studies should focus on early detection and a standardized approach for the treatment EHE. Literature pertaining to EHE is limited with case reports and case series comprising the majority. This can largely be attributed to the low prevalence of EHE, reported as approximately 1 in 1 million . Sites of primary and metastatic involvement in EHE most commonly involve the liver, lung, and bone; however, the disease has been reported in nearly every part of the body. When cutaneous EHE is discovered, it typically represents metastatic disease rather than primary malignancy. The appearance, location, and characteristics of cutaneous EHE vary immensely from case to case, with no clear consensus available . The extreme heterogeneity of this disease complicates detection and diagnosis . Histopathology and immunohistochemistry are often crucial for diagnosis of cutaneous disease. Histologically, tumors typically show circumcised nodules with an overlying acanthotic epidermis. A mixture of pleomorphic spindle and epithelioid cells with sharply eosinophilic cytoplasm will be present, typically embedded in a myxoid or hyaline matrix . Cells typically stain positive for CD31, CD34, factor VIII-related antigen, α-smooth muscle actin, and cytokeratin . When unable to be clearly differentiated from other vascular tumors, the presence of the WWTR1-CAMTA1 translocation can aid the diagnosis of EHE . This translocation dysregulates the Hippo pathway, promoting cancer proliferation and survival . The prognosis of strictly cutaneous EHE is not readily available. In a small case series of 30 patients with cutaneous EHE, at 36 months follow-up, 21% of cases had metastatic disease, 13% had local recurrence, and 17% had died from the disease . In all cases of EHE irrespective of site, 1-year overall survival is 90% with a 5-year overall survival of 73% . Given the low prevalence of EHE, no randomized clinical trials exist regarding the optimal treatment strategy . Patients with cutaneous EHE should receive additional imaging to evaluate for metastatic disease. When no metastatic disease is found, the treatment is surgical resection . A variety of treatments such as cytotoxic chemotherapy, immunotherapy, targeted therapies, and organ transplantation have been used for metastatic disease . With reports of spontaneous disease regression , watchful waiting can also be proposed as a reasonable course following EHE diagnosis, especially if the nature of the disease is not yet understood or the risks of treatment outweigh benefits. The heterogeneity of EHE is also demonstrated in its variable course; EHE can be unpredictable, at times being indolent and at other times very aggressive . Given the uncertain course of the disease, joint decision-making between the patient and physician is necessary. Active surveillance includes monitoring progression, and the decision to treat with radiation or surgery often follows once the nature of the tumor is better understood . Systemic treatments have been recorded, but not enough data are currently available to determine a standard approach . Regardless of the course of management, close follow-up for local recurrence and metastatic disease is essential. Future studies should focus on early detection and a standardized approach for the treatment EHE. |
Bibliometric analysis of Naunyn–Schmiedeberg’s Archives of Pharmacology (1947–1974) | 1c5c326d-1ea3-474e-a7c5-6792a7c8ae4f | 11422447 | Pharmacology[mh] | In 1873, Bernhard Naunyn, Oswald Schmiederberg, and Edwin Klebs founded the “Archiv für experimentelle Pathologie und Pharmakologie,” which has evolved into the Naunyn–Schmiedeberg’s Archives of Pharmacology. Their collaboration in Dorpat, where Naunyn and Schmiedeberg served as professors, was foundational in creating a journal that would integrate the fields of pathology and pharmacology. Schmiedeberg is particularly noted as a pioneer of experimental pharmacology, with a lasting influence that extends throughout the global pharmacological community (Starke ). Quickly after its founding, Naunyn-Schmiedeberg’s Archives of Pharmacology was recognized as one of the most important journals in its field, being the oldest pharmacological journal. This reputation has been consistently upheld (Koch-Weser and Schechter ; Starke ). However, the aftermath of the Second World War presented significant challenges, including a 2-year publication hiatus that marked a period of re-evaluation and recovery for the journal (Starke ). This era coincided with the emergence of a new international scientific community, from which Germany was initially isolated (Ahlers et al. ). A recent bibliometric analysis (Dats et al. ), which primarily examined the early twenty-first century using distinct datasets, did not address the post-war period in detail, leaving a gap in the literature. Despite Klaus Starke’s comprehensive historical overview on the first 125 years of Naunyn–Schmiedeberg’s Archives of Pharmacology (Starke ), there is a lack of focused analysis on Naunyn–Schmiedeberg’s Archives of Pharmacology’s evolution in the post-World War II period. This era, critical for the journal’s shift towards internationalization and English publication, significantly boosted its citations and global presence. By providing an overview of the journal’s development from the post-war period up to 1974, this paper aims to contribute to this existing gap. To offer insights into Germany’s efforts to reassert its scientific contributions on the global stage, this paper places the bibliometric findings within a historical, political, and scientific context. Extraction process for publication data The bibliometric analysis of Naunyn–Schmiedeberg’s Archives of Pharmacology was conducted using Python and Beautiful Soup, focusing on extracting publication details from the official SpringerLink website ( https://link.springer.com/journal/210/volumes-and-issues ; Python Software Foundation ; Richardson ; Springer Link ). This methodological choice was driven by the need for a comprehensive and automated approach to data collection over a significant historical span, specifically targeting volumes 204 (1947) to 286 (1974) and encompassing a range of publication metrics from a total of 4839 publications identified across 224 cities and 44 countries (Fig. ). The previously used data extraction using Excel yielded substantial data gaps for the period from 1947 to 1971 (Dats et al. ) which were now closed with the comprehensive Phython and Beautiful Soup approach. Identification of publication metrics The extracted data included various publication metrics such as the SN type (original papers, abstracts, short papers, announcements, discussions, editorial notes, erratum, opening speeches, main topics, keynote lectures, short communications, symposiums, DGPT spring meetings, demonstrations), titles, author names (first, second, and last), affiliations, DOI numbers, issue dates and years, volume numbers, and citation counts. Citation numbers were obtained via CrossRef ( https://www.crossref.org/about/ ) which is integrated with the SpringerLink page, ensuring the accuracy of citation data. For this analysis, the focus was narrowed to “Original Paper” SN type, resulting in a dataset of 3244 publications (Fig. ). The Orignal Papers were then analyzed over the years (Fig. ). These data were then organized into a table using the Pandas library in Python and saved in an Excel (.xlsx) format. To visualize the findings, charts and tables were generated from the Excel spreadsheet. Data structuring and accuracy assurance To validate the accuracy of the extracted data, a Python unit test was conducted. This test compared the data in the Excel table against the original information on the SpringerLink page, ensuring the reliability of the data collection process. This rigorous methodological approach not only underscores the thoroughness of the analysis but also guarantees the credibility of the findings derived from the bibliometric study. Language trends in publications Leveraging Python’s langdetect module, the language of each publication (Original Papers) was discerned from the titles once the data had been structured into an Excel spreadsheet. This enabled a longitudinal analysis of the languages used in the publications spanning from 1947 to 1974. It was observed that articles were published in French and Italian on a few occasions—six times and once, respectively. However, to ensure clarity and focus on the visual representation of the data, these instances were omitted from the graphical analysis presented in Fig. . Citation analysis For citation analysis, the number of citations for each publication (Original Papers) was retrieved through CrossRef, utilizing Python and Beautiful Soup, as of January 2, 2024. CrossRef, which is updated daily and directly linked to the SpringerLink page, provided a comprehensive view of the citation patterns across the studied period (1947–1974). This dataset facilitated a granular examination of citation behavior over the years, allowing for an in-depth understanding of the factors influencing citation frequency, such as publication number and citation quotient (Fig. , Tab. ). Analysis of the 100 most-cited articles Further analysis was conducted on the 100 most-cited articles (Original Papers) within the specified timeframe, offering insights into the reasons behind their frequent citation. This examination included the years of publication (Fig. ), impact of the publication language (Fig. ), and the thematic focus (Fig. ), on their citation rates. Additionally, the geographic origins of these highly cited articles were analyzed (Fig. ), with the findings summarized in a detailed table (Table ). Topics For simplicity, the entire dataset, excluding publication titles, was translated into English. The titles themselves are crucial for identifying the thematic trends and author focus within the period analyzed. To systematically categorize these themes, the study utilized the textbook “Basiswissen Pharmakologie” as a framework for grouping (Seifert ). The textbook’s main topics provided a structured basis, with the addition of “Purinergic system,” “Substance P,” and “Toxicology” to encompass frequently occurring themes not originally listed. This thematic grouping facilitated trend analyses and the exploration of topic-related dynamics over time (Fig. S2). Furthermore, a pie chart analysis (Fig. ) highlights the distribution of topics within the 100 most cited articles, offering insights into the most influential research areas. Authors To dissect the authorship patterns within Naunyn–Schmiedeberg’s Archives of Pharmacology from 1947 to 1974, a comprehensive analysis was conducted not only on the first authors but also on the second and last authors of the papers. This methodical examination uncovered distinct trends in publication frequency among authors, culminating in the identification of the top 15 contributors to the journal during this period (Fig. ). An in-depth analysis of the top five authors was subsequently carried out, focusing on their affiliations, thematic interests, and rates of publication (see Tab. S6). Geographical analysis of publication origins Additionally, the geographical origins of the publications (Original Papers) were analyzed. City and country information extracted from the SpringerLink page was anglicized for consistency. A Python script facilitated the classification of countries into their respective continents, enabling a focused analysis of the geographical distribution of the research contributions. This analysis was narrowed down to the top five contributing countries, which together accounted for 97% of the journal’s publications between 1947 and 1974. The cities leading in publication volume were highlighted in a bar chart, providing a visual representation of the geographical trends in research output (Figs. S3, S4 and S5). Furthermore, to offer a granular view of the contributions within Germany, a heat map was created using Plotly in Python. This visual tool effectively showcased the density of publications across German cities, offering insightful perspectives on the regional distribution of research within the country. This layered approach, combining authorship trends with geographical analysis, provides a nuanced understanding of the journal’s contributions to the pharmacologcial community over the specified period. The bibliometric analysis of Naunyn–Schmiedeberg’s Archives of Pharmacology was conducted using Python and Beautiful Soup, focusing on extracting publication details from the official SpringerLink website ( https://link.springer.com/journal/210/volumes-and-issues ; Python Software Foundation ; Richardson ; Springer Link ). This methodological choice was driven by the need for a comprehensive and automated approach to data collection over a significant historical span, specifically targeting volumes 204 (1947) to 286 (1974) and encompassing a range of publication metrics from a total of 4839 publications identified across 224 cities and 44 countries (Fig. ). The previously used data extraction using Excel yielded substantial data gaps for the period from 1947 to 1971 (Dats et al. ) which were now closed with the comprehensive Phython and Beautiful Soup approach. The extracted data included various publication metrics such as the SN type (original papers, abstracts, short papers, announcements, discussions, editorial notes, erratum, opening speeches, main topics, keynote lectures, short communications, symposiums, DGPT spring meetings, demonstrations), titles, author names (first, second, and last), affiliations, DOI numbers, issue dates and years, volume numbers, and citation counts. Citation numbers were obtained via CrossRef ( https://www.crossref.org/about/ ) which is integrated with the SpringerLink page, ensuring the accuracy of citation data. For this analysis, the focus was narrowed to “Original Paper” SN type, resulting in a dataset of 3244 publications (Fig. ). The Orignal Papers were then analyzed over the years (Fig. ). These data were then organized into a table using the Pandas library in Python and saved in an Excel (.xlsx) format. To visualize the findings, charts and tables were generated from the Excel spreadsheet. To validate the accuracy of the extracted data, a Python unit test was conducted. This test compared the data in the Excel table against the original information on the SpringerLink page, ensuring the reliability of the data collection process. This rigorous methodological approach not only underscores the thoroughness of the analysis but also guarantees the credibility of the findings derived from the bibliometric study. Leveraging Python’s langdetect module, the language of each publication (Original Papers) was discerned from the titles once the data had been structured into an Excel spreadsheet. This enabled a longitudinal analysis of the languages used in the publications spanning from 1947 to 1974. It was observed that articles were published in French and Italian on a few occasions—six times and once, respectively. However, to ensure clarity and focus on the visual representation of the data, these instances were omitted from the graphical analysis presented in Fig. . For citation analysis, the number of citations for each publication (Original Papers) was retrieved through CrossRef, utilizing Python and Beautiful Soup, as of January 2, 2024. CrossRef, which is updated daily and directly linked to the SpringerLink page, provided a comprehensive view of the citation patterns across the studied period (1947–1974). This dataset facilitated a granular examination of citation behavior over the years, allowing for an in-depth understanding of the factors influencing citation frequency, such as publication number and citation quotient (Fig. , Tab. ). Further analysis was conducted on the 100 most-cited articles (Original Papers) within the specified timeframe, offering insights into the reasons behind their frequent citation. This examination included the years of publication (Fig. ), impact of the publication language (Fig. ), and the thematic focus (Fig. ), on their citation rates. Additionally, the geographic origins of these highly cited articles were analyzed (Fig. ), with the findings summarized in a detailed table (Table ). For simplicity, the entire dataset, excluding publication titles, was translated into English. The titles themselves are crucial for identifying the thematic trends and author focus within the period analyzed. To systematically categorize these themes, the study utilized the textbook “Basiswissen Pharmakologie” as a framework for grouping (Seifert ). The textbook’s main topics provided a structured basis, with the addition of “Purinergic system,” “Substance P,” and “Toxicology” to encompass frequently occurring themes not originally listed. This thematic grouping facilitated trend analyses and the exploration of topic-related dynamics over time (Fig. S2). Furthermore, a pie chart analysis (Fig. ) highlights the distribution of topics within the 100 most cited articles, offering insights into the most influential research areas. To dissect the authorship patterns within Naunyn–Schmiedeberg’s Archives of Pharmacology from 1947 to 1974, a comprehensive analysis was conducted not only on the first authors but also on the second and last authors of the papers. This methodical examination uncovered distinct trends in publication frequency among authors, culminating in the identification of the top 15 contributors to the journal during this period (Fig. ). An in-depth analysis of the top five authors was subsequently carried out, focusing on their affiliations, thematic interests, and rates of publication (see Tab. S6). Additionally, the geographical origins of the publications (Original Papers) were analyzed. City and country information extracted from the SpringerLink page was anglicized for consistency. A Python script facilitated the classification of countries into their respective continents, enabling a focused analysis of the geographical distribution of the research contributions. This analysis was narrowed down to the top five contributing countries, which together accounted for 97% of the journal’s publications between 1947 and 1974. The cities leading in publication volume were highlighted in a bar chart, providing a visual representation of the geographical trends in research output (Figs. S3, S4 and S5). Furthermore, to offer a granular view of the contributions within Germany, a heat map was created using Plotly in Python. This visual tool effectively showcased the density of publications across German cities, offering insightful perspectives on the regional distribution of research within the country. This layered approach, combining authorship trends with geographical analysis, provides a nuanced understanding of the journal’s contributions to the pharmacologcial community over the specified period. Publication activity Following World War II, Naunyn–Schmiedeberg’s Archives of Pharmacology faced a publication halt for 2 years, resuming in February 1947 with its first post-war volume (204), delayed due to US intelligence control oversight (Starke ; Herken ). Initially, the journal’s productivity was low, publishing around 70 articles each in 1947 and 1948, a direct consequence of war-induced destruction of academic facilities and the emigration of scientists during the Nazi regime, hindering immediate post-war scientific activities (Philippu ; Löffelholz ; Weise-Pötschke ; Heinsohn and Nicolaysen ; Dats et al. ; Hattori and Seifert ; Fig. ). By 1949, publication numbers surged to 122, reflecting the release of wartime research (Dats et al. ). This growth continued, notably doubling around 1951 and 1953, partly due to celebrating pharmacologist Otto Loewi’s 80th birthday (Starke ). Despite a peak in 1964 honoring Otto Krayer’s 65th birthday, the journal experienced a gradual decline in the number of publications until 1970, influenced by the preference for publishing in higher-impact English-language journals (Zehetbauer et al. ; Dats et al. ; Gzoyan et al. ). The shift towards English, essential for international scientific communication, led to the marginalization of German-language journals (Gzoyan et al. ). Recognizing the need for internationalization, the journal’s editors initiated significant changes in the late 1960s, including the internationalization of the editorial board and the transition to publishing in English, culminating in the journal’s name change in 1971 and the mandate for English publications from 1973 onwards. These measures revitalized the journal, as evidenced by a 149% increase in publications from 1970 to 1973, stabilizing its contribution to the international research community thereafter (Starke ; Dats et al. ; Hattori and Seifert ; Fig. ). Language Naunyn–Schmiedeberg’s Archives of Pharmacology predominantly featured German-language publications until the late 1960s, reflecting its national orientation (Fig. ). This linguistic homogeneity meant its research largely remained within German-speaking circles, limiting international engagement and recognition (Bajerski ). Analysis of publication trends (Fig. ) against the language of publication (Fig. ) reveals a parallel between the overall publication volume and German-language articles up to 1970. Notably, spikes in English publications in 1953 and 1964 corresponded with contributions from the USA, including works dedicated to emigrant pharmacologists Otto Loewi and Otto Krayer, highlighting brief periods of international collaboration (Starke ). Loewi, a former student of Oswald Schmiedeberg, fled Nazi Germany in 1938, eventually settling in the USA at New York University (Bettendorf ; Philippu ; McCoy and Tan ). Similarly, Krayer, who refused a position vacated under Nazi policies, led Harvard University’s pharmacology institute from 1939 to 1966 (Starke ; Philippu ; Rubin ). The contributions of Loewi and Krayer underscore the journal’s intermittent international reach. The sharp increase in English publications from 1970 onwards (Fig. ) reflects editorial efforts to promote English, transitioning from five English articles in 1967 to 164 by 1973, with a corresponding decline in German articles. This shift was instrumental in repositioning the journal within the global scientific community (Starke ; Francisco ; Hattori and Seifert ). Post-WWII, Germany’s diminished academic stature and UNESCO’s advocacy for multilingualism in scientific publishing prompted questions about delayed internationalization efforts. The reluctance of the journal’s publisher, Dr. Springer, to adopt multilingualism until the late 1960s contributed to this lag (DGPT archive in Göttingen 01.09.1949). The ascendancy of English as the primary scientific language post-1970s highlights a broader shift towards internationalization in German science, necessitated by a globalizing research landscape (Winkmann et al. ; Heinsohn and Nicolaysen ). The debate over the late adoption of English in German journals touches on the broader discourse around multilingualism’s role in ensuring high-quality, globally communicable science (Tardy ; Billings 2015; Davydova ). The imposition of English as the lingua franca raises ethical questions about linguistic equity and the potential marginalization of non-English speakers (Phillipson ; O’Neil ). Moreover, the dominance of English reflects colonial legacies, risking the devaluation of other languages and the sidelining of non-English scientific contributions (Phillipson ; Ahn et al. ). The international neglect of non-English journals, resulting in fewer citations and recognition, underscores the challenges of linguistic isolation in the global scientific community (Phillipson ; Gzoyan et al. ). Citations While high publication volumes post-1970s indicate increased scientific productivity, assessing the international recognition of Naunyn–Schmiedeberg’s Archives of Pharmacology and its articles requires examining citation patterns (Francisco ). Figure presents annual publication counts alongside total citations from 1947 to 1974, calculating an average citation rate per article. This metric reveals a gradual increase in citation impact, with a notable surge from 1963 onwards. Specifically, the citation quotient jumped significantly in the early 1970s, from 17.9 in 1971 to 29.4 in 1972, compared to 7.3 in 1947 and 28.3 in 1974 (see Fig. ). This trend suggests the journal’s internationalization efforts in the 1970s significantly bolstered its standing within the global research community (Starke ; Francisco ). Although high citation rates and impact factors (IFs) are commonly associated with a journal’s academic prestige, IFs have been critiqued as an imperfect measure for evaluating journal quality (Seglen ). Nonetheless, citations do play a role in enhancing a publication’s international visibility, with English-language articles typically receiving more citations than non-English papers. This advantage shows that publishing in English can significantly benefit both journals and authors in terms of both international and national recognition (Vinther and Rosenberg ). The internationalization measures of the 1970s, therefore, not only enhanced the visibility of Naunyn–Schmiedeberg’s Archives of Pharmacology but also contributed to a broader acknowledgment of the work published within its pages (Starke ; Francisco ). 100 most quoted original papers Citation behavior significantly impacts both journals and their authors, as high citation counts are often associated with scientific prominence and success (Seglen ; Francisco ). A detailed bibliometric analysis of the 100 most-cited publications (Original Papers) within Naunyn–Schmiedeberg’s Archives of Pharmacology highlights the outcomes of the journal’s 1970s internationalization efforts. Notably, 50 of these pivotal publications were produced in the years 1972 and 1974 alone (Fig. ), underscoring the beneficial impact of publishing in English on citation numbers—a key factor for enduring relevance in the global scientific community (Starke ; Francisco ; Gzoyan et al. ). The analysis reveals a linguistic shift in the composition of the most-cited list, transitioning from a predominance of German-language articles before the late 1960s to a decisive majority of English-language articles post-1970. By 1971, English-language publications constituted over half of the top 100 cited works (Fig. ), reflecting the journal’s successful adaptation to the international scientific publishing landscape (Starke ). Among these highly cited articles, research on the cholinergic and adrenergic systems is particularly prominent, representing 21% of the list, followed by studies on pharmacodynamics and the dopaminergic system (Fig. ; Fig. S2). This thematic emphasis is in part due to the foundational discoveries by Otto Loewi and Sir Henry Dale on neurotransmission, which have sparked extensive research into these biological systems (Tansey ; McCoy and Tan ). Technological advancements in the 1950s that improved acetylcholine detection also played a critical role, facilitating expanded research that coincided with heightened interest due to the military application of nerve agents during and post-World War II (Dacre ; Warburton and Wesnes ; John et al. ; Amend et al. ; Hrvat and Kovarik 2020). Interest in the cholinergic and adrenergic systems surged until the 1990s before Naunyn–Schmiedeberg’s Archives of Pharmacology moved towards immunopharmacology and drugs for the treatment of malignant tumor diseases in the new millennium (Hornykiewicz ; Dats et al. ). The journal’s 1970s internationalization initiatives played a crucial role in enhancing citation metrics and expanding its international stature (Starke ; Francisco ; Hattori and Seifert ). Earlier periods also witnessed citation surges, notably in 1963 and 1968, thanks to seminal works by Huković and Muscholl and the influential article by Thoenen and Tranzer on dopaminergic neurons, which is the period’s most cited work with 717 citations (Table , last accessed 02.01.2024). Importantly, alongside the journal’s significant contributions to neuroscience, this study highlights a pivotal moment in Parkinson’s disease research. Consequently, the mid-1960s are marked by an improvement in Parkinson’s therapy, significantly contributing to the scientific understanding of dopamine’s function. The Thoenen and Tranzer article specifically explored the effects of 6-hydroxydopamine (6-OHDA) on dopaminergic neurons, demonstrating its potential to induce Parkinson’s syndrome in experimental models. This critical insight into 6-OHDA has since been instrumental in the study and treatment of Parkinson’s disease, further emphasizing the journal’s enduring influence on neurological research and therapy (Hornykiewicz ; Thoenen and Tranzer ; Simola et al. ; Fahn ; Li and Le ). Authors Between 1947 and 1974, Naunyn–Schmiedeberg’s Archives of Pharmacology saw contributions from 2065 first authors. Detailed analysis including first, second, and last authorships identified those who published (Original Papers) most frequently within this timeframe. Leading the count was Manfred Kiese (1910–1983) with 47 publications, followed closely by Gerhard Zetler (1921–2007) and Gustav Kuschinsky (1904–1992), each with 29 publications. Ernst Habermann (1926–2001) and Peter Holtz (1902–1970) also made significant contributions with 27 and 26 publications, respectively (Fig. ; Tab. S2). Manfred Kiese, who completed his doctorate under Wolfgang Heubner (1877–1957) in 1935, showed remarkable productivity, especially between 1947 and 1949, by publishing 17 articles while leading the pharmacology laboratory at the University Hospital in Kiel, a position he assumed in 1947. Kiese’s research predominantly focused on the pharmacodynamics and kinetics of methemoglobin and hemoglobin, contributing 33 articles on these topics over the years (Philippu ). Importantly, Kiese demonstrated early international engagement through his publications in English as early as 1963. One of his publications achieved remarkable recognition and is listed among the 100 most cited papers in the journal, in the 73rd place (Table ). Gerhard Zetler began publishing consistently in 1951, maintaining an average of two publications per year until 1974. Starting his career at Christian-Albrecht University in Kiel in 1949, Zetler moved to the Institute for Experimental and Clinical Pharmacology and Toxicology at the University Medical Center Schleswig–Holstein in Lübeck in 1964, serving as its first director (Philippu ). His work predominantly explored substance P and resulted in a total of 18 publications in Kiel and 11 in Lübeck. Gustav Kuschinsky, another prominent contributor with 29 articles, started publishing in the journal in 1947 and continued until 1968. After beginning his career under Paul Trendelenburg in Berlin, he moved to Tung Chi University in Shanghai in 1934 and later became a full professor at the German University in Prague in 1939 (Philippu ). His research, which did not include any English publications, focused on the cholinergic and adrenergic systems, with a significant portion of his work being published during his tenure at the Johannes Gutenberg University in Mainz. Ernst Habermann’s contributions, spanning from 1954 to 1974, were primarily in toxicology, including groundbreaking studies on bee venom (Apis mellifera). His work spanned two institutions: from 1954 to 1966 at the University of Würzburg and from 1966 to 1974 at the Justus Liebig University Giessen, with a transition to English publications starting in 1971 (Philippu ). Two of his works were cited particularly often, which underlines their importance in the scientific community. These works are ranked 28th and 49th of the 100 most cited publications (Table ). Peter Holtz, with 26 articles, focused on the cholinergic and adrenergic system, particularly on noradrenaline, earning him the National Prize of the GDR. His work at the University of Rostock and later at Goethe University in Frankfurt am Main contributed significantly to the field. Two of his papers are among the 100 most cited articles in the journal, which also indicates that his work is highly recognized. These are ranked 15th and 62nd in the list of most cited papers (Table ). Continent, countries, and cities The geographical distribution of the publications (Original Papers) further highlights the journal’s initial European, particularly German, orientation, with 94% of articles coming from Europe until 1974 (Fig. S3). This dominance underscores the limited global reach of publications in the German language, which rarely gained significance beyond Europe (Bajerski ). North America’s 4% contribution is primarily attributed to works dedicated to Otto Krayer and Otto Loewi, indicating a Western-centric publication trend (Starke ). The concentration of publications in Germany, especially in the immediate post-war era, can be seen as an effort to mitigate the country’s international isolation and foster reintegration into the global scientific community. This period also saw a shift in focus from national prestige to the pursuit of international recognition among German scientists after the second world war (Ahlers et al. ). A comparison of the present data with the results of Dats et al. for the period 1990 to 2020 reveals an increasing internationalization of the journal until 2020. Particularly, a significant increase in the representation of Asian publications can be observed. While the Asian continent was clearly underrepresented until 1974 with only 47 publications (Fig. S4), Asia positioned itself as the second most represented continent from 1990 to 2020. The analysis by Dats et al. also reveals an increase in publications from South America, with the number of publications increasing from an under-representation of only 20 publications up to 1974 to around 250 in the period from 1990 to 2020, with the majority of these publications coming from Brazil. In contrast, the number of publications from Africa and Australia did not change relatively in the observed period by Dats et al. . In the analyzed period from 1947 to 1974, an overwhelming 84% of publications in Naunyn–Schmiedeberg’s Archives of Pharmacology originated from Germany, with Austria and Switzerland contributing 5% and 4%, respectively (see Fig. S4). These data underscore that during this period, 93% of the journal's publications were from German-speaking countries, aligning with the perception of the journal as predominantly German in its focus (Fig. ). Post-World War II, a modest 10% of contributions came from non-German-speaking countries. However, the 1970s marked a pivotal shift towards internationalization, leading to a significant increase in contributions from outside the German-speaking countries. By the late 1990s, international contributions constituted 60% of the total publications, reflecting the journal’s successful global integration (Starke ; Hamel ). Figures and show that most publications (Original Papers) in the period from 1947 to 1974, with a total of 1971 publications, came from West-Germany. Whereas 616 publications came from East-Germany, a closer look shows that 328 of these works were published in West-Berlin. Taking this categorization into consideration, only about 12% of the publications, namely 288, came from the geographical area of the former GDR; 55 of these works were published in East-Berlin. This publication distribution across German cities indicates a pronounced clustering in Western Germany (Figs. , , and ). In 1974, the population of the GDR was 16.891 millions and the FRG 61.99 millions (Statistisches Bundesamt Demografische Aspekte Deutschland ). West-Germany produced 1971 publications (Original Papers) in total, whereas East-Germany produced 288 publications (Original Papers; Fig. ). In terms of population size, the FRG, including West-Berlin, produced around 31.8 publications per million inhabitants, while the GDR, including East-Berlin, recorded around 17.1 publications per million inhabitants. Although the population in West-Germany is almost four times as high as in East Germany, East Germany was still quite productive with 17.1 publications (original paper) (Fig. and Table ). However, the discrepancy increases when Berlin is considered individually in East and West. West-Berlin had a population of around 2.1 million people (1974), while East Berlin had a population of around 1.1 million (1974) (Statistisches Bundesamt Demografische Aspekte Deutschland ). During the time between 1947–1974, publications per 1 million inhabitants in West-Berlin amounted to 155, whereas publications per 1 million inhabitants in East-Berlin in 1974 amounted to 51 (Table ). These figures illustrate the higher scientific productivity in the FRG compared to the GDR (Fig. ). Nonetheless, the figures also show that even with limited financial resources, one can be scientifically productive. Berlin emerged as the leading city in terms of publication volume (Original Papers) with 383, significantly outpacing Göttingen (rank 2) and Mainz (rank 3) (Fig. ; Fig. S5 and S7). A closer look at Berlin allows us to recognize differences between the pharmacological institutes in West-Berlin and East-Berlin and they publication rates over time (Fig. ). Between 1947 and 1974, the pharmacology institutes in West-Berlin were responsible for over 85% of publications (328 Original Papers), clearly surpassing the institutes in East-Berlin (55 Original Papers) (Fig. ). Excluding 1952, when East-Berlin institutes in the GDR contributed to approximately 36% of that year’s publications, West-Berlin institutes led in scientific output. Post-1952, East-Berlin’s publication numbers dwindled, halting almost entirely after 1962, with rare exceptions in 1964 and 1967 (Fig. ). This decline is linked to the Cold War’s deepening, particularly after 1963, when East German pharmacologists were barred from participating in DGPT meetings. This restriction severely curtailed, if not entirely severed, scientific collaborations between pharmacologists from East and West Germany (Starke ). In the aftermath of World War II, the DGPT showcased remarkable resilience and inclusivity, keeping its membership unified across the East–West divide. Early post-war gatherings, like the significant 1948 Düsseldorf meeting, saw participation from both East- and West-German pharmacologists. However, the intensifying Cold War tensions and Soviet policies made it impossible for East German members to participate, prompting them to form a separate society. Despite these challenges, the DGPT remained committed to fostering scientific dialogue and cooperation across the geopolitical divide, until external pressures necessitated a separate organization for East-German pharmacologists (Starke ). Between 1951 and 1967, the presence of East-German pharmacologists in the Naunyn–Schmiedeberg’s Archives of Pharmacology significantly diminished, with a notable focus on research on drugs of treatment of heart failure and coronary heart disease, evidenced by 22 out of 49 original papers (Fig. ). This focus mirrors the rise in cardiovascular disease mortality rates noted in the Soviet Union, a trend that the GDR likely experienced due to comparable socio-economic and environmental factors (Cooper ; Jargin ). With the intensification of the Cold War and the consequent isolation, East-German researchers shifted their contributions to “Die Pharmazie,” a journal established in 1946 in the GDR under the challenging conditions of the Soviet Occupation Zone. Remarkably, despite initial contributions from both Eastern and Western authors, political tensions also led to a decline in cross-border collaborations, particularly after the erection of the Berlin Wall. Nonetheless, “Die Pharmazie” continued to prioritize scientific inquiry over political content, gradually increasing its English-language publications and maintaining its significance as a scientific platform originating from the GDR (Friedrich and Helmstädter ). Shortly after World War II, Germany was divided into four zones controlled by the Allies, leading to the formation of two separate states in 1949. The Federal Republic of Germany (FRG), formed from the western zones occupied by the United States, the United Kingdom, and France, adopted a capitalist system and embraced western democratic values, rapidly becoming integrated into the western bloc. Established in the Soviet-occupied zone, the German Democratic Republic (GDR) adopted a socialist economy and centralized governance and aligned itself with the communist ideology of the Soviet Union. This partition embodied the ideological fissure of the Cold War, with the FRG and the GDR each serving as outpost states for the competing Western and Eastern blocs (Kalberg ; Kastner ; Berger ). West-Berlin’s prominence is attributed to the city’s vigorous push towards internationalization post-war and substantial financial support from the American occupation zone aimed at revitalizing academic activities. This strategic and financial backing significantly contributed to the resurgence of scientific work in West-Berlin, positioning it as a central hub for pharmacological research and publication (Philippu ; Heinsohn and Nicolaysen ; Dats et al. ). Dats et al. document a notable decline in Berlin’s output from 1990 to 2020 compared to the earlier period examined. Nonetheless, during this latter period, Berlin remains the epicenter for scholarly output in Eastern Germany. The enduring effects of post-World War II financial investments have had a long-term beneficial influence on academic institutions well into the millennium transition (Philippu ; Heinsohn and Nicolaysen ; Dats et al. ). Moreover, Freiburg ascended to the first position in the 1990–2020 timeframe, surpassing Berlin as the preeminent city. Conversely, Göttingen, previously ranked second, saw a significant reduction in its publication activity. Bonn, ranked twelfth up to 1974, advanced to second place during the 1990–2020 span (Dats et al. ; Fig. S7). This analysis reveals dynamic shifts in publication rates (Original Papers) across various cities, with significant fluctuations observed. Despite these changes, the overarching pattern of publication concentration in Western Germany has persisted through many decades (Dats et al. ; Fig. ). Limitations and future studies The bibliometric analysis conducted through data extraction from SpringerLink with Phython and Beautiful Soup provided a comprehensive yet focused insight into Naunyn–Schmiedeberg’s Archives of Pharmacology, centering on “Original Papers”. This approach, however, offered a limited view, excluding a variety of content such as short papers, DGPT spring meetings, abstracts, reviews, and reports etc. that also form an integral part of the journal’s content. The exclusion of reviews (SN-Type) was a strategic decision for this study, setting the stage for a subsequent study dedicated to exploring reviews in Naunyn–Schmiedeberg’s Archives of Pharmacology from its establishment in 1873 to the current day. The methodology described in this paper permits an in-depth evaluation over the journal’s full 150-year history. Investigating the initial resistance to English language adoption by Dr. Springer, the publisher, could also shed light on internal dynamics and resistance to internationalization within the journal’s editorial board, which saw the journal as distinctly German. Following World War II, Naunyn–Schmiedeberg’s Archives of Pharmacology faced a publication halt for 2 years, resuming in February 1947 with its first post-war volume (204), delayed due to US intelligence control oversight (Starke ; Herken ). Initially, the journal’s productivity was low, publishing around 70 articles each in 1947 and 1948, a direct consequence of war-induced destruction of academic facilities and the emigration of scientists during the Nazi regime, hindering immediate post-war scientific activities (Philippu ; Löffelholz ; Weise-Pötschke ; Heinsohn and Nicolaysen ; Dats et al. ; Hattori and Seifert ; Fig. ). By 1949, publication numbers surged to 122, reflecting the release of wartime research (Dats et al. ). This growth continued, notably doubling around 1951 and 1953, partly due to celebrating pharmacologist Otto Loewi’s 80th birthday (Starke ). Despite a peak in 1964 honoring Otto Krayer’s 65th birthday, the journal experienced a gradual decline in the number of publications until 1970, influenced by the preference for publishing in higher-impact English-language journals (Zehetbauer et al. ; Dats et al. ; Gzoyan et al. ). The shift towards English, essential for international scientific communication, led to the marginalization of German-language journals (Gzoyan et al. ). Recognizing the need for internationalization, the journal’s editors initiated significant changes in the late 1960s, including the internationalization of the editorial board and the transition to publishing in English, culminating in the journal’s name change in 1971 and the mandate for English publications from 1973 onwards. These measures revitalized the journal, as evidenced by a 149% increase in publications from 1970 to 1973, stabilizing its contribution to the international research community thereafter (Starke ; Dats et al. ; Hattori and Seifert ; Fig. ). Naunyn–Schmiedeberg’s Archives of Pharmacology predominantly featured German-language publications until the late 1960s, reflecting its national orientation (Fig. ). This linguistic homogeneity meant its research largely remained within German-speaking circles, limiting international engagement and recognition (Bajerski ). Analysis of publication trends (Fig. ) against the language of publication (Fig. ) reveals a parallel between the overall publication volume and German-language articles up to 1970. Notably, spikes in English publications in 1953 and 1964 corresponded with contributions from the USA, including works dedicated to emigrant pharmacologists Otto Loewi and Otto Krayer, highlighting brief periods of international collaboration (Starke ). Loewi, a former student of Oswald Schmiedeberg, fled Nazi Germany in 1938, eventually settling in the USA at New York University (Bettendorf ; Philippu ; McCoy and Tan ). Similarly, Krayer, who refused a position vacated under Nazi policies, led Harvard University’s pharmacology institute from 1939 to 1966 (Starke ; Philippu ; Rubin ). The contributions of Loewi and Krayer underscore the journal’s intermittent international reach. The sharp increase in English publications from 1970 onwards (Fig. ) reflects editorial efforts to promote English, transitioning from five English articles in 1967 to 164 by 1973, with a corresponding decline in German articles. This shift was instrumental in repositioning the journal within the global scientific community (Starke ; Francisco ; Hattori and Seifert ). Post-WWII, Germany’s diminished academic stature and UNESCO’s advocacy for multilingualism in scientific publishing prompted questions about delayed internationalization efforts. The reluctance of the journal’s publisher, Dr. Springer, to adopt multilingualism until the late 1960s contributed to this lag (DGPT archive in Göttingen 01.09.1949). The ascendancy of English as the primary scientific language post-1970s highlights a broader shift towards internationalization in German science, necessitated by a globalizing research landscape (Winkmann et al. ; Heinsohn and Nicolaysen ). The debate over the late adoption of English in German journals touches on the broader discourse around multilingualism’s role in ensuring high-quality, globally communicable science (Tardy ; Billings 2015; Davydova ). The imposition of English as the lingua franca raises ethical questions about linguistic equity and the potential marginalization of non-English speakers (Phillipson ; O’Neil ). Moreover, the dominance of English reflects colonial legacies, risking the devaluation of other languages and the sidelining of non-English scientific contributions (Phillipson ; Ahn et al. ). The international neglect of non-English journals, resulting in fewer citations and recognition, underscores the challenges of linguistic isolation in the global scientific community (Phillipson ; Gzoyan et al. ). While high publication volumes post-1970s indicate increased scientific productivity, assessing the international recognition of Naunyn–Schmiedeberg’s Archives of Pharmacology and its articles requires examining citation patterns (Francisco ). Figure presents annual publication counts alongside total citations from 1947 to 1974, calculating an average citation rate per article. This metric reveals a gradual increase in citation impact, with a notable surge from 1963 onwards. Specifically, the citation quotient jumped significantly in the early 1970s, from 17.9 in 1971 to 29.4 in 1972, compared to 7.3 in 1947 and 28.3 in 1974 (see Fig. ). This trend suggests the journal’s internationalization efforts in the 1970s significantly bolstered its standing within the global research community (Starke ; Francisco ). Although high citation rates and impact factors (IFs) are commonly associated with a journal’s academic prestige, IFs have been critiqued as an imperfect measure for evaluating journal quality (Seglen ). Nonetheless, citations do play a role in enhancing a publication’s international visibility, with English-language articles typically receiving more citations than non-English papers. This advantage shows that publishing in English can significantly benefit both journals and authors in terms of both international and national recognition (Vinther and Rosenberg ). The internationalization measures of the 1970s, therefore, not only enhanced the visibility of Naunyn–Schmiedeberg’s Archives of Pharmacology but also contributed to a broader acknowledgment of the work published within its pages (Starke ; Francisco ). Citation behavior significantly impacts both journals and their authors, as high citation counts are often associated with scientific prominence and success (Seglen ; Francisco ). A detailed bibliometric analysis of the 100 most-cited publications (Original Papers) within Naunyn–Schmiedeberg’s Archives of Pharmacology highlights the outcomes of the journal’s 1970s internationalization efforts. Notably, 50 of these pivotal publications were produced in the years 1972 and 1974 alone (Fig. ), underscoring the beneficial impact of publishing in English on citation numbers—a key factor for enduring relevance in the global scientific community (Starke ; Francisco ; Gzoyan et al. ). The analysis reveals a linguistic shift in the composition of the most-cited list, transitioning from a predominance of German-language articles before the late 1960s to a decisive majority of English-language articles post-1970. By 1971, English-language publications constituted over half of the top 100 cited works (Fig. ), reflecting the journal’s successful adaptation to the international scientific publishing landscape (Starke ). Among these highly cited articles, research on the cholinergic and adrenergic systems is particularly prominent, representing 21% of the list, followed by studies on pharmacodynamics and the dopaminergic system (Fig. ; Fig. S2). This thematic emphasis is in part due to the foundational discoveries by Otto Loewi and Sir Henry Dale on neurotransmission, which have sparked extensive research into these biological systems (Tansey ; McCoy and Tan ). Technological advancements in the 1950s that improved acetylcholine detection also played a critical role, facilitating expanded research that coincided with heightened interest due to the military application of nerve agents during and post-World War II (Dacre ; Warburton and Wesnes ; John et al. ; Amend et al. ; Hrvat and Kovarik 2020). Interest in the cholinergic and adrenergic systems surged until the 1990s before Naunyn–Schmiedeberg’s Archives of Pharmacology moved towards immunopharmacology and drugs for the treatment of malignant tumor diseases in the new millennium (Hornykiewicz ; Dats et al. ). The journal’s 1970s internationalization initiatives played a crucial role in enhancing citation metrics and expanding its international stature (Starke ; Francisco ; Hattori and Seifert ). Earlier periods also witnessed citation surges, notably in 1963 and 1968, thanks to seminal works by Huković and Muscholl and the influential article by Thoenen and Tranzer on dopaminergic neurons, which is the period’s most cited work with 717 citations (Table , last accessed 02.01.2024). Importantly, alongside the journal’s significant contributions to neuroscience, this study highlights a pivotal moment in Parkinson’s disease research. Consequently, the mid-1960s are marked by an improvement in Parkinson’s therapy, significantly contributing to the scientific understanding of dopamine’s function. The Thoenen and Tranzer article specifically explored the effects of 6-hydroxydopamine (6-OHDA) on dopaminergic neurons, demonstrating its potential to induce Parkinson’s syndrome in experimental models. This critical insight into 6-OHDA has since been instrumental in the study and treatment of Parkinson’s disease, further emphasizing the journal’s enduring influence on neurological research and therapy (Hornykiewicz ; Thoenen and Tranzer ; Simola et al. ; Fahn ; Li and Le ). Between 1947 and 1974, Naunyn–Schmiedeberg’s Archives of Pharmacology saw contributions from 2065 first authors. Detailed analysis including first, second, and last authorships identified those who published (Original Papers) most frequently within this timeframe. Leading the count was Manfred Kiese (1910–1983) with 47 publications, followed closely by Gerhard Zetler (1921–2007) and Gustav Kuschinsky (1904–1992), each with 29 publications. Ernst Habermann (1926–2001) and Peter Holtz (1902–1970) also made significant contributions with 27 and 26 publications, respectively (Fig. ; Tab. S2). Manfred Kiese, who completed his doctorate under Wolfgang Heubner (1877–1957) in 1935, showed remarkable productivity, especially between 1947 and 1949, by publishing 17 articles while leading the pharmacology laboratory at the University Hospital in Kiel, a position he assumed in 1947. Kiese’s research predominantly focused on the pharmacodynamics and kinetics of methemoglobin and hemoglobin, contributing 33 articles on these topics over the years (Philippu ). Importantly, Kiese demonstrated early international engagement through his publications in English as early as 1963. One of his publications achieved remarkable recognition and is listed among the 100 most cited papers in the journal, in the 73rd place (Table ). Gerhard Zetler began publishing consistently in 1951, maintaining an average of two publications per year until 1974. Starting his career at Christian-Albrecht University in Kiel in 1949, Zetler moved to the Institute for Experimental and Clinical Pharmacology and Toxicology at the University Medical Center Schleswig–Holstein in Lübeck in 1964, serving as its first director (Philippu ). His work predominantly explored substance P and resulted in a total of 18 publications in Kiel and 11 in Lübeck. Gustav Kuschinsky, another prominent contributor with 29 articles, started publishing in the journal in 1947 and continued until 1968. After beginning his career under Paul Trendelenburg in Berlin, he moved to Tung Chi University in Shanghai in 1934 and later became a full professor at the German University in Prague in 1939 (Philippu ). His research, which did not include any English publications, focused on the cholinergic and adrenergic systems, with a significant portion of his work being published during his tenure at the Johannes Gutenberg University in Mainz. Ernst Habermann’s contributions, spanning from 1954 to 1974, were primarily in toxicology, including groundbreaking studies on bee venom (Apis mellifera). His work spanned two institutions: from 1954 to 1966 at the University of Würzburg and from 1966 to 1974 at the Justus Liebig University Giessen, with a transition to English publications starting in 1971 (Philippu ). Two of his works were cited particularly often, which underlines their importance in the scientific community. These works are ranked 28th and 49th of the 100 most cited publications (Table ). Peter Holtz, with 26 articles, focused on the cholinergic and adrenergic system, particularly on noradrenaline, earning him the National Prize of the GDR. His work at the University of Rostock and later at Goethe University in Frankfurt am Main contributed significantly to the field. Two of his papers are among the 100 most cited articles in the journal, which also indicates that his work is highly recognized. These are ranked 15th and 62nd in the list of most cited papers (Table ). The geographical distribution of the publications (Original Papers) further highlights the journal’s initial European, particularly German, orientation, with 94% of articles coming from Europe until 1974 (Fig. S3). This dominance underscores the limited global reach of publications in the German language, which rarely gained significance beyond Europe (Bajerski ). North America’s 4% contribution is primarily attributed to works dedicated to Otto Krayer and Otto Loewi, indicating a Western-centric publication trend (Starke ). The concentration of publications in Germany, especially in the immediate post-war era, can be seen as an effort to mitigate the country’s international isolation and foster reintegration into the global scientific community. This period also saw a shift in focus from national prestige to the pursuit of international recognition among German scientists after the second world war (Ahlers et al. ). A comparison of the present data with the results of Dats et al. for the period 1990 to 2020 reveals an increasing internationalization of the journal until 2020. Particularly, a significant increase in the representation of Asian publications can be observed. While the Asian continent was clearly underrepresented until 1974 with only 47 publications (Fig. S4), Asia positioned itself as the second most represented continent from 1990 to 2020. The analysis by Dats et al. also reveals an increase in publications from South America, with the number of publications increasing from an under-representation of only 20 publications up to 1974 to around 250 in the period from 1990 to 2020, with the majority of these publications coming from Brazil. In contrast, the number of publications from Africa and Australia did not change relatively in the observed period by Dats et al. . In the analyzed period from 1947 to 1974, an overwhelming 84% of publications in Naunyn–Schmiedeberg’s Archives of Pharmacology originated from Germany, with Austria and Switzerland contributing 5% and 4%, respectively (see Fig. S4). These data underscore that during this period, 93% of the journal's publications were from German-speaking countries, aligning with the perception of the journal as predominantly German in its focus (Fig. ). Post-World War II, a modest 10% of contributions came from non-German-speaking countries. However, the 1970s marked a pivotal shift towards internationalization, leading to a significant increase in contributions from outside the German-speaking countries. By the late 1990s, international contributions constituted 60% of the total publications, reflecting the journal’s successful global integration (Starke ; Hamel ). Figures and show that most publications (Original Papers) in the period from 1947 to 1974, with a total of 1971 publications, came from West-Germany. Whereas 616 publications came from East-Germany, a closer look shows that 328 of these works were published in West-Berlin. Taking this categorization into consideration, only about 12% of the publications, namely 288, came from the geographical area of the former GDR; 55 of these works were published in East-Berlin. This publication distribution across German cities indicates a pronounced clustering in Western Germany (Figs. , , and ). In 1974, the population of the GDR was 16.891 millions and the FRG 61.99 millions (Statistisches Bundesamt Demografische Aspekte Deutschland ). West-Germany produced 1971 publications (Original Papers) in total, whereas East-Germany produced 288 publications (Original Papers; Fig. ). In terms of population size, the FRG, including West-Berlin, produced around 31.8 publications per million inhabitants, while the GDR, including East-Berlin, recorded around 17.1 publications per million inhabitants. Although the population in West-Germany is almost four times as high as in East Germany, East Germany was still quite productive with 17.1 publications (original paper) (Fig. and Table ). However, the discrepancy increases when Berlin is considered individually in East and West. West-Berlin had a population of around 2.1 million people (1974), while East Berlin had a population of around 1.1 million (1974) (Statistisches Bundesamt Demografische Aspekte Deutschland ). During the time between 1947–1974, publications per 1 million inhabitants in West-Berlin amounted to 155, whereas publications per 1 million inhabitants in East-Berlin in 1974 amounted to 51 (Table ). These figures illustrate the higher scientific productivity in the FRG compared to the GDR (Fig. ). Nonetheless, the figures also show that even with limited financial resources, one can be scientifically productive. Berlin emerged as the leading city in terms of publication volume (Original Papers) with 383, significantly outpacing Göttingen (rank 2) and Mainz (rank 3) (Fig. ; Fig. S5 and S7). A closer look at Berlin allows us to recognize differences between the pharmacological institutes in West-Berlin and East-Berlin and they publication rates over time (Fig. ). Between 1947 and 1974, the pharmacology institutes in West-Berlin were responsible for over 85% of publications (328 Original Papers), clearly surpassing the institutes in East-Berlin (55 Original Papers) (Fig. ). Excluding 1952, when East-Berlin institutes in the GDR contributed to approximately 36% of that year’s publications, West-Berlin institutes led in scientific output. Post-1952, East-Berlin’s publication numbers dwindled, halting almost entirely after 1962, with rare exceptions in 1964 and 1967 (Fig. ). This decline is linked to the Cold War’s deepening, particularly after 1963, when East German pharmacologists were barred from participating in DGPT meetings. This restriction severely curtailed, if not entirely severed, scientific collaborations between pharmacologists from East and West Germany (Starke ). In the aftermath of World War II, the DGPT showcased remarkable resilience and inclusivity, keeping its membership unified across the East–West divide. Early post-war gatherings, like the significant 1948 Düsseldorf meeting, saw participation from both East- and West-German pharmacologists. However, the intensifying Cold War tensions and Soviet policies made it impossible for East German members to participate, prompting them to form a separate society. Despite these challenges, the DGPT remained committed to fostering scientific dialogue and cooperation across the geopolitical divide, until external pressures necessitated a separate organization for East-German pharmacologists (Starke ). Between 1951 and 1967, the presence of East-German pharmacologists in the Naunyn–Schmiedeberg’s Archives of Pharmacology significantly diminished, with a notable focus on research on drugs of treatment of heart failure and coronary heart disease, evidenced by 22 out of 49 original papers (Fig. ). This focus mirrors the rise in cardiovascular disease mortality rates noted in the Soviet Union, a trend that the GDR likely experienced due to comparable socio-economic and environmental factors (Cooper ; Jargin ). With the intensification of the Cold War and the consequent isolation, East-German researchers shifted their contributions to “Die Pharmazie,” a journal established in 1946 in the GDR under the challenging conditions of the Soviet Occupation Zone. Remarkably, despite initial contributions from both Eastern and Western authors, political tensions also led to a decline in cross-border collaborations, particularly after the erection of the Berlin Wall. Nonetheless, “Die Pharmazie” continued to prioritize scientific inquiry over political content, gradually increasing its English-language publications and maintaining its significance as a scientific platform originating from the GDR (Friedrich and Helmstädter ). Shortly after World War II, Germany was divided into four zones controlled by the Allies, leading to the formation of two separate states in 1949. The Federal Republic of Germany (FRG), formed from the western zones occupied by the United States, the United Kingdom, and France, adopted a capitalist system and embraced western democratic values, rapidly becoming integrated into the western bloc. Established in the Soviet-occupied zone, the German Democratic Republic (GDR) adopted a socialist economy and centralized governance and aligned itself with the communist ideology of the Soviet Union. This partition embodied the ideological fissure of the Cold War, with the FRG and the GDR each serving as outpost states for the competing Western and Eastern blocs (Kalberg ; Kastner ; Berger ). West-Berlin’s prominence is attributed to the city’s vigorous push towards internationalization post-war and substantial financial support from the American occupation zone aimed at revitalizing academic activities. This strategic and financial backing significantly contributed to the resurgence of scientific work in West-Berlin, positioning it as a central hub for pharmacological research and publication (Philippu ; Heinsohn and Nicolaysen ; Dats et al. ). Dats et al. document a notable decline in Berlin’s output from 1990 to 2020 compared to the earlier period examined. Nonetheless, during this latter period, Berlin remains the epicenter for scholarly output in Eastern Germany. The enduring effects of post-World War II financial investments have had a long-term beneficial influence on academic institutions well into the millennium transition (Philippu ; Heinsohn and Nicolaysen ; Dats et al. ). Moreover, Freiburg ascended to the first position in the 1990–2020 timeframe, surpassing Berlin as the preeminent city. Conversely, Göttingen, previously ranked second, saw a significant reduction in its publication activity. Bonn, ranked twelfth up to 1974, advanced to second place during the 1990–2020 span (Dats et al. ; Fig. S7). This analysis reveals dynamic shifts in publication rates (Original Papers) across various cities, with significant fluctuations observed. Despite these changes, the overarching pattern of publication concentration in Western Germany has persisted through many decades (Dats et al. ; Fig. ). The bibliometric analysis conducted through data extraction from SpringerLink with Phython and Beautiful Soup provided a comprehensive yet focused insight into Naunyn–Schmiedeberg’s Archives of Pharmacology, centering on “Original Papers”. This approach, however, offered a limited view, excluding a variety of content such as short papers, DGPT spring meetings, abstracts, reviews, and reports etc. that also form an integral part of the journal’s content. The exclusion of reviews (SN-Type) was a strategic decision for this study, setting the stage for a subsequent study dedicated to exploring reviews in Naunyn–Schmiedeberg’s Archives of Pharmacology from its establishment in 1873 to the current day. The methodology described in this paper permits an in-depth evaluation over the journal’s full 150-year history. Investigating the initial resistance to English language adoption by Dr. Springer, the publisher, could also shed light on internal dynamics and resistance to internationalization within the journal’s editorial board, which saw the journal as distinctly German. The decline in publication numbers (Original Papers) in the early 1950s (Fig. ) marked a critical juncture for Naunyn–Schmiedeberg’s Archives of Pharmacology, reflecting broader shifts within the scientific community and highlighting the journal’s struggle with post-war isolation due to its adherence to German. This period underscored the limitations of a national focus amidst a rapidly globalizing research landscape (Starke ; Francisco ; Gzoyan et al. ). Recognizing these challenges, the journal embarked on a path towards internationalization in the late 1960s, a move symbolized by the adoption of English for publications by 1973, which was pivotal in revitalizing the journal’s relevance and broadening its audience (Starke ; Dats et al. ; Hattori and Seifert ). By the end of the 1990s, the composition of the journal’s authorship had dramatically shifted, transitioning from a predominantly German-centric outlet to one enriched by diverse international contributions (Dats et al. ). This transformation not only mirrors the evolving dynamics of scientific inquiry but also highlights the essential role of internationalization in fostering scientific collaboration and communication. The adoption of English as the lingua franca of science, while facilitating wider discourse, also introduces challenges for non-English speaking scientists, prompting reflections on access and equity within the global knowledge community (Tardy ; Di Bitetti and Ferreras ; O’Neil ). This narrative extends beyond bibliometric analysis to engage with the discourses of internationality and scientific legitimacy, challenging the community to address the complexities of linguistic dominance. It underscores the necessity for journals like Naunyn–Schmiedeberg’s Archives of Pharmacology to navigate globalization thoughtfully, promoting scientific excellence and inclusivity (Phillipson ; Ahn et al. ; O’Neil ; Ahlers et al. ). From its origins as a German journal to its current status as an internationally recognized publication, Naunyn–Schmiedeberg’s Archives of Pharmacology illustrates this interplay between tradition and innovation in scientific publishing. Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 456 KB) |
Reconciling emergences: An information-theoretic approach to identify causal emergence in multivariate data | 2e3309ab-2801-4f55-b564-34b63f245edc | 7833221 | Physiology[mh] | While most of our representations of the physical world are hierarchical, there is still no agreement on how the co-existing “layers” of this hierarchy interact. On the one hand, reductionism claims that all levels can always be explained based on sufficient knowledge of the lowest scale and, consequently—taking an intentionally extreme example—that a sufficiently accurate theory of elementary particles should be able to predict the existence of social phenomena like communism. On the other hand, emergentism argues that there can be autonomy between layers, i.e. that some phenomena at macroscopic layers might only be accountable in terms of other macroscopic phenomena. While emergentism might seem to better serve our intuition, it is not entirely clear how a rigorous theory of emergence could be formulated within our modern scientific worldview, which tends to be dominated by reductionist principles. Emergent phenomena are usually characterised as either strong or weak . Strong emergence corresponds to the somewhat paradoxical case of supervenient properties with irreducible causal power ; i.e. properties that are fully determined by microscopic levels but can nevertheless exert causal influences that are not entirely accountable from microscopic considerations (the case of strong emergence most commonly argued in the literature is the one of conscious experiences with respect to their corresponding physical substrate ). Strong emergence has been as much a cause of wonder as a perennial source of philosophical headaches, being described as “uncomfortably like magic” while accused of being logically inconsistent and sustained on illegitimate metaphysics . Weak emergence has been proposed as a more docile alternative to strong emergence, where macroscopic features have irreducible causal power in practice but not in principle. A popular formulation of weak emergence is due to Bedau , and corresponds to properties generated by elements at microscopic levels in such complicated ways that they cannot be derived via explanatory shortcuts, but only by exhaustive simulation. While this formulation is usually accepted by the scientific community, it is not well-suited to address mereological questions about emergence in scenarios where parts-whole relationships are the primary interest. Part of the difficulty in building a deeper understanding of strong emergence is the absence of simple but clear analytical models that can serve the community to guide discussions and mature theories. Efforts have been made to introduce quantitative metrics of weak emergence , which enable fine-grained data-driven alternatives to traditional all-or-none classifications. In this vein, an attractive alternative comes from the work on causal emergence introduced in Ref. and later developed in Refs. , which showed that macroscopic observables can sometimes exhibit more causal power (as understood within the framework of Pearl’s do-calculus ) than microscopic variables. However, this framework relies on strong assumptions that are rarely satisfied in practice, which severely hinders its applicability (this point is further elaborated in Section Relationship with other quantitative theories of emergence ). Inspired by Refs. , here we introduce a practically useful and philosophically innocent framework to study causal emergence in multivariate data. Building on previous work , we take the perspective of an experimentalist who has no prior knowledge of the underlying phenomenon of interest, but has sufficient data of all relevant variables that allows an accurate statistical description of the phenomenon. In this context, we put forward a formal definition of causal emergence that doesn’t rely on coarse-graining functions as Ref. , but addresses the “paradoxical” properties of strong emergence based on the laws of information flow in multivariate systems. The main contribution of this work is to enable a rigorous, quantitative definition of downward causation , and introduce a novel notion of causal decoupling as a complementary modality of causal emergence. Another contribution is to extend the domain of applicability of causal emergence analyses to include cases of observational data, in which case causality ought to be understood in the Granger sense, i.e. as predictive ability . Furthermore, our framework yields practical criteria that can be effectively applied to large systems, bypassing prohibitive estimation issues that severely restrict previous approaches. The rest of this paper is structured as follows. First, Section Fundamental intuitions discusses minimal examples of emergence. Then, Section A formal theory of causal emergence presents the core of our theory, and Section Measuring emergence discusses practical methods to quantify emergence from experimental data. Our framework is then illustrated on a number of case studies, presented in Section Case studies . Finally, the Section Discussion concludes the paper with a discussion of some of the implications of our findings.
To ground our intuitions, let us introduce minimal examples that embody a few key notions of causally emergent behaviour. Throughout this section, we consider systems composed of n parts described by a binary vector X t = ( X t 1 , … , X t n ) ∈ { 0 , 1 } n , which undergo Markovian stochastic dynamics following a transition probability p X t + 1 | X t . For simplicity, we assume that at time t the system is found in an entirely random configuration (i.e. p X t ( x t ) = 2 - n ). From there, we consider three evolution rules. Example 1 . Consider a temporal evolution where the parity of X t is preserved with probability γ ∈ (0, 1). Mathematically , p X t + 1 | X t ( x t + 1 | x t ) = { γ 2 n - 1 if ⊕ j = 1 n x t + 1 j = ⊕ j = 1 n x t j , 1 - γ 2 n - 1 otherwise , for all t ∈ N , where ⊕ j = 1 n a j ≔ 1 if ∑ j = 1 n a j is even and zero otherwise. Put simply : x t +1 is a random sample from the set of all strings with the same parity as x t with probability γ; and is a sample from the strings with opposite parity with probability 1 − γ . This evolution rule has a number of interesting properties. First, the system has a non-trivial causal structure, since some properties of the future state (its parity) can be predicted from the past state. However, this structure is noticeable only at the collective level, as no individual variable has any predictive power over the evolution of itself or any other variable (see ). Furthermore, even the complete past of the system X t has no predictive power over any individual future X t + 1 j . This case shows an extreme kind of causal emergence that we call “causal decoupling,” in which the parity predicts its own evolution but no element (or subset of elements) predicts the evolution of any other element . Example 2 . Consider now a system where the parity of X t determines X t + 1 1 (i.e . X t + 1 1 = ⊕ i = 1 n X t i ), and X t + 1 j for j ≠ 1 is a fair coin flip independent of X t (see ). In this scenario X t predicts X t + 1 1 with perfect accuracy, while it can be verified that X t i ⫫ X t + 1 1 for all i ∈ {1, …, n }. Therefore, under this evolution rule the whole system has a causal effect over a particular element, although this effect cannot be attributed to any individual part (for a related discussion, see Ref . ), being a minimal example of downward causation . Example 3 . Let us now study an evolution rule that includes the mechanisms of both Examples 1 and 2. Concretely, consider p X t + 1 | X t ( x t + 1 | x t ) = { 0 if x t + 1 1 ≠ ⊕ j = 1 n x t j , γ 2 n - 2 if x t + 1 1 = ⊕ j = 1 n x t j and ⊕ j = 1 n x t + 1 j = ⊕ j = 1 n x t j , 1 - γ 2 n - 2 otherwise . As in Example 1, the parity of X t is transfered to X t +1 with probability γ; additionally, it is guaranteed that X t + 1 1 = ⊕ i = 1 n X t i . Hence, in this case not only is there a macroscopic effect that cannot be explained from the parts, but at the same time there is another effect going from the whole to one of the parts. Importantly, both effects co-exist independently of each other . The above are minimal examples of dynamical laws that cannot be traced from the interactions between their elementary components: Example 1 shows how a collective property can propagate without interacting with its underlying substrate; Example 2 how a collective property can influence the evolution of specific parts; and Example 3 how these two kinds of phenomena take place in the same system. All these issues are formalised by the theory developed in the next section.
This section presents the main body of our theory of causal emergence. To fix ideas, we consider a scientist measuring a system composed of n parts. The scientist is assumed to measure the system regularly over time, and the results of those measurements are denoted by X t = ( X t 1 , … , X t n ) , with X t i ∈ X i corresponding to the state of the i th part at time t ∈ N with phase space X i . When referring to a collection of parts, we use the notation X t α = ( X t i 1 , … , X t i K ) for α = { i 1 , …, i K }⊂{1, …, n }. We also use the shorthand notation [ n ] ≔ {1, …, n }. Supervenience Our analysis considers two time points of the evolution of the system, denoted as t and t ′, with t < t ′. The corresponding dynamics are encoded in the transition probability p X t ′ | X t ( x t ′ | x t ) . We consider features V t ∈ V generated via a conditional probability p V t | X t that are supervenient on the underlying system; i.e. that does not provide any predictive power for future states at times t ′ > t if the complete state of the system at time t is known with perfect precision. We formalise this in the following definition. Definition 1 . A stochastic process V t is said to be supervenient over X t if V t − X t − X t ′′ form a Markov chain for all t ′′ ≠ t . The above condition is equivalent to require V t to be statistically independent of X t ′′ when X t is given. The relationship between supervenient features and the underlying system is illustrated in . This formalisation of supervenience characterises features V t that are fully determined by the state of the system at a given time t , but also allows the feature to be noisy—which is not critical for our results, but is useful for extending their domain of applicability to practical scenarios. In effect, Definition 1 includes as particular cases deterministic functions F : ∏ j = 1 n X j → V such that V t = F ( X t ), as well as features calculated under observational noise—e.g. V t = F ( X t ) + ν t , where ν t is independent of X t for all t . In contrast, features that are computed using the values of X t at multiple timepoints (e.g. the Fourier transform of X t ) generally fail to be supervenient. Partial information decomposition Our theory is based on the Partial Information Decomposition (PID) framework , which provides powerful tools to reason about information in multivariate systems. In a nutshell, PID decomposes the information that n sources X = ( X 1 , …, X n ) provide about a target variable Y in terms of information atoms as follows: I ( X ; Y ) = ∑ α ∈ A I ∂ α ( X ; Y ) , (1) with A = { { α 1 , … , α L } : α i ⊆ [ n ] , α i ¬ ⊄ α j ∀ i , j } being the set of antichain collections . Intuitively, I ∂ α for α = { α 1 , …, α L } represents the information that the collection of variables X α 1 , … , X α L provide redundantly, but their sub-collections don’t. For example, for n = 2 source variables, α = {{1}{2}} corresponds to the information about Y that is provided by both of them, α = {{ i }} to the information provided uniquely by X i , and, most interestingly, α = {{12}} corresponds to the information provided by both sources jointly but not separately—commonly referred to as informational synergy . One of the drawbacks of PID is that the number of atoms (i.e. the cardinality of A ) grows super-exponentially with the number of sources, and hence it is useful to coarse-grain the decomposition according to specific criteria. Here we introduce the notion of k th -order synergy between n variables, which is calculated as Syn ( k ) ( X ; Y ) ∑ α ∈ S ( k ) I ∂ α ( X ; Y ) , with S ( k ) = { { α 1 , … , α L } ∈ A : min j | α j | > k } . Intuitively, Syn ( k ) ( X ; Y ) corresponds to the information about the target that is provided by the whole X but is not contained in any set of k or less parts when considered separately from the rest. Accordingly, S ( k ) only contains collections with groups of more than k sources. Similarly, we introduce the unique information of X β with β ⊂ [ n ] with respect to sets of at most k other variables, which is calculated as Un ( k ) ( X β ; Y | X - β ) ∑ α ∈ U ( k ) ( β ) I ∂ α ( X ; Y ) . Above, U ( k ) ( β ) = { α ∈ A : β ∈ α , ∀ α ≠ β ∈ α , α ⊆ [ n ] \ β , | α | > k } , and X − β being all the variables in X whose indices are not in β . Put simply, Un ( k ) ( X t β ; Y | X - β ) represents the information carried by X β about Y that no group of k or less variables within X − β has on its own. Note that these coarse-grained terms can be used to build a general decomposition of I ( X , Y ) described in (Section 1), the properties of which are proven in (Section 2). One peculiarity of PID is that it postulates the structure of information atoms and the relations between them, but it does not prescribe a particular functional form to compute I ∂ α . In fact, only one of the information atoms must be specified to determine the whole PID—usually the redundancy between all individual elements . There have been multiple proposals for specific functional forms of I ∂ α in the PID literature; see e.g. Refs. . A particular method for fully computing the information atoms based on a recent PID is discussed in Section Measuring emergence via synergistic channels . Conveniently, our theory doesn’t rely on a specific functional form of PID, but only on a few basic properties that are precisely formulated in (Section 2). Therefore, the theory can be instantiated using any PID—as long as those properties are satisfied. Importantly, as shown in Section Practical criteria for large systems , the theory allows the derivation of practical metrics that are valid independently of the PID chosen. Defining causal emergence With the tools of PID at hand, now we introduce our formal definition of causal emergence. Definition 2 . For a system described by X t , a supervenient feature V t is said to exhibit causal emergence of order k if Un ( k ) ( V t ; X t ′ | X t ) > 0 . (2) Accordingly, causal emergence takes place when a supervenient feature V t has irreducible causal power, i.e. when it exerts causal influence that is not mediated by any of the parts of the system . In other words, V t represents some emergent collective property of the system if: 1) contains information that is dynamically relevant (in the sense that it predicts the future evolution of the system); and 2) this information is beyond what is given by the groups of k parts in the system when considered separately. To better understand the implications of this definition, let us study some of its basic properties. Lemma 1 . Consider a feature V t that exhibits causal emergence of order 1 over X t . Then , The dimensionality of the system satisfies n ≥ 2. There exists no deterministic function g (⋅) such that V t = g ( X t j ) for any j = 1, …, n . Proof . See , Section 3. These two properties establish causal emergence as a fundamentally collective phenomenon. In effect, property (i) states that causal emergence is a property of multivariate systems, and property (ii) that V t cannot have emergent behaviour if it can be perfectly predicted from a single variable. In order to use Definition 2, one needs a candidate feature V t to be tested. However, in some cases there are no obvious candidates for an emergent feature, for which Definition 2 might seem problematic. Our next result provides a criterion for the existence of emergent features based solely on the system’s dynamics. Theorem 1 . A system X t has a causally emergent feature of order k if and only if Syn ( k ) ( X t ; X t ′ ) > 0 . (3) Proof . See , Section 2. Corollary 1 . The following bound holds for any supervenient feature V t : Un ( k ) ( V t ; X t ′ | X t ) ≤ Syn ( k ) ( X t ; X t ′ ) . This result shows that the capability of exhibiting emergence is closely related to how synergistic the system components are with respect to their future evolution. Importantly, this result enables us to determine whether or not the system admits any emergent features by just inspecting the synergy between its parts— without knowing what those features might be . Conversely, this result also allows us to discard the existence of causal emergence by checking a single condition: the lack of dynamical synergy. Furthermore, Corollary 1 implies that the quantity Syn ( k ) ( X t ; X t ′ ) serves as a measure of the emergence capacity of the system, as it upper-bounds the unique information of all possible supervenient features. Theorem 1 establishes a direct link between causal emergence and the system’s statistics, avoiding the need for the observer to propose a particular feature of interest. It is important to remark that the emergence capacity of a system depends on the system’s partition into microscopic elements—in fact, it is plausible that a system might have emergence capacity under one microscopic representation, but not with respect to another after a change of variables. Therefore, emergence in the context of our theory always refers to “emergence with respect to a given microscopic partition.” A taxonomy of emergence Our theory, so far, is able to detect whether there is emergence taking place; the next step is to be able to characterise which kind of emergence it is. For this purpose, we combine our feature-agnostic criterion of emergence presented in Theorem 1 with Integrated Information Decomposition, ΦID, a recent extension of PID to multi-target settings . Using ΦID, one can decompose a PID atom as I ∂ α ( X t ; X t ′ ) = ∑ β ∈ A I ∂ α → β ( X t ; X t ′ ) . (4) For example, if n = 2 then I ∂ { 1 } { 2 } → { 1 } { 2 } represents the information shared by both time series at both timesteps (for example, when X t 1 , X t 2 , X t ′ 1 , X t ′ 2 are all copies of each other); and I ∂ { 12 } → { 1 } corresponds to the synergistic causes in X t that have a unique effect on X t ′ 1 (for example, when X t ′ 1 = X t 1 ⊕ X t 2 ). More details and intuitions on ΦID can be found in Ref. . With the fine-grained decomposition provided by ΦID one can discriminate between different kinds of synergies. In particular, we introduce the downward causation and causal decoupling indices of order k , denoted by D ( k ) and G ( k ) respectively, as D ( k ) ( X t ; X t ′ ) ≔ ∑ α ∈ S ( k ) β ∈ A \ S ( k ) I ∂ α → β ( X t ; X t ′ ) , (5) G ( k ) ( X t ; X t ′ ) ≔ ∑ α , β ∈ S ( k ) I ∂ α → β ( X t ; X t ′ ) . (6) From these definitions and , one can verify that Syn ( k ) ( X t ; X t ′ ) = G ( k ) ( X t ; X t ′ ) + D ( k ) ( X t ; X t ′ ) . (7) Therefore, the emergence capacity of a system naturally decomposes in two different components: information about k -plets of future variables, and information about future collective properties beyond k -plets. The ΦID atoms that belong to these two terms are illustrated within the ΦID lattice for two time series in . The rest of this section shows that D ( k ) and G ( k ) are natural metrics of downward causation and causal decoupling, respectively. Downward causation Intuitively, downward causation occurs when collective properties have irreducible causal power over individual parts. More formally: Definition 3 . A supervenient feature V t exhibits downward causation of order k if, for some α with | α | = k : Un ( k ) ( V t ; X t ′ α | X t ) > 0 . (8) Note that, in contrast with Definition 2, downward causation requires the feature V t to have unique predictive power over the evolution of specific subsets of the whole system. In particular, an emergent feature V t that has predictive power over e.g. X t ′ j is said to exert downward causation, as it predicts something about X t ′ j that could not be predicted from any particular X t i for i ∈ [ n ]. Put differently, in a system with downward causation the whole has an effect on the parts that cannot be reduced to low-level interactions. A minimal case of this is provided by Example 2 in Section Fundamental intuitions . Our next result formally relates downward causation with the index D ( k ) introduced in . Theorem 2 . A system X t admits features that exert downward causation of order k iff D ( k ) ( X t ; X t ′ ) > 0 . Proof . See , Section 3. Causal decoupling In addition to downward causation, causal decoupling takes place when collective properties have irreducible causal power over other collective properties. In technical terms: Definition 4 . A supervenient feature V t is said to exhibit causal decoupling of order k if Un ( k ) ( V t ; V t ′ | X t , X t ′ ) > 0 . (9) Furthermore, V t is said to have pure causal decoupling if Un ( k ) ( V t ; X t ′ | X t ) > 0 and Un ( k ) ( V t ; X t ′ α | X t ) = 0 for all α ⊂ [ n ] with | α | = k . Finally, a system is said to be perfectly decoupled if all the emergent features exhibit pure causal decoupling . Above, the term Un ( k ) ( V t ; V t ′ | X t , X t ′ ) refers to information that V t and V t ′ share that cannot be found in any microscopic element, either at time t or t ′ (note that Un ( k ) ( V t ; V t ′ | X t , X t ′ ) is information shared between V t and V t ′ that no combination of k or less variables from X t or X t ′ has in its own). Features that exhibit causal decoupling could still exert influence over the evolution of individual elements, while features that exhibit pure decoupling cannot. In effect, the condition Un ( V t ; X t ′ j | X t ) = 0 implies that the high-order causal effect does not affect any particular part – only the system as a whole. Interestingly, a feature that exhibits pure causal decoupling can be thought of as having “a life of its own;” a sort of statistical ghost , that perpetuates itself over time without any individual part of the system influencing or being influenced by it. The system’s parity, in the first example of Section Fundamental intuitions , constitutes a simple example of perfect causal decoupling. Importantly, the case studies presented in Section Case studies show that causal decoupling can take place not only in toy models but also in diverse scenarios of practical relevance. We close this section by formally establishing the connection between causal decoupling and the index G ( k ) introduced in . Theorem 3 . A system possesses features that exhibit causal decoupling if and only if G ( k ) ( X t ; X t ′ ) > 0 . Additionally, the system is perfectly decoupled if G ( k ) ( X t ; X t ′ ) > 0 and D ( k ) ( X t ; X t ′ ) = 0 . Proof . See , Section 3.
Our analysis considers two time points of the evolution of the system, denoted as t and t ′, with t < t ′. The corresponding dynamics are encoded in the transition probability p X t ′ | X t ( x t ′ | x t ) . We consider features V t ∈ V generated via a conditional probability p V t | X t that are supervenient on the underlying system; i.e. that does not provide any predictive power for future states at times t ′ > t if the complete state of the system at time t is known with perfect precision. We formalise this in the following definition. Definition 1 . A stochastic process V t is said to be supervenient over X t if V t − X t − X t ′′ form a Markov chain for all t ′′ ≠ t . The above condition is equivalent to require V t to be statistically independent of X t ′′ when X t is given. The relationship between supervenient features and the underlying system is illustrated in . This formalisation of supervenience characterises features V t that are fully determined by the state of the system at a given time t , but also allows the feature to be noisy—which is not critical for our results, but is useful for extending their domain of applicability to practical scenarios. In effect, Definition 1 includes as particular cases deterministic functions F : ∏ j = 1 n X j → V such that V t = F ( X t ), as well as features calculated under observational noise—e.g. V t = F ( X t ) + ν t , where ν t is independent of X t for all t . In contrast, features that are computed using the values of X t at multiple timepoints (e.g. the Fourier transform of X t ) generally fail to be supervenient.
Our theory is based on the Partial Information Decomposition (PID) framework , which provides powerful tools to reason about information in multivariate systems. In a nutshell, PID decomposes the information that n sources X = ( X 1 , …, X n ) provide about a target variable Y in terms of information atoms as follows: I ( X ; Y ) = ∑ α ∈ A I ∂ α ( X ; Y ) , (1) with A = { { α 1 , … , α L } : α i ⊆ [ n ] , α i ¬ ⊄ α j ∀ i , j } being the set of antichain collections . Intuitively, I ∂ α for α = { α 1 , …, α L } represents the information that the collection of variables X α 1 , … , X α L provide redundantly, but their sub-collections don’t. For example, for n = 2 source variables, α = {{1}{2}} corresponds to the information about Y that is provided by both of them, α = {{ i }} to the information provided uniquely by X i , and, most interestingly, α = {{12}} corresponds to the information provided by both sources jointly but not separately—commonly referred to as informational synergy . One of the drawbacks of PID is that the number of atoms (i.e. the cardinality of A ) grows super-exponentially with the number of sources, and hence it is useful to coarse-grain the decomposition according to specific criteria. Here we introduce the notion of k th -order synergy between n variables, which is calculated as Syn ( k ) ( X ; Y ) ∑ α ∈ S ( k ) I ∂ α ( X ; Y ) , with S ( k ) = { { α 1 , … , α L } ∈ A : min j | α j | > k } . Intuitively, Syn ( k ) ( X ; Y ) corresponds to the information about the target that is provided by the whole X but is not contained in any set of k or less parts when considered separately from the rest. Accordingly, S ( k ) only contains collections with groups of more than k sources. Similarly, we introduce the unique information of X β with β ⊂ [ n ] with respect to sets of at most k other variables, which is calculated as Un ( k ) ( X β ; Y | X - β ) ∑ α ∈ U ( k ) ( β ) I ∂ α ( X ; Y ) . Above, U ( k ) ( β ) = { α ∈ A : β ∈ α , ∀ α ≠ β ∈ α , α ⊆ [ n ] \ β , | α | > k } , and X − β being all the variables in X whose indices are not in β . Put simply, Un ( k ) ( X t β ; Y | X - β ) represents the information carried by X β about Y that no group of k or less variables within X − β has on its own. Note that these coarse-grained terms can be used to build a general decomposition of I ( X , Y ) described in (Section 1), the properties of which are proven in (Section 2). One peculiarity of PID is that it postulates the structure of information atoms and the relations between them, but it does not prescribe a particular functional form to compute I ∂ α . In fact, only one of the information atoms must be specified to determine the whole PID—usually the redundancy between all individual elements . There have been multiple proposals for specific functional forms of I ∂ α in the PID literature; see e.g. Refs. . A particular method for fully computing the information atoms based on a recent PID is discussed in Section Measuring emergence via synergistic channels . Conveniently, our theory doesn’t rely on a specific functional form of PID, but only on a few basic properties that are precisely formulated in (Section 2). Therefore, the theory can be instantiated using any PID—as long as those properties are satisfied. Importantly, as shown in Section Practical criteria for large systems , the theory allows the derivation of practical metrics that are valid independently of the PID chosen.
With the tools of PID at hand, now we introduce our formal definition of causal emergence. Definition 2 . For a system described by X t , a supervenient feature V t is said to exhibit causal emergence of order k if Un ( k ) ( V t ; X t ′ | X t ) > 0 . (2) Accordingly, causal emergence takes place when a supervenient feature V t has irreducible causal power, i.e. when it exerts causal influence that is not mediated by any of the parts of the system . In other words, V t represents some emergent collective property of the system if: 1) contains information that is dynamically relevant (in the sense that it predicts the future evolution of the system); and 2) this information is beyond what is given by the groups of k parts in the system when considered separately. To better understand the implications of this definition, let us study some of its basic properties. Lemma 1 . Consider a feature V t that exhibits causal emergence of order 1 over X t . Then , The dimensionality of the system satisfies n ≥ 2. There exists no deterministic function g (⋅) such that V t = g ( X t j ) for any j = 1, …, n . Proof . See , Section 3. These two properties establish causal emergence as a fundamentally collective phenomenon. In effect, property (i) states that causal emergence is a property of multivariate systems, and property (ii) that V t cannot have emergent behaviour if it can be perfectly predicted from a single variable. In order to use Definition 2, one needs a candidate feature V t to be tested. However, in some cases there are no obvious candidates for an emergent feature, for which Definition 2 might seem problematic. Our next result provides a criterion for the existence of emergent features based solely on the system’s dynamics. Theorem 1 . A system X t has a causally emergent feature of order k if and only if Syn ( k ) ( X t ; X t ′ ) > 0 . (3) Proof . See , Section 2. Corollary 1 . The following bound holds for any supervenient feature V t : Un ( k ) ( V t ; X t ′ | X t ) ≤ Syn ( k ) ( X t ; X t ′ ) . This result shows that the capability of exhibiting emergence is closely related to how synergistic the system components are with respect to their future evolution. Importantly, this result enables us to determine whether or not the system admits any emergent features by just inspecting the synergy between its parts— without knowing what those features might be . Conversely, this result also allows us to discard the existence of causal emergence by checking a single condition: the lack of dynamical synergy. Furthermore, Corollary 1 implies that the quantity Syn ( k ) ( X t ; X t ′ ) serves as a measure of the emergence capacity of the system, as it upper-bounds the unique information of all possible supervenient features. Theorem 1 establishes a direct link between causal emergence and the system’s statistics, avoiding the need for the observer to propose a particular feature of interest. It is important to remark that the emergence capacity of a system depends on the system’s partition into microscopic elements—in fact, it is plausible that a system might have emergence capacity under one microscopic representation, but not with respect to another after a change of variables. Therefore, emergence in the context of our theory always refers to “emergence with respect to a given microscopic partition.”
Our theory, so far, is able to detect whether there is emergence taking place; the next step is to be able to characterise which kind of emergence it is. For this purpose, we combine our feature-agnostic criterion of emergence presented in Theorem 1 with Integrated Information Decomposition, ΦID, a recent extension of PID to multi-target settings . Using ΦID, one can decompose a PID atom as I ∂ α ( X t ; X t ′ ) = ∑ β ∈ A I ∂ α → β ( X t ; X t ′ ) . (4) For example, if n = 2 then I ∂ { 1 } { 2 } → { 1 } { 2 } represents the information shared by both time series at both timesteps (for example, when X t 1 , X t 2 , X t ′ 1 , X t ′ 2 are all copies of each other); and I ∂ { 12 } → { 1 } corresponds to the synergistic causes in X t that have a unique effect on X t ′ 1 (for example, when X t ′ 1 = X t 1 ⊕ X t 2 ). More details and intuitions on ΦID can be found in Ref. . With the fine-grained decomposition provided by ΦID one can discriminate between different kinds of synergies. In particular, we introduce the downward causation and causal decoupling indices of order k , denoted by D ( k ) and G ( k ) respectively, as D ( k ) ( X t ; X t ′ ) ≔ ∑ α ∈ S ( k ) β ∈ A \ S ( k ) I ∂ α → β ( X t ; X t ′ ) , (5) G ( k ) ( X t ; X t ′ ) ≔ ∑ α , β ∈ S ( k ) I ∂ α → β ( X t ; X t ′ ) . (6) From these definitions and , one can verify that Syn ( k ) ( X t ; X t ′ ) = G ( k ) ( X t ; X t ′ ) + D ( k ) ( X t ; X t ′ ) . (7) Therefore, the emergence capacity of a system naturally decomposes in two different components: information about k -plets of future variables, and information about future collective properties beyond k -plets. The ΦID atoms that belong to these two terms are illustrated within the ΦID lattice for two time series in . The rest of this section shows that D ( k ) and G ( k ) are natural metrics of downward causation and causal decoupling, respectively. Downward causation Intuitively, downward causation occurs when collective properties have irreducible causal power over individual parts. More formally: Definition 3 . A supervenient feature V t exhibits downward causation of order k if, for some α with | α | = k : Un ( k ) ( V t ; X t ′ α | X t ) > 0 . (8) Note that, in contrast with Definition 2, downward causation requires the feature V t to have unique predictive power over the evolution of specific subsets of the whole system. In particular, an emergent feature V t that has predictive power over e.g. X t ′ j is said to exert downward causation, as it predicts something about X t ′ j that could not be predicted from any particular X t i for i ∈ [ n ]. Put differently, in a system with downward causation the whole has an effect on the parts that cannot be reduced to low-level interactions. A minimal case of this is provided by Example 2 in Section Fundamental intuitions . Our next result formally relates downward causation with the index D ( k ) introduced in . Theorem 2 . A system X t admits features that exert downward causation of order k iff D ( k ) ( X t ; X t ′ ) > 0 . Proof . See , Section 3. Causal decoupling In addition to downward causation, causal decoupling takes place when collective properties have irreducible causal power over other collective properties. In technical terms: Definition 4 . A supervenient feature V t is said to exhibit causal decoupling of order k if Un ( k ) ( V t ; V t ′ | X t , X t ′ ) > 0 . (9) Furthermore, V t is said to have pure causal decoupling if Un ( k ) ( V t ; X t ′ | X t ) > 0 and Un ( k ) ( V t ; X t ′ α | X t ) = 0 for all α ⊂ [ n ] with | α | = k . Finally, a system is said to be perfectly decoupled if all the emergent features exhibit pure causal decoupling . Above, the term Un ( k ) ( V t ; V t ′ | X t , X t ′ ) refers to information that V t and V t ′ share that cannot be found in any microscopic element, either at time t or t ′ (note that Un ( k ) ( V t ; V t ′ | X t , X t ′ ) is information shared between V t and V t ′ that no combination of k or less variables from X t or X t ′ has in its own). Features that exhibit causal decoupling could still exert influence over the evolution of individual elements, while features that exhibit pure decoupling cannot. In effect, the condition Un ( V t ; X t ′ j | X t ) = 0 implies that the high-order causal effect does not affect any particular part – only the system as a whole. Interestingly, a feature that exhibits pure causal decoupling can be thought of as having “a life of its own;” a sort of statistical ghost , that perpetuates itself over time without any individual part of the system influencing or being influenced by it. The system’s parity, in the first example of Section Fundamental intuitions , constitutes a simple example of perfect causal decoupling. Importantly, the case studies presented in Section Case studies show that causal decoupling can take place not only in toy models but also in diverse scenarios of practical relevance. We close this section by formally establishing the connection between causal decoupling and the index G ( k ) introduced in . Theorem 3 . A system possesses features that exhibit causal decoupling if and only if G ( k ) ( X t ; X t ′ ) > 0 . Additionally, the system is perfectly decoupled if G ( k ) ( X t ; X t ′ ) > 0 and D ( k ) ( X t ; X t ′ ) = 0 . Proof . See , Section 3.
Intuitively, downward causation occurs when collective properties have irreducible causal power over individual parts. More formally: Definition 3 . A supervenient feature V t exhibits downward causation of order k if, for some α with | α | = k : Un ( k ) ( V t ; X t ′ α | X t ) > 0 . (8) Note that, in contrast with Definition 2, downward causation requires the feature V t to have unique predictive power over the evolution of specific subsets of the whole system. In particular, an emergent feature V t that has predictive power over e.g. X t ′ j is said to exert downward causation, as it predicts something about X t ′ j that could not be predicted from any particular X t i for i ∈ [ n ]. Put differently, in a system with downward causation the whole has an effect on the parts that cannot be reduced to low-level interactions. A minimal case of this is provided by Example 2 in Section Fundamental intuitions . Our next result formally relates downward causation with the index D ( k ) introduced in . Theorem 2 . A system X t admits features that exert downward causation of order k iff D ( k ) ( X t ; X t ′ ) > 0 . Proof . See , Section 3.
In addition to downward causation, causal decoupling takes place when collective properties have irreducible causal power over other collective properties. In technical terms: Definition 4 . A supervenient feature V t is said to exhibit causal decoupling of order k if Un ( k ) ( V t ; V t ′ | X t , X t ′ ) > 0 . (9) Furthermore, V t is said to have pure causal decoupling if Un ( k ) ( V t ; X t ′ | X t ) > 0 and Un ( k ) ( V t ; X t ′ α | X t ) = 0 for all α ⊂ [ n ] with | α | = k . Finally, a system is said to be perfectly decoupled if all the emergent features exhibit pure causal decoupling . Above, the term Un ( k ) ( V t ; V t ′ | X t , X t ′ ) refers to information that V t and V t ′ share that cannot be found in any microscopic element, either at time t or t ′ (note that Un ( k ) ( V t ; V t ′ | X t , X t ′ ) is information shared between V t and V t ′ that no combination of k or less variables from X t or X t ′ has in its own). Features that exhibit causal decoupling could still exert influence over the evolution of individual elements, while features that exhibit pure decoupling cannot. In effect, the condition Un ( V t ; X t ′ j | X t ) = 0 implies that the high-order causal effect does not affect any particular part – only the system as a whole. Interestingly, a feature that exhibits pure causal decoupling can be thought of as having “a life of its own;” a sort of statistical ghost , that perpetuates itself over time without any individual part of the system influencing or being influenced by it. The system’s parity, in the first example of Section Fundamental intuitions , constitutes a simple example of perfect causal decoupling. Importantly, the case studies presented in Section Case studies show that causal decoupling can take place not only in toy models but also in diverse scenarios of practical relevance. We close this section by formally establishing the connection between causal decoupling and the index G ( k ) introduced in . Theorem 3 . A system possesses features that exhibit causal decoupling if and only if G ( k ) ( X t ; X t ′ ) > 0 . Additionally, the system is perfectly decoupled if G ( k ) ( X t ; X t ′ ) > 0 and D ( k ) ( X t ; X t ′ ) = 0 . Proof . See , Section 3.
This section explores methods to operationalise the framework presented in the previous section. We discuss two approaches: first, Section Practical criteria for large systems introduces sufficiency criteria that are practical for use in large systems; then, Section Measuring emergence via synergistic channels illustrates how further considerations can be made if one adopts a specific method of computing ΦID atoms. The latter approach provides accurate discrimination at the cost of being data-intensive and hence only applicable to small systems; the former can be computed in large systems and its results hold independently of the chosen PID, but is vulnerable to misdetections (i.e. false negatives). Practical criteria for large systems While theoretically appealing, our proposed framework suffers from the challenge of estimating joint probability distributions over many random variables, and the computation of the ΦID atoms themselves. As an alternative, we consider approximation techniques that do not require the adoption of any particular PID or ΦID function and are data-efficient, since they are based on pairwise distributions only. As practical criteria to measure causal emergence of order k , we introduce the quantities Ψ t , t ′ ( k ) , Δ t , t ′ ( k ) , and Γ t , t ′ ( k ) . For simplicity, we write here the special case k = 1, and provide full formulae for arbitrary k and accompanying proofs in , Section 4: Ψ t , t ′ ( 1 ) ( V ) ≔ I ( V t ; V t ′ ) - ∑ j I ( X t j ; V t ′ ) , (10a) Δ t , t ′ ( 1 ) ( V ) ≔ max j ( I ( V t ; X t ′ j ) - ∑ i I ( X t i ; X t ′ j ) ) , (10b) Γ t , t ′ ( 1 ) ( V ) ≔ max j I ( V t ; X t ′ j ) . (10c) Our next result links these quantities with the formal definitions in Section A formal theory of causal emergence , showing their value as practical criteria to detect causal emergence. Proposition 1 . Ψ t , t ′ ( k ) ( V ) > 0 is a sufficient condition for V t to be causally emergent. Similarly , Δ t , t ′ ( k ) ( V ) > 0 is a sufficient condition for V t to exhibit downward causation. Finally , Ψ t , t ′ ( k ) ( V ) > 0 and Γ t , t ′ ( k ) ( V ) = 0 is sufficient for causal decoupling . Proof . See , Section 4. Although calculating whether a system has emergent features via Proposition 1 may be computationally challenging, if one has a candidate feature V one believes may be emergent, one can compute the simple quantities in which depend only on standard mutual information and bivariate marginals, and scales linearly with system size (for k = 1). These quantities are easy to compute and test for significance using standard information-theoretic tools . Moreover, the outcome of these measures is valid for any choice of PID and ΦID that is compatible with the properties specified in , Section 2. In a broader context, Ψ t , t ′ ( k ) and Δ t , t ′ ( k ) belong to the same whole-minus-sum family of measures as the interaction information , the redundancy-synergy index and, more recently, the O-information Ω —which cannot measure synergy by itself, but only the balance between synergy and redundancy. In practice, this means that if there is redundancy in the system it will be harder to detect emergence, since redundancy will drive Ψ t , t ′ ( k ) and Δ t , t ′ ( k ) more negative. Furthermore, by summing all marginal mutual informations (e.g. I ( X t j ; V t ′ ) in the case of Ψ t , t ′ ( 1 ) ), these measures effectively double-count redundancy up to n times, further penalising the criteria. This problem of double-counting can be avoided if one is willing to commit to a particular PID or ΦID function, as we show next. It is worth noticing that the value of k can be tuned to explore emergence with respect to different “scales.” For example, k = 1 corresponds to emergence with respect to individual microscopic elements, while k = 2 refers to emergence with respect to all couples—i.e. individual elements and their pairwise interactions. Accordingly, the criteria in Proposition 1 are, in general, harder to satisfy for larger values of k . In addition, from a practical perspective, considering large values of k requires estimating information-theoretic quantities in high-dimensional distributions, which usually requires exponentially larger amounts of data. Measuring emergence via synergistic channels This section leverages recent work on information decomposition reported in Ref. , and presents a way of directly measuring the emergence capacity and the indices of downward causation and causal decoupling. The key takeaway of this section is that if one adopts a particular ΦID, then it is possible to evaluate D ( k ) and G ( k ) directly, providing a direct route to detect emergence without double-counting redundancy, as the methods introduced in Section Practical criteria for large systems do. Moreover, additional properties may become available due to the characteristics of the particular ΦID chosen. Let us first introduce the notion of k -synergistic channels: mappings p V | X that convey information about X but not about any of the parts X α for all | α | = k . The set of all k -synergistic channels is denoted by C k ( X ) = { p V | X | V ⫫ X α , ∀ α ⊆ [ n ] , | α | = k } . (11) A variable V generated via a k -synergistic channel is said to be a k -synergistic observable. With this definition, we can consider the k th -order synergy to be the maximum information extractable from a k -synergistic channel: Syn ⋆ ( k ) ( X t ; X t ′ ) ≔ sup p V | X t ∈ C k ( X t ) : V - X t - X t ′ I ( V ; X t ′ ) . (12) This idea can be naturally extended to the case of causal decoupling by requiring synergistic channels at both sides, i.e. G ⋆ ( k ) ( X t ; X t ′ ) ≔ sup p V | X t ∈ C k ( X t ) , p U | X t ′ ∈ C k ( X t ′ ) : V - X t - X t ′ - U I ( V ; U ) . (13) Finally, the downward causation index can be computed from the difference D ⋆ ( k ) ( X t ; X t ′ ) ≔ Syn ⋆ ( k ) ( X t ; X t ′ ) - G ⋆ ( k ) ( X t ; X t ′ ) . (14) Note that Syn ⋆ ( k ) ≥ G ⋆ ( k ) , which is a direct consequence of the data processing inequality applied on V − X t − X t ′ − U , and therefore D ⋆ ( k ) , G ⋆ ( k ) ≥ 0 . By exploiting the properties of this specific way of measuring synergy, one can prove the following result. For this, let us say that a feature V t is auto-correlated if I ( V t ; V t ′ )>0. Proposition 2 . If X t is stationary, all auto-correlated k-synergistic observables are k th -order emergent . Proof . See , Section 4. In summary, D ⋆ ( k ) and G ⋆ ( k ) provide data-driven tools to test—and possibly reject—hypotheses about emergence in scenarios of interest. Efficient algorithms to compute these quantities are discussed in Ref. . Although current implementations allow only relatively small systems, this line of thinking shows that future advances in PID might make the computation of emergence indices more scalable, avoiding the limitations of .
While theoretically appealing, our proposed framework suffers from the challenge of estimating joint probability distributions over many random variables, and the computation of the ΦID atoms themselves. As an alternative, we consider approximation techniques that do not require the adoption of any particular PID or ΦID function and are data-efficient, since they are based on pairwise distributions only. As practical criteria to measure causal emergence of order k , we introduce the quantities Ψ t , t ′ ( k ) , Δ t , t ′ ( k ) , and Γ t , t ′ ( k ) . For simplicity, we write here the special case k = 1, and provide full formulae for arbitrary k and accompanying proofs in , Section 4: Ψ t , t ′ ( 1 ) ( V ) ≔ I ( V t ; V t ′ ) - ∑ j I ( X t j ; V t ′ ) , (10a) Δ t , t ′ ( 1 ) ( V ) ≔ max j ( I ( V t ; X t ′ j ) - ∑ i I ( X t i ; X t ′ j ) ) , (10b) Γ t , t ′ ( 1 ) ( V ) ≔ max j I ( V t ; X t ′ j ) . (10c) Our next result links these quantities with the formal definitions in Section A formal theory of causal emergence , showing their value as practical criteria to detect causal emergence. Proposition 1 . Ψ t , t ′ ( k ) ( V ) > 0 is a sufficient condition for V t to be causally emergent. Similarly , Δ t , t ′ ( k ) ( V ) > 0 is a sufficient condition for V t to exhibit downward causation. Finally , Ψ t , t ′ ( k ) ( V ) > 0 and Γ t , t ′ ( k ) ( V ) = 0 is sufficient for causal decoupling . Proof . See , Section 4. Although calculating whether a system has emergent features via Proposition 1 may be computationally challenging, if one has a candidate feature V one believes may be emergent, one can compute the simple quantities in which depend only on standard mutual information and bivariate marginals, and scales linearly with system size (for k = 1). These quantities are easy to compute and test for significance using standard information-theoretic tools . Moreover, the outcome of these measures is valid for any choice of PID and ΦID that is compatible with the properties specified in , Section 2. In a broader context, Ψ t , t ′ ( k ) and Δ t , t ′ ( k ) belong to the same whole-minus-sum family of measures as the interaction information , the redundancy-synergy index and, more recently, the O-information Ω —which cannot measure synergy by itself, but only the balance between synergy and redundancy. In practice, this means that if there is redundancy in the system it will be harder to detect emergence, since redundancy will drive Ψ t , t ′ ( k ) and Δ t , t ′ ( k ) more negative. Furthermore, by summing all marginal mutual informations (e.g. I ( X t j ; V t ′ ) in the case of Ψ t , t ′ ( 1 ) ), these measures effectively double-count redundancy up to n times, further penalising the criteria. This problem of double-counting can be avoided if one is willing to commit to a particular PID or ΦID function, as we show next. It is worth noticing that the value of k can be tuned to explore emergence with respect to different “scales.” For example, k = 1 corresponds to emergence with respect to individual microscopic elements, while k = 2 refers to emergence with respect to all couples—i.e. individual elements and their pairwise interactions. Accordingly, the criteria in Proposition 1 are, in general, harder to satisfy for larger values of k . In addition, from a practical perspective, considering large values of k requires estimating information-theoretic quantities in high-dimensional distributions, which usually requires exponentially larger amounts of data.
This section leverages recent work on information decomposition reported in Ref. , and presents a way of directly measuring the emergence capacity and the indices of downward causation and causal decoupling. The key takeaway of this section is that if one adopts a particular ΦID, then it is possible to evaluate D ( k ) and G ( k ) directly, providing a direct route to detect emergence without double-counting redundancy, as the methods introduced in Section Practical criteria for large systems do. Moreover, additional properties may become available due to the characteristics of the particular ΦID chosen. Let us first introduce the notion of k -synergistic channels: mappings p V | X that convey information about X but not about any of the parts X α for all | α | = k . The set of all k -synergistic channels is denoted by C k ( X ) = { p V | X | V ⫫ X α , ∀ α ⊆ [ n ] , | α | = k } . (11) A variable V generated via a k -synergistic channel is said to be a k -synergistic observable. With this definition, we can consider the k th -order synergy to be the maximum information extractable from a k -synergistic channel: Syn ⋆ ( k ) ( X t ; X t ′ ) ≔ sup p V | X t ∈ C k ( X t ) : V - X t - X t ′ I ( V ; X t ′ ) . (12) This idea can be naturally extended to the case of causal decoupling by requiring synergistic channels at both sides, i.e. G ⋆ ( k ) ( X t ; X t ′ ) ≔ sup p V | X t ∈ C k ( X t ) , p U | X t ′ ∈ C k ( X t ′ ) : V - X t - X t ′ - U I ( V ; U ) . (13) Finally, the downward causation index can be computed from the difference D ⋆ ( k ) ( X t ; X t ′ ) ≔ Syn ⋆ ( k ) ( X t ; X t ′ ) - G ⋆ ( k ) ( X t ; X t ′ ) . (14) Note that Syn ⋆ ( k ) ≥ G ⋆ ( k ) , which is a direct consequence of the data processing inequality applied on V − X t − X t ′ − U , and therefore D ⋆ ( k ) , G ⋆ ( k ) ≥ 0 . By exploiting the properties of this specific way of measuring synergy, one can prove the following result. For this, let us say that a feature V t is auto-correlated if I ( V t ; V t ′ )>0. Proposition 2 . If X t is stationary, all auto-correlated k-synergistic observables are k th -order emergent . Proof . See , Section 4. In summary, D ⋆ ( k ) and G ⋆ ( k ) provide data-driven tools to test—and possibly reject—hypotheses about emergence in scenarios of interest. Efficient algorithms to compute these quantities are discussed in Ref. . Although current implementations allow only relatively small systems, this line of thinking shows that future advances in PID might make the computation of emergence indices more scalable, avoiding the limitations of .
Let us summarise our results so far. We began by formulating a rigorous definition of emergent features based on PID (Section Defining causal emergence ), and then used ΦID to break down the emergence capacity into the causal decoupling and downward causation indices (Section A taxonomy of emergence ). Although these are not straightforward to compute, the ΦID framework allows us to formulate readily computable sufficiency conditions (Section Practical criteria for large systems ). This section illustrates the usage of those conditions in various case studies. Code to compute all emergence criteria in is provided in an online open-source repository ( https://github.com/pmediano/ReconcilingEmergences ). Canonical examples of putative emergence Here we present an evaluation of our practical criteria for emergence (Proposition 1) in two well-known systems: Conway’s Game of Life (GoL) , and Reynolds’ flocking boids model . Both are widely regarded as paradigmatic examples of emergent behaviour, and have been thoughtfully studied in the complexity and artificial life literature . Accordingly, we use these models as test cases for our methods. Technical details of the simulations are provided in , Section 5. Conway’s Game of Life A well-known feature of GoL is the presence of particles : coherent, self-sustaining structures known to be responsible for information transfer and modification . These particles have been the object of extensive study, and detailed taxonomies and classifications exist . To test the emergent properties of particles, we simulate the evolution of 15x15 square cell arrays, which we regard as a binary vector X t ∈ {0, 1} n with n = 225. As initial condition, we consider configurations that correspond to a “particle collider” setting, with two particles of known type facing each other . In each trial, the system is randomised by changing the position, type, and relative displacement of the particles. After an intial configuration has been selected, the well-known GoL evolution rule is applied 1000 times, leading to a final state X t ′ . Simulations showed that this interval is enough for the system to settle in a stable state after the collision. To use the criteria from , we need to choose a candidate emergent feature V t . In this case, we consider a symbolic, discrete-valued vector that encodes the type of particle(s) present in the board. Specifically, we consider V t = ( V t 1 , … , V t L ) , where V t j = 1 iff there is a particle of type j at time t —regardless of its position or orientation. With these variables, we compute the quantities in using Bayesian estimators of mutual information . The result is that, as expected, the criterion for causal emergence is met with Ψ t , t ′ ( 1 ) ( V ) = 0 . 58 ± 0 . 02 . Furthermore, we found that Γ t , t ′ ( 1 ) ( V ) = 0 . 009 ± 0 . 0002 , which is orders of magnitude smaller than I ( V t ; V t ′ ) = 0.99±0.02. Errors represent the standard deviation over surrogate data, as described in , Section 5. Using Proposition 1, these two results suggest that particle dynamics in GoL may not only be emergent, but causally decoupled with respect to their substrate. Reynolds’ flocking model As a second test case, we consider Reynolds’ model of flocking behaviour. This model is composed by boids (bird-oid objects), with each boid represented by three numbers: its position in 2D space and its heading angle. As candidate feature for emergence, we use the 2D coordinates of the center of mass of the flock, following Seth . In this model boids interact with one another following three rules, each regulated by a scalar parameter : aggregation ( a 1 ), as they fly towards the center of the flock; avoidance ( a 2 ), as they fly away from their closest neighbour; and alignment ( a 3 ), as they align their flight direction to that of their neighbours. Following Ref. , we study small flocks of N = 10 boids with different parameter settings to showcase some properties of our practical criterion of emergence. Note that this study is meant as an illustration of the proposed theory, and not as a thorough exploration of the flocking model, for which a vast literature exists (see e.g. the work of Vicsek and references therein). shows the results of a parameter sweep over the avoidance parameter, a 2 , while keeping a 1 and a 3 fixed. When there is no avoidance, boids orbit around a slowly-moving center of mass, in what could be called an ordered regime. Conversely, for high values of a 2 neighbour repulsion is too strong for lasting flocks to form, and isolated boids spread across the space avoiding one another. For intermediate values, the center of mass traces a smooth trajectory, as flocks form and disintegrate. In line with the findings of Seth , our criterion indicates that the flock exhibits causally emergent behaviour in this intermediate range. By studying separately the two terms that make up Ψ we found that the criterion of emergence fails for both low and high a 2 , but for different reasons (see ). In effect, for high a 2 the self-predictability of the center of mass (i.e. I ( V t ; V t ′ )) is low; while for low a 2 it is high, yet lower than the mutual information from individual boids (i.e. ∑ i I ( X t i ; V t ′ ) ). These results suggest that the low-avoidance scenario is dominated not by a reduction in synergy, but by an increase in redundancy, which effectively increases the synergy threshold needed to detect emergence. However, note that, due to the limitations of the criterion, the fact that Ψ t , t ′ ( 1 ) < 0 is inconclusive and does not rule out the possibility of emergence. This is a common limitation of whole-minus-sum estimators like Ψ ; further refinements may provide bounds that are less susceptible to these issues and perform accurately in these scenarios. Mind from matter: Emergence, behaviour, and neural dynamics A tantalising outcome of having a formal theory of emergence is the capability of bringing a quantitative angle to the archetype of emergence: the mind-matter relationship . As a first step in this direction, we conclude this section with an application of our emergence criteria to neurophysiological data. We study simultaneous electrocorticogram (ECoG) and motion capture (MoCap) data of Japanese macaques performing a reaching task , obtained from the online Neurotycho database. Note that the MoCap data cannot be assumed to be a supervenient feature of the available ECoG data, since it doesn’t satisfy the conditional independence conditions required by our definition of supervenience (see Section A formal theory of causal emergence ). This is likely to be the case, because the neural system is only partially observed—i.e. the ECoG does not capture every source of relevant activity in the macaque’s cortex. Note that non-supervenient features are of limited interest within our framework, as they can satisfy Proposition 1 in trivial ways (e.g. time series which are independent of the underlying system satisty Ψ > 0 if they are auto-correlated). Instead, we focus on the portion of neural activity encoded in the ECoG signal that is relevant to predict the macaque’s behaviour, and conjecture this information to be an emergent feature of the underlying neural activity . To test this hypothesis, we take the neural activity (as measured by 64 ECoG channels distributed across the left hemisphere) to be the system of interest, and consider a memoryless predictor of the 3D coordinates of the macaque’s right wrist based on the ECoG signal. Therefore, in this scenario X t ∈ R 64 and V t = F ( X t ) ∈ R 3 . To build V t , we used Partial Least Squares (PLS) and a Support Vector Machine (SVM) regressor, the details of which can be found in , Section 6. After training the decoder and evaluating on a held-out test set, results show that Ψ > 0, confirming our conjecture that the motor-related information is an emergent feature of the macaque’s cortical activity. For short timescales ( t ′ − t = 8 ms), we find Γ t , t ′ ( 1 ) ( V ) = 0 . 049 ± 0 . 002 , which is orders of magnitude smaller than Ψ t , t ′ ( 1 ) ( V ) = 1 . 275 ± 0 . 002 , suggesting that the behaviour may have an important component decoupled from individual ECoG channels (errors are standard deviations estimated over time-shuffled data). Furthermore, the emergence criterion is met for multiple timescales t ′ − t of up to ≈0.2s, beyond which the predictive power in V t and individual electrodes decrease and become nearly identical. As a control, we performed a surrogate data test to confirm the results in were not driven by the autocorrelation in the ECoG time series. To this end, we re-run the analysis (including training and testing the PLS-SVM) using the same ECoG data, but time-shuffling the wrist position—resulting in a V t that does not extract any meaningful information from the ECoG, but has the same properties induced by autocorrelation, filtering and regularisation. As expected, the resulting surrogate Ψ t , t ′ ( 1 ) is significantly lower than the one using the un-shuffled wrist position, confirming the measured Ψ t , t ′ ( 1 ) is positive and higher than what would be expected from a similar, random projection of the ECoG (details in , Section 6). This analysis, while just a proof of concept, helps us quantify how and to what extent behaviour emerges from collective neural activity; and opens the door to further tests and quantitative empirical explorations of the mind-matter relationship.
Here we present an evaluation of our practical criteria for emergence (Proposition 1) in two well-known systems: Conway’s Game of Life (GoL) , and Reynolds’ flocking boids model . Both are widely regarded as paradigmatic examples of emergent behaviour, and have been thoughtfully studied in the complexity and artificial life literature . Accordingly, we use these models as test cases for our methods. Technical details of the simulations are provided in , Section 5. Conway’s Game of Life A well-known feature of GoL is the presence of particles : coherent, self-sustaining structures known to be responsible for information transfer and modification . These particles have been the object of extensive study, and detailed taxonomies and classifications exist . To test the emergent properties of particles, we simulate the evolution of 15x15 square cell arrays, which we regard as a binary vector X t ∈ {0, 1} n with n = 225. As initial condition, we consider configurations that correspond to a “particle collider” setting, with two particles of known type facing each other . In each trial, the system is randomised by changing the position, type, and relative displacement of the particles. After an intial configuration has been selected, the well-known GoL evolution rule is applied 1000 times, leading to a final state X t ′ . Simulations showed that this interval is enough for the system to settle in a stable state after the collision. To use the criteria from , we need to choose a candidate emergent feature V t . In this case, we consider a symbolic, discrete-valued vector that encodes the type of particle(s) present in the board. Specifically, we consider V t = ( V t 1 , … , V t L ) , where V t j = 1 iff there is a particle of type j at time t —regardless of its position or orientation. With these variables, we compute the quantities in using Bayesian estimators of mutual information . The result is that, as expected, the criterion for causal emergence is met with Ψ t , t ′ ( 1 ) ( V ) = 0 . 58 ± 0 . 02 . Furthermore, we found that Γ t , t ′ ( 1 ) ( V ) = 0 . 009 ± 0 . 0002 , which is orders of magnitude smaller than I ( V t ; V t ′ ) = 0.99±0.02. Errors represent the standard deviation over surrogate data, as described in , Section 5. Using Proposition 1, these two results suggest that particle dynamics in GoL may not only be emergent, but causally decoupled with respect to their substrate. Reynolds’ flocking model As a second test case, we consider Reynolds’ model of flocking behaviour. This model is composed by boids (bird-oid objects), with each boid represented by three numbers: its position in 2D space and its heading angle. As candidate feature for emergence, we use the 2D coordinates of the center of mass of the flock, following Seth . In this model boids interact with one another following three rules, each regulated by a scalar parameter : aggregation ( a 1 ), as they fly towards the center of the flock; avoidance ( a 2 ), as they fly away from their closest neighbour; and alignment ( a 3 ), as they align their flight direction to that of their neighbours. Following Ref. , we study small flocks of N = 10 boids with different parameter settings to showcase some properties of our practical criterion of emergence. Note that this study is meant as an illustration of the proposed theory, and not as a thorough exploration of the flocking model, for which a vast literature exists (see e.g. the work of Vicsek and references therein). shows the results of a parameter sweep over the avoidance parameter, a 2 , while keeping a 1 and a 3 fixed. When there is no avoidance, boids orbit around a slowly-moving center of mass, in what could be called an ordered regime. Conversely, for high values of a 2 neighbour repulsion is too strong for lasting flocks to form, and isolated boids spread across the space avoiding one another. For intermediate values, the center of mass traces a smooth trajectory, as flocks form and disintegrate. In line with the findings of Seth , our criterion indicates that the flock exhibits causally emergent behaviour in this intermediate range. By studying separately the two terms that make up Ψ we found that the criterion of emergence fails for both low and high a 2 , but for different reasons (see ). In effect, for high a 2 the self-predictability of the center of mass (i.e. I ( V t ; V t ′ )) is low; while for low a 2 it is high, yet lower than the mutual information from individual boids (i.e. ∑ i I ( X t i ; V t ′ ) ). These results suggest that the low-avoidance scenario is dominated not by a reduction in synergy, but by an increase in redundancy, which effectively increases the synergy threshold needed to detect emergence. However, note that, due to the limitations of the criterion, the fact that Ψ t , t ′ ( 1 ) < 0 is inconclusive and does not rule out the possibility of emergence. This is a common limitation of whole-minus-sum estimators like Ψ ; further refinements may provide bounds that are less susceptible to these issues and perform accurately in these scenarios.
A well-known feature of GoL is the presence of particles : coherent, self-sustaining structures known to be responsible for information transfer and modification . These particles have been the object of extensive study, and detailed taxonomies and classifications exist . To test the emergent properties of particles, we simulate the evolution of 15x15 square cell arrays, which we regard as a binary vector X t ∈ {0, 1} n with n = 225. As initial condition, we consider configurations that correspond to a “particle collider” setting, with two particles of known type facing each other . In each trial, the system is randomised by changing the position, type, and relative displacement of the particles. After an intial configuration has been selected, the well-known GoL evolution rule is applied 1000 times, leading to a final state X t ′ . Simulations showed that this interval is enough for the system to settle in a stable state after the collision. To use the criteria from , we need to choose a candidate emergent feature V t . In this case, we consider a symbolic, discrete-valued vector that encodes the type of particle(s) present in the board. Specifically, we consider V t = ( V t 1 , … , V t L ) , where V t j = 1 iff there is a particle of type j at time t —regardless of its position or orientation. With these variables, we compute the quantities in using Bayesian estimators of mutual information . The result is that, as expected, the criterion for causal emergence is met with Ψ t , t ′ ( 1 ) ( V ) = 0 . 58 ± 0 . 02 . Furthermore, we found that Γ t , t ′ ( 1 ) ( V ) = 0 . 009 ± 0 . 0002 , which is orders of magnitude smaller than I ( V t ; V t ′ ) = 0.99±0.02. Errors represent the standard deviation over surrogate data, as described in , Section 5. Using Proposition 1, these two results suggest that particle dynamics in GoL may not only be emergent, but causally decoupled with respect to their substrate.
As a second test case, we consider Reynolds’ model of flocking behaviour. This model is composed by boids (bird-oid objects), with each boid represented by three numbers: its position in 2D space and its heading angle. As candidate feature for emergence, we use the 2D coordinates of the center of mass of the flock, following Seth . In this model boids interact with one another following three rules, each regulated by a scalar parameter : aggregation ( a 1 ), as they fly towards the center of the flock; avoidance ( a 2 ), as they fly away from their closest neighbour; and alignment ( a 3 ), as they align their flight direction to that of their neighbours. Following Ref. , we study small flocks of N = 10 boids with different parameter settings to showcase some properties of our practical criterion of emergence. Note that this study is meant as an illustration of the proposed theory, and not as a thorough exploration of the flocking model, for which a vast literature exists (see e.g. the work of Vicsek and references therein). shows the results of a parameter sweep over the avoidance parameter, a 2 , while keeping a 1 and a 3 fixed. When there is no avoidance, boids orbit around a slowly-moving center of mass, in what could be called an ordered regime. Conversely, for high values of a 2 neighbour repulsion is too strong for lasting flocks to form, and isolated boids spread across the space avoiding one another. For intermediate values, the center of mass traces a smooth trajectory, as flocks form and disintegrate. In line with the findings of Seth , our criterion indicates that the flock exhibits causally emergent behaviour in this intermediate range. By studying separately the two terms that make up Ψ we found that the criterion of emergence fails for both low and high a 2 , but for different reasons (see ). In effect, for high a 2 the self-predictability of the center of mass (i.e. I ( V t ; V t ′ )) is low; while for low a 2 it is high, yet lower than the mutual information from individual boids (i.e. ∑ i I ( X t i ; V t ′ ) ). These results suggest that the low-avoidance scenario is dominated not by a reduction in synergy, but by an increase in redundancy, which effectively increases the synergy threshold needed to detect emergence. However, note that, due to the limitations of the criterion, the fact that Ψ t , t ′ ( 1 ) < 0 is inconclusive and does not rule out the possibility of emergence. This is a common limitation of whole-minus-sum estimators like Ψ ; further refinements may provide bounds that are less susceptible to these issues and perform accurately in these scenarios.
A tantalising outcome of having a formal theory of emergence is the capability of bringing a quantitative angle to the archetype of emergence: the mind-matter relationship . As a first step in this direction, we conclude this section with an application of our emergence criteria to neurophysiological data. We study simultaneous electrocorticogram (ECoG) and motion capture (MoCap) data of Japanese macaques performing a reaching task , obtained from the online Neurotycho database. Note that the MoCap data cannot be assumed to be a supervenient feature of the available ECoG data, since it doesn’t satisfy the conditional independence conditions required by our definition of supervenience (see Section A formal theory of causal emergence ). This is likely to be the case, because the neural system is only partially observed—i.e. the ECoG does not capture every source of relevant activity in the macaque’s cortex. Note that non-supervenient features are of limited interest within our framework, as they can satisfy Proposition 1 in trivial ways (e.g. time series which are independent of the underlying system satisty Ψ > 0 if they are auto-correlated). Instead, we focus on the portion of neural activity encoded in the ECoG signal that is relevant to predict the macaque’s behaviour, and conjecture this information to be an emergent feature of the underlying neural activity . To test this hypothesis, we take the neural activity (as measured by 64 ECoG channels distributed across the left hemisphere) to be the system of interest, and consider a memoryless predictor of the 3D coordinates of the macaque’s right wrist based on the ECoG signal. Therefore, in this scenario X t ∈ R 64 and V t = F ( X t ) ∈ R 3 . To build V t , we used Partial Least Squares (PLS) and a Support Vector Machine (SVM) regressor, the details of which can be found in , Section 6. After training the decoder and evaluating on a held-out test set, results show that Ψ > 0, confirming our conjecture that the motor-related information is an emergent feature of the macaque’s cortical activity. For short timescales ( t ′ − t = 8 ms), we find Γ t , t ′ ( 1 ) ( V ) = 0 . 049 ± 0 . 002 , which is orders of magnitude smaller than Ψ t , t ′ ( 1 ) ( V ) = 1 . 275 ± 0 . 002 , suggesting that the behaviour may have an important component decoupled from individual ECoG channels (errors are standard deviations estimated over time-shuffled data). Furthermore, the emergence criterion is met for multiple timescales t ′ − t of up to ≈0.2s, beyond which the predictive power in V t and individual electrodes decrease and become nearly identical. As a control, we performed a surrogate data test to confirm the results in were not driven by the autocorrelation in the ECoG time series. To this end, we re-run the analysis (including training and testing the PLS-SVM) using the same ECoG data, but time-shuffling the wrist position—resulting in a V t that does not extract any meaningful information from the ECoG, but has the same properties induced by autocorrelation, filtering and regularisation. As expected, the resulting surrogate Ψ t , t ′ ( 1 ) is significantly lower than the one using the un-shuffled wrist position, confirming the measured Ψ t , t ′ ( 1 ) is positive and higher than what would be expected from a similar, random projection of the ECoG (details in , Section 6). This analysis, while just a proof of concept, helps us quantify how and to what extent behaviour emerges from collective neural activity; and opens the door to further tests and quantitative empirical explorations of the mind-matter relationship.
A large fraction of the modern scientific literature considers strong emergence to be impossible or ill-defined. This judgement is not fully unfounded: a property that is simultaneously supervenient (i.e. that can be computed from the state of the system) and that has irreducible causal power (i.e. that “tells us something” that the parts don’t) can indeed seem to be an oxymoron . Nonetheless, by linking supervenience to static relationships and causal power to dynamical properties, our framework shows that these two phenomena are perfectly compatible within the—admittedly counterintuitive – laws of multivariate information dynamics , providing a tentative solution to this paradox. Our theory of causal emergence is about predictive power, not “explicability” , and therefore is not related to views on strong emergence such as Chalmers’ . Nevertheless, our framework embraces aspects that are commonly associated with strong emergence—such as downward causation—and renders them quantifiable. Our framework also does not satisfy conventional definitions of weak emergence (systems studied in Section Fundamental intuitions are not weakly emergent in the sense of Bedau , being simple and susceptible to explanatory shortcuts) but is compatible with more general notions of weak emergence, e.g. the one introduced by Seth (see Section Relationship with other quantitative theories of emergence ). Hence, our theory can be seen as an attempt at reconciling these approaches , showing how “strong” a “weak” framework can be. An important consequence of our theory is the fundamental connection established between causal emergence and statistical synergy: the system’s capacity to host emergent features was found to be determined by how synergistic its elements are with respect to their future evolution. Although previous ideas about synergy have been loosely linked to emergence in the past , this is (to the best of our knowledge) the first time such ideas have been formally laid out and quantified using recent advances in multivariate information theory. Next, we examine a few caveats regarding the applicability of the proposed theory, its relation with prior work, and some open problems. Scope of the theory Our theory focuses on synchronic aspects of emergence, analysing the interactions between the elements of dynamical systems and collective properties of them as they jointly evolve over time. As such, our theory directly applies to any system with well-defined dynamics, including systems described by deterministic dynamical systems with random initial conditions and stochastic systems described by Fokker-Planck equations . In contrast, the application of our theory to systems in thermodynamic equilibrium may not be straightforward, as their dynamics are often not uniquely specified by the corresponding Gibbs distributions (for an explicit example, when considering the Ising model, Kawasaki and Glauber dynamics are known to behave differently even when the system is in equilibrium ; and thus may provide quite different values of the measures described in Section Measuring emergence ). Finding principled approaches to guide the application of our theory to those cases is an interesting challenge for future studies. In addition, given the breadth of the concept of “emergence,” there are a number of other theories leaning more towards philosophy that are orthogonal to our framework. This includes, for example, theories of emergence as radical novelty (in the sense of features not previously observed in the system) , most prominently encapsulated in the aphorism “more is different” by Anderson—see Refs. , particularly his approach to emergence in biology (note that some of Anderson’s views, particularly the ones related to rigidity, are nevertheless closely related to the approach developed by our framework), and also articulated in the work of Kauffman . Also, contextual emergence emphasises a role for macro-level contexts that cannot be described at the micro-level, but which impose constraints on the micro-level for the emergence of the macro . These are valuable philosophical positions, which have been studied from a statistical mechanics perspective in Ref. . Future work shall attempt to unify these other approaches with our proposed framework. Causality The de facto way to assess the causal structure of a system is to analyse its response to controlled interventions or to build intervention models (causal graphs) based on expert knowledge, which leads to the well-known do-calculus spearheaded by Judea Pearl . This approach is, unfortunately, not applicable in many scenarios of interest, as interventions may incur prohibitive costs or even be impossible, and expert knowledge may not be available. These scenarios can still be assessed via the Wiener-Granger theory of statistical causation , which studies the blueprint of predictive power across the system of interest by accounting non-mediated correlations between past and future events . Both frameworks provide similar results when all the relevant variables have been measured, but can neverthelss differ radically when there are unobserved interacting variables . The debate between the Wiener-Granger and the Pearl schools has been discussed in other related contexts—see e.g. Refs. for a discussion regarding Integrated Information Theory (IIT), and Ref. for a discussion about effective and functional connectivity in the context of neuroimaging time series analysis (in a nutshell, effective connectivity aims to uncover the minimal physical causal mechanism underlying the observed data, while functional connectivity describes directed or undirected statistical dependences ). In our theory, the main object of analysis is Shannon’s mutual information, I ( X t ; X t ′ ), which depends on the joint probability distribution p X t , X t ′ . The origin of this distribution (whether it was obtained by passive observation or by active intervention) will change the interpretation of the quantities presented above, and will speak differently to the Pearl and the Wiener-Granger schools of thought; some of the implications of these differences are addressed when discussing Ref. below. Nonetheless, since both methods of obtaining p X t , X t ′ allow synergy to take place, our results are in principle applicable in both frameworks—which allows us to formulate our theory of causal emergence without taking a rigid stance on a theory of causality itself. Relationship with other quantitative theories of emergence This work is part of a broader movement towards formalising theories of complexity through information theory. In particular, our framework is most directly inspired by the work of Seth and Hoel et al . , and also related to recent work by Chang et al . . This section gives a brief account of these theories, and discusses how they differ from our proposal. Seth proposes that a process V t is Granger-emergent (or G-emergent ) with respect to X t if two conditions are met: (i) V t is autonomous with respect to X t (i.e. I ( V t ; V t ′ | X t )>0), and (ii) V t is G-caused by X t (i.e. I ( X t ; V t ′ | V t )>0). The latter condition is employed to guarantee a relationship between X t and V t ; in our framework an equivalent role is taken by the requirement of supervenience. The condition of autonomy is certainly related with our notion of causal decoupling. However, as shown in Ref. , the conditional mutual information conflates unique and synergistic information, which can give rise to undesirable situations: for example, it could be that I ( V t ; V t ′ | X t )>0 while, at the same time, I ( V ; V t ′ ) = 0, meaning that the dynamics of the feature V t are only visible when considering it together with the full system X t , but not on its own. Our framework avoids this problem by refining this criterion via PID, and uses only the unique information for the definition of emergence. Our work is also strongly influenced by the framework put forward by Hoel and colleagues in Ref. . Their approach is based on a coarse-graining function F (⋅) relating a feature of interest to the system, V t = F ( X t ), which is a particular case of our more general definition of supervenience. Emergence is declared when the dependency between V t and V t ′ is “stronger” than the one between X t and X t ′ . Note that V t − X t − X t ′ − V t ′ is a Markov chain, and hence I ( V t ; V t ′ )≤ I ( X t ; X t ′ ) due to the data processing inequality; therefore, a direct usage of Shannon’s mutual information would make the above criterion impossible to fulfill. Instead, this framework focuses on the transition probabilities p V t ′ | V t and p X t ′ | X t , and hence the mutual information terms are computed using maximum entropy distributions instead of the stationary marginals. By doing this, Hoel et al . account not for what the system actually does , but for all the potential transitions the system could do . However, in our view this approach is not well-suited to assess dynamical systems, as it might account for transitions that are never actually explored. The difference between stationary and maximum entropy distributions can be particularly dramatic in non-ergodic systems with multiple attractors—for a related discussion in the context of Integrated Information Theory, see Ref. . Additionally, this framework relies on having exact knowledge about the microscopic transitions as encoded by p X t ′ | X t , which is not possible to obtain in most applications. Finally, Chang et al . consider supervenient variables that are “non-trivially informationally closed” (NTIC) to their corresponding microscopic substrate. NTIC is based on a division of X t into a subsystem of interest, X t α , and its “environment” given by X t - α . Interestingly, a system being NTIC requires V t to be supervenient only with respect to X t α (i.e. V t = F ( X t α ) ), as well as information flow from the environment to the feature (i.e. I ( X t - α ; V t ′ ) > 0 ) mediated by the feature itself, so that X t − V t − V t ′ is a Markov chain. Hence, NTIC requires features that are sufficient statistics for their own dynamics, which is akin to our concept of causal decoupling but focused on the interaction between a macroscopic feature, an agent, and its environment. Extending our framework to agent-environment systems involved in active inference is part of our future work. Limitations and open problems The framework presented in this paper focuses on features from fully observable systems with Markovian dynamics. These assumptions, however, often do not hold when dealing with experimental data—especially in biological and social systems. As an important extension, future work should investigate the effect of unobserved variables on our measures. This could be done, for example, leveraging Takens’ embedding theorem or other methods . An interesting feature of our framework is that, although it depends on the choice of PID and ΦID, its practical application via the criteria discussed in Section Practical criteria for large systems is agnostic to those choices. However, they incur the cost of a limited sensitivity to detect emergence due to an overestimation of the microscopic redundancy; so they can detect emergence when it is substantial, but might miss it in more subtle cases. Additionally, these criteria are unable to rule out emergence, as they are sufficient but not necessary conditions. Therefore, another avenue of future work should search for improved practical criteria for detecting emergence from data. One interesting line of research is providing scalable approximations for Syn ⋆ ( k ) and G ⋆ ( k ) as introduced in Section Measuring emergence via synergistic channels , which could be computed in large systems. Another open question is how the emergence capacity is affected by changes in the microscopic partition of the system (c.f. Section Defining causal emergence ). Interesting applications of this includes scenarios where elements of interest have been subject to a mixing process, such as the case of electroencephalography where each electrode detects a mixture of brain sources. Other interesting questions include studying systems with non-zero emergence capacity for all reasonable microscopic partitions, which may correspond to a stronger type of emergence.
Our theory focuses on synchronic aspects of emergence, analysing the interactions between the elements of dynamical systems and collective properties of them as they jointly evolve over time. As such, our theory directly applies to any system with well-defined dynamics, including systems described by deterministic dynamical systems with random initial conditions and stochastic systems described by Fokker-Planck equations . In contrast, the application of our theory to systems in thermodynamic equilibrium may not be straightforward, as their dynamics are often not uniquely specified by the corresponding Gibbs distributions (for an explicit example, when considering the Ising model, Kawasaki and Glauber dynamics are known to behave differently even when the system is in equilibrium ; and thus may provide quite different values of the measures described in Section Measuring emergence ). Finding principled approaches to guide the application of our theory to those cases is an interesting challenge for future studies. In addition, given the breadth of the concept of “emergence,” there are a number of other theories leaning more towards philosophy that are orthogonal to our framework. This includes, for example, theories of emergence as radical novelty (in the sense of features not previously observed in the system) , most prominently encapsulated in the aphorism “more is different” by Anderson—see Refs. , particularly his approach to emergence in biology (note that some of Anderson’s views, particularly the ones related to rigidity, are nevertheless closely related to the approach developed by our framework), and also articulated in the work of Kauffman . Also, contextual emergence emphasises a role for macro-level contexts that cannot be described at the micro-level, but which impose constraints on the micro-level for the emergence of the macro . These are valuable philosophical positions, which have been studied from a statistical mechanics perspective in Ref. . Future work shall attempt to unify these other approaches with our proposed framework.
The de facto way to assess the causal structure of a system is to analyse its response to controlled interventions or to build intervention models (causal graphs) based on expert knowledge, which leads to the well-known do-calculus spearheaded by Judea Pearl . This approach is, unfortunately, not applicable in many scenarios of interest, as interventions may incur prohibitive costs or even be impossible, and expert knowledge may not be available. These scenarios can still be assessed via the Wiener-Granger theory of statistical causation , which studies the blueprint of predictive power across the system of interest by accounting non-mediated correlations between past and future events . Both frameworks provide similar results when all the relevant variables have been measured, but can neverthelss differ radically when there are unobserved interacting variables . The debate between the Wiener-Granger and the Pearl schools has been discussed in other related contexts—see e.g. Refs. for a discussion regarding Integrated Information Theory (IIT), and Ref. for a discussion about effective and functional connectivity in the context of neuroimaging time series analysis (in a nutshell, effective connectivity aims to uncover the minimal physical causal mechanism underlying the observed data, while functional connectivity describes directed or undirected statistical dependences ). In our theory, the main object of analysis is Shannon’s mutual information, I ( X t ; X t ′ ), which depends on the joint probability distribution p X t , X t ′ . The origin of this distribution (whether it was obtained by passive observation or by active intervention) will change the interpretation of the quantities presented above, and will speak differently to the Pearl and the Wiener-Granger schools of thought; some of the implications of these differences are addressed when discussing Ref. below. Nonetheless, since both methods of obtaining p X t , X t ′ allow synergy to take place, our results are in principle applicable in both frameworks—which allows us to formulate our theory of causal emergence without taking a rigid stance on a theory of causality itself.
This work is part of a broader movement towards formalising theories of complexity through information theory. In particular, our framework is most directly inspired by the work of Seth and Hoel et al . , and also related to recent work by Chang et al . . This section gives a brief account of these theories, and discusses how they differ from our proposal. Seth proposes that a process V t is Granger-emergent (or G-emergent ) with respect to X t if two conditions are met: (i) V t is autonomous with respect to X t (i.e. I ( V t ; V t ′ | X t )>0), and (ii) V t is G-caused by X t (i.e. I ( X t ; V t ′ | V t )>0). The latter condition is employed to guarantee a relationship between X t and V t ; in our framework an equivalent role is taken by the requirement of supervenience. The condition of autonomy is certainly related with our notion of causal decoupling. However, as shown in Ref. , the conditional mutual information conflates unique and synergistic information, which can give rise to undesirable situations: for example, it could be that I ( V t ; V t ′ | X t )>0 while, at the same time, I ( V ; V t ′ ) = 0, meaning that the dynamics of the feature V t are only visible when considering it together with the full system X t , but not on its own. Our framework avoids this problem by refining this criterion via PID, and uses only the unique information for the definition of emergence. Our work is also strongly influenced by the framework put forward by Hoel and colleagues in Ref. . Their approach is based on a coarse-graining function F (⋅) relating a feature of interest to the system, V t = F ( X t ), which is a particular case of our more general definition of supervenience. Emergence is declared when the dependency between V t and V t ′ is “stronger” than the one between X t and X t ′ . Note that V t − X t − X t ′ − V t ′ is a Markov chain, and hence I ( V t ; V t ′ )≤ I ( X t ; X t ′ ) due to the data processing inequality; therefore, a direct usage of Shannon’s mutual information would make the above criterion impossible to fulfill. Instead, this framework focuses on the transition probabilities p V t ′ | V t and p X t ′ | X t , and hence the mutual information terms are computed using maximum entropy distributions instead of the stationary marginals. By doing this, Hoel et al . account not for what the system actually does , but for all the potential transitions the system could do . However, in our view this approach is not well-suited to assess dynamical systems, as it might account for transitions that are never actually explored. The difference between stationary and maximum entropy distributions can be particularly dramatic in non-ergodic systems with multiple attractors—for a related discussion in the context of Integrated Information Theory, see Ref. . Additionally, this framework relies on having exact knowledge about the microscopic transitions as encoded by p X t ′ | X t , which is not possible to obtain in most applications. Finally, Chang et al . consider supervenient variables that are “non-trivially informationally closed” (NTIC) to their corresponding microscopic substrate. NTIC is based on a division of X t into a subsystem of interest, X t α , and its “environment” given by X t - α . Interestingly, a system being NTIC requires V t to be supervenient only with respect to X t α (i.e. V t = F ( X t α ) ), as well as information flow from the environment to the feature (i.e. I ( X t - α ; V t ′ ) > 0 ) mediated by the feature itself, so that X t − V t − V t ′ is a Markov chain. Hence, NTIC requires features that are sufficient statistics for their own dynamics, which is akin to our concept of causal decoupling but focused on the interaction between a macroscopic feature, an agent, and its environment. Extending our framework to agent-environment systems involved in active inference is part of our future work.
The framework presented in this paper focuses on features from fully observable systems with Markovian dynamics. These assumptions, however, often do not hold when dealing with experimental data—especially in biological and social systems. As an important extension, future work should investigate the effect of unobserved variables on our measures. This could be done, for example, leveraging Takens’ embedding theorem or other methods . An interesting feature of our framework is that, although it depends on the choice of PID and ΦID, its practical application via the criteria discussed in Section Practical criteria for large systems is agnostic to those choices. However, they incur the cost of a limited sensitivity to detect emergence due to an overestimation of the microscopic redundancy; so they can detect emergence when it is substantial, but might miss it in more subtle cases. Additionally, these criteria are unable to rule out emergence, as they are sufficient but not necessary conditions. Therefore, another avenue of future work should search for improved practical criteria for detecting emergence from data. One interesting line of research is providing scalable approximations for Syn ⋆ ( k ) and G ⋆ ( k ) as introduced in Section Measuring emergence via synergistic channels , which could be computed in large systems. Another open question is how the emergence capacity is affected by changes in the microscopic partition of the system (c.f. Section Defining causal emergence ). Interesting applications of this includes scenarios where elements of interest have been subject to a mixing process, such as the case of electroencephalography where each electrode detects a mixture of brain sources. Other interesting questions include studying systems with non-zero emergence capacity for all reasonable microscopic partitions, which may correspond to a stronger type of emergence.
This paper introduces a quantitative definition of causal emergence, which addresses the apparent paradox of supervenient macroscopic features with irreducible causal power using principles of multivariate statistics. We provide a formal, quantitative theory that embodies many of the principles attributed to strong emergence, while being measurable and compatible with the established scientific worldview. Perhaps the most important contribution of this work is to bring the discussion of emergence closer to the realm of quantitative, empirical scientific investigation, complementing the ongoing philosophical inquiries on the subject. Mathematically, the theory is based on the Partial Information Decomposition (PID) framework , and on a recent extension, Integrated Information Decomposition (ΦID) . The theory allows the derivation of sufficiency criteria for the detection of emergence that are scalable, easy to compute from data, and based only on Shannon’s mutual information. We illustrated the use of these practical criteria in three case studies, and concluded that: i) particle collisions are an emergent feature in Conway’s Game of Life, ii) flock dynamics are an emergent feature of simulated birds; and iii) the representation of motor behaviour in the cortex is emergent from neural activity. Our theory, together with these practical criteria, enables novel data-driven tools for scientifically addressing conjectures about emergence in a wide range of systems of interest. Our original aim in developing this theory, beyond the contribution to complexity theory, is to help bridge the gap between the mental and the physical, and ultimately understand how mind emerges from matter. This paper provides formal principles to explore the idea that psychological phenomena could emerge from collective neural patterns, and interact with each other dynamically in a causally decoupled fashion—perhaps akin to the “statistical ghosts” mentioned in Section Causal decoupling . Put simply: just as particles in the Game of Life have their own collision rules, we wonder if thought patterns could have their own emergent dynamical laws, operating at a larger scale with respect to their underlying neural substrate (similar ideas have been recently explored by Kent ). Importantly, the theory presented in this paper not only provides conceptual tools to frame this conjecture rigorously, but also provides practical tools to test it from data. The exploration of this conjecture is left as an exciting avenue for future research.
S1 Appendix Provides the mathematical proofs of our results, and further details about simulations and preprocessing pipelines. (PDF) Click here for additional data file.
|
Relationship between digital health literacy, distrust in the health system and health anxiety in health sciences students | 9f0313d4-026a-443b-93df-4f4d2e4869eb | 11890528 | Health Literacy[mh] | With the development of technology, individual Internet use has also increased worldwide . In 2023, 5 billion 45 million people around the world are Internet users. Similarly, 65% of women and 70% of men are Internet users globally. When analyzing Internet users by age group, people in the 25–34 age group are in first place with 35.6%, people in the 35–44 age group are in second place with 24%, and people in the 18–24 age group are in third place with 19%. When the Internet usage rates are analyzed by country, China, India and the United States of America (USA) are in the first three places . In Türkiye, adult Internet usage increased from 87.1% in 2023 to 88.8% in 2024. In 2024, this rate is 92.2% for adult males and 85.4% for adult females . Because such intensive use of the Internet provides free access to information, people seek health information online when dealing with a health-related situation, as with any subject . Online health information resources can meet the needs of patients before consulting a physician in terms of diagnosis, treatment, symptoms, indications and contraindications for diseases . This behavior, which is usually performed via general search engines, can be effective in protecting and improving individuals’ health, reaching the appropriate branch and physician for diagnosis and treatment, and coping with problems. Health information search behavior from trusted websites can benefit individuals in terms of the urgency of the problem, understanding the symptoms, and applying to the right branch and physician. However, information obtained as a result of searches from unreliable sites can cause individual health anxiety . Health anxieties refer to fear and worry from an individual’s hypersensitivity to his/her own health status . In the Diagnostic and Statistical Manual of Mental Disorders (DSM-5), the concept of health anxiety consists of two categories . Health anxiety with symptoms is characterized as a somatic symptom disorder, whereas health anxiety without somatic symptoms is characterized as a hypochondriasis disorder. The main characteristic of individuals with health anxiety is constant anxiety, and they are worried that they have a serious illness, which is an interpretation of concerns arising from one’s symptoms . Individuals with health anxiety tend to engage in more online health information-seeking behaviors to find solutions to these problems. This condition is referred to as Cyberchondria in the literature . Cyberchondria is an important health problem and is associated with hypochondriasis, health anxiety, obsessive-compulsive disorders and health anxiety . In addition to an increase in the demand for health services, an increase in the average number of applications to health institutions and new costs and dissatisfaction with the service increased the level of anxiety . However, the negative behavior of individuals is thought to improve as the level of health literacy and e-health literacy increases . Health literacy is defined as the ability to access, process and understand basic health information and services necessary for individual health decisions and is generally accepted as an important determinant of health behaviors and outcomes . Digital health literacy, also known as e-health literacy, is defined as the process of searching for, finding, understanding and evaluating health-related information from electronic platforms in line with technological advances and addressing and solving health-related problems using this information. Unlike other types of literacy recognized in the literature, Digital Health literacy (DHL) involves combinations of different components: computer literacy, tradition literacy, information literacy, health literacy, scientific literacy and medical literacy . Inadequate health literacy impedes timely access to health services, leading to misuse or unnecessary utilization. Limited health literacy is a key element of health and well-being and has negative impacts on health protection and disease prevention. Moreover, low health literacy affects many areas of health, including the management of chronic diseases, decision-making about healthy living habits, misunderstanding and misuse of medicines, frequent hospitalizations and mortality rates, and the use of expensive health services, especially specialists and emergency services. The negative outcomes increase the burden on and cost of health systems . Young people entering university struggle with the challenges of adolescence as they try to adapt to their new environment. Living away from family, adjusting to university life, developing new social circles, and coping with financial problems are important factors that affect students’ physical and mental health. In this process, especially when close friendships have not yet been established, young people tend to seek online information on sexual and mental health issues. In these situations where health anxiety is high, young people with low e-health literacy may experience more anxiety due to the information they receive from unreliable sources. Students in the health sciences, in particular, face higher levels of stress due to factors such as heavy academic workloads, intense course schedules, and the search for identity . Unlike other members of society, students studying in health sciences should have a higher level of digital health literacy, because this group is the most important professional group in the formation of the basis of public health, the protection and improvement of the health of society. For this reason, the aim of this study is to reveal the relationships between digital health literacy, health anxiety and distrust in the health system of students studying in health sciences departments and the effects of digital health literacy and distrust in the health system on health anxiety. In the literature review, no study in which these three factors were studied together was found. This situation reveals the originality of the study.
Study design The study was cross-sectional. The population of the study consisted of the Faculty of Medicine, Faculty of Dentistry, Faculty of Health Sciences (Nursing, Health Management, Nutrition and Dietetics, Midwifery, Physiotherapy and Rehabilitation, Speech Therapy, Social Work, Ergotherapy) and Vocational School of Health Services (Medical Services and Techniques, Pharmacy Services, Child Development, Health Care Services, Therapy and Rehabilitation and Dental Services Department) of Kutahya Health Sciences University, which was established with the theme of health in Türkiye. It consists of 5,455 students between the ages of 18–40 enrolled in 2024–2025. The convenience sampling method was used in the study. Turkish validity and reliability of the digitally prepared questionnaire forms have been previously conducted. The prepared questionnaire form was shared with the students via the WhatsApp program through class representatives. The data were collected between 20.09.2024 and 31.10.2024. In the explanation section of the digital questionnaire, the purpose, possible risks and benefits of the study were explained, and the first question “do you agree to participate in this study?” was asked to obtain consent from the participants. The forms of the participants who answered “yes” to this question were included in the study. To prevent participants from completing more than one questionnaire, the block feature from the same IP and the same e-mail address was used. Of the questionnaires sent digitally to the entire population, 752 completed questionnaires were analyzed. Data collection tools A digital questionnaire consisting of 4 sections was used to collect the research data. The first part of the questionnaire includes 10 statements to measure the socio-demographic characteristics and online health information seeking behavior of students. Digital health literacy scale was used in the second part, health anxiety scale in the third part and mistrust in health system scale in the fourth part. Permission to use the scales in the study was obtained by e-mail from the academicians who conducted the Turkish validity and reliability study. Digital health literacy scale This scale was developed by Çetin and Gümüş (2023). It consists of 18 statements and 6 subdimensions. The subdimensions of Searching for Information, Assessing Reliability, Determining the Level of Interest and Adding Content are scored in 3 statements and a 4-point Likert scale “4 -Quite Easy, 1 - Quite Difficult”. The direction-finding skills and privacy protection subdimensions were scored in 3 statements and a 4-point Likert scale “4 - Never, 1 - Frequently”. The Cronbach’s alpha value of the scale is 0.907. The scale mean is minimum 1, maximum 4. When the average score approaches 4, the level of digital health literacy is high . In this study, the Cronbach’s alpha value of the scale was 0.773. Health anxiety inventory The Health Anxiety Inventory was developed by Salkovskis et al. (2002) to measure a wide range of health anxiety symptoms. The inventory consists of 18 self-reported questions. The main section (first 14 items), related to the dimension of hypersensitivity to somatic symptoms and distress, contains four ordered statements questioning mental state. The negative consequences section (the remaining 4 items) asks about people’s attitudes about how bad it would be if they were to develop a serious illness. Each item of the inventory is scored on a 0–3 scale, and the scores for each item are summed to obtain a total score ranging from 0 to 54. A higher score indicates a higher level of health anxiety. The short form can be used in health screening; the inventory can also help differentiate hypochondriasis from anxiety disorders and identify true physical illnesses. The Health Anxiety Inventory was adapted into Turkish by Aydemir et al. (2013), and reliability and validity analyses of the Turkish version of the inventory were conducted on a sample of outpatients with somatoform disorders and panic disorders and university students . In this study, the Cronbach’s alpha value of the scale was 0.887. Distrust in health systems scale The “Distrust in Health Systems Scale” developed by Rose et al. (2004) was adapted to Türkiye by Yeşildal (2020) et al. The scale consists of 10 items and one dimension. The scale items are scored using 5-point Likert-type, and the participants are asked to mark the most appropriate option in the range of “1 - Strongly disagree, 2 - Disagree, 3 - Undecided, 4 - Agree, 5 - Strongly agree”. The scale is calculated via the arithmetic mean method. A scale mean approaching 5 indicates an increase in mistrust of health systems, while a scale mean approaching 1 indicates a decrease in mistrust. The Cronbach’s alpha of the research conducted by Yeşildal et al. (2020) was calculated as 0.789 . In this study, the Cronbach’s alpha value of the scale was 0.712. Statistical analysis We used the SPSS 26.00 program for data analysis. The skewness and kurtosis values of the data were examined first. Since the skewness and kurtosis values were found to be between + 1.5 and − 1.5, it was assumed that the data were normally distributed. Therefore, Pearson correlation analysis was performed to determine the relationships among the variables. Multiple regression analysis was used to determine the effect between variables. Ethical approval Permission was obtained from the Noninterventional Research Ethics Committee of Kütahya Health Sciences University with the code number of 2024/07–18 and a date of 17.05.2024.
The study was cross-sectional. The population of the study consisted of the Faculty of Medicine, Faculty of Dentistry, Faculty of Health Sciences (Nursing, Health Management, Nutrition and Dietetics, Midwifery, Physiotherapy and Rehabilitation, Speech Therapy, Social Work, Ergotherapy) and Vocational School of Health Services (Medical Services and Techniques, Pharmacy Services, Child Development, Health Care Services, Therapy and Rehabilitation and Dental Services Department) of Kutahya Health Sciences University, which was established with the theme of health in Türkiye. It consists of 5,455 students between the ages of 18–40 enrolled in 2024–2025. The convenience sampling method was used in the study. Turkish validity and reliability of the digitally prepared questionnaire forms have been previously conducted. The prepared questionnaire form was shared with the students via the WhatsApp program through class representatives. The data were collected between 20.09.2024 and 31.10.2024. In the explanation section of the digital questionnaire, the purpose, possible risks and benefits of the study were explained, and the first question “do you agree to participate in this study?” was asked to obtain consent from the participants. The forms of the participants who answered “yes” to this question were included in the study. To prevent participants from completing more than one questionnaire, the block feature from the same IP and the same e-mail address was used. Of the questionnaires sent digitally to the entire population, 752 completed questionnaires were analyzed.
A digital questionnaire consisting of 4 sections was used to collect the research data. The first part of the questionnaire includes 10 statements to measure the socio-demographic characteristics and online health information seeking behavior of students. Digital health literacy scale was used in the second part, health anxiety scale in the third part and mistrust in health system scale in the fourth part. Permission to use the scales in the study was obtained by e-mail from the academicians who conducted the Turkish validity and reliability study. Digital health literacy scale This scale was developed by Çetin and Gümüş (2023). It consists of 18 statements and 6 subdimensions. The subdimensions of Searching for Information, Assessing Reliability, Determining the Level of Interest and Adding Content are scored in 3 statements and a 4-point Likert scale “4 -Quite Easy, 1 - Quite Difficult”. The direction-finding skills and privacy protection subdimensions were scored in 3 statements and a 4-point Likert scale “4 - Never, 1 - Frequently”. The Cronbach’s alpha value of the scale is 0.907. The scale mean is minimum 1, maximum 4. When the average score approaches 4, the level of digital health literacy is high . In this study, the Cronbach’s alpha value of the scale was 0.773. Health anxiety inventory The Health Anxiety Inventory was developed by Salkovskis et al. (2002) to measure a wide range of health anxiety symptoms. The inventory consists of 18 self-reported questions. The main section (first 14 items), related to the dimension of hypersensitivity to somatic symptoms and distress, contains four ordered statements questioning mental state. The negative consequences section (the remaining 4 items) asks about people’s attitudes about how bad it would be if they were to develop a serious illness. Each item of the inventory is scored on a 0–3 scale, and the scores for each item are summed to obtain a total score ranging from 0 to 54. A higher score indicates a higher level of health anxiety. The short form can be used in health screening; the inventory can also help differentiate hypochondriasis from anxiety disorders and identify true physical illnesses. The Health Anxiety Inventory was adapted into Turkish by Aydemir et al. (2013), and reliability and validity analyses of the Turkish version of the inventory were conducted on a sample of outpatients with somatoform disorders and panic disorders and university students . In this study, the Cronbach’s alpha value of the scale was 0.887. Distrust in health systems scale The “Distrust in Health Systems Scale” developed by Rose et al. (2004) was adapted to Türkiye by Yeşildal (2020) et al. The scale consists of 10 items and one dimension. The scale items are scored using 5-point Likert-type, and the participants are asked to mark the most appropriate option in the range of “1 - Strongly disagree, 2 - Disagree, 3 - Undecided, 4 - Agree, 5 - Strongly agree”. The scale is calculated via the arithmetic mean method. A scale mean approaching 5 indicates an increase in mistrust of health systems, while a scale mean approaching 1 indicates a decrease in mistrust. The Cronbach’s alpha of the research conducted by Yeşildal et al. (2020) was calculated as 0.789 . In this study, the Cronbach’s alpha value of the scale was 0.712.
This scale was developed by Çetin and Gümüş (2023). It consists of 18 statements and 6 subdimensions. The subdimensions of Searching for Information, Assessing Reliability, Determining the Level of Interest and Adding Content are scored in 3 statements and a 4-point Likert scale “4 -Quite Easy, 1 - Quite Difficult”. The direction-finding skills and privacy protection subdimensions were scored in 3 statements and a 4-point Likert scale “4 - Never, 1 - Frequently”. The Cronbach’s alpha value of the scale is 0.907. The scale mean is minimum 1, maximum 4. When the average score approaches 4, the level of digital health literacy is high . In this study, the Cronbach’s alpha value of the scale was 0.773.
The Health Anxiety Inventory was developed by Salkovskis et al. (2002) to measure a wide range of health anxiety symptoms. The inventory consists of 18 self-reported questions. The main section (first 14 items), related to the dimension of hypersensitivity to somatic symptoms and distress, contains four ordered statements questioning mental state. The negative consequences section (the remaining 4 items) asks about people’s attitudes about how bad it would be if they were to develop a serious illness. Each item of the inventory is scored on a 0–3 scale, and the scores for each item are summed to obtain a total score ranging from 0 to 54. A higher score indicates a higher level of health anxiety. The short form can be used in health screening; the inventory can also help differentiate hypochondriasis from anxiety disorders and identify true physical illnesses. The Health Anxiety Inventory was adapted into Turkish by Aydemir et al. (2013), and reliability and validity analyses of the Turkish version of the inventory were conducted on a sample of outpatients with somatoform disorders and panic disorders and university students . In this study, the Cronbach’s alpha value of the scale was 0.887.
The “Distrust in Health Systems Scale” developed by Rose et al. (2004) was adapted to Türkiye by Yeşildal (2020) et al. The scale consists of 10 items and one dimension. The scale items are scored using 5-point Likert-type, and the participants are asked to mark the most appropriate option in the range of “1 - Strongly disagree, 2 - Disagree, 3 - Undecided, 4 - Agree, 5 - Strongly agree”. The scale is calculated via the arithmetic mean method. A scale mean approaching 5 indicates an increase in mistrust of health systems, while a scale mean approaching 1 indicates a decrease in mistrust. The Cronbach’s alpha of the research conducted by Yeşildal et al. (2020) was calculated as 0.789 . In this study, the Cronbach’s alpha value of the scale was 0.712.
We used the SPSS 26.00 program for data analysis. The skewness and kurtosis values of the data were examined first. Since the skewness and kurtosis values were found to be between + 1.5 and − 1.5, it was assumed that the data were normally distributed. Therefore, Pearson correlation analysis was performed to determine the relationships among the variables. Multiple regression analysis was used to determine the effect between variables.
Permission was obtained from the Noninterventional Research Ethics Committee of Kütahya Health Sciences University with the code number of 2024/07–18 and a date of 17.05.2024.
The mean age of the participants (Min.-18, Max.-36) was X̄=21.42 ± 1.68. A total of 15.16% of the participants were studying in the Department of Physiotherapy and Rehabilitation, and 45.88% of them had families residing in the city (Table ). A total of 97.47% of the participants accessed the Internet on their mobile phones, 65.69% searched for online health information less than once a month, 39.50% stated that they would visit a physician when they had a health problem, and 23% would search the Internet (Table ). The mean digital health literacy score of the participants was X̄=2.93 ± 0.32. Among the subdimensions of digital health literacy, the highest mean was in the direction X̄=3,00 ± 0,51, and the lowest mean was in the reliability subdimension X̄=2,75 ± 0,57. The mean health anxiety score was X̄=17.19 ± 8.33, and the mean level of distrust in health services was X̄=2.95 ± 0.51 (Table ). A statistically significant, negative and moderate relationship was found between digital health literacy and health anxiety and distrust in health services (r˃0.500, p < 0.05) (Table ). The multiple regression model used to identify health anxiety and its determinants was statistically significant (F (5,746) = 11.967, p = 0.000). According to the results of the analysis, the independent variables (digital health literacy, distrust in the health system and online information seeking) explained 21.9% of the change in the dependent variable (health anxiety). These results indicate that health anxiety is negatively affected by digital health literacy and positively affected by distrust in the health system and online health information-seeking behavior (Table ). According to the standardized regression coefficient (β), the relative order of importance of the predictor variables in predicting health anxiety is as follows: digital health literacy (β = -0.276, t = -7.324), distrust in the health system (β = 0.049, t = 1.342) and seeking online health information behavior (β = 0.151, t = 4.289) (Table ).
In this study, the relationships between digital health literacy, distrust in the health system and health anxiety and the effects of digital health literacy and distrust in the health system on health anxiety were revealed. Since there are no studies in which these three concepts were evaluated together in previous research, the results of the study will be discussed with similar studies. The average age of the participants was 21.42 ± 1.68, 15.6% of them were studying in the Department of Physiotherapy and Rehabilitation, 97.47% of them provided internet access via mobile phone, and 23% of them searched for online health information when they had a health problem. In the study conducted by Batı et al. (2018), which examined the relationship between health anxiety and cyberchondria in health science students, the average age of the students was determined to be 20.33 ± 1.4, and 83.7% of the students provided Internet access from their mobile phones . In the studies conducted in the literature on the Internet health information-seeking behavior of health sciences students, the majority of students believe that accessing health resources on the Internet is very important and that health information on the Internet helps them make their health information-seeking behavior . These results suggest that university students should have good digital health literacy levels to access accurate information, and considering the importance of the study group, a high level of digital literacy is very important for both themselves and public health. The mean digital health literacy score of the participants was X̄=2.93 ± 0.32, the mean level of distrust in health systems was X̄=2.95 ± 0.51, and the mean level of health anxiety was X̄=17.19 ± 8.33. In the study conducted by Göde and Kuşcu (2022), in which they examined the relationship between distrust in the health system and e-pulseus with university students, the means of distrust in the health system were X̄=2.46 ± 0.67 and X̄=3.04 ± 0.66 in the study conducted by Kızılkaya (2024) in Türkiye . In the study conducted by Uslu-Sahan and Purtul (2023), the mean health anxiety score was X̄=18.93 ± 10.78, and in the study conducted by Baştürk et al. (2023) with university students, the mean health anxiety score was X̄=19.29 ± 7.25 . In studies in which digital health literacy was different from that in this study, the digital health literacy level of university students was higher than that in other segments of society, and in a study conducted with medical faculty students in Europe, 53.2% of the students had poor or very poor digital health literacy levels. 40. In a study conducted by Çetin and Gümüş (2023) in Türkiye with with 1274 adults between the ages of 18 and 64, the average digital health literacy was found to be X̄=3.04 ± 0.50 . In the study conducted by Frings et al. (2022), the difference in the level of digital health literacy between students receiving education and other students was determined . The fact that the levels of digital health literacy in studies on digital health literacy differ from each other can be explained by the fact that the selected samples are different from each other and that their demographic and sociocultural characteristics are different from each other. High levels of digital health literacy in all societies are associated with health protection, health promotion, chronic disease and healthy life development . In another study, a positive relationship was found between health anxiety and distrust in the health system, and a negative relationship was found between health anxiety and digital literacy. According to the results of multiple regression analysis, health anxiety was found to be negatively influenced by digital health literacy and positively influenced by distrust in the health system and online health information-seeking behavior. Age and sex were found to have no effect on health anxiety. In studies conducted with students in the literature, health literacy increases with increasing age . This may be explained by the increased experience of older students, who receive more health services and are more likely to communicate with health professionals . A study conducted by Sansakorn et al. (2024) in Pakistan revealed that health anxiety was positively affected by cyberchondria and negatively affected by health literacy . Previous studies have shown that online health information-seeking behavior positively predicts health anxiety because it increases the level of anxiety . The results of this study show that digital health literacy is the most important factor in reducing the health anxiety of individuals in this age of incredible advancement in technology. This situation becomes even more important, especially considering the importance of the group we studied. To the best of our knowledge, this is the first study to examine the relationship between digital health literacy, mistrust of the health system, and health anxiety in health science students in a cross-sectional manner. Therefore, it is believed that the results of the study may be effective in organizing digital health courses in the curriculum of health science students. In addition, it can be said that it is even more important to provide digital health education nowadays when the application of artificial intelligence (AI) in health education is rapidly increasing. Furthermore, the results of the study can also contribute to developing new policies regionally by comparing similar studies conducted in different regions of the world. As with all studies, this study has some limitations. First, although the sample size meets the adequacy criteria, the study was conducted on a relatively small group considering the number of health science students in Türkiye. Therefore, these results cannot be generalized to all students. Second, the results of the study may have been influenced by the subjective views and perceptions of the participants. Also, this study only focused on the level of digital health literacy, mistrust of the health system, and health anxiety among health science students at a university in Türkiye. Therefore, the results should be interpreted with caution. Third, it is difficult to conclude about the long term because it is a cross-sectional study. Finally, there is a possibility of selection bias because the study was conducted with an online survey method.
This study revealed that students’ digital health literacy, distrust in the health system and health anxiety levels were above average. In addition, health anxiety was affected by digital health literacy and distrust in the health system. The students’ level of distrust in the health system is above average because the negative news about the health system in Türkiye recently affects students as well as society in general. It is thought that there is a need to plan and evaluate the criteria of health literacy by taking into account the factors that are subject to increases in health anxiety and distrust in the health system. For this reason, policy makers should take the necessary measures before such negative situations arise, and people who are not experts in their field should be prevented from publishing on health-related issues in print and visual media. Moreover, students studying in health sciences constitute the most important human resource in the protection and development of health in the future. For this reason, students studying in these faculties should be taught courses on digital systems throughout their education life and should be provided with the necessary training on this subject.
Below is the link to the electronic supplementary material. Supplementary Material 1
|
Under the Microscope: A Case Report of Thoracic SMARCA4-Deficient Undifferentiated Tumor with Review of the Literature | ea3fc782-fb82-4d1d-982a-de84ba032a7e | 11131568 | Anatomy[mh] | SMARCA4-deficient thoracic undifferentiated tumors (SMARCA4-UT) represent a recently discovered and characterized pathological entity. This disorder was initially documented by Loarer et al. in 2015 . SMARCA4-UT is primarily distinguished by the inactivation of the SMARCA4 gene situated at 19p13, which encodes the Brahma related gene-1 (BRG1) protein. This protein is a critical constituent of the switch/sucrose-nonfermenting (SWI/SNF) chromatin remodeling complex. SMARCA4-UT is strongly associated with smoking and displays a molecular profile similar to that of smoking-related non-small cell lung carcinoma (NSCLC) and also exhibits a focal expression of NSCLC markers such as TTF1 and p40, as reported by Rekhtman et al. suggesting that SMARCA4 deficient thoracic sarcomas are conceptually similar to sarcomatoid carcinomas and have undergone epithelial-mesenchymal transition . SMARCA4-UT predominantly affects young and middle-aged individuals, with a slight male predominance. These tumors exclusively occur in the thoracic region and commonly present as masses in the mediastinum, lung, and/or pleura. The clinical prognosis of this disease is known to be unfavorable . Morphologically, the tumor displays an undifferentiated and/or rhabdoid phenotype and exhibits the expression of one or more stem cell markers, including CD34, SOX2, and SALL4. Thymic, lung, and mesothelial markers are absent, and there is a complete loss of BRG1 protein, as confirmed by immunohistochemistry (IHC) . Herein, we present a case report of SMARCA4-deficient thoracic undifferentiated tumor (SMARCA4-UT) that posed challenges in its diagnosis based on examination of the lymph node biopsy specimen. The encountered difficulties in utilizing a panel of immunohistochemistry (IHC) markers and conducting a comprehensive literature review to establish an accurate diagnosis for this particular case of SMARCA4-UT are presented. A 55-year-old gentleman presented with symptoms of difficulty in breathing, a dry cough, intermittent fever, and enlarged lymph nodes on the right side of the neck. The computed tomography (CT) and positron emission tomography (PET) scan revealed the presence of a necrotic mass in the upper lobe of the right lung, measuring approximately 22×21 mm, with the surrounding lung tissue showing subpleural fibrosis and bronchiectatic changes . Other multiple enlarged and necrotic lymph nodes were detected in the mediastinum (measuring 64×49 mm), neck (measuring 26×24 mm), and the left paraaortic region. An excision biopsy was performed on a right cervical lymph node, revealing a complete loss of the normal nodal architecture replaced by neoplastic cells arranged in dyscohesive sheets, accompanied by extensive areas of necrosis. The neoplastic cells exhibited a morphology ranging from epithelioid to undifferentiated cells, characterized by round to oval nuclei, vesicular chromatin, prominent nucleoli, and a moderate amount of eosinophilic cytoplasm with high mitotic count . Undifferentiated morphology raised the possibility of multiple differential diagnoses of poorly differentiated carcinoma, rhabdomyosarcoma, neuroendocrine carcinoma, melanoma, and lymphoma. Immunohistochemical (IHC) panel, including pan-cytokeratin (PCK), TTF1, p40, CD56, INSM1, Desmin, S100, SOX10, LCA, CD117, and CD30 markers, were tested, but they all yielded negative results. IHC staining for vimentin and synaptophysin was positive, CD34 and EMA showed very focal positivity in the tumor cells, while SMARCB1 (INI-1) was retained . The diagnosis was confirmed by SMARCA4 (BRG-1) IHC, which showed a complete loss of SMARCA4 protein in the neoplastic cells. The treatment protocol included chemotherapeutic agents, paclitaxel, and carboplatin. CT scan done after three months of treatment revealed a moderate reduction in the size of the lung lesion and a slight decrease in lymph node volume with an increase in necrotic tissue. Thoracic SMARCA4-UT tumors are rare neoplasms, with approximately 100 cases reported globally . They commonly manifest during the fourth and fifth decades of life, exhibiting a broad age range (27 to 90 years). Furthermore, these tumors show a notable male predilection, with a male-to-female ratio of 9:1, and are strongly associated with smoking . Most patients present with symptoms attributable to the mass effect and compression exerted by the tumor on neighbouring structures. These symptoms typically include pain, dyspnea, cough, haemoptysis, and superior vena cava syndrome . Given the aggressive nature of the neoplasm, initial presentation with metastasis to lymph nodes, skeletal bones, brain, or the abdominal cavity/pelvis is not an uncommon occurrence . Radiological imaging scans indicate that these masses predominantly occur in the upper and middle mediastinum, frequently involving adjacent structures such as the esophagus, bronchus, thymus, and major blood vessels. Moreover, contiguous involvement of the lung parenchyma is frequently observed . summarizes the clinicopathological features and survival data of cases in the published literature. As exemplified by the current case, the histomorphology of thoracic tumors, regardless of metastasis, exhibits a high-grade undifferentiated or epithelioid to rhabdoid cell phenotype, characterized by a relatively uniform dyscohesive arrangement in sheets, accompanied by brisk mitosis and necrosis. Regarding immunoprofiling, these tumors typically exhibit stem cell markers such as CD34, SALL4, and SOX2. There is usually either an absence or varying expression of PCK, EMA, and neuroendocrine markers, except for synaptophysin, which may show positivity. Additionally, focal expression of non-small cell lung cancer (NSCLC) markers like p63, p40, and TTF1 may be observed, while INI-1 expression remains intact. Immunonegativity is observed for calretinin, WT1, NUT, CD30, ALK, HMB-45, Desmin, and LCA. The characteristic feature of these tumors is the complete loss or significant underexpression of SMARCA4 (BRG-1), along with SMARCA2 (BRM) loss in more than 95% of the cases . Since the most common morphological presentation is as undifferentiated malignancy or with rhabdoid features, it warrants to rule out other mimics like carcinoma, lymphoma (large cell phenotype), germ cell tumor, melanoma, epithelioid sarcoma, and large cell neuroendocrine tumors. Differentiating these tumors on small biopsy is very challenging due to undifferentiated morphology and IHC promiscuity. As mentioned earlier, markers like CD34, PCK/EMA, and synaptophysin can be focally present and may not be helpful in small biopsies. Immunohistochemistry is of help if the before-mentioned markers are positive in a summative pattern. BRG1 IHC is helpful but it is not available at all centers. However, this antibody is now becoming essential in the repertoire of IHC panels to diagnose these neoplasms. Another point to remember is distinguishing SMARCA4-UT from SMARCA4-deficient non-small cell lung cancer (SMARCA4-NSCLC), as the latter is relatively more prevalent. These two entities can be differentiated based on distinct histomorphological characteristics. SMARCA4-NSCLC typically presents with clear-cut adenocarcinoma (AdCC) features or, less frequently, squamous cell carcinoma (SCC). Additionally, the expression of CK7 and the absence or focal expression of CD34, SOX2, and SALL4 are helpful markers for differentiation . The loss of SMARCA4 (BRG-1) can exceptionally occur in tumors such as SMARCB1/INI1-retained epithelioid sarcoma (ES); in this context, the absence of SOX2 and SALL4 expression aids in distinguishing ES from SMARCA4-UT . More prevalent entities such as large cell neuroendocrine carcinoma and small cell carcinoma may initially elicit misdiagnosis due to crush artifacts and tissue necrosis, along with the expression of synaptophysin and a high ki-67 index . These cases would derive an advantage from a composite approach involving a first-generation neuroendocrine marker (Synaptophysin and chromogranin) in conjunction with a second-generation neuroendocrine marker such as INSM1 and BRG-1 IHC loss to differentiate them from SMARCA4-UT. Additional diagnostic techniques include next generation sequencing (NGS) in identifying SMARCA4 mutations. However, in some cases where the loss of BRG-1 expression was observed using IHC, it was not detected by NGS. This could be because of structural variations (translocations) involving the intronic regions. Fluorescence in-situ hybridization (FISH) also has a limited role in diagnosis due to the truncating mutations coupled with the loss of heterozygosity (LOH), which is frequently copy-neutral (accompanied by the duplication of the mutant allele) . Therefore, it is crucial to use IHC to diagnose SMARCA4 mutations accurately. SMARCA4-UT are aggressive and are associated with a poor prognosis, with median overall survival ranging from 4 to 7 months . Improved outcomes have been reported in some instances of SMARCA4-UT treated with immunotherapy agents such as pembrolizumab , atezolizumab , and nivolumab . Notably, one documented case demonstrated a remarkable survival period of up to 22 months . SMARCA4-UT represents a rare subset of highly aggressive, poorly differentiated thoracic tumors primarily observed in middle-aged individuals with a history of smoking. These tumors are presently classified as epithelial tumors in the WHO classification of thoracic tumors . The diagnosis is established by integrating clinical and pathological features, with particular emphasis on undifferentiated morphology and loss of the BRG1 protein. Despite their unfavorable prognosis, there is potential for therapeutic interventions such as immunotherapy and SMARCA4-targeted therapies, offering promising prospects for the future. The authors declare that they have no conflict of interest for this article. |
Longitudinal urine metabolic profiling and gestational age prediction in human pregnancy | f01b447f-0881-47bf-a39e-8158c2ef4754 | 11830194 | Biochemistry[mh] | Accurate gestational age (GA) dating is essential for guiding prenatal care. Current methods, like using the last menstrual period, can be unreliable due to imprecise recall and symptoms such as early pregnancy bleeding that may be mistaken for a period . Although fetal ultrasound is the most precise method, it is limited by timing and resource availability and is most effective when performed before 20 weeks . Ultrasound also requires advanced equipment and skilled personnel , highlighting the need for a more accessible and precise GA dating method, especially in diverse socioeconomic contexts. Advances in omics technology offer new ways to explore the dynamic changes in pregnancy, capturing shifts in the maternal transcriptome, proteome, and metabolome . The metabolome, reflecting biochemical reactions, particularly responds to metabolic changes during pregnancy . Investigation of longitudinal maternal metabolomic alternations over the course of pregnancy has the potential to be a highly informative approach for mechanistic investigation and a breakthrough tool for GA dating. This approach has recently attracted more attention but has relied mostly on maternal blood samples . The use of maternal urine for GA dating and metabolic profiling has yet to be comprehensively explored, and it may provide a cost-effective and non-invasive method that could be easily translated into clinical settings. If found to be useful, it would transform prenatal care, especially in under-resourced regions. At present, the cost of implementing and maintaining liquid chromatography–mass spectrometry (LC–MS)-based untargeted metabolomics far exceeds that of ultrasound. In the future, as specific metabolite biomarkers are identified and validated, the clinical application of metabolomics could shift toward more cost-effective approaches, such as targeted metabolomics methods or simple immunoassays based on these biomarkers could be developed. These methods require significantly less infrastructure and expertise than LC–MS, making them more accessible and affordable, particularly in under-resourced settings. So far, research has focused on metabolic biomarkers for risks like preeclampsia and preterm birth (PTB) . However, metabolomic profiling throughout pregnancy could improve understanding of maternal metabolic changes, allowing for better risk stratification and insight into the pregnancy process . In this study, we analyzed longitudinal urine samples from 36 pregnant women, identifying numerous metabolites linked to pregnancy progression. We examined shifts in maternal metabolic pathways and developed a model to predict GA, identifying individualized metabolic alterations throughout pregnancy.
The SMART Diaphragm pregnancy cohort This observational study aimed to determine if the urine metabolome could identify metabolic changes and predict GA. We analyzed urine samples from 36 pregnant women recruited in San Francisco as part of the SMART Diaphragm (SMART-D) study . SMART-D developed a vaginal device to detect cervical changes for early PTB risk prediction. Samples, including urine and cervicovaginal swabs, were collected longitudinally during pregnancy and postpartum. For this study, at least one urine sample was collected per trimester from each participant, resulting in 3–13 samples per participant (median: 10). Participants in the SMART-D study represented diverse backgrounds. The 36 participants were of four races (Asian, Black, Pacific Islander, and White), aged 21–39 years. Pre-pregnancy BMI ranged from 19.5 to 57.2, and parity ranged from 1 to 9 . Detailed characteristics are in . Data is also shown at: http://47.100.52.12:3838/smartd-shiny/ . The urine metabolome accurately reflects metabolic alterations during pregnancy Untargeted metabolomics was performed, identifying 20 314 chemical signals (peaks; ). After removing 44 outlier samples, 302 samples remained for analysis . The batch effect has been largely mitigated, as indicated by the PCA score plot , confirming high data quality. The SMART-D study’s frequent sampling enabled detailed tracking of metabolome changes throughout pregnancy . PCA of peaks (QC RSD < 30%) showed a clear separation between early and late GA samples, with postpartum samples closely resembling early GA samples . Most participants followed this overall pattern . Significance analysis for microarrays (SAM) and linear regression identified 14.87% of peaks as significantly altered during pregnancy . Unsupervised k-means clustering revealed three clusters correlated with GA: cluster 1 (10–26 weeks), cluster 2 (26–32 weeks), and cluster 3 (32–42 weeks; ; ). Consistent with PCA, postpartum samples mostly fell into cluster 1. These findings confirm that the urine metabolome reliably reflects metabolic shifts during pregnancy. Alterations of functional metabolic network and pathways during pregnancy An important strength of this study is the high-density sampling, providing detailed insights into metabolic regulation at each pregnancy stage. Samples were assigned to 14 GA ranges based on sampling times, with each range including at least ten subjects and samples to ensure robust analysis . Altered peaks were identified using the Wilcoxon signed-rank test [false discovery rate (FDR)-adjusted P < .05] compared to the baseline . Notably, 84.83% of altered peaks remained significant across all subsequent GA ranges, indicating a consistent pattern of metabolic changes throughout pregnancy . The number of altered peaks significantly increased from early to late pregnancy , aligning with the PCA and k-means clustering results. After childbirth, the number of altered peaks dramatically decreased compared to baseline. Based on the number of altered metabolic signatures, pregnancy metabolic signals were classified into four distinct periods: 10–18 weeks, 18–26 weeks, 26–34 weeks, and 34–42 weeks. These findings correspond with the clustering patterns observed . To investigate changes in specific metabolic networks across GA ranges, altered peaks were analyzed using PIUMet , identifying a network of altered metabolites for each range. All annotated metabolites from PIUMet were then used to build a cross-sectional correlation network . The network included 160 nodes (metabolites) and 1148 edges (correlations), with 80.4% of annotated metabolites represented, suggesting dense interactions and a coordinated regulatory network for metabolic changes during pregnancy . Using community analysis based on edge betweenness centrality revealed 20 clusters with a modularity of 0.30. Seven clusters (> 5 modes) were selected for further analysis . These clusters retained 76.25% of nodes and 96.95% of edges from the original network, indicating that they captured most of the correlations. These clusters likely represent physiologically related and correlated metabolites during pregnancy. We analyzed alterations in the seven clusters during pregnancy by cluster and peak intensity . Only clusters 2 and 3 showed consistent changes at both levels. Cluster 2, the largest with 75 nodes and 963 edges, mainly includes lipids and lipid-like molecules (51/75, ), indicating a lipid-related regulatory network during pregnancy. The top pathways for cluster 2 metabolites were steroid hormone biosynthesis, ovarian steroidogenesis, cortisol synthesis, aldosterone synthesis, prolactin signaling, aldosterone-regulated sodium reabsorption, and bile secretion . Metabolite levels in cluster 2 increased throughout pregnancy, with rapid increases at weeks 18 and 26, aligning with periods defined in . Cluster 3, with five metabolites (3-methylguanine, 7-methylguanine, L-phenylalanine, asymmetric dimethylarginine, and (S)-3-hydroxy-N-methylcoclaurine), displayed similar trends. While no pathway mapped to more than one metabolite, four of five metabolites related to amino acid modification, suggesting cluster 3’s involvement in amino acid metabolism . Pathway enrichment analysis was conducted with PIUMet for each GA range to investigate pregnancy-related metabolic pathways further. Thirteen pathways showed enrichment in at least one GA range, with most increasing during pregnancy (FDR-adjusted P < .05, overlap ≥3; ). Six pathways were consistent at both the metabolite and pathway levels . Five of these six pathways overlapped with those in the regulated network of cluster 2. Prediction of gestational age using the urine metabolome Next, we explored whether the urine metabolome could estimate GA, which could improve prenatal and neonatal care in cases of uncertain dating. Urine samples were divided into training (16 subjects, 125 samples) and validation (20 subjects, 156 samples) datasets . Demographics and birth characteristics did not significantly differ between these datasets ( P > .05, ). A random forest (RF) prediction model was built using 28 selected peaks identified through the Boruta algorithm and peak shape filtering . The training dataset was utilized as the internal dataset to validate prediction accuracy using the bootstrap method. The root mean squared error (RMSE) between actual and predicted GA was found to be 2.35 weeks, and adjusted R 2 was 0.86 (Pearson correlation r = 0.93; P < 2.2 × 10 −6 ; ). External validation yielded an RMSE of 2.66 weeks and adjusted R 2 of 0.79 (r = 0.89; P < 2.2 × 10 −6 ; ), indicating the model was not overfitting. Overall, our results demonstrated that the urine metabolome may be useful for accurately predicting GA. The impact of patient demographics on prediction accuracy was also assessed. Maternal BMI, age, parity, and race were included with 28 peaks to construct a prediction model. The RMSE of this model was 2.70, and the adjusted R 2 was 0.76, which demonstrated no significant differences compared to the prediction model utilizing 28 peaks. The inclusion of subject demographics minimally improved prediction accuracy. Prediction of gestational age at the individual level Our study demonstrated that the pregnancy urine metabolome, using 28 peaks, can accurately predict GA. Peaks were annotated through an in-house MS2 pipeline , with 875 of 20 314 peaks annotated at level 1 or 2, though only 5 of the final 28 peaks were annotated . Using these 875 metabolites, we selected 32 for prediction, refining to 21 biomarkers after excluding those with poor peak shapes and MS 2 matches ( and ). Most biomarkers were lipids and lipid-like molecules , aligning with maternal plasma findings . Most biomarkers ranked highly in the prediction model . The 21 metabolite biomarkers achieved similar accuracy to the peak model, with adjusted R 2 of 0.81 (r = 0.90, P < 2.2 × 10 −6 ) for internal validation and 0.77 (r = 0.87, P < 2.2 × 10 −6 ) for external validation datasets . The RMSE was 2.89 weeks for internal and 2.97 weeks for external validation . A 1000-time permutation test confirmed no overfitting . Notably, model performance improved over pregnancy, with RMSE decreasing from the first to the third trimester for both training (4.71 in T1, 2.81 in T2, 2.82 in T3) and validation datasets (7.30 in T1, 3.14 in T2, 2.81 in T3; ). There was also no significant difference in accuracy between the metabolite and peak models, particularly in the validation dataset (RMSE = 2.97 versus 2.66 weeks, adjusted R 2 = 0.77 versus 0.79). These results indicate that urine metabolites can effectively predict GA and have promising clinical applications. When applied to individual participants in the external validation dataset, 16 out of 20 achieved an adjusted R 2 >0.75 , demonstrating the robustness of our model for individual predictions. Our cohort, which includes women with diverse demographic and clinical characteristics , suggests the model’s utility across varied backgrounds. We then assessed the impact of individual characteristics on prediction accuracy, calculating correlations between RMSE/adjusted R 2 and continuous variables. Surprisingly, continuous factors showed no significant correlation with prediction accuracy (all r < 0.5, all P > .05; ). Notably, three participants were outliers for birth weight, BMI, and parity. Participant S1760 had a BMI of 57.23 (mean: 27.09, P < .001) with high prediction accuracy (RMSE = 1.05, adjusted R 2 = 0.93; ). Participant S1762, with parity of nine (mean: 2.92, P < .001), also achieved good accuracy (RMSE = 2.94, adjusted R 2 = 0.90; ). For S1562, with a birth weight of 6.185 kg (mean: 3.397 kg, P < .001), internal validation accuracy was similarly high (RMSE = 3.70, adjusted R 2 = 0.95; ). We also examined whether categorical characteristics influenced prediction accuracy. Results indicated that prediction accuracy was unaffected by these factors ( , analysis of variance test, all P > .05). Overall, these findings demonstrate that the GA prediction model based on metabolite biomarkers is highly robust and adaptable to individual diversity. Prediction of time to delivery using the urine metabolome We next evaluated whether the urine metabolome could predict time-to-delivery, defined as the difference between GA at sample collection and GA at delivery, independent of ultrasound estimates. Participants with scheduled Cesarean sections were excluded, leaving 20 participants . The model included 21 metabolites, 18 of which overlapped with those in the GA prediction model . Predicted values aligned well with actual values in both the training (RMSE = 2.58 weeks; adjusted R 2 = 0.83; r = 0.94, P < 2.2 × 10 −6 ; ) and validation datasets (RMSE = 2.87; adjusted R 2 = 0.77; r = 0.88, P = 4.91 × 10 −15 ; ). A permutation test confirmed no overfitting . Prediction accuracy was also unaffected by patient demographics, similar to the GA model . These findings show that the time-to-delivery model is robust and effective across diverse individual characteristics. Altered metabolic signatures during pregnancy We also examined the biological function of the 24 metabolite markers. Most of the markers (9 of 24; 8 are unknown) were lipids and lipid-like molecules (primarily hormones, ; ), aligning with earlier findings at the peak level. To capture shifting metabolic signatures, hierarchical and fuzzy c-mean clustering grouped the 24 markers into two clusters with contrasting regulation patterns . The first group showed downregulation during pregnancy, returning to normal postpartum , while the second group increased with pregnancy and normalized postpartum . This second group was enriched in pathways related to glucocorticoid and mineralocorticoid biosynthesis, growth hormones, and lipid metabolism. Given progesterone’s clinical relevance in PTB treatment , other similarly regulated steroids may serve as diagnostic or therapeutic targets. Correlation analysis across GA periods revealed significant metabolomic shifts as the pregnancy progressed . Early pregnancy showed a positive correlation between metabolite intensity and GA, shifting to a negative correlation in later stages, indicating that urine metabolome alterations may help predict delivery timing. Further analysis showed positive correlations among many markers and delivery-related factors, such as maternal BMI and birth weight, suggesting co-regulated metabolic pathways, with the exception of a negative correlation between BMI and pregnenolone . This aligns with studies linking obesity to higher PTB risk , suggesting that pregnenolone levels might aid in GA prediction and preterm risk assessment. Although BMI showed no significant correlation with most lipid biomarkers (except BMI-pregnenolone, FDR adjusted P < .05), a non-significant trend of negative correlation with other lipids was observed .
This observational study aimed to determine if the urine metabolome could identify metabolic changes and predict GA. We analyzed urine samples from 36 pregnant women recruited in San Francisco as part of the SMART Diaphragm (SMART-D) study . SMART-D developed a vaginal device to detect cervical changes for early PTB risk prediction. Samples, including urine and cervicovaginal swabs, were collected longitudinally during pregnancy and postpartum. For this study, at least one urine sample was collected per trimester from each participant, resulting in 3–13 samples per participant (median: 10). Participants in the SMART-D study represented diverse backgrounds. The 36 participants were of four races (Asian, Black, Pacific Islander, and White), aged 21–39 years. Pre-pregnancy BMI ranged from 19.5 to 57.2, and parity ranged from 1 to 9 . Detailed characteristics are in . Data is also shown at: http://47.100.52.12:3838/smartd-shiny/ .
Untargeted metabolomics was performed, identifying 20 314 chemical signals (peaks; ). After removing 44 outlier samples, 302 samples remained for analysis . The batch effect has been largely mitigated, as indicated by the PCA score plot , confirming high data quality. The SMART-D study’s frequent sampling enabled detailed tracking of metabolome changes throughout pregnancy . PCA of peaks (QC RSD < 30%) showed a clear separation between early and late GA samples, with postpartum samples closely resembling early GA samples . Most participants followed this overall pattern . Significance analysis for microarrays (SAM) and linear regression identified 14.87% of peaks as significantly altered during pregnancy . Unsupervised k-means clustering revealed three clusters correlated with GA: cluster 1 (10–26 weeks), cluster 2 (26–32 weeks), and cluster 3 (32–42 weeks; ; ). Consistent with PCA, postpartum samples mostly fell into cluster 1. These findings confirm that the urine metabolome reliably reflects metabolic shifts during pregnancy.
An important strength of this study is the high-density sampling, providing detailed insights into metabolic regulation at each pregnancy stage. Samples were assigned to 14 GA ranges based on sampling times, with each range including at least ten subjects and samples to ensure robust analysis . Altered peaks were identified using the Wilcoxon signed-rank test [false discovery rate (FDR)-adjusted P < .05] compared to the baseline . Notably, 84.83% of altered peaks remained significant across all subsequent GA ranges, indicating a consistent pattern of metabolic changes throughout pregnancy . The number of altered peaks significantly increased from early to late pregnancy , aligning with the PCA and k-means clustering results. After childbirth, the number of altered peaks dramatically decreased compared to baseline. Based on the number of altered metabolic signatures, pregnancy metabolic signals were classified into four distinct periods: 10–18 weeks, 18–26 weeks, 26–34 weeks, and 34–42 weeks. These findings correspond with the clustering patterns observed . To investigate changes in specific metabolic networks across GA ranges, altered peaks were analyzed using PIUMet , identifying a network of altered metabolites for each range. All annotated metabolites from PIUMet were then used to build a cross-sectional correlation network . The network included 160 nodes (metabolites) and 1148 edges (correlations), with 80.4% of annotated metabolites represented, suggesting dense interactions and a coordinated regulatory network for metabolic changes during pregnancy . Using community analysis based on edge betweenness centrality revealed 20 clusters with a modularity of 0.30. Seven clusters (> 5 modes) were selected for further analysis . These clusters retained 76.25% of nodes and 96.95% of edges from the original network, indicating that they captured most of the correlations. These clusters likely represent physiologically related and correlated metabolites during pregnancy. We analyzed alterations in the seven clusters during pregnancy by cluster and peak intensity . Only clusters 2 and 3 showed consistent changes at both levels. Cluster 2, the largest with 75 nodes and 963 edges, mainly includes lipids and lipid-like molecules (51/75, ), indicating a lipid-related regulatory network during pregnancy. The top pathways for cluster 2 metabolites were steroid hormone biosynthesis, ovarian steroidogenesis, cortisol synthesis, aldosterone synthesis, prolactin signaling, aldosterone-regulated sodium reabsorption, and bile secretion . Metabolite levels in cluster 2 increased throughout pregnancy, with rapid increases at weeks 18 and 26, aligning with periods defined in . Cluster 3, with five metabolites (3-methylguanine, 7-methylguanine, L-phenylalanine, asymmetric dimethylarginine, and (S)-3-hydroxy-N-methylcoclaurine), displayed similar trends. While no pathway mapped to more than one metabolite, four of five metabolites related to amino acid modification, suggesting cluster 3’s involvement in amino acid metabolism . Pathway enrichment analysis was conducted with PIUMet for each GA range to investigate pregnancy-related metabolic pathways further. Thirteen pathways showed enrichment in at least one GA range, with most increasing during pregnancy (FDR-adjusted P < .05, overlap ≥3; ). Six pathways were consistent at both the metabolite and pathway levels . Five of these six pathways overlapped with those in the regulated network of cluster 2.
Next, we explored whether the urine metabolome could estimate GA, which could improve prenatal and neonatal care in cases of uncertain dating. Urine samples were divided into training (16 subjects, 125 samples) and validation (20 subjects, 156 samples) datasets . Demographics and birth characteristics did not significantly differ between these datasets ( P > .05, ). A random forest (RF) prediction model was built using 28 selected peaks identified through the Boruta algorithm and peak shape filtering . The training dataset was utilized as the internal dataset to validate prediction accuracy using the bootstrap method. The root mean squared error (RMSE) between actual and predicted GA was found to be 2.35 weeks, and adjusted R 2 was 0.86 (Pearson correlation r = 0.93; P < 2.2 × 10 −6 ; ). External validation yielded an RMSE of 2.66 weeks and adjusted R 2 of 0.79 (r = 0.89; P < 2.2 × 10 −6 ; ), indicating the model was not overfitting. Overall, our results demonstrated that the urine metabolome may be useful for accurately predicting GA. The impact of patient demographics on prediction accuracy was also assessed. Maternal BMI, age, parity, and race were included with 28 peaks to construct a prediction model. The RMSE of this model was 2.70, and the adjusted R 2 was 0.76, which demonstrated no significant differences compared to the prediction model utilizing 28 peaks. The inclusion of subject demographics minimally improved prediction accuracy.
Our study demonstrated that the pregnancy urine metabolome, using 28 peaks, can accurately predict GA. Peaks were annotated through an in-house MS2 pipeline , with 875 of 20 314 peaks annotated at level 1 or 2, though only 5 of the final 28 peaks were annotated . Using these 875 metabolites, we selected 32 for prediction, refining to 21 biomarkers after excluding those with poor peak shapes and MS 2 matches ( and ). Most biomarkers were lipids and lipid-like molecules , aligning with maternal plasma findings . Most biomarkers ranked highly in the prediction model . The 21 metabolite biomarkers achieved similar accuracy to the peak model, with adjusted R 2 of 0.81 (r = 0.90, P < 2.2 × 10 −6 ) for internal validation and 0.77 (r = 0.87, P < 2.2 × 10 −6 ) for external validation datasets . The RMSE was 2.89 weeks for internal and 2.97 weeks for external validation . A 1000-time permutation test confirmed no overfitting . Notably, model performance improved over pregnancy, with RMSE decreasing from the first to the third trimester for both training (4.71 in T1, 2.81 in T2, 2.82 in T3) and validation datasets (7.30 in T1, 3.14 in T2, 2.81 in T3; ). There was also no significant difference in accuracy between the metabolite and peak models, particularly in the validation dataset (RMSE = 2.97 versus 2.66 weeks, adjusted R 2 = 0.77 versus 0.79). These results indicate that urine metabolites can effectively predict GA and have promising clinical applications. When applied to individual participants in the external validation dataset, 16 out of 20 achieved an adjusted R 2 >0.75 , demonstrating the robustness of our model for individual predictions. Our cohort, which includes women with diverse demographic and clinical characteristics , suggests the model’s utility across varied backgrounds. We then assessed the impact of individual characteristics on prediction accuracy, calculating correlations between RMSE/adjusted R 2 and continuous variables. Surprisingly, continuous factors showed no significant correlation with prediction accuracy (all r < 0.5, all P > .05; ). Notably, three participants were outliers for birth weight, BMI, and parity. Participant S1760 had a BMI of 57.23 (mean: 27.09, P < .001) with high prediction accuracy (RMSE = 1.05, adjusted R 2 = 0.93; ). Participant S1762, with parity of nine (mean: 2.92, P < .001), also achieved good accuracy (RMSE = 2.94, adjusted R 2 = 0.90; ). For S1562, with a birth weight of 6.185 kg (mean: 3.397 kg, P < .001), internal validation accuracy was similarly high (RMSE = 3.70, adjusted R 2 = 0.95; ). We also examined whether categorical characteristics influenced prediction accuracy. Results indicated that prediction accuracy was unaffected by these factors ( , analysis of variance test, all P > .05). Overall, these findings demonstrate that the GA prediction model based on metabolite biomarkers is highly robust and adaptable to individual diversity.
We next evaluated whether the urine metabolome could predict time-to-delivery, defined as the difference between GA at sample collection and GA at delivery, independent of ultrasound estimates. Participants with scheduled Cesarean sections were excluded, leaving 20 participants . The model included 21 metabolites, 18 of which overlapped with those in the GA prediction model . Predicted values aligned well with actual values in both the training (RMSE = 2.58 weeks; adjusted R 2 = 0.83; r = 0.94, P < 2.2 × 10 −6 ; ) and validation datasets (RMSE = 2.87; adjusted R 2 = 0.77; r = 0.88, P = 4.91 × 10 −15 ; ). A permutation test confirmed no overfitting . Prediction accuracy was also unaffected by patient demographics, similar to the GA model . These findings show that the time-to-delivery model is robust and effective across diverse individual characteristics.
We also examined the biological function of the 24 metabolite markers. Most of the markers (9 of 24; 8 are unknown) were lipids and lipid-like molecules (primarily hormones, ; ), aligning with earlier findings at the peak level. To capture shifting metabolic signatures, hierarchical and fuzzy c-mean clustering grouped the 24 markers into two clusters with contrasting regulation patterns . The first group showed downregulation during pregnancy, returning to normal postpartum , while the second group increased with pregnancy and normalized postpartum . This second group was enriched in pathways related to glucocorticoid and mineralocorticoid biosynthesis, growth hormones, and lipid metabolism. Given progesterone’s clinical relevance in PTB treatment , other similarly regulated steroids may serve as diagnostic or therapeutic targets. Correlation analysis across GA periods revealed significant metabolomic shifts as the pregnancy progressed . Early pregnancy showed a positive correlation between metabolite intensity and GA, shifting to a negative correlation in later stages, indicating that urine metabolome alterations may help predict delivery timing. Further analysis showed positive correlations among many markers and delivery-related factors, such as maternal BMI and birth weight, suggesting co-regulated metabolic pathways, with the exception of a negative correlation between BMI and pregnenolone . This aligns with studies linking obesity to higher PTB risk , suggesting that pregnenolone levels might aid in GA prediction and preterm risk assessment. Although BMI showed no significant correlation with most lipid biomarkers (except BMI-pregnenolone, FDR adjusted P < .05), a non-significant trend of negative correlation with other lipids was observed .
Accurate GA estimation is essential for preventive prenatal care and timely interventions as maternal and fetal needs change throughout pregnancy . While metabolic changes in pregnancy have been extensively studied using blood samples , the comprehensive dynamics of the pregnancy urine metabolome remain less explored. In this study, we used comprehensive metabolic profiling of urine samples to better understand prenatal and postnatal metabolic changes in maternal urine. We developed models to predict GA and time-to-delivery with strong predictive accuracy at both cohort and individual levels (training: adjusted R 2 = 0.81, RMSE = 2.89; validation: R 2 = 0.77, RMSE = 2.97). The GA model tended to overestimate in early pregnancy and underestimate in later stages, possibly due to diverse biological processes that require further study in larger cohorts. Overall, our findings demonstrate the potential of the urine metabolome for accurate GA estimation and time-to-delivery prediction. Pregnenolone, progesterone, and corticoids were all upregulated in the glucocorticoid pathways during pregnancy, and related metabolites used in the time-to-delivery prediction model were enriched for glucocorticoid and CMP-N-acetylneuraminate biosynthesis pathways. These hormones have been reported to play key roles in pregnancy regulation . For instance, progesterone has been approved for the treatment of amenorrhea, metrorrhagia, and infertility . In our previous study of maternal plasma collected in an independent cohort, we also identified tetrahydrodeoxycorticosterone, estriol glucuronide, and progesterone as markers for GA estimation . Other identified derivatives in the same steroid hormone group of estrogens and progesterone derivatives, as well as uncharacterized steroid-like compounds discovered in this study, may also play roles in pregnancy, although their functions remain unclear. Furthermore, N-acetylmannosamine and N-acetylneuraminate were both significantly upregulated in the CMP-N-acetylneuraminate biosynthesis pathway, although the impact of these signaling molecules on pregnancy-related processes remains to be explored. As a proof of principle, our results show that urine metabolomic profiles can be used to track gestation throughout pregnancy. By applying an RF model, we successfully predicted GA based on 21 urine metabolites , including diverse glucocorticoids, lipids, glucuronide, and amino acid derivatives, indicating comprehensive regulation of glucocorticoid biosynthesis and CMP-N-acetylneuraminate biosynthesis by pregnancy. Compared to previous research , which employed targeted or restricted metabolomic approaches, our study utilizes untargeted LC–MS to capture a broader spectrum of metabolic features. By analyzing longitudinal samples across the entire pregnancy (11–40 weeks), we identified key metabolic shifts and constructed a GA prediction model. While prior studies demonstrated the feasibility of urinary metabolomics for GA estimation, our study advances this field by offering enhanced predictive performance and a more comprehensive understanding of pregnancy-associated metabolic changes. While our study provides valuable insights into the potential of urine metabolomics for GA prediction, our study still has several limitations that need to be addressed in future studies. First, the cohort size is relatively small (36 participants), which limits the generalizability of the findings. Future studies should aim to validate these findings in larger and more diverse cohorts to ensure broader applicability and to assess the robustness of the model across different populations and settings. Additionally, To mitigate overfitting risks, we used the Boruta algorithm for robust feature selection and employed cross-validation within the training dataset. However, the small cohort size and high-dimensional nature of metabolomics data could still pose overfitting risks, warranting further validation in larger, independent cohorts. Second, while our analysis identified significant metabolites and pathways associated with GA, and we took additional steps to enhance the quality of selected biomarkers by excluding metabolites with poor peak shapes or weak MS 2 matches, the biological roles of some unannotated metabolites remain unclear. This underscores the need for further research to characterize these unknown features and explore their potential contributions to pregnancy-related metabolic changes in urine samples. Future studies could leverage advanced algorithms and emerging methodologies to improve the characterization and annotation of these unannotated metabolites. Additionally, although only a small proportion of peaks were annotated, the metabolites selected for the predictive model underwent rigorous statistical validation using the Boruta algorithm and cross-validation to minimize overfitting and maximize reproducibility. We also excluded metabolites with poor peak shapes or weak MS 2 matches to enhance the quality of the selected biomarkers further. Third, while we controlled for key demographic and clinical factors such as BMI and parity, other potential confounders, including dietary intake, medication use, and environmental exposures, were not accounted for due to the lack of available data. These factors may influence metabolomic profiles and could contribute to variability in our findings. Future studies should aim to incorporate detailed dietary, medication, and environmental exposure data to comprehensively assess their effects on metabolomic profiles and improve the robustness of predictive models. Fourth, although normalization was applied to minimize variability, residual batch effects cannot be completely excluded, particularly given the complexity of untargeted metabolomics datasets. Advanced techniques may further reduce batch-related noise in future studies. Fifth, while the FDR-adjusted threshold minimizes false positives, it may still overlook subtle but biologically meaningful changes, particularly in high-dimensional datasets where signal-to-noise ratios can vary. Finally, different omics data provide unique insights into the human body, each capturing distinct aspects of biological processes. The findings presented here are based solely on metabolomics data. In the future, integrating metabolomics data with other omics layers, such as proteomics and genomics, could significantly enhance the accuracy and robustness of GA estimation. Indeed, several studies have demonstrated that combining metabolomics with other omics data in blood samples improves GA prediction accuracy . We anticipate that integrating multi-omics data from urine and other biological samples will not only enhance the precision of GA prediction but also provide a more comprehensive understanding of the molecular mechanisms underlying pregnancy progression. The characterized alterations in the maternal urine metabolome showed a strong correlation with GA and pregnancy progression. Accurate and convenient GA determination and time-to-delivery prediction can enhance fetal development monitoring and support timely interventions to improve maternal and infant health. Monitoring maternal metabolic changes and their association with GA could offer deeper insights into fetal development regulation and pregnancy disorders. The non-invasive and accessible nature of urine metabolomics makes it a valuable tool for GA assessment in diverse clinical settings, especially in resource-limited areas with restricted access to early prenatal care.
Participant enrolment and urine sample collection A total of 346 urine samples were collected from 36 ethnically diverse women throughout pregnancy (11.8–40.7 weeks) and postpartum . The information for ‘duction’, ‘child sex’, and ‘child weight’ is missing for 10 participants in our cohort due to incomplete data collection . Samples were collected longitudinally in two batches , with GA dating based on American Congress of Obstetricians and Gynecologists guidelines. Urine samples were collected as random spot urine samples at various time points during pregnancy and postpartum. No 24-hour urine collections were performed. Chemical material and internal standard preparation MS-grade solvents (water, methanol, acetonitrile) were from Fisher Scientific. MS-grade acetic acid and analytical-grade internal standards were from Sigma Aldrich. The internal standard mixture (acetyl-d3-carnitine, phenylalanine-3,3-d2, tiapride, trazodone, reserpine, phytosphingosine, and chlorpromazine) was diluted 1:50 with a 3:1 acetonitrile-water solution for HILIC and with water for RPLC. Urine sample preparation Thawed urine samples were centrifuged at 17 000 rcf for 10 minutes; 250 μL of supernatant was diluted with 750 μL of internal standard. After a 10-second vortex and another 10-minute centrifuge at 17 000 rpm (4°C), the supernatant was used for LC–MS. A pooled QC sample was injected every 10 samples to ensure consistent retention time and signal intensity. All samples were randomized during sample preparation and data acquisition to minimize bias introduced by batch effects or systematic trends. Liquid chromatography–mass spectrometry data acquisition A Hypersil GOLD HPLC and guard columns (Thermo Scientific, San Jose, CA) were used for RPLC. The mobile phases were 0.06% acetic acid in water (A) and methanol with 0.06% acetic acid (B), with a flow rate of 0.25 mL/min and backpressure of 120–160 bar at 99% phase A. A linear gradient from 1% to 80% phase B was applied >9–10 minutes. The column was heated to 60°C, and the sample injection volume was 5 μL. HILIC experiments were performed using a ZIC-HILIC column (Merck Millipore) and mobile phase solvents consisting of 10 mM ammonium acetate in 50/50 acetonitrile/water (A) and 10 mM ammonium acetate in 95/5 acetonitrile/water (B). Metabolites were eluted from the columns at 0.5 ml/min using a 1–50% phase A gradient >15 min. The Thermo Q Exactive Hybrid Quadrupole-Orbitrap Plus and Q Exactive mass spectrometers (Xcalibur, Thermo Scientific, San Jose, CA, USA) were operated in full MS-scan mode for data acquisition, covering an m/z range of 50 to 1000. For MS 1 acquisition, the instruments were set with a scan rate of ~1.5 Hz, a resolution of 10 000 (at m/z 127), an injection time of 50 ms, and an automatic gain control (AGC) target of 3 × 10 6 . MS/MS spectra for the QC sample were acquired using the top 10 parent ions, fragmented at normalized collision energies (NCE) of 25 and 50, with an injection time of 100 ms and an AGC target of 1 × 10 5 . Data processing and cleaning MS raw data were converted to .mzXML (MS 1 ) and .mgf (MS 2 ) formats using ProteoWizard . Peak detection and alignment were done via an in-house pipeline , producing an MS 1 peak table. Across the dataset, an average of 12.5% of data points per feature were missing prior to imputation. Features with >20% missing values in QC samples were excluded from the analysis to minimize the potential for bias. We also excluded samples with >50% missing values (outlier samples) and imputed remaining gaps using the k-nearest neighbors algorithm (KNN). We selected KNN imputation as it is a widely used method in metabolomics studies due to its ability to estimate missing values based on the similarity of observed features . Outlier samples were defined as the samples with >50% missing values. Outlier samples were removed to prevent potential biases in downstream statistical analyses and machine learning modeling. These samples exhibited aberrant metabolomic profiles likely caused by technical errors, such as incomplete sample preparation, instrument fluctuations, or contamination. Peak intensity was normalized by mean, with batch mean ratios applied for data integration. To account for variability in urine concentration, the mean normalization was applied to the dataset. This approach was selected due to its simplicity and effectiveness in controlling for systematic biases across samples. Although more specific methods, such as probabilistic quotient normalization (PQN), creatinine adjustment, or specific gravity correction, are available, we chose the mean normalization to maintain consistency with prior studies and ensure robust analysis across a diverse cohort . General statistical analysis and data visualization Most statistical analysis and data visualization were conducted in R (version 3.6.0, ). Before analysis, data were log10 transformed and auto-scaled. Categorical data are reported as counts and percentages, while continuous variables are presented as the mean ± standard deviation or standard error of the mean. Due to the right-skewed distribution of most peaks, nonparametric methods were used for statistical tests, with all P -values adjusted using the FDR. The FDR-adjusted p-value threshold of 0.05 was selected to balance sensitivity and specificity in identifying significant features while controlling for the high-dimensional nature of metabolomics data. This threshold is widely accepted in metabolomics and omics studies as a rigorous standard for multiple testing correction. SAM test and linear regression model to detect overall altered peaks during pregnancy To identify peaks that significantly changed with GA, we applied a SAM and a linear regression model . SAM uses permutations of repeated measurements to estimate the FDR for peaks exceeding an adjustable threshold. The linear regression model was built using the R ‘lm’ function, adjusting for acquisition batch, BMI, maternal age, parity, and race. Peaks with an FDR-adjusted P < .05 were considered significant. K means consensus-clustering Unsupervised K-means consensus clustering was conducted using the R packages CancerSubtypes and ConsensusClusterPlus. Sample clusters were identified with K-means clustering, Euclidean distance, and 1000 resampling repetitions across 2 to 6 clusters. The empirical cumulative distribution function plot suggested 2 or 3 clusters as optimal for all urine samples. Consensus matrix heatmaps also indicated that 2, 3, and 4 clusters showed good separation. To determine the optimal cluster number, we extracted silhouette scores using the silhouette_SimilarityMatrix function. Comparing k = 2, 3, and 4, we found that k = 3 provided the highest clustering stability . PIUMet analysis PIUMet, a network-based tool, links peaks to potential metabolites and related dysregulated molecular mechanisms, effectively converting peak data into network information . For each GA range, altered peaks were saved as .txt files and uploaded to the PIUMet website. Results from PIUMet were then processed through an in-house pipeline, where annotation results across GA ranges were combined. Peaks appearing in fewer than two GA ranges were removed, and mean values of matched peaks were used as quantitative values for each metabolite. Correlation network and community analysis For the combined dataset of metabolites and clinical variables, correlations between variables were calculated, and only pairs with an absolute correlation >0.5 and FDR-adjusted P < 0.05 were used to construct correlation networks. Community analysis was performed with edge betweenness-based detection (Girvan–Newman method). This method iteratively removes edges with the highest edge betweenness, which likely connect separate modules until individual nodes remain. The result is a dendrogram, with leaves as individual nodes and the root as the whole network. Modularity analysis was used to identify communities within the network, maximizing modularity at each iteration to identify statistically significant structures . Only clusters with at least three nodes were retained for further analysis. Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis Pathway enrichment analysis was conducted using the KEGG database, a widely used resource in metabolomics, which initially contains 275 pathways divided into metabolic and disease pathways based on the ‘Class’ information. Enrichment analysis was performed using a hypergeometric distribution test, with P -values adjusted by the FDR method and a significance cutoff of .05. Metabolite annotation The metabolite annotation was performed using the metID package . For the metabolite annotated using the in-house database , the annotations are level 1 according to MSI . For metabolites annotated using the public databases, the annotation is level 2, according to MSI . The random forest prediction model Feature selection. The Boruta algorithm was used to identify biomarkers by creating shadow features through shuffling and then comparing them to real features via RF classification, repeated 100 times for robustness. Selected features served as potential biomarkers for the RF model. Parameter optimization. Default settings were used except for ntree (trees) and mtry (variables per split) in the RF model, optimized in the training set by minimizing MSE. GA prediction model. Samples from batch 1 (16 subjects, 125 samples) were used for training, and batch 2 (20 subjects, 156 samples) for validation. Feature selection was performed on the training dataset, followed by the construction of the RF model. A linear regression model was then applied between predicted and actual GA to correct predictions . This two-part model was validated on external samples, with RMSE and adjusted R 2 used to assess accuracy. 1000 bootstraps resampled 63% of training data, using 37% for validation; final GA predictions averaged per sample, calculating MSE and adjusted R 2 . Time-to-delivery prediction model. Calculated as weeks between sample and delivery dates, using the same approach as the GA model. Permutation test First, responses (GA or time to delivery) were randomly shuffled in the training and validation datasets. Potential biomarkers were selected, and RF parameters were optimized in the training dataset. An RF model was then built with selected features and applied to the validation dataset, yielding null RMSE and adjusted R 2 values. This process was repeated 1000 times to generate distributions of null RMSE and adjusted R 2 values. Using maximum likelihood estimation, these values were modeled with a Gamma distribution, and cumulative distribution functions were calculated to derive p-values for the actual RMSE and adjusted R 2 . Fuzzy c-means clustering The fuzzy c-means clustering algorithm was used to classify metabolite biomarkers by GA. Samples were grouped into time ranges from 11 to 41 weeks in two-week increments, with postpartum samples assigned to a ‘PP’ group. For each time range, the mean intensity of each metabolite was calculated, creating a new dataset with 16 observations. First, we optimized the parameter ‘m’ using Mfuzz and determined the optimal cluster number based on the within-cluster sum of squared errors. Using default parameters, fuzzy c-means clustering was performed, and only features with a membership score > 0.5 were retained. No smoothing was applied to the output results. Key points This study applied Liquid chromatography–mass spectrometry-based untargeted metabolomics and metabolic peak-based pathway analysis to analyze longitudinal changes in maternal urine, highlighting significant alterations in glucocorticoids, lipids, and amino acid derivatives during pregnancy. A machine learning model based on urinary metabolites accurately predicts gestational age (GA), offering a non-invasive pregnancy dating method that is especially useful in resource-limited settings. The findings underscore the potential clinical utility of urine metabolomics in improving GA estimation and monitoring maternal metabolic health. The study provides a detailed, comprehensive characterization of the maternal urine metabolome, which could enhance prenatal care and maternal–fetal health management.
A total of 346 urine samples were collected from 36 ethnically diverse women throughout pregnancy (11.8–40.7 weeks) and postpartum . The information for ‘duction’, ‘child sex’, and ‘child weight’ is missing for 10 participants in our cohort due to incomplete data collection . Samples were collected longitudinally in two batches , with GA dating based on American Congress of Obstetricians and Gynecologists guidelines. Urine samples were collected as random spot urine samples at various time points during pregnancy and postpartum. No 24-hour urine collections were performed.
MS-grade solvents (water, methanol, acetonitrile) were from Fisher Scientific. MS-grade acetic acid and analytical-grade internal standards were from Sigma Aldrich. The internal standard mixture (acetyl-d3-carnitine, phenylalanine-3,3-d2, tiapride, trazodone, reserpine, phytosphingosine, and chlorpromazine) was diluted 1:50 with a 3:1 acetonitrile-water solution for HILIC and with water for RPLC.
Thawed urine samples were centrifuged at 17 000 rcf for 10 minutes; 250 μL of supernatant was diluted with 750 μL of internal standard. After a 10-second vortex and another 10-minute centrifuge at 17 000 rpm (4°C), the supernatant was used for LC–MS. A pooled QC sample was injected every 10 samples to ensure consistent retention time and signal intensity. All samples were randomized during sample preparation and data acquisition to minimize bias introduced by batch effects or systematic trends.
A Hypersil GOLD HPLC and guard columns (Thermo Scientific, San Jose, CA) were used for RPLC. The mobile phases were 0.06% acetic acid in water (A) and methanol with 0.06% acetic acid (B), with a flow rate of 0.25 mL/min and backpressure of 120–160 bar at 99% phase A. A linear gradient from 1% to 80% phase B was applied >9–10 minutes. The column was heated to 60°C, and the sample injection volume was 5 μL. HILIC experiments were performed using a ZIC-HILIC column (Merck Millipore) and mobile phase solvents consisting of 10 mM ammonium acetate in 50/50 acetonitrile/water (A) and 10 mM ammonium acetate in 95/5 acetonitrile/water (B). Metabolites were eluted from the columns at 0.5 ml/min using a 1–50% phase A gradient >15 min. The Thermo Q Exactive Hybrid Quadrupole-Orbitrap Plus and Q Exactive mass spectrometers (Xcalibur, Thermo Scientific, San Jose, CA, USA) were operated in full MS-scan mode for data acquisition, covering an m/z range of 50 to 1000. For MS 1 acquisition, the instruments were set with a scan rate of ~1.5 Hz, a resolution of 10 000 (at m/z 127), an injection time of 50 ms, and an automatic gain control (AGC) target of 3 × 10 6 . MS/MS spectra for the QC sample were acquired using the top 10 parent ions, fragmented at normalized collision energies (NCE) of 25 and 50, with an injection time of 100 ms and an AGC target of 1 × 10 5 .
MS raw data were converted to .mzXML (MS 1 ) and .mgf (MS 2 ) formats using ProteoWizard . Peak detection and alignment were done via an in-house pipeline , producing an MS 1 peak table. Across the dataset, an average of 12.5% of data points per feature were missing prior to imputation. Features with >20% missing values in QC samples were excluded from the analysis to minimize the potential for bias. We also excluded samples with >50% missing values (outlier samples) and imputed remaining gaps using the k-nearest neighbors algorithm (KNN). We selected KNN imputation as it is a widely used method in metabolomics studies due to its ability to estimate missing values based on the similarity of observed features . Outlier samples were defined as the samples with >50% missing values. Outlier samples were removed to prevent potential biases in downstream statistical analyses and machine learning modeling. These samples exhibited aberrant metabolomic profiles likely caused by technical errors, such as incomplete sample preparation, instrument fluctuations, or contamination. Peak intensity was normalized by mean, with batch mean ratios applied for data integration. To account for variability in urine concentration, the mean normalization was applied to the dataset. This approach was selected due to its simplicity and effectiveness in controlling for systematic biases across samples. Although more specific methods, such as probabilistic quotient normalization (PQN), creatinine adjustment, or specific gravity correction, are available, we chose the mean normalization to maintain consistency with prior studies and ensure robust analysis across a diverse cohort .
Most statistical analysis and data visualization were conducted in R (version 3.6.0, ). Before analysis, data were log10 transformed and auto-scaled. Categorical data are reported as counts and percentages, while continuous variables are presented as the mean ± standard deviation or standard error of the mean. Due to the right-skewed distribution of most peaks, nonparametric methods were used for statistical tests, with all P -values adjusted using the FDR. The FDR-adjusted p-value threshold of 0.05 was selected to balance sensitivity and specificity in identifying significant features while controlling for the high-dimensional nature of metabolomics data. This threshold is widely accepted in metabolomics and omics studies as a rigorous standard for multiple testing correction.
To identify peaks that significantly changed with GA, we applied a SAM and a linear regression model . SAM uses permutations of repeated measurements to estimate the FDR for peaks exceeding an adjustable threshold. The linear regression model was built using the R ‘lm’ function, adjusting for acquisition batch, BMI, maternal age, parity, and race. Peaks with an FDR-adjusted P < .05 were considered significant.
Unsupervised K-means consensus clustering was conducted using the R packages CancerSubtypes and ConsensusClusterPlus. Sample clusters were identified with K-means clustering, Euclidean distance, and 1000 resampling repetitions across 2 to 6 clusters. The empirical cumulative distribution function plot suggested 2 or 3 clusters as optimal for all urine samples. Consensus matrix heatmaps also indicated that 2, 3, and 4 clusters showed good separation. To determine the optimal cluster number, we extracted silhouette scores using the silhouette_SimilarityMatrix function. Comparing k = 2, 3, and 4, we found that k = 3 provided the highest clustering stability .
PIUMet, a network-based tool, links peaks to potential metabolites and related dysregulated molecular mechanisms, effectively converting peak data into network information . For each GA range, altered peaks were saved as .txt files and uploaded to the PIUMet website. Results from PIUMet were then processed through an in-house pipeline, where annotation results across GA ranges were combined. Peaks appearing in fewer than two GA ranges were removed, and mean values of matched peaks were used as quantitative values for each metabolite.
For the combined dataset of metabolites and clinical variables, correlations between variables were calculated, and only pairs with an absolute correlation >0.5 and FDR-adjusted P < 0.05 were used to construct correlation networks. Community analysis was performed with edge betweenness-based detection (Girvan–Newman method). This method iteratively removes edges with the highest edge betweenness, which likely connect separate modules until individual nodes remain. The result is a dendrogram, with leaves as individual nodes and the root as the whole network. Modularity analysis was used to identify communities within the network, maximizing modularity at each iteration to identify statistically significant structures . Only clusters with at least three nodes were retained for further analysis.
Pathway enrichment analysis was conducted using the KEGG database, a widely used resource in metabolomics, which initially contains 275 pathways divided into metabolic and disease pathways based on the ‘Class’ information. Enrichment analysis was performed using a hypergeometric distribution test, with P -values adjusted by the FDR method and a significance cutoff of .05.
The metabolite annotation was performed using the metID package . For the metabolite annotated using the in-house database , the annotations are level 1 according to MSI . For metabolites annotated using the public databases, the annotation is level 2, according to MSI .
Feature selection. The Boruta algorithm was used to identify biomarkers by creating shadow features through shuffling and then comparing them to real features via RF classification, repeated 100 times for robustness. Selected features served as potential biomarkers for the RF model. Parameter optimization. Default settings were used except for ntree (trees) and mtry (variables per split) in the RF model, optimized in the training set by minimizing MSE. GA prediction model. Samples from batch 1 (16 subjects, 125 samples) were used for training, and batch 2 (20 subjects, 156 samples) for validation. Feature selection was performed on the training dataset, followed by the construction of the RF model. A linear regression model was then applied between predicted and actual GA to correct predictions . This two-part model was validated on external samples, with RMSE and adjusted R 2 used to assess accuracy. 1000 bootstraps resampled 63% of training data, using 37% for validation; final GA predictions averaged per sample, calculating MSE and adjusted R 2 . Time-to-delivery prediction model. Calculated as weeks between sample and delivery dates, using the same approach as the GA model.
First, responses (GA or time to delivery) were randomly shuffled in the training and validation datasets. Potential biomarkers were selected, and RF parameters were optimized in the training dataset. An RF model was then built with selected features and applied to the validation dataset, yielding null RMSE and adjusted R 2 values. This process was repeated 1000 times to generate distributions of null RMSE and adjusted R 2 values. Using maximum likelihood estimation, these values were modeled with a Gamma distribution, and cumulative distribution functions were calculated to derive p-values for the actual RMSE and adjusted R 2 .
The fuzzy c-means clustering algorithm was used to classify metabolite biomarkers by GA. Samples were grouped into time ranges from 11 to 41 weeks in two-week increments, with postpartum samples assigned to a ‘PP’ group. For each time range, the mean intensity of each metabolite was calculated, creating a new dataset with 16 observations. First, we optimized the parameter ‘m’ using Mfuzz and determined the optimal cluster number based on the within-cluster sum of squared errors. Using default parameters, fuzzy c-means clustering was performed, and only features with a membership score > 0.5 were retained. No smoothing was applied to the output results. Key points This study applied Liquid chromatography–mass spectrometry-based untargeted metabolomics and metabolic peak-based pathway analysis to analyze longitudinal changes in maternal urine, highlighting significant alterations in glucocorticoids, lipids, and amino acid derivatives during pregnancy. A machine learning model based on urinary metabolites accurately predicts gestational age (GA), offering a non-invasive pregnancy dating method that is especially useful in resource-limited settings. The findings underscore the potential clinical utility of urine metabolomics in improving GA estimation and monitoring maternal metabolic health. The study provides a detailed, comprehensive characterization of the maternal urine metabolome, which could enhance prenatal care and maternal–fetal health management.
This study applied Liquid chromatography–mass spectrometry-based untargeted metabolomics and metabolic peak-based pathway analysis to analyze longitudinal changes in maternal urine, highlighting significant alterations in glucocorticoids, lipids, and amino acid derivatives during pregnancy. A machine learning model based on urinary metabolites accurately predicts gestational age (GA), offering a non-invasive pregnancy dating method that is especially useful in resource-limited settings. The findings underscore the potential clinical utility of urine metabolomics in improving GA estimation and monitoring maternal metabolic health. The study provides a detailed, comprehensive characterization of the maternal urine metabolome, which could enhance prenatal care and maternal–fetal health management.
Table_S1_bbaf059 Table_S2_bbaf059 Table_S3_bbaf059 Table_S4_bbaf059 Table_S5_bbaf059 Table_S6_bbaf059 supplementary_bbaf059
|
Assessment of the operation of a pilot program of preventive medicine for adults in three primary care centers of Athens and Piraeus: a cross-sectional study | 29d255af-cf2a-4130-84ce-26afc8501442 | 7888401 | Preventive Medicine[mh] | In Greece evidence shows that metabolic syndrome (MetS) affects 20-30% of the adult population , meanwhile 98 per 100,000 persons die each year from chronic heart disease and 200 per 100,000 persons die from malignant neoplasms . Lifestyle and family history have been shown to contribute to the development of both acute and chronic disease . Health Screening Programs are the most important aspects of prevention, representing basic tools of modern public health, aiming at early detection of disease . Several screening programs have been implemented over the years worldwide for targeting risk factors and specific populations at risk . Preventive medicine and screening have gained a growing interest in Greece with the Ministry of Health launching in 2008 the National Plan of Action on Public Health, aiming at the effective protection and health promotion, through measures, programs, structures and new prevention strategies . In Greece there are organizational barriers in the provision of screening services, with only a small percentage of the Greek population receiving screening . There are only few nationwide screening programs for cancer or chronic diseases and few more local programs in rural areas , so preventive screening uptake is low, making timely treatment problematic. Due to the lack of a nationally organized invitational screening programs, screening is carried out mostly on the advice of general practitioners (GP) or based on the individual’s request. In this context, in 2009 the pilot Program of Preventive Medicine for Adults (PPMA) was established by the National Organization for Healthcare Provision (EOPYY) providing primary health care to citizens aged 40-55 years. The program aims to raise awareness among citizens regarding annual screening for the early diagnosis of diseases such as heart disease, malignant neoplasms, metabolic syndrome and their effective management. To our knowledge, no national or international study to date has evaluated the operation of the PPMA and aimed to analyse which from the health services offered within this primary health care program of EOPYY are important for its evaluation. Program evaluation permits to identify program strengths, weaknesses, and areas for improvement. DESIGN An observational, cross-sectional study was undertaken to evaluate the operation of the pilot PPMA. The study took place in primary health care centers of Athens and Piraeus (primary health care centers of Kallithea and Peristeri in Athens and Agia Sofia in Piraeus), densely populated districts of the capital and the largest seaport of Greece, covering approximately 300,000 citizens (according to the 2011 census). A questionnaire was specifically developed for data collection. ETHICAL CONSIDERATIONS The study was approved by the Department of Prevention and Health Promotion of EOPYY (protocol number 3917). DESCRIPTION OF PILOT PROGRAM OF PREVENTIVE MEDICINE FOR ADULTS (PPMA) Each primary health care center had a physician or a cardiologist, a midwife and a health visitor (healthcare professional who deals mainly with the protection and promotion of health for the most vulnerable population groups) or nurse. The midwife and health visitor or nurse worked mainly by protocol. The doctor met with participants for counselling and health promotion activities, with the exception of the Kallithea health care center where the physician had minimal involvement unless a medical problem occurred. The program offers participants annual screening including clinical assessments and laboratory tests. Eligible participants should be between 40-55 years of age, overall healthy and should not have undergone any type of screening in the past 12 months. Exclusion criteria were a recorded diagnosis of hypertension, diabetes, cardiovascular disease, thyroidopathy, hyperlipidaemia, prostatitis and anaemia. Participants who met the inclusion criteria were scheduled to undergo the recommended screening tests which included: Cell Blood Count (CBC), urine analysis, erythrocyte sedimentation rate (ESR), thyroid stimulating hormone (TSH), lipid panel screening, urea, fasting blood sugar and electrocardiogram. Women also received a mammography and Pap smear and men received a prostate-specific antigen test (PSA). Demographic characteristics, anthropometric, clinical data and test results for all participants were obtained by a health visitor/nurse then registered on the participant’s medical record so that the doctor could make a medical evaluation. If during the screening a medical problem was discovered the participant was referred to the appropriate specialty for further evaluation and treatment . PARTICIPANT RECRUITMENT The target population comprised participants who were already taking part in the program and had completed the entire screening process. Participants were informed about the program only through announcements in the primary health care centers or through relatives and friends. Assuming an effect size (f 2 ) = 0.2, α error probability = 5%, statistical power = 95% and number of predictors = 6, an a priori power analysis using the G power software (version 3.1) revealed that data would have to be collected from N = 111 participants. Overall 149 people had verbally consented to participate in this study, 7 participants left the questionnaires incomplete. In total, 142 respondents, 50 participants from Kallithea, 50 from the Peristeri health care center and 42 from Agia Sophia, who completed the questionnaire were included in quantitative analysis. The overall response rate was 95.3%. PROCEDURE AND DATA COLLECTION A structured questionnaire for patients was developed to evaluate primary care in a number of key areas ranging from the access to care, the helpfulness of visitors/nurses, the doctors’ communication skills and overall program evaluation on validated questionnaires used in primary care in Greece, UK, US and Europe . To increase confidence a pre-study evaluation of the questionnaire (with 5 patients, 2 nurses and 1 physician) was performed to test validity in terms of language and understanding. The questionnaire consisted of 37 questions and was divided into four sections: socio-demographic features, ease of access to the program, attitude of health care personnel and general program evaluation. The majority were closed-ended questions using five-point Likert scale, ranging from 1 (very dissatisfied), 2 (dissatisfied), 3 (undecided), 4 (satisfied), 5 (very satisfied), so that higher scores indicate greater satisfaction. The remainder were in yes/no format and multiple choice questions based on guidelines reported in other studies . Evaluation of health visitor/nurse and evaluation of doctor is defined as the participant’s opinion about the existence of good communication, courtesy and concern of the health visitor/nurse and the doctor with participant’s questions/worries. The evaluation of the pilot PPMA is defined as the participant’s opinion about organizational structure, accessibility to the program, physical environment, ease of arranging appointment(s) and good service attitude. Each participant who had completed the entire screening process was interviewed after visiting the physician. The interviews took 10 to 15 minutes on average to complete and were scheduled on specific days. The primary author (T.M) of this study collected the data independently. No monetary benefit was given to the respondents for participation in this study. STATISTICAL ANALYSIS Statistical analysis was carried out with SPSS (Version 22.0. Chicago: SPSS Inc.). A descriptive analysis for categorical variables was carried out using Pearson’s Chi-square test, to evaluate the associations between each health care center. The results are presented as counts and percentages. Continuous variables were expressed as means and standard deviation (SD) using analysis of variance (ANOVA) for measuring the difference between means. To identify which variable(s) is the best predictor of the evaluation of the pilot PPMA, a standard multiple linear regression analysis was performed including variables whose association were significant at a value P < 0.2. After excluding collinearity, the best subset of variables was selected. The predictor (independent) variables for this study were demographic characteristics (age and gender), attention and interest of health visitors/nurse, evaluation of doctor, communication of results to participants and evaluation of health visitor/nurse. The evaluation of the pilot PPMA was the dependent variable. The statistically significant threshold was set at .05 probability value. An observational, cross-sectional study was undertaken to evaluate the operation of the pilot PPMA. The study took place in primary health care centers of Athens and Piraeus (primary health care centers of Kallithea and Peristeri in Athens and Agia Sofia in Piraeus), densely populated districts of the capital and the largest seaport of Greece, covering approximately 300,000 citizens (according to the 2011 census). A questionnaire was specifically developed for data collection. The study was approved by the Department of Prevention and Health Promotion of EOPYY (protocol number 3917). Each primary health care center had a physician or a cardiologist, a midwife and a health visitor (healthcare professional who deals mainly with the protection and promotion of health for the most vulnerable population groups) or nurse. The midwife and health visitor or nurse worked mainly by protocol. The doctor met with participants for counselling and health promotion activities, with the exception of the Kallithea health care center where the physician had minimal involvement unless a medical problem occurred. The program offers participants annual screening including clinical assessments and laboratory tests. Eligible participants should be between 40-55 years of age, overall healthy and should not have undergone any type of screening in the past 12 months. Exclusion criteria were a recorded diagnosis of hypertension, diabetes, cardiovascular disease, thyroidopathy, hyperlipidaemia, prostatitis and anaemia. Participants who met the inclusion criteria were scheduled to undergo the recommended screening tests which included: Cell Blood Count (CBC), urine analysis, erythrocyte sedimentation rate (ESR), thyroid stimulating hormone (TSH), lipid panel screening, urea, fasting blood sugar and electrocardiogram. Women also received a mammography and Pap smear and men received a prostate-specific antigen test (PSA). Demographic characteristics, anthropometric, clinical data and test results for all participants were obtained by a health visitor/nurse then registered on the participant’s medical record so that the doctor could make a medical evaluation. If during the screening a medical problem was discovered the participant was referred to the appropriate specialty for further evaluation and treatment . The target population comprised participants who were already taking part in the program and had completed the entire screening process. Participants were informed about the program only through announcements in the primary health care centers or through relatives and friends. Assuming an effect size (f 2 ) = 0.2, α error probability = 5%, statistical power = 95% and number of predictors = 6, an a priori power analysis using the G power software (version 3.1) revealed that data would have to be collected from N = 111 participants. Overall 149 people had verbally consented to participate in this study, 7 participants left the questionnaires incomplete. In total, 142 respondents, 50 participants from Kallithea, 50 from the Peristeri health care center and 42 from Agia Sophia, who completed the questionnaire were included in quantitative analysis. The overall response rate was 95.3%. A structured questionnaire for patients was developed to evaluate primary care in a number of key areas ranging from the access to care, the helpfulness of visitors/nurses, the doctors’ communication skills and overall program evaluation on validated questionnaires used in primary care in Greece, UK, US and Europe . To increase confidence a pre-study evaluation of the questionnaire (with 5 patients, 2 nurses and 1 physician) was performed to test validity in terms of language and understanding. The questionnaire consisted of 37 questions and was divided into four sections: socio-demographic features, ease of access to the program, attitude of health care personnel and general program evaluation. The majority were closed-ended questions using five-point Likert scale, ranging from 1 (very dissatisfied), 2 (dissatisfied), 3 (undecided), 4 (satisfied), 5 (very satisfied), so that higher scores indicate greater satisfaction. The remainder were in yes/no format and multiple choice questions based on guidelines reported in other studies . Evaluation of health visitor/nurse and evaluation of doctor is defined as the participant’s opinion about the existence of good communication, courtesy and concern of the health visitor/nurse and the doctor with participant’s questions/worries. The evaluation of the pilot PPMA is defined as the participant’s opinion about organizational structure, accessibility to the program, physical environment, ease of arranging appointment(s) and good service attitude. Each participant who had completed the entire screening process was interviewed after visiting the physician. The interviews took 10 to 15 minutes on average to complete and were scheduled on specific days. The primary author (T.M) of this study collected the data independently. No monetary benefit was given to the respondents for participation in this study. Statistical analysis was carried out with SPSS (Version 22.0. Chicago: SPSS Inc.). A descriptive analysis for categorical variables was carried out using Pearson’s Chi-square test, to evaluate the associations between each health care center. The results are presented as counts and percentages. Continuous variables were expressed as means and standard deviation (SD) using analysis of variance (ANOVA) for measuring the difference between means. To identify which variable(s) is the best predictor of the evaluation of the pilot PPMA, a standard multiple linear regression analysis was performed including variables whose association were significant at a value P < 0.2. After excluding collinearity, the best subset of variables was selected. The predictor (independent) variables for this study were demographic characteristics (age and gender), attention and interest of health visitors/nurse, evaluation of doctor, communication of results to participants and evaluation of health visitor/nurse. The evaluation of the pilot PPMA was the dependent variable. The statistically significant threshold was set at .05 probability value. CHARACTERISTICS OF THE PARTICIPANTS Most participants were female (75.4%), aged 40-45 years (43.7%), Greek nationals (88%), married (78.2%), employed (62,7%) and high school graduates (52.8%). In addition, we found a significant association between the 3 primary health care centers and participant characteristics: nationality ( X 2 = 19.918, p < .001, Cramer’s V = .375), age ( X 2 = 14.289, p = .006, Cramer’s V = .224) and gender ( X 2 = 8.086, p = .017, Cramer’s V = .239) . Concerning access to the program, 68.3% of participants scheduled an appointment in the first two days, while 58.9% declared a waiting time in the waiting room of up to 5 minutes. Moreover, all participants totally agreed (100%) that they will continue undergoing screening and that they will recommend the program to family and friends. The majority of participants (66.4%) learned about the program through friends and relatives, while a smaller number of participants (9.4%) were informed by their doctor . With regards to the counselling, the participants declared that they preferred the doctor to perform health promotion counselling (43.7 %), where it was available (see discussion) or health visitor/nurse (48.6%) . Moreover, we found a significant association between the 3 primary health care centers and the amount of waiting time ( X 2 = 16.687, p < .001, Cramer’s V = 0.342) and the health promotion counselling ( X 2 = 45.806, p < .001, Cramer’s V = .401). Concerning the attitude of the health care personnel, analysis of variance showed greater satisfaction with the health visitor/nurse of Kallithea and Peristeri. Participants stated that they received more attention and interest and that the health visitors/nurses in these centres were more capable, available and helpful compared to the health professionals at Agia Sofia . The doctor of Agia Sophia received the highest score (4.90) in the doctor’s evaluation and the doctor of Peristeri the lowest (4.32). There was no statistically significant difference in the general evaluation of the pilot PPMA between the primary health care centers. Allhealth care centers received very high evaluation showing impressive satisfaction rates from the operation of the pilot PPMA. Multiple linear regression was used to identify independent determinants of the evaluation of the pilot PPMA . The results of linear regression revealed attention and interest of health visitors/nurse, evaluation of doctor, gender and age not to be statistically significant predictors to the model (p > .05). However, the results of multiple linear regression analysis revealed a statistically significant association between communication of results to participants and evaluation of health visitor/nurse (R2 = .355, adjusted R2 = .326, F (6, 135) = 12.381, p < .001). Most participants were female (75.4%), aged 40-45 years (43.7%), Greek nationals (88%), married (78.2%), employed (62,7%) and high school graduates (52.8%). In addition, we found a significant association between the 3 primary health care centers and participant characteristics: nationality ( X 2 = 19.918, p < .001, Cramer’s V = .375), age ( X 2 = 14.289, p = .006, Cramer’s V = .224) and gender ( X 2 = 8.086, p = .017, Cramer’s V = .239) . Concerning access to the program, 68.3% of participants scheduled an appointment in the first two days, while 58.9% declared a waiting time in the waiting room of up to 5 minutes. Moreover, all participants totally agreed (100%) that they will continue undergoing screening and that they will recommend the program to family and friends. The majority of participants (66.4%) learned about the program through friends and relatives, while a smaller number of participants (9.4%) were informed by their doctor . With regards to the counselling, the participants declared that they preferred the doctor to perform health promotion counselling (43.7 %), where it was available (see discussion) or health visitor/nurse (48.6%) . Moreover, we found a significant association between the 3 primary health care centers and the amount of waiting time ( X 2 = 16.687, p < .001, Cramer’s V = 0.342) and the health promotion counselling ( X 2 = 45.806, p < .001, Cramer’s V = .401). Concerning the attitude of the health care personnel, analysis of variance showed greater satisfaction with the health visitor/nurse of Kallithea and Peristeri. Participants stated that they received more attention and interest and that the health visitors/nurses in these centres were more capable, available and helpful compared to the health professionals at Agia Sofia . The doctor of Agia Sophia received the highest score (4.90) in the doctor’s evaluation and the doctor of Peristeri the lowest (4.32). There was no statistically significant difference in the general evaluation of the pilot PPMA between the primary health care centers. Allhealth care centers received very high evaluation showing impressive satisfaction rates from the operation of the pilot PPMA. Multiple linear regression was used to identify independent determinants of the evaluation of the pilot PPMA . The results of linear regression revealed attention and interest of health visitors/nurse, evaluation of doctor, gender and age not to be statistically significant predictors to the model (p > .05). However, the results of multiple linear regression analysis revealed a statistically significant association between communication of results to participants and evaluation of health visitor/nurse (R2 = .355, adjusted R2 = .326, F (6, 135) = 12.381, p < .001). The aim of the current study was to evaluate the operation of the pilot PPMA organized by EOPYY in 3 primary health care centers of Athens and Piraeus. To the best of our knowledge, the present study is the first to evaluate the operation of the program. The results of our study demonstrate great evaluation of participants regarding access to the program. Most participants were informed about the program from family and friends with only a small number referred by their doctor. Research has shown that Greek doctors have limited awareness of screening so it is crucial to educate physicians in using effective strategies for the implementation of prevention . Also, protocols and guidelines should be established to improve doctors screening attitudes . All respondents agreed that they will continue undergoing screening and that they will recommend the program to family and friends. Research findings show substantial benefits from undergoing annual examinations especially with patients receiving lipid screening and gynaecological screening . Communication of health results and health promotion could be performed by nurses or health visitors due to the confidence, proximity and comfort participants feel for them . Our study revealed a higher satisfaction rate for the health visitors/nurses than with doctors. Participants stated that they received more attention and interest from the paramedical personnel (midwives, health visitors and nurses). The negative rating of the doctor’s evaluation could possibly be attributed to the lack of time due to part-time occupation and/or lack of interest. Despite the relatively low evaluation of the program’s doctor in our study, participants gave great evaluation for the operation of the pilot PPMA as a whole in all three primary health care centers. This is attributed to the courtesy of personnel, the provision of free screening tests and ease of appointments. The latter confers with findings from other studies showing that participants will evaluate health care services with high scores provided they are satisfied with the organizational structure, the waiting room, the waiting time and the behaviour of the health care personnel . Importantly, using a multiple linear regression analysis in order to identify variables important for the evaluation of the program, demonstrated that the communication of results to participants and evaluation of health visitor/nurse were associated with higher general evaluation of the PPMA from the participants. The present study has certain limitations. The survey was conducted in only 3 of the 6 primary health care centers where the program runs including non-randomly selected participants. The selection of the health care centers was made taking into account the attendance rate in each health care center and the long-distance travel from Athens due to funding constraints. The participants were interviewed on a scheduled day when the rate of appointment was higher. Due to the lack of random selection it is difficult to generalize findings. Bias also may arise from the gratitude and satisfaction participants feel towards the health care personnel resulting in skewed values. We must also consider the probability of bias arising from the participants that were eliminated from the program, due to the inclusion/exclusion criteria, that may change the general evaluation rate. Despite these limitations this study is the first to evaluate the operation of the pilot PPMA and it is worth mentioning the really high rate of positive response. In Greece implemented prevention focuses mainly on cancer screening programs. The operation of this program is innovative due to the coverage of a wider range of diseases, so the benefit for citizens is greater. Additionally, in the current years of economic crisis, free of charge screening is considered a blessing for low- and middle-class population. These findings are valid and should be taken into account for the implementation of the program in larger general population groups. The findings of the current study demonstrate high evaluation both for the operation of the pilot PPMA and the health care personnel in the all health care centers and the desire for continuation of the screening program on a general basis. In Greece the attendance rate in screening programs is extremely low due to the lack of centralized invitation system, guidelines and protocols and the reduced interest of primary care doctors for prevention. The results of our study can be used to inform preventive medicine program managers about the benefits of the program for the general population so it could lead to nation-wide implementation providing free or low-cost screening and should emphasize the contribution of health visitors and nurses in the successful operation and its acceptance. |
Unlocking breast cancer in Brazilian public health system: Using tissue microarray for accurate immunohistochemical evaluation with limitations in subtyping | c4a748a7-ce8f-4908-8a2f-0bb1be246689 | 11694303 | Anatomy[mh] | Breast cancer (BC) is the most common cancer in women worldwide, with 70% of deaths from the disease occurring in low- and middle-income countries, such as Brazil. The National Cancer Institute estimates that approximately 18,000 Brazilian women die of BC every year. Around 75% of the population has no private health insurance and relies exclusively on the Universal Health System (SUS), the largest public health system in the world that provides free healthcare to all Brazilians, regardless of their socioeconomic status. BC is more frequently diagnosed in its symptomatic and in more advanced stages in SUS than in private health systems or high-income countries. Brazilian public hospitals face enormous pressure to optimize healthcare services and reduce costs. The Hospital de Clínicas de Porto Alegre, a tertiary public hospital in the South of Brazil, processes approximately 540 immunohistochemical (IHC) tests of BC biomarkers per year at a cost of around 31,000 USD (154,400.00 BRL—Brazilian reais). This is the most common and expensive individual test offered in our laboratory and includes analysis of the expression of estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor 2 (HER2), and the proliferation marker Ki-67. These biomarkers are combined for BC subtyping into luminal (A and B), HER2-positive, and triple-negative tumors and guide systemic therapy. Proposing strategies to increase access to BC diagnosis and treatment is a priority in the Brazilian public health context. The IHC tests of BC biomarkers are traditionally done on surgical specimens or biopsies on whole individual glass slides. The tissue microarray (TMA) approach, which combines multiple cylindrical fragments of tumor tissue from different patients in the same glass slide, has been extensively used in pathology research. TMA saves working time, standardizes reactions, allows for comparative interpretation of cases, and reduces the total cost of tissue analyses. However, the use of TMA in clinical practice remains controversial worldwide, and its feasibility and cost-benefit have never been evaluated in the Brazilian public health system before. BC was chosen as the prototype for this type of study due to its high regional prevalence at the regional level and throughout the country. This study aimed to assess the diagnostic accuracy of TMA as a cost-effective alternative to evaluating the IHC status of ER, PR, HER2, Ki-67, and BC subtyping and maximize its potential use in clinical practice. Patients The study is a retrospective cohort analysis that evaluates the diagnostic accuracy of TMA in BC IHC evaluation. Two hundred forty-two women diagnosed with invasive BC in Hospital de Clínicas de Porto Alegre between 2010 and 2015 were consecutively included in the study. The patient eligible criteria are BC diagnosis and previous IHC evaluation for ER, PR, HER2, and Ki-67 available in medical records. Formalin-fixed tissue blocks from all patients were retrieved from the Laboratory of Pathology archive in accordance with ethical guidelines. We consistently follow the established preanalytical handling guidelines of the College of American Pathologists. The clinical data and the original IHC scores of ER, PR, HER2, and Ki-67 were obtained from the anatomopathological reports through analyses of the whole slide and medical records. The average age was 58.2 years (range 24–92 years), and invasive carcinoma of the non-special type was the most frequent histopathological type of tumor. Pathological staging was determined using the AJCC TNM System, and was distributed as follows: 131 patients in stage I, 57 in stage II, 42 in stage III, and 12 in stage IV. In 237 of the 242 cases, the IHC scores were fully available in the pathology report, making it possible to define the IHC subtype (BC subtype): 101 tumors were classified as luminal A (ER + and/or PR + , HER2 − and Ki-67 ⩽20%), 87 as luminal B (ER + and/or PR + , HER2 + or Ki-67 >20%), 19 as HER2 positive (ER − , PR − and HER2 + ) and 30 as triple negative (ER − , PR − , and HER2 − ). Cases with tumor areas smaller than 2 cm, treated with neoadjuvant chemotherapy prior to surgical resection or without IHC evaluation for the four markers evaluated were excluded. Only excisional samples were utilized, as core biopsies were not employed to ensure the preservation of the patient’s archived tissue and to mitigate the risk of material depletion during TMA assembly. TMA assembly and immunohistochemistry The most representative area of the tumor was carefully circled by an expert breast pathologist (MSG) on the hematoxylin-eosin-stained slide in areas with high tumor cellularity. For TMAs assembly, we used the manual TMA T-Sue system (Simport ® Scientific, Beloeil, Canada) to extract two cores of 2.0 mm of each tumor using the principles first described by Kononen et al. . Briefly, the procedures began with the preparation of the TMA grid to correctly identify the position of each sample and the organization of the donor blocks. Then, two cylindrical tissue cores were extracted from the donor block with a 2.00 mm punch needle, no more than 3 mm deep, and precisely placed in the recipient block, which was previously prepared using the M473-60 mold. This mold has a capacity for 60 cores, distributed over 6 rows and 10 columns, allowing to include 24 duplicate tumors/cases per TMA. For guidance in reading the TMA, a core containing placental tissue was included in each TMA block. The cores were fixed with light pressure followed by brief heating, cooled overnight, and sectioned (4 µm). The sections were then mounted on slides for H&E staining and analysis. The total tumor area size analyzed is 3.14 mm 2 for each 2.0 mm core. The minimum number of tumor cells sufficient for scoring was ⩾100 per core. For immunohistochemistry, the TMA blocks were cut into 3 µm sections and placed on glass slides with positive and negative controls. The sections were processed on Ventana automation equipment (BenchMark AutoStainer; Ventana Medical Systems, Tucson, AZ, USA) using the following antibodies: ER (clone SP1; Ventana, Tucson, AZ, USA), PR (clone 1E2; Ventana Medical Systems), HER2 (clone 4B5; Ventana Medical Systems) and Ki-67 (clone 30-9; Ventana Medical Systems) . This immunostaining method is the same one used in the laboratory’s routine work, with quality attested to by the Joint Commission International Accreditation Seal in 2017. Microscopic analysis of TMA TMA consolidated multiple tissue samples into a single slide for simultaneous analysis. In contrast, in the traditional IHC, a whole section of the tumor tissue is analyzed in slides individually. The reliability of TMA microscopic analysis depended on the quality of the TMA, the alignment of the cores, and the pathologist’s ability to orient themselves and identify precise samples according to the grid. Then, following an initial overall TMA quality evaluation and positioning, the pathologist proceeded to assign a core-specific IHC score to each tissue core by traversing the slide in an up-and-down motion. The same criteria of immunostaining evaluation in the whole section were applied to TMA. The evaluation of ER, PR, and HER2 expression was carried out in accordance with the guidelines of the American Society of Clinical Oncology. Nuclear staining was considered positive for ER and/or PR when detected in at least 1% of tumor cells at any intensity ( and . For HER2, staining in the membranes of tumor cells was classified as follows : 0, when no tumor cells showed HER2-positive staining or incomplete and weakly perceptible membrane staining in ⩽10% of tumor cells; 1+, incomplete and weakly perceptible staining in ⩾10% of tumor cells; 2+, weak to moderate complete staining observed in ⩾10% of tumor cells; and 3+, circumferential and strong complete staining in ⩾10% of tumor cells. Cases with a score of 3+ were considered HER2-positive. Cases 2+ are considered indeterminate. All other cases (0 or 1+) were considered HER2-negative . For Ki-67, the IHC score was determined using the St. Gallen International Expert Consensus . Tumor cells were evaluated for Ki-67 and scored with the percentage of positively stained nuclei. A cut-off point >20% was considered high (“positive”) for Ki-67, while values ⩽20% were considered low (“negative”) . The TMA slides were read by a breast specialist pathologist (MSG) who read the first core of each case. If it was impossible to read the first core due to selection errors or loss of material during the procedure, the second core was analyzed. Informative cores were those that allowed the pathologist to interpret and determine the IHC score successfully in TMA. When cores were missing or in the absence of a tumor, they were considered non-informative. A second breast specialist pathologist (DMU) evaluated the TMA slides independently to assess agreement between observers. To assess intratumoral heterogeneity, two cores from the same case introduced into the TMA were evaluated in a randomly selected subset of cases ( n = 12). The combined analysis of the four biomarker readings on the TMA described before was used to determine the BC subtype in each case. The IHC scores for ER, PR, HER2, Ki-67, and the BC subtype resulting from the TMA reading were compared to those obtained in the original pathology report for the respective case by consulting the medical records. In cases of disagreement, the original slide of the case was re-analyzed by the leading pathologist (MSG) to determine the final IHC score. Statistical analysis The sample size was calculated using data from Hospital de Clínicas in Porto Alegre, considering a proportion ( P ) of positivity of 60% for PR/RE, 20% for HER-2-enriched, and 20% for triple-negative BC. The estimation precision ( D ) used considered the spectrum of the 10% confidence interval, with semi-amplitude (0.05 above or 0.05 below) as the maximum acceptable error. The confidence interval used was 95% ( Z = 1.96, for α = 0.05). By applying the formula N = Z * Z ( P (1 − P ))/( D * D ), the N of 96 samples were obtained. All statistical analysis was carried out using SPSS version 18 (SPSS IBM, New York, NY, USA). The agreement between the IHC score in the TMA versus the medical records and between the different observers was determined by calculating Cohen’s kappa. Sensitivity, specificity, disease prevalence, positive and negative predictive value, and accuracy are expressed as percentages and in Clopper-Pearson confidence intervals. p -Values of and less than 0.05 were considered statistically significant. We consistently followed the STARD2015 as the appropriate reporting guidelines when preparing our manuscript and submitted the completed checklist as Supplemental Material . The study is a retrospective cohort analysis that evaluates the diagnostic accuracy of TMA in BC IHC evaluation. Two hundred forty-two women diagnosed with invasive BC in Hospital de Clínicas de Porto Alegre between 2010 and 2015 were consecutively included in the study. The patient eligible criteria are BC diagnosis and previous IHC evaluation for ER, PR, HER2, and Ki-67 available in medical records. Formalin-fixed tissue blocks from all patients were retrieved from the Laboratory of Pathology archive in accordance with ethical guidelines. We consistently follow the established preanalytical handling guidelines of the College of American Pathologists. The clinical data and the original IHC scores of ER, PR, HER2, and Ki-67 were obtained from the anatomopathological reports through analyses of the whole slide and medical records. The average age was 58.2 years (range 24–92 years), and invasive carcinoma of the non-special type was the most frequent histopathological type of tumor. Pathological staging was determined using the AJCC TNM System, and was distributed as follows: 131 patients in stage I, 57 in stage II, 42 in stage III, and 12 in stage IV. In 237 of the 242 cases, the IHC scores were fully available in the pathology report, making it possible to define the IHC subtype (BC subtype): 101 tumors were classified as luminal A (ER + and/or PR + , HER2 − and Ki-67 ⩽20%), 87 as luminal B (ER + and/or PR + , HER2 + or Ki-67 >20%), 19 as HER2 positive (ER − , PR − and HER2 + ) and 30 as triple negative (ER − , PR − , and HER2 − ). Cases with tumor areas smaller than 2 cm, treated with neoadjuvant chemotherapy prior to surgical resection or without IHC evaluation for the four markers evaluated were excluded. Only excisional samples were utilized, as core biopsies were not employed to ensure the preservation of the patient’s archived tissue and to mitigate the risk of material depletion during TMA assembly. The most representative area of the tumor was carefully circled by an expert breast pathologist (MSG) on the hematoxylin-eosin-stained slide in areas with high tumor cellularity. For TMAs assembly, we used the manual TMA T-Sue system (Simport ® Scientific, Beloeil, Canada) to extract two cores of 2.0 mm of each tumor using the principles first described by Kononen et al. . Briefly, the procedures began with the preparation of the TMA grid to correctly identify the position of each sample and the organization of the donor blocks. Then, two cylindrical tissue cores were extracted from the donor block with a 2.00 mm punch needle, no more than 3 mm deep, and precisely placed in the recipient block, which was previously prepared using the M473-60 mold. This mold has a capacity for 60 cores, distributed over 6 rows and 10 columns, allowing to include 24 duplicate tumors/cases per TMA. For guidance in reading the TMA, a core containing placental tissue was included in each TMA block. The cores were fixed with light pressure followed by brief heating, cooled overnight, and sectioned (4 µm). The sections were then mounted on slides for H&E staining and analysis. The total tumor area size analyzed is 3.14 mm 2 for each 2.0 mm core. The minimum number of tumor cells sufficient for scoring was ⩾100 per core. For immunohistochemistry, the TMA blocks were cut into 3 µm sections and placed on glass slides with positive and negative controls. The sections were processed on Ventana automation equipment (BenchMark AutoStainer; Ventana Medical Systems, Tucson, AZ, USA) using the following antibodies: ER (clone SP1; Ventana, Tucson, AZ, USA), PR (clone 1E2; Ventana Medical Systems), HER2 (clone 4B5; Ventana Medical Systems) and Ki-67 (clone 30-9; Ventana Medical Systems) . This immunostaining method is the same one used in the laboratory’s routine work, with quality attested to by the Joint Commission International Accreditation Seal in 2017. TMA consolidated multiple tissue samples into a single slide for simultaneous analysis. In contrast, in the traditional IHC, a whole section of the tumor tissue is analyzed in slides individually. The reliability of TMA microscopic analysis depended on the quality of the TMA, the alignment of the cores, and the pathologist’s ability to orient themselves and identify precise samples according to the grid. Then, following an initial overall TMA quality evaluation and positioning, the pathologist proceeded to assign a core-specific IHC score to each tissue core by traversing the slide in an up-and-down motion. The same criteria of immunostaining evaluation in the whole section were applied to TMA. The evaluation of ER, PR, and HER2 expression was carried out in accordance with the guidelines of the American Society of Clinical Oncology. Nuclear staining was considered positive for ER and/or PR when detected in at least 1% of tumor cells at any intensity ( and . For HER2, staining in the membranes of tumor cells was classified as follows : 0, when no tumor cells showed HER2-positive staining or incomplete and weakly perceptible membrane staining in ⩽10% of tumor cells; 1+, incomplete and weakly perceptible staining in ⩾10% of tumor cells; 2+, weak to moderate complete staining observed in ⩾10% of tumor cells; and 3+, circumferential and strong complete staining in ⩾10% of tumor cells. Cases with a score of 3+ were considered HER2-positive. Cases 2+ are considered indeterminate. All other cases (0 or 1+) were considered HER2-negative . For Ki-67, the IHC score was determined using the St. Gallen International Expert Consensus . Tumor cells were evaluated for Ki-67 and scored with the percentage of positively stained nuclei. A cut-off point >20% was considered high (“positive”) for Ki-67, while values ⩽20% were considered low (“negative”) . The TMA slides were read by a breast specialist pathologist (MSG) who read the first core of each case. If it was impossible to read the first core due to selection errors or loss of material during the procedure, the second core was analyzed. Informative cores were those that allowed the pathologist to interpret and determine the IHC score successfully in TMA. When cores were missing or in the absence of a tumor, they were considered non-informative. A second breast specialist pathologist (DMU) evaluated the TMA slides independently to assess agreement between observers. To assess intratumoral heterogeneity, two cores from the same case introduced into the TMA were evaluated in a randomly selected subset of cases ( n = 12). The combined analysis of the four biomarker readings on the TMA described before was used to determine the BC subtype in each case. The IHC scores for ER, PR, HER2, Ki-67, and the BC subtype resulting from the TMA reading were compared to those obtained in the original pathology report for the respective case by consulting the medical records. In cases of disagreement, the original slide of the case was re-analyzed by the leading pathologist (MSG) to determine the final IHC score. The sample size was calculated using data from Hospital de Clínicas in Porto Alegre, considering a proportion ( P ) of positivity of 60% for PR/RE, 20% for HER-2-enriched, and 20% for triple-negative BC. The estimation precision ( D ) used considered the spectrum of the 10% confidence interval, with semi-amplitude (0.05 above or 0.05 below) as the maximum acceptable error. The confidence interval used was 95% ( Z = 1.96, for α = 0.05). By applying the formula N = Z * Z ( P (1 − P ))/( D * D ), the N of 96 samples were obtained. All statistical analysis was carried out using SPSS version 18 (SPSS IBM, New York, NY, USA). The agreement between the IHC score in the TMA versus the medical records and between the different observers was determined by calculating Cohen’s kappa. Sensitivity, specificity, disease prevalence, positive and negative predictive value, and accuracy are expressed as percentages and in Clopper-Pearson confidence intervals. p -Values of and less than 0.05 were considered statistically significant. We consistently followed the STARD2015 as the appropriate reporting guidelines when preparing our manuscript and submitted the completed checklist as Supplemental Material . TMA performance In order to incorporate the 242 duplicate BC cases, we constructed 10 TMA blocks, each containing 2 cores of 2.0 mm per case. Each BC case contributes 4 cores (one for each antibody: ER, PR, HER2, Ki-67), resulting in a total of 968 cores. These 968 cores represent the total number of potential cores to be scored in the 10 TMA slides (242 cases × 4 markers). Regarding the overall quality of the TMA, the immunostaining on the TMA slides showed consistent results with no discrepancies between central and peripheral nuclei. The proper alignment of the nuclei, the inclusion of positive and negative controls, and the orientation of nuclei (such as the placenta) ensured an effective and safe reading by the pathologists. In TMA slides IHC evaluation, out of the total 968 cores, 97% (940) provided informative results, showing high immunostaining quality and sufficient tumor cellularity (>100 tumor nuclei per core) for adequate scoring . In 91% of cases, the reading of the first core of the duplicate was sufficient to determine the IHC score. However, in 79 cases, the second core had to be assessed to complete the analysis, highlighting the importance of including duplicate tumors in the TMAs. Uninformative cores were minimal at 2.9% , primarily due to errors in tumor area selection where both cores lacked tumor tissue. Loss of both cores during processing occurred in only 1% of cases. Importantly, there were no differences observed in the quality of TMA slides stained with different antibodies. Inter-examiner variability and intratumoral heterogeneity For all the antibodies evaluated, there was almost perfect and statistically significant agreement in determining the IHC score by reading the TMAs by two different pathologists. Kappa values ranged from 0.85 for HER2 to 0.91 for ER , classified as “almost perfect” by Cohen’s criteria. With regard to intratumoral heterogeneity, the agreement between the IHC scores assigned to the two cores from the same case included in the TMA varied by antibody. For ER and HER2, agreement was almost perfect (100%), with kappa values of 1.0 for both markers. For PR and Ki-67, there was less agreement, classified as moderate for PR ( k = 0.47) and substantial for Ki-67 ( k = 0.68). Among the discordant cases, two PR-positive cases in the first core were assessed as negative in the second, and two Ki-67-high cases in the first core were classified as low in the second. Comparison of TMA results versus original report Overall, there was a high agreement between the IHC scores obtained in TMA cores and those in the original report, based on the evaluation of the whole section. In the first analysis, 828 of the 940 (88%) IHC scores were concordant, and 112 were discordant. The discordant cases had the original slide containing the whole section reviewed by the study’s leading pathologist (MSG), who then reissued the final IHC score. Forty-three IHC scores with initially discordant results were considered concordant after a whole section review using the same immunostaining interpretation criteria. Thus, final agreement was observed between the TMA versus the original report in 871 of the 940 IHC scores (93%) evaluated, being classified as almost perfect and statistically significant ( k = 0.81, p < 0.001) . Some differences could be observed when the concordance rates were compared among the antibodies . There was an almost perfect agreement for ER and PR, while for HER2 and Ki-67, this was slightly lower and classified as substantial. The final comparative analysis of the 69 discordant IHC scores showed that in the evaluation of ER and PR, there was a lower and similar frequency of false-positive and false-negative cases in the TMA. For HER2 and Ki-67, there were more discordant cases, with a higher frequency of false negatives (4.5 and 6.7%, respectively) than false positives in the TMA. The most significant discrepancy in results was observed for Ki-67, where 24 of the 235 IHC scores were discordant in the TMA compared to the original report. Working time and cost analysis shows a comparative analysis of the time and cost spent on the technical procedures and evaluation of results using the TMA versus the traditional procedure. The TMA approach reduced the time of IHC evaluation (for the four markers) from 8.5 to 0.5 h per case. This estimated time included glass slide preparation, TMA assembly, IHC staining, and the pathologist’s IHC scoring process of an individual or TMA glass slides. Considering the current values, the cost of the IHC panel with four biomarkers (including labor and materials) is $53.61 per case compared to $4.58 spent per case in the TMA approach, a reduction of approximately 11 times. In a TMA of 24 cases, the apparent saving is $1146.52 in total or $47.77 per case. BC subtyping Defining the BC subtype is of great clinical relevance in therapeutic management and disease outcomes. Overall, BC subtyping was possible in 97% (237/242) of the cases using the traditional method compared with 89% (217/242) using TMA. In 20 cases (8.4%), the IHC subtype could not be determined due to failure to read the IHC score on the TMA for one or more of the biomarkers analyzed. Between the 217 remaining cases, there was agreement in the BC subtype in 162 (75%) by the two methods. presents a detailed analysis of the sensibility, specificity, and overall accuracy of TMA in BC subtyping. Among the 55 discordant cases, 41 (74%) were luminal tumors classified incorrectly as luminal A or B, 5 were HER2 tumor classified incorrectly as luminal A, and 4 triple-negative tumors were incorrectly classified as luminal A, luminal B, or HER-2 subtypes using TMA. In order to incorporate the 242 duplicate BC cases, we constructed 10 TMA blocks, each containing 2 cores of 2.0 mm per case. Each BC case contributes 4 cores (one for each antibody: ER, PR, HER2, Ki-67), resulting in a total of 968 cores. These 968 cores represent the total number of potential cores to be scored in the 10 TMA slides (242 cases × 4 markers). Regarding the overall quality of the TMA, the immunostaining on the TMA slides showed consistent results with no discrepancies between central and peripheral nuclei. The proper alignment of the nuclei, the inclusion of positive and negative controls, and the orientation of nuclei (such as the placenta) ensured an effective and safe reading by the pathologists. In TMA slides IHC evaluation, out of the total 968 cores, 97% (940) provided informative results, showing high immunostaining quality and sufficient tumor cellularity (>100 tumor nuclei per core) for adequate scoring . In 91% of cases, the reading of the first core of the duplicate was sufficient to determine the IHC score. However, in 79 cases, the second core had to be assessed to complete the analysis, highlighting the importance of including duplicate tumors in the TMAs. Uninformative cores were minimal at 2.9% , primarily due to errors in tumor area selection where both cores lacked tumor tissue. Loss of both cores during processing occurred in only 1% of cases. Importantly, there were no differences observed in the quality of TMA slides stained with different antibodies. For all the antibodies evaluated, there was almost perfect and statistically significant agreement in determining the IHC score by reading the TMAs by two different pathologists. Kappa values ranged from 0.85 for HER2 to 0.91 for ER , classified as “almost perfect” by Cohen’s criteria. With regard to intratumoral heterogeneity, the agreement between the IHC scores assigned to the two cores from the same case included in the TMA varied by antibody. For ER and HER2, agreement was almost perfect (100%), with kappa values of 1.0 for both markers. For PR and Ki-67, there was less agreement, classified as moderate for PR ( k = 0.47) and substantial for Ki-67 ( k = 0.68). Among the discordant cases, two PR-positive cases in the first core were assessed as negative in the second, and two Ki-67-high cases in the first core were classified as low in the second. Overall, there was a high agreement between the IHC scores obtained in TMA cores and those in the original report, based on the evaluation of the whole section. In the first analysis, 828 of the 940 (88%) IHC scores were concordant, and 112 were discordant. The discordant cases had the original slide containing the whole section reviewed by the study’s leading pathologist (MSG), who then reissued the final IHC score. Forty-three IHC scores with initially discordant results were considered concordant after a whole section review using the same immunostaining interpretation criteria. Thus, final agreement was observed between the TMA versus the original report in 871 of the 940 IHC scores (93%) evaluated, being classified as almost perfect and statistically significant ( k = 0.81, p < 0.001) . Some differences could be observed when the concordance rates were compared among the antibodies . There was an almost perfect agreement for ER and PR, while for HER2 and Ki-67, this was slightly lower and classified as substantial. The final comparative analysis of the 69 discordant IHC scores showed that in the evaluation of ER and PR, there was a lower and similar frequency of false-positive and false-negative cases in the TMA. For HER2 and Ki-67, there were more discordant cases, with a higher frequency of false negatives (4.5 and 6.7%, respectively) than false positives in the TMA. The most significant discrepancy in results was observed for Ki-67, where 24 of the 235 IHC scores were discordant in the TMA compared to the original report. shows a comparative analysis of the time and cost spent on the technical procedures and evaluation of results using the TMA versus the traditional procedure. The TMA approach reduced the time of IHC evaluation (for the four markers) from 8.5 to 0.5 h per case. This estimated time included glass slide preparation, TMA assembly, IHC staining, and the pathologist’s IHC scoring process of an individual or TMA glass slides. Considering the current values, the cost of the IHC panel with four biomarkers (including labor and materials) is $53.61 per case compared to $4.58 spent per case in the TMA approach, a reduction of approximately 11 times. In a TMA of 24 cases, the apparent saving is $1146.52 in total or $47.77 per case. Defining the BC subtype is of great clinical relevance in therapeutic management and disease outcomes. Overall, BC subtyping was possible in 97% (237/242) of the cases using the traditional method compared with 89% (217/242) using TMA. In 20 cases (8.4%), the IHC subtype could not be determined due to failure to read the IHC score on the TMA for one or more of the biomarkers analyzed. Between the 217 remaining cases, there was agreement in the BC subtype in 162 (75%) by the two methods. presents a detailed analysis of the sensibility, specificity, and overall accuracy of TMA in BC subtyping. Among the 55 discordant cases, 41 (74%) were luminal tumors classified incorrectly as luminal A or B, 5 were HER2 tumor classified incorrectly as luminal A, and 4 triple-negative tumors were incorrectly classified as luminal A, luminal B, or HER-2 subtypes using TMA. In the present study, we propose using a TMA constructed with two 2.0 mm diameter cores of tumor tissue as an alternative to the traditional procedure for the IHC evaluation of ER, PR, HER2, and Ki-67 in BC. The study results suggest that TMA is a fast, highly accurate, and cost-effective method for testing individual BC biomarkers. However, based on the combined analysis of the four antibodies, we do not recommend using TMA with two cores for BC subtyping unless a reduction in costs is necessary to continue testing patients and providing them with treatment. Regarding TMA feasibility and overall performance, we observed that after an initial and time-consuming period of training for the technical staff, high-quality TMAs were constructed in our laboratory in a satisfactory way. There was a high retention rate of informative cores in the TMAs, with only 2.9% lost , similar to that reported in a previous study. The absence of a tumor is the main cause of non-informative cores, probably due to an error in selecting the area to be punctured in the original block, with the capture of more peripheral cores where the tumor may not be represented. This finding reflects the need for attention and adequate training for the tumor selection stage. The loss of the two cores during the IHC process occurred in only 1% of cases, a frequency similar to that observed in previous studies , and lower than the 10% loss reported by Visser et al., who used 0.6 mm cores. Thus, our practice of using 2.0 mm cores in duplicate seems to be ideal to avoid the need to recolor the entire slide due to the infeasibility of analyzing non-informative IHC scores in the TMA. In microscopic TMA analyses, when the 4 antibodies were analyzed individually (core-by-core), the comparison of 940 IHC scores in the 242 cases showed a high overall agreement (93%, almost perfect) between the TMA results and the original report. In general, an accuracy rate of 90% or above is often deemed high and acceptable for clinical implementation. Many widely accepted diagnostic tests, such as mammography for BC screening, often have sensitivity and specificity rates in the range of 80%–90%. In summary, a 93% accuracy rate is considered high and reliable for clinical practice, meeting or exceeding the standards of well-established diagnostic methods. However, as described by other authors, , our study confirms that the TMA performance is not the same for all antibodies. The concordance of IHC scores was higher for ER and PR and lower for HER2 and Ki-67. Our hypothesis to explain these differences is mainly based on intratumoral heterogeneity, which has also been reported as a limitation of the use of TMA in routine IHC evaluation. – To investigate this hypothesis, we performed an intratumoral heterogeneity analysis and detected that there is a high agreement between the IHC scores obtained by comparing the two cores of the same case for the ER and HER2 markers and slightly lower for PR and Ki-67. So, we suggest that intratumoral heterogeneity partially explains the occurrence of false-negative and false-positive IHC scores observed in our study (4%, 5%, 10%, and 10% for ER, PR, HER2, and Ki-67, respectively). These results align with previous studies, where discordant results in 2%, 7%, and 8% of cases for ER, PR, and HER2, respectively and 18% for Ki-67 were also associated with intratumoral heterogeneity. It is well known that increasing the number of cores for each case in the TMA to cover a larger area of the tumor may reduce its impact. Taken together, our results indicate that, especially for HER2 and Ki-67, the addition of more than two cores (2.0 mm each) per case in the TMA or the use of whole-slide staining to IHC analyses should be considered to decrease the chance of discrepant results. It was expected that the TMA approach would drastically reduce the work time and the cost of evaluating BC markers spent in Brazilian women’s healthcare. Indeed, this is the first study to detail potential savings related to implementing TMA technology in Brazil, specifically inside the public health system (SUS). Importantly, we showed a reduction of 17-fold in time and 11-fold in cost of an individual BC IHC scoring, considering labor time and direct and indirect costs. Taken our current demand of 540 requests per year, TMA would allow us to save 2,500,000 USD per year in BC diagnosis, a reduction of 91% in the amount originally spent on this test in our hospital. It is important to remember that cost-effectiveness is directly linked to the volume of tests performed in each laboratory. So, the time spent to gather sufficient cases to fulfill the TMA should be considered to avoid delays in the release of results. An applicable alternative would be to use TMAs with a smaller number of cases (12, 24, or 36 cores per TMA), which could be produced weekly. “Urgent” cases would be processed immediately using the traditional method and reported in less than 48 h. Even so, TMA may not be a viable method for laboratories with a low volume of tests. However, to properly decide on implementing new technology as an alternative method in clinical practice, it was crucial to know the TMA accuracy in predicting the BC subtype through the combined analyses of the four BC markers. Based on our results, two arguments can demonstrate that IHC analyses of TMA and whole section are not equivalent and their potential for predicting BC subtype: (1) Accuracy and reliability concerns: the results indicate that TMA yielded a lower rate of successful BC subtyping compared to the traditional method. While the traditional method achieved BC subtyping in 97% of cases, TMA only achieved it in 89% of cases. This discrepancy in success rates suggests that TMA may be less accurate or reliable in determining BC subtypes. Our findings highlight the variability in TMAs accuracy across BC subtypes, with notable differences in sensitivity, specificity, and overall accuracy. Among Luminal A cases, TMA demonstrated the highest agreement rate but showed relatively lower accuracy in identifying Luminal B cases. In the same way, TMA performance was better in identifying triple-negative cases than HER2-positive. These results highlight the limitations of TMA in accurately capturing the heterogeneity of BC in each individual marker, which can potentially be amplified when they are combined to predict the BC subtype, with serious implications for treatment decisions and patient outcomes. (2) Technical challenges and limitations: the inability to determine the BC subtype in 8.4% of cases due to failure to read the IHC score on the TMA indicates technical challenges associated with this method. Despite the high number of informative cores in our TMAs, issues such as inadequate tissue sampling or technical errors during slide preparation, staining, or interpretation significantly impacted the BC subtyping in those cases The entire IHC process may need to be redone, potentially resulting in significant financial and time losses. When we highlighted luminal tumors, the predominant subtype in the Brazilian population, we observed that they were frequently misclassified using TMA, with 74% of the discordant cases incorrectly classified as luminal A or B. The correct Ki-67 IHC scoring is crucial for distinguishing between luminal A and B tumors, and the misclassification of Ki-67 as “low” (⩽20%) or “high” (>20%) can be the cause of the higher rate of false-positive or false-negative results for the Ki-67 marker and consequently elevated rate of luminal B incorrectly subtyped in TMA (~30% in our study). Our study aligns with the previous showing a high discordance rate of 38% in Ki-67 scoring in TMAs using the same cut-off of ⩾20%. Our data reinforce the existence of possible reproducibility flaws in the Ki-67 evaluation in TMAs depending on the Ki-67 cut-off point applied in the analyses. Among the HER2 BC group, despite the high specificity, TMA failed to detect 5 in 13 HER2-positive cases, representing the lowest sensibility (58%) compared to the other subtypes. This result differs from the 98% sensitivity detected previously in a similar study that recommends TMA for HER2 subtyping, a rate that we can’t confirm in our study. This could be due to differences in the HER2 scoring methods used in both studies and our patient population composition, including all BC subtypes, which need further investigation. Detecting HER2-positive BC accurately is crucial because it significantly impacts treatment decisions and patient outcomes. They tend to be more aggressive than HER2-negative BC, and they require targeted therapy with drugs like trastuzumab or other HER2-targeted therapies. Finally, our data support previous findings that TMA-based IHC results should be used with caution in BC subtype classification, especially when distinguishing luminal A from luminal B and when interpreting findings for HER2-enriched cancers. Finally, this study has some limitations that should be acknowledged. Technical challenges, such as misclassification of Ki-67 scores and core selection errors, may have contributed to false-negative and false-positive results. Additionally, our findings are based on a single institution’s patient population, which restricts their applicability to broader contexts. Future research should involve larger, multi-center cohorts to enhance the reliability of TMA in BC subtyping in clinical practice. While TMA offers a fast and cost-effective method for testing individual ER, PR, HER2, and Ki-67 biomarkers in BC, caution is needed when using them for BC subtyping. Challenges include tissue loss during construction and varying performance across markers due to tumor heterogeneity. TMAs perform well in identifying certain BC subtypes, like luminal A and triple-negative, but show less reliability in classifying luminal B tumors. Of concern is their lower sensitivity in detecting HER2-positive BC, impacting treatment decisions. Despite the benefits of efficiency and cost, careful consideration of limitations is crucial in clinical practice, requiring further research to optimize TMA use in BC diagnosis and subtyping. sj-docx-1-whe-10.1177_17455057241304654 – Supplemental material for Unlocking breast cancer in Brazilian public health system: Using tissue microarray for accurate immunohistochemical evaluation with limitations in subtyping Supplemental material, sj-docx-1-whe-10.1177_17455057241304654 for Unlocking breast cancer in Brazilian public health system: Using tissue microarray for accurate immunohistochemical evaluation with limitations in subtyping by Rubia Denise Ruppenthal, Emily Ferreira Salles Pilar, Jordan Boeira dos Santos, Rafael Correa Coelho, Carina Machado Costamilan Henriques, Diego de Mendonça Uchôa and Marcia Silveira Graudenz in Women’s Health sj-docx-2-whe-10.1177_17455057241304654 – Supplemental material for Unlocking breast cancer in Brazilian public health system: Using tissue microarray for accurate immunohistochemical evaluation with limitations in subtyping Supplemental material, sj-docx-2-whe-10.1177_17455057241304654 for Unlocking breast cancer in Brazilian public health system: Using tissue microarray for accurate immunohistochemical evaluation with limitations in subtyping by Rubia Denise Ruppenthal, Emily Ferreira Salles Pilar, Jordan Boeira dos Santos, Rafael Correa Coelho, Carina Machado Costamilan Henriques, Diego de Mendonça Uchôa and Marcia Silveira Graudenz in Women’s Health |
Evaluating the strengths and weaknesses of large language models in answering neurophysiology questions | 456651b5-b845-4ab4-9813-32393cc489cd | 11088627 | Physiology[mh] | The world is currently experiencing significant transformations as new tools and technology permeating every corner and aspect of our lives. People are shocked, contemplating the pros, cons and wondering how these advancements will impact us. Can we rely on these innovations? To find answers, researchers are delving into various approaches. They enter artificial intelligence (AI), a captivating and significant phenomenon of our time, with versatile capabilities applicable to a wide range of tasks. Recently, there have been remarkable advancements in natural language processing (NLP). This progress has given rise to sophisticated large language models (LLMs) that can engage with humans in a remarkably human-like manner. Specifically, chatbot platforms have made strides, providing accurate and contextually appropriate responses to users’ queries . With this ongoing progress, there is a growing demand for reliable and efficient question-answering systems in specialized domains like neurophysiology. The rapid advancements in conversational AI have given rise to advanced language models capable of generating humanlike writing. With their wide range of functionalities, including generating human-like responses, proficiency in professional exams, complex problem-solving, and more, these models have captivated interest . Large language models (LLMs) are becoming increasingly popular in both academia and industry owing to their unprecedented performance in various applications. As LLMs continue to play a vital role in both research and everyday activities, their evaluation becomes increasingly critical, not only at the task level but also at the society level to better comprehend their potential risks. In recent years, substantial efforts have been devoted to examine LLMs from diverse perspectives . With the popularization of software like OpenAI’s ChatGPT, Google’s Bard and Anthropic’s Claude, LLMs have permeated various aspects of life and work. They are used to provide customized recipes, suggesting substitutions for missing ingredients. It can be used to draft research proposals, write working code in many programming languages, translate text between languages, assist in policy making, and more. Users interact with LLMs through “prompts” or natural language instructions. Carefully designed prompts can significantly enhance the quality of better outputs . These models, designed to emulate human intelligence, employ statistical analyses to understand patterns and connections among words and phrases . Neurophysiology, a key branch of neuroscience, is dedicated to unraveling the complex mechanisms governing the nervous system's operations. Investigating neurophysiological phenomena necessitates a deep grasp of diverse concepts, theories, and experimental approaches. Consequently, having a highly competent question-answering system capable of addressing neurophysiology inquiries is of utmost importance to researchers, clinicians, and students in this field. Questions in the system can be categorized into two categories, lower-order and higher-order questions, aligned with Bloom's taxonomy, enabling the assessment of language models' ability to respond to queries in each category. Bloom's taxonomy, a widely utilized framework in educational contexts, classifies cognitive levels into six domains: knowledge, comprehension, application, analysis, synthesis, and evaluation . By applying Bloom’s taxonomy to evaluate LLMs, their efficacy in answering questions spanning various cognitive levels, including those in neurophysiology, can be gauged . By considering how well ChatGPT, Bard, and Claude perform at different topics and different levels of Bloom's taxonomy, their abilities to comprehensively and accurately address neurophysiology questions. Previous publications evaluating LLMs across various disciplines have covered fields such as, gastroenterology , pathology , neurology , physiology , , and solving case vignettes in physiology . In a cross-sectional study, the performance of LLMs on neurology board–style examinations were assessed using a question bank approved by the American Board of Psychiatry and Neurology. The questions were categorized into lower-order and higher-order based on the Bloom taxonomy for learning and assessment . To the best of our knowledge there was no study specifically on evaluating LLMs in the field of neurophysiology. Additionally, in studies within similar domains, most studies have investigated the ability of LLMs to provide accurate answers for multiple-choice questions – . To comprehensively understand the strengths and weaknesses of these models in a sophisticated field like neurophysiology, it is essential to evaluate the capabilities of these models in responding to essay questions, across all cognitive levels. Neurophysiology presents a diverse range of questions levels, making it a valuable area for assessing the strengths and limitations of LLMs. This study compares the performance of three language models, namely, ChatGPT, Bard, and Claude, in answering neurophysiology questions in both the Persian and English languages. It focuses on various cognitive levels based on Bloom's Taxonomy and evaluates the models' reasoning process by asking for the rationale behind their responses. The study aims to evaluate the performance of the LLMs in addressing neurophysiology questions in different cognitive levels, along with determining whether the models rely on memorization or demonstrate analytical reasoning and logical explanations. Moreover, it offers insights into the capabilities of the LLMs by identifying potential reasons for incorrect answers to determine their weaknesses in responding to neurophysiology questions. This exploratory, applicational and cross-sectional study was carried out using AI-driven chat applications, including ChatGPT (chat.openai.com), Claude (claude.ai), and Bard (bard.google.com), which offer free services for researchers. The researchers aimed to assess the strengths and weaknesses of the selected LLMs in their ability to answer neurophysiology questions. Questions A total of 20 questions were chosen from four topics in neurophysiology, including general, sensory, motor, and integrative systems, with each topic comprising 5 questions. The LLMs were asked to provide explanations for their selected answers for all questions, which encompassed true/false, multiple-choice, and essay formats. Therefore, all the questions were effectively essay questions allowing for a scoring range of 0–5 points for the responses. Furthermore, the questions were categorized based on cognitive skills into lower-order and higher-order categories, with each topic included 3 lower-order and 2 higher-order questions. It is worth noting that, according to Bloom’s taxonomy, memorization and recall are categorized as lower-level cognitive skills, necessitating only a minimal degree of comprehension. In contrast, the application of knowledge and critical thinking fall under the category of higher-level cognitive skills, requiring deep conceptual understating . A panel of three skilled physiologists was chosen to validate the questions and evaluate the answers of the LLMs to the questions. They were university lecturers who had at least 2 years of teaching experience in neurophysiology to medical students. The questions, topics, and cognitive skills are listed in Table 1 of Supplementary 1. Data collection The latest versions of ChatGPT 3.5 (November 21, 2023), Claude 2 (December 5, 2023), and Bard (November 21, 2023) were prompted with questions in both Persian and English languages. These versions are undergoing public testing for academic research. The Persian and English questions, along with the answers generated by the three selected LLMs, were stored in separate files for evaluation by the physiologists. Notably, prompt engineering is essential to improve the efficiency of LLMs. It includes strategies such as chain-of-thought (CoT) prompting and structured prompting . The CoT prompting has achieved the state-of-the-art performances in arithmetic and symbolic reasoning , . The model is instructed in the CoT prompting to provide step-by-step reasoning in generating a final answer, which could be few-shot or zero-shot . Utilizing structured prompting, which includes important components such as context, the expected behavior, and the format of the output, is another strategy for achieving optimal outcomes. In this study, zero-shot CoT was employed by adding "let's think step by step" into the questions. Also, the following structured prompt was used for all the questions: “Imagine you are an expert physiologist with a specializing in neurophysiology. Answer the following question. {question…}. Explain the steps and reasons that lead you to the answer. write your final answer. Let’s think step by step”. The panel of three physiologists was asked to score each question on a scale of 0 to 5 points, where a score of 5 indicated a full and comprehensive response to the question. All data were recorded in an Excel file for further analysis. Statistical analysis The statistical analysis employed mean, median and standard deviation to provide a comprehensive overview of the data. The Friedman test was used to assess if there were statistically significant variations in the scores of LLMs between Persian and English languages, with each group comprising 20 questions. Furthermore, the Kruskal‒Wallis’s test was carried out to assess the significance of score differences across four topics and two levels of cognitive skills. The intraclass correlation coefficient (ICC), a two-way random model with absolute agreement, was used to evaluate the level of agreement among the physiologists' scores. Furthermore, the Wilcoxon signed rank test was applied to ascertain the significant difference between the scores of LLMs in Persian and English. A p value of below 0.05 was considered statistically significant. All statistical analyses were performed using SPSS software, version 22. A total of 20 questions were chosen from four topics in neurophysiology, including general, sensory, motor, and integrative systems, with each topic comprising 5 questions. The LLMs were asked to provide explanations for their selected answers for all questions, which encompassed true/false, multiple-choice, and essay formats. Therefore, all the questions were effectively essay questions allowing for a scoring range of 0–5 points for the responses. Furthermore, the questions were categorized based on cognitive skills into lower-order and higher-order categories, with each topic included 3 lower-order and 2 higher-order questions. It is worth noting that, according to Bloom’s taxonomy, memorization and recall are categorized as lower-level cognitive skills, necessitating only a minimal degree of comprehension. In contrast, the application of knowledge and critical thinking fall under the category of higher-level cognitive skills, requiring deep conceptual understating . A panel of three skilled physiologists was chosen to validate the questions and evaluate the answers of the LLMs to the questions. They were university lecturers who had at least 2 years of teaching experience in neurophysiology to medical students. The questions, topics, and cognitive skills are listed in Table 1 of Supplementary 1. The latest versions of ChatGPT 3.5 (November 21, 2023), Claude 2 (December 5, 2023), and Bard (November 21, 2023) were prompted with questions in both Persian and English languages. These versions are undergoing public testing for academic research. The Persian and English questions, along with the answers generated by the three selected LLMs, were stored in separate files for evaluation by the physiologists. Notably, prompt engineering is essential to improve the efficiency of LLMs. It includes strategies such as chain-of-thought (CoT) prompting and structured prompting . The CoT prompting has achieved the state-of-the-art performances in arithmetic and symbolic reasoning , . The model is instructed in the CoT prompting to provide step-by-step reasoning in generating a final answer, which could be few-shot or zero-shot . Utilizing structured prompting, which includes important components such as context, the expected behavior, and the format of the output, is another strategy for achieving optimal outcomes. In this study, zero-shot CoT was employed by adding "let's think step by step" into the questions. Also, the following structured prompt was used for all the questions: “Imagine you are an expert physiologist with a specializing in neurophysiology. Answer the following question. {question…}. Explain the steps and reasons that lead you to the answer. write your final answer. Let’s think step by step”. The panel of three physiologists was asked to score each question on a scale of 0 to 5 points, where a score of 5 indicated a full and comprehensive response to the question. All data were recorded in an Excel file for further analysis. The statistical analysis employed mean, median and standard deviation to provide a comprehensive overview of the data. The Friedman test was used to assess if there were statistically significant variations in the scores of LLMs between Persian and English languages, with each group comprising 20 questions. Furthermore, the Kruskal‒Wallis’s test was carried out to assess the significance of score differences across four topics and two levels of cognitive skills. The intraclass correlation coefficient (ICC), a two-way random model with absolute agreement, was used to evaluate the level of agreement among the physiologists' scores. Furthermore, the Wilcoxon signed rank test was applied to ascertain the significant difference between the scores of LLMs in Persian and English. A p value of below 0.05 was considered statistically significant. All statistical analyses were performed using SPSS software, version 22. The responses from three LLMs, ChatGPT, Bard, and Claude, were collected for both Persian and English languages questions. Three experienced physiologists evaluated the responses. Each question was given to the LLMs only once, simulating a student answering neurophysiology questions in an exam setting. As a result, the ambiguity of the questions or the LLMs lack of understanding of the question content or the unimportant content that should not be mentioned in the responses could affect the scores that the LLMs received from each question. The Persian questions along with the answers of LLMs to these questions are shown in Supplementary 2, while the English questions along with the answers of LLMs are shown in Supplementary 3 and the evaluation results from the experts, including the average scores they assigned, are summarized in Supplementary 1 Table . The evaluation results using by ICC, showed good agreement among the physiologists in scoring. The ICC values for various topics ranged from 0.935 to 0.993. The ICC value for all questions was 0.978 (F = 51.217, p < 0.001). This high level of agreement in the physiologists' scores signifies the reliability of expert opinions. The results of the ICC test among the physiologists are shown in Table . Given the good agreement between the raters, the mean of their scores was used as the score for each question in the subsequent analysis. The evaluation results from the physiologists showed that the overall performance of selected LLMs in responding to the questions, as well as the performance of each of LLMs in both English and Persian languages, were deemed satisfactory (Table ). The overall mean score obtained for the questions was 3.87 ± 1.7. As illustrated in Fig. , the mean scores for various LLMs in the Persian and English languages ranged from 3.35 (Bard in Persian) to 4.50 (Bard in English). Nevertheless, the results of the Friedman test did not reveal any statistically significant difference in LLMs scores between Persian (p = 0.794) and English (p = 0.281). Overall, the average scores in English (Mean = 4.18, Median = 4.64) surpassed those in Persian (Mean = 3.56, Median = 4.72). However, the Wilcoxon signed rank test showed that this difference was not statistically significant (p = 0.222). Regarding different topics, the highest scores were associated with the motor system topic, while the lowest score was obtained for the integrative topic (Table ). Based on the results, the performance of LLMs can be generally evaluated as excellent for general and motor system topics, good for sensory system and integrative topics. The best scores for the English questions were attributed to the general topic, whereas the weakest scores for the Persian questions were linked to the sensory topic (Fig. ). The results of the Kruskal‒Wallis’s test revealed a significant difference in the scores for the integrative topic compared to other topics (p < 0.001). Moreover, regarding the cognitive level of the questions, the results of the Kruskal‒Wallis’s test indicated that there was no significant difference between the scores (p = 0.613). The lowest score of 3.51 was recorded for higher-order questions in Persian, while the highest score of 4.38 was achieved for lower-order questions in English (Fig. ). Figures and show the mean scores for different questions in the Persian and English languages. The proximity of the curves indicates the similarity in scores in different LLMs, while the closer the curves are to the outer edge of the diagram signifies higher scores for those question. The diagrams suggest that for most questions, there is a comparable performance level among different LLMs. However, this consistency is not observed for certain questions. For instance, in the Persian questions, for the Sensory_1 question, ChatGPT and Claude were provided nearly complete answers, but Bard received a score of zero. In addition, for Sensory_3, the scores of ChatGPT and Claude achieved fairly scores, while Bard was unable to answer the question. In contrast, for Integrative_3, both ChatGPT and Claude were unable to provide an answer, but Bard managed to receive a perfect score for the question (Fig. ). For English questions, there are also questions where there is no similarity in performance among LLMs. For example, both Bard and Claude received almost full scores for General_5, but ChatGPT struggled to provide correct answers to these questions. Moreover, for Motor_4, both ChatGPT and Claude were unable to offer a satisfactory response, whereas Bard's answer was almost complete. In contrast, for Integrative_4, both ChatGPT and Claude fell short in providing a good answer, but Bard managed to achieve a perfect score for the question (Fig. ). In addition to the inconsistency in responses, in some questions, almost none of the LLMs were able to adequately respond to the question. For further analysis, the questions to which LLMs couldn’t respond adequately were identified. The total possible scores of the three language models for each question in Persian and English were 15. Questions with a mean score of 3 or less for each LLM were selected based on the criterion. Therefore, questions for which the total score of all LLMs were equal or less than 9 were chosen. In Persian, the selected questions included General_5, Sensory_2, Sensory_3, Sensory_4, Integrative_1 and Integrative_4 questions. Additionally, for the English questions, the total score was below 9 for Motor_4, Integrative_1 and Integrative_4. General_5 question: Is myelination of postganglionic sympathetic fibers done by Schwann cells? The correct answer to this question is that postganglionic sympathetic fibers lack myelin. The use of the phrase “by Schwann cells” in the question stem is a misleading phrase. In the Persian language, none of the language models could provide the correct answer even after removing the misleading phrase from the question. Through further questions, it became clear that in Persian, postganglionic sympathetic fibers were incorrectly categorized as type A instead of type C. Also, none of the models had sufficient information regarding which types of fibers are myelinated. Hence, the cause of the wrong answer in the Persian language can be considered as "having inaccurate information" in the LLMs, but by removing the misleading phrase from the question, all LLMs were able to provide the correct answer in the English language. Therefore, the cause of the initial incorrect answer in English in the ChatGPT can be attributed to the presence of a “misleading phrase in the question”. Sensory_2 question: Are sexual sensations mostly transmitted through the posterior column—medial lemniscus? The correct answer to this question is “No”. In Persian, Bard did not provide a response to the question and instead wrote: “I am a language model and do not have the capacity to understand or respond to this query”. Probably the Persian equivalent of the phrase “sexual sensations” has led to this response. Two other LLMs also failed to provide a correct response. By changing the question and using the English phrase equivalent to the ‘posterior column-medial lemniscus’ in Persian all LLMs were able to provide the correct answer in Persian. Therefore, the reason for the wrong answer to this question in Persian can be expressed as “incorrect translation for phrases in Persian”. Sensory_3 question: State key components, including nuclei and neurotransmitters, in the central nervous system analgesic pathway? The correct answer is “the PAG projects enkephalinergic neurons to the Raphe, and after stimulation, the serotonergic projections go to the spine and stimulate the enkephalinergic neurons that cause pain inhibition”. In response to this essay question, the LLMs failed to mention some important nuclei or mentioned nuclei that were of lesser importance. This means that the most important phrase in the question was not considered. This lack of attention to importance was present in both the Persian and English languages responses, with a more pronounced effect in Persian. Thus, the reason for the incorrect response to this question can be stated as “not considering the importance and priority” and providing “insignificant additional explanation” compared to a knowledgeable individual in this field. Sensory_4 question: Which sensation is NOT transmitted through the anterolateral pathway? A) Chronic pain B) Cold sensation C) Touch sensation from Meissner receptor D) Touch sensation from Ruffini receptors. The sensation that is not transmitted through the anterolateral pathway is (C) Touch sensation from Meissner receptors. LLMs in English provided the correct answer to this question, whereas LLMs in Persian answered it incorrectly. Claude stated that Meissner receptors transmit the sensation of pressure to the brain, while Ruffini receptors transmit the sensations of contact and vibration. However, the opposite of this statement is correct. Moreover, ChatGPT and Bard offered general rather than specialized information with details regards to this question. Hence, the reason for the incorrect response in the Persian language can be attributed to as “inaccurate information” and “insufficient specialized knowledge” in Persian language concerning this question. Motor_4 question: Does microinjection of glutamate into the medullary reticular nucleus cause relaxation of axial muscles? The correct answer is “Yes”. ChatGPT and Claude failed to provide the accurate response to this question. Research indicates that neural projections can exhibit both excitatory and inhibitory functions. So, these two LLMs focused on the excitatory aspect. However, stronger evidence from textbooks supports the idea that neural projections can indeed be inhibitory. Therefore, the reason for the incorrect response can be attributed to “neglecting the significance of available evidence”. Integrative_1 question: In medical science and neurophysiology, is knowing “my birthday is January 10, 1998” an example of semantic explicit memory? The correct answer is “No”. Because stating my birthday date is only a claim about a past event, which can be considered a verified fact if supported by evidence confirming that event. None of the LLMs, except for Claude, managed to provide the correct response in either Persian or English. They mistakenly treated this statement as a fact. Most likely, the reason for that is the absence of a similar sentence in the training texts used for the LLMs. Therefore, the reason for the incorrect answer to this question can be considered “using non-existent example” and “lack of reasoning ability” for questions that require reasoning based on prior knowledge and applying that knowledge to the current context. Integrative_4 question: In medical science and neurophysiology, which of the following represents explicit memory? A) The Shahnameh is the masterpiece of the great Iranian poet named Ferdowsi B) Today I arrived about 7 minutes late to physiology class. I'm usually late for classes. C) In 2010 my house had a major fire D) One of my elementary school friends’ last names ended in “Abadi” or “Abadian” The correct answers are A and C. ChatGPT correctly identified that option A is a fact and pertains to semantic memory. Also, it initially stated that the explicit memory consists of semantic and episodic types. However, in the final summary, despite initially identifying it as semantic, it failed to categorize it as explicit memory. Regarding option B, it also correctly mentioned that it does not pertain to long-term memory and therefore, cannot be explicit memory. Yet in the final summary, it categorized it as explicit memory. For option D, the lack of accurate recollection of the past, a complete memory has not formed and therefore it is not explicit, which most LLMs failed to identify. Therefore, the reason for the incorrect answer can be considered as “insufficient specialized information” and “lack of reasoning ability”. The facts are correctly stated step-by-step, but combining these facts and deducing conclusions from them is not executed effectively. The correct answer to this question is that postganglionic sympathetic fibers lack myelin. The use of the phrase “by Schwann cells” in the question stem is a misleading phrase. In the Persian language, none of the language models could provide the correct answer even after removing the misleading phrase from the question. Through further questions, it became clear that in Persian, postganglionic sympathetic fibers were incorrectly categorized as type A instead of type C. Also, none of the models had sufficient information regarding which types of fibers are myelinated. Hence, the cause of the wrong answer in the Persian language can be considered as "having inaccurate information" in the LLMs, but by removing the misleading phrase from the question, all LLMs were able to provide the correct answer in the English language. Therefore, the cause of the initial incorrect answer in English in the ChatGPT can be attributed to the presence of a “misleading phrase in the question”. The correct answer to this question is “No”. In Persian, Bard did not provide a response to the question and instead wrote: “I am a language model and do not have the capacity to understand or respond to this query”. Probably the Persian equivalent of the phrase “sexual sensations” has led to this response. Two other LLMs also failed to provide a correct response. By changing the question and using the English phrase equivalent to the ‘posterior column-medial lemniscus’ in Persian all LLMs were able to provide the correct answer in Persian. Therefore, the reason for the wrong answer to this question in Persian can be expressed as “incorrect translation for phrases in Persian”. The correct answer is “the PAG projects enkephalinergic neurons to the Raphe, and after stimulation, the serotonergic projections go to the spine and stimulate the enkephalinergic neurons that cause pain inhibition”. In response to this essay question, the LLMs failed to mention some important nuclei or mentioned nuclei that were of lesser importance. This means that the most important phrase in the question was not considered. This lack of attention to importance was present in both the Persian and English languages responses, with a more pronounced effect in Persian. Thus, the reason for the incorrect response to this question can be stated as “not considering the importance and priority” and providing “insignificant additional explanation” compared to a knowledgeable individual in this field. The sensation that is not transmitted through the anterolateral pathway is (C) Touch sensation from Meissner receptors. LLMs in English provided the correct answer to this question, whereas LLMs in Persian answered it incorrectly. Claude stated that Meissner receptors transmit the sensation of pressure to the brain, while Ruffini receptors transmit the sensations of contact and vibration. However, the opposite of this statement is correct. Moreover, ChatGPT and Bard offered general rather than specialized information with details regards to this question. Hence, the reason for the incorrect response in the Persian language can be attributed to as “inaccurate information” and “insufficient specialized knowledge” in Persian language concerning this question. The correct answer is “Yes”. ChatGPT and Claude failed to provide the accurate response to this question. Research indicates that neural projections can exhibit both excitatory and inhibitory functions. So, these two LLMs focused on the excitatory aspect. However, stronger evidence from textbooks supports the idea that neural projections can indeed be inhibitory. Therefore, the reason for the incorrect response can be attributed to “neglecting the significance of available evidence”. The correct answer is “No”. Because stating my birthday date is only a claim about a past event, which can be considered a verified fact if supported by evidence confirming that event. None of the LLMs, except for Claude, managed to provide the correct response in either Persian or English. They mistakenly treated this statement as a fact. Most likely, the reason for that is the absence of a similar sentence in the training texts used for the LLMs. Therefore, the reason for the incorrect answer to this question can be considered “using non-existent example” and “lack of reasoning ability” for questions that require reasoning based on prior knowledge and applying that knowledge to the current context. The correct answers are A and C. ChatGPT correctly identified that option A is a fact and pertains to semantic memory. Also, it initially stated that the explicit memory consists of semantic and episodic types. However, in the final summary, despite initially identifying it as semantic, it failed to categorize it as explicit memory. Regarding option B, it also correctly mentioned that it does not pertain to long-term memory and therefore, cannot be explicit memory. Yet in the final summary, it categorized it as explicit memory. For option D, the lack of accurate recollection of the past, a complete memory has not formed and therefore it is not explicit, which most LLMs failed to identify. Therefore, the reason for the incorrect answer can be considered as “insufficient specialized information” and “lack of reasoning ability”. The facts are correctly stated step-by-step, but combining these facts and deducing conclusions from them is not executed effectively. Three LLMs, ChatGPT, Bard, and Claude, were used to assess their capacity in providing comprehensive and logical answers to neurophysiology essay prompts in both Persian and English languages. These LLMs can respond to complex commands by analyzing and comprehending the supplied text, utilizing their highly advanced natural language processing capabilities and their vast training datasets . The results showed that, overall, the models demonstrated commendable proficiency in addressing neurophysiology queries. However, certain variations among the models were observed depending upon the specific topic of the inquiries. Across the various topics analyzed, the LLMs performed the best on queries concerning to the motor system and general neurophysiology, indicating their strength in addressing fundamental principles. In terms of sensory system topics, the performance was moderately solid, suggesting that the models can comprehend and explain sensory neurophysiology to a certain degree. However, when faced with integrative questions, the scores significantly dropped. This underscores a present constraint of the models in tackling complex, multi-step reasoning requiring integration of knowledge across neurophysiology topics. Tailored training focusing on integrative concepts could help improve LLMs’ capabilities in this realm . Interestingly, although there were no significant disparities in the performance of the models in Persian and English or between lower-order and higher-order questions, a detailed analysis revealed some inconsistencies. A qualitative analysis of the responses unveiled deficiencies in reasoning capabilities, particularly evident in unfamiliar question scenarios that necessitate adaptable application of knowledge. For certain questions, one model excelled, while others faltered, without a discernible pattern. This lack of uniformity implies knowledge gaps and variances in the training of the distinct models . Additionally, all three models struggled with several complex questions in both languages, yielding subpar scores. This further underscore the limitations of these models in advanced reasoning and handling ambiguous and multifaceted questions. When comparing languages, the scores were mostly comparable for all the LLMs. The models appeared to have acquired sufficient linguistic knowledge proficiency to comprehend and provide accurate responses in both languages. Nonetheless, a few incorrect answers unique to Persian emphasized deficiencies in the information encoded in the models for that language. Overall, the outcomes confirm the effectiveness of LLMs for addressing neurophysiology inquiries in various languages. An in-depth review of the incorrect responses shed light on the specific limitations of the LLMs. Providing flawed information and the inability to discern key aspects of questions emerged as some of the deficiencies. However, some studies have reported a satisfactory reasoning level in LLMs , and a deficiency in reasoning for unfamiliar scenarios has been identified as one of the deficiencies in providing correct answers in various questions. These gaps need to be addressed through more extensive training of the models utilizing high-quality data encompassing diverse neurophysiology topics, contexts, and linguistic nuances. The subpar performance on integrative questions can be attributed to the models' reliance on memorization and pattern recognition from the training data rather than a profound comprehension of the concepts. Although large datasets help them to remember facts and terminology, it is still difficult for LLMs to integrate knowledge across topics to solve new problems. Although previous studies demonstrating that CoT prompting improves the reasoning abilities of the LLMs – , in this study, the utilization of zero-shot CoT prompting resulted in instances where the steps to arrive at an answer were correctly outlined, but the final conclusion based on these steps was inaccurate for certain neurophysiology questions. Therefore, it seems that in the field of neurophysiology, one of the main weaknesses of the LLMs lies in their reasoning capabilities. Further training focused on constructing causal models of physiology could address this issue more effectively than relying solely on statistical associations. The results of Mahowald et al. and Tuckute et al. align with the results we found in our study, indicating that LLMs excel in formal language abilities but exhibit limitations in real-world language understanding and cognitive skills. The Models lack reasoning skills, world knowledge, situation modeling, and social cognition , . Moreover, Schubert et al. concluded that higher-order cognitive tasks posed significant challenging for both GPT-4 and GPT-3.5 . While some researchers express cautious optimism in these cases and express their opinions such as Puchert et al., LLMs have transformed natural language processing and their impressive capabilities, concerns are raised regarding their tendency to generate hallucinations, providing inaccurate information in their responses. It is emphasized that rigorous evaluation methods are essential to ensure accurate assessment of LLM performance. Evaluations of LLM performance in specific knowledge domains, based on question-and-answer datasets, often rely on a single accuracy metric for the entire field, which hampers transparency and model enhancement . Loconte et al. claimed that while ChatGPT was well known to exhibit outstanding performance in generative linguistic tasks, its performance on prefrontal tests exhibited variability, as they reached the results, with some tests yielding results well above average, others falling in the lower range, and some showing significant impairment . These diverse perspectives underscore the need for a nuanced understanding of LLMs capabilities and limitations across different cognitive tasks and domains. Overall, the study findings demonstrate that LLMs like ChatGPT, Bard, and Claude have achieved impressive proficiency in responding neurophysiology questions, however, they still face challenges in some aspects of knowledge application, reasoning, and integration. It is evident that there is room for improvement in how these models operate, particularly in answering complex and ambiguous questions that require multistep reasoning and integration of knowledge across diverse topics. The variability observed among different models also highlights the need for ongoing evaluation. As LLMs continue to evolve, rigorous assessment across various knowledge domains will be essential for their continued enhancement and effectiveness. This study provides insights into the capabilities of LLMs in answering neurophysiology questions. The results indicate that ChatGPT, Bard, and Claude can successfully address numerous fundamental concepts but face challenges when it comes to more complex reasoning and integration and synthesizing information of knowledge across different topics. Overall, the models demonstrated relatively strong performance on general neurophysiology and motor system questions with moderate proficiency in sensory neurophysiology. However, they struggled with integrative questions requiring multistep inference. There was no significant difference between languages or cognitive levels. Nevertheless, qualitative analysis revealed inconsistencies and deficiencies, indicating that the models rely heavily on memorization rather than a profound conceptual grasp. The incorrect responses underscore shortcomings in reasoning, discerning key information, considering the level of importance and priority levels, lack of sufficient information specially in Persian and handling unfamiliar questions. Tailored training focusing on causal physiologic models instead of statistical associations and utilizing reliable sources in various languages could help overcome these limitations. As LLMs advance, rigorous multidisciplinary assessments will be essential to gauge progress and measure advancements. This study provides a robust evaluation methodology and benchmark for future research aimed at enhancing the neurophysiology knowledge and reasoning competence of these models. The insights can inform efforts to refine LLMs through advanced training techniques and the evaluation of complex integrative tasks. By focusing on targeted improvements, these models hold immense promise in advancing neurophysiology education, research, and clinical practice. The study's findings pave the way for further advancements in LLM technology, ultimately benefiting the field of neurophysiology and beyond. Supplementary Tables. Supplementary Information 2. |
The Amount of Orthodontic Force Reaching the Dental Pulp and Neuro-Vascular Bundle During Orthodontic Movements in the Intact Periodontium | 7ee887fc-2542-4c4b-9555-129e9394aef0 | 11727835 | Dentistry[mh] | The amount of orthodontic force applied during treatment can induce various local circulatory disturbances in the dental pulp and its neuro-vascular bundle (NVB) . Their severity and intensity depend on the amount of the applied orthodontic force that reaches these tissue structures . However, the same local circulatory disturbances also trigger orthodontic movement . It must be emphasized that much of the orthodontic load is absorbed and dissipated by the tooth and periodontal ligament before reaching the pulp and NVB . Nevertheless, this ability has not yet been individually assessed . The local physiological maximum hydrostatic pressure (MHP) present at this level is about 16–22 KPa (about 80% of the systolic pressure) . If this amount is exceeded, local ischemia is induced, which triggers the local remodeling processes related to orthodontic movement. The time and intensity highly influence the outcomes of these circulatory disturbances . In healthy intact tissue, usually under light and moderate orthodontic forces, there is limited ischemia, with no further consequences due to the physiological tissue’s ability to sustain, limit, and recover after tissular damage . However, if there has been previous occlusal trauma with a strong impact on the NVB and/or coronal pulpal injuries due to direct and indirect pulp capping , this ability to sustain damage is severely diminished and ischemia is triggered, closely followed by local tissular necrosis and further degenerative and resorptive processes . These previously mentioned injuries and trauma are rarely clinically visible and usually pass unnoticed . Their effects become clinically visible only after the irreversible pathological processes start . Thus, the same amount of orthodontic force can be safely applied in healthy intact tissue or induce severe effects in other tissue types with a clinically healthy appearance . Thus, there is a need to investigate the individual biomechanical behavior of the pulp and its NVB under orthodontic forces and movements in healthy intact tissue. It must be emphasized that no similar studies can be found except our previous ones . The tissular biomechanical behavior of dental tissue can be assessed only through numerical studies . No such studies can be found, mainly due to the complexity of the anatomical model’s reconstructions . However, multiple numerical studies of the periodontal ligament– and bone–implant interface are available in the current research flow, reporting various results due to accuracy issues . It must be emphasized that the periodontal ligament anatomically holds the NVB structure in its apical third, with a strong impact on its biomechanical behavior . Numerical studies derived from the engineering field are widely used and recognized for their remarkable accuracy . In the dental field, they were introduced in the past decade but have provided multiple contradictory reports with accuracy issues due to the misunderstanding of the principles underlying numerical studies . In our previous studies, our team identified the main accuracy issues and provided methodological improvements, reporting accurate results . These were related to the use of brittle and hydrostatic failure criteria for ductile dental tissue, low-accuracy tissular models, boundary assumptions that are inconsistent with the clinical reality, and a lack of correlation with the MHP. However, all data needed to overcome these above-mentioned problems are also available in the engineering field . The current numerical dental studies provide contradictory results on the safest amount of force to be applied in an intact periodontium during orthodontic treatment, ranging between 0.2 and 6 N, while the most stressful movements reported are rotation, intrusion, and translation. Moreover, recent reports are conflicting regarding the poor quality of multiple in vivo studies due to methodology issues. Thus, there is a need for newer numerical studies to assess the individual biomechanical behavior of dental tissue, following the methodological requirements to report correct results. The objective of this study was to assess the amount of orthodontic force applied at the bracket level that produces effects at the NVB and dental pulp levels during five orthodontic movements performed in an intact periodontium.
Our study was part of a larger, stepwise research work, with clinical protocol 158/02.04.2018, continuing to study the effects of orthodontic forces and movements on dental tissue during periodontal breakdown . This study focused on the biomechanical behavior of the pulp and NVB in the intact periodontium, focusing on the second lower premolars of nine patients, namely four males and five females with a mean age of 29.81 ± 1.45 years, totaling 180 numerical simulations. The inclusion criteria consisted of a complete mandibular dental arch in the region of interest (i.e., the two premolars and first lower molar), no malposition, intact healthy teeth, 1–2 mm bone loss in the region of interest, a healthy periodontium, orthodontic treatment indication, and good oral hygiene. The exclusion criteria consisted of a particular root geometry, an abnormal crown shape, abnormal root surface defects, radiologically visible bone defects, an abnormal pulp chamber and root canals, more than 2–3 mm bone loss, and poor oral hygiene after acceptance. By using a CBCT scan (ProMax 3DS, Planmeca, Helsinki, Finland; voxel size of 0.075 mm), the lower premolar regions of the nine patients were investigated. The CBCT scans were then loaded into AMIRA 5.4.0 (Visage Imaging Inc., Andover, MA, USA), an image reconstruction software program, where each individual tissular component was identified and manually reconstructed. Thus, the enamel, dentine, dental pulp, periodontal ligament (PDL), neuro-vascular bundle (NVB), and trabecular and cortical bone were identified and reconstructed into 3D models. The alveolar sockets of the first lower molar and premolar were filled with trabecular and cortical bone. The cementum could not be separated from the dentine; thus, it was reconstructed as dentine due to their similar physical properties; see . The PDL was reconstructed with the anatomical variable thickness of 0.15–0.225 mm. The missing bone and periodontal ligament were reconstructed to obtain 3D models with an intact periodontium. The base of a stainless-steel bracket was reconstructed on the enamel of the crown. The results provided by AMIRA 5.4.0 were nine intact periodontium mesh models with 5.06–6.05 million C3D4 tetrahedral elements, 0.97–1.07 million nodes, and a global element size of 0.08–0.116 mm. No element errors were found in the mesh models. However, a few element warnings were present in non-essential areas, while the essential areas were continuous. In the pulp and NVB of one of the models, four element warnings were found, amounting to 0.0158% of the total of 25,252 elements. In the same mesh model, the tooth mesh displayed thirty-nine element warnings, amounting to 0.00589% of the total of 661,137 elements. All internal algorithms checks were successfully passed, and the mesh was prepared to be imported into the numerical analysis software. ABAQUS 6.13-1(Dassault Systèmes Simulia Corp., Maastricht, The Netherlands) was used for the numerical study. The simulated orthodontic movements consisted of intrusion, extrusion, rotation, tipping, and translation. The orthodontic forces were 0.5 N/5 KPa and 4 N/40 KPa. The employed failure criteria were those for ductile-like materials, namely Von Mises (overall, homogenous) and Tresca (shear, non-homogenous), which are suitable for dental tissue. As boundary assumptions, isotropy, homogeneity/non-homogeneity, linear elasticity, perfectly bonded interfaces, and zero displacements in the bases of the models were assumed (similarly to most numerical studies). The numerical simulations’ results were displayed as color-coded projections of various intensities and consisted of red–orange high, yellow–green moderate, and blue low shades. The mean average NVB and pulpal stress for each movement and failure criterion were registered. The amount of stress was compared with the local maximum physiological hydrostatic pressure of 16–22 KPa. The results were then correlated with known clinical biomechanical behavior and other numerical studies.
The numerical simulations with the two methods showed that, in the intact periodontium, only a small amount of the initial orthodontic load produced effects in the NVB and the dental pulp ( and ). Both numerical methods displayed similar qualitative results, while the quantitative ones were slightly smaller for Von Mises when compared with Tresca. The quantitative difference between the Tresca and Von Mises results, for both forces, was an average of 1.11 times (1.17 times for NVB and 1.05 times for pulp), falling within the scientifically specified average interval of 1.15 times (around 10%). In the intact periodontium, the orthodontic forces induced NVB stresses that were 5.7 times higher than in the dental pulp. The orthodontic loads produced visible effects at the NVB level and almost none at the pulp level (except in the translation movement). Thus, from the initial 4 N/40 KPa of orthodontic load, only around 1 KPa was manifested as stress at the NVB level. Both methods similarly showed the rotation, followed by the extrusion and intrusion, displaying stresses of around 1 KPa at the NVB level, i.e., about 2.85% of the initial orthodontic load of 40 KPa applied at the bracket level. This implies that the remaining 97.15% of the orthodontic load was absorbed and dissipated by the other dental components—the dentine, enamel, stainless-steel bracket base, and PDL—before arriving at the NVB. Since this amount was lower than the physiological maximum hydrostatic circulatory pressure of 16–22 KPa at this level, a load of 40 KPa seems to be safe for the NVB and pulp. During the rotation movement, the pulp received only 0.17–0.20 KPa of the initial 40 KPa applied at the bracket level, i.e., about 0.5% of the applied load, while the remaining 99.5% seemed to be absorbed and dissipated by the enamel, dentine, and stainless-steel bracket. The 0.5 N/5 KPa load produced similar stress results, with 2.8% of the initial load producing effects at the NVB level and 0.02% at the pulp level, showing a tissular absorption–dissipation capacity of 97.2–99.98% (by the dentine, enamel, stainless-steel bracket base, and PDL) of the initial applied amount of orthodontic force. Both the Tresca and Von Mises methods showed similar color-coded stress displays for both loads. The NVB tissular deformations were more visible for intrusion and extrusion movements. Translation and rotation movements were the only ones to display coronal pulp stress , which was higher on the proximal sides (mesial and distal) and lower on the vestibular one. These amounts of stress, due to being lower than the maximum physiological hydrostatic circulatory pressure, had no impact on the healthy and intact pulp and NVB (since this tissue has a natural ability to sustain damage). Quantitatively, the rotation movement induced the highest stresses in the NVB, closely followed by intrusion and extrusion, appearing to be the most stressful for the NVB. For the dental pulp, rotation remained the most stressful, closely followed by tipping and translation.
This analysis aimed to assess, in the intact periodontium, the amount of load reaching the dental pulp and its NVB when compared with the initial applied orthodontic load. To obtain correct results, only two numerical methods are suitable for these tissue types . The individual study of the pulp and NVB is possible only through numerical methods . Five orthodontic movements and two loads (0.5 N—light and 4 N—moderate to large) were used in a total of 180 simulations. We must emphasize that our study is the first to assess this issue and the benefits of this approach. No other studies, except our earlier ones , related to these issues can be found in the current research flow. Both used methods were reported to be the most correct and suitable for dental tissue, being especially designed for ductile resemblance in homogenous and non-homogenous materials . Dental tissue has been reported to have ductile resemblance but with a certain brittle flow mode . Both the Von Mises and Tresca methods reported similar color-coded stress display projections in the pulp and NVB, with higher tissular deformation visible in the NVB during the five orthodontic movements and two forces, in line with our earlier reports . The quantitative results displayed a similar pattern, with the amounts of stress being lower than the physiological maximum hydrostatic pressure of 16–22 KPa, and being 5.7 times higher in the NVB than in the pulp, in line with earlier reports . From the total amount of the applied load, only around 2.8% induced stress in the NVB and 0.02–0.5% in the dental pulp. Thus, 97.2–99.9% of the load was absorbed and dissipated by the enamel, dentine, and stainless-steel bracket, in line with earlier reports . All previously mentioned tissue types and materials are considered to have ductile biomechanical behavior . Only two studies related to this subject can be found in the current research flow, both belonging to our team. In an earlier report , we found an absorption–dissipation capacity of 40–93% for the dentine, 40–65% for the enamel, and 16% for the stainless-steel bracket, in line with the present work. The same study reported that, from the total amount of the orthodontic load, only a small percentage reached the PDL (0.3–8.4%), NVB (0.2–0.7%), and dental pulp (0.02–0.12%), also in agreement with the present work. The other study reported the tooth structure’s absorption–dissipation capacity to be 85% of the stress before reaching the sensitive circulatory tissue, with 86.66–97.5% of the stress being dissipated before reaching the PDL, 98% before reaching the NVB, and 99.6–99.94% before reaching the pulp, in agreement with the present work. Ductility is the ability of a material to deform under a load without breaking and to recover its original form when the load is removed . We must emphasize that, under the orthodontic forces and during the movements, the displacements and tissular deformations were extremely small when compared with those described by physical mechanics in the engineering field . Forces of 0.5 and 4 N are extremely small from an engineering point of view . Contradictory results were obtained when using other numerical methods (maximum and minimum principle—brittle-like materials and hydrostatic pressure—liquid/gas), since these failure criteria were especially designed in the engineering field for the study of a specific type of biomechanical behavior. Earlier studies by our group reported these methods to be less exact than the Tresca and Von Mises methods . Since no similar studies could be found, the only possible correlations could be performed with PDL numerical studies (since, anatomically, the NVB is held by the periodontal ligament in the apical third). However, these PDL numerical studies mostly used the maximum and minimum principle, as well as hydrostatic pressure criteria (proven by our earlier studies to suffer from accuracy issues), reporting results that varied from one study to another or simply contradicting the clinical knowledge (but without any discussions about these issues). In particular, Wu et al. employed hydrostatic pressure when assessing the optimal orthodontic force in premolar rotation. They reported 2.1–2.9 N rotation to be optimal, showing amounts of stress lower than the MHP and finding only PDL apical third stress and no cervical third stress, which contradicts the existing clinical knowledge. Hofman et al. reported values of 80 KPa for 1 N of intrusion and around 40 KPa for 3–6 N lingual torque, implying degenerative and resorptive processes, which clinically is incorrect. Even new studies persist in adopting the same approach of employing brittle and hydrostatic liquid criteria for the study of dental tissue with a ductile-like resemblance, without considering the abovementioned issues . From a biomechanical point of view, during the orthodontic movements, intrusion and extrusion displayed higher NVB tissular deformations, in agreement with Minch et al. and Hofman et al. . Translation (barely visible in rotation) displayed coronal pulp stress areas, found on the mesial and distal sides (higher) as well as on the vestibular side (lower). In both the NVB and dental pulp, rotation displayed the highest amounts of stress, in agreement with Wu et al.’s numerical reports . However, the abovementioned factors seemed not to induce any effects in the healthy intact tissue, in agreement with other reports . Regarding the stress absorption–distribution pattern, 4 N/40 KPa of rotation, tipping, and translation displayed values of 0.13–0.20 Kpa in the coronal pulp (0.5% of the applied load), showing an absorption–dissipation capacity of 99.5% for the dental tissue surrounding the pulp, i.e., the enamel and dentine (the bracket also might have a contribution), in line with previous reports . The same amount of 40 KPa during rotation, tipping, extrusion, and intrusion induced 0.93–1.14 Kpa in the NVB (2.8% of the applied load), showing an absorption–dissipation capacity of 97.5% in the surrounding dental tissue, i.e., the PDL, dentine, and enamel (the bracket might also contribute), in line with other reports . This percentage difference between the pulp and NVB is probably linked to the PDL component, since the neuro-vascular bundle is held in the apical third of the periodontal ligament, suffering from important tissular deformations during movement . It must be emphasized that the orthodontic movements and tissular deformations are extremely small (as herein reported), with no induced ischemic risks due to their natural ability to adapt to trauma , as reported in our previous works and other reports . It must also be emphasized that the previously mentioned findings refer to intact healthy tissue with an intact ability to adapt to ischemia and to prevent local resorptive and degenerative processes. If these tissue types were previously traumatized and/or injured, as is clinically possible through occlusal trauma and direct or indirect pulpal capping , their functional ability to adapt would be diminished due to internal pathological and histomorphological changes . Thus, if, in the intact healthy periodontium, a light or moderate (as herein proven) orthodontic force induces no ischemic risks when triggering a movement, the same force could pose ischemic risks in the presence of a traumatized and/or injured NVB and dental pulp , and the used amount must be considered with care. To provide accurate results, the numerical methods must follow the engineering field’s requirements: failure criteria must be selected on the basis of the analyzed type of material, with boundary assumptions describing the behavioral conditions and accurate and detailed 3D models . Nevertheless, most of the numerical dental studies found in the current research field do not acknowledge the abovementioned aspects , with their reports showing various accuracy issues . The individual study of the dental pulp and NVB can be achieved only through numerical studies. For dental studies, the only suitable failure criteria are the Von Mises and Tresca ones, since they are the only ones to be specifically designed for ductile-like materials such as dental tissue . All other methods are designed for brittle and liquid-like substances and provide less accurate results when employed, as previously reported . The boundary assumptions for dental studies are related to homogeneity/non-homogeneity, isotropy–anisotropy, and linear/non-linear elasticity, as well as the amount of applied load . Natural tissue is non-homogenous, anisotropic, and non-linear elastic. Numerical studies cannot completely reproduce the clinical reality, but they are the only possible option to individually study these small tissue types . The applied load is extremely important when describing the tissular biomechanical behavior, since, in biomechanical physics, under small forces, all materials show linear elasticity . It must be emphasized that forces of up to 4 N are considered extremely small from an engineering point of view . Moreover, there are multiple comparative reports assessing various amounts of force ranging from 0.2 to 4.8 N, as conducted by our team, reporting similar stress distribution areas for these forces, confirming the view of the engineering field . These abovementioned findings are also supported by the clinically identified extremely small tissular deformations and displacements . Thus, under these circumstances, the linear elasticity assumption used in previous dental studies is correct . Zhang et al. reported that, when using the abovementioned assumed boundary conditions, the results were closer to those of clinical and animal studies. In dental studies, only the PDL is assessed as non-linear or linear, reporting various results, while all other components are seen as linear. Toms et al. investigated 1 N of extrusion in an intact PDL and reported higher amounts for linear than non-linear when using the maximum and minimum principle method. In contrast, Hemanth et al. assessed these differences in an intact periodontium, reporting, for the maxillary central incisor PDL, values ranging between 20 and 50% (148,097 elements and 239,666 nodes). Nevertheless, they considered 0.2–1 N for intrusion and tipping but used the maximum and minimum principle method (specifically designed for brittle materials) to describe the biomechanical behavior of ductile-like tissue. The homogeneity and non-homogeneity issue is also approached herein, since the Von Mises method assesses homogenous materials while Tresca considers non-homogenous ones . The average difference between them in scientific engineering is around 1.15 times (about a 10–15% difference), with Tresca being higher than VM, with the present results (1.11 times/11% difference) falling within the specified interval, in line with our previous reports . This issue has not been addressed in existing numerical studies . In our study, rotation, closely followed by tipping, extrusion, and intrusion, was more stressful for the NVB, while, for the dental pulp, rotation remained the most stressful, followed by tipping and translation. Our findings are in line with Wu et al. and Hofman et al. , despite their use of brittle-like failure criteria for ductile dental tissue. However, there are many other numerical studies reporting various contradictory results regarding the most stressful orthodontic movement, although none of them includes a comparison of the five movements or employs ductile-like materials . The levels of detail in the 3D model are of extreme importance, especially when studying small and complex dental tissue . Most numerical studies have employed low levels of detail in anatomically idealized models, with a reduced number of elements and nodes and a larger global element size, strongly affecting the accuracy of the results. Thus, most studies have employed low-anatomical-accuracy 3D models (without NVB or dental pulp tissular reconstructions), with a large global element size of up to 1.2 mm and a small number of elements/nodes, e.g., 1674/5205-23563/32812-1.67 million. The present study employed models with a reduced global element size of 0.08–0.116 mm and a large number of elements/nodes of 5.06–6.05 million/0.97–1.07 million, better meeting the anatomical accuracy criteria. The sample size is specific to the numerical study due to the great variability in the experimental conditions, leading to different results . Most dental numerical studies have used a sample size of one . To enhance our results’ accuracy, we used a sample size of nine. The limitation of the numerical method lies in not being able to entirely reproduce the clinical conditions. In the clinical setting, there are often associations among movements and not a pure movement, as in the present analysis. This could mean that the clinical amounts of stress are in fact even lower than those identified, but with no impact on their accuracy. Moreover, this is the only available possibility for individual tissular studies. To ensure accuracy, the method needs anatomically correct 3D models (which are extremely difficult to reconstruct), and it is necessary to employ failure criteria that are suitable for the analyzed material, while the boundary assumptions should agree with the clinical conditions (as previously discussed). Any deviations from these mandatory requirements will lead to incorrect results and can be considered as limiting the method. Thus, the reduced number of numerical studies meeting the abovementioned criteria, as well as reports regarding the poor quality of multiple in vivo studies used for indirect validation, emphasizes the need for more numerical studies in order complete the data. Moreover, since the PDL is reported to play a key role in the load’s absorption–dissipation during tooth biomechanics, further numerical simulations are needed to assess this ability.
The numerical simulations with the two methods showed that, in the intact periodontium, only a small amount of the initial orthodontic load produced effects in the NVB and dental pulp. Only about 2.85% of the initial orthodontic load of 40 KPa/4 N applied at the bracket level induced stresses in the NVB, while this was 0.5% for the dental pulp. A similar distribution was seen for the force of 5 KPa/0.5 N. The absorption–dissipation ability of the dental tissue varied between 97.15 and 99.98%. Both forces displayed quantitative stress amounts that were lower than the maximum physiological hydrostatic circulatory pressure, appearing not to produce harmful effects on the healthy and intact pulp and NVB. Quantitatively, the rotation movement seems the most stressful for the NVB, closely followed by intrusion and extrusion. For the dental pulp, rotation remained the most stressful, closely followed by tipping and translation. Tissular deformations were visible for the NVB area during intrusion and extrusion. The dental pulp showed pulpal stresses under translation and rotation. Clinically, it appears that up to 4 N of applied force is safe for the NVB and dental pulp, since only 0.5–2.85% of the load reaches these tissue types. Practical Implications Numerical studies of the dental pulp and NVB are scarce; only a few are available, with most of them belonging to our research team. This is the first numerical study to directly investigate the amount of stress reaching these small dental tissue types (about 0.02–2.85%) when compared with the total amount of applied orthodontic force. This study, conducted on the intact periodontium and healthy tissue, proved that both light and moderate orthodontic forces have similar quantitative tissular absorption–dissipation patterns, with an absorption rate of 97.15–99.98%. Moreover, forces of 0.5 and 4 N displayed similar biomechanical behavior, with tissular deformations of the NVB and coronal pulpal stress areas induced by translation and rotation movements, with no impact on the healthy intact tissue but with importance for previously traumatized tissue. Moreover, using the only two numerical methods suitable for the dental tissue, our study proved that, up to 4 N, there is no NVB stress exceeding the local physiological maximum hydrostatic pressure; this is of importance for the clinical phase of orthodontic treatment. For researchers, this study provides information on how to conduct an accurate numerical dental study, as well as a clear picture of the problems related to the boundary assumptions and the method selection.
Numerical studies of the dental pulp and NVB are scarce; only a few are available, with most of them belonging to our research team. This is the first numerical study to directly investigate the amount of stress reaching these small dental tissue types (about 0.02–2.85%) when compared with the total amount of applied orthodontic force. This study, conducted on the intact periodontium and healthy tissue, proved that both light and moderate orthodontic forces have similar quantitative tissular absorption–dissipation patterns, with an absorption rate of 97.15–99.98%. Moreover, forces of 0.5 and 4 N displayed similar biomechanical behavior, with tissular deformations of the NVB and coronal pulpal stress areas induced by translation and rotation movements, with no impact on the healthy intact tissue but with importance for previously traumatized tissue. Moreover, using the only two numerical methods suitable for the dental tissue, our study proved that, up to 4 N, there is no NVB stress exceeding the local physiological maximum hydrostatic pressure; this is of importance for the clinical phase of orthodontic treatment. For researchers, this study provides information on how to conduct an accurate numerical dental study, as well as a clear picture of the problems related to the boundary assumptions and the method selection.
|
Stent-graft implantation for late postpancreatectomy hemorrhage after pancreatoduodenectomy | 11015547-c246-44b5-8a5b-5738b61621c3 | 11840307 | Surgical Procedures, Operative[mh] | Introduction Pancreatoduodenectomy (PD) is the standard surgical procedure for resectable pancreatic head lesions. With the improvement of surgical technique and perioperative management, the mortality rate of PD dropped significantly to less than 5% in high-volume centers. , Postpancreatectomy hemorrhage (PPH) is a life-threatening complication of PD. According to its time of onset, PPH was classified as early (within 24 h after the end of the index operation) or late (over 24 h) according to the International Study Group of Pancreatic Surgery (ISGPS) definition. The underlying pathophysiological process of late PPH was different from early PPH, possible mechanisms including enzymatic digestion of the blood vessel, intraabdominal infection with involvement of peripancreatic vessels, and vascular injury during resection that leads to a pseudoaneurysm. The incidence of late PPH ranged from 3% to 16%, while its mortality rate was up to 21%. Invasive treatment options for late PPH includes relaparotomy, endoscopic intervention, and endovascular intervention. Relaparotomy is recommended for hemodynamically unstable patients, and endoscopy is recommended for hemodynamically stable patients in certain settings. Endovascular intervention is the recommended treatment option for initial management of late PPH in both hemodynamically stable and unstable patients. Transarterial embolization (TAE) and stent-graft implantation are the 2 major methods of endovascular intervention. , TAE raises the concern of impaired perfusion of distal organ, especially when hepatic artery is embolized, the physiological function of the liver could be influenced. With the development of covered stent-graft, stent-graft implantation becomes the preferred technique to treat late PPH in our center. The advantages of covered stent-graft includes: (1) The procedure is minimally invasive; (2) angiography has high sensitivity in identifying the bleeding site; (3) the hemostasis is conclusive after excluding the bleeding site and could be reconfirmed with angiography; (4) a conduit is built with a covered stent-graft for perfusion of the distal organ. With this study, we aim to report the single-center experience in treating PPH with covered stent-graft implantation. Methods 2.1 Study population Between April 2020 and December 2023, a total of 1723 pancreatectomies were performed, including 1068 PD cases and 655 total pancreatectomy cases. Among those, 16 patients underwent endovascular intervention owing to late PPH, including 4 cases treated with TAE and 12 cases treated with covered stent-graft implantation. These patients were followed and outcomes were recorded. Written informed consent was obtained from each patient. This study conformed to the ethical principles of the Declaration of Helsinki and was approved by the institutional review board. An electronic, prospective database is regularly updated, collecting demographics, intra-operative data, and postoperative outcomes for all pancreatic resections performed at our institution. Data were retrieved from the database and retrospectively reviewed. 2.2 Endovascular procedure Local anesthesia was used in all cases. The left/right common femoral artery was incised as an access route based on previous contrast-enhanced CT. The guidewire was introduced from the access route together with a 6 or 7 Fr catheter (Destination, Terumo, Tokyo, Japan). After arriving at the plane above the celiac artery, a 5 Fr pigtail angiographic catheter (COOK, Bloomington, IN, USA) was exchanged to perform angiography. The bleeding site and anatomical pattern of relevant arteries were identified according to the distribution and leakage of the contrast medium. After the initial angiography, the 5 Fr pigtail angiographic catheter was exchanged to 0.035 (Microport, Shanghai, China) or 0.018 (Boston Scientific, MA, USA) guidewire together with suitable catheter (Cobra; Medikit, Tokyo, Japan) to select the celiac artery and advance to the bleeding artery (common hepatic artery, right hepatic artery, proper hepatic artery, splenic artery, gastroduodenal artery stump, and superior mesenteric artery). The guidewire was advanced to the distal region of the bleeding site to allow sufficient zone for stent-graft deployment, and sufficient proximal/distal landing zone. Self-expandable covered stent-graft Viabahn (TBE, WL Gore, Flagstaff, AZ, USA) was used in all cases, the diameter of which ranged from 5 to 8 mm. Angiography would be performed after the deployment of stent-graft. If persistent contrast extravasation, incomplete exclusion of pseudoaneurysm, or insufficient expansion of the stent-graft was detected, balloon dilatation or extra bridging stent-graft implantation would be performed. Final angiography was performed to reconfirm the complete exclusion of pseudoaneurysm, no leakage of contrast, normal distal organ perfusion, ideal stent-graft shape, and no sign of endoleak. 2.3 Perioperative routine After the endovascular intervention, all patients were sent to the intensive care unit and monitored with cardiac monitoring for at least 24 h. Complete blood count and liver function test were performed immediately after the operation and daily thereafter. Only after a comprehensive evaluation of the patients’ vital signs, laboratory findings, and volume and color of drainage, the patient would be transferred back to the general ward. If any sign of recurrent bleeding was observed, the contrast-enhanced CT would be immediately performed, and endovascular intervention or relaparotomy would be considered. Coagulation was not considered in concern of rebleeding or high risk of bleeding. After discharge, dual antiplatelet therapy with aspirin (100 mg) and clopidogrel (75 mg) was performed according to the judgment of individual physicians. 2.4 Definition PPH was recorded according to the ISGPS classification. Based on the severity of hemorrhage, PPH was defined as mild (small or medium volume blood loss; mild clinical volume impairment; therapeutic consequence; no need for reoperation or interventional angiographic embolization) or severe (large volume blood loss; clinically significant impairment; need for blood transfusion; need for invasive treatment). Postoperative pancreatic fistula (POPF) was defined as a drain output of any measurable volume of fluid with an amylase level over 3 times the upper limit of institutional normal serum amylase activity, associated with a clinically relevant development/condition related directly to the POPF. Technical success was defined as angiography showing complete cessation of bleeding, patent stent-graft, and normal end-organ perfusion of the affected aorta. Clinical success was defined as resolution of hemorrhagic symptoms (based on vital signs, laboratory tests, and hemorrhagic drainage) without additional endovascular or surgical intervention. Procedure-related complications were evaluated according to the Claviene-Dindo classification. Survival or any other adverse events were defined as those occurring during hospitalization (in-hospital mortality), within 90 days postoperatively (90-day mortality), and after 90 days postoperatively (follow-up mortality). The reason for death and adverse events were described in detail and recorded as indeterminate if not available. After discharge, patients were followed every 3 months at the outpatient. Contrast-enhanced CT was performed at 3-month intervals for 1 year and every 6 months thereafter. Stent-graft patency and its end-organ perfusion were evaluated with contrast-enhanced CT during follow-up. Primary patency was patent vessel without endovascular reintervention, and secondary patency was patent vessel after endovascular reintervention. 2.5 Statistical analysis Continuous data are reported as means ± standard deviation (SD) when normally distributed or as median (Q 1 , Q 3 ) when the assumption of normality was not met according to Shapiro-Wilk tests. Categorical data are reported as n . A p < 0.05 was considered statistically significant. Kaplan-Meier estimates were used for stent patency and patients’ survival. The data were analyzed using SPSS version 21 (SPSS Inc, Chicago, IL, USA) and GraphPad Prism (GraphPad Software Inc, La Jolla, CA, USA). Study population Between April 2020 and December 2023, a total of 1723 pancreatectomies were performed, including 1068 PD cases and 655 total pancreatectomy cases. Among those, 16 patients underwent endovascular intervention owing to late PPH, including 4 cases treated with TAE and 12 cases treated with covered stent-graft implantation. These patients were followed and outcomes were recorded. Written informed consent was obtained from each patient. This study conformed to the ethical principles of the Declaration of Helsinki and was approved by the institutional review board. An electronic, prospective database is regularly updated, collecting demographics, intra-operative data, and postoperative outcomes for all pancreatic resections performed at our institution. Data were retrieved from the database and retrospectively reviewed. Endovascular procedure Local anesthesia was used in all cases. The left/right common femoral artery was incised as an access route based on previous contrast-enhanced CT. The guidewire was introduced from the access route together with a 6 or 7 Fr catheter (Destination, Terumo, Tokyo, Japan). After arriving at the plane above the celiac artery, a 5 Fr pigtail angiographic catheter (COOK, Bloomington, IN, USA) was exchanged to perform angiography. The bleeding site and anatomical pattern of relevant arteries were identified according to the distribution and leakage of the contrast medium. After the initial angiography, the 5 Fr pigtail angiographic catheter was exchanged to 0.035 (Microport, Shanghai, China) or 0.018 (Boston Scientific, MA, USA) guidewire together with suitable catheter (Cobra; Medikit, Tokyo, Japan) to select the celiac artery and advance to the bleeding artery (common hepatic artery, right hepatic artery, proper hepatic artery, splenic artery, gastroduodenal artery stump, and superior mesenteric artery). The guidewire was advanced to the distal region of the bleeding site to allow sufficient zone for stent-graft deployment, and sufficient proximal/distal landing zone. Self-expandable covered stent-graft Viabahn (TBE, WL Gore, Flagstaff, AZ, USA) was used in all cases, the diameter of which ranged from 5 to 8 mm. Angiography would be performed after the deployment of stent-graft. If persistent contrast extravasation, incomplete exclusion of pseudoaneurysm, or insufficient expansion of the stent-graft was detected, balloon dilatation or extra bridging stent-graft implantation would be performed. Final angiography was performed to reconfirm the complete exclusion of pseudoaneurysm, no leakage of contrast, normal distal organ perfusion, ideal stent-graft shape, and no sign of endoleak. Perioperative routine After the endovascular intervention, all patients were sent to the intensive care unit and monitored with cardiac monitoring for at least 24 h. Complete blood count and liver function test were performed immediately after the operation and daily thereafter. Only after a comprehensive evaluation of the patients’ vital signs, laboratory findings, and volume and color of drainage, the patient would be transferred back to the general ward. If any sign of recurrent bleeding was observed, the contrast-enhanced CT would be immediately performed, and endovascular intervention or relaparotomy would be considered. Coagulation was not considered in concern of rebleeding or high risk of bleeding. After discharge, dual antiplatelet therapy with aspirin (100 mg) and clopidogrel (75 mg) was performed according to the judgment of individual physicians. Definition PPH was recorded according to the ISGPS classification. Based on the severity of hemorrhage, PPH was defined as mild (small or medium volume blood loss; mild clinical volume impairment; therapeutic consequence; no need for reoperation or interventional angiographic embolization) or severe (large volume blood loss; clinically significant impairment; need for blood transfusion; need for invasive treatment). Postoperative pancreatic fistula (POPF) was defined as a drain output of any measurable volume of fluid with an amylase level over 3 times the upper limit of institutional normal serum amylase activity, associated with a clinically relevant development/condition related directly to the POPF. Technical success was defined as angiography showing complete cessation of bleeding, patent stent-graft, and normal end-organ perfusion of the affected aorta. Clinical success was defined as resolution of hemorrhagic symptoms (based on vital signs, laboratory tests, and hemorrhagic drainage) without additional endovascular or surgical intervention. Procedure-related complications were evaluated according to the Claviene-Dindo classification. Survival or any other adverse events were defined as those occurring during hospitalization (in-hospital mortality), within 90 days postoperatively (90-day mortality), and after 90 days postoperatively (follow-up mortality). The reason for death and adverse events were described in detail and recorded as indeterminate if not available. After discharge, patients were followed every 3 months at the outpatient. Contrast-enhanced CT was performed at 3-month intervals for 1 year and every 6 months thereafter. Stent-graft patency and its end-organ perfusion were evaluated with contrast-enhanced CT during follow-up. Primary patency was patent vessel without endovascular reintervention, and secondary patency was patent vessel after endovascular reintervention. Statistical analysis Continuous data are reported as means ± standard deviation (SD) when normally distributed or as median (Q 1 , Q 3 ) when the assumption of normality was not met according to Shapiro-Wilk tests. Categorical data are reported as n . A p < 0.05 was considered statistically significant. Kaplan-Meier estimates were used for stent patency and patients’ survival. The data were analyzed using SPSS version 21 (SPSS Inc, Chicago, IL, USA) and GraphPad Prism (GraphPad Software Inc, La Jolla, CA, USA). Results 3.1 Baseline characteristics There were 12 patients with 9 male and 3 female, aged 63 (44, 80) years who had Viabahn implantation owing to late PPH been involved. Patients’ characteristics are summarized in . The most common underlying disease was pancreatic cancer ( n = 8, 66.7%). Eleven patients underwent open PD, and 1 patient underwent robot-assisted PD. All patients had 3−4 drain tubes placed in the abdomen cavity during the operation. The median onset of hemorrhage was 24.9 (10.0, 50.0) days after surgery. All patients had late, extraluminal, severe PPH, which was Grade C according to the ISGPS grading. All patients underwent contrast-enhanced CT before the procedure, and all had positive findings revealing contrast extravasation or pseudoaneurysm at the bleeding site. 3.2 Procedure outcomes The median duration of the operation was 68.3 (30.0, 120.0) min. A pseudoaneurysm and/or contrast extravasation was identified in all patients. The pancreatic fistula was identified in 6 (50.0%) cases, and pseudoaneurysm was identified in 3 (25.0%) cases, including pancreatic fistula together with pseudoaneurysm in 1 case. The common hepatic artery was the most common bleeding site ( n = 6), after which were splenic artery ( n = 3), gastroduodenal artery (GDA) stump ( n = 2), and superior mesenteric artery ( n = 1). In 3 cases, 2 Viabahns were implanted for complete hemostasis. The diameter of the stent-graft was 5 mm ( n = 1), 6 mm ( n = 4), 7 mm ( n = 4), and 8 mm ( n = 5). In 1 case, post-stent balloon dilatation was required. In another case with pseudoaneurysm in the hepatic artery, the GDA stump was embolized with coils, and the pseudoaneurysm was excluded with Viabahn stent-graft. In 2 cases, Pulsar stent-graft was deployed distal to the Viabahn stent-graft, with several segments overlapping due to lack of Viabahn of appropriate size. Technical success was achieved in all cases, which is, bleeding successfully was treated with covered stent-graft, and distal organ perfusion remained patent . 3.3 Clinical outcomes The median hospital stay was 40.0 (18.0, 92.0) days, and the median ICU stay was 10.8 (1.0, 29.0) days. During hospitalization, 1 patient developed pulmonary embolism after the operation and was considered as contraindicated to coagulation. The patient underwent inferior vena-cava filter implantation to prevent disease progression. One patient had herniation of the small intestine into the thoracic cavity, which caused broad thoracic and abdominal infection. The patient refused further surgical intervention and died of septicemia . The median follow-up was 316.5 (156.3, 682.0) days. During the follow-up, 2 patients died of malignancy recurrence at 10 and 12 months after the operation, whose stent-grafts’ patency was indetermined. Rebleeding occurred in 1 case after stent-graft implantation for the splenic artery. The patient was sent to the hospital because of abdominal pain and hemorrhagic shock 2 months after the initial operation. The newly-occurred bleeding site was GDA stump and another Viabahn was implanted and achieved hemostasis. At the latest image follow-up (27 months after PD), both stent-grafts were patent. Asymptomatic stent-graft occlusion was confirmed with follow-up CT at 26.3 and 24.6 months after the operation, respectively . Baseline characteristics There were 12 patients with 9 male and 3 female, aged 63 (44, 80) years who had Viabahn implantation owing to late PPH been involved. Patients’ characteristics are summarized in . The most common underlying disease was pancreatic cancer ( n = 8, 66.7%). Eleven patients underwent open PD, and 1 patient underwent robot-assisted PD. All patients had 3−4 drain tubes placed in the abdomen cavity during the operation. The median onset of hemorrhage was 24.9 (10.0, 50.0) days after surgery. All patients had late, extraluminal, severe PPH, which was Grade C according to the ISGPS grading. All patients underwent contrast-enhanced CT before the procedure, and all had positive findings revealing contrast extravasation or pseudoaneurysm at the bleeding site. Procedure outcomes The median duration of the operation was 68.3 (30.0, 120.0) min. A pseudoaneurysm and/or contrast extravasation was identified in all patients. The pancreatic fistula was identified in 6 (50.0%) cases, and pseudoaneurysm was identified in 3 (25.0%) cases, including pancreatic fistula together with pseudoaneurysm in 1 case. The common hepatic artery was the most common bleeding site ( n = 6), after which were splenic artery ( n = 3), gastroduodenal artery (GDA) stump ( n = 2), and superior mesenteric artery ( n = 1). In 3 cases, 2 Viabahns were implanted for complete hemostasis. The diameter of the stent-graft was 5 mm ( n = 1), 6 mm ( n = 4), 7 mm ( n = 4), and 8 mm ( n = 5). In 1 case, post-stent balloon dilatation was required. In another case with pseudoaneurysm in the hepatic artery, the GDA stump was embolized with coils, and the pseudoaneurysm was excluded with Viabahn stent-graft. In 2 cases, Pulsar stent-graft was deployed distal to the Viabahn stent-graft, with several segments overlapping due to lack of Viabahn of appropriate size. Technical success was achieved in all cases, which is, bleeding successfully was treated with covered stent-graft, and distal organ perfusion remained patent . Clinical outcomes The median hospital stay was 40.0 (18.0, 92.0) days, and the median ICU stay was 10.8 (1.0, 29.0) days. During hospitalization, 1 patient developed pulmonary embolism after the operation and was considered as contraindicated to coagulation. The patient underwent inferior vena-cava filter implantation to prevent disease progression. One patient had herniation of the small intestine into the thoracic cavity, which caused broad thoracic and abdominal infection. The patient refused further surgical intervention and died of septicemia . The median follow-up was 316.5 (156.3, 682.0) days. During the follow-up, 2 patients died of malignancy recurrence at 10 and 12 months after the operation, whose stent-grafts’ patency was indetermined. Rebleeding occurred in 1 case after stent-graft implantation for the splenic artery. The patient was sent to the hospital because of abdominal pain and hemorrhagic shock 2 months after the initial operation. The newly-occurred bleeding site was GDA stump and another Viabahn was implanted and achieved hemostasis. At the latest image follow-up (27 months after PD), both stent-grafts were patent. Asymptomatic stent-graft occlusion was confirmed with follow-up CT at 26.3 and 24.6 months after the operation, respectively . Discussion PPH is a life-threatening complication of PD. With the improvement in surgical technique and perioperative management, the incidence of PPH has decreased significantly. The incidence of PPH varies in different centers, which could be as low as 3% in high-volume centers, and up to 10% in several reports. , , , Despite timely diagnosis and treatment, the mortality rate of PPH remains high, ranging from 21% to 34% in previous reports. According to the timing of onset, PPH was classified as early and late. The etiology of early and late PPH was different. Early PPH was mostly caused by technical failure of hemostasis during the index operation or coagulopathy. Late PPH is more complicated and multifactorial. During operation, ligation of the arteries, lymphadenectomy, and manipulation of peripancreatic vessels could lead to injury and damage to the protective tissues, rendering the vessel more vulnerable to subsequent damage. POPF was associated with leakage of enzyme rich fluid into the abdomen, causing vessel erosion and formation of pseudoaneurysm. In our series, POPF was identified in 6 late PPH cases, which was the major cause of PPH. Pseudoaneurysm was another main cause of late PPH. The reason for pseudoaneurysm formation was commonly believed to be associated with pancreatic fistula or anastomotic dehiscence, while its pathological process was not fully understood. For 3 cases with pseudoaneurysms, only 1 case had both POPF and splenic artery pseudoaneurysm, where POPF was considered to cause pseudoaneurysm. Continuous vessel erosion by the enzyme rich fluid eventually led to rupture of pseudoaneurysm, causing major hemorrhage. Early stratification of PPH risk was beneficial for patients’ management. Birgin et al. developed and validated a prediction model for PPH. PPH was associated with sentinel bleeding, drain fluid culture positive for Candida species, and radiologic proof of rim enhancement of or gas within a peripancreatic fluid collection. Similarly, Palumbo et al. found postoperative CT evidence of fluid collections, air bubbles, and posterior pancreaticojejunostomy defect associated with PPH. Robust prediction models could identify PPH risk at an early stage, and allow different management strategies for patients with different risks. Management for late PPH included relaparotomy, endoscopic intervention, and endovascular intervention. Immediate relaparotomy used to be the mainstream treatment was now recommended for those with negative findings, insufficient hemostasis in angiography and endoscopy, and hemodynamically unstable patients. Adhesion within the abdominal cavity after the initial PD increased the difficulty in exposure and identification of the bleeding site. Relaparotomy was highly invasive, requiring a longer time for full recovery and higher costs. Survival after relaparotomy was also a concern. A systematic review reported the survival rate after primary relaparotomy as 37%. The indication of endoscopic intervention was limited, which was mostly used in hemodynamically stable patients with intra-luminal hemorrhage. Endovascular intervention was composed of TAE and stent-graft implantation. At the early stage, TAE was the most used approach in our center. However, after embolization of the hepatic artery, liver infarction, liver abscess, and impaired liver function had been observed in early cases. The blood flow was impaired and caused distal organ malperfusion. Hassold et al. compared TAE and covered stent-graft implantation for delayed PPH. In the TAE group, infarction distal to the embolized vessel occurred in 6/11 cases; in the covered stent-graft group, ischemia distal to the occluded stent-graft was observed in 2/14 cases. The 1- and 2-year survival rate was also higher in the covered stent-graft group. After Viabahn stent-graft became available, stent-graft implantation has become the preferred treatment option in our center. The major limitation of stent-graft implantation was its requirement for suitable anatomy. Torturous access artery increases the difficulty in the selection of bleeding arteries, and some bleeding artery was too angulated for stent-graft implantation. For those with unsuitable anatomy, TAE would be considered to achieve hemostasis. A major concern of stent-graft implantation was rebleeding. Maccabe et al. summarized outcomes of PPH stratified according to ISGPS grading. For Grade B PPH, the reintervention rate after initial endovascular treatment was higher than initial surgical intervention (9.7% (3/31) vs . 0 (0/9)). For Grade C PPH, the reintervention rate after initial endovascular treatment was 9.2% (11/119), similar to that of Grade B PPH. In our series, rebleeding occurred in 1 case. The initial bleeding site was a splenic artery, and the rebleeding site was a GDA stump. Another Viabahn was implanted, and hemostasis was achieved after the deployment of the stent-graft. From our experience, Viabahn stent-graft was effective in hemostasis. Endovascular intervention allowed precise adjustment within the artery, and its technical and clinical success rate was high. In cases where rebleeding occurs, another endovascular or surgical intervention could be considered, providing more optimal options compared with initial surgical intervention. , Another concern after stent-graft implantation was its long-term patency. Occlusion or stenosis of stent-graft would impair the distal organ perfusion. In the previous reports, Viabahn stent-graft patency at 1 year varied significantly (40%−80%), while all data were from retrospective studies with small sample sizes. , , Interestingly, Min et al. reported no liver ischemia despite a high stent failure rate (50%). The explanation was that the process of stent-graft occlusion was thrombotic, which was long enough for the formation of collateral blood vessels. Antiplatelet therapy is recommended after implantation of stent-graft with a small diameter to prevent thrombosis, including all that used in this series (5−8 mm). Anticoagulation or antiplatelet therapy was not routinely used during the perioperative period considering the rebleeding risk after PPH, and antiplatelet therapy was recommended after discharge in our center. For those with available follow-up contrast-enhanced CT, stent-grafts (4 cases, 5 stent-grafts) were all patent. Izumi et al. suggested that anticoagulation was not a necessity since there was no association between the administration of anticoagulation or antiplatelet and stent-graft patency. Instead, recurrence of primary malignancy was associated with stent failure, suggesting that physical compression of the tumor was the major reason for thrombotic occlusion. Current evidence for the relation between rebleeding, coagulation, stent-graft patency, and distal organ perfusion was low level, and further randomized study was needed to guide practice. This was a single-center, retrospective study with a small sample size, where selection bias was inevitable. Owing to the short survival period of pancreatic cancer, the follow-up period was limited. Only patients with suitable anatomy would be considered for stent-graft implantation, while the other would be treated with TAE, thus the technical success rate could be over-evaluated. The bleeding site varies, and preoperative evaluation was performed by an experienced physician. A randomized, multicenter study with longer follow-up was needed to further verify the safety and feasibility of Viabahn implantation to treat PPH with different bleeding sites. In conclusion, covered stent-graft implantation with Viabahn is an effective and safe treatment option for PPH with various bleeding sites and causes. Once the covered stent-graft was successfully deployed at the bleeding site, immediate hemostasis would be achieved, and distal organ perfusion was preserved with a low risk of rebleeding. The technical and clinical success rate was high. The major reason for death was a recurrence of primary malignancy during follow-up. Xiaoye Li: Conceptualization, Data curation, Formal analysis, Writing - original draft, Writing - review & editing. Shibo Xia: Data curation, Formal analysis. Liangxi Yuan: Data curation, Formal analysis, Investigation, Methodology, Project administration. Lei Zhang: Data curation, Formal analysis, Methodology, Project administration. Chao Song: Data curation, Formal analysis, Supervision. Xiaolong Wei: Data curation, Project administration. Qingsheng Lu: Conceptualization, Funding acquisition, Methodology, Supervision, Validation, Writing–original draft. Written informed consent was obtained from each patient. This study conformed to the ethical principles of the Declaration of Helsinki and was approved by the institutional review board. This study was supported by the 234 Discipline Climbing Program, A multicenter study of robotic treatment of dilated aortic disease by endovascular repair, 2019YXK048. There is no conflict of interest or funding to disclosure. |
Sociodemographic Characteristics and Digital Behaviors Associated with the Use of Fitness and Diet Apps Among Adolescents | 5da64770-2eed-4aa7-af3e-a3635f89ee97 | 11380736 | Health Literacy[mh] | Smartphones are increasingly used among adolescents not only to keep in contact with classmates and peers but also to browse the Internet and YouTube and access software applications (apps). A recent study on adolescents from Portugal has identified that features such as food and physical activity recommendations are critical for the use of a diet app to prevent overweight and obesity. This study has also found that the utility and interest in a particular health app were the main factors in favor of their continued use. Presently, although numerous health apps focusing on fitness, nutrition, and physical activity are available, many adolescents have never used them. To help adolescents be more physically active and learn about healthy diet and nutrition, the use of technology and online digital platforms could be a valuable tool in efforts to improve adolescents’ health and health-related behaviors. Bearing in mind that positive health-related attitudes and behaviors established in adolescence set the trajectory for healthy lifestyles later in life, the examination of Internet use could offer important evidence about digital health information needs among young people. Accordingly, this study aimed to assess the prevalence of the use of fitness and diet apps and to examine contributing socio-demographic factors and digital behaviors among high school students.
A cross-sectional study was performed from December 2016 to January 2017. Four out of 21 public high schools were randomly selected from the inner urban area of Belgrade, the capital city of Serbia. In all schools, both girls and boys are enrolled together, so classrooms always include students of both genders. High schools have four-year programs and offer general education in the natural sciences, humanities, foreign languages, arts, and physical education. After the completion of high school, the majority of students continue their education at universities. For this study, the online sample size calculator was used, with a margin of error of 5%, a 99% confidence level, the suggested high school population size of 20 000 from the 2011 Census, and a population distribution of 50%. The calculated minimum sample size was 655, which was increased by approximately 10% because potential students refused to participate. All students who were invited also consented to participation, as high school students were quite interested in taking part in this study. Therefore, the response rate was 100%. Ethical approval for the study was obtained from the Institutional Review Board of the Faculty of Medicine, University of Belgrade (approval No. 747-I). As the study participants were minors, the schools informed the parents about the study. The parents were offered the possibility of opting out, implying that the parents could have informed the school in case they did not want their child to participate in the survey. Assent (i.e., consent for participation by participants who were minors) was implied by returning the completed questionnaire. Observed outcome The students were asked to circle the apps that they used at the time of the survey for (1) fitness, running, counting of steps/distance, and the like, as well as (2) healthy diet, nutrition, meal preparation, calorie counting, and the like. The use of fitness and diet apps was categorized as a binary value (yes vs. no) based on whether the specific type of app was circled. A space was considered for students to write the name of an app that they utilized in case they were unsure how to classify it. Covariates The students were asked about their socio-demographic characteristics (gender and age), study year (1 st to 4 th ), type of high school program (science-mathematics vs. humanities-languages), grade point average (GPA), parental education level, and household monthly income. In the Serbian school system, the GPA is based on a numerical scale from 1 to 5 (1 = fail, 2 = pass, 3 = good, 4 = very good, and 5 = excellent). Higher grades correspond to higher academic achievement. The possible range of GPA is 2.0 (poor) – 5.0 (excellent). Students are graded at the end of the winter and the spring semester. However, only grades at the end of the spring semester (which is the end of the academic year) are registered as the final grading score for that school year. Higher grades correspond to higher academic achievement. Students were asked to provide the highest education attainment for both parents. Because there were < 5 observations for the primary education category, primary and secondary education were merged into one. Household monthly income was categorized into < 405, 405–810, and > 810 Euros per month based on the average income in the capital city area. To examine the use of the Internet, the students were asked, “Do you use the Internet?” (yes vs. no). They were also asked to write about the age at which they first began using the Internet and specifically asked whether they browsed webpages about fitness/exercise and nutrition/diet (yes vs. no). In addition to webpages, students were asked whether they used YouTube as a source of health-related information (yes vs. no). To examine the use of smartphones, the students were questioned, “Do you use a smartphone?” (yes vs. no). E-health literacy, defined as the ability to search, comprehend, and assess online health information to make informed health decisions, was quantified using the e-health literacy scale (eHEALS). The scale is composed of 8 questions that examine students’ confidence when searching, assessing, and using health information on the Internet. Each question offers 5 answers (ranks), whereby rank 1 = strongly disagree and rank 5 = strongly agree. The sum of all 8 ranks represents the total eHEALS score (range 8–40). Better e-health literacy is quantified by a higher score. The eHEALS was previously translated and validated in the local language. The internal consistency, as measured by Cronbach’s alpha coefficient, was 0.849. One factorial structure was corroborated using the confirmatory factor analysis (goodness of fit index = 0.983, comparative fit index = 0.964, Tucker Lewis index = 0.949, root mean square error of approximation = 0.090). Finally, the students were asked to quantify the degree of influence of online health information on their health decisions. Their answers were graded on a 4-point scale (1 = not at all, 2 = a little, 3 = quite a bit, and 4 = a lot). Data analysis Mann-Whitney (for continuous variables) and Chi-square (for categorical variables) tests were used to assess differences between variables. Considering that it has been consistently observed that females tend to practice more favorable health behaviors than males, the interaction terms (gender x age) in the two regression models ( , Tables S1 and S2) were tested using the PROCESS macro by Andrew Hayes. The PROCESS macro is a software add-on for SPSS and SAS that can be downloaded from the open-source website http://processmacro.org . The purpose of the program is to facilitate the moderation and mediation analysis of variables in different regression models. Two series of multiple logistic regression models were built to examine factors associated with the use of fitness and diet apps. In the first one, the dependent variable was the ‘use of fitness apps’ (yes vs. no). In the second series, the dependent variable was the ‘use of diet apps’ (yes vs. no). The independent variables were classified according to Model 1 (with demographic variables only), Model 2 (demographic variables + digital consumer behaviors), and Model 3 (demographic variables + digital consumer behaviors + use of apps). All analyses were performed using Statistical Package for Social Sciences (SPSS Inc., Chicago, IL, USA), version 20, and P < 0.05 was considered statistically significant.
The students were asked to circle the apps that they used at the time of the survey for (1) fitness, running, counting of steps/distance, and the like, as well as (2) healthy diet, nutrition, meal preparation, calorie counting, and the like. The use of fitness and diet apps was categorized as a binary value (yes vs. no) based on whether the specific type of app was circled. A space was considered for students to write the name of an app that they utilized in case they were unsure how to classify it.
The students were asked about their socio-demographic characteristics (gender and age), study year (1 st to 4 th ), type of high school program (science-mathematics vs. humanities-languages), grade point average (GPA), parental education level, and household monthly income. In the Serbian school system, the GPA is based on a numerical scale from 1 to 5 (1 = fail, 2 = pass, 3 = good, 4 = very good, and 5 = excellent). Higher grades correspond to higher academic achievement. The possible range of GPA is 2.0 (poor) – 5.0 (excellent). Students are graded at the end of the winter and the spring semester. However, only grades at the end of the spring semester (which is the end of the academic year) are registered as the final grading score for that school year. Higher grades correspond to higher academic achievement. Students were asked to provide the highest education attainment for both parents. Because there were < 5 observations for the primary education category, primary and secondary education were merged into one. Household monthly income was categorized into < 405, 405–810, and > 810 Euros per month based on the average income in the capital city area. To examine the use of the Internet, the students were asked, “Do you use the Internet?” (yes vs. no). They were also asked to write about the age at which they first began using the Internet and specifically asked whether they browsed webpages about fitness/exercise and nutrition/diet (yes vs. no). In addition to webpages, students were asked whether they used YouTube as a source of health-related information (yes vs. no). To examine the use of smartphones, the students were questioned, “Do you use a smartphone?” (yes vs. no). E-health literacy, defined as the ability to search, comprehend, and assess online health information to make informed health decisions, was quantified using the e-health literacy scale (eHEALS). The scale is composed of 8 questions that examine students’ confidence when searching, assessing, and using health information on the Internet. Each question offers 5 answers (ranks), whereby rank 1 = strongly disagree and rank 5 = strongly agree. The sum of all 8 ranks represents the total eHEALS score (range 8–40). Better e-health literacy is quantified by a higher score. The eHEALS was previously translated and validated in the local language. The internal consistency, as measured by Cronbach’s alpha coefficient, was 0.849. One factorial structure was corroborated using the confirmatory factor analysis (goodness of fit index = 0.983, comparative fit index = 0.964, Tucker Lewis index = 0.949, root mean square error of approximation = 0.090). Finally, the students were asked to quantify the degree of influence of online health information on their health decisions. Their answers were graded on a 4-point scale (1 = not at all, 2 = a little, 3 = quite a bit, and 4 = a lot).
Mann-Whitney (for continuous variables) and Chi-square (for categorical variables) tests were used to assess differences between variables. Considering that it has been consistently observed that females tend to practice more favorable health behaviors than males, the interaction terms (gender x age) in the two regression models ( , Tables S1 and S2) were tested using the PROCESS macro by Andrew Hayes. The PROCESS macro is a software add-on for SPSS and SAS that can be downloaded from the open-source website http://processmacro.org . The purpose of the program is to facilitate the moderation and mediation analysis of variables in different regression models. Two series of multiple logistic regression models were built to examine factors associated with the use of fitness and diet apps. In the first one, the dependent variable was the ‘use of fitness apps’ (yes vs. no). In the second series, the dependent variable was the ‘use of diet apps’ (yes vs. no). The independent variables were classified according to Model 1 (with demographic variables only), Model 2 (demographic variables + digital consumer behaviors), and Model 3 (demographic variables + digital consumer behaviors + use of apps). All analyses were performed using Statistical Package for Social Sciences (SPSS Inc., Chicago, IL, USA), version 20, and P < 0.05 was considered statistically significant.
A total of 702 high school students filled out the questionnaire. All students used the Internet. Of 702 students, 670 (95.4%) utilized smartphones and were able to access fitness and diet apps from their mobile phones, so only their data underwent analysis. Socio-demographic characteristics and digital behaviors of students who employed smartphones are presented in . There were significantly more students in the science-mathematics program compared to those in the humanities-languages program, and girls had a higher GPA than boys. Boys started using the Internet at an earlier age than girls. Girls used fitness and diet apps more than boys. Girls browsed fitness and diet webpages more than boys. No difference was observed in the use of YouTube between girls and boys . Use of fitness apps A total of 117 (117/394, 29.7%) girls and 51 (51/276, 17.3%) boys who had smartphones utilized fitness apps (χ 2 = 10.870, P = 0.001) listed in the questionnaire. Overall, only three students used fitness apps that were labeled as “other” (i.e., not listed in the questionnaire). The multiplicative interaction of gender and age was tested to examine the role of gender in fitness app use. The product term of age and gender in the logistic regression model was deemed significant (Table S1). Accordingly, the study sample was stratified by gender. A total of three logistic regression models were tested in each stratum. The series of three regression models among the girls is presented in . Model 1 showed that girls whose families had higher monthly incomes were more likely to use fitness apps. Model 2 demonstrated that, in addition to higher monthly income, having better e-health literacy, browsing websites about fitness and diet, and a lack of use of YouTube were associated with the use of fitness apps. Model 3 indicated that, in addition to previous variables, the use of diet apps was associated with the use of fitness apps among girls. The series of three regression models among the boys is provided in . Model 1 revealed that younger boys were more likely to use fitness apps. Model 2 represented that, in addition to younger age, browsing fitness websites and YouTube was associated with the use of fitness apps among boys. The results did not materially change after the adjustment for the use of diet apps (Model 3). Use of diet apps A total of 38 (38/394, 9.6%) girls and 10 (10/276, 3.6%) boys who used smartphones also employed diet apps (χ 2 = 8.848, P = 0.003). The interaction terms were tested to examine the role of gender as a moderator in the association between students’ age and the use of diet apps. The product term between age and gender was deemed significant (Table S2). For this reason, the study sample was stratified by gender. Because of the small proportion of boys who reported the use of diet apps (3.6%), the logistic regression models were not deemed sufficiently robust. Thus, factors associated with the use of diet apps were investigated only among the girls. lists the series of three regression models among the girls. Model 1 demonstrated that younger girls were more likely to use diet apps. Model 2 showed that being younger, having a higher family income, and browsing websites about diet and fitness were associated with the use of diet apps. After the adjustment for the use of fitness apps, Model 3 indicated that girls who were younger, browsed websites about diet, and used fitness apps were more likely to utilize diet apps.
A total of 117 (117/394, 29.7%) girls and 51 (51/276, 17.3%) boys who had smartphones utilized fitness apps (χ 2 = 10.870, P = 0.001) listed in the questionnaire. Overall, only three students used fitness apps that were labeled as “other” (i.e., not listed in the questionnaire). The multiplicative interaction of gender and age was tested to examine the role of gender in fitness app use. The product term of age and gender in the logistic regression model was deemed significant (Table S1). Accordingly, the study sample was stratified by gender. A total of three logistic regression models were tested in each stratum. The series of three regression models among the girls is presented in . Model 1 showed that girls whose families had higher monthly incomes were more likely to use fitness apps. Model 2 demonstrated that, in addition to higher monthly income, having better e-health literacy, browsing websites about fitness and diet, and a lack of use of YouTube were associated with the use of fitness apps. Model 3 indicated that, in addition to previous variables, the use of diet apps was associated with the use of fitness apps among girls. The series of three regression models among the boys is provided in . Model 1 revealed that younger boys were more likely to use fitness apps. Model 2 represented that, in addition to younger age, browsing fitness websites and YouTube was associated with the use of fitness apps among boys. The results did not materially change after the adjustment for the use of diet apps (Model 3).
A total of 38 (38/394, 9.6%) girls and 10 (10/276, 3.6%) boys who used smartphones also employed diet apps (χ 2 = 8.848, P = 0.003). The interaction terms were tested to examine the role of gender as a moderator in the association between students’ age and the use of diet apps. The product term between age and gender was deemed significant (Table S2). For this reason, the study sample was stratified by gender. Because of the small proportion of boys who reported the use of diet apps (3.6%), the logistic regression models were not deemed sufficiently robust. Thus, factors associated with the use of diet apps were investigated only among the girls. lists the series of three regression models among the girls. Model 1 demonstrated that younger girls were more likely to use diet apps. Model 2 showed that being younger, having a higher family income, and browsing websites about diet and fitness were associated with the use of diet apps. After the adjustment for the use of fitness apps, Model 3 indicated that girls who were younger, browsed websites about diet, and used fitness apps were more likely to utilize diet apps.
The results of this study revealed that high school girls used fitness apps and diet apps more than high school boys. Similar results were observed among adolescents in Flanders (Belgium), where fitness apps garner more interest than diet apps, as well as in the United Kingdom. However, in Vietnam, few teenagers use this type of app. This difference is expected, given that access to smartphones and, subsequently, health apps might be lower in resource-limited settings. The stratification of the study sample according to gender suggested that browsing fitness websites was associated with the use of fitness apps among high school girls and boys. The finding that high school girls used fitness and diet apps more frequently compared with high school boys is in line with a well-acknowledged difference between genders relative to self-care and health-promoting behavior. , Females tend to practice more favorable health behaviors and are more interested in health compared to males. Women are also more inclined to seek advice or help. Thus, the results of our study support the previous evidence , and indicate that the gender gap in digital health-information-seeking begins as early as adolescence. A previous study reported that the use of health apps helps set health goals and enable self-monitoring and self-perception. For an app to be used by adolescents, several researchers agree that the content needs to be easy to navigate, provide understandable information, and offer a personalized regimen, as well as the possibility to connect with peers and other people through social networks. , , Although health apps have the potential for health promotion among youth, not all health apps are equally effective in helping the users to achieve higher levels of fitness, desired body image/composition or ensure long-term health benefits. For example, one randomized controlled trial reported that, after a 2-month follow-up, the cardio-respiratory fitness of adolescents who used a specific fitness app did not differ from those who did not utilize such an app. Use of fitness and diet apps has also been observed to have a negative impact on some young adults because adjustment to the scheduled app program led them to feel socially isolated (due to a certain eating and fitness regimen), controlled by the app, or afraid of not being able to achieve certain targets. Younger girls in our study were more likely to use diet apps. This finding could be explained by the notion that girls become aware of their body image at a younger age and strive to be thin compared to boys. Overall, it is estimated that one-half of adolescent girls are not satisfied with their weight and are more likely to diet and restrict caloric intake. Parental attitudes and behaviors toward food and dieting are closely related to those of their adolescent children. Furthermore, parental feedback about fitness and diet is the strongest contributor to body dissatisfaction among adolescents. Considering the risk of becoming underweight or developing eating disorders, the nutrition and diet of adolescent girls need to be closely monitored by adults (parents and school teachers). In fact, it is essential that parents be healthy role models for their adolescent children. Therefore, the parental role must be continuously emphasized in the effort to set the trajectory for adolescents’ healthy eating habits. In this study, it was found that girls coming from higher-income families were more likely to use fitness and diet apps. This was not observed in boys. Evidence suggests that there is an inverse association between adolescent girls’ weight and socio-economic status, with girls of higher socio-economic status more likely to be in the healthy weight range. A longitudinal study in the Czech Republic reported that adolescents from higher socio-economic groups were more likely to eat fruits and vegetables every day compared to those from lower socio-economic groups. Although socio-economic inequalities have a strong influence on adolescents’ food choices, the importance of in-person education about proper nutrition should be prioritized in schools using formal and informal teaching methods. Our findings revealed that the use of fitness and diet websites was consistently associated with the use of fitness and diet apps, respectively. This was identified among both girls and boys. Students in our sample began using the Internet at the age of 9 (boys) and 10 (girls), suggesting that adolescents in our study practically grew up using the Internet for various purposes. As a result, a high level of digital literacy is expected. For this reason, the association of better e-health literacy with the use of fitness apps among girls was not surprising. Given that adolescents’ use of the Internet in general preceded owning and using a personal smartphone, the use of various digital platforms, such as in our study (use of web pages, YouTube, and smartphone app) can be considered a ‘digital package deal’. Use of YouTube was associated with the use of fitness and diet apps among adolescent boys, but not among girls. Over the past years, the number of YouTube users has been growing. This digital platform offers unique audio-visual content that adolescents can relate to, and it has become a popular medium for sharing various health information, including data about fitness and diet. While YouTube content may be visually appealing for adolescent viewers, a recent study on YouTube videos aimed at achieving fitness reported that the content of fitness creators on YouTube was not health-promoting. Because of the visually appealing content, often documenting fitness transformations and ‘how to’ recommendations, it is reasonable to expect that the use of fitness and diet apps on YouTube is being endorsed. Considering the simultaneous use of various digital media, the importance of a balance between on-screen and off-screen time, parental involvement, and the improvement of digital literacy has been widely emphasized due to their relevance for the promotion of adolescents’ health. The limitations of this study are related to the fact that high school students living in areas less urbanized than the capital city and rural areas have less access to the Internet, given that 63.8% of households have Internet access. We targeted high schools in the inner urban area, whereas high schools in municipalities on the outskirts of the metropolitan area were not included in our study. High school students in this region may have different behaviors and practices for smartphone and app use. In a similar vein, students from rural areas might have less access to smartphones due to costs. All fitness and nutrition-related apps were observed jointly. We have not stratified apps according to specific targets, such as weight loss/calorie counting, consistency of a certain diet, or exercise regimen. Nevertheless, because of the relatively low prevalence of diet app use, particularly among boys, further stratification according to types of fitness and diet apps would not allow a regression model to converge. We are aware that this response rate may be open to social acceptability bias, implying that students who were hesitant to participate did so because their classmates opted to participate. Due to the cross-sectional study design, we could not make definite causal inferences between the examined variables and the outcomes of interest. Finally, because of the fast pace of innovations in the digital landscape, the data presented in this study may not be entirely applicable to all school settings. However, despite the efforts to keep up, digital approaches to learning in schools in Serbia still face significant challenges. For this reason, our results could remain relevant in the forthcoming years. Highlights The prevalence of app use was higher among high school girls than high school boys. Fitness apps were used more than diet apps. The use of YouTube was associated with fitness apps among boys. The use of Internet websites was related to fitness and diet apps.
The prevalence of app use was higher among high school girls than high school boys. Fitness apps were used more than diet apps. The use of YouTube was associated with fitness apps among boys. The use of Internet websites was related to fitness and diet apps.
In general, fitness and diet apps are being used among high school students. The prevalence of app use was higher among high school girls compared to high school boys. The use of fitness and diet websites was consistently associated with the use of fitness and diet apps across genders. Additionally, the use of YouTube was associated with the use of health apps among boys. Given that younger girls are more likely to be the users of diet apps, parents and other adults are advised to monitor girls’ diet regimens to prevent unhealthy dieting. Apps could be a useful tool to improve physical activity and nutrition. It is recommended that certain classes at school be devoted to the use of digital health resources. Practical demonstrations of the use of certain health apps could be an additional opportunity to support positive health behaviors among adolescents.
The authors are grateful to all high school students who participated in this study.
Conceptualization: Tatjana Gazibara, Milica Cakic, Jelena Cakic, Anita Grgurevic, Tatjana Pekmezovic. Data curation: Tatjana Gazibara, Milica Cakic, Jelena Cakic. Formal analysis: Tatjana Gazibara. Funding acquisition: Tatjana Pekmezovic. Investigation: Tatjana Gazibara, Milica Cakic, Jelena Cakic, Anita Grgurevic, Tatjana Pekmezovic. Methodology: Tatjana Gazibara, Milica Cakic, Jelena Cakic, Anita Grgurevic, Tatjana Pekmezovic. Project administration: Tatjana Gazibara. Resources: Tatjana Pekmezovic. Supervision: Anita Grgurevic. Validation: Tatjana Gazibara, Milica Cakic, Jelena Cakic, Anita Grgurevic, Tatjana Pekmezovic. Visualization: Tatjana Gazibara. Writing–original draft: Tatjana Gazibara. Writing–review & editing: Milica Cakic, Jelena Cakic, Anita Grgurevic, Tatjana Pekmezovic.
The authors declare that they have no conflict of interests.
The Institutional Review Board of the Faculty of Medicine at the University of Belgrade approved this study (approval No. 747/I).
The study was supported by the Ministry of Education, Sciences, and Technological Development of the Republic of Serbia (grant No. 200100).
Supplementary file 1 contains Tables S1 and S2.
|
Development and evaluation of a health literacy scale for parasitic diseases | 8a5943c8-0b39-40a3-9c14-a94739307fd1 | 11395214 | Health Literacy[mh] | Parasitic diseases are caused by parasites that invade humans and the pathology and symptoms of individual parasitic diseases vary depending on the species involved . Seven of the world’s top-ten tropical diseases are parasitic diseases included Dracunculiasis, Echinococcosis, Clonorchiasis, Fascioliasis, Paragonimiasis, Human African trypanosomiasis, and Leishmaniasis, which are widespread, diverse, and serious infectious diseases that require priority control as designated by the WHO . From 2014 to 2015, a national parasitic diseases survey was carried out in China . The survey data covered three major features in the spread of parasitic diseases in China. Firstly, the rate of key parasitic-disease infections has declined markedly in comparison to the rates recorded in last national survey conducted in 2001–2004. Helminth and protozoan decreased from 21.38 to 3.41%, 19.34–3.38%, respectively. Secondly, it exhibits unique characteristics in terms of the spatial spread of parasitic diseases. Tapeworm infections generally declined from west to east. The endemic areas of clonorchis mainly distributed in South China and Northeast China. Thirdly, the number of key parasitic-disease infections was large, as high as 38.59 million. The rate of key parasitic-disease infections was 5.96%. There is a big gap compared with developed countries. Study indicated that poverty and parasitic disease are closely linked. For example, China has a high prevalence of echinococcosis, with 0.38 million cases reported, making up 40% of the worldwide burden of cystic echinococcosis DALYs and over 90% of global alveolar echinococcosis DALYs . Opportunistic parasitic diseases have increased due to the rise in diseases involving the immune system. Imported parasitic diseases, such as malaria and schistosomiasis have increased , as well as food-borne parasitic diseases associated with higher incomes and living standards, and the rise in the consumption of exotic foods and delicacies . The number and proportion of imported malaria cases increased from 18.26% (7310/42 319) in 2005 to 99.88% (844/845) in 2022 . From 1979 to 2017, a total of 384 imported cases of schistosomiasis were reported in China . Hence, parasitic diseases remain a public health problem in China. It was found that knowledge activated a belief system, causing emotions, which in turn led to an intention to engage in a specific behavior . Modifying people’s behaviors may disrupt the parasite’s life cycle, thereby reducing the risk of transmission . Parasitic diseases are associated with people’s knowledge of health and their behaviours. Therefore, education designed to disseminate health-related knowledge and promote healthy behaviours play an important role in the prevention and control of parasitic diseases. The risk of infection and super-infection with parasitic diseases can be reduced substantially by implementing health education . The National Academy of Medicine, a non-governmental organisation, defines health literacy as the degree to which individuals have the capacity to obtain, process, and understand basic health information and the services needed to make appropriate health-related decisions . Sorensen and Okan revisited the concept of health literacy, emphasizing the ability to access, understand, evaluate, and apply health-related information to decision-making . Study suggests that health literacy is a stronger predictor of health than age, income, employment, or education are, and that it is considered a cost-effective health-promotion intervention . Many assessment scales targeting special populations and diseases have emerged since the concept of health literacy was first introduced at an international conference in 1974. Examples include the Parental Health Literacy Questionnaire for Caregivers , the Multiple Sclerosis Health Literacy Questionnaire , the Chinese Health Literacy Scale for Diabetes , and for lesbian, gay and bisexual (LBG) indviduals, the LGB-Specific Health Literacy Scale .In 2008, China’s National Health Commission published a survey with 66 items on health literacy (i.e. the Chinese Resident Health Literacy—Basic Knowledge and Skills (Trial)) , but only one item pertaining to schistosomiasis was included. Similarly, parasitic diseases were not covered in the Chinese Health Literacy Scale, which included 56 questions. The European Centre for Disease Prevention and Control emphasises the importance of health literacy for infectious diseases . Study suggests that limited health literacy is associated with the adoption of less protective behaviour, such as poor understanding of antibiotics leads to misuse of antibiotics. Efforts to promote health literacy have been made for influenza, MMR immunizations, viral hepatitis, and other infections , but investigations on parasite diseases are lacking. Given the importance of health literacy for the prevention and control of parasitic diseases, it is necessary to develop a reliable tool to measure individuals’ health literacy of parasitic diseases. Therefore, based on the definition of health literacy, we try to define the concept of parasitic diseases health literacy. It is a comprehensive conception refer that the individual’s ability to access, understand, evaluate information about parasitic diseases, make appropriate health-related decisions and adopt healthy behaviors and prevent parasitic diseases. The present study focused on the concept of ‘parasitic disease health literacy’ and used the concept to develop a tool to measure the parasitic disease health literacy of residents with the goal of generating ideas for implementing targeted interventions to promote the health of China’s residents.
The study was conducted in three stages from 2021 to 2022 using qualitative and quantitative methods (Fig. ). Data were collected using the free online Sojump survey template. The phases of each stage are as follows. The Ethics Committee of the JIPD approved the study (JIPD-2022-009). Scale development Phase 1: Indicator construction The health literacy indicators for the parasitic diseases in this study are based on the conceptual framework developed by the National Academy of Medicine (NAM). Three levels of indicators were generated. First, the first-level indicators were constructed in accordance with the definition of health literacy and the Chinese residents’ framework of health literacy, including basic knowledge and awareness, capacity for healthy behaviour, and health-related skills. The second-level indicators were formulated using the Health Literacy Evaluation Index System for Infectious Diseases, which included sources of infection, transmission, and the prevention of infectious diseases . The third-level indicators were based on the results of the National Parasitic Diseases Survey . The data from that the national survey revealed that helminths, nematodes and food-borne parasitic diseases were mainly the content of the third-level indicators. Second, the original indicators were developed through discussions among professionals from the Jiangsu Institute of Parasitic Diseases and the Centres for Disease Control and Prevention (CDC), who were charged with the prevention and control of parasitic diseases in the city and county through focus-group discussions. Third, we selected 14 experts in parasitic disease prevention and control, clinical diagnosis and treatment of parasitic diseases, health literacy monitoring, and public health education to complete a two-round Delphi consultation to confirm the judgements of the importance, and familiarity of the indicators. These three activities resulted in the identification of three first-level indicators, 12 s-level indicators, and 48 third-level indicators by consensus. Phase 2: Questionnaire development The questionnaire was based on the 48 third-level indicators. One indicator was revised to serve as a question on the scale after a discussion among the research group. As a result, the 48-question Parasitic Disease Health Literacy Questionnaire (PDHLQ) was developed, which addressed three factors: information processing of assessments, appraisals, and applications. The weight of each indicator was developed using a 5-point Likert scale, which was appraised during the last round of the Delphi consultation using the Analytic Hierarchy Process . The total score was converted to a percentage grade, with a perfect score of 100. The scale’s passing score was determined using the receiver operating characteristic curve (ROC) . The ROC is a curve with sensitivity as the ordinate and 1-specificity as the abscissa. Therefore, each question had a different score, based on its weight. The original version of the 48-question PDHLQ was reviewed by three staff members who worked at the Jiangsu Institute of Parasitic Diseases (JIPD), to determine whether the questions were consistent with the indicators. 15 questionnaires were distributed to the JIPD cleaning and security staff, to enhance the items’ clarity and comprehension. The following topics were covered: (a) whether the wording was appropriate and easily understood; (b) which items they had difficulty responding to and why; and (c) suggestions for items they believed were not clear. Although the cleaning and security staff had no problems responding to the items, some modifications were suggested to ensure the clarity and simplicity of the items and answers. Evaluation of the questionnaire Phase 1: Participants and data collection Anthoine et al. suggested that (1) a sample size between 2 and 20 subjects is appropriate for each question; (2) a total sample of 500 participants is an adequate number; and (3) 1,000 or more subjects is an excellent number of participants . This study used a methodological design with multistage sampling and a household survey. A cross-sectional survey was conducted in six districts of the prefecture of Jiangsu.First, Jiangsu Province was divided into three regions: northern Jiangsu (Xuzhou Lianyungang, Huai’an, Suqian–4 cities), central Jiangsu (Nantong, Yancheng, Yangzhou, Taizhou–4 cities), and southern Jiangsu (Nanjing, Wuxi, Changzhou, Suzhou, Zhenjiang–5 cities), according to their geographical orientations, cultural traditions, and social and economic development . Therefore, one city was chosen from each of the three regions, and three counties were randomly selected from the three areas. Second, two sub-districts (county-level city or district) were randomly selected from each county, yielding a total of six sub-districts. Third, three townships were randomly selected from each of the six sub-districts in step two, yielding a total of 18 townships. Fourth, 100 residents age 14–69 years in the participating townships were randomly selected based on their location. Only 55 residents in each of the 18 townships were asked to complete the questionnaire. Fifth, a random sample of household members was selected using Kish Table . The survey was administrated by 12-trained investigators, who worked in the township health centres. To reduce bias, investigators received specialist training and were assisted by staff from the local CDC, who were familiar with parasitic diseases. A telephone appointment before the household survey was implemented for quality-assurance and for efficient data collection. The survey was anonymous and confidential. Participants completed the questionnaire with the help of an investigator during a face-to-face interview. These questionnaires were returned to the staff from the local CDC after the survey was completed. Data on demographics were collected from the participants, including their age, gender, educational level, and family income. Phase 2: Pilot study A pilot test was conducted in the northern, central, and southern regions of Jiangsu Province at 2021. Questionnaires were sent to residents, age 14–69 years, and 990 valid questionnaires were returned. Participants were sampled from the multi-stage sample in Phase 1. Statistical analysis A semi-structured interview guide was used to collect information through focus group discussions. Colaizzi’s seven-step method was used for analyse the interview data. And previous research had explained the qualitative analysis approach . Internal consistency was measured using Cronbach’s α, and split-half reliability was measured using the Spearman-Brown coefficient between the odd and even numbered questions. Cronbach’s α coefficient ranges from 0 to 1. A larger Cronbach’s α coefficient indicates better internal consistency, and a value > 0.7 is considered to be good. A split-half reliability coefficient > 0.7 is considered an acceptable level of reliability . Exploratory factor analysis (EFA) was used to re-evaluate and filter the items. The Bartlett test of sphericity was performed for all items and the Kaiser–Meyer–Olkin (KMO) Index was calculated. The Bartlett test of sphericity was significant ( P < 0.05) and the KMO scores were > 0.7, which were considered appropriate for factor analysis . Principal component analyses (PCA) with Varimax rotation were performed. Items are generally retained unless their factor loading is > 0.4, the commonality is > 0.2, the eigenvalue > 1, and deviation of the factor loading is < 0.2 between different factors . Based on the results of the EFA of the pilot study, confirmatory factor analysis (CFA) was used to verify the construct validity of the questionnaire . The model fit was considered acceptable when the following criteria were met: 2/df lower than 3.00, a root mean square error of approximation (RMSEA) lower than 0.05, a goodness of fit index (GFI) greater than 0.85, and a comparative fit index (CFI) greater than 0.90 . The content validity index (CVI) of the questionnaire was confirmed after a two-round Delphi consultation. Descriptive statistics for participants’ characteristics were tabulated. The relationships between scores and demographic characteristics were examined using one-way analysis of variance (ANOVA). Parametric tests, including the EFA, CFA, internal consistency, and split-half reliability, stepwise regression analysis, were analyzed using SPSS Statistics for Windows, Version 25.0. Armonk, NY: IBM Corp. The significance level was set at P < 0.01.
Phase 1: Indicator construction The health literacy indicators for the parasitic diseases in this study are based on the conceptual framework developed by the National Academy of Medicine (NAM). Three levels of indicators were generated. First, the first-level indicators were constructed in accordance with the definition of health literacy and the Chinese residents’ framework of health literacy, including basic knowledge and awareness, capacity for healthy behaviour, and health-related skills. The second-level indicators were formulated using the Health Literacy Evaluation Index System for Infectious Diseases, which included sources of infection, transmission, and the prevention of infectious diseases . The third-level indicators were based on the results of the National Parasitic Diseases Survey . The data from that the national survey revealed that helminths, nematodes and food-borne parasitic diseases were mainly the content of the third-level indicators. Second, the original indicators were developed through discussions among professionals from the Jiangsu Institute of Parasitic Diseases and the Centres for Disease Control and Prevention (CDC), who were charged with the prevention and control of parasitic diseases in the city and county through focus-group discussions. Third, we selected 14 experts in parasitic disease prevention and control, clinical diagnosis and treatment of parasitic diseases, health literacy monitoring, and public health education to complete a two-round Delphi consultation to confirm the judgements of the importance, and familiarity of the indicators. These three activities resulted in the identification of three first-level indicators, 12 s-level indicators, and 48 third-level indicators by consensus. Phase 2: Questionnaire development The questionnaire was based on the 48 third-level indicators. One indicator was revised to serve as a question on the scale after a discussion among the research group. As a result, the 48-question Parasitic Disease Health Literacy Questionnaire (PDHLQ) was developed, which addressed three factors: information processing of assessments, appraisals, and applications. The weight of each indicator was developed using a 5-point Likert scale, which was appraised during the last round of the Delphi consultation using the Analytic Hierarchy Process . The total score was converted to a percentage grade, with a perfect score of 100. The scale’s passing score was determined using the receiver operating characteristic curve (ROC) . The ROC is a curve with sensitivity as the ordinate and 1-specificity as the abscissa. Therefore, each question had a different score, based on its weight. The original version of the 48-question PDHLQ was reviewed by three staff members who worked at the Jiangsu Institute of Parasitic Diseases (JIPD), to determine whether the questions were consistent with the indicators. 15 questionnaires were distributed to the JIPD cleaning and security staff, to enhance the items’ clarity and comprehension. The following topics were covered: (a) whether the wording was appropriate and easily understood; (b) which items they had difficulty responding to and why; and (c) suggestions for items they believed were not clear. Although the cleaning and security staff had no problems responding to the items, some modifications were suggested to ensure the clarity and simplicity of the items and answers.
The health literacy indicators for the parasitic diseases in this study are based on the conceptual framework developed by the National Academy of Medicine (NAM). Three levels of indicators were generated. First, the first-level indicators were constructed in accordance with the definition of health literacy and the Chinese residents’ framework of health literacy, including basic knowledge and awareness, capacity for healthy behaviour, and health-related skills. The second-level indicators were formulated using the Health Literacy Evaluation Index System for Infectious Diseases, which included sources of infection, transmission, and the prevention of infectious diseases . The third-level indicators were based on the results of the National Parasitic Diseases Survey . The data from that the national survey revealed that helminths, nematodes and food-borne parasitic diseases were mainly the content of the third-level indicators. Second, the original indicators were developed through discussions among professionals from the Jiangsu Institute of Parasitic Diseases and the Centres for Disease Control and Prevention (CDC), who were charged with the prevention and control of parasitic diseases in the city and county through focus-group discussions. Third, we selected 14 experts in parasitic disease prevention and control, clinical diagnosis and treatment of parasitic diseases, health literacy monitoring, and public health education to complete a two-round Delphi consultation to confirm the judgements of the importance, and familiarity of the indicators. These three activities resulted in the identification of three first-level indicators, 12 s-level indicators, and 48 third-level indicators by consensus.
The questionnaire was based on the 48 third-level indicators. One indicator was revised to serve as a question on the scale after a discussion among the research group. As a result, the 48-question Parasitic Disease Health Literacy Questionnaire (PDHLQ) was developed, which addressed three factors: information processing of assessments, appraisals, and applications. The weight of each indicator was developed using a 5-point Likert scale, which was appraised during the last round of the Delphi consultation using the Analytic Hierarchy Process . The total score was converted to a percentage grade, with a perfect score of 100. The scale’s passing score was determined using the receiver operating characteristic curve (ROC) . The ROC is a curve with sensitivity as the ordinate and 1-specificity as the abscissa. Therefore, each question had a different score, based on its weight. The original version of the 48-question PDHLQ was reviewed by three staff members who worked at the Jiangsu Institute of Parasitic Diseases (JIPD), to determine whether the questions were consistent with the indicators. 15 questionnaires were distributed to the JIPD cleaning and security staff, to enhance the items’ clarity and comprehension. The following topics were covered: (a) whether the wording was appropriate and easily understood; (b) which items they had difficulty responding to and why; and (c) suggestions for items they believed were not clear. Although the cleaning and security staff had no problems responding to the items, some modifications were suggested to ensure the clarity and simplicity of the items and answers.
Phase 1: Participants and data collection Anthoine et al. suggested that (1) a sample size between 2 and 20 subjects is appropriate for each question; (2) a total sample of 500 participants is an adequate number; and (3) 1,000 or more subjects is an excellent number of participants . This study used a methodological design with multistage sampling and a household survey. A cross-sectional survey was conducted in six districts of the prefecture of Jiangsu.First, Jiangsu Province was divided into three regions: northern Jiangsu (Xuzhou Lianyungang, Huai’an, Suqian–4 cities), central Jiangsu (Nantong, Yancheng, Yangzhou, Taizhou–4 cities), and southern Jiangsu (Nanjing, Wuxi, Changzhou, Suzhou, Zhenjiang–5 cities), according to their geographical orientations, cultural traditions, and social and economic development . Therefore, one city was chosen from each of the three regions, and three counties were randomly selected from the three areas. Second, two sub-districts (county-level city or district) were randomly selected from each county, yielding a total of six sub-districts. Third, three townships were randomly selected from each of the six sub-districts in step two, yielding a total of 18 townships. Fourth, 100 residents age 14–69 years in the participating townships were randomly selected based on their location. Only 55 residents in each of the 18 townships were asked to complete the questionnaire. Fifth, a random sample of household members was selected using Kish Table . The survey was administrated by 12-trained investigators, who worked in the township health centres. To reduce bias, investigators received specialist training and were assisted by staff from the local CDC, who were familiar with parasitic diseases. A telephone appointment before the household survey was implemented for quality-assurance and for efficient data collection. The survey was anonymous and confidential. Participants completed the questionnaire with the help of an investigator during a face-to-face interview. These questionnaires were returned to the staff from the local CDC after the survey was completed. Data on demographics were collected from the participants, including their age, gender, educational level, and family income. Phase 2: Pilot study A pilot test was conducted in the northern, central, and southern regions of Jiangsu Province at 2021. Questionnaires were sent to residents, age 14–69 years, and 990 valid questionnaires were returned. Participants were sampled from the multi-stage sample in Phase 1.
Anthoine et al. suggested that (1) a sample size between 2 and 20 subjects is appropriate for each question; (2) a total sample of 500 participants is an adequate number; and (3) 1,000 or more subjects is an excellent number of participants . This study used a methodological design with multistage sampling and a household survey. A cross-sectional survey was conducted in six districts of the prefecture of Jiangsu.First, Jiangsu Province was divided into three regions: northern Jiangsu (Xuzhou Lianyungang, Huai’an, Suqian–4 cities), central Jiangsu (Nantong, Yancheng, Yangzhou, Taizhou–4 cities), and southern Jiangsu (Nanjing, Wuxi, Changzhou, Suzhou, Zhenjiang–5 cities), according to their geographical orientations, cultural traditions, and social and economic development . Therefore, one city was chosen from each of the three regions, and three counties were randomly selected from the three areas. Second, two sub-districts (county-level city or district) were randomly selected from each county, yielding a total of six sub-districts. Third, three townships were randomly selected from each of the six sub-districts in step two, yielding a total of 18 townships. Fourth, 100 residents age 14–69 years in the participating townships were randomly selected based on their location. Only 55 residents in each of the 18 townships were asked to complete the questionnaire. Fifth, a random sample of household members was selected using Kish Table . The survey was administrated by 12-trained investigators, who worked in the township health centres. To reduce bias, investigators received specialist training and were assisted by staff from the local CDC, who were familiar with parasitic diseases. A telephone appointment before the household survey was implemented for quality-assurance and for efficient data collection. The survey was anonymous and confidential. Participants completed the questionnaire with the help of an investigator during a face-to-face interview. These questionnaires were returned to the staff from the local CDC after the survey was completed. Data on demographics were collected from the participants, including their age, gender, educational level, and family income.
A pilot test was conducted in the northern, central, and southern regions of Jiangsu Province at 2021. Questionnaires were sent to residents, age 14–69 years, and 990 valid questionnaires were returned. Participants were sampled from the multi-stage sample in Phase 1.
A semi-structured interview guide was used to collect information through focus group discussions. Colaizzi’s seven-step method was used for analyse the interview data. And previous research had explained the qualitative analysis approach . Internal consistency was measured using Cronbach’s α, and split-half reliability was measured using the Spearman-Brown coefficient between the odd and even numbered questions. Cronbach’s α coefficient ranges from 0 to 1. A larger Cronbach’s α coefficient indicates better internal consistency, and a value > 0.7 is considered to be good. A split-half reliability coefficient > 0.7 is considered an acceptable level of reliability . Exploratory factor analysis (EFA) was used to re-evaluate and filter the items. The Bartlett test of sphericity was performed for all items and the Kaiser–Meyer–Olkin (KMO) Index was calculated. The Bartlett test of sphericity was significant ( P < 0.05) and the KMO scores were > 0.7, which were considered appropriate for factor analysis . Principal component analyses (PCA) with Varimax rotation were performed. Items are generally retained unless their factor loading is > 0.4, the commonality is > 0.2, the eigenvalue > 1, and deviation of the factor loading is < 0.2 between different factors . Based on the results of the EFA of the pilot study, confirmatory factor analysis (CFA) was used to verify the construct validity of the questionnaire . The model fit was considered acceptable when the following criteria were met: 2/df lower than 3.00, a root mean square error of approximation (RMSEA) lower than 0.05, a goodness of fit index (GFI) greater than 0.85, and a comparative fit index (CFI) greater than 0.90 . The content validity index (CVI) of the questionnaire was confirmed after a two-round Delphi consultation. Descriptive statistics for participants’ characteristics were tabulated. The relationships between scores and demographic characteristics were examined using one-way analysis of variance (ANOVA). Parametric tests, including the EFA, CFA, internal consistency, and split-half reliability, stepwise regression analysis, were analyzed using SPSS Statistics for Windows, Version 25.0. Armonk, NY: IBM Corp. The significance level was set at P < 0.01.
Statistics for the PDHLQ Three first-, 12 s-, and 48 third-level indicators were identified after a two-round Delphi consultation, after which, the 48-item PDHLQ was constructed using the 48 third-level indicators . The questionnaire was amended based on the results of the pilot study. Finally, the health literacy indicator system for parasitic diseases included three first-level indicators, nine second-level indicators, and 23 third-level indicators (Table ). A total of 25 items were deleted based on the results of the EFA (Table ). The questionnaire consisted of five true or false items and 18 single-choice items (Additional file ). The area under the ROC was 98.9%, P < 0.001, and the questionnaire’s passing score was 60. PDHLQ evaluation The 23-item PDHLQ had good internal consistency (Cronbach’s α = 0.774) and a good split-half reliability (Spearman-Brown coefficient = 0.778). The results of the CFA showed a relatively good fit of the three first-level indicators of parasitic disease health literacy (2/df = 3.0, GFI = 0.937, MSA = 0.047, CFI = 0.813). A CVI of 0.88 was achieved by adhering to the scientific process of the Delphi consultation. A total of 990 valid questionnaires were collected from participants in three cities in Jiangsu (Fig. ). Their ages ranged from 15 to 69 years, with 32.2% in the 55 to 64 years age group. The yearly income of the families was mostly less than 50,000 RMB at 49.7%. Most of the participants in the survey had a low educational level, and more than half of the participants’ educational level was below the junior high school (75.7%) level. Farmers comprised the main population at 52.6%. The health literacy for parasitic diseases among the residents in the three cities of Jiangsu was 15.8% (Table ). After comparing the average score on parasitic disease health literacy in the different socio-demographic groups, the score differences in all groups were found to be significant. The average score for males was 43, which was higher than the average score for females. As participants’ education and income increased, so did their scores for parasitic disease health literacy. Civil servants achieved the highest scores, and farmers’ scores were the lowest among all the career groups. The factors affecting the participants’ health literacy scores for parasitic diseases were analyzed using stepwise regression analysis (Table ). The main factors were illiteracy, having a bachelor’s degree, an income less than 50,000 RMB, being a farmer, and being a civil servant. A significant difference in the rate of correct responses was observed between the first-level indicators and second-level indicators (Table ). The correct rates of the three first-level indicators were 51% for basic knowledge and awareness, 80% for the capacity for healthy behaviours, and 65% for healthy skills. The correct rates of the second-level indicators were low, including knowledge of the prevalence of parasitic diseases (24%), awareness of medical policies related to parasitic diseases (31%), and understanding the transmission of parasitic diseases (41%).
Three first-, 12 s-, and 48 third-level indicators were identified after a two-round Delphi consultation, after which, the 48-item PDHLQ was constructed using the 48 third-level indicators . The questionnaire was amended based on the results of the pilot study. Finally, the health literacy indicator system for parasitic diseases included three first-level indicators, nine second-level indicators, and 23 third-level indicators (Table ). A total of 25 items were deleted based on the results of the EFA (Table ). The questionnaire consisted of five true or false items and 18 single-choice items (Additional file ). The area under the ROC was 98.9%, P < 0.001, and the questionnaire’s passing score was 60.
The 23-item PDHLQ had good internal consistency (Cronbach’s α = 0.774) and a good split-half reliability (Spearman-Brown coefficient = 0.778). The results of the CFA showed a relatively good fit of the three first-level indicators of parasitic disease health literacy (2/df = 3.0, GFI = 0.937, MSA = 0.047, CFI = 0.813). A CVI of 0.88 was achieved by adhering to the scientific process of the Delphi consultation. A total of 990 valid questionnaires were collected from participants in three cities in Jiangsu (Fig. ). Their ages ranged from 15 to 69 years, with 32.2% in the 55 to 64 years age group. The yearly income of the families was mostly less than 50,000 RMB at 49.7%. Most of the participants in the survey had a low educational level, and more than half of the participants’ educational level was below the junior high school (75.7%) level. Farmers comprised the main population at 52.6%. The health literacy for parasitic diseases among the residents in the three cities of Jiangsu was 15.8% (Table ). After comparing the average score on parasitic disease health literacy in the different socio-demographic groups, the score differences in all groups were found to be significant. The average score for males was 43, which was higher than the average score for females. As participants’ education and income increased, so did their scores for parasitic disease health literacy. Civil servants achieved the highest scores, and farmers’ scores were the lowest among all the career groups. The factors affecting the participants’ health literacy scores for parasitic diseases were analyzed using stepwise regression analysis (Table ). The main factors were illiteracy, having a bachelor’s degree, an income less than 50,000 RMB, being a farmer, and being a civil servant. A significant difference in the rate of correct responses was observed between the first-level indicators and second-level indicators (Table ). The correct rates of the three first-level indicators were 51% for basic knowledge and awareness, 80% for the capacity for healthy behaviours, and 65% for healthy skills. The correct rates of the second-level indicators were low, including knowledge of the prevalence of parasitic diseases (24%), awareness of medical policies related to parasitic diseases (31%), and understanding the transmission of parasitic diseases (41%).
Principal findings This study reported the development and evaluation of a parasitic disease health scale using both qualitative and quantitative approaches. The results indicate that the PDHLQ has good reliability and validity, and could be a useful tool for assessing individuals’ health literacy for parasitic diseases. The overall PDHLQ is reliable, as indicated by its high internal consistency and split-half reliability (all coefficients > 0.7). The results of the CFA suggested that the constructs of the questionnaire fit well with the theoretical model, which represented an acceptable fit. Various methods were used to ensure the questionnaire’s content validity, including the literature review, professional consultation, and the pilot study. A 23-question scale was developed to evaluate parasitic disease health literacy. The PDHLQ covered key knowledge of parasitic epidemics, such as route of transmission, typical symptoms, and preventive measures for diseases (e.g. schistosomiasis, malaria, and foodborne parasitic diseases). The questionnaire’s passing score of 60 points, which was determined by the ROC, indicated a satisfactory level of parasitic disease health literacy. The present study developed and applied the first measurement for assessing parasitic disease health literacy to residents of China. Overall, this study offers three tips for promoting the health education about parasitic diseases. First, health education should be targeted to different people and regions. The risk factors for parasitic diseases and their necessary precautions are key knowledge points. Second, the results suggest that residents might not know why they do something, but they do it all the time. Therefore, residents should receive more information about what to do and less information about why they should do something. Third, parasitic diseases are among the neglected diseases, and sometimes even referred to as a rare disease in China. Therefore, it is important for residents to talk positively about their healthy behaviours with a doctor. Doing so may provide the key information for a diagnosis. Comparison with prior work Wuxi, Taizhou, and Lianyungang locate in southern, central, and northern Jiangsu. The most developed city among them was Wuxi, followed by Taizhou, and Lianyungang. The total number of people in Wuxi was approximately 4.52 million, in Taizhou the number was 4.60 million, and in Lianyungang it was 7.48 million people. The gross domestic product of the three cities in 2021 was 1.49 trillion yuan in Wuxi, 640.2 billion yuan in Taizhou, and 400.5 billion yuan, in Lianyungang . A study showed that soil-derived nematode diseases are still the main parasitic disease in Jiangsu Province . The rate of infection by soil-derived nematodes was 0.2-2% in northern Jiangsu, 0.1-0.5% in central Jiangsu, and 0-0.6% in southern Jiangsu according to the National Parasitic Diseases Survey from 2014 to 2015 . The PDHLQ score of the residents from Wuxi was the highest (44.18), followed by the score of the Taizhou (43.45), and Lianyungang residents (37.48), and the differences between them were statistically significant. This study found that the health literacy of parasitic diseases was positively correlated with city development and negatively correlated with parasitic infection. Furthermore, the capacity for healthy behaviour was associated with the location of the city. For example, Wuxi locate in the southern region of Jiangsu and surrounded by water. Residents who lived in Wuxi preferred to swim and play in the wild areas that were uninhabited and to eat raw or semi-raw fish, shrimp, and crab unlike their counterparts from Taizhou and Lianyungang. According to the survey, the accuracy rates of participants living in Wuxi, Taizhou, and Lianyungang, who swam and played in the wild, were 27%, 12%, and 6%, respectively. The accuracy rates of participants living in Wuxi, Taizhou, and Lianyungang who ate raw or semi-raw fish, shrimp, or crab were 40%, 16%, and 5%, respectively. The capacity for healthy behaviours of participants from Wuxi was much lower than that of the participants from the other two cities, and it was reported to be different from those with other infectious diseases . Table shows that the highest correct rate was the capacity for healthy behaviour and the lowest correct rate was basic knowledge and awareness. The correct rate of capacity for healthy behaviour was much higher than that of basic knowledge and awareness. These findings indicate a weak correlation between knowledge and behaviour, which is not consistent with the findings of other studies on health literacy . This inconsistent finding may be because parasitic diseases were considered ‘rare’ diseases and residents often neglected them. Another reason could be that healthy behaviour was a benefit of many infectious diseases, not only parasitic diseases. The low correct rate of the second-level indicators indicated that understanding ‘prevalence’, ‘medical policies’, and ‘transmission’ should be main content areas covered in parasitic disease health education. The present study showed that participants’ score on the PDHLQ was related to their age, income, employment, and educational level (Table ). Higher educational level and income were associated with a higher score, whereas older age was associated with a lower score. The main factors affecting scores included illiteracy, a bachelor’s degree, income less than 50,000 RMB, being a farmer, and being a civil servant, as found by the stepwise regression analysis. The results suggest that the health literacy of parasitic diseases should be an integrated indicator rather than one piece of demographic information. The results are consistent with previous study . Limitations This study has some limitations. First, all participants were from Jiangsu, which is one of the developed areas of China. To be able to generalize the results, participants from other regions and settings of China should be included in future studies using the PDHLQ. Second, response bias might be present because participants’ self-report responses to questions about their capacity for healthy behaviours and skills might have contained response bias.
This study reported the development and evaluation of a parasitic disease health scale using both qualitative and quantitative approaches. The results indicate that the PDHLQ has good reliability and validity, and could be a useful tool for assessing individuals’ health literacy for parasitic diseases. The overall PDHLQ is reliable, as indicated by its high internal consistency and split-half reliability (all coefficients > 0.7). The results of the CFA suggested that the constructs of the questionnaire fit well with the theoretical model, which represented an acceptable fit. Various methods were used to ensure the questionnaire’s content validity, including the literature review, professional consultation, and the pilot study. A 23-question scale was developed to evaluate parasitic disease health literacy. The PDHLQ covered key knowledge of parasitic epidemics, such as route of transmission, typical symptoms, and preventive measures for diseases (e.g. schistosomiasis, malaria, and foodborne parasitic diseases). The questionnaire’s passing score of 60 points, which was determined by the ROC, indicated a satisfactory level of parasitic disease health literacy. The present study developed and applied the first measurement for assessing parasitic disease health literacy to residents of China. Overall, this study offers three tips for promoting the health education about parasitic diseases. First, health education should be targeted to different people and regions. The risk factors for parasitic diseases and their necessary precautions are key knowledge points. Second, the results suggest that residents might not know why they do something, but they do it all the time. Therefore, residents should receive more information about what to do and less information about why they should do something. Third, parasitic diseases are among the neglected diseases, and sometimes even referred to as a rare disease in China. Therefore, it is important for residents to talk positively about their healthy behaviours with a doctor. Doing so may provide the key information for a diagnosis.
Wuxi, Taizhou, and Lianyungang locate in southern, central, and northern Jiangsu. The most developed city among them was Wuxi, followed by Taizhou, and Lianyungang. The total number of people in Wuxi was approximately 4.52 million, in Taizhou the number was 4.60 million, and in Lianyungang it was 7.48 million people. The gross domestic product of the three cities in 2021 was 1.49 trillion yuan in Wuxi, 640.2 billion yuan in Taizhou, and 400.5 billion yuan, in Lianyungang . A study showed that soil-derived nematode diseases are still the main parasitic disease in Jiangsu Province . The rate of infection by soil-derived nematodes was 0.2-2% in northern Jiangsu, 0.1-0.5% in central Jiangsu, and 0-0.6% in southern Jiangsu according to the National Parasitic Diseases Survey from 2014 to 2015 . The PDHLQ score of the residents from Wuxi was the highest (44.18), followed by the score of the Taizhou (43.45), and Lianyungang residents (37.48), and the differences between them were statistically significant. This study found that the health literacy of parasitic diseases was positively correlated with city development and negatively correlated with parasitic infection. Furthermore, the capacity for healthy behaviour was associated with the location of the city. For example, Wuxi locate in the southern region of Jiangsu and surrounded by water. Residents who lived in Wuxi preferred to swim and play in the wild areas that were uninhabited and to eat raw or semi-raw fish, shrimp, and crab unlike their counterparts from Taizhou and Lianyungang. According to the survey, the accuracy rates of participants living in Wuxi, Taizhou, and Lianyungang, who swam and played in the wild, were 27%, 12%, and 6%, respectively. The accuracy rates of participants living in Wuxi, Taizhou, and Lianyungang who ate raw or semi-raw fish, shrimp, or crab were 40%, 16%, and 5%, respectively. The capacity for healthy behaviours of participants from Wuxi was much lower than that of the participants from the other two cities, and it was reported to be different from those with other infectious diseases . Table shows that the highest correct rate was the capacity for healthy behaviour and the lowest correct rate was basic knowledge and awareness. The correct rate of capacity for healthy behaviour was much higher than that of basic knowledge and awareness. These findings indicate a weak correlation between knowledge and behaviour, which is not consistent with the findings of other studies on health literacy . This inconsistent finding may be because parasitic diseases were considered ‘rare’ diseases and residents often neglected them. Another reason could be that healthy behaviour was a benefit of many infectious diseases, not only parasitic diseases. The low correct rate of the second-level indicators indicated that understanding ‘prevalence’, ‘medical policies’, and ‘transmission’ should be main content areas covered in parasitic disease health education. The present study showed that participants’ score on the PDHLQ was related to their age, income, employment, and educational level (Table ). Higher educational level and income were associated with a higher score, whereas older age was associated with a lower score. The main factors affecting scores included illiteracy, a bachelor’s degree, income less than 50,000 RMB, being a farmer, and being a civil servant, as found by the stepwise regression analysis. The results suggest that the health literacy of parasitic diseases should be an integrated indicator rather than one piece of demographic information. The results are consistent with previous study .
This study has some limitations. First, all participants were from Jiangsu, which is one of the developed areas of China. To be able to generalize the results, participants from other regions and settings of China should be included in future studies using the PDHLQ. Second, response bias might be present because participants’ self-report responses to questions about their capacity for healthy behaviours and skills might have contained response bias.
Health literacy of parasitic diseases is an integrated indicator rather than just one piece of knowledge or behavior information. The PDHLQ has been developed with good reliability and validity and could be a useful tool for assessing the health literacy of parasitic diseases. This study reported a low percentage (15.8%) of residents from Jiangsu Province with health literacy for parasitic diseases. The correlation between knowledge and behavior was weak. The capacity for healthy behavior of parasitic disease was associated with the location and culture of the city. Based on this study’s results, we have recommended three tips for the health education of individuals that may foster their health literacy for parasitic diseases. People should receive more information about what to do and less information about why they should do something. For neglected diseases, it is important for people to talk positively about their behaviors with a doctor.
Below is the link to the electronic supplementary material. Supplementary Material 1
|
Targeted DNA sequencing and | 864881c7-8712-4f1c-954d-91973c9a7a48 | 5247573 | Pathology[mh] | Quantification of RCA products with mobile phone microscopy First, we aimed to image individual rolling circle amplification (RCA)-amplified single molecules, generated on glass slides and inside preserved cells and tissues using a mobile-phone-based microscope. For these goals, we designed and 3D-printed a light-weight optomechanical attachment that is integrated with the existing camera module of a mobile phone . This optical attachment contains two compact laser diodes (at 532 and 638 nm) for multicolour fluorescence imaging and a white light-emitting diode (LED) for bright-field transmission imaging . An integrated sample holder, consisting of a z-movement stage ( , red part) and an x–y-movement stage ( , blue part), enables three dimensional movement and alignment of the inserted sample slide. This optomechanical attachment to the mobile phone is also equipped with a cost-effective external lens module (focal length: 2.6 mm) providing a half-pitch resolution of 0.98 μm and an imaging field of view of ∼0.8 mm 2 . For the molecular analysis, we developed targeted sequencing library preparation schemes based on selector probes (refs , ) and padlock probes in situ (refs , ) , and RCA to generate micron-sized DNA coils that consist of hundreds of concatemerized repeats of the circular template and that can each be brightly labelled with fluorescent hybridization probes or sequenced . We then established that individual RCA products (RCPs) can be discriminated and precisely quantified by our mobile-phone-based microscope over a 4-log dynamic range (1 fM–10 pM), demonstrating its utility to image and analyse individual RCA-amplified single molecules . Targeted DNA sequencing for point mutation analysis RCPs can be used as sequencing libraries in next-generation sequencing (NGS) applications . To investigate whether our mobile-phone- based microscope can be used to read NGS reactions, we imaged and quantified sequencing by ligation (SBL) reactions of RCPs generated on standard microscopy slides . To confirm that SBL reactions are base-specific, we generated sequencing libraries from synthetic KRAS fragments with either wild-type sequence, resulting in base G-specific Cy3 stain , or with a codon 12 mutation, generating a base A-specific Cy5 stain . To test the feasibility of detecting small amounts of mutant KRAS DNA within the background of wild-type DNA, synthetic fragments were sequenced in a ratio of 1:1,000 mutant:wild type . Our imaging results show that nine mutant RCPs with base A-specific stain (Cy5 labelled) were successfully detected using the mobile phone microscope among 1,552 wild-type RCPs with G-specific sequencing signal (Cy3 labelled) ( , inset). At such high sequencing depths (>1,000-fold per field of view on our mobile microscope), a high mutation detection sensitivity (<1%) can potentially be achieved that is comparable to FDA-approved PCR-based KRAS diagnostic tests (the therascreen KRAS RGQ PCR Kit, https://www.qiagen.com/us/resources/technologies/oncology-companion-diagnostics/therascreen-kras-test-usa-labs/#performance ). Next, we validated the targeted sample preparation and sequencing scheme on genomic DNA extracted from cell lines. Sequencing the second base of codon 12 in genomic DNA extracted from A427 cells (heterozygous for a codon 12 mutation) with our mobile phone microscope resulted in 52% Cy3- and 48% Cy5-stained RCPs, well representing the expected ratio of mutant and wild-type molecules . Mobile sequencing of genomic DNA from Onco-DG1 cells (homozygous KRAS wild type) resulted in predominantly wild-type-specific sequencing signals . Regular benchtop fluorescence microscopy, combined with a custom-developed sequencing analysis pipeline , generated sequencing results that are very well consistent with our mobile-phone-based measurements, demonstrating the utility of our mobile targeted DNA sequencing approach. We further sequenced the extracted DNA from three different colon cancer biopsies. All three tumour samples were measured KRAS wild type through our mobile-phone-based targeted sequencing , also confirmed by both regular microscopy-based sequencing analysis and diagnostic PCR analysis . Single cell in situ analysis can find rare KRAS mutant cells An alternative mutation analysis scheme, without the need for DNA extraction and sequencing, is genotyping directly in situ within tumour tissue sections. Padlock probes, combined with RCA, enable in situ analysis of point mutations directly in preserved cells and tissues, adding molecular information to tissue morphology . These in situ generated RCPs can be imaged and quantified using the presented mobile-phone-based multimodal microscopy platform ( and ). We also performed an in situ padlock probe-based single base discrimination assay to target the most prevalent mutations in KRAS codon 12 and 13 (ref. ) occurring in 30–40% of human colon cancers . The multiplex padlock probe panel was first validated on cell lines for high mutant–wild type discrimination of the most recurrent KRAS codon 12 and 13 alleles . In situ RCA conditions were then optimized for mobile phone microscopy-based imaging with an optimal RCP intensity after a reaction time of 120 min . To quantify RCPs using our mobile phone microscope, we also developed a machine-learning-based image analysis algorithm that permits automated RCP recognition and digital counting (see Methods, and ). In total, 14 recognition features (including for example, mean, maximum and minimum intensities) were extracted from individual RCPs in the training set, which included 1,136 RCPs in total. Using this machine learning and training process, an average detection accuracy of 94.9±4.5% was achieved using our mobile phone microscope images, compared to 96.9±1.8% using the images obtained through a regular benchtop microscope. Next, we tested the utility of this mobile microscope combined with the in situ genotyping assay to detect rare cancer cells within a high background of normal cells through a cell spike-in experiment. Cells from the A549 cell line, carrying a mutation in KRAS codon 12, were spiked into a background of KRAS wild type Onco-DG1 cells in ratios of 1:100 and 1:1,000 ( and ). After the in situ padlock assay, the cells were imaged with the mobile phone microscope and the acquired images were processed as described above. KRAS mutant cells were detected in low ratios of 1:1,000, yielding a mutant:wild type RCP ratio of 0.49%, compared to 0.15% measured in KRAS wild-type cells only . These data match well with the results obtained using a conventional benchtop microscope , demonstrating the utility of our mobile platform to identify cancer cells with oncogene mutations among the majority of wild-type cells. Tissue in situ analysis accurately scores tumour samples Finally, we tested the clinical applicability of our approach for detecting KRAS mutations directly in colon tumour tissue sections . Our mobile-phone-based multimodal microscope delivers dual-colour fluorescence images that very well match with conventional microscope images of the same samples as illustrated in . We imaged several randomly chosen areas on 6 different colon cancer tumour samples and quantified RCPs , based on which the genotype is determined . Our results revealed that mobile phone microscope-based in situ genotyping resulted in 100% concordance to clinical NGS analysis and whole tissue scanning with an automated benchtop fluorescence microscop e . Using mobile-phone-based in situ genotyping, a minimum of six images of randomly chosen areas were required to detect a sufficient number of mutant RCPs and score all the mutant samples in concordance with clinical NGS data and whole tissue scanning . These results provide evidence that our mobile phone multimodal microscopy platform can be used to analyse RCA-based in situ genotyping assays and accurately genotype cancer patient biopsies directly in situ , which may facilitate integrating molecular diagnostics with tumour morphology directly in pathologists' offices and POC.
First, we aimed to image individual rolling circle amplification (RCA)-amplified single molecules, generated on glass slides and inside preserved cells and tissues using a mobile-phone-based microscope. For these goals, we designed and 3D-printed a light-weight optomechanical attachment that is integrated with the existing camera module of a mobile phone . This optical attachment contains two compact laser diodes (at 532 and 638 nm) for multicolour fluorescence imaging and a white light-emitting diode (LED) for bright-field transmission imaging . An integrated sample holder, consisting of a z-movement stage ( , red part) and an x–y-movement stage ( , blue part), enables three dimensional movement and alignment of the inserted sample slide. This optomechanical attachment to the mobile phone is also equipped with a cost-effective external lens module (focal length: 2.6 mm) providing a half-pitch resolution of 0.98 μm and an imaging field of view of ∼0.8 mm 2 . For the molecular analysis, we developed targeted sequencing library preparation schemes based on selector probes (refs , ) and padlock probes in situ (refs , ) , and RCA to generate micron-sized DNA coils that consist of hundreds of concatemerized repeats of the circular template and that can each be brightly labelled with fluorescent hybridization probes or sequenced . We then established that individual RCA products (RCPs) can be discriminated and precisely quantified by our mobile-phone-based microscope over a 4-log dynamic range (1 fM–10 pM), demonstrating its utility to image and analyse individual RCA-amplified single molecules .
RCPs can be used as sequencing libraries in next-generation sequencing (NGS) applications . To investigate whether our mobile-phone- based microscope can be used to read NGS reactions, we imaged and quantified sequencing by ligation (SBL) reactions of RCPs generated on standard microscopy slides . To confirm that SBL reactions are base-specific, we generated sequencing libraries from synthetic KRAS fragments with either wild-type sequence, resulting in base G-specific Cy3 stain , or with a codon 12 mutation, generating a base A-specific Cy5 stain . To test the feasibility of detecting small amounts of mutant KRAS DNA within the background of wild-type DNA, synthetic fragments were sequenced in a ratio of 1:1,000 mutant:wild type . Our imaging results show that nine mutant RCPs with base A-specific stain (Cy5 labelled) were successfully detected using the mobile phone microscope among 1,552 wild-type RCPs with G-specific sequencing signal (Cy3 labelled) ( , inset). At such high sequencing depths (>1,000-fold per field of view on our mobile microscope), a high mutation detection sensitivity (<1%) can potentially be achieved that is comparable to FDA-approved PCR-based KRAS diagnostic tests (the therascreen KRAS RGQ PCR Kit, https://www.qiagen.com/us/resources/technologies/oncology-companion-diagnostics/therascreen-kras-test-usa-labs/#performance ). Next, we validated the targeted sample preparation and sequencing scheme on genomic DNA extracted from cell lines. Sequencing the second base of codon 12 in genomic DNA extracted from A427 cells (heterozygous for a codon 12 mutation) with our mobile phone microscope resulted in 52% Cy3- and 48% Cy5-stained RCPs, well representing the expected ratio of mutant and wild-type molecules . Mobile sequencing of genomic DNA from Onco-DG1 cells (homozygous KRAS wild type) resulted in predominantly wild-type-specific sequencing signals . Regular benchtop fluorescence microscopy, combined with a custom-developed sequencing analysis pipeline , generated sequencing results that are very well consistent with our mobile-phone-based measurements, demonstrating the utility of our mobile targeted DNA sequencing approach. We further sequenced the extracted DNA from three different colon cancer biopsies. All three tumour samples were measured KRAS wild type through our mobile-phone-based targeted sequencing , also confirmed by both regular microscopy-based sequencing analysis and diagnostic PCR analysis .
in situ analysis can find rare KRAS mutant cells An alternative mutation analysis scheme, without the need for DNA extraction and sequencing, is genotyping directly in situ within tumour tissue sections. Padlock probes, combined with RCA, enable in situ analysis of point mutations directly in preserved cells and tissues, adding molecular information to tissue morphology . These in situ generated RCPs can be imaged and quantified using the presented mobile-phone-based multimodal microscopy platform ( and ). We also performed an in situ padlock probe-based single base discrimination assay to target the most prevalent mutations in KRAS codon 12 and 13 (ref. ) occurring in 30–40% of human colon cancers . The multiplex padlock probe panel was first validated on cell lines for high mutant–wild type discrimination of the most recurrent KRAS codon 12 and 13 alleles . In situ RCA conditions were then optimized for mobile phone microscopy-based imaging with an optimal RCP intensity after a reaction time of 120 min . To quantify RCPs using our mobile phone microscope, we also developed a machine-learning-based image analysis algorithm that permits automated RCP recognition and digital counting (see Methods, and ). In total, 14 recognition features (including for example, mean, maximum and minimum intensities) were extracted from individual RCPs in the training set, which included 1,136 RCPs in total. Using this machine learning and training process, an average detection accuracy of 94.9±4.5% was achieved using our mobile phone microscope images, compared to 96.9±1.8% using the images obtained through a regular benchtop microscope. Next, we tested the utility of this mobile microscope combined with the in situ genotyping assay to detect rare cancer cells within a high background of normal cells through a cell spike-in experiment. Cells from the A549 cell line, carrying a mutation in KRAS codon 12, were spiked into a background of KRAS wild type Onco-DG1 cells in ratios of 1:100 and 1:1,000 ( and ). After the in situ padlock assay, the cells were imaged with the mobile phone microscope and the acquired images were processed as described above. KRAS mutant cells were detected in low ratios of 1:1,000, yielding a mutant:wild type RCP ratio of 0.49%, compared to 0.15% measured in KRAS wild-type cells only . These data match well with the results obtained using a conventional benchtop microscope , demonstrating the utility of our mobile platform to identify cancer cells with oncogene mutations among the majority of wild-type cells.
in situ analysis accurately scores tumour samples Finally, we tested the clinical applicability of our approach for detecting KRAS mutations directly in colon tumour tissue sections . Our mobile-phone-based multimodal microscope delivers dual-colour fluorescence images that very well match with conventional microscope images of the same samples as illustrated in . We imaged several randomly chosen areas on 6 different colon cancer tumour samples and quantified RCPs , based on which the genotype is determined . Our results revealed that mobile phone microscope-based in situ genotyping resulted in 100% concordance to clinical NGS analysis and whole tissue scanning with an automated benchtop fluorescence microscop e . Using mobile-phone-based in situ genotyping, a minimum of six images of randomly chosen areas were required to detect a sufficient number of mutant RCPs and score all the mutant samples in concordance with clinical NGS data and whole tissue scanning . These results provide evidence that our mobile phone multimodal microscopy platform can be used to analyse RCA-based in situ genotyping assays and accurately genotype cancer patient biopsies directly in situ , which may facilitate integrating molecular diagnostics with tumour morphology directly in pathologists' offices and POC.
The sample processing steps in our approach are relatively simple to perform and do not require advanced equipment or infrastructure. Yet, in manual operation, preparation of reaction mixes and pipetting steps require a lab technician. The bottleneck in molecular pathology, however, is usually not the availability of lab technicians, but of pathologists. Nevertheless, further automation may facilitate broader implementation and use of our approach in clinical settings. To explore the initial feasibility of integration with micro-fluidics, we also imaged in situ RCA assays using the mobile phone microscope directly through a flow cell attached to a sample slide , which showed quite promising results, suggesting that liquid handling can also be integrated onto the same mobile-phone-based interface. However, preparation of tumour cryo-sections, as used in our work, still requires skilled personnel and infrastructure, which may not be available in resource-limited settings. As an alternative, tumour touch imprints can be used as a simple method to prepare samples at POC, especially in resource-limited environments . In this rapid procedure, tumours are gently pressed onto a microscope glass slide leaving a layer of cells attached to the surface with partially preserved morphological structure . These tissue fragments can then be subjected to the point mutation assay and quantified using our mobile platform. The sensitivity of our mobile-phone-based targeted sequencing method for extracted tumour DNA is currently limited by the sequencing depth and accuracy to ∼5% mutant:wild-type ratio, which is similar to diagnostic KRAS PCR Kits (the therascreen KRAS RGQ PCR Kit, https://www.qiagen.com/us/resources/technologies/oncology-companion-diagnostics/therascreen-kras-test-usa-labs/#performance ). Our sequencing depth (currently 100–200 ×) can be increased by using higher DNA concentrations. Higher accuracy can be achieved by adding a third fluorescent channel on our mobile microscope and perform a common RCP anchor stain, which aids in discarding auto-fluorescent objects that currently contribute to nonspecific signals and limit the accuracy. Ultimately, sequencing several bases will increase our accuracy by discarding auto-fluorescent objects that do not change colour over sequencing cycles. In conclusion, mobile-phone-enabled molecular diagnostic analysis may provide a simple, cost-effective and yet powerful means to integrate molecular marker information with traditional morphology analysis and might further help digital molecular pathology become widely accessible at POC offices and even in resource-limited settings. The impact of this approach goes beyond molecular pathology. Other important applications may include for example, infectious disease diagnostics, where pathogen identity and load, as well as antibiotic resistance markers, can potentially be measured using the same mobile platform. With a simple DNA sequencing library preparation scheme and the capability to image NGS reactions, mobile-phone-enabled imaging and sensing tools may soon be used for targeted DNA sequencing in clinical settings and POC offices, with the potential to dramatically decrease the cost of NGS-based diagnostics globally.
Ethical permission on human samples The in situ somatic mutation analyses of cancer and patient-matched normal tissues were approved by the Regional Ethical Review Board of Uppsala (2007/116). The tumour biopsy material used for targeted sequencing analysis and basic clinicopathological data were anonymously provided by pathologists without any patient identity or related information, and do not require ethical permission, since only research with biological material that can be traced to a person is subject to ethical approval (The Swedish ethical review act; 2003:460, section 4;3). Design of the multimodal cellphone-based microscope A handheld cellphone-based multimodal microscope was developed by integrating two battery-powered laser diodes (532 nm at 75 mW, Z-Bolt, and 638 nm at 180 mW, Mitsubishi Electric) into a 3D printed optomechanical attachment. Both of these laser diodes were mounted on a focusing stage, each with a tilted incidence angle of ∼75° creating a strong background rejection suitable for dark-field imaging and their illumination spots coincided on the sample plane. The fluorescence signals of the samples were collected by an external lens (focal length: 2.6 mm, UCTronics) added in front of the cellphone camera module and passed through a dual-band emission filter (577/690 nm, Semrock) before reaching the cellphone CMOS image sensor. Using the same cellphone-based microscope design, bright-field images of the samples can also be obtained by illuminating the sample slide vertically with a white LED (897-1183-ND, DigiKey) that was mounted above the sample. All the light sources were powered by a rechargeable battery pack (3.7 V, 1,700 mAh, Vivitar). The 3D movable sample stage consisted of two moving parts (that is, red and blue parts as shown in ) which were also created by 3D printing. The red piece was connected to the basement of the mobile phone attachment via a miniature dovetail stage (DT12, Thorlabs), and served as the z-stage, moving the sample slide in the z direction for focusing adjustment. The lateral movement of the sample slide ( x and y directions) was controlled by the blue part which was connected to the red piece via a 2-axis dovetail translation stage (DT12XY, Thorlabs). A Nokia Lumia 1020 mobile phone was used in our prototype. It is equipped with a 1/1.5 inch camera sensor that provides 38 megapixels per image (7,152 × 5,368, 4:3 mode) with a pixel size of ∼1.1 μm. The camera also has a long focal length of 6.86 mm and together with the external lens that we have used in our microscope design, it provides an effective magnification factor of ∼2.6X. Mobile-phone-based image acquisition All the three illumination sources (532 nm laser diode, 638 nm laser diode, and white LED) of the mobile phone microscope can be turned on and off independently. Each laser diode provides an oblique illumination angle (∼75°) with their illumination spots coinciding on the focal plane of the mobile microscope. Multicolour imaging was achieved by switching one laser diode on at a time, and fluorescent signals of the samples were separated from the excitation by using a multi-band fluorescence emission filter. For imaging of individual RCPs, a sample slide of interest was inserted into the mobile device from the side. First, the white LED was turned on to align and bring the sample slide into focus using the bright-field imaging mode. Then, 532 and 638 nm laser diodes were turned on sequentially and multiple cellphone fluorescence images ( N ≥5) of the same ROI at each fluorescent channel were captured. All the mobile phone images were recorded in a lossless digital negative (DNG) format by using the default smartphone camera application (that is, Nokia Pro Cam) with the same settings (focusing, white balance, ISO and integration time: 4 s per frame). Digital counting of RCPs using machine learning A machine-learning-based RCP counting algorithm was developed to process the acquired mobile phone microscope images . The image processing starts with the conversion of the acquired DNG images into 16-bit single-channel TIFF images (using the green channel for Cy3 staining and the red channel for Cy5 staining, respectively). Next, these cellphone image frames of the same ROIs (N≥5) are averaged for both Cy3 and Cy5 channels in order to improve the signal-to-noise ratio of the fluorescent images. The background noise of the averaged images is further reduced by Gaussian filtering. Finally, the Cy3, Cy5, and the transmission/bright-field images of the mobile phone microscope are co-registered and aligned with respect to each other using common spatial features to create digitally superimposed multimodal images. Our machine learning algorithm utilizes a random forest approach to differentiate real RCP signals from background noise, which combines ‘bagging' strategy and the randomness of the features to reduce overfitting. The training libraries for machine learning were established by extraction of characteristic image features from individually validated RCPs using a benchtop fluorescence microscope (Leica TCS SP5, HCX PL APO CS × 40 objective, NA=1.25, oil). A total of 14 different image features including for example, mean, maximum and minimum intensities of individual RCPs were extracted from 20 Cy3 channel images and 20 Cy5 channel images, respectively, and used as a training set for blind counting and analysis of newly acquired Cy3 and/or Cy5 mobile phone microscope images in future mobile microscopy experiments. The gold standard RCP counts used in our comparison were obtained from benchtop microscope images (for example, ) by using an intensity threshold-based automated algorithm. While this automated counting in the current study was done using a desktop computer, for both the benchtop and mobile-phone-based microscope images, it is also possible to conduct the same analysis on the smartphone itself using a custom-written application . On-slide RCA Biotinylated circles were prepared by mixing 500 pM biotinylated ligation template, 100 pM padlock probe , 0.2 mg ml −1 BSA (NEB), 1 × phi29 polymerase reaction buffer (33 mM Tris-acetate (pH 7.9), 10 mM Mg-acetate, 66 mM K-acetate, 0.1% (v/v) Tween 20, 1 mM DTT, Thermo Scientific), 680 nM ATP and 20 mU μl −1 T4 ligase in a final volume of 25 μl. The mix was incubated for 15 min at 37 °C, followed by inactivation of the enzyme at 65 °C for 2 min. Circles were diluted 1:10 from 100 pM to 1 fM in binding buffer (10 mM Tris-HCl pH 8, 10 mM EDTA, 0.05% Tween 20, 1 M NaCl). 50 μl secure seals (Grace Biolabs) were mounted on Neutravidin glass slides (Poly-An). 50 μl circle dilutions in duplicates from 10 pM to 1 fM were added and incubated for 3 h at room temperature, followed by two washes with PBS+Tween 0.05% (PBS-T). Rolling circle amplification was performed on slide by adding RCA reaction mix, containing 0.2 mg ml −1 BSA, 1x phi29 polymerase reaction buffer (33 mM Tris-acetate (pH 7.9), 10 mM Mg-acetate, 66 mM K-acetate, 0.1% Tween-20 (v/v), 125 mM dNTPs (DNA Gdansk) and 500 mU μl −1 phi29 polymerase (Olink Biosciences). The reaction was incubated overnight at room temperature. After washing twice with PBS-T, RCA products were fluorescently labelled by addition of 100 nM Cy3 and 100 nM Cy5-modified detection oligonucleotides in hybridization buffer (2 × SSC, 20% Formamide) and incubation for 1 h at room temperature. After three washes with PBS-T the secure seals were detached from the slide, followed by an ethanol series (70% 1′, 80% 1′, 100% 1′). The slides were mounted in SlowFade Gold Antifade (Thermo scientific) mounting medium with a 25 × 60 mm coverslip (Themo Scientific). For quantification, three randomly chosen positions per duplicate were imaged with the mobile phone microscope and a regular epifluorescence microscope (Zeiss Axio Imager Z2) using a 20 × microscope objective. Finally, images were analysed through CellProfiler 2.1.2 (Broad Institute) software with an in-house script for RCP recognition and quantification . Targeted sequencing of KRAS codon 12 and 13 To prepare surfaces for sequencing reactions, Neutravidin modified slides (Polyan, Germany) were functionalized with a dense layer of biotinylated selector probes by incubation in 200 μl secure seal chambers (Grace Biolabs) in binding buffer containing 1 μM biotinylated selector probe , 10 mM Tris-HCl pH7.5, 5 mM EDTA, 0.1% Tween-20 and 1 M NaCl for 20 min. The mix was removed and the slide washed with PBS (0.05% Tween-20) twice. The secure seal chamber was removed. An in-house PDMS fabricated mini-well gasket with ∼2 mm 2 area wells was applied on top of the slide and clamped so that the micro-well chambers seal on top of the slide. The mini wells serve as confined reaction chambers for the parallel processing of several samples on one slide, which is optional to increase throughput of analysis. Sequencing libraries were prepared from genomic DNA isolated from cell lines and from colon cancer tumour biopsies by column-based DNA purification (Mini-preparation Kit, Qiagen, Germany). For that purpose approximately 5 × 10 6 cells were centrifuged down and the pellet resolved in lysis buffer. Tumour DNA was extracted from 10–20 mg tumour biopsy material. Approximately 100 ng μl −1 DNA concentrations were obtained. 500 ng DNA was subjected to restriction digestion in NEB cutsmart buffer (50 mM Potassium Acetate, 20 mM Tris-acetate, 10 mM Magnesium Acetate), 300 μg ml −1 BSA and 0.25 U μl −1 MseI at 37 °C for 1 h. Before application to the selector probe functionalized slides, DNA was heat denaturated at 100 °C for 10 min and immediately put on ice. DNA samples were mixed with Ampligase buffer (20 mM Tris-HCl pH 8.3, 250 mM KCl, 100 mM MgCl 2 , 5 mM NAD, and 0.1% Triton X-100), 0.2 mg ml −1 BSA (NEB) and 250 mU μl −1 Ampligase (Epicentre) to a final volume of 20 μl. The samples were then added onto the selector probe functionalized slides in the mini wells, prepared as described above. The wells were sealed and the reaction incubated overnight at 45 °C. During the ligation reaction the genomic MseI KRAS fragments diffuse to the slide surface, the fragment ends hybridize onto the selector probes and are ligated ( , for illustration). The reaction mixtures were removed and the mini wells were detached. 200 μl secure seal reaction chambers were mounted on top of the area, on which the samples were spotted with help of the mini wells. The continuous steps were then performed in 200 μl chambers to process all the spotted samples in parallel from there on. The spotted samples were washed twice in PBS-Tween 0.05%. The ligated DNA fragments were amplified by RCA through addition of RCA reaction mixture, as described above, with addition of 25 nM KRAS specific compaction oligonucleotide . The reaction was incubated overnight at room temperature and washed twice with PBS-Tween 0.05%. Sequencing by ligation reactions were performed by first hybridizing a common anchor probe to RCA products by addition of 100 nM Alexa750 labelled anchor probe in hybridization buffer (2 × SSC, 20% Formamide) and incubation for 1 h at room temperature. After two washes in hybridization buffer without anchor probe and 1 wash in PBS-Tween 0.05%, sequencing reaction mix was applied, containing fluorescently labelled 9mer oligonucleotides with degenerated base composition on all positions except for the position that was sequenced . On the query position nucleotide A was coded by fluorescent Cy5 label, T by FITC, G by Cy3 and G by Texas red (or blank). Sequencing reaction mix contained 1 × phi29 reaction buffer, 250 nM ATP (Fermentas), 0.2 mg ml −1 BSA, 100 nM sequencing library each and 100 mU μl −1 T4 ligase (DNA GDansk). The reaction was incubated at room temperature (22 °C) for 1 h. The reaction was stopped by removing the reaction mix and washing three times in PBS-Tween 0.05%. To determine the sequenced nucleotide, the reactions were fluorescently imaged as described below. To sequence the next nucleotide position, the ligated sequencing library and anchor probe are enzymatically removed through UNG treatment, which removes the Uracils in the anchor probe and weakens the hybridization. For that purpose 20 mU μl −1 UNG (Fermentas) in phi29 reaction buffer and 0.2 mg ml −1 BSA were applied and incubated at room temperature for 10 min. The mixture was removed, the slides washed in PBS-Tween 0.05% twice and then incubated in hybridization buffer (2 × SSC, 20% Formamide) for 20 min at 45 °C. The slides were washed twice in PBS-Tween 0.05%. After that the sequencing cycle is repeated by hybridization of the same anchor probe and addition of sequencing reaction mix containing sequencing library probes that interrogate the second nucleotide position. To image the sequencing reactions the secure seals were detached from the slide, followed by an ethanol series (70% 1′, 80% 1′, 100% 1′). The slides were then mounted in SlowFade Gold mounting medium with coverslip covering the sample area. Mobile-phone-based microscopic imaging was performed as described above. Initially, to validate that sequencing reactions were successful, the reactions were imaged on Zeiss Axio Imager Z2 by fluorescent imaging of sequenced RCPs in Alexa 750, Cy3, Cy5, FITC and Texas red channels using a × 20 microscope objective. Alexa750 stain commonly identifies all RCPs and helps discarding non-RCPs that are not labelled with Alexa750. Mobile-phone-based microscopic imaging was performed with Cy5 and Cy3 excitation sources as detailed above. Sequencing reactions were quantified in an imaging analysis pipeline with Cell Profiler 2.1.2 (Broad Institute), ImageJ and Matlab (Mathworks) software, using a customized version of an analysis pipeline reported previously (Pipeline and scripts available in , for mobile images and benchtop images, respectively). Double-stained fluorescent signals, with a ratio that is higher than 0.3, were discarded as auto-fluorescent objects. Cell line and tumour section sample preparations ONCO-DG-1 and A-427 cell lines were cultured in RPMI culture medium (Sigma) without L -Glutamine supplemented with 10% FBS (Sigma) 2 mM L-Glutamine (Sigma) and 1 × Penicillin-Streptomycin (PEST, Sigma). A-549 was cultured in DMEM (Sigma) supplemented 10% FBS and 1 × PEST. When confluent, all cell lines were seeded on Superfrost Plus slides (Thermo Scientific) and allowed to attach for 12 h. The cells were then fixed in 3% paraformaldehyde (Sigma) in DEPC-treated PBS (DEPC-PBS) for 15 min at room temperature. After fixation, slides were washed twice in DEPC-PBS and dehydrated through an ethanol series of 70, 85 and 100% for 4 min each. Secure seal chambers were mounted on the slides, the cells were hydrated by a brief wash with PBS-T (DEPC-PBS with 0.05% Tween 20 (Sigma)) followed by a permeabilization with 0.1 HCl in H 2 O for 1 min at room temperature. All cell lines were obtained from DSMZ, and were tested negative for mycoplasma infection. According to the International Cell Line Authentication Committee, the Onco-DG1 cell line may be contaminated with OVCAR-3, an ovarian cancer cell line. We used Onco-DG1 in our work, because it has an elevated expression of wild-type KRAS . A potential contamination with OVCAR-3, likewise KRAS wild type, would not have affected our results and conclusions. Sensitivity and specificity of the KRAS mutation detection assay were validated in cell line dilution experiments. To test for the lowest detectable ratio of mutant cells in a majority of wild-type cells a spike-in experiment was performed. A-549 ( KRAS mutant G12S) cells were spiked into Onco-DG1 KRAS wild-type cells in ratio 1:100 and 1:1,000. Cell mixtures were seeded on Superfrost Plus slides and left to attach for 12 h. The slides were washed in cold PBS and fixed in 3% PFA for 15 min and further processed as described above. Fresh frozen human tumour tissues from colorectal cancer patients were anonymously obtained from Biobank Uppsala, Sweden, without a link to any patient related information. The somatic mutation analyses of cancer and patient-matched normal tissues were approved by the Regional Ethical Review Board of Uppsala (2007/116). Tape transfer sections (4 μm thick) were fixed in 3% paraformaldehyde in DEPC-PBS for 45 min at room temperature followed by washing with DEPC-PBS. The samples were then treated with 0.01% pepsin (Sigma) in 0.1 M HCl at 37% for 90 s. The digestion was stopped by two washes in DEPC-PBS-T for 2 min. Slides were then dehydrated through an ethanol series of 70, 85 and 100% for 2 min each. In situ mutation detection assay and tumour genotyping All in situ reactions were performed in Secure-Seal Chambers (Grace Bio-Labs Inc.), with 50 μl reaction volumes (size 9 mm diameter, 0.8 mm deep) for cells, and either 100 μl (size 13 diameter, 0.8 mm deep) or 200 μl (size 22 mm diameter, 0.8 deep) for tissues, depending on the size of the tissue sample. Reverse transcription reaction mixture was added to the chambers, containing 1 μm cDNA primer , 20 U μl −1 TranscriptMe reverse transcriptase (DNA Gdansk), 500 μM dNTP (Thermo Scientific), 0.2 mg ml −1 BSA (NEB), and 1 U μl −1 RiboLock RNase Inhibitor (Thermo Scientific) in TranscriptMe reaction buffer (DNA Gdansk). Cell slides were incubated for 3 h and tissue sections overnight at 45 °C. After brief wash in PBS-T, cells were postfixated in 3% PFA for 10 min and tissues for 30 min at room temperature. The padlock probe assay was performed as follows: In brief, 100 nM of each padlock probe was added in a mix of 1 U μl −1 Ampligase (Epicentre), 0.4 U μl −1 RNase H (DNA Gdansk), 1 U μl −1 RiboLock RNase Inhibitor, 50 mM KCl, 20% formamide in Ampligase buffer. Incubation was performed first at 37 °C for 30 min, followed by 45 min at 45 °C. After ligation, slides were washed by flushing the chambers with PBS-T. After a brief wash in PBS-T, RCA was performed, in a mix containing 1 U μl −1 phi29 DNA polymerase (Olink Biosciences) in phi29 buffer (Thermo Scientific), 250 μM dNTP, 0.2 mg ml −1 BSA and 5% Glycerol. RCA was performed on cell slides at 37 °C for 3 h (30, 60, 120 and 300 min during RCA time course experiment). Tissue samples were incubated overnight at room temperature. After RCA, the slides were washed with PBS-T. RCPs were visualized by hybridizing 0.5 μM of each corresponding detection probe in 2 × SSC and 20% formamide (Sigma) for 30 min at 37 °C. Cell nuclei were stained with DAPI during the same step. Secure-seals were removed and the slides were dehydrated through an ethanol series of 70, 85 and 100% for 2 min each. Slides were then mounted using Slowfade Antifade reagent (Life Technologies). Samples were imaged with our mobile phone microscope as described earlier. For comparison, the same samples were also imaged with Zeiss Axio Imager Z2 using a × 20 objective lens with fluorescent channels for DAPI (cell nuclei), Cy3 (mutant specific RCPs), Cy5 (wild-type-specific RCPs), and Alexa750 (Actb specific RCPs). Mobile-phone-based microscopy images of cell line dilution experiments were analysed and RCPs quantified as described above. For tumour section analysis with mobile phone microscopy, several randomly chosen regions and some predefined regions were imaged. tumour sections were additionally scanned in its full size using regular benchtop microscopy with an automated scanning stage (Zeiss Axio imager Z2). RCPs in both mobile phone and regular microscope images were analysed using the open source software CellProfiler (Pipeline available in ). RCPs were identified and quantified using different thresholds according to the intensity of the RCPs and the auto-fluorescence level of the tissue section. Images were filtered using the ‘Enhance of Suppress Features' module and RCPs then identified with the ‘Identify Objects' module using size thresholds of 2–8 pixels and intensity thresholds of 0.1–0.4, depending on the sample and the level of auto-fluorescent structures. Double stained auto-fluorescent structures in RCP size and intensity range were identified and filtered out using the Image Math module in CellProfiler. Finally, wild type and mutant specific RCPs were counted and replotted onto the tumour section image. Ratios of mutant / wild type RCPs per individual ROI and for all the ROIs combined were determined. The threshold for scoring a tumour section as wild type was set at 8%, under which tumour sections were scored as wild type, and above which tumour sections were scored as mutant. The 8% threshold is based on the clinically scored wild-type section that resulted in the highest mutant/wild type ratio (tumour section E1: 7.8% on whole tissue scanning). This background arises from strong auto-fluorescent structures, especially in necrotic regions, that can contribute to elevated unspecific mutant RCP counts. Data availability Data can be obtained by request from the authors. Image analysis codes and pipelines are available in the .
The in situ somatic mutation analyses of cancer and patient-matched normal tissues were approved by the Regional Ethical Review Board of Uppsala (2007/116). The tumour biopsy material used for targeted sequencing analysis and basic clinicopathological data were anonymously provided by pathologists without any patient identity or related information, and do not require ethical permission, since only research with biological material that can be traced to a person is subject to ethical approval (The Swedish ethical review act; 2003:460, section 4;3).
A handheld cellphone-based multimodal microscope was developed by integrating two battery-powered laser diodes (532 nm at 75 mW, Z-Bolt, and 638 nm at 180 mW, Mitsubishi Electric) into a 3D printed optomechanical attachment. Both of these laser diodes were mounted on a focusing stage, each with a tilted incidence angle of ∼75° creating a strong background rejection suitable for dark-field imaging and their illumination spots coincided on the sample plane. The fluorescence signals of the samples were collected by an external lens (focal length: 2.6 mm, UCTronics) added in front of the cellphone camera module and passed through a dual-band emission filter (577/690 nm, Semrock) before reaching the cellphone CMOS image sensor. Using the same cellphone-based microscope design, bright-field images of the samples can also be obtained by illuminating the sample slide vertically with a white LED (897-1183-ND, DigiKey) that was mounted above the sample. All the light sources were powered by a rechargeable battery pack (3.7 V, 1,700 mAh, Vivitar). The 3D movable sample stage consisted of two moving parts (that is, red and blue parts as shown in ) which were also created by 3D printing. The red piece was connected to the basement of the mobile phone attachment via a miniature dovetail stage (DT12, Thorlabs), and served as the z-stage, moving the sample slide in the z direction for focusing adjustment. The lateral movement of the sample slide ( x and y directions) was controlled by the blue part which was connected to the red piece via a 2-axis dovetail translation stage (DT12XY, Thorlabs). A Nokia Lumia 1020 mobile phone was used in our prototype. It is equipped with a 1/1.5 inch camera sensor that provides 38 megapixels per image (7,152 × 5,368, 4:3 mode) with a pixel size of ∼1.1 μm. The camera also has a long focal length of 6.86 mm and together with the external lens that we have used in our microscope design, it provides an effective magnification factor of ∼2.6X.
All the three illumination sources (532 nm laser diode, 638 nm laser diode, and white LED) of the mobile phone microscope can be turned on and off independently. Each laser diode provides an oblique illumination angle (∼75°) with their illumination spots coinciding on the focal plane of the mobile microscope. Multicolour imaging was achieved by switching one laser diode on at a time, and fluorescent signals of the samples were separated from the excitation by using a multi-band fluorescence emission filter. For imaging of individual RCPs, a sample slide of interest was inserted into the mobile device from the side. First, the white LED was turned on to align and bring the sample slide into focus using the bright-field imaging mode. Then, 532 and 638 nm laser diodes were turned on sequentially and multiple cellphone fluorescence images ( N ≥5) of the same ROI at each fluorescent channel were captured. All the mobile phone images were recorded in a lossless digital negative (DNG) format by using the default smartphone camera application (that is, Nokia Pro Cam) with the same settings (focusing, white balance, ISO and integration time: 4 s per frame).
A machine-learning-based RCP counting algorithm was developed to process the acquired mobile phone microscope images . The image processing starts with the conversion of the acquired DNG images into 16-bit single-channel TIFF images (using the green channel for Cy3 staining and the red channel for Cy5 staining, respectively). Next, these cellphone image frames of the same ROIs (N≥5) are averaged for both Cy3 and Cy5 channels in order to improve the signal-to-noise ratio of the fluorescent images. The background noise of the averaged images is further reduced by Gaussian filtering. Finally, the Cy3, Cy5, and the transmission/bright-field images of the mobile phone microscope are co-registered and aligned with respect to each other using common spatial features to create digitally superimposed multimodal images. Our machine learning algorithm utilizes a random forest approach to differentiate real RCP signals from background noise, which combines ‘bagging' strategy and the randomness of the features to reduce overfitting. The training libraries for machine learning were established by extraction of characteristic image features from individually validated RCPs using a benchtop fluorescence microscope (Leica TCS SP5, HCX PL APO CS × 40 objective, NA=1.25, oil). A total of 14 different image features including for example, mean, maximum and minimum intensities of individual RCPs were extracted from 20 Cy3 channel images and 20 Cy5 channel images, respectively, and used as a training set for blind counting and analysis of newly acquired Cy3 and/or Cy5 mobile phone microscope images in future mobile microscopy experiments. The gold standard RCP counts used in our comparison were obtained from benchtop microscope images (for example, ) by using an intensity threshold-based automated algorithm. While this automated counting in the current study was done using a desktop computer, for both the benchtop and mobile-phone-based microscope images, it is also possible to conduct the same analysis on the smartphone itself using a custom-written application .
Biotinylated circles were prepared by mixing 500 pM biotinylated ligation template, 100 pM padlock probe , 0.2 mg ml −1 BSA (NEB), 1 × phi29 polymerase reaction buffer (33 mM Tris-acetate (pH 7.9), 10 mM Mg-acetate, 66 mM K-acetate, 0.1% (v/v) Tween 20, 1 mM DTT, Thermo Scientific), 680 nM ATP and 20 mU μl −1 T4 ligase in a final volume of 25 μl. The mix was incubated for 15 min at 37 °C, followed by inactivation of the enzyme at 65 °C for 2 min. Circles were diluted 1:10 from 100 pM to 1 fM in binding buffer (10 mM Tris-HCl pH 8, 10 mM EDTA, 0.05% Tween 20, 1 M NaCl). 50 μl secure seals (Grace Biolabs) were mounted on Neutravidin glass slides (Poly-An). 50 μl circle dilutions in duplicates from 10 pM to 1 fM were added and incubated for 3 h at room temperature, followed by two washes with PBS+Tween 0.05% (PBS-T). Rolling circle amplification was performed on slide by adding RCA reaction mix, containing 0.2 mg ml −1 BSA, 1x phi29 polymerase reaction buffer (33 mM Tris-acetate (pH 7.9), 10 mM Mg-acetate, 66 mM K-acetate, 0.1% Tween-20 (v/v), 125 mM dNTPs (DNA Gdansk) and 500 mU μl −1 phi29 polymerase (Olink Biosciences). The reaction was incubated overnight at room temperature. After washing twice with PBS-T, RCA products were fluorescently labelled by addition of 100 nM Cy3 and 100 nM Cy5-modified detection oligonucleotides in hybridization buffer (2 × SSC, 20% Formamide) and incubation for 1 h at room temperature. After three washes with PBS-T the secure seals were detached from the slide, followed by an ethanol series (70% 1′, 80% 1′, 100% 1′). The slides were mounted in SlowFade Gold Antifade (Thermo scientific) mounting medium with a 25 × 60 mm coverslip (Themo Scientific). For quantification, three randomly chosen positions per duplicate were imaged with the mobile phone microscope and a regular epifluorescence microscope (Zeiss Axio Imager Z2) using a 20 × microscope objective. Finally, images were analysed through CellProfiler 2.1.2 (Broad Institute) software with an in-house script for RCP recognition and quantification .
KRAS codon 12 and 13 To prepare surfaces for sequencing reactions, Neutravidin modified slides (Polyan, Germany) were functionalized with a dense layer of biotinylated selector probes by incubation in 200 μl secure seal chambers (Grace Biolabs) in binding buffer containing 1 μM biotinylated selector probe , 10 mM Tris-HCl pH7.5, 5 mM EDTA, 0.1% Tween-20 and 1 M NaCl for 20 min. The mix was removed and the slide washed with PBS (0.05% Tween-20) twice. The secure seal chamber was removed. An in-house PDMS fabricated mini-well gasket with ∼2 mm 2 area wells was applied on top of the slide and clamped so that the micro-well chambers seal on top of the slide. The mini wells serve as confined reaction chambers for the parallel processing of several samples on one slide, which is optional to increase throughput of analysis. Sequencing libraries were prepared from genomic DNA isolated from cell lines and from colon cancer tumour biopsies by column-based DNA purification (Mini-preparation Kit, Qiagen, Germany). For that purpose approximately 5 × 10 6 cells were centrifuged down and the pellet resolved in lysis buffer. Tumour DNA was extracted from 10–20 mg tumour biopsy material. Approximately 100 ng μl −1 DNA concentrations were obtained. 500 ng DNA was subjected to restriction digestion in NEB cutsmart buffer (50 mM Potassium Acetate, 20 mM Tris-acetate, 10 mM Magnesium Acetate), 300 μg ml −1 BSA and 0.25 U μl −1 MseI at 37 °C for 1 h. Before application to the selector probe functionalized slides, DNA was heat denaturated at 100 °C for 10 min and immediately put on ice. DNA samples were mixed with Ampligase buffer (20 mM Tris-HCl pH 8.3, 250 mM KCl, 100 mM MgCl 2 , 5 mM NAD, and 0.1% Triton X-100), 0.2 mg ml −1 BSA (NEB) and 250 mU μl −1 Ampligase (Epicentre) to a final volume of 20 μl. The samples were then added onto the selector probe functionalized slides in the mini wells, prepared as described above. The wells were sealed and the reaction incubated overnight at 45 °C. During the ligation reaction the genomic MseI KRAS fragments diffuse to the slide surface, the fragment ends hybridize onto the selector probes and are ligated ( , for illustration). The reaction mixtures were removed and the mini wells were detached. 200 μl secure seal reaction chambers were mounted on top of the area, on which the samples were spotted with help of the mini wells. The continuous steps were then performed in 200 μl chambers to process all the spotted samples in parallel from there on. The spotted samples were washed twice in PBS-Tween 0.05%. The ligated DNA fragments were amplified by RCA through addition of RCA reaction mixture, as described above, with addition of 25 nM KRAS specific compaction oligonucleotide . The reaction was incubated overnight at room temperature and washed twice with PBS-Tween 0.05%. Sequencing by ligation reactions were performed by first hybridizing a common anchor probe to RCA products by addition of 100 nM Alexa750 labelled anchor probe in hybridization buffer (2 × SSC, 20% Formamide) and incubation for 1 h at room temperature. After two washes in hybridization buffer without anchor probe and 1 wash in PBS-Tween 0.05%, sequencing reaction mix was applied, containing fluorescently labelled 9mer oligonucleotides with degenerated base composition on all positions except for the position that was sequenced . On the query position nucleotide A was coded by fluorescent Cy5 label, T by FITC, G by Cy3 and G by Texas red (or blank). Sequencing reaction mix contained 1 × phi29 reaction buffer, 250 nM ATP (Fermentas), 0.2 mg ml −1 BSA, 100 nM sequencing library each and 100 mU μl −1 T4 ligase (DNA GDansk). The reaction was incubated at room temperature (22 °C) for 1 h. The reaction was stopped by removing the reaction mix and washing three times in PBS-Tween 0.05%. To determine the sequenced nucleotide, the reactions were fluorescently imaged as described below. To sequence the next nucleotide position, the ligated sequencing library and anchor probe are enzymatically removed through UNG treatment, which removes the Uracils in the anchor probe and weakens the hybridization. For that purpose 20 mU μl −1 UNG (Fermentas) in phi29 reaction buffer and 0.2 mg ml −1 BSA were applied and incubated at room temperature for 10 min. The mixture was removed, the slides washed in PBS-Tween 0.05% twice and then incubated in hybridization buffer (2 × SSC, 20% Formamide) for 20 min at 45 °C. The slides were washed twice in PBS-Tween 0.05%. After that the sequencing cycle is repeated by hybridization of the same anchor probe and addition of sequencing reaction mix containing sequencing library probes that interrogate the second nucleotide position. To image the sequencing reactions the secure seals were detached from the slide, followed by an ethanol series (70% 1′, 80% 1′, 100% 1′). The slides were then mounted in SlowFade Gold mounting medium with coverslip covering the sample area. Mobile-phone-based microscopic imaging was performed as described above. Initially, to validate that sequencing reactions were successful, the reactions were imaged on Zeiss Axio Imager Z2 by fluorescent imaging of sequenced RCPs in Alexa 750, Cy3, Cy5, FITC and Texas red channels using a × 20 microscope objective. Alexa750 stain commonly identifies all RCPs and helps discarding non-RCPs that are not labelled with Alexa750. Mobile-phone-based microscopic imaging was performed with Cy5 and Cy3 excitation sources as detailed above. Sequencing reactions were quantified in an imaging analysis pipeline with Cell Profiler 2.1.2 (Broad Institute), ImageJ and Matlab (Mathworks) software, using a customized version of an analysis pipeline reported previously (Pipeline and scripts available in , for mobile images and benchtop images, respectively). Double-stained fluorescent signals, with a ratio that is higher than 0.3, were discarded as auto-fluorescent objects.
ONCO-DG-1 and A-427 cell lines were cultured in RPMI culture medium (Sigma) without L -Glutamine supplemented with 10% FBS (Sigma) 2 mM L-Glutamine (Sigma) and 1 × Penicillin-Streptomycin (PEST, Sigma). A-549 was cultured in DMEM (Sigma) supplemented 10% FBS and 1 × PEST. When confluent, all cell lines were seeded on Superfrost Plus slides (Thermo Scientific) and allowed to attach for 12 h. The cells were then fixed in 3% paraformaldehyde (Sigma) in DEPC-treated PBS (DEPC-PBS) for 15 min at room temperature. After fixation, slides were washed twice in DEPC-PBS and dehydrated through an ethanol series of 70, 85 and 100% for 4 min each. Secure seal chambers were mounted on the slides, the cells were hydrated by a brief wash with PBS-T (DEPC-PBS with 0.05% Tween 20 (Sigma)) followed by a permeabilization with 0.1 HCl in H 2 O for 1 min at room temperature. All cell lines were obtained from DSMZ, and were tested negative for mycoplasma infection. According to the International Cell Line Authentication Committee, the Onco-DG1 cell line may be contaminated with OVCAR-3, an ovarian cancer cell line. We used Onco-DG1 in our work, because it has an elevated expression of wild-type KRAS . A potential contamination with OVCAR-3, likewise KRAS wild type, would not have affected our results and conclusions. Sensitivity and specificity of the KRAS mutation detection assay were validated in cell line dilution experiments. To test for the lowest detectable ratio of mutant cells in a majority of wild-type cells a spike-in experiment was performed. A-549 ( KRAS mutant G12S) cells were spiked into Onco-DG1 KRAS wild-type cells in ratio 1:100 and 1:1,000. Cell mixtures were seeded on Superfrost Plus slides and left to attach for 12 h. The slides were washed in cold PBS and fixed in 3% PFA for 15 min and further processed as described above. Fresh frozen human tumour tissues from colorectal cancer patients were anonymously obtained from Biobank Uppsala, Sweden, without a link to any patient related information. The somatic mutation analyses of cancer and patient-matched normal tissues were approved by the Regional Ethical Review Board of Uppsala (2007/116). Tape transfer sections (4 μm thick) were fixed in 3% paraformaldehyde in DEPC-PBS for 45 min at room temperature followed by washing with DEPC-PBS. The samples were then treated with 0.01% pepsin (Sigma) in 0.1 M HCl at 37% for 90 s. The digestion was stopped by two washes in DEPC-PBS-T for 2 min. Slides were then dehydrated through an ethanol series of 70, 85 and 100% for 2 min each.
mutation detection assay and tumour genotyping All in situ reactions were performed in Secure-Seal Chambers (Grace Bio-Labs Inc.), with 50 μl reaction volumes (size 9 mm diameter, 0.8 mm deep) for cells, and either 100 μl (size 13 diameter, 0.8 mm deep) or 200 μl (size 22 mm diameter, 0.8 deep) for tissues, depending on the size of the tissue sample. Reverse transcription reaction mixture was added to the chambers, containing 1 μm cDNA primer , 20 U μl −1 TranscriptMe reverse transcriptase (DNA Gdansk), 500 μM dNTP (Thermo Scientific), 0.2 mg ml −1 BSA (NEB), and 1 U μl −1 RiboLock RNase Inhibitor (Thermo Scientific) in TranscriptMe reaction buffer (DNA Gdansk). Cell slides were incubated for 3 h and tissue sections overnight at 45 °C. After brief wash in PBS-T, cells were postfixated in 3% PFA for 10 min and tissues for 30 min at room temperature. The padlock probe assay was performed as follows: In brief, 100 nM of each padlock probe was added in a mix of 1 U μl −1 Ampligase (Epicentre), 0.4 U μl −1 RNase H (DNA Gdansk), 1 U μl −1 RiboLock RNase Inhibitor, 50 mM KCl, 20% formamide in Ampligase buffer. Incubation was performed first at 37 °C for 30 min, followed by 45 min at 45 °C. After ligation, slides were washed by flushing the chambers with PBS-T. After a brief wash in PBS-T, RCA was performed, in a mix containing 1 U μl −1 phi29 DNA polymerase (Olink Biosciences) in phi29 buffer (Thermo Scientific), 250 μM dNTP, 0.2 mg ml −1 BSA and 5% Glycerol. RCA was performed on cell slides at 37 °C for 3 h (30, 60, 120 and 300 min during RCA time course experiment). Tissue samples were incubated overnight at room temperature. After RCA, the slides were washed with PBS-T. RCPs were visualized by hybridizing 0.5 μM of each corresponding detection probe in 2 × SSC and 20% formamide (Sigma) for 30 min at 37 °C. Cell nuclei were stained with DAPI during the same step. Secure-seals were removed and the slides were dehydrated through an ethanol series of 70, 85 and 100% for 2 min each. Slides were then mounted using Slowfade Antifade reagent (Life Technologies). Samples were imaged with our mobile phone microscope as described earlier. For comparison, the same samples were also imaged with Zeiss Axio Imager Z2 using a × 20 objective lens with fluorescent channels for DAPI (cell nuclei), Cy3 (mutant specific RCPs), Cy5 (wild-type-specific RCPs), and Alexa750 (Actb specific RCPs). Mobile-phone-based microscopy images of cell line dilution experiments were analysed and RCPs quantified as described above. For tumour section analysis with mobile phone microscopy, several randomly chosen regions and some predefined regions were imaged. tumour sections were additionally scanned in its full size using regular benchtop microscopy with an automated scanning stage (Zeiss Axio imager Z2). RCPs in both mobile phone and regular microscope images were analysed using the open source software CellProfiler (Pipeline available in ). RCPs were identified and quantified using different thresholds according to the intensity of the RCPs and the auto-fluorescence level of the tissue section. Images were filtered using the ‘Enhance of Suppress Features' module and RCPs then identified with the ‘Identify Objects' module using size thresholds of 2–8 pixels and intensity thresholds of 0.1–0.4, depending on the sample and the level of auto-fluorescent structures. Double stained auto-fluorescent structures in RCP size and intensity range were identified and filtered out using the Image Math module in CellProfiler. Finally, wild type and mutant specific RCPs were counted and replotted onto the tumour section image. Ratios of mutant / wild type RCPs per individual ROI and for all the ROIs combined were determined. The threshold for scoring a tumour section as wild type was set at 8%, under which tumour sections were scored as wild type, and above which tumour sections were scored as mutant. The 8% threshold is based on the clinically scored wild-type section that resulted in the highest mutant/wild type ratio (tumour section E1: 7.8% on whole tissue scanning). This background arises from strong auto-fluorescent structures, especially in necrotic regions, that can contribute to elevated unspecific mutant RCP counts.
Data can be obtained by request from the authors. Image analysis codes and pipelines are available in the .
How to cite this article: Kühnemund, M. et al . Targeted DNA sequencing and in situ mutation analysis using mobile phone microscopy. Nat. Commun. 8, 13913 doi: 10.1038/ncomms13913 (2017). Publisher's note : Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information Supplementary Figures, Supplementary Tables, Supplementary Notes, and Supplementary References Supplementary Software Image analysis software folder, containing Matlab codes and Cell profiler analysis pipelines.
|
Complementary and Alternative Medicine Attitudes of Gynecologic Patients: Experience in a Tertiary Clinic | 65739de9-f869-44e8-be33-9e70da7443eb | 10183933 | Gynaecology[mh] | Complementary and alternative medicine (CAM) methods are a popular form of medicine all over the world. By definition, while complementary medicine is the treatment used by patients together with conventional treatments, alternative medicine is the treatment approach used instead of conventional treatments. In an updated systematic review, the estimated rate of CAM use in the past 12 months ranged from 9.8 to 76%, with the highest prevalence in eastern Asian countries. In Turkey, there are studies among cancer patients, and the frequency of CAM use among these patients varies between 22 and 84%. Complementary and alternative medicine methods are generally divided into five categories: phytotherapy (herbals, special diets, etc.); alternative medicine (traditional Chinese medicine, acupuncture, Ayurveda, dry or wet cupping, etc.); mind-body therapies (meditation, hypnosis, biofeedback, etc.); external energy therapies (reiki, electromagnetic therapy, etc.); and body-based therapies (massage, chiropractic, etc.). A small proportion of CAM methods have been scientifically tested, while the remaining majority is widely used, despite not being proven. Therefore, healthcare professionals should note the methods used by patients and obtain information about CAM applications. Scientifically proven methods that could be used alongside conventional treatments should be distinguished from other unsafe and scientifically unproven methods. It has been shown in most of the studies that women prefer these practices more often than men. In addition, most CAM users do not notify their physicians about CAM use. In this respect, the present survey was conducted to evaluate the knowledge, attitudes, and behaviors of women about CAM applications who were admitted to our gynecology outpatient clinics. The present survey was approved by the Ethics Committee of the Ankara Zekai Tahir Burak Women's Health Education and Research Hospital (77/2019). Informed consent was obtained from all the participants. In the present cross-sectional research, a questionnaire on CAM practices was applied on 1,000 women who were admitted to the Ankara Zekai Tahir Burak Women's Health Education and Research Hospital Gynecology Outpatient Clinics between July and August 2019. The study was conducted in a tertiary maternity hospital and there were ∼ 10,000 admissions to gynecology outpatient clinics during the study period. Eligible patients were women aged ≥ 18 years old who could understand and speak Turkish without difficulty. Patients with an oncologic disease or any neurologic or psychiatric disease that might interfere with the understanding of questions and responses were excluded. The questionnaire included questions on age, education, residential area, income level, presence of chronic disease, whether they had information on CAM methods and the sources where they obtained this information, whether they found these applications useful, whether they used them before, if so what, how, and for what purpose they used them, whether they benefited from the CAM, and whether they would like to receive training about CAM and, if so, from whom. The questionnaire was filled out in person. Complementary and alternative medicine applications were divided into five subtitles: phytotherapy (herbals, special diets, etc.); alternative medicine (acupuncture, dry or wet cupping, Ayurveda, etc.); mind-body therapies (meditation, hypnosis, biofeedback, etc.); external energy therapies (reiki, electromagnetic therapy, etc.); and body-based therapies (massage, chiropractic, etc.). Each category and its examples were all written in common language that could be easily understood. There was another open-ended question under each category so that participants could enter unlisted CAM practices. For illiterate women, all the questions were read by one of the authors (Öztürk U. K.). IBM SPSS Statistics for Windows version 22.0 (IBM Corp., Armonk, NY, USA) was used for statistical analysis. The suitability of age and CAM usage variables to normal distribution were examined with the Shapiro-Wilk test. The median (interquartile range [IQR]) value was given for the age variable that did not fit normal distribution. Age analysis between CAM users and nonusers was performed with the Mann-Whitney test. Descriptors of the questionnaires are shown in numbers and percentages. Crosstables were created for the necessary questions. The chi-squared test was applied to compare CAM use according to the place of residence, income, education, smoking behavior, and chronic disease. The number of total CAM methods used in chronic disease and smoking behavior was also analyzed with the Mann-Whitney test. The relationship between age and the number of total CAM methods used was evaluated with Spearman rank correlation (Rho). The chi-squared automatic interaction detection (CHAID) method was used to determine factors affecting the use of CAM. Data were analyzed at a 95% confidence level and p < 0.05 was considered to be significant. The research was performed with 1,000 volunteers whose ages ranged from 18 to 83 years old, with a median age of 38.0 (IQR = 15.0). The demographic and clinical characteristics of the participants are given in . While 37.5% ( n = 375) of the participants stated that they used one of the CAM methods at least once, 68.5% ( n = 257) of them used ≥ 2 CAM methods. The most preferred CAM method was phytotherapy (91.2%) followed by acupuncture (7.9%), detoxification (4.1%), wet cupping (2.4%), hypnosis (2.2%), ozone (2.2%), and leeches (0.5%). The median age of CAM users was higher than that of the nonusers (39.0 [IQR = 15.0] versus 36.0 [IQR = 16], Z = 3.191; p = 0.001, respectively). No significant relationship was found between the age of the participants and the number of total CAM methods used (Rho = - 0.072; p = 0.166). The percentage of CAM users was higher in the moderate-income group (41.0; χ 2 = 31.785; p < 0.001), and in women with chronic disease (44.5; χ 2 = 6.765; p = 0.009). There were no statistically significant differences in the percentage of CAM use according to residential area, education level, or smoking behavior (χ 2 = 1.975; p = 0.373; χ 2 = 4.188; p = 0.242; and χ 2 = 2.038; p = 0.153, respectively). The number of total CAM methods used was similar according to residential area, income status, educational level, smoking behavior, and the presence of chronic disease ( p > 0.05). When the factors affecting the use of CAM were examined with the CHAID method, having knowledge about CAM (χ 2 = 353.197; p < 0.001) was determined to be the most effective factor. Complementary and alternative medicine use was significantly higher among those who had information about CAM (χ 2 = 31.311; p < 0.001). Complementary and alternative medicine usage according to the factors is shown as a decision tree in . It was found that the established decision tree CHAID model correctly classified CAM users in 70.7% of cases and nonusers in 77.8%. The correct classification rate of the CHAID model was determined to be 78.5%. shows the opinions of the participants about the usefulness of CAM, the percentages of use according to their opinions, and whether they have benefited from CAM or not. While 80.7% of the participants thought that CAM was useful, only 44.5% of them stated that they have used CAM before. also shows the demographic and clinical characteristics of women according to their opinion about the usefulness of CAM. Of the women who thought that CAM was not useful, 6.7% (13/193) had used CAM before. Of the women who used CAM, 40.8% stated that they have benefited from CAM, and 53.4% stated that they have partially benefited from it. shows the benefit rates of different CAM modalities. Phytotherapy was the most preferred CAM method. The most common herbals were onion (18.9%) followed by black seed (12.3%), ginger (11.6%), garlic (11.4%), and stinging nettle (6.1%). St. John's wort, yarrow, echinacea, and the fruit of vitex agnus-castus were used by < 5% of the respondents; however, there were a total of 60 different herbs (mixed herbal tea, sage, parsley, aloe vera, flaxseed etc.) among the answers given . Out of the 336 patients that used phytotherapy, 133 (39.5%) of them used ≥ 2 herbals in various forms. When the plants used in phytotherapy were examined according to their intended use, the herbals with a use frequency of < 3% were included under the title of others . The most common purposes of use were stress (15.4%) and fatigue (15.3%) followed by weight problems (14.5%), menstrual irregularities (13.0%), and body care (11.1%). Phytotherapy was used due to ovarian cysts for 6.8%, while the percentage of use due to uterine leiomyoma and sexual health were 3.3% for each. Phytotherapy was used for a total of 40 different reasons. A total of 72.4% of the participants stated that they would like to get information on CAM methods, while 27.6% stated that they did not want to. Of the women who wished to get information, 65.6% of them wanted to get information on phytotherapy, followed by acupuncture (20.4%), hypnosis (14.5%), detoxification therapy (12.9%), and ozone therapy (10.7%). They wished to be informed by physicians (93.8%), followed by newspapers and television (17.7%), internet (9.8%), and herbalists (6.1%) . Although CAM is widely used all over the world, the use rates and the methods of CAM differ among countries due to social, cultural, and economic factors. Sociodemographic features such as age, gender, education level, income, and health status have all been found to be different in studies conducted around the world. In Europe, the rate of CAM use was reported as ranging from 10 to 40% across countries and was more common among women, people with higher education, and those with good health. In addition, the nature of health problems and income status were found to affect CAM methods. According to the results of a national survey in the United States, single CAM use was common among adults, and CAM practices differed by age, gender, ethnic descent, and educational background. In Japan, the prevalence of CAM use was 62.1% and female gender and health anxiety were found to be associated with higher CAM use. In our study, we aimed to evaluate knowledge, attitudes, and behaviors about CAM practices of women admitted to our gynecology outpatient clinic. The results of our study showed that 37.5% of the women used CAM before and that CAM use was more common in women with moderate income and in those who had a chronic disease. The median age of CAM users was 39 years old, which was higher than that of nonusers. Similarly, Sarris et al. reported that middle-aged women (aged between 40 and 64 years old) used any kind of CAM significantly more than older women (≥ 65 years old). Considering the frequent use of CAM in the reproductive age of women, and considering that this use is generally without the knowledge of healthcare professionals, both obstetricians and gynecologists should be careful and vigilant about CAM use. In Germany, the rate of CAM use was of 61.4% among third-trimester pregnant women. The use and interactions of CAM are extensively researched especially among gynecologic oncology patients. In adolescents, primary dysmenorrhea draws attention as the primary reason. In a survey, women were asked to report the use of ≥ 1 CAM methods especially for the treatment of obstetric or gynecologic problems, and the reported rate was of 54.5%. In our study, we asked the reason for CAM use without any restriction, and to the best of our knowledge, there is no study in the literature reflecting all reasons for the gynecological patient group. In Turkey, most of the studies have been conducted among gynecological cancer patients. The frequency of CAM use in these studies varies between 47 and 61%. However, in the study of Turkish women with urinary incontinence, the frequency of CAM use was found to be very low, at 7%. Our results showed that the most common sources of information on CAM methods were newspapers/television and family/friends. Similar results have been reported from different countries. In our study, the least common source of information about CAM was physicians, with 8.5%, although having knowledge about CAM was determined as the most effective factor in CAM use. In the aforementioned studies, the rate of physicians as a source of information ranged from 6.4 to 15.7%. In another study from our country, Kav et al. found that 10% of the patients received information from a physician. The main conclusion arising from these studies is that health professionals, especially physicians, tend not to inform patients on this subject. Also, > 90% of the participants in our survey stated that they wanted to get information from physicians. In this regard, Münstedt et al. emphasized that information about CAM applications is rarely given in universities, especially in medical faculties, and that healthcare professionals must obtain this information from other places. In our study, the most used CAM method was phytotherapy, with 90%. As stated, due to sociocultural and socioeconomic factors, the preferred method of CAM varies considerably between countries. Yet, in the study by Sirois et al., phytotherapy stands out as the most frequently used method. In other studies conducted with gynecologic cancer patients in our country, phytotherapy was found to be the most used CAM method. We determined that 60 different herbals were used by the participants and that the most popular was onion, followed by black seed. Onion has been shown to have antioxidant, anticarcinogenic, hypolipidemic, hypoglycemic, and antiaggregatory effects. The reason for onion being the most preferred herb might be the popularity of discourses on onion cures, especially in visual media in Turkey nowadays. Nevertheless, these preferences differ from oncological patients, since stinging nettle was the preferred herbal for cancer patients in our country. On the other hand, even pregnant women do not worry about using herbal remedies, as they feel safe in using them. Pallivalapila et al. reported that 40 different plant varieties have been used by 62% of pregnant women in the 3 rd trimester. Thus, the point that must be taken into consideration is the possible negative interactions between prescribed drugs and herbs. For example, black seed, the second most preferred herbal in our study, is an herb known to have antidiabetic properties. However, it was claimed to cause acute renal failure in a diabetic woman. Another interesting finding of our study was that approximately one-fourth of women used phytotherapy because of gynecologic problems, including menstrual irregularities (13%), ovarian cysts (6.8%), uterine leiomyoma (3.3%), and sexual health (3.3%). In Germany, the gynecological reasons for gynecologists to recommend phytotherapy were postmenopausal symptoms (63%), premenstrual symptoms (56%), infertility (23%), genital inflammation (22%), incontinence (18%), polycystic ovary syndrome (16%), and uterine leiomyoma (12%). The major limitation of the present survey was that it was conducted in a single center. The results of the present survey cannot be attributed to the general population of women in Turkey. However, the survey was conducted in one of the largest maternity hospitals of our country. Additionally, it has the advantage of having a larger number of participants compared with similar studies in the literature and having participants fill out the questionnaire in person. The results of our survey have shown that more than one out of every three women who applied to the gynecology outpatient clinic have used a CAM method at least once before. The most used CAM method was phytotherapy, which can be attributed to the stereotypical and conservative view that might be based on traditional experience lasting for centuries. Considering that the most effective factor in the use of CAM is to have knowledge about these methods and the most desired source of information on these methods is physicians, we gynecologists and obstetricians have more responsibilities. We need to be more knowledgeable of these methods to provide correct guidance for women to access accurate and effective information. |
Establishing a Virtual Corneal Clinic: A Real-Time Teleophthalmology Approach | d3e506de-f201-4401-9ab5-2e28b94d8a90 | 9894129 | Ophthalmology[mh] | This retrospective study was conducted on 72 consecutive patients (84 eyes) who attended VCC sessions, which involved coming to a JEC hospital to meet the appointed doctor and undergoing prespecified preliminary investigations. Then followed by clinical examination by the attending JEC subspecialists, with simultaneous real-time video consultation with the primary corneal surgeon from ECS (DTHT), Singapore, using ZOOM Meeting software (Zoom Video Communication Inc, San Jose, CA). The process involved a real-time video from slit-lamp examinations and reviewing of clinical case notes and investigations using video-linked EMR systems. Synchronous bidirectional communication was achieved through a ubiquitous video communication with appropriate audiovisual technologies. The study inclusion criterion was ECS Indonesian corneal patients with a history of keratoplasty and keratoprosthesis surgery who were unable to come to Singapore for their scheduled clinical appointments due to COVID-19 travel restrictions. Sessions included in this study were performed over a 10-month period between June 2020 and April 2021 and were recorded with patient consent. The study was approved by the Medical and Health Research Ethics Committee of Faculty of Medicine, Public Health and Nursing of Universitas Gadjah Mada–Dr. Sardjito General Hospital (Ref. No: KE/FK/0690/EC/2021) and adhered to the tenets of the Declaration of Helsinki. VCC data included patient demographics, a concise but detailed summary of previous corneal and keratoplasty history from the referring Singapore corneal specialist, all VCC examination and investigations performed, and clinical decisions made at each VCC appointment. Virtual Corneal Clinic Protocol Appointment Scheduling, Medical Summaries, and Pretesting Requirements ECS Indonesian corneal patients with a history of PKP, DMEK, DSAEK, DALK, and keratoprosthesis surgery performed who were unable to come to Singapore for their scheduled clinical appointments due to COVID-19 travel restrictions were contacted by ECS through emails to ascertain whether they were keen to physically attend clinic visits at JEC locations in Jakarta, involving their Singapore corneal specialist participating on a teleophthalmology consultation, along with collaborating Indonesian ophthalmologists. Patients received detailed information regarding the VCC process, including confidentiality issues and financial counseling, and on full informed consent, ECS emailed relevant patient details to JEC. Patients were made aware at the outset that they would have the privilege of 2 simultaneous clinical/specialist consultations but would also have to pay consultation charges for both clinics, which generally mirrored the normal clinical consultation charges of the respective clinics involved (the cost-savings of avoiding air travel far outweighed clinical charges). VCC appointments were then coordinated and scheduled between ECS, JEC, and patients—in both ECS and JEC, these sessions were booked and dovetailed into the specialists' respective corneal clinics which coincided. Detailed medical summaries and preexamination investigation requests from ECS were sent to JEC before the appointments to enable receiving JEC ophthalmologists to review the patient's medical history and for JEC clinic staff to plan for relevant preexamination test (PET) requests from the Singapore corneal specialist ( Fig. ). Preexamination Investigations Where necessary, preexamination investigations were requested by the Singapore corneal specialist, as part of the original continuation of care of the particular corneal patient, and these tests were booked at JEC, with patients requested to come earlier at specific time points just before the VCC. Examples of PET are provided in Table . The results of these investigations were then reviewed by the doctors simultaneously during the VCC. For pediatric patients who would be not compliant with slit-lamp examination, the VCC surgeons discussed sedation or examination under anesthesia findings, with the parents also present so that they were informed of the child's progress. Where possible, prior sedations or general anesthetic examinations were requested and preplanned before the VCC session. Electronic Hardware and Software Required to Establish a VCC The establishment of a VCC between ECS and JEC was facilitated by the fact that both sites had fully computerized clinics with full internet access and large screen monitors in the clinics, and well-established electronic medical record (EMR) support and remote desktop software, with the list of electronic hardware and software listed in Table . Good internet access was important, and dedicated omnidirectional microphones were used to reduce ambient noise levels. ZOOM Meeting software (Zoom Video Communication Inc., San Jose, CA) was used with remote desktop software to synchronize the teleophthalmology consults, with images received from the slit-lamp biomicroscope saved into the patients’ EMR. Each institution used its own electronic records to document VCC sessions. As the patients were seen at JEC, JEC's electronic records were usually viewed by ECS live and online, but on occasion, where needed, previous ECS medical records were also shared with JEC doctors by screen-sharing. Where clinical judgments were made, to subjectively grade or evaluate the severity or extent of clinical findings, these issues were subjectively described and deliberated on (usually until a consensus was reached) and subsequently recorded as such in both sets of medical records by the individual doctors. We used internet bandwidths with a minimum download speed of 20 Mbps and a minimum upload speed of 3 Mbps. Two thousand one hundred pixels were presented on the desktop monitor with a resolution of 1920 pixels by 1080 pixels. We did not specify monitor screen sizes but simply used the available 17-inch desktop computer screens with a standard resolution, present in the clinics. Teleophthalmology Informed Consent and Patient Confidentiality All patients agreed to have their ECS medical records shared with the JEC VCC staff and were required to sign a teleophthalmology consent form which enabled full disclosure of patient demographics and identifiers, contact details, and full medical details. In addition, consent for video recording of the teleconsultations was available as an option for patients. Care was taken to ensure that patient confidentiality was respected and fully conformed to the Singapore data protection laws, including the Personal Data Protection Act (PDPA) Law No. 26 of 2012, and also complied with the Singapore Ministry of Health National Telemedicine Guidelines (2015) and Indonesian medical records and data protection laws Peraturan Menteri Kesehatan No. 269 tahun 2018 tentang Rekam Medis and Peraturan Menteri Komunikasi dan Informatika No. 20 tahun 2016 tentang Perlindungan Data Pribadi dalam Sistem Elektronik. To ensure full privacy and security, only authorized parties were provided with the Zoom user ID and password, and once parties were admitted to the session, the meeting room was “closed” to ensure no additional log-ons were permitted. In addition, VCC session recordings were stored on JEC's internal servers, not on Zoom, and all clinical and administrative staff (including IT staff) of both JEC and ECS are bound by medical confidentiality. Post-VCC At the conclusion of the VCC, a short medical summary was sent to the patient, detailing the various changes to treatment. In addition, any medications which were required and not available in Indonesia were then couriered to the patient from Singapore. VCC processes are shown in Fig. .
Appointment Scheduling, Medical Summaries, and Pretesting Requirements ECS Indonesian corneal patients with a history of PKP, DMEK, DSAEK, DALK, and keratoprosthesis surgery performed who were unable to come to Singapore for their scheduled clinical appointments due to COVID-19 travel restrictions were contacted by ECS through emails to ascertain whether they were keen to physically attend clinic visits at JEC locations in Jakarta, involving their Singapore corneal specialist participating on a teleophthalmology consultation, along with collaborating Indonesian ophthalmologists. Patients received detailed information regarding the VCC process, including confidentiality issues and financial counseling, and on full informed consent, ECS emailed relevant patient details to JEC. Patients were made aware at the outset that they would have the privilege of 2 simultaneous clinical/specialist consultations but would also have to pay consultation charges for both clinics, which generally mirrored the normal clinical consultation charges of the respective clinics involved (the cost-savings of avoiding air travel far outweighed clinical charges). VCC appointments were then coordinated and scheduled between ECS, JEC, and patients—in both ECS and JEC, these sessions were booked and dovetailed into the specialists' respective corneal clinics which coincided. Detailed medical summaries and preexamination investigation requests from ECS were sent to JEC before the appointments to enable receiving JEC ophthalmologists to review the patient's medical history and for JEC clinic staff to plan for relevant preexamination test (PET) requests from the Singapore corneal specialist ( Fig. ). Preexamination Investigations Where necessary, preexamination investigations were requested by the Singapore corneal specialist, as part of the original continuation of care of the particular corneal patient, and these tests were booked at JEC, with patients requested to come earlier at specific time points just before the VCC. Examples of PET are provided in Table . The results of these investigations were then reviewed by the doctors simultaneously during the VCC. For pediatric patients who would be not compliant with slit-lamp examination, the VCC surgeons discussed sedation or examination under anesthesia findings, with the parents also present so that they were informed of the child's progress. Where possible, prior sedations or general anesthetic examinations were requested and preplanned before the VCC session. Electronic Hardware and Software Required to Establish a VCC The establishment of a VCC between ECS and JEC was facilitated by the fact that both sites had fully computerized clinics with full internet access and large screen monitors in the clinics, and well-established electronic medical record (EMR) support and remote desktop software, with the list of electronic hardware and software listed in Table . Good internet access was important, and dedicated omnidirectional microphones were used to reduce ambient noise levels. ZOOM Meeting software (Zoom Video Communication Inc., San Jose, CA) was used with remote desktop software to synchronize the teleophthalmology consults, with images received from the slit-lamp biomicroscope saved into the patients’ EMR. Each institution used its own electronic records to document VCC sessions. As the patients were seen at JEC, JEC's electronic records were usually viewed by ECS live and online, but on occasion, where needed, previous ECS medical records were also shared with JEC doctors by screen-sharing. Where clinical judgments were made, to subjectively grade or evaluate the severity or extent of clinical findings, these issues were subjectively described and deliberated on (usually until a consensus was reached) and subsequently recorded as such in both sets of medical records by the individual doctors. We used internet bandwidths with a minimum download speed of 20 Mbps and a minimum upload speed of 3 Mbps. Two thousand one hundred pixels were presented on the desktop monitor with a resolution of 1920 pixels by 1080 pixels. We did not specify monitor screen sizes but simply used the available 17-inch desktop computer screens with a standard resolution, present in the clinics. Teleophthalmology Informed Consent and Patient Confidentiality All patients agreed to have their ECS medical records shared with the JEC VCC staff and were required to sign a teleophthalmology consent form which enabled full disclosure of patient demographics and identifiers, contact details, and full medical details. In addition, consent for video recording of the teleconsultations was available as an option for patients. Care was taken to ensure that patient confidentiality was respected and fully conformed to the Singapore data protection laws, including the Personal Data Protection Act (PDPA) Law No. 26 of 2012, and also complied with the Singapore Ministry of Health National Telemedicine Guidelines (2015) and Indonesian medical records and data protection laws Peraturan Menteri Kesehatan No. 269 tahun 2018 tentang Rekam Medis and Peraturan Menteri Komunikasi dan Informatika No. 20 tahun 2016 tentang Perlindungan Data Pribadi dalam Sistem Elektronik. To ensure full privacy and security, only authorized parties were provided with the Zoom user ID and password, and once parties were admitted to the session, the meeting room was “closed” to ensure no additional log-ons were permitted. In addition, VCC session recordings were stored on JEC's internal servers, not on Zoom, and all clinical and administrative staff (including IT staff) of both JEC and ECS are bound by medical confidentiality. Post-VCC At the conclusion of the VCC, a short medical summary was sent to the patient, detailing the various changes to treatment. In addition, any medications which were required and not available in Indonesia were then couriered to the patient from Singapore. VCC processes are shown in Fig. .
ECS Indonesian corneal patients with a history of PKP, DMEK, DSAEK, DALK, and keratoprosthesis surgery performed who were unable to come to Singapore for their scheduled clinical appointments due to COVID-19 travel restrictions were contacted by ECS through emails to ascertain whether they were keen to physically attend clinic visits at JEC locations in Jakarta, involving their Singapore corneal specialist participating on a teleophthalmology consultation, along with collaborating Indonesian ophthalmologists. Patients received detailed information regarding the VCC process, including confidentiality issues and financial counseling, and on full informed consent, ECS emailed relevant patient details to JEC. Patients were made aware at the outset that they would have the privilege of 2 simultaneous clinical/specialist consultations but would also have to pay consultation charges for both clinics, which generally mirrored the normal clinical consultation charges of the respective clinics involved (the cost-savings of avoiding air travel far outweighed clinical charges). VCC appointments were then coordinated and scheduled between ECS, JEC, and patients—in both ECS and JEC, these sessions were booked and dovetailed into the specialists' respective corneal clinics which coincided. Detailed medical summaries and preexamination investigation requests from ECS were sent to JEC before the appointments to enable receiving JEC ophthalmologists to review the patient's medical history and for JEC clinic staff to plan for relevant preexamination test (PET) requests from the Singapore corneal specialist ( Fig. ).
Where necessary, preexamination investigations were requested by the Singapore corneal specialist, as part of the original continuation of care of the particular corneal patient, and these tests were booked at JEC, with patients requested to come earlier at specific time points just before the VCC. Examples of PET are provided in Table . The results of these investigations were then reviewed by the doctors simultaneously during the VCC. For pediatric patients who would be not compliant with slit-lamp examination, the VCC surgeons discussed sedation or examination under anesthesia findings, with the parents also present so that they were informed of the child's progress. Where possible, prior sedations or general anesthetic examinations were requested and preplanned before the VCC session.
The establishment of a VCC between ECS and JEC was facilitated by the fact that both sites had fully computerized clinics with full internet access and large screen monitors in the clinics, and well-established electronic medical record (EMR) support and remote desktop software, with the list of electronic hardware and software listed in Table . Good internet access was important, and dedicated omnidirectional microphones were used to reduce ambient noise levels. ZOOM Meeting software (Zoom Video Communication Inc., San Jose, CA) was used with remote desktop software to synchronize the teleophthalmology consults, with images received from the slit-lamp biomicroscope saved into the patients’ EMR. Each institution used its own electronic records to document VCC sessions. As the patients were seen at JEC, JEC's electronic records were usually viewed by ECS live and online, but on occasion, where needed, previous ECS medical records were also shared with JEC doctors by screen-sharing. Where clinical judgments were made, to subjectively grade or evaluate the severity or extent of clinical findings, these issues were subjectively described and deliberated on (usually until a consensus was reached) and subsequently recorded as such in both sets of medical records by the individual doctors. We used internet bandwidths with a minimum download speed of 20 Mbps and a minimum upload speed of 3 Mbps. Two thousand one hundred pixels were presented on the desktop monitor with a resolution of 1920 pixels by 1080 pixels. We did not specify monitor screen sizes but simply used the available 17-inch desktop computer screens with a standard resolution, present in the clinics.
All patients agreed to have their ECS medical records shared with the JEC VCC staff and were required to sign a teleophthalmology consent form which enabled full disclosure of patient demographics and identifiers, contact details, and full medical details. In addition, consent for video recording of the teleconsultations was available as an option for patients. Care was taken to ensure that patient confidentiality was respected and fully conformed to the Singapore data protection laws, including the Personal Data Protection Act (PDPA) Law No. 26 of 2012, and also complied with the Singapore Ministry of Health National Telemedicine Guidelines (2015) and Indonesian medical records and data protection laws Peraturan Menteri Kesehatan No. 269 tahun 2018 tentang Rekam Medis and Peraturan Menteri Komunikasi dan Informatika No. 20 tahun 2016 tentang Perlindungan Data Pribadi dalam Sistem Elektronik. To ensure full privacy and security, only authorized parties were provided with the Zoom user ID and password, and once parties were admitted to the session, the meeting room was “closed” to ensure no additional log-ons were permitted. In addition, VCC session recordings were stored on JEC's internal servers, not on Zoom, and all clinical and administrative staff (including IT staff) of both JEC and ECS are bound by medical confidentiality.
At the conclusion of the VCC, a short medical summary was sent to the patient, detailing the various changes to treatment. In addition, any medications which were required and not available in Indonesia were then couriered to the patient from Singapore. VCC processes are shown in Fig. .
Eighty-two patients involved in this teleophthalmology program were documented during the period June 2020-April 2021. Of these, 10 patients were excluded from this study analysis because they were not corneal patients but were patients with other ophthalmic morbidities including retinal and glaucoma cases. Of the 72 corneal patients (84 eyes) who met the inclusion criteria, 44 (61%) were women and 28 (39%) were men. The mean age of the patients was 55.4 ± 18.6 years, ranging from 1 to 88 years. The purpose of VCC was that of postkeratoplasty follow-up in 63 patients (87.5%) under the care of the primary corneal specialist (DTHT), whereas 9 patients (12.5%) were first consulted who either requested for consultation with the Singapore corneal specialist or were referred by the Indonesian specialists. In addition to DTHT as the primary ECS corneal specialist, 4 other Singapore ECS subspecialists (in glaucoma, oculoplastics, medical, and surgical retina) were also co-opted into the VCC sessions where required. On the JEC side, a total of 3 corneal specialists and 4 additional doctors from other subspecialties were also co-opted to participate. Thirty-six patients (50.0%) originated from within Jakarta city, whereas the other half were from wide-ranging cities dispersed throughout the Indonesian archipelago including other cities such as Surabaya (5 patients (6.9%)), Bandung (4 patients (5.6%)), and Bekasi (4 patients (5.6%)). Patients often used internal domestic flights to fly to Jakarta for their VCC appointment because the various cities were located far apart across Indonesian islands—the furthest distance between these cities was approximately 5245 km between east and west, a distance exceeding that between Los Angeles and New York (4489 km). All forms of keratoplasty, including keratoprosthesis, were represented in this patient group—Descemet membrane endothelial keratoplasty (DMEK, 37 eyes, 44.0%) was the most common form of keratoplasty, followed by Descemet stripping automated endothelial keratoplasty (DSAEK, 14 eyes, 16.7%), penetrating keratoplasty (PKP, 11 eyes, 13.1%), deep anterior lamellar keratoplasty (DALK, 4 eyes, 4.8%), Boston keratoprosthesis type 1 (4 eyes, 4.8%), and osteo-odonto-keratoprosthesis (OOKP, 1 eye, 1.2%). Details are provided in Table . Video real-time slit-lamp biomicroscopic examination (SLE) was the most important clinical procedure during VCC which provided the maximum benefit of teleophthalmology because the corneal specialist in Jakarta performed SLE real-time, which was viewed in Singapore by the referring corneal specialist. Communication between surgeons during SLE enabled detailed discussion around examination of specific areas of the grafted cornea, with the ability of the viewing Singapore surgeon to request higher magnification of specific areas of the graft, specify closer examination of areas such as the corneal state adjacent to a glaucoma tube and the scleral or conjunctival state around the tube plate or trabeculectomy bleb, or request for the removal of bandage contact lenses or for corneal staining. Overall, the video-linked resolution was clearly sufficient to detect minor corneal changes such as faint stromal nebulae; very mild, localized areas of stromal edema; or superficial epitheliopathy without surface staining, which was generally adequate to determine the overall graft status, be it DMEK, DSEK, or DALK. However, with the resolution of the video image clearly not as ideal as actual SLE, examination for the presence of very fine keratic precipitates (KPs), or anterior chamber cellular activity, was generally not easily seen, although larger, more obvious pigmented and unpigmented KPs were easily spotted, and in many instances, the Singapore surgeon requested the Indonesian examiner to specifically look for the presence of fine KPs or intraocular inflammation and depended on the findings of the examiner as slit-lamp video-resolution clearly was insufficient to see fine KPs or AC activity. A particular advantage of VCC, however, was that surgeons were able to discuss the current findings and state of the graft collectively so as to enable an integrated clinical response or management, which was also heard by and noted by the patient or attending relative, as part of an integrated examination and patient management decision. This was especially significant in the cases where the JEC specialist had referred the patient in the original instance to the Singapore specialist for a medical opinion and further treatment. The ability to review the preexamination tests (PETs) which had just been performed before the VCC, which classifies as an additional hybrid telemedicine approach, also greatly facilitated and expedited the patient management process. For example, the ability to review the most current corneal endothelial cell count after SLE (and also see the digital specular microscopy image) and compare this to previous cell counts which had been performed in Singapore greatly facilitated the spatial trajectory of the surviving graft, while reviewing (and comparing previous) high-resolution AS-OCT and pachymetric data, both quantitative and qualitative, was also very informative. Similarly, the evaluation of adjunctive tests, such as automated perimetry, and posterior segment investigations, such as ultrasound B-scans, and OCTs also was invaluable in the holistic management of patients, especially as many were complex cases with secondary glaucoma and previous glaucoma surgery. Where needed, preplanned, simultaneous involvement of additional specialists (glaucoma, retina, oculoplastic, and pediatric) was a significant additional advantage to expedite efficient clinical care, with these additional specialists joining in the VCC from their respective clinic rooms or hospital locations, something which patients also greatly appreciated because ordinarily, they would have to be shunted sequentially to corneal, glaucoma, or retinal clinics during 1 visit or at other specific visits. This was however not always possible, depending on the clinic timings or the various specialists. One further advantage to patients was the ability to involve additional family members not colocated with the patient, as and when needed—1 specific VCC involved important decision making in the management of a one-eyed patient with end-stage corneal and glaucoma disease, whereby the son, a senior decision maker of the family, joined the VCC from the United States. We evaluated the overall clinical impact of VCC by determining the clinical management changes to the treatment regimen of the VCC patients within the study period. Of the 72 patients, 57 patients (79.2%) had their topical and/or systemic medications altered in some form or other after VCC, mainly pertaining to tapering (or increasing) of topical steroid dosing regimens and variation of glaucoma medication combinations. Twelve patients (16.6%) required additional follow-up referrals to other specialties (glaucoma: 6 (8.3%), retina: 5 (6.9%), and oculoplastic 1 (1.4%))—these comprised patients with de novo comorbidities diagnosed during VCC or cases in which the particular subspecialist was not able to attend the VCC. In all instances, the results/opinions of the subsequent subspecialty referral were subsequently conveyed to the VCC team by group email, leading to email discussions between managing surgeons and specific changes to the clinical management detailed in the VCC clinical records of the patient. Finally, VCC also affected on a small, but clinically significant group of patients who required subsequent surgical intervention after VCC. Four patients (5.5%) had additional therapeutic interventions which were performed at JEC, which included glaucoma tube surgery, blepharoplasty, suture removal, and Nd:YAG laser capsulotomy. An additional 4 patients (5.5%) required surgery which was subsequently performed in Singapore—in 3 cases, this was following specific urgent medical requests to the Singapore Ministry of Health for them to fly to Singapore and perform their quarantine before surgery. These included a pediatric 6-year-old patient with congenital glaucoma and mesodermal dysgenesis and multiple failed grafts, who had artificial iris implantation in Singapore, followed by PK; an advanced Fuchs endothelial corneal dystrophy patient with painful bullous keratopathy, who then had DMEK performed in Singapore; and a Boston keratoprosthesis (type 1) patient who required vitrectomy and membrane peeling in Singapore for myopic retinoschisis in an only eye. All patients had successful outcomes up to the present time. One final patient fortuitously came to Singapore on a work visit pass and had Nd:YAG laser capsulotomy. Some operational limitations of the VCC program were experienced on occasion, mostly pertaining to issues with transient transmission difficulties, presumably related to high network usage on either side, but these were relatively rare occurrences, and no clinical sessions had to be canceled or postponed because of transmission difficulties.
Countries in the South-East Asia are currently at the epicenter of the resurgence of new waves of COVID-19 linked to the Delta and Omicron variants of COVID-19, with the worst affected country in the region being Indonesia. As of September 2021, the Singapore Ministry of Health has not rescinded the directive prohibiting foreign medical patients to travel to Singapore, and recently in early February 2022, the Minister of Health of Indonesia has announced a third wave of the pandemic in Indonesia due to the Omicron surge. The challenge for patients from high-risk COVID-19 countries to seek medical care overseas will thus continue, at least until the pandemic is controlled globally. Postkeratoplasty patients are at particular risk, partly because timely follow-up clinic visits are essential to long-term graft survival, and many corneal patients seek corneal transplants in neighboring countries and would routinely travel overseas for postoperative follow-up care, which has now been disrupted by COVID-19 and imposed travel restrictions. Our approach to mitigate the severe consequences of graft failure among our keratoplasty patients has been to develop a specific form of physician-to-physician teleophthalmology which best mimics or enables surrogate clinical evaluation by the corneal surgeon. The VCC program involves the development of a real-time video conferencing approach enabling synchronous bidirectional communication between physicians, patients, and even patients' relatives, with the real-time synchronous slit-lamp examination, between the patient's originating surgeon and the receiving specialist in the patient's host country, using audiovisual technology. In addition, the VCC concept also integrates a preceding asynchronous hybrid component, in which specific investigations are preordered and performed immediately before VCC video-consultation or performed after that visit in preparation for evaluation at the next VCC session. , One example of this approach is our ability to temporally track corneal endothelial cell status after DMEK, DSAEK, or PK which provides a significant prognostic value to physicians and patients. The additional hybrid component allows patients to be seen and managed by multiple subspecialties simultaneously applied in VCC which greatly enhances concurrent comorbidity management within a comprehensive subspecialist team approach. The concept of the VCC program also works for retinal examinations using the retinal lens and the slitlamp, or even using an indirect ophthalmoscope if it can be video linked. Video real-time simultaneous consultation from each individual subspecialty clinics, from either Singapore or Jakarta, also represents a significant time-saving advantage over conventional interreferrals, where patients necessarily have to attend different specialty clinics on the same day or on different days. This approach, although requiring significant planning and coordination, has the effect of greatly expediting clinical decision making, coupled with the value of real-time holistic conferencing of all subspecialists concerned with the patient's care. Studies on teleophthalmology tend to focus on telemedicine initiatives involving physician–patient interactions for patients without access to local ophthalmologists or, in this COVID-19 era, to avoid patients traveling to attend medical clinic appointments, and as such, there seem to be few examples in the literature of this form of real-time synchronous teleophthalmology for the anterior segment because most home-based clinical examinations would be highly limiting. – By contrast, our approach involves a physician-to-physician or clinic-to-clinic coconsultation for postsurgical management and long-term follow-up of keratoplasty patients. Although this teleophthalmology program was derived out of necessity in this current COVID-19 pandemic which restricted air travel between countries, our VCC concept essentially enables a teleophthalmology approach for close collaboration between non–co-located ophthalmology subspecialists, which could still be relevant beyond this COVID-19 crisis, and may even constitute part of the “new-norm” of clinical medicine. In the instance of corneal disease, many ophthalmologists refer corneal patients to corneal specialists who may be located in another state or region for consultation or surgery—if keratoplasty is performed, then the patient needs to either travel regularly for his postoperative follow-up care or rely on his general ophthalmologist who may be less familiar with managing keratoplasty patients. In this instance, repeat VCC management will enable such patients to continue to be followed up by their local ophthalmologist, but simultaneously enjoy coconsultation with their distant corneal surgeon. In our series, most of these postkeratoplasty patients were extremely complex, and hence the advantage of several corneal and other subspeciality surgeons discussing together to come to a joint management plan. Similar scenarios could be envisaged in the case of other complex or subspecialty conditions, where highly specialized subspecialist expertise would be desired. To cater to the relatively higher demands of a physician–physician teleophthalmology setup, where detailed subspecialist examination and investigations would be needed, it is clear that VCC transmission of real-time slit-lamp video examination, coupled with the ability to review EMR-based imaging investigations, and a synchronous or asynchronous prior hybrid adjunctive approach would be essential. The recent familiarity of current audiovisual telecommunication technologies, such as Zoom, the common cloud-based communications application, coupled with EMR-based ophthalmology clinic software and clinic equipment essentially transforms what would have been a 1 ophthalmologist clinic visit into a multispecialist real-time consultation also known as a hybrid approach. One obvious limitation of VCC, therefore, is the ability of both clinic locations to have the same relatively high level of cloud access and clinical technologies, which is probably available in most developed countries. The limitations of VCC include the need for good telecommunication links, good transmission capabilities of the transmitting clinic, and the reliance on a stable network as well as the potential for cybersecurity risks imposed by data transfer of medical information across networks. , However, we consider these limitations to be far outweighed by the much improved and time-sensitive patient care afforded to our keratoplasty patients, to say nothing of the significant reduction in travel costs to patients, and the benefit of multiphysician assessments. As it is, although Singapore is undergoing even more stringent local restrictions and quarantine orders due to the increasing number of Delta and Omicron variant COVID-19 cases, ECS has currently adopted a similar VCC concept as part of its contingency plans in case 1 or more of our physicians become COVID-19-positive and are forced to undergo home quarantine—using VCC will still enable our patients to be managed from home by our quarantined physician, with the assistance of his or her colleagues in the clinic. Finally, it is highly likely that our institutions will continue with VCC after this current COVID-19 pandemic abates as a new collaborative clinical approach to patient management.
|
Numbers, graphs and words – do we really understand the lab test results accessible via the patient portals? | 73870105-623e-48d7-8ce2-50b5c12045d1 | 7592036 | Health Communication[mh] | The heavy reliance on remote patient care (RPC) during the COVID-19 health crisis may have expedited the use of digital health tools to safely and effectively move the locus of care from the hospital to the community, and even the home . Enthusiasts see this as a long-overdue opportunity to reengineer care processes so as to reap the full benefits of health information technologies . Health information technologies (HIT) are broadly defined as “the electronic systems that health care providers and increasingly, patients, use to store, share and analyze information” . Direct patient use of test data is consistent with trends toward patient-centered care and the medical home concept, which aims to achieve greater patient involvement in both medical decision making and health self-management . These care approaches expect the “digitally engaged patient” to self-monitor and self-care for themselves and their families through the skillful use of enhanced digital technologies (see also ). Yet managing their own health has been shown to place heavy demands on laypersons, who are now expected to correctly interpret their test results, evaluate the pros and cons of different treatments, and decide on a preferred course of action . Correspondingly, clinicians have expressed concern that patients often experience great difficulty in comprehending, interpreting, and correctly responding to personalized health information, partly due to inappropriate presentation of the information in patient portals . In particular, misunderstanding test results leads to confusion, frustration, and disruptions in healthcare processes, including delays in seeking care, overutilization of services, medication errors, and inappropriate healthcare decision-making . A stream of evidence shows that laypersons differ from experts in how they assess the meaning of healthcare information and evaluate its trustworthiness . For instance, in the case of antenatal screening tests, practitioners tend to frame risk information numerically, as the probability of a genetic condition, even though laypersons display a better understanding of the information when a verbal or narrative format is used . Yet patients are more likely to take up genetic testing when presented with numeric risk information . Compared with medical professionals, laypersons are more readily influenced by the attractiveness of a site’s design and have even been found to reject high-quality content because of poor visual design, confusing displays, and a low density of relevant information . Since laypersons can and often do view the results of their check-ups and medical tests prior to interacting with their healthcare providers, the meaning they attach to these results can significantly affect their decision-making and subsequent follow-up care. Recent definitions of e-health literacy consider the set of individual capacities that allow the person to acquire and use new information, as well as the cognitive competencies required to make judgments and decisions in everyday life concerning health . E-health literacy has been shown to combine knowledge and skills from a wide variety of domains, and is inherently contingent upon the social contexts wherein it is developed and expected to be put into use. As a result, it can be affected by a variety of personal and socio-demographic factors, including age, gender, education, acute and chronic health concerns, general health literacy, and technological proficiency . However, it is still difficult to assess whether people’s interpretation of information such as test results is more affected by these socio-demographic variables, or by how the information is displayed. Thus, the first aim of the present study is to help address this gap by examining respondents’ interpretation of medical information while controlling for various demographic characteristics (age, education level, socio-demographic status, health status, attitudes towards self-care, and responsiveness to medical or health recommendations). The present research is also motivated by the fact that despite growing interest in laypersons’ comprehension and use of medical knowledge to reach appropriate medical decisions, little is still known of the extent to which different visual displays help people discriminate between test results that do or do not require urgent action . Zikmund-Fisher et al. asked their participants to imagine that they were viewing the results of a set of blood tests on an online patient portal. They varied the format in which participants viewed these test results – as a number line graph, a table, or raw numbers – and tested the relationship between the display format and respondents’ perceptions of urgency and inclination to contact health care providers. They found that when test results were abnormal (indicated extreme values), perceived urgency was universally high, regardless of which display format was shown. However, perceptions regarding near-normal values varied substantially across formats. The pattern they found was consistent: participants who saw their near-normal values in a tabular display rated those results as most urgent, while those who saw a gradient line display perceived the results as least urgent. The authors attributed these differences in interpretation to both the amount of information conveyed by each format, and to the cognitive skills each format requires to deduce the bottom-line implications of the information. Thus, the second aim of the present study is to test whether the findings of Zikmund-Fisher et al. are conceptually replicated in a different sample and study design. Following Zikmund-Fisher et al. , we examine three aspects of knowing related to the interpretation of lab test results in patient health portals: knowing, uncertainty, and accuracy. Knowing is defined here according to the standard dictionary definition, as a perception of clear and certain mental apprehension. Uncertainty is defined as a cognitive claim of insufficient knowledge, or a lack of understanding regarding the meaning of the information presented . Accuracy is defined as a condition or quality of being correct or exact with relation to a standard, again based on the standard dictionary definition. More concretely, we examine how different displays of information were related to different interpretations of severity, “do not know” answers and inaccuracies in judgment. We propose three hypotheses: H1: The three information formats (verbal, numeric, and graphic) will differ in the accuracy of participants’ assessments. H2: The three information formats (verbal, numeric, and graphic) will yield different levels of uncertainty with regard to the condition’s level of gravity. H3: The higher the perceived gravity of the health condition, the more proactive people are likely to be in seeking help or information. The conceptual model which forms the basis for the hypotheses is displayed in Fig. .
Design and sample We employed a survey to assess respondents’ reactions to 10 medical decision-making scenarios, where the same information was presented using different formats. In each scenario, respondents were presented with real (anonymized) patient lab results using either numeric expressions, graphs, or verbal expressions. Participants were asked to assess the gravity of the hypothetical patient’s condition and the course of action they would follow if they were that patient. Participants were recruited through convenience sampling, mainly via the authors’ networks. This sampling method is commonly employed in healthcare-related surveys . Each respondent was asked to virally distribute the link to others in their network (snowball sampling). The link was operative for a period of 2 weeks, and we monitored the response rate daily. By the end of the data collection period the link had been distributed to over 300 individuals. We used all valid responses obtained. That is, rather than trimming or imputing, we worked with different sample sizes for each analysis, calculating the means for each respondent where necessary. In total, 225 participants returned questionnaires suitable for analysis, meaning they responded to at least some of the scenarios, and 220 returned fully complete questionnaires. Approximately 83% of those who began the survey (i.e., who completed the demographic questions) submitted usable answers – a satisfactory percentage, given that the questionnaire was relatively long and contained 10 different scenarios. Figure describes our missing values policy in detail. Missing data analysis revealed that missing values were random for most variables A Pearson Chi-Square Test showed that those who did not complete the questionnaire differed from those who did in three demographic variables: income (X2 (3, N = 270) =15.09 value, p = 0.002), and family status (X2 (2, N = 270) =18.386 value, p < 0.002). An independent t-test revealed that those who did not complete the questionnaire were significantly older (M = 41.87; std. = 12.144) than those who did (M = 35.19, std. = 13.781) F = 2.459, P = 0.003). These differences are consistent, meaning that those who did not provide usable responses were significantly older, had slightly higher incomes, and were married. No differences were found in health status. Finally, we applied a binomial test of equal proportions or two-proportion z-test to determine the minimum required sample size. In this study, a random sample of 43 pairs (where the mean difference is 0.22 and the standard deviation of the difference is 0.5) would allow us to declare with 80% power that the mean of the paired differences is significantly different from zero (i.e., a two sided p -value is less than 0.05). The sample size of this study is 225, 114 for version A and 109 for version B – comfortably above 43 pairs (see Table ). Participants were randomly assigned to receive one of two versions of the questionnaire. Both versions contained the same scenarios and questions, differing only in the presentation of the test results. Of the 225 surveys which contained usable responses, 114 represented the first version of the survey, and 111 the second. For each scenario, two of the three formats were contrasted between the two versions (i.e., sometimes verbal vs. graphic, sometimes graphic vs. numeric, etc.; see under “Procedure and materials” below). An independent-samples t-test showed no differences in demographic measures between respondents who received the two versions of the questionnaire (see Table ). Procedure and materials Using Qualtrics software we produced anonymized links to the two versions of the survey, each containing 10 medical decision-making scenarios. The order of the scenarios was the same in both versions. In each scenario, respondents were presented with lab results relating to important but non-life-threatening health conditions using either numeric expressions, graphs, or verbal expressions. Participants were asked to assess the gravity of the hypothetical patient’s condition and the course of action they would follow if they were that patient. The lab tests (blood work or cultures) were extracted from authentic patient portals, and had been ordered to investigate or test for one of the following: erratic menstrual cycles; low hemoglobin; hepatitis B; streptococcus (a throat infection); or a routine cholesterol check. These conditions and tests were chosen because they are relatively common, likely to be only moderately serious, and in most cases potentially applicable to both men and women. Each scenario contained a short description of the patient’s symptoms and the possible consequences of poor treatment. Two physicians, both of them general practitioners working both in the community and hospitals, independently reviewed the scenarios and confirmed the accuracy and reliability of the information provided. An example of one scenario is shown in Fig. . After reading the scenarios and viewing the results, participants were asked to assess the gravity of the condition, and what course of action they would recommend for the patient. Our aim was not to test our participants’ medical proficiency or knowledge, but solely their interpretation of the lab results presented to them. For each scenario, the test results were presented in two of the three different formats (numeric, graphic, or verbal), one in version A and a different one in version B (see Fig. ). Presentations in the verbal format contained either a diagnosis, or a short explanation and/or recommendation. The numeric format contained a single measure or a series of measures presented in table form, sometimes with an indication of a norm. The graphic format contained a line graph showing the current measurement, previous measurements, and an indication of the norm. Measures Knowing. After reading each scenario, participants were asked to assess the gravity of the hypothetical health condition, based on the results of the lab tests provided. Perceived gravity ranged from 1 = very low to 5 = very high, with 6 = don’t know. Knowing was operationalized as any response from 1 to 5 (as opposed to 6, meaning “don’t know”). Uncertainty was operationalized as a choice of the sixth option in the gravity scale, namely “don’t know.” A choice of the “don’t know” response indicated that the participant had difficulty interpreting the lab results. Accuracy, the quality of being correct with respect to a standard, was also measured in relation to responses on the gravity scale, and was operationalized by comparing all responses of 1 to 5 against the physicians’ assessment of the gravity of the health condition. We were thus able to assess who underestimated the gravity of the condition, who correctly assessed the gravity of the condition, and who overestimated the gravity of the condition. This variable ranged from − 0.35 to 3.75. We then turned this into a categorical variable named level of accuracy (> 0 = 1; 0 = 2; < 0 = 3). Preferred course of action For each scenario, we asked participants to indicate how likely they would be, if they were the patient, to do each of the following upon seeing the presented lab results: 1. immediately contact their doctor; 2. search out more information on the Internet; 3. wait for their physician to contact them; and 4. wait until their next visit to the doctor to verify the meaning of the results. For each course of action, respondents were asked their likelihood of taking that path on a 5-point scale, where 1 = very unlikely and 5 = very likely. Demographics and controls Demographic measures collected were age, gender, HMO membership, income, education, religiosity, family status, and country of birth (see Table ). To offset differential response rates by age, we divided the respondents into three distinct age samples (18–39, 40–59, and 60 and older). Income was measured on a three-point scale (“The average income is 7500 NIS [about $2500 a month]. Is your income higher than, equal to, or less than 7500 a month?”). We also controlled for use of EPR systems, respondents’ self-reported health status, attitudes towards self-care, and responsiveness to medical or health recommendations (details on these measures are given under Results below). We carried out reliability analyses on the scales assessing participants’ level of EPR use (8 items) and beliefs about health and healthcare (7 items). Both reached acceptable reliabilities, α = 0.88 and α = 0.620 for the EPR and health beliefs scales, respectively.
We employed a survey to assess respondents’ reactions to 10 medical decision-making scenarios, where the same information was presented using different formats. In each scenario, respondents were presented with real (anonymized) patient lab results using either numeric expressions, graphs, or verbal expressions. Participants were asked to assess the gravity of the hypothetical patient’s condition and the course of action they would follow if they were that patient. Participants were recruited through convenience sampling, mainly via the authors’ networks. This sampling method is commonly employed in healthcare-related surveys . Each respondent was asked to virally distribute the link to others in their network (snowball sampling). The link was operative for a period of 2 weeks, and we monitored the response rate daily. By the end of the data collection period the link had been distributed to over 300 individuals. We used all valid responses obtained. That is, rather than trimming or imputing, we worked with different sample sizes for each analysis, calculating the means for each respondent where necessary. In total, 225 participants returned questionnaires suitable for analysis, meaning they responded to at least some of the scenarios, and 220 returned fully complete questionnaires. Approximately 83% of those who began the survey (i.e., who completed the demographic questions) submitted usable answers – a satisfactory percentage, given that the questionnaire was relatively long and contained 10 different scenarios. Figure describes our missing values policy in detail. Missing data analysis revealed that missing values were random for most variables A Pearson Chi-Square Test showed that those who did not complete the questionnaire differed from those who did in three demographic variables: income (X2 (3, N = 270) =15.09 value, p = 0.002), and family status (X2 (2, N = 270) =18.386 value, p < 0.002). An independent t-test revealed that those who did not complete the questionnaire were significantly older (M = 41.87; std. = 12.144) than those who did (M = 35.19, std. = 13.781) F = 2.459, P = 0.003). These differences are consistent, meaning that those who did not provide usable responses were significantly older, had slightly higher incomes, and were married. No differences were found in health status. Finally, we applied a binomial test of equal proportions or two-proportion z-test to determine the minimum required sample size. In this study, a random sample of 43 pairs (where the mean difference is 0.22 and the standard deviation of the difference is 0.5) would allow us to declare with 80% power that the mean of the paired differences is significantly different from zero (i.e., a two sided p -value is less than 0.05). The sample size of this study is 225, 114 for version A and 109 for version B – comfortably above 43 pairs (see Table ). Participants were randomly assigned to receive one of two versions of the questionnaire. Both versions contained the same scenarios and questions, differing only in the presentation of the test results. Of the 225 surveys which contained usable responses, 114 represented the first version of the survey, and 111 the second. For each scenario, two of the three formats were contrasted between the two versions (i.e., sometimes verbal vs. graphic, sometimes graphic vs. numeric, etc.; see under “Procedure and materials” below). An independent-samples t-test showed no differences in demographic measures between respondents who received the two versions of the questionnaire (see Table ).
Using Qualtrics software we produced anonymized links to the two versions of the survey, each containing 10 medical decision-making scenarios. The order of the scenarios was the same in both versions. In each scenario, respondents were presented with lab results relating to important but non-life-threatening health conditions using either numeric expressions, graphs, or verbal expressions. Participants were asked to assess the gravity of the hypothetical patient’s condition and the course of action they would follow if they were that patient. The lab tests (blood work or cultures) were extracted from authentic patient portals, and had been ordered to investigate or test for one of the following: erratic menstrual cycles; low hemoglobin; hepatitis B; streptococcus (a throat infection); or a routine cholesterol check. These conditions and tests were chosen because they are relatively common, likely to be only moderately serious, and in most cases potentially applicable to both men and women. Each scenario contained a short description of the patient’s symptoms and the possible consequences of poor treatment. Two physicians, both of them general practitioners working both in the community and hospitals, independently reviewed the scenarios and confirmed the accuracy and reliability of the information provided. An example of one scenario is shown in Fig. . After reading the scenarios and viewing the results, participants were asked to assess the gravity of the condition, and what course of action they would recommend for the patient. Our aim was not to test our participants’ medical proficiency or knowledge, but solely their interpretation of the lab results presented to them. For each scenario, the test results were presented in two of the three different formats (numeric, graphic, or verbal), one in version A and a different one in version B (see Fig. ). Presentations in the verbal format contained either a diagnosis, or a short explanation and/or recommendation. The numeric format contained a single measure or a series of measures presented in table form, sometimes with an indication of a norm. The graphic format contained a line graph showing the current measurement, previous measurements, and an indication of the norm.
Knowing. After reading each scenario, participants were asked to assess the gravity of the hypothetical health condition, based on the results of the lab tests provided. Perceived gravity ranged from 1 = very low to 5 = very high, with 6 = don’t know. Knowing was operationalized as any response from 1 to 5 (as opposed to 6, meaning “don’t know”). Uncertainty was operationalized as a choice of the sixth option in the gravity scale, namely “don’t know.” A choice of the “don’t know” response indicated that the participant had difficulty interpreting the lab results. Accuracy, the quality of being correct with respect to a standard, was also measured in relation to responses on the gravity scale, and was operationalized by comparing all responses of 1 to 5 against the physicians’ assessment of the gravity of the health condition. We were thus able to assess who underestimated the gravity of the condition, who correctly assessed the gravity of the condition, and who overestimated the gravity of the condition. This variable ranged from − 0.35 to 3.75. We then turned this into a categorical variable named level of accuracy (> 0 = 1; 0 = 2; < 0 = 3). Preferred course of action For each scenario, we asked participants to indicate how likely they would be, if they were the patient, to do each of the following upon seeing the presented lab results: 1. immediately contact their doctor; 2. search out more information on the Internet; 3. wait for their physician to contact them; and 4. wait until their next visit to the doctor to verify the meaning of the results. For each course of action, respondents were asked their likelihood of taking that path on a 5-point scale, where 1 = very unlikely and 5 = very likely. Demographics and controls Demographic measures collected were age, gender, HMO membership, income, education, religiosity, family status, and country of birth (see Table ). To offset differential response rates by age, we divided the respondents into three distinct age samples (18–39, 40–59, and 60 and older). Income was measured on a three-point scale (“The average income is 7500 NIS [about $2500 a month]. Is your income higher than, equal to, or less than 7500 a month?”). We also controlled for use of EPR systems, respondents’ self-reported health status, attitudes towards self-care, and responsiveness to medical or health recommendations (details on these measures are given under Results below). We carried out reliability analyses on the scales assessing participants’ level of EPR use (8 items) and beliefs about health and healthcare (7 items). Both reached acceptable reliabilities, α = 0.88 and α = 0.620 for the EPR and health beliefs scales, respectively.
For each scenario, we asked participants to indicate how likely they would be, if they were the patient, to do each of the following upon seeing the presented lab results: 1. immediately contact their doctor; 2. search out more information on the Internet; 3. wait for their physician to contact them; and 4. wait until their next visit to the doctor to verify the meaning of the results. For each course of action, respondents were asked their likelihood of taking that path on a 5-point scale, where 1 = very unlikely and 5 = very likely.
Demographic measures collected were age, gender, HMO membership, income, education, religiosity, family status, and country of birth (see Table ). To offset differential response rates by age, we divided the respondents into three distinct age samples (18–39, 40–59, and 60 and older). Income was measured on a three-point scale (“The average income is 7500 NIS [about $2500 a month]. Is your income higher than, equal to, or less than 7500 a month?”). We also controlled for use of EPR systems, respondents’ self-reported health status, attitudes towards self-care, and responsiveness to medical or health recommendations (details on these measures are given under Results below). We carried out reliability analyses on the scales assessing participants’ level of EPR use (8 items) and beliefs about health and healthcare (7 items). Both reached acceptable reliabilities, α = 0.88 and α = 0.620 for the EPR and health beliefs scales, respectively.
Descriptive statistics – demographics, health status, health behaviors, and EPR use Table presents the descriptive statistics of the final sample. As the table shows, the final sample was fairly heterogeneous in its socio-demographic characteristics. With respect to HMO membership, the proportions represented in our sample resemble the proportions in the Israeli population as a whole, with a slight over-representation for one HMO, Maccabi (Clalit = 53.967% vs. 47% in sample; Maccabi = 26.007% vs. 35.4%; Leumit = 7.772% vs. 7.5; Meuhedet = 12.254 vs. 10.1; all data from Social security 2020). As for education, our sample has a relatively high proportion of educated participants. However, it should be noted that 50% of all Israeli citizens aged 25–64 have either tertiary or academic education . Our sample underrepresents Haredi and religious participants, as well as other minority groups. The limitations of the chosen sampling method will be discussed later in the paper. Based on the full sample, 5% of our respondents claimed to be in poor health, while 88% reported being in good, very good or excellent health. Fifty-five respondents (18%) reported suffering from a chronic illness, and 22 respondents (5%) reported suffering from some type of physical limitation. Being in good or very good health was negatively correlated with age (r [255] = − 0.240, P < 0.01). Age was positively correlated with feeling responsible for one’s health (r [213] = 0.139, P < 0.001), and negatively correlated with postponing regular checkups (r [213] = − 0.162, P < 0.001). Most indicated that they felt responsible for keeping healthy (M = 4.85, std. = 0.468, on a 5-point scale where 1 = strongly disagree, 5 = strongly agree) and that maintaining a healthy lifestyle was important to them (M = 4.69, std. = 0.57; 1 = strongly disagree, 5 = strongly agree). They also reported generally complying with their doctor’s recommended regime (M = 4.36, std. = 0.844). When asked how they respond when they feel sick, 50% of the respondents who answered this question ( N = 130) said they turn to their doctor for a consultation, while 25% ( N = 65) turn to a family member, 22.7% ( N = 59) consult medical websites for information, and only 2.3% ( N = 6) consult online health forums. With respect to EPR use, 71% of our participants ( N = 173) reported that they frequently access their lab results via the EPR. Ten percent ( N = 30) claimed to have never viewed their lab results via the EPR, and an additional 7.7% ( N = 23) of our participants were not aware of being able to view their lab results via the EPR. In general, women tend to use the EPR significantly more than men (t [216] = − 3.6, P < 0.001). However, these differences disappear when focusing on use of the EPR to view lab results and health recommendations (i.e., women more than men use the EPR for administrative purposes such as scheduling doctors’ appointments and filing requests for prescription drugs for themselves and other family members). There was a significant main effect of age on EPR use, F (2, 246) = 4.718, P < 0.000. Participants aged 18–39 were significantly less inclined to consult the EPR than those aged 40–59 and those aged 60+ ( P < 0.000). Hypothesis testing H1: The three information formats (verbal, numeric, and graphic) will differ in the accuracy of participants’ assessments. In general, both the participants and the physicians interpreted the conditions as mildly serious or not very serious. A Wilcoxon signed-rank test showed that physicians’ assessments of gravity were significantly lower than those of the laypersons (Z = − -2.828, p = 0.005). A follow-up Pearson chi-square test confirmed these differences (chi-square, 7.222, df = 2, p < 0.027). The results suggest that the participants were fairly accurate in the general trend, but tended to overestimate the conditions’ gravity in all three formats (See Fig. - Participants’ and Experts’ assessments of gravity, for each information format). Looking dipper, a paired sample t-test revealed that accuracy is higher when results are explained verbally, rather than having a number stand on its own (see Table ). Accuracy is greater when a numeric value appears in tabular form, as opposed to only a value. And overall, accuracy is greater when results are presented in a tabular form, rather than in a line graph, even though both represent deviations from the norm. Finally, we examined whether accuracy can be explained by demographic variables or participants’ general health status and familiarity with EPR use. Results of a multiple linear regression to predict level of accuracy point to a collective significant effect of gender, age, education, health status, income, EPR use, and uncertainty (F (7,201) = 24.442, p < .001, R2 = .460). However, only age (Beta = .186; t = − 3.063, p = .002), and uncertainty (Beta = −.285; t = − 5.437, P < 0.000) were significant predictors in the model. A one-way ANOVA revealed differences between the three age groups with regard to accuracy (F (2,208) =14.455, p = .000). A Tukey post-hoc test revealed that accuracy was significantly lower among those aged 18–39 (M = 0.59; std. = 0.8; p = 0.00) than among those aged 40–59 (M = 0.89, std. = 1.16, p = 0.00) and those aged 60+ (M = 1.26, std. = 1.26, p = 0.000). No significant differences were found between those aged 40–59 and those 60 years old and older. These findings suggest that age-related familiarity with different health conditions could be related to accuracy. H2: The three information formats (verbal, numeric, and graphic) will yield different levels of uncertainty with regards to being able to assess the condition’s level of severity. We measured the proportion of respondents who chose the “don’t know” response for any of the 10 scenarios. Slightly more than half (50.7%) chose the “don’t know” option at least once. Of those, only 15% chose the “don’t know” response in more than eight scenarios. These findings indicate that the “don’t know” option was generally not chosen automatically, and without reflection. Figure shows the percentage of respondents who chose the “don’t know” response in each of the 10 scenarios. Figure demonstrates the proportion of “don’t know” responses for each scenario, and level of accuracy. The graph reveals no discernable association between either “don’t know” responses or accuracy and health condition, leading us to believe that the display of information plays an important role in both. The only visible exception relates to levels of progesterone (scenarios 1 and 3). In both scenarios the rate of “don’t know” responses is relatively high, and the level of accuracy is relatively low, compared to all other scenarios. We hypothesized that gender and age could explain these findings, and conducted two separate two-way ANOVAs to examine the effect of gender and age on accuracy and on “don’t know” responses in those two scenarios. However, no statistically significant main effects nor interaction were found. More generally, we performed a hierarchical linear regression to predict the level of “don’t know” responses based on various demographic variables (age, gender, education, family status, and income), along with EPR use and health status. Variables were entered into the equation using the stepwise method. These variables explained a relatively small proportion of variance in uncertainty scores (“don’t know” responses). In the first model, R 2 = .021, F (1,244) =6.214, p < .001. In the second model, R 2 = .035, F (1,244) =5.92, p < .001. In the first model, income alone significantly predicted uncertainty scores, B = 0.512, t (244) = 6.53, p < .001. In the second model, both income, B = 0.512, t (244) = 6.53, p < .001, and gender, B = 0.512, t (244) = 6.53, p < .001, significantly predicted uncertainty scores. We then conducted a hierarchical linear regression to predict the level of accuracy based on demographic variables (age, gender, education, and income), along with EPR use and health status. Variables were entered into the equation using the stepwise method. In the first model, income significantly predicted accuracy, B = .228, t (205) = 3.196, R2 = .107, F (1, 205) = 2.791, p < .001. However, once age was entered into the equation, income was no longer significant. In the second model, age alone significantly predicted accuracy, B = .326, t (205) = 3.233, R2 = .0.92, F (1, 205) = 20.735, p < .001. These findings suggest that women more than men, and those of higher versus lower income, indicated that they could not assess the conditions’ gravity based on the information displayed. However, those who did were more accurate than those who were younger and of lesser means. Gender had no effect on accuracy, suggesting that women were more comfortable indicating that they were unsure of the answer than the men participating in the study. We then performed a one-way between-subjects ANOVA to compare the effect of information format on “don’t know” responses. We found a significant effect of format type on “don’t know” responses, F (2,17) = 9.789, p = .001. Post hoc comparisons using the Tukey HSD test show that the mean score for the graph condition (M = 5.06, std. = 2.17) is significantly lower than that for the numeric condition (M = 15.90, std. = 4.05), and also lower than the mean score for the verbal condition (M = 9.87, std. = 5.33). On average, the numeric format produced the highest number of “don’t know” responses and the graphic format the least, indicating that respondents found the numeric format most difficult to understand and the graphic format the easiest. These findings confirmed our hypothesis that the three information formats differ in the ease with which they were understood. Yet as reported earlier, those participants who assessed the gravity of the health conditions were slightly but significantly more accurate when results appeared in a table, than in the line graph format, even though in both appeared a scale showing normal and abnormal results (see Fig. ). H3: The higher the perceived gravity of the health condition, the more proactive people are likely to be in seeking help or information. First, we performed a linear regression to predict the level of proactivity based on various demographic variables (age, sex, education, family status, having children under the age of 18, and income), and perceived gravity of the health condition. Variables were entered into the equation using the stepwise method, starting with perceived gravity and then adding the control variables one by one. Two models were found significant. In the first model, the only independent variable to predict level of proactivity was perceived gravity of the health condition, F (4,207) = 15.901, P < 000, R 2 = .73. In the second model, income was found significant in addition to gravity, F (2,201) = 9.731, P < 001, R 2 = .99. We conducted a one-way between-subjects ANOVA to compare the effect of information format on prefered course of action. Interestingly, we found that information format had a significant effect only on respondents’ tendency to choose “search the Internet” as a preferred course of action, F (2,16) = 3.159, p < .0.05. In post hoc comparisons using the LSD test, the mean score for the numeric condition (M = 62.58, std. = 8.43) was significantly higher than for the verbal condition (M = 54.8, std. = 12.69). No significant difference was found between the numeric condition and the graph condition (M = 69.71, std. = 6.55). This finding is congruent with our earlier finding that the numeric format produced the highest number of “don’t know” responses and the graphic format the least, suggesting that respondents found the numeric format hardest to understand and the graphic format the easiest. We assume that our respondents expected that searching the Internet would clarify the situation. Interestingly, however, the graphic presentation, which supposedly offers the greatest amount of contextualized information, also precipitated relatively high scores for Internet search, perhaps because of its complexity. Finally, we expected that those who did not understand the information (“don’t know” responses) would favor more proactive measures, defined as immediately contacting their doctor or searching for information on the Internet. A Pearson correlation revealed an inverse relationship between “don’t know” responses and participants’ tendency to call a doctor (r = − 0.184, p < 001), and a positive relationship between “don’t know” and the other three courses of action: searching the Internet (r = 0.438, P < 001), waiting for the doctor to call them (r = 0.442, P < .005), or waiting for their next visit to the doctor (r = 0.488, P < .005). Thus, the less understandable the information presented, the less likely the participants were to immediately call their family doctors. Rather, they were more likely to search the Internet for information, wait for their next doctor’s appointment, or wait for their doctor to contact them. Consulting the Internet for information was positively correlated with waiting for the doctor to call and waiting for one’s next visit to the doctor (r [298] = 0.482, P < 0.001; r [298] = 0.467, P < 0.001, respectively). Thus, our hypothesis was not supported (see Table ).
Table presents the descriptive statistics of the final sample. As the table shows, the final sample was fairly heterogeneous in its socio-demographic characteristics. With respect to HMO membership, the proportions represented in our sample resemble the proportions in the Israeli population as a whole, with a slight over-representation for one HMO, Maccabi (Clalit = 53.967% vs. 47% in sample; Maccabi = 26.007% vs. 35.4%; Leumit = 7.772% vs. 7.5; Meuhedet = 12.254 vs. 10.1; all data from Social security 2020). As for education, our sample has a relatively high proportion of educated participants. However, it should be noted that 50% of all Israeli citizens aged 25–64 have either tertiary or academic education . Our sample underrepresents Haredi and religious participants, as well as other minority groups. The limitations of the chosen sampling method will be discussed later in the paper. Based on the full sample, 5% of our respondents claimed to be in poor health, while 88% reported being in good, very good or excellent health. Fifty-five respondents (18%) reported suffering from a chronic illness, and 22 respondents (5%) reported suffering from some type of physical limitation. Being in good or very good health was negatively correlated with age (r [255] = − 0.240, P < 0.01). Age was positively correlated with feeling responsible for one’s health (r [213] = 0.139, P < 0.001), and negatively correlated with postponing regular checkups (r [213] = − 0.162, P < 0.001). Most indicated that they felt responsible for keeping healthy (M = 4.85, std. = 0.468, on a 5-point scale where 1 = strongly disagree, 5 = strongly agree) and that maintaining a healthy lifestyle was important to them (M = 4.69, std. = 0.57; 1 = strongly disagree, 5 = strongly agree). They also reported generally complying with their doctor’s recommended regime (M = 4.36, std. = 0.844). When asked how they respond when they feel sick, 50% of the respondents who answered this question ( N = 130) said they turn to their doctor for a consultation, while 25% ( N = 65) turn to a family member, 22.7% ( N = 59) consult medical websites for information, and only 2.3% ( N = 6) consult online health forums. With respect to EPR use, 71% of our participants ( N = 173) reported that they frequently access their lab results via the EPR. Ten percent ( N = 30) claimed to have never viewed their lab results via the EPR, and an additional 7.7% ( N = 23) of our participants were not aware of being able to view their lab results via the EPR. In general, women tend to use the EPR significantly more than men (t [216] = − 3.6, P < 0.001). However, these differences disappear when focusing on use of the EPR to view lab results and health recommendations (i.e., women more than men use the EPR for administrative purposes such as scheduling doctors’ appointments and filing requests for prescription drugs for themselves and other family members). There was a significant main effect of age on EPR use, F (2, 246) = 4.718, P < 0.000. Participants aged 18–39 were significantly less inclined to consult the EPR than those aged 40–59 and those aged 60+ ( P < 0.000).
H1: The three information formats (verbal, numeric, and graphic) will differ in the accuracy of participants’ assessments. In general, both the participants and the physicians interpreted the conditions as mildly serious or not very serious. A Wilcoxon signed-rank test showed that physicians’ assessments of gravity were significantly lower than those of the laypersons (Z = − -2.828, p = 0.005). A follow-up Pearson chi-square test confirmed these differences (chi-square, 7.222, df = 2, p < 0.027). The results suggest that the participants were fairly accurate in the general trend, but tended to overestimate the conditions’ gravity in all three formats (See Fig. - Participants’ and Experts’ assessments of gravity, for each information format). Looking dipper, a paired sample t-test revealed that accuracy is higher when results are explained verbally, rather than having a number stand on its own (see Table ). Accuracy is greater when a numeric value appears in tabular form, as opposed to only a value. And overall, accuracy is greater when results are presented in a tabular form, rather than in a line graph, even though both represent deviations from the norm. Finally, we examined whether accuracy can be explained by demographic variables or participants’ general health status and familiarity with EPR use. Results of a multiple linear regression to predict level of accuracy point to a collective significant effect of gender, age, education, health status, income, EPR use, and uncertainty (F (7,201) = 24.442, p < .001, R2 = .460). However, only age (Beta = .186; t = − 3.063, p = .002), and uncertainty (Beta = −.285; t = − 5.437, P < 0.000) were significant predictors in the model. A one-way ANOVA revealed differences between the three age groups with regard to accuracy (F (2,208) =14.455, p = .000). A Tukey post-hoc test revealed that accuracy was significantly lower among those aged 18–39 (M = 0.59; std. = 0.8; p = 0.00) than among those aged 40–59 (M = 0.89, std. = 1.16, p = 0.00) and those aged 60+ (M = 1.26, std. = 1.26, p = 0.000). No significant differences were found between those aged 40–59 and those 60 years old and older. These findings suggest that age-related familiarity with different health conditions could be related to accuracy. H2: The three information formats (verbal, numeric, and graphic) will yield different levels of uncertainty with regards to being able to assess the condition’s level of severity. We measured the proportion of respondents who chose the “don’t know” response for any of the 10 scenarios. Slightly more than half (50.7%) chose the “don’t know” option at least once. Of those, only 15% chose the “don’t know” response in more than eight scenarios. These findings indicate that the “don’t know” option was generally not chosen automatically, and without reflection. Figure shows the percentage of respondents who chose the “don’t know” response in each of the 10 scenarios. Figure demonstrates the proportion of “don’t know” responses for each scenario, and level of accuracy. The graph reveals no discernable association between either “don’t know” responses or accuracy and health condition, leading us to believe that the display of information plays an important role in both. The only visible exception relates to levels of progesterone (scenarios 1 and 3). In both scenarios the rate of “don’t know” responses is relatively high, and the level of accuracy is relatively low, compared to all other scenarios. We hypothesized that gender and age could explain these findings, and conducted two separate two-way ANOVAs to examine the effect of gender and age on accuracy and on “don’t know” responses in those two scenarios. However, no statistically significant main effects nor interaction were found. More generally, we performed a hierarchical linear regression to predict the level of “don’t know” responses based on various demographic variables (age, gender, education, family status, and income), along with EPR use and health status. Variables were entered into the equation using the stepwise method. These variables explained a relatively small proportion of variance in uncertainty scores (“don’t know” responses). In the first model, R 2 = .021, F (1,244) =6.214, p < .001. In the second model, R 2 = .035, F (1,244) =5.92, p < .001. In the first model, income alone significantly predicted uncertainty scores, B = 0.512, t (244) = 6.53, p < .001. In the second model, both income, B = 0.512, t (244) = 6.53, p < .001, and gender, B = 0.512, t (244) = 6.53, p < .001, significantly predicted uncertainty scores. We then conducted a hierarchical linear regression to predict the level of accuracy based on demographic variables (age, gender, education, and income), along with EPR use and health status. Variables were entered into the equation using the stepwise method. In the first model, income significantly predicted accuracy, B = .228, t (205) = 3.196, R2 = .107, F (1, 205) = 2.791, p < .001. However, once age was entered into the equation, income was no longer significant. In the second model, age alone significantly predicted accuracy, B = .326, t (205) = 3.233, R2 = .0.92, F (1, 205) = 20.735, p < .001. These findings suggest that women more than men, and those of higher versus lower income, indicated that they could not assess the conditions’ gravity based on the information displayed. However, those who did were more accurate than those who were younger and of lesser means. Gender had no effect on accuracy, suggesting that women were more comfortable indicating that they were unsure of the answer than the men participating in the study. We then performed a one-way between-subjects ANOVA to compare the effect of information format on “don’t know” responses. We found a significant effect of format type on “don’t know” responses, F (2,17) = 9.789, p = .001. Post hoc comparisons using the Tukey HSD test show that the mean score for the graph condition (M = 5.06, std. = 2.17) is significantly lower than that for the numeric condition (M = 15.90, std. = 4.05), and also lower than the mean score for the verbal condition (M = 9.87, std. = 5.33). On average, the numeric format produced the highest number of “don’t know” responses and the graphic format the least, indicating that respondents found the numeric format most difficult to understand and the graphic format the easiest. These findings confirmed our hypothesis that the three information formats differ in the ease with which they were understood. Yet as reported earlier, those participants who assessed the gravity of the health conditions were slightly but significantly more accurate when results appeared in a table, than in the line graph format, even though in both appeared a scale showing normal and abnormal results (see Fig. ). H3: The higher the perceived gravity of the health condition, the more proactive people are likely to be in seeking help or information. First, we performed a linear regression to predict the level of proactivity based on various demographic variables (age, sex, education, family status, having children under the age of 18, and income), and perceived gravity of the health condition. Variables were entered into the equation using the stepwise method, starting with perceived gravity and then adding the control variables one by one. Two models were found significant. In the first model, the only independent variable to predict level of proactivity was perceived gravity of the health condition, F (4,207) = 15.901, P < 000, R 2 = .73. In the second model, income was found significant in addition to gravity, F (2,201) = 9.731, P < 001, R 2 = .99. We conducted a one-way between-subjects ANOVA to compare the effect of information format on prefered course of action. Interestingly, we found that information format had a significant effect only on respondents’ tendency to choose “search the Internet” as a preferred course of action, F (2,16) = 3.159, p < .0.05. In post hoc comparisons using the LSD test, the mean score for the numeric condition (M = 62.58, std. = 8.43) was significantly higher than for the verbal condition (M = 54.8, std. = 12.69). No significant difference was found between the numeric condition and the graph condition (M = 69.71, std. = 6.55). This finding is congruent with our earlier finding that the numeric format produced the highest number of “don’t know” responses and the graphic format the least, suggesting that respondents found the numeric format hardest to understand and the graphic format the easiest. We assume that our respondents expected that searching the Internet would clarify the situation. Interestingly, however, the graphic presentation, which supposedly offers the greatest amount of contextualized information, also precipitated relatively high scores for Internet search, perhaps because of its complexity. Finally, we expected that those who did not understand the information (“don’t know” responses) would favor more proactive measures, defined as immediately contacting their doctor or searching for information on the Internet. A Pearson correlation revealed an inverse relationship between “don’t know” responses and participants’ tendency to call a doctor (r = − 0.184, p < 001), and a positive relationship between “don’t know” and the other three courses of action: searching the Internet (r = 0.438, P < 001), waiting for the doctor to call them (r = 0.442, P < .005), or waiting for their next visit to the doctor (r = 0.488, P < .005). Thus, the less understandable the information presented, the less likely the participants were to immediately call their family doctors. Rather, they were more likely to search the Internet for information, wait for their next doctor’s appointment, or wait for their doctor to contact them. Consulting the Internet for information was positively correlated with waiting for the doctor to call and waiting for one’s next visit to the doctor (r [298] = 0.482, P < 0.001; r [298] = 0.467, P < 0.001, respectively). Thus, our hypothesis was not supported (see Table ).
It has been shown that engaged patients – those who actively seek to know more about and manage their own health – are more likely than others to participate in preventive and healthy practices, self-manage their conditions, and achieve better outcomes . Studies also show that engaged patients are better able to understand whether or not a result is worrisome, and what actions, if any, should be followed . In our study, respondents found the numeric format hardest to understand and the graphic format the easiest. Yet they displayed a slightly higher level of inaccuracy in the graphic format, less so in the numeric, and the least in the verbal. In other words, for those respondents who hazarded a gravity assessment (as opposed to those who chose the “don’t know” option), information was most difficult to interpret correctly when presented in a line graph, and easiest to interpret correctly when presented numerically (in a tabular form). However, these differences should be explored further. Our findings concur with previous studies which suggest that graphs may appear as an appealing alternative to numbers because visualization allows for quick and intuitive assessment. However, some aspects of graph interpretation may require effortful cognitive skills that must be learned . Formats that leave respondents less able to understand the results – namely, the verbal and graphic formats – produced lower inclination to actively seek professional help. Low levels of understanding (operationalized through respondents’ choice of “don’t know” when asked to assess the gravity of the information) were negatively correlated with an expectation of immediately calling the doctor, and positively correlated with searching the Internet, waiting for the doctor to contact the patient, and waiting for one’s next visit to the doctor. Thus, uncertainty regarding the meaning of the lab results drove participants to shift the burden of responsibility to their doctors, as well as to delay actively seeking medical services . As Zikmund-Fisher and his colleagues found in their study , high perceived gravity of the health condition was the only predictor of immediately calling the doctor. Our findings suggest that age is an important predictor of both accuracy and uncertainty, indicating that familiarity with a wide range of health conditions and with the healthcare system may enhance accurate interpretation of the results. This finding is in line with the literature indicating that a broad acceptance of personal health record (PHR) technology may not be related to education or income, but to the patient’s health literacy . However, the participants in this study were not required to operate the EPR to elicit the test results. It is possible that low proficiency in navigating patient portals can affect older people’s effective use of these technologies. A semantic approach to knowledge transfer posits that even if a common syntax or language is present, differences of interpretation can impede communication between experts and laypersons . As suggested by Witteman and Zikmund-Fisher , patients viewing laboratory results may not care about the number itself. Instead, they wish to know: “Is this good or bad?” or, more personally, “Am I OK?” or “Do I need to do anything?” (, p. 360). Moreover, simply understanding the plain meaning of medical information may not be enough to interpret the information’s significance. For instance, it is rarely enough to understand the meaning of each isolated result, to truly assess the gravity of a health condition, patients need to grasp the comprehensive meaning of the results. To move patients from adherence to engagement, personalized information must be presented in a way that ensures precision of interpretation, not only informing patients, but allowing them to act on the information . And so, to engage individuals in their health, it is critical that the numbers, values, terms, and units have meaning for the person receiving them, and can easily become actionable. A well-designed results sheet can and should encourage patients to take an active role in interpreting their test results in ways that will allow them to follow up on their health. We suggest that graphs, tables, and charts could be made easier to interpret if coupled with a brief and concise verbal explanation, using language that is familiar to readers. Moreover, instead of indicating a diversion from the norm, it might be more helpful to indicate an overall level of urgency, and include a recommendation for the type of follow-up required (i.e., a consultation with a doctor, more tests, or certain types of monitoring). Limitations Our findings should be considered in light of the study’s limitations. First, though the information presented was drawn from authentic records, our respondents encountered it in the context of hypothetical scenarios. This method has been documented in the literature on EPR design (e.g., ). Nonetheless, addressing hypothetical scenarios meant that the participants lacked the personal relevance that such test results have for patients attempting to manage these conditions. Relatedly, we did not measure respondents’ familiarity with the specific tests presented. We assumed that many of our respondents were familiar with at least some of the conditions in the scenarios, and had perhaps even managed them in the past. We also assumed that there are many scenarios in which patients with no prior knowledge of a given laboratory test might view laboratory results in a patient portal. We therefore believe that our study design successfully simulated realistic circumstances. However, it is possible that our results may not accurately reflect how people respond to their own personalized health information. Future inquiries should consider further how people might interpret their own health information, using a combination of quantitative and qualitative methods. Second, the sample size, and level of attrition in studies such as this need be addressed in further studies. It is possible that the length of the questionnaire and its complexity contributed to both. Furthermore, despite our sample’s diversity in terms of demographics, experience with the EPR system and health beliefs, our research design prevented us from reaching minority groups within the Israeli population, such as Arabs, ultraorthodox Jews, and immigrants from the former Soviet Union or of Ethiopian or French descent. Their omission from the study was partly due to methodological complications. In particular, the language used in the EPR is Hebrew, but for many Israeli minorities Hebrew is a second language. Asking minority respondents to fill in the questionnaire in Hebrew would have required a control for language proficiency that was beyond the scope of this study; while the alternative, translating the scenarios into Russian, Amharic, Arabic or French, might have reduced the authenticity of the information. In the present case, we can assume that if native Israelis, highly proficient in Hebrew, demonstrated significant deficiencies in their comprehension of personalized medical information, members of these populations would do so as well. However, we encourage scholars to study the role of cultural and educational diversity in the interpretation of personalized health information. To further our understanding on how information presentation affects laypersons’ understanding, perceptions and actions, future studies should design methodologies that can survey larger and more diversified populations.
Our findings should be considered in light of the study’s limitations. First, though the information presented was drawn from authentic records, our respondents encountered it in the context of hypothetical scenarios. This method has been documented in the literature on EPR design (e.g., ). Nonetheless, addressing hypothetical scenarios meant that the participants lacked the personal relevance that such test results have for patients attempting to manage these conditions. Relatedly, we did not measure respondents’ familiarity with the specific tests presented. We assumed that many of our respondents were familiar with at least some of the conditions in the scenarios, and had perhaps even managed them in the past. We also assumed that there are many scenarios in which patients with no prior knowledge of a given laboratory test might view laboratory results in a patient portal. We therefore believe that our study design successfully simulated realistic circumstances. However, it is possible that our results may not accurately reflect how people respond to their own personalized health information. Future inquiries should consider further how people might interpret their own health information, using a combination of quantitative and qualitative methods. Second, the sample size, and level of attrition in studies such as this need be addressed in further studies. It is possible that the length of the questionnaire and its complexity contributed to both. Furthermore, despite our sample’s diversity in terms of demographics, experience with the EPR system and health beliefs, our research design prevented us from reaching minority groups within the Israeli population, such as Arabs, ultraorthodox Jews, and immigrants from the former Soviet Union or of Ethiopian or French descent. Their omission from the study was partly due to methodological complications. In particular, the language used in the EPR is Hebrew, but for many Israeli minorities Hebrew is a second language. Asking minority respondents to fill in the questionnaire in Hebrew would have required a control for language proficiency that was beyond the scope of this study; while the alternative, translating the scenarios into Russian, Amharic, Arabic or French, might have reduced the authenticity of the information. In the present case, we can assume that if native Israelis, highly proficient in Hebrew, demonstrated significant deficiencies in their comprehension of personalized medical information, members of these populations would do so as well. However, we encourage scholars to study the role of cultural and educational diversity in the interpretation of personalized health information. To further our understanding on how information presentation affects laypersons’ understanding, perceptions and actions, future studies should design methodologies that can survey larger and more diversified populations.
To conclude, this study makes three unique contributions. First, it is concerned not only with assessments of urgency, but with the accuracy of patients’ assessments. Second, it uncovers how people react when they are unsure what the results encountered in electronic records mean. Third, it examines which follow-up actions laypersons are likely to take in response to their interpretation of the results. As such, the paper deals with the core problem of digitation – namely, how to make medical information understandable so that it can be translated into appropriate and timely action. Addressing these concerns is key to designing health information technologies that can improve laypersons’ engagement and self-care, as well as reduce both under- and over-utilization of health services.
|
Economic evaluation of pharmacogenomic-guided antiplatelet treatment in Spanish patients suffering from acute coronary syndrome participating in the U-PGx PREPARE study | af64eff4-6505-4a67-99b0-6f8da5b39401 | 10249170 | Pharmacology[mh] | Cardiovascular diseases (CVDs) are one of the leading causes of mortality all over the world. Being classified as a non-communicable disease, CVDs are a dominant health issue with both social and economic burdens. In numbers, CVDs are the leading cause of death in Spain accounting for almost 120.000 deaths per year while it is the second reason for hospital admissions estimating to reach almost 592.000 hospitalizations per year . Acute coronary syndrome (ACS) constitutes a life-threatening CVD type, associated with high risk of morbidity and mortality. It includes a range of heart conditions related to sudden, reduced blood flow to the heart. Myocardial infarction (MI) (both ST elevation myocardial infarction (STEMI) and non-ST (NSTEMI)) and unstable angina are a few examples of ACS . Unfortunately, ACS incidence rate is rapidly increasing all over the world due to modifiable factors such as smoking, obesity, extensive alcohol consumption, diabetes mellitus, hypertension, etc. . Therapy with antiplatelets is the first-line treatment strategy for ACS since dual antiplatelet therapy with aspirin and a P2Y12 receptor antagonist is usually recommended . Clopidogrel is a well-known P2Y(12) receptor antagonist commonly prescribed to ACS patients with high bleeding risk . This antiplatelet prodrug is metabolized by CYP450 in liver to active metabolite (clop-H4) that inhibits platelet aggregation and subsequent thrombogenesis by binding to ADP platelet receptor P2Y12 . Less than 15% of the prodrug is transformed into an active form, while the remaining 85% is hydrolyzed by esterases to inactive forms, subsequently excreted from the human body . CYP2C19 enzyme is encoded from the CYP2C19 gene, which is highly polymorphic, and it shows great variability (approximately 12%) among populations due to inter-individual and inter-ethnic differences in the genetic background, resulting in significant variation in the drug metabolizing status of the CYP2C19 enzyme, both in terms of drug efficacy and toxicity . Being involved in the whole bioactivation process of clopidogrel, CYP2C19 genetic variation exerts a significant impact on the formation of active metabolite. Indeed, approximately 5–40% of patients treated with conventional doses of clopidogrel display inadequate antiplatelet responses owing to low inhibition of ADP-induced platelet activation, which could lead to severe cardiovascular and cerebrovascular complications . Evidently, such phenomena are mainly attributed to genetic variants in CYP2C19 , resulting in poor or intermediate metabolizer phenotypes (PM and IM) that are receiving a sub-optimal therapy and have a high on treatment platelet reactivity . Many randomized controlled clinical trials like POPular Genetics and TAILOR-PCI have established the correlation between CYP2C19 genotype and clopidogrel response primarily in the cohort of ACS patients undergoing PCI showing the importance of de-escalation of dose based on genetic results and switching to different P2Y12 receptor antagonist, a fact that highlights the potential impact of PGx testing in antiplatelet treatment . Clopidogrel is one of the first medications to be associated with pharmacogenomic (PGx) biomarkers and clinical guidelines . Based on the Dutch Pharmacogenomics Working Group (DPWG), patients with an actionable phenotype due to a genetic variation in the CYP2C19 gene are recommended to either undertake higher drug dosage (IM) or switch to an alternative antiplatelet therapy (PM) (e.g., prasugrel or ticagrelor) in case of no other contraindication to avoid adverse drug reactions . Antiplatelet treatment and P2Y12 receptor antagonists have high risk of severe ADRs that can lead to a person’s hospitalization and are commonly related with severe bleeding events . In accordance with the latest European estimations, the annual CVDs cost to European Union economy can reach up to 210 billion euros per year and 53% of those are related to healthcare costs due to hospitalization . Hospitalization costs along with high incidence of ADRs, population aging, and scarcity of available resources put viability of European healthcare systems at risk. PGx is a promising technology that can improve the overall flow of drug and disease management by tailoring one’s medication according to individual’s genomic profile, and consequently to reduce the risk of ADRs and at the same time maximize treatment’s efficacy . Evidently, PGx-guided strategy in antiplatelets is shown to bring fruitful results in ACS disease management (reduced MACE, bleeding events) in accordance with several randomized clinical trials and it has been characterized as a “reasonable alternative for standard P2Y12 inhibitor therapy based on European Society of Cardiology guidelines published in 2021 . By taking into consideration the disease prevalence and incidence, even a small improvement thanks to the adoption of PGx-guided treatment is likely to be translated into meaningful population-level health gains . Nonetheless, physicians treating CVD patients haven’t widely adopted this initiative and PGx strategy hasn’t been implemented in the clinical setting . Given that the available resources are rather scarce, the aim of the present study is to estimate whether a PGx-guided clopidogrel treatment is cost-effective compared to conventional clopidogrel treatment in patients diagnosed with ACS in the Spanish healthcare setting. Data collection Both clinical and economic data derived from the PREPARE (PREemptive Pharmacogenomic testing for preventing Adverse drug REactions study), a prospective, open-label, randomized controlled clinical trial having taken place at the University Hospital of San Cecilio, the University hospital Virgen de las Nieves, the Zaidin South Primary Care Centre, and the Zaidin Specialty Centre, Granada Spain, from May 2017 until June 2020 . PREPARE is the first and largest multinational, open-label, controlled, cluster-randomized, crossover implementation study that investigates the clinical and cost-effectiveness of implementing preemptive genotyping testing in the population using a PGx panel . PREPARE protocol was previously reported elsewhere . The present analysis refers to data collected from Spanish sites participating in the study. The study analysis was undertaken based on 243 participants for both arms, 113 subjects in the PGx group and 130 in the control group, for whom detailed medical records were documented in source documents and in study’s electronic case report system (eCRF). Study design All inclusion and exclusion criteria of the study are briefly described below. Subjects of any ethnicity, ≥ 18 years of age with a clinical diagnosis of a type of ACS (i.e., MI, unstable angina, ST and non-ST elevation, STEMI and NSTEMI) that were primer naïve to clopidogrel, hadn’t undertaken any genetic testing in the past for CYP2C19 , consented to be followed up for at least 12 weeks and could give blood or saliva sample were eligible to participate in the study. Patients were excluded in case that (a) they were reluctant to give signed informed consent, (b) were pregnant or breastfeeding, (c) were suffering from advanced liver failure (stage Child–Pugh C) or had an existing impaired hepatic or renal function, (d) their estimated life expectancy was less than 3 months and (e) had no fixed address or an assigned general practitioner. Physicians participating in the study established the diagnosis of ACS, the life expectancy of patient and medical history of each patient relying on all available clinical data . In Spain, PGx-guided treatment group run from April 2017 until September 2018 and the other group from November 2018 until June 2020. All study participants were followed up for a minimum of 12 weeks and no more than 18 months. Control group followed a non-tailored treatment strategy based on the common clinical routine related to clopidogrel whereas PGx-guided group received a PGx-guided treatment strategy based on each patient’s CYP2C19 genotyping results. During the study, subjects were asked to complete two online questionnaires at week 2 and at week 8 and to perform four interviews called nurse assessments on baseline, week 4, week 12 and upon 18 months. Those nurse assessments were conducted either via phone calls remotely or on-site interviews by trained research personnel and included questions about disease progression, subject’s quality of life, the occurrence of any adverse event, use of any concomitant medication or procedure and any hospitalization event. On baseline visit, well-trained physicians discussed with participants all study’s requirements including saliva sample, follow-up visits, and interviews and provided them with the informed consent form. Upon giving informed consent, patients donated saliva sample were randomized and prescribed clopidogrel in 75 mg/per day as loading dose. Genetic results of PGx group were available within 7 days upon sample collection day. Then, physicians reviewed each patient’s results to tailor individual’s clopidogrel treatment either by adjusting the dosage or by changing medication in accordance with DPWG relevant guidelines . Therefore, PGx treatment strategy and maintenance dose were finalized a week upon patient’s enrolment. Basic participants’ demographics information including gender, age, body-mass index (BMI), smoking and alcohol consumption status along with clinical data such as comorbidities and co-medication use was recorded at the baseline visit (see Table ). Data related to adverse events, utilities, visits to emergency units, and hospital admissions were collected via the nurse assessments as mentioned above. All available data of the present analysis were collected by clinical staff trained in study’s protocol and systems. Data were reviewed and reconciliated by two of the main authors of the paper for any typos or discrepancies between source documents and eCRF. Upon reviewing database, 243 patients were included in the analysis. PREPARE trial was performed in compliance with the 1964 Helsinki declaration. It was approved by Comité Coordinador de Ética de la Investigación Biomédica de Andalucía (CCEIBA)—ethics committee in Spain , and it is registered on clinicaltrials.gov (NCT03093818). Perspective of analysis The analysis perspective of this study was that of sickness fund . All type of direct medical costs (hospitalization costs, emergency costs, follow-up costs, genetic testing cost) along with the relevant induced costs were included. Those costs were reimbursed by the payers in Spanish Prefecture of Andalusia. Other direct costs borne and paid by the patients (diet costs, travel expenses, home nurse aide, etc.) or indirect costs such as loss of productivity due to absenteeism , albeit important, were not taken into consideration for this analysis. Missing data analysis Dealing with missing data is a common issue in economic analysis, and their proper handling might improve the cost-effectiveness conclusions . Following Faria and coworkers, a descriptive analysis was undertaken to provide details regarding the percentage of missing values in individuals’ answers in nurse assessments including both details about utilities and assessment dates . Then, a logistic regression was run to gain insight into the association among missingness—which represented as a binary variable—and (a) baseline characteristics (such as age, gender, BMI, etc.) and (b) final outcomes (total cost and quality-adjusted life years (QALYs)) . Missing baseline values can have a great impact on the analysis, on the ground that it might be necessary to use those missing values to predict subsequent outcomes . Indeed, single imputation method was applied for baseline utility in each treatment arm, by filling the missing values with the average of the observed cases . For intermittent missing data in quality-of-life answers, when possible, linear interpolation method was used between measurement points , while multiple imputation method with five imputed datasets was done for the rest of them . Right censored cost data Right censoring for cost data is a specific case of missingness in which some individuals are lost to follow-up within the study period or still alive at the time of study completion and, thus, their complete/total costs are not available for statistical analysis . To deal with this issue, the nonparametric, unbiased and consistent Bang–Tsiatis estimator was employed , adding a correction term to improve efficiency (Zhao-Tian estimator). In short, this estimator calculates the weighted cost for each uncensored individual per group, based on the inverse probability of being censored at the time of failure. For computational purposes, a more intuitive replace-from-the-right algorithm was used as an equivalent alternative to the Zhao-Tian estimator . Briefly, the cost of each censored individual was replaced by the average of costs of those individuals who survived longer than him/her, taking also into account the mean cost of each censored patient into account at the time of censoring and projecting this cost to the estimated unobserved survival . Utility values Utility values describe the health-related quality of life (QoL) associated with different health states. In the original analysis plan, time-trade-off method was applied, but this plan was abandoned due to the low response rate of participants . Hence, utility weights were also extracted from the literature . In particular, the ‘‘well’’ state was set at 0.87, while for those experiencing any major event, utility decrements and the correspondence duration were used (see Table ). In sensitivity analysis, the QoL was estimated by means of participants’ VAS score given at baseline visit, week 4, week 12 and 18 months from baseline. Quality-adjusted life years (QALYs) were measured by calculating the integral of the product of individual’s life expectancy multiplied by weighted VAS score and adjusting the baseline measures of utility within a covariate regression framework . Costing methodology and economic analysis Treatment’ effectiveness was determined by mean survival and it was estimated based on the official start date of clopidogrel to (a) death related to the CVDs (complete cases), (b) death from any cause (complete cases), (c) loss to follow-up (censored patients) or (d) to the end of study period (censored patients). Total cost included (a) the cost of ADRs, (b) hospitalization’s costs, (c) follow-up costs and (d) the cost of genetic testing applicable only for PGx-guided group. Cost of index drug (clopidogrel) itself was not taken into account in the analysis owing to the fact that both groups represent a pool of individuals with different health status and comorbidities and only ADRs’ cost can make a difference. Similarly with a previous pharmacoeconomic analysis , patient-level resource utilization data were combined with unit cost data and then aggregated to compute total treatment cost per patient. The following ADRs were considered for cost evaluation: ACS, gastrointestinal pain, heart failure, dizziness, chest pain, cardiac arrest, stroke, MI, atrial fibrillation, diarrhea, cardiac arrest, oral hemorrhage, rectal hemorrhage and colonic hemorrhage (see Table ). Variation in resource utilization among individuals reflects differences related to hospitalization, health complications, unplanned operations, laboratory tests, etc. Reimbursement tariffs used were obtained from the official sources and were applicable to all public hospitals and public payers of Andalucía community region in Spain. All components’ costs are presented in Table . Due to limited time horizon of this observational study, discount rate was not applied. In addition, due to lack of official price, genetic test’s cost was extracted from the literature and is consistent to those reported in another Mediterranean country (Italy) . Finally, a generalized linear model (GLM) was employed to estimate the effect of covariates (patients characteristics) in total cost to achieve greater flexibility in the presence of heteroskedasticity and right skewness in cost data . In particular, a tweedie distribution was assumed for cost and a logarithmic as a link function. Moreover, a multivariate seemingly unrelated regression equation was employed to provide the necessary information for statistical inference in cost-effectiveness analyses, namely differences in costs and QALYs along with the correlation between the estimations . Incremental Cost-Effectiveness Ratio (ICER) was determined as the ratio of the difference in costs between PGx-guided group vs control group divided by the difference in QALYs. Uncertainty A probabilistic sensitivity analysis was undertaken to test data robustness and to identify how the deterministic results vary under uncertainty . In particular, a new dataset with 5000 nonparametric bootstrap replications with replacement was constructed to determine confidence intervals for the main variables. In the present analysis, the straightforward percentile method was applied . Cost-Effectiveness Acceptability Curve was used to illustrate probabilistic results, which shows the probability (on the y-axis) that PGx-guided group may be cost-effective compared to control group for a range (on the x -axis) of maximum monetary values that a decision-maker might be willing to pay per QALY. Based on the assumption of bivariate normality, an ellipse and its contour were constructed to represent the 95% confidence intervals . As a last step, a Value of Information Analysis was performed to investigate the monetary value that can be adjusted to eliminate uncertainty in the decision-making process . The main metrics used were the Expected Value of Perfect Information (EVPI) value, for three different willingness-to-pay thresholds for a QALY, except of partially EVPI. Both clinical and economic data derived from the PREPARE (PREemptive Pharmacogenomic testing for preventing Adverse drug REactions study), a prospective, open-label, randomized controlled clinical trial having taken place at the University Hospital of San Cecilio, the University hospital Virgen de las Nieves, the Zaidin South Primary Care Centre, and the Zaidin Specialty Centre, Granada Spain, from May 2017 until June 2020 . PREPARE is the first and largest multinational, open-label, controlled, cluster-randomized, crossover implementation study that investigates the clinical and cost-effectiveness of implementing preemptive genotyping testing in the population using a PGx panel . PREPARE protocol was previously reported elsewhere . The present analysis refers to data collected from Spanish sites participating in the study. The study analysis was undertaken based on 243 participants for both arms, 113 subjects in the PGx group and 130 in the control group, for whom detailed medical records were documented in source documents and in study’s electronic case report system (eCRF). All inclusion and exclusion criteria of the study are briefly described below. Subjects of any ethnicity, ≥ 18 years of age with a clinical diagnosis of a type of ACS (i.e., MI, unstable angina, ST and non-ST elevation, STEMI and NSTEMI) that were primer naïve to clopidogrel, hadn’t undertaken any genetic testing in the past for CYP2C19 , consented to be followed up for at least 12 weeks and could give blood or saliva sample were eligible to participate in the study. Patients were excluded in case that (a) they were reluctant to give signed informed consent, (b) were pregnant or breastfeeding, (c) were suffering from advanced liver failure (stage Child–Pugh C) or had an existing impaired hepatic or renal function, (d) their estimated life expectancy was less than 3 months and (e) had no fixed address or an assigned general practitioner. Physicians participating in the study established the diagnosis of ACS, the life expectancy of patient and medical history of each patient relying on all available clinical data . In Spain, PGx-guided treatment group run from April 2017 until September 2018 and the other group from November 2018 until June 2020. All study participants were followed up for a minimum of 12 weeks and no more than 18 months. Control group followed a non-tailored treatment strategy based on the common clinical routine related to clopidogrel whereas PGx-guided group received a PGx-guided treatment strategy based on each patient’s CYP2C19 genotyping results. During the study, subjects were asked to complete two online questionnaires at week 2 and at week 8 and to perform four interviews called nurse assessments on baseline, week 4, week 12 and upon 18 months. Those nurse assessments were conducted either via phone calls remotely or on-site interviews by trained research personnel and included questions about disease progression, subject’s quality of life, the occurrence of any adverse event, use of any concomitant medication or procedure and any hospitalization event. On baseline visit, well-trained physicians discussed with participants all study’s requirements including saliva sample, follow-up visits, and interviews and provided them with the informed consent form. Upon giving informed consent, patients donated saliva sample were randomized and prescribed clopidogrel in 75 mg/per day as loading dose. Genetic results of PGx group were available within 7 days upon sample collection day. Then, physicians reviewed each patient’s results to tailor individual’s clopidogrel treatment either by adjusting the dosage or by changing medication in accordance with DPWG relevant guidelines . Therefore, PGx treatment strategy and maintenance dose were finalized a week upon patient’s enrolment. Basic participants’ demographics information including gender, age, body-mass index (BMI), smoking and alcohol consumption status along with clinical data such as comorbidities and co-medication use was recorded at the baseline visit (see Table ). Data related to adverse events, utilities, visits to emergency units, and hospital admissions were collected via the nurse assessments as mentioned above. All available data of the present analysis were collected by clinical staff trained in study’s protocol and systems. Data were reviewed and reconciliated by two of the main authors of the paper for any typos or discrepancies between source documents and eCRF. Upon reviewing database, 243 patients were included in the analysis. PREPARE trial was performed in compliance with the 1964 Helsinki declaration. It was approved by Comité Coordinador de Ética de la Investigación Biomédica de Andalucía (CCEIBA)—ethics committee in Spain , and it is registered on clinicaltrials.gov (NCT03093818). The analysis perspective of this study was that of sickness fund . All type of direct medical costs (hospitalization costs, emergency costs, follow-up costs, genetic testing cost) along with the relevant induced costs were included. Those costs were reimbursed by the payers in Spanish Prefecture of Andalusia. Other direct costs borne and paid by the patients (diet costs, travel expenses, home nurse aide, etc.) or indirect costs such as loss of productivity due to absenteeism , albeit important, were not taken into consideration for this analysis. Dealing with missing data is a common issue in economic analysis, and their proper handling might improve the cost-effectiveness conclusions . Following Faria and coworkers, a descriptive analysis was undertaken to provide details regarding the percentage of missing values in individuals’ answers in nurse assessments including both details about utilities and assessment dates . Then, a logistic regression was run to gain insight into the association among missingness—which represented as a binary variable—and (a) baseline characteristics (such as age, gender, BMI, etc.) and (b) final outcomes (total cost and quality-adjusted life years (QALYs)) . Missing baseline values can have a great impact on the analysis, on the ground that it might be necessary to use those missing values to predict subsequent outcomes . Indeed, single imputation method was applied for baseline utility in each treatment arm, by filling the missing values with the average of the observed cases . For intermittent missing data in quality-of-life answers, when possible, linear interpolation method was used between measurement points , while multiple imputation method with five imputed datasets was done for the rest of them . Right censoring for cost data is a specific case of missingness in which some individuals are lost to follow-up within the study period or still alive at the time of study completion and, thus, their complete/total costs are not available for statistical analysis . To deal with this issue, the nonparametric, unbiased and consistent Bang–Tsiatis estimator was employed , adding a correction term to improve efficiency (Zhao-Tian estimator). In short, this estimator calculates the weighted cost for each uncensored individual per group, based on the inverse probability of being censored at the time of failure. For computational purposes, a more intuitive replace-from-the-right algorithm was used as an equivalent alternative to the Zhao-Tian estimator . Briefly, the cost of each censored individual was replaced by the average of costs of those individuals who survived longer than him/her, taking also into account the mean cost of each censored patient into account at the time of censoring and projecting this cost to the estimated unobserved survival . Utility values describe the health-related quality of life (QoL) associated with different health states. In the original analysis plan, time-trade-off method was applied, but this plan was abandoned due to the low response rate of participants . Hence, utility weights were also extracted from the literature . In particular, the ‘‘well’’ state was set at 0.87, while for those experiencing any major event, utility decrements and the correspondence duration were used (see Table ). In sensitivity analysis, the QoL was estimated by means of participants’ VAS score given at baseline visit, week 4, week 12 and 18 months from baseline. Quality-adjusted life years (QALYs) were measured by calculating the integral of the product of individual’s life expectancy multiplied by weighted VAS score and adjusting the baseline measures of utility within a covariate regression framework . Treatment’ effectiveness was determined by mean survival and it was estimated based on the official start date of clopidogrel to (a) death related to the CVDs (complete cases), (b) death from any cause (complete cases), (c) loss to follow-up (censored patients) or (d) to the end of study period (censored patients). Total cost included (a) the cost of ADRs, (b) hospitalization’s costs, (c) follow-up costs and (d) the cost of genetic testing applicable only for PGx-guided group. Cost of index drug (clopidogrel) itself was not taken into account in the analysis owing to the fact that both groups represent a pool of individuals with different health status and comorbidities and only ADRs’ cost can make a difference. Similarly with a previous pharmacoeconomic analysis , patient-level resource utilization data were combined with unit cost data and then aggregated to compute total treatment cost per patient. The following ADRs were considered for cost evaluation: ACS, gastrointestinal pain, heart failure, dizziness, chest pain, cardiac arrest, stroke, MI, atrial fibrillation, diarrhea, cardiac arrest, oral hemorrhage, rectal hemorrhage and colonic hemorrhage (see Table ). Variation in resource utilization among individuals reflects differences related to hospitalization, health complications, unplanned operations, laboratory tests, etc. Reimbursement tariffs used were obtained from the official sources and were applicable to all public hospitals and public payers of Andalucía community region in Spain. All components’ costs are presented in Table . Due to limited time horizon of this observational study, discount rate was not applied. In addition, due to lack of official price, genetic test’s cost was extracted from the literature and is consistent to those reported in another Mediterranean country (Italy) . Finally, a generalized linear model (GLM) was employed to estimate the effect of covariates (patients characteristics) in total cost to achieve greater flexibility in the presence of heteroskedasticity and right skewness in cost data . In particular, a tweedie distribution was assumed for cost and a logarithmic as a link function. Moreover, a multivariate seemingly unrelated regression equation was employed to provide the necessary information for statistical inference in cost-effectiveness analyses, namely differences in costs and QALYs along with the correlation between the estimations . Incremental Cost-Effectiveness Ratio (ICER) was determined as the ratio of the difference in costs between PGx-guided group vs control group divided by the difference in QALYs. A probabilistic sensitivity analysis was undertaken to test data robustness and to identify how the deterministic results vary under uncertainty . In particular, a new dataset with 5000 nonparametric bootstrap replications with replacement was constructed to determine confidence intervals for the main variables. In the present analysis, the straightforward percentile method was applied . Cost-Effectiveness Acceptability Curve was used to illustrate probabilistic results, which shows the probability (on the y-axis) that PGx-guided group may be cost-effective compared to control group for a range (on the x -axis) of maximum monetary values that a decision-maker might be willing to pay per QALY. Based on the assumption of bivariate normality, an ellipse and its contour were constructed to represent the 95% confidence intervals . As a last step, a Value of Information Analysis was performed to investigate the monetary value that can be adjusted to eliminate uncertainty in the decision-making process . The main metrics used were the Expected Value of Perfect Information (EVPI) value, for three different willingness-to-pay thresholds for a QALY, except of partially EVPI. The number and proportion of complete data in each nurse assessment are shown in Table . In addition, Table summarizes the multiple logistic regression model which explore the relationship between the presence of censoring and baseline characteristics. The log of the odds of a censored case was found to be positively associated with hypertension (OR 0.20, 95% CI: 0.05–0.75, p value = 0.017). The Hosmer & Lemeshow (H–L) goodness of fit test was estimated at x 2 (8) = 12.24, p = 0.141, while the Nagelkerke (pseudo) R 2 was 17.6%. The overall predictive score of the model was very high, estimated at 94.2%. In contrast, analysis indicated that the association of censoring with final outcomes (total cost and QALYs) were not statistically significant when adjusted for baseline characteristics (not shown in tables, available on request). Thus, it was concluded that there was a Missing-at-Random covariance-depended context and the correction of censoring was applied separately for those suffering from hypertension in each group. In a similar manner, missing data were also covariance-dependent, and consequently, a missing-at-random hypothesis was adopted for the multiple imputation analysis (see Tables and ). In general, PGx-guided group shared better results in several parameters. More precisely, it was found that PGx-guided group was associated with fewer visits in emergency units, less ADRs and fewer hospital admissions compared to the control group and subsequently, lower costs. However, no statistically significant difference was found between the groups in terms of QALYs and life-years (LYs) The mean estimate for QALYs (base case scenario) in the PGx-guided group was 1.07 (95% CI, 1.04–1.10) versus 1.06 (95% CI, 1.03–1.09) for the control group, while LYs for both groups were estimated at 1.24 (95% CI, 1.20–1.26) and 1.23 (95% CI, 1.19–1.26), respectively. PGx-guided group shared better results in terms of VAS score. In particular, PGx-guided group shared 0.84 QALYs (95% CI, 0.80–0.88) in comparison with 0.76 QALYs (95% CI, 0.72–0.79) of the control group. Furthermore, the mean total cost of PGx-guided group was €883 (95% UI, €316–€1582), while control group shared a mean cost of €1755 (95% UI, €765–€2949), a finding suggesting that PGx-guided treatment might be a cost-saving option with a mean difference of €873 (95% CI, €− 389–€2189). Furthermore, and importantly, health utilization costs were much less (35.05%) in the PGx-guided group (8408.44 EUR) compared to the control group (12,939.29 EUR; see Table ). Hospital admission costs accounted for most of the expenses in both groups (65% in the PGx-guided group versus 77.9% in the control group) followed by emergency units (13.3% in the PGx-guided group versus 17% in the control group) and follow-up costs (2.8% in the PGx-guided group versus 5.1% in the control group. PGx-guided group had an additional cost dedicated to genotyping testing that represented a 13.6% of the total group’s costs. It is noteworthy that there were a few extreme values because some costs in patients of both groups were as low as €130 or even lower, while others as much as €12,000 or even more due to more expensive resources consumed to deal with their adverse events. The results of the GLM illustrated in Table highlighted that variables such as study group, diabetes and hypertension were statistically significant, while the rest of variables didn’t provide any additional predicted value to the model and thus, were excluded from the analysis. Based on parameters’ estimates for seemingly unrelated regression model, the main cost-effectiveness parameters were: (ΔC = €1229.3 ± 566.7, ΔΕ = − 0.199 (572.4) and r = 0.001). Since the standard deviations (SD) were high for both groups, with very low correlation coefficient, a nonparametric bootstrap replication was preferred. In particular, Fig. depicts the joint distribution of 5,000 bootstrap experiments of the difference in the total cost and in the effectiveness (measured in QALYs), between the two study groups. It was assumed that the depicted ellipse followed the bivariate normal distribution, and its contour represented the 95% confidence intervals. Most bootstrap pairs fell into IV quadrant, in which PGx-guided treatment option is more effective and simultaneously less costly. Hence, there is a neutrality between the two alternatives in terms of QALYs as the 5000 dots were scattered almost evenly around the x- axis. In this aspect, since the concept of cost-effective represents a subjective assessment, a willingness-to-pay-threshold was determined to estimate the probability of acceptance or rejection of the PGx technology in the Spanish healthcare setting. Probabilistic results were illustrated using a cost-effectiveness acceptability curve (see Fig. ) in which PGx-guided group (on the y -axis) may be cost-effective compared to control for a range (on the x -axis) of maximum monetary values that a decision-maker might be willing to pay per QALY. This explains that the acceptability curve is relatively independent of the value of the ceiling ratio and in favor of PGx technology. Indeed, the probability of PGx-guided treatment of being cost-effective increased significantly at a lower willingness-to-pay (WTP) threshold. Notably, at €50,000 per QALY, the probability of being cost-effectiveness was higher than 50%, at €30,000 was almost 62%, in case of WTP < €20,000, the probability overcome the 71%. EVPI analysis indicated that the cost of information for 50,000/QALY, 30,000/QALY and 20,000/QALY was determined at €659.4, €287.9 and €136.5, respectively (see Fig. ). Clopidogrel is the oldest and the most popular antiplatelet drug, used by millions of ACS patients every year. Besides its health benefits, it is demonstrated to be associated with increased risk of ADRs in a considerable number of patients. Apart from drug–drug interactions and the effect of concomitant medications, it was shown that individuals’ genotype affects clopidogrel metabolism leading to ADRs. CYP2C19 genomic variants have been linked to variable response of CVD patients to clopidogrel, a fact that implies the necessity for more personalized treatment schemes following PGx testing. In the present analysis, a cost-effectiveness analysis (CEA) of PGx-guided clopidogrel treatment in individuals suffering from different forms of ACS was conducted. This is one of the few studies that aims to compare the cost-effectiveness of PGx-guided treatment of this antiplatelet agent versus non-PGx-guided one in a cohort of Spanish patients with several comorbidities and without strict eligibility criteria. Our analysis concluded that PGx-guided treatment strategy cost two times less than conventional strategy and has a marginally higher effectiveness. At first, it is noteworthy that this study is based on raw clinical and economic data derived from the PREPARE study which is the largest, multinational, controlled, cluster-randomized, crossover implementation study focusing exclusively in investigating the impact of preemptive PGx testing, a fact that differentiates it from most available studies in the literature that used simulated data . This is very important since real-world evidence regarding each drug–gene pair is limited and RCT data are lacking population diversity and inclusion, a fact that raise concerns about data validity and health equality. PREPARE trial shares then a unique trial design that meets the scientific needs and can enhance the clinical significance of PGx. As it was indicated, PGx-guided treatment represented a cost-saving option compared to a non-tailored one, sharing almost 50% less hospital admissions, less emergency visits and almost 13% less ADRs. All these clinical aspects imply an improvement in disease management that is also translated into costs’ reduction; PGx-guided treatment approach had 50% less total cost compared to the most conventional approach while most costs were related to hospitalization in both arms. These important findings are congruent with the literature. Indeed, Fragoulakis and coworkers, demonstrated that in a cohort of Spanish patients undergoing percutaneous coronary intervention (PCI), PGx-guided treatment was dominant over standard of care with 0.9446 QALYs gained and €2971 cost compared to 0.9379 QALYs and €3205 at 1-year horizon . Hospitalization was also the main type of costs and accounted for the majority of expenses for both arms . Moreover, Claassens and coworkers in POPular Genetics concluded that a CYP2C19 genotype-guided strategy was dominant over conventional therapy (prasugrel or ticagrelor) with 8.98 QALYs gained and €725 k cost savings in a simulated cohort of 1000 patients suffering from ACS . In Dong and coworkers, it was shown that CYP2C19 genotype-guided strategy was a cost-effective approach compared to a non-tailored one, on the grounds that it brought cost savings per patient (($4785/person vs. $5311/person) and a gain of 0.0027 QALYs . The occurrence of ADRs was also reduced by 13% a fact that is in alliance with our findings and there was a significant decrease in medication costs by 20% Moreover, another study highlighted that in China, CYP2C19 genotyping strategy to guide antiplatelet treatment was a cost-effective approach with an ICER of CNY 13,552.74 (US$1930.59) per QALY gained compared to standard of clinical care . Probabilistic analysis demonstrated that in 95.7% of simulations, PGx-guided treatment was cost-effective in a WTP ranging from $0–$175,000 . According to Reese and coworkers, in a simulated cohort of patients suffering from ACS, genotyping-driven group was the dominant treatment strategy owing to its higher clinical effectiveness and lower cost in comparison with universal prescription of clopidogrel to all patients, no matter their genetic makeup, while they focused on the number of adverse events prevented to express effectiveness . In that study, calculated ICER was estimated at (ICER–$6,760, [95% CI –$6,720 to –$6,790]). In another study involving US patients conducted by Borse and coworkers, in which effectiveness was also measured in major cardiovascular events (MACE), it was indicated that PGx testing was cost-effective in 62% of the simulations when WTP threshold was set to US$ 50,000 while universal clopidogrel wasn’t . Moreover, Limdi and coworkers in a simulated cohort demonstrated that PGx treatment was cost-effective ($42,365/QALY) and they pinpointed that it was more likely for PGx treatment to be cost-effective in different WTP thresholds in contrast to non-genotyping-driven strategies . Moreover, in Singapore setting, Kim and coworkers concluded that genetic-driven treatment shared better QALYs and was less costly in the long-run, a conclusion that comes in accordance with Lala and coworkers . Following the results of a systematic review conducted by Verbelen and coworkers, in general most economic evaluations poses a positive attitude toward PGx-guided treatment . More precisely, PGx-guided strategy was presented as dominant in 27% of published economic analysis while 30% of the studies concluded that PGx option is cost-effective . Even if the Verbelen and coworkers study included publications of economic evaluations for all type of studies, it was highlighted that PGx-driven treatment was potentially a cost-effective option that could improve disease and drug management to great extent by diminishing healthcare expenditures . In all above-mentioned studies , all data derived from literature or simulations and no raw data from clinical trials were used except of the Cai and coworkers study. In addition to it, both direct and induced costs were taken into consideration including medications and prescription costs while the analysis perspective was mainly those of payers (ie. sickness fund). The superiority of PGx-guided treatment was also demonstrated in terms of less hospital admissions, less emergency visits and 50% reduction in ADRs occurrence and especially in those of high grade. All these features imply that PGx testing can offer a more optimal disease management and constitute a promising treatment strategy for ACS. These findings are in accordance with other studies. Reese and coworkers mentioned that tailoring one’s clopidogrel treatment following his/her genotyping results resulted in 450 less events . In other words, one additional adverse event was prevented for every 23 individuals. This clinical endpoint can be translated in costs savings, less healthcare resources and less deaths. In a risk–benefit assessment, Guzauskas and coworkers came up with similar results as well . They showed that PGx-guided treatment of ACS patients could reduce the incidence risk of suffering a health complication such as MI, stroke or death by 6.8% compared to universal clopidogrel use and decrease the risk of experiencing a MACE . Given that clopidogrel is correlated with high incidence of MACE, the great difference between preemptive PGx testing and universal clopidogrel in terms of MACE incidence is very positive and illustrates the clinical significance of dosage adjustment, suggests that this approach could improve platelet inhibition and enhances the role of PGx implementation in drug and disease management. Finally, according to the literature demographic factors such as age, BMI, and comorbidities such as obesity, diabetes and hypertension were proven to affect clopidogrel pharmacokinetics and drug metabolism. Indeed, Jiang and coworkers highlighted that there was scientific evidence regarding the association of age and prevalence of ADRs such as bleeding upon the use of clopidogrel, whereas it was shown that cardiac drugs belong to the top three classes of drugs that were responsible for even fatal ADR’s . Those factors may exert an impact on clinical effectiveness of clopidogrel, but it is shown to influence cost-effectiveness results. As reported by Nicolic and coworkers, age and gender had a slight affect in cost-effectiveness findings, but it wasn’t statistically significant while other parameters such as cardiac events could impact the analysis . In this analysis, no demographic or individual characteristics had a significant influence in CEA. Only hypertension was indicated to affect ADR, a fact that comply with Nicolic and coworkers since hypertension was the only parameter with significant impact and not age or gender. This study has a few limitations related to trial design. Not being a randomized controlled clinical trial in the strict sense, PREPARE comes with less strict inclusion/exclusion criteria and thus lower compliance. This fact led to low response rate and missing data, in terms of utility details. From the one hand, this fact seems to be an important drawback for study’s analysis, but from the other hand it is relatable to real-world data and represents a better cohort of patients that every clinician will meet during his/ her exercise. This CEA based on raw data from the PREPARE study in Spanish clinical sites suggests that preemptive genotyping before prescribing clopidogrel could add more value in the clinical practice and improve decision-making process for healthcare professionals. As genotyping was conducted as part of clinical care, unlike previous CEAs, our analysis was not limited by assumptions regarding the availability of genotype data in a timeline conducive for clinical care. Finally, PGx testing lowers the risk of ADRs occurrence and especially of life-threatening events and is possible to decrease the overall healthcare cost. Using raw clinical data that are closer to real-world situations gives an insight into the cohort of individuals eligible to receive clopidogrel treatment and the relevant costs. PGx will play an important role in CVD medicine by broaden the horizon for more efficient and cost-effective medications. Finally, all cost data (except of the genetic test price) derived from the official cost of Andalucía region representing sickness fund perspective, while a wider socioeconomic analysis could be the scope of the future research. |
PLOD2 Is a Prognostic Marker in Glioblastoma That Modulates the Immune Microenvironment and Tumor Progression | 28d05d72-24ce-4102-bbfc-109e9440012d | 9181500 | Anatomy[mh] | GBM is the most common and fatal malignant primary brain tumor in adults . Despite the extensive standard of care therapy including maximal safe surgical resection followed by radiation and chemotherapy, the relative 5-year survival of GBM patients is less than 7% with a median survival of 14 months . The diffuse infiltration in the surrounding brain parenchyma makes complete surgical resection impossible and is the main reason for recurrence and therapy resistance of GBM (reviewed in ). Therefore, the focus of research regarding therapeutic approaches for GBM has shifted towards immunotherapy and individualized therapy. In particular, numerous vaccine approaches, oncolytic viruses and immune-checkpoint inhibitors are in preclinical and clinical trials . Furthermore, specific targeted therapies gained importance through a better understanding of the underlying molecular heterogeneity of GBM. Despite these multimodal approaches, GBM remains an incurable disease at present. Thus, there is still an urgent need to identify novel cellular and molecular mechanisms that control the progression of GBM and could serve as therapeutic targets in this type of cancer. Accumulating evidence indicates that the interplay between tumor cells and the tumor microenvironment (TME) plays a crucial role in tumor migration, invasion and progression . The TME is largely determined by the extracellular matrix (ECM) with collagen as the most abundant protein . During tumor progression, increased collagen crosslinking promotes stiffening of the extracellular matrix, thus enhancing invasion and metastasis . The main enzyme mediating stabilized collagen crosslinks is Procollagen-Lysine,2-Oxoglutarat 5-Dioxygenase 2 (PLOD2) . This membrane-bound homodimeric enzyme hydroxylates lysine residues in the telopeptides of procollagens and thus, plays a crucial role in the post-translational modification of collagen biosynthesis . The resulting hydroxyl groups are essential for the formation of stable crosslinks by lysyl oxidases . PLOD2 is upregulated in various cancers and is associated with poor outcomes in bladder cancer , hepatocellular carcinoma and breast cancer (and reviewed in ). Exploratory studies on a small cohort of 28 GBM patients indicated that PLOD2 was also associated with poor survival in this type of cancer . Furthermore, Xu et al. found that high gene expression of PLOD2 was significantly associated with poor overall- and progression-free survival in glioblastoma patients . The same study also showed that increasing PLOD2 protein levels were associated with increasing tumor grade in glioma . At the molecular level, PLOD2 induces epithelial-mesenchymal transition and activates the PI3K-Akt , JAK-STAT and FAK signaling pathways. Although the exact mechanisms are still largely unclear, PLOD2 can be expected to affect key signaling pathways in tumor cells, thereby modulating tumor progression. The TME also contains a variety of infiltrating and resident immune cells that interact with the tumor cells and modulate their biology and functions. Recent studies showed that the GBM microenvironment hosts a large number of tumor-infiltrating neutrophils, which are actively recruited by GBM cells through the expression of IL-8 and IL-1b (and reviewed in ). Importantly, the presence of infiltrating neutrophils in GBM was significantly associated with a poor outcome in these patients (reviewed in ). These findings suggest that neutrophils are substantially involved in the progression of GBM. As an important modulator of the extracellular matrix and TME, PLOD2 may also activate the tumor-infiltrating neutrophils. The role of PLOD2 in the pathophysiology of GBM still requires extensive characterization. This study aimed to determine (1) the association between PLOD2 expression and the clinical outcome of IDH wild-type GBM patients; (2) the involvement of PLOD2 in the modulation of GBM tumor cell functions and (3) the effect of PLOD2 on the biology and function of neutrophils.
2.1. PLOD2 Associates with and Predicts Poor Overall Survival of GBM Patients Previous studies by Xu et al. using the TCGA database showed that high gene expression of PLOD2 is significantly associated with a poor outcome in GBM patients . Here, we investigated whether the protein levels of PLOD2 in tumor tissue are associated with overall survival (OS) or progression-free survival (PFS) of GBM patients with confirmed IDH wild-type (IDH WT) status. To this end, the levels of PLOD2 were assessed by immunohistochemistry (see Material and Methods section) in two independent patient cohorts. PLOD2 expression was subsequently dichotomized into “low“ and “high“ based on the median-split method. The survival curves were plotted according to the Kaplan–Meier method and the statistical significance was assessed with the log-rank test. In the Hannover cohort, GBM patients with high tumor levels of PLOD2 (PLOD2 high ) had a significantly shorter OS compared to patients with low levels of PLOD2 (PLOD2 low ) ( p = 0.020; log-rank) ( A). These findings were confirmed in the Magdeburg cohort of GBM patients ( p < 0.001; log-rank) ( B). In both cohorts, PLOD2 high patients had a shorter PFS compared to their PLOD2 low counterparts, but statistical significance was only reached in the Magdeburg cohort ( p = 0.001; log-rank) ( C, D). We further analyzed the OS and PFS of IDH WT GBM patients using Cox proportional-hazard models adjusted for factors known to influence the patients’ outcome, such as age , Karnofsky Performance Scale (KPS) , extent of surgical resection , therapy and MGMT methylation status . Initial analysis of the time-dependent covariate (T_COV_) for PLOD2 showed that the proportional hazard assumption of these models had been satisfied: Hannover cohort_OS: p = 0.272; Magdeburg cohort_OS: p = 0.448; Hannover cohort_PFS: p = 0.440; Magdeburg cohort_PFS: p = 0.402. In both cohorts, high expression of PLOD2 predicted poor OS in GBM patients (Hannover cohort: HR = 1.401, CI [95%] = 1.009–1.946, p = 0.044; Magdeburg cohort: HR = 1.493, CI [95%] = 1.042–2.140, p = 0.029) ( A). For PFS, PLOD2 high patients had an increased hazard ratio compared to PLOD2 low patients but statistical significance was only reached in the Magdeburg cohort (HR = 1.645, CI [95%] = 1.040–2.603, p = 0.033) ( B). These data indicate that PLOD2 could serve as an independent prognostic biomarker, at least for the overall survival of GBM patients. 2.2. PLOD2 Promotes the Invasion, Proliferation and Anchorage-Independent Growth of GBM Cells Using U87 and U251 GBM cells, recent studies showed that PLOD2 enhanced the aggressiveness of GBM cells by promoting tumor invasion . Here we investigated the effect of PLOD2 on the biology and functions of GBM using the H4 GBM cell line. To this end, the cells were stably transfected with a sh-RNA plasmid to downregulate the levels of PLOD2 (sh-PLOD2) or with a control plasmid (sh-control) (see Material and Methods section). All functional assays were performed in the absence of the selection antibiotic puromycin. Control western blot analysis confirmed that PLOD2 knock-downs remained stable until at least day 10, which was the last time point of the longest assay . Tumor invasion was assessed by the degree of “gap” closure (red line) in a 3D collagen matrix, using the Oris TM system ( A). The results showed that PLOD2 knock-down significantly reduced the invasiveness of H4 GBM cells ( B). We additionally determined the activity of matrix metalloproteases (MMP2 and MMP9), since MMPs are critical for tumor invasion in many types of cancer, including GBM . To this end, sh-control and sh-PLOD2 GBM cells were incubated in a culture medium and the supernatants were collected 48 h later. A culture medium without cells was used as a control. We found that H4 GBM cells released MMP2 but only negligible levels of MMP9 ( C). Importantly, PLOD2 knock-down cells released significantly lower levels of MMP2 compared to their control-transfected counterparts ( D). These findings indicate that PLOD2 enhances the invasiveness of GBM cells, possibly via MMP2. In further studies, we investigated the role of PLOD2 in GBM proliferation by assessing the metabolic activity of the transfected H4 cells using the MTT assay. To this end, the concentration of metabolized MTT was measured at different time points in one setup with 2000 cells and another with 4000 cells. The results showed that PLOD2 knock-down decreased the metabolic activity of H4 cells in both setups ( A,B). However, statistical significance was only reached for certain time points. We additionally determined the anchorage-independent growth of transfected GBM cells by allowing the cells to form colonies in low-gelling agarose for 10 days ( C). We found that PLOD2 knock-down cells formed significantly fewer colonies than their sh-control counterparts ( D). Taken together, these data indicate that PLOD2 promotes the invasion, proliferation and anchorage-independent growth of GBM cells. 2.3. PLOD2 Modulates the Expression of Catenin D1, CD44, CD99 and MT1-MMP in GBM Cells As shown above, PLOD2 modulates the biological functions of GBM cells. To obtain further insight into the molecular mechanisms downstream of PLOD2 in GBM cells, we assessed by western blot the protein expression of several markers associated with tumor proliferation and invasion, such as Catenin D1, CD44, CD99, CDK6, EGFR, HIF1-beta, Integrin beta-1, MT1-MMP and PRAS40. We found that the levels of Catenin D1, CD44, CD99 and MT1-MMP were significantly lower in PLOD2 knock-down cells compared to their control-transfected counterparts ( A–E). 2.4. GBM-Associated PLOD2 Induces Neutrophil Granulocytes to Aquire a Pro-Tumor Phenotype Accumulating evidence indicates that the GBM microenvironment contains significant numbers of neutrophils and that high neutrophil infiltration is associated with poor outcomes in GBM patients (reviewed in ). Furthermore, very recent studies found a correlation between high tumor levels of PLOD2 and high neutrophil infiltration in cervical and hepatocellular carcinoma tissues. Based on these findings, we hypothesized that GBM cells modulate the biology and functions of neutrophils via PLOD2. To test this hypothesis, we produced conditioned supernatants (SN) from sh-control and sh-PLOD2 GBM cells. Subsequently, we stimulated peripheral blood neutrophils with these supernatants and determined neutrophil survival as well as the release of MMP9—both indicators of a pro-tumor neutrophil phenotype ( A). The results showed that the sh-control SN from H4 cells prolonged the survival of neutrophils at 24 h post-stimulation ( B). This effect was significantly lower upon stimulation with sh-PLOD2 H4 SN ( B). To confirm these findings, we additionally stimulated neutrophils with SN from a second GBM cell line (U251). Similar to the H4 SN, the sh-control U251 SN prolonged neutrophil survival while sh-PLOD2 U251 SN had a significantly weaker effect ( B). To test whether GBM cells induce neutrophils to release MMP9, we stimulated neutrophils with sh-control or sh-PLOD2 SN for 1 h and determined MMP9 release by gelatin zymography. We found that the sh-control SN from both H4 and U251 cells induced neutrophils to release MMP9 ( C,D). The release of MMP9 by neutrophils was significantly lower upon stimulation with sh-PLOD2 SN ( C,D). GBM SN without neutrophils had only negligible levels of MMP9 (data not shown). To exclude potential clonal effects, we repeated this set of studies with SN derived from different sh-PLOD2 clones—for both H4 and U251 cells—and obtained similar results . Taken together these findings indicate that GBM cells release soluble factors via PLOD2, which stimulate neutrophils to acquire a tumor-promoting phenotype. To test the clinical relevance of these findings, we stained GBM tissues against the neutrophilic marker CD66b. The patients were subsequently divided into four groups according to the combined expression of CD66b and PLOD2: CD66b low /PLOD2 low , CD66b low /PLOD2 high , CD66b high /PLOD2 low and CD66b high /PLOD2 high . Kaplan–Meier analysis revealed that CD66b high /PLOD2 high patients had the shortest overall survival of all GBM patients ( E). These results were confirmed in a multivariate Cox regression model adjusted for age, KPS, therapy, resection efficiency and MGMT status where CD66b high /PLOD2 high patients had a significantly increased hazard ratio compared to the other groups of GBM patients (HR = 1.703, CI [95%] = 1.067–2.720, p = 0.026) ( F).
Previous studies by Xu et al. using the TCGA database showed that high gene expression of PLOD2 is significantly associated with a poor outcome in GBM patients . Here, we investigated whether the protein levels of PLOD2 in tumor tissue are associated with overall survival (OS) or progression-free survival (PFS) of GBM patients with confirmed IDH wild-type (IDH WT) status. To this end, the levels of PLOD2 were assessed by immunohistochemistry (see Material and Methods section) in two independent patient cohorts. PLOD2 expression was subsequently dichotomized into “low“ and “high“ based on the median-split method. The survival curves were plotted according to the Kaplan–Meier method and the statistical significance was assessed with the log-rank test. In the Hannover cohort, GBM patients with high tumor levels of PLOD2 (PLOD2 high ) had a significantly shorter OS compared to patients with low levels of PLOD2 (PLOD2 low ) ( p = 0.020; log-rank) ( A). These findings were confirmed in the Magdeburg cohort of GBM patients ( p < 0.001; log-rank) ( B). In both cohorts, PLOD2 high patients had a shorter PFS compared to their PLOD2 low counterparts, but statistical significance was only reached in the Magdeburg cohort ( p = 0.001; log-rank) ( C, D). We further analyzed the OS and PFS of IDH WT GBM patients using Cox proportional-hazard models adjusted for factors known to influence the patients’ outcome, such as age , Karnofsky Performance Scale (KPS) , extent of surgical resection , therapy and MGMT methylation status . Initial analysis of the time-dependent covariate (T_COV_) for PLOD2 showed that the proportional hazard assumption of these models had been satisfied: Hannover cohort_OS: p = 0.272; Magdeburg cohort_OS: p = 0.448; Hannover cohort_PFS: p = 0.440; Magdeburg cohort_PFS: p = 0.402. In both cohorts, high expression of PLOD2 predicted poor OS in GBM patients (Hannover cohort: HR = 1.401, CI [95%] = 1.009–1.946, p = 0.044; Magdeburg cohort: HR = 1.493, CI [95%] = 1.042–2.140, p = 0.029) ( A). For PFS, PLOD2 high patients had an increased hazard ratio compared to PLOD2 low patients but statistical significance was only reached in the Magdeburg cohort (HR = 1.645, CI [95%] = 1.040–2.603, p = 0.033) ( B). These data indicate that PLOD2 could serve as an independent prognostic biomarker, at least for the overall survival of GBM patients.
Using U87 and U251 GBM cells, recent studies showed that PLOD2 enhanced the aggressiveness of GBM cells by promoting tumor invasion . Here we investigated the effect of PLOD2 on the biology and functions of GBM using the H4 GBM cell line. To this end, the cells were stably transfected with a sh-RNA plasmid to downregulate the levels of PLOD2 (sh-PLOD2) or with a control plasmid (sh-control) (see Material and Methods section). All functional assays were performed in the absence of the selection antibiotic puromycin. Control western blot analysis confirmed that PLOD2 knock-downs remained stable until at least day 10, which was the last time point of the longest assay . Tumor invasion was assessed by the degree of “gap” closure (red line) in a 3D collagen matrix, using the Oris TM system ( A). The results showed that PLOD2 knock-down significantly reduced the invasiveness of H4 GBM cells ( B). We additionally determined the activity of matrix metalloproteases (MMP2 and MMP9), since MMPs are critical for tumor invasion in many types of cancer, including GBM . To this end, sh-control and sh-PLOD2 GBM cells were incubated in a culture medium and the supernatants were collected 48 h later. A culture medium without cells was used as a control. We found that H4 GBM cells released MMP2 but only negligible levels of MMP9 ( C). Importantly, PLOD2 knock-down cells released significantly lower levels of MMP2 compared to their control-transfected counterparts ( D). These findings indicate that PLOD2 enhances the invasiveness of GBM cells, possibly via MMP2. In further studies, we investigated the role of PLOD2 in GBM proliferation by assessing the metabolic activity of the transfected H4 cells using the MTT assay. To this end, the concentration of metabolized MTT was measured at different time points in one setup with 2000 cells and another with 4000 cells. The results showed that PLOD2 knock-down decreased the metabolic activity of H4 cells in both setups ( A,B). However, statistical significance was only reached for certain time points. We additionally determined the anchorage-independent growth of transfected GBM cells by allowing the cells to form colonies in low-gelling agarose for 10 days ( C). We found that PLOD2 knock-down cells formed significantly fewer colonies than their sh-control counterparts ( D). Taken together, these data indicate that PLOD2 promotes the invasion, proliferation and anchorage-independent growth of GBM cells.
As shown above, PLOD2 modulates the biological functions of GBM cells. To obtain further insight into the molecular mechanisms downstream of PLOD2 in GBM cells, we assessed by western blot the protein expression of several markers associated with tumor proliferation and invasion, such as Catenin D1, CD44, CD99, CDK6, EGFR, HIF1-beta, Integrin beta-1, MT1-MMP and PRAS40. We found that the levels of Catenin D1, CD44, CD99 and MT1-MMP were significantly lower in PLOD2 knock-down cells compared to their control-transfected counterparts ( A–E).
Accumulating evidence indicates that the GBM microenvironment contains significant numbers of neutrophils and that high neutrophil infiltration is associated with poor outcomes in GBM patients (reviewed in ). Furthermore, very recent studies found a correlation between high tumor levels of PLOD2 and high neutrophil infiltration in cervical and hepatocellular carcinoma tissues. Based on these findings, we hypothesized that GBM cells modulate the biology and functions of neutrophils via PLOD2. To test this hypothesis, we produced conditioned supernatants (SN) from sh-control and sh-PLOD2 GBM cells. Subsequently, we stimulated peripheral blood neutrophils with these supernatants and determined neutrophil survival as well as the release of MMP9—both indicators of a pro-tumor neutrophil phenotype ( A). The results showed that the sh-control SN from H4 cells prolonged the survival of neutrophils at 24 h post-stimulation ( B). This effect was significantly lower upon stimulation with sh-PLOD2 H4 SN ( B). To confirm these findings, we additionally stimulated neutrophils with SN from a second GBM cell line (U251). Similar to the H4 SN, the sh-control U251 SN prolonged neutrophil survival while sh-PLOD2 U251 SN had a significantly weaker effect ( B). To test whether GBM cells induce neutrophils to release MMP9, we stimulated neutrophils with sh-control or sh-PLOD2 SN for 1 h and determined MMP9 release by gelatin zymography. We found that the sh-control SN from both H4 and U251 cells induced neutrophils to release MMP9 ( C,D). The release of MMP9 by neutrophils was significantly lower upon stimulation with sh-PLOD2 SN ( C,D). GBM SN without neutrophils had only negligible levels of MMP9 (data not shown). To exclude potential clonal effects, we repeated this set of studies with SN derived from different sh-PLOD2 clones—for both H4 and U251 cells—and obtained similar results . Taken together these findings indicate that GBM cells release soluble factors via PLOD2, which stimulate neutrophils to acquire a tumor-promoting phenotype. To test the clinical relevance of these findings, we stained GBM tissues against the neutrophilic marker CD66b. The patients were subsequently divided into four groups according to the combined expression of CD66b and PLOD2: CD66b low /PLOD2 low , CD66b low /PLOD2 high , CD66b high /PLOD2 low and CD66b high /PLOD2 high . Kaplan–Meier analysis revealed that CD66b high /PLOD2 high patients had the shortest overall survival of all GBM patients ( E). These results were confirmed in a multivariate Cox regression model adjusted for age, KPS, therapy, resection efficiency and MGMT status where CD66b high /PLOD2 high patients had a significantly increased hazard ratio compared to the other groups of GBM patients (HR = 1.703, CI [95%] = 1.067–2.720, p = 0.026) ( F).
An increased effort has been made to identify the cellular/molecular factors that modulate the pathophysiology of GBM and that could provide information regarding diagnosis, prognosis and therapy in this type of cancer. PLOD2 is a promising biomarker and a target for cancer therapy, but its exact role in GBM still requires characterization. Several studies found an association between PLOD2 overexpression and poor outcome in multiple types of cancer, such as sarcoma , breast cancer , hepatocellular carcinoma and bladder cancer . Previous studies on a small ( n = 28) cohort of GBM patients suggested that PLOD2 may serve as a biomarker in this type of cancer . Song et al. found an association between high PLOD2 expression and poor outcomes in glioma patients. However, their study did not distinguish between high grade and low grade gliomas . In a comprehensive study on glioma and GBM patients, Xu et al. demonstrated that increasing PLOD2 protein levels are associated with increasing tumor grade. Furthermore, the gene expression of PLOD2 was significantly higher in GBM than in healthy tissues and correlated with overall and progression-free survival . Using two independent cohorts of IDH WT GBM patients, our study shows that high protein levels of PLOD2 (PLOD2 high ) are significantly associated with and predicted for poor overall survival of these patients. Together with the study by Xu et al., these findings indicate that PLOD2 could be a robust biomarker for the survival of GBM patients. We next sought to characterize the biological functions of PLOD2 in GBM cells. We found that PLOD2 promoted the invasiveness of H4 GBM cells. These data are in line with previous studies showing that PLOD2 modulates the migration and invasion of glioma cells. Specifically, Song et al. showed that PLOD2 knock-down suppressed the migration and invasion of U87 and U251 GBM cells, while Xu et al. showed in the same cell lines that the depletion of PLOD2 decreased invasion in vitro and in vivo, possibly by remodeling the stiffness of the ECM and decreasing the focal adhesion plaques . We additionally found that PLOD2 promoted the release of ECM-degrading MMP2—a mechanism associated with enhanced invasiveness and worse outcomes in different types of cancer, including glioma . Furthermore, we demonstrate that PLOD2 promotes the metabolic activity and the anchorage-independent growth of GBM cells. The effect of PLOD2 on the anchorage-independent tumor growth was especially striking since PLOD2 knock-down led to almost a complete inhibition of colony formation in GBM cells. Together, these findings are of particular importance for the pathophysiology of GBM, since high tumor invasiveness into the adjacent brain tissue and rapid growth are the main reasons why these tumors remain incurable at present. It should be mentioned at this point, that regulation of the ECM and tumor invasion is extremely complex. PLOD2 alone seems to play multiple roles in this process, since it can both degrade the basement membrane via MMP2 release (our own data), as well as induce collagen crosslinking/stiffening, thereby creating a “highway” for local invasion and activating different signaling pathways by mechanotransduction . Furthermore, as elegant studies by Georgescu et al. recently showed, there are many other factors modulating the ECM program in the microenvironment of GBM . Thus, the mechanisms involved in GBM invasion and the exact role of PLOD2 in this process still require further characterization. Previous studies found an association between PLOD2 and epithelial-mesenchymal transition (EMT), hypoxia-induced activation of PI3K-Akt signaling, as well as FAK phosphorylation in GBM cells . Here, we found that PLOD2 knock-down cells had decreased levels of MT1-MMP, which is known to be a key regulator of cell migration and invasion in GBM . Furthermore, MT1-MMP is an important activator of MMP2 leading to an enhanced invasion and progression in different tumor types including GBM . These findings support our data on the PLOD2-mediated release of MMP2 in GBM cells (see above). We additionally found that PLOD2 knock-down decreased the levels of CD44 indicating that PLOD2 regulates CD44 expression. CD44 is known to promote tumor formation through interactions with the tumor microenvironment and is involved in various cellular processes including invasion, proliferation and apoptosis in many types of cancer including GBM (reviewed in ). Increased CD44 expression was associated with worse survival in GBM . By which mechanisms PLOD2 affects CD44 expression in GBM remains unclear. However, previous studies in laryngeal carcinoma suggested that PLOD2 enhanced CD44 expression via activation of the Wnt-signaling pathway , which may be a possible explanation for GBM as well. Our study further found that PLOD2 knock-down decreased the levels of CD99 in GBM cells. CD99 is known to alter the structure of the cytoskeleton, thus facilitating cell migration . Indeed, studies on the GBM cell line U87 showed that CD99 overexpression increased the migration and invasion of these cells . The exact mechanisms of CD99 regulation by PLOD2 are, however, currently unknown and remain to be characterized in future studies. Finally, our data showed that Catenin D1 levels were lower upon PLOD2 knock-down. Catenin D1 has been previously linked to oncogenic signaling pathways important for anchorage-independent cell growth . This supports our findings that PLOD2 knock-down cells were almost completely unable to form colonies in soft agar clonogenic assays. Increasing evidence suggests that the interplay between tumor cells and the immune system is a key modulator of tumor biology and determines cancer pathogenesis and progression. Recent studies found a significant correlation between PLOD2 and infiltrating immune cells including neutrophils in cervical, hepatocellular and lung cancer . These findings suggest that PLOD2 is an important modulator of the tumor immune microenvironment. The GBM microenvironment contains high numbers of infiltrating neutrophil granulocytes , which were found to associate with a poor outcome in GBM patients (and reviewed in ). Neutrophils release various factors in the microenvironment, which can promote tumor progression. For instance, neutrophils have strong pro-angiogenetic activity via the release of MMP9 and vascular endothelial growth factor (VEGF). Additionally, neutrophils promote tumor motility, migration and invasion via the release of neutrophil elastase, cathepsin G, proteinase 3, MMP8 and MMP9. Under physiological conditions, neutrophils rapidly undergo apoptosis. However, their lifespan can be prolonged by tumor-derived factors resulting in enhanced local inflammation and, ultimately, tumor progression . Our study shows that PLOD2 controls the production of (currently unidentified) soluble factors by GBM cells, which subsequently enhance neutrophil survival and the release of MMP9. Moreover, patients with synchronous high expression of PLOD2 and the neutrophilic marker CD66b had a significantly shorter overall survival compared to the other groups of GBM patients. These data suggest that PLOD2 modulates the immune microenvironment of GBM leading to the progression of this cancer. All of the above supports the fact that PLOD2 promotes GBM progression and, thus, might serve as a potential therapeutic target. Several pharmacological inhibitors of PLOD2 are available at present. In particular, Minoxidil was found to reduce sarcoma migration and metastasis in vitro and in vivo by inhibiting PLOD family members . Similarly, inhibition of PLOD2 by Minoxidil reduced tumor migration in lung carcinoma and might prevent metastasis in this type of cancer . Interestingly, several studies additionally found that Minoxidil enhanced drug delivery, including that of Temozolomide, by permeabilizing the blood-tumor barrier (BTB) in GBM . It would be, therefore, tempting to speculate that GBM patients with high expression of PLOD2 might benefit from individualized therapies with PLOD2 inhibitors, such as Minoxidil. In summary, our study identifies PLOD2 as an independent prognostic biomarker in GBM. Furthermore, we demonstrate that PLOD2 mediates important biological functions of GBM cells, such as proliferation, invasion and anchorage-independent growth. We additionally show that PLOD2 regulates the expression of MMP2, MT1-MMP, CD44, CD99 and Catenin D1 in GBM cells. Importantly, we link PLOD2 with the immune modulation of neutrophils in the microenvironment of GBM. The main results of our study are summarized in . These findings contribute to a better understanding of GBM pathophysiology and may ultimately foster the development of novel therapeutic strategies against this type of cancer.
4.1. Study Subjects In this study, we retrospectively analyzed tissues from two independent cohorts of adult patients with histopathologically confirmed, newly diagnosed GBM. The tumors were clinically classified as primary GBM, as no lower grade glioma had been documented in the patient’s medical history. The patients in the Hannover cohort were treated at the Department of Neurosurgery, Nordstadt Hospital Hannover between 2004–2014 and had a median age of 66 years. The patients in the Magdeburg cohort were treated at the Department of Neurosurgery, University Hospital Magdeburg between 2005–2018 and had a median age of 64 years. All studies were carried out in accordance with the Declaration of Helsinki of 1975, revised in 2013 and approved by the ethics committees of the Medical School Hannover (Study Nr. 6864, 2015) and Otto-von-Guericke University Magdeburg (Study Nr. 146, 2019), respectively. The ethics committees additionally provided a waiver for the need for informed consent. The clinical characteristics of the patients including sex, post-operative Karnofsky Performance Scale (KPS), therapy, extent of surgical resection, MGMT methylation status and IDH mutation status are summarized in . The survival analysis was performed only on patients with confirmed IDH wild-type (WT) status. The clinical characteristics of the IDH WT patients are separately summarized in . 4.2. Tissue Microarrays (TMA): Immunhistochemistry and Scoring TMA blocks were built using the Arraymold kit E (Riverton, UT, USA) as previously described and cut into 2 µm sections. The sections were incubated with 667 ng/mL PLOD2-specific polyclonal antibodies (Proteintech Europe, Manchester, UK) or 500 ng/mL anti-human CD66b antibodies (BioLegend, San Diego, CA, USA) at 4 °C overnight. Secondary and colorimetric reactions were performed using the UltraVision TM Detection System according to the manufacturer’s instructions (Thermo Scientific, Freemont, CA, USA). Nuclei were counterstained with Haematoxylin (Carl Roth, Karlsruhe, Germany) and the sections were covered with Mountex ® embedding medium (Medite, Burgdorf, Germany). All stained TMAs were digitalized with an Aperio VERSA 8 high-resolution whole-slide scanner and the digital images were viewed with the Aperio ImageScope software (Leica Biosystems, Nussloch, Germany). Authors C.A.D., H.S. and N.K. independently performed blinded histological analysis. PLOD2 exhibited mainly a cytoplasmic subcellular localization. The expression intensity of the marker was categorized as “weak”, “medium” or “strong” and assigned 1, 2, or 3 points, respectively . As a number of samples exhibited heterogeneous staining, the expression was subsequently graded using the H-Score according to the formula: (1 × X) + (2 × Y) + (3 × Z), where X + Y + Z = 100% of the total tumor area (1) CD66b (neutrophilic marker) was assessed by counting the number of positive cells at 20× magnification in at least two different fields per TMA spot. Samples with an average of ≤5 cells/field were considered as “CD66b low ” and samples with >5 cells/field as “CD66b high ” . 4.3. Cell Lines and Stable Transfection Both H4 and U251 cell lines were a kind gift from Prof A. Temme (University Hospital Dresden) but are also commercially available. The main characteristics of these cells are shown in . The cells were cultured in Dulbecco’s Modified Eagle Medium (Gibco ® DMEM; Thermo Fisher Scientific, Dreieich, Germany) supplemented with 10% fetal calf serum (FCS; Pan Biotech, Aidenbach, Germany), and 1% Penicillin-Streptomycin (Gibco ® DMEM; Thermo Fisher Scientific). The cells were transfected with the OmicsLink TM shRNA clone HSH013271-nH1-b against PLOD2 or with CSHCTR001-nH1 as a control (both from GeneCopoeia, Rockville, MD, USA) using Panfect A-plus transfection reagent (Pan Biotech) according to the manufacturer’s protocol. Transfected cells were selected with 1 µg/mL Puromycin (InvivoGen, Toulouse, France) and then maintained in cell culture medium containing 0.3 µg/mL Puromycin. The efficiency of PLOD2 knock-down (sh-PLOD2) compared to control transfection (sh-control) was assessed both at protein and mRNA levels . Based on these results, clone 6 (C6) for H4 and clone 3 (C3) for U251 cells were used in all subsequent experiments. To exclude potential clonal effects, selected experiments were performed using clone 4 (C4) for H4 and clone 1 (C1) for U251 cells . In all figures depicting in vitro studies, “+” indicates the sample where the respective cells (either sh-PLOD2 or sh-control) were used. 4.4. SDS-PAGE and Western Blot GBM cells were lysed with a commercially available buffer containing Triton X-100 and protease/phosphatase inhibitors (both from Cell Signaling Technology, Frankfurt am Main, Germany). Cell debris was removed by centrifugation and the lysates were incubated with an SDS-Loading buffer containing 4% glycerin, 0.8% SDS, 1.6% beta-mercaptoethanol and 0.04% bromophenol blue (all from Carl Roth). Samples were separated by SDS-PAGE followed by transfer to Immobilon-P (Merck Millipore) or Roti ® -Fluoro (Carl Roth) PVDF membranes. The membranes were incubated with the following primary antibodies: anti-Catenin D1, anti-CD44, anti-CD99 (from Cell Signaling Technology), anti-MT1-MMP and anti-PLOD2 (from Proteintech) overnight at 4 °C. Secondary reactions were performed for 1h at room temperature using HRP-, AlexaFluor ® 488- or AlexaFluor ® 647-coupled antibodies (all from Cell Signaling Technology). All antibodies were diluted as recommended by the respective manufacturer using the SignalBoost™ Immunoreaction Enhancer Kit (Merck Millipore). Signal detection was performed on a ChemoStar imaging system (Intas Science Imaging, Göttingen, Germany). 4.5. Gene Expression Analysis The mRNA from sh-PLOD2 and sh-control GBM cells was isolated with the InnuPREP RNA Mini Kit 2.0 (Analytik Jena AG, Jena, Germany) according to the manufacturer’s instructions. Subsequently, reverse transcription was performed with the LunaScript RT SuperMix Kit (New England Biolabs, Frankfurt am Main, Germany). The samples were incubated with primers against PLOD2 or GAPDH in the presence of Luna Universal qPCR Mix (New England Biolabs). The following primers were used: PLOD2 forward 5′- CATGGACACAGGATAATGGCTG-3′ PLOD2 reverse 5′-AGGGGTTGGTTGCTCAATAAAA-3′ GAPDH forward 5′-AGGGCTGCTTTTAACTCTGGT-3′ GAPDH reverse 5′-CCCCACTTGATTTTGGAGGGA-3′ 4.6. MTT Assay GBM cells were seeded at a density of 2000 cells/well and 4000 cells/well in 96-well plates. At the indicated time-points, fresh medium containing 10% MTT (3-(4, 5-dimethylthiazol-2-yl)-2, 5-diphenyltetrazolium bromide) (Carl Roth) was added and the samples were incubated for 4 h at 37 °C to allow for the formation of formazan crystals. After lysis with a solution containing isopropanol and hydrochloric acid (both from Carl Roth), colorimetric detection was performed at OD 540 -OD 690 on a TECAN plate reader (Tecan, Männedorf, Switzerland). 4.7. Soft Agar Clonogenic Assay Ninety-six-well plates were coated with 1% high-gelling agarose (Carl Roth). GBM cells (1000 cells/well) were mixed with low-gelling agarose (Carl Roth) at a final concentration of 0.3% and were added on top of the first layer. The low-gelling agarose was allowed to solidify for 1 h at 4 °C and culture medium was added to each well. The samples were incubated at 37 °C for 10 days with medium change every 3–4 days. The samples were subsequently stained with a solution containing 0.05% Crystal Violet (Carl Roth). Colonies with a diameter of at least 50 µm were counted using a BZ-X810 microscope (Keyence, Neu-Isenburg, Germany). 4.8. Invasion Assay The invasion of GBM cells was assessed with the ORIS TM cell invasion system (Platypus Technologies LLC, Madison, WI, USA) according to the manufacturer’s instructions. The GBM cells were allowed to invade for 72 h in a matrix containing 1 mg/mL collagen I. The degree of “gap-closure” was quantified with the ImageJ 1.48v software. 4.9. Gelatin Zymography The release of matrix metalloproteases (MMPs) by GBM cells was analyzed by gelatin zymography, as described previously . Briefly, 10 5 cells/mL were incubated at 37 °C in DMEM medium, supplemented as above. As serum-supplemented cell culture medium also contains MMPs, medium without cells was used as control. The supernatants were collected at 48 h and mixed with Zymogram sample buffer at a final concentration of 80 mM Tris pH 6.8, 1% SDS, 4% glycerol and 0.006% bromophenol blue. Proteins were separated by SDS-PAGE containing 0.2% gelatin 180 Bloom and then renatured in 2.5% Triton-X-100 for 1 h at room temperature. The enzymatic reaction was performed overnight at 37 °C in a buffer containing 50 mM Tris pH 7.5, 200 mM NaCl, 5 mM CaCl 2 and 1% Triton-X-100. The gels were stained with a solution containing 0.5% Coomassie blue, 30% methanol and 10% acetic acid for 1 h at room temperature. Finally, the gels were de-stained with 30% methanol and 10% acetic acid until the digested bands became visible. All chemicals were from Carl Roth (Karlsruhe, Germany). The gelatinolytic bands were quantified with ImageJ 1.48v software. The release of MMPs by neutrophils was assessed as above, except for using a different cell number (10 6 cells/mL) and duration of stimulation (1 h). 4.10. Isolation of Neutrophils from Peripheral Blood Diluted blood (1:1, v / v in phosphate buffered saline (PBS)) was subjected to density gradient centrifugation using Pancoll (Pan Biotech). The mononuclear cell fraction was discarded, and the neutrophil fraction was collected in a fresh test tube. Erythrocytes were removed by sedimentation with a solution containing 1% polyvinyl alcohol (Sigma-Aldrich, Burlington, MA, USA) and, subsequently, by lysis with pre-warmed Aqua Braun (B. Braun, Melsungen, Germany). The resulting neutrophils were cultured in DMEM medium supplemented as above. The purity of the neutrophil population after isolation was routinely >98%. 4.11. Apoptosis Assays Neutrophils (10 6 cells/mL) were stimulated as indicated and were stained 24 h later with FITC Annexin V/propidium iodide according to the manufacturer’s instructions (BioLegend). Quantification was performed with a BD FACSCanto II flow cytometer (BD Biosciences, Heidelberg, Germany). 4.12. Statistical Analysis Clinical data were analyzed with the SPSS statistical software version 26 (IBM Corporation). Survival curves (5-year, 3-year or 1-year cut-off) were plotted according to the Kaplan–Meier method. Significance was initially tested by univariate analysis using the log-rank test. Multivariate analysis was subsequently used to determine the prognostic value of selected variables using Cox’s proportional hazard linear regression models adjusted for age, Karnofsky Performance Scale (KPS), therapy, extent of surgical resection and MGMT methylation status. The in vitro data were analyzed with the paired student’s t -test. In all studies, the level of significance was set at p ≤ 0.05.
In this study, we retrospectively analyzed tissues from two independent cohorts of adult patients with histopathologically confirmed, newly diagnosed GBM. The tumors were clinically classified as primary GBM, as no lower grade glioma had been documented in the patient’s medical history. The patients in the Hannover cohort were treated at the Department of Neurosurgery, Nordstadt Hospital Hannover between 2004–2014 and had a median age of 66 years. The patients in the Magdeburg cohort were treated at the Department of Neurosurgery, University Hospital Magdeburg between 2005–2018 and had a median age of 64 years. All studies were carried out in accordance with the Declaration of Helsinki of 1975, revised in 2013 and approved by the ethics committees of the Medical School Hannover (Study Nr. 6864, 2015) and Otto-von-Guericke University Magdeburg (Study Nr. 146, 2019), respectively. The ethics committees additionally provided a waiver for the need for informed consent. The clinical characteristics of the patients including sex, post-operative Karnofsky Performance Scale (KPS), therapy, extent of surgical resection, MGMT methylation status and IDH mutation status are summarized in . The survival analysis was performed only on patients with confirmed IDH wild-type (WT) status. The clinical characteristics of the IDH WT patients are separately summarized in .
TMA blocks were built using the Arraymold kit E (Riverton, UT, USA) as previously described and cut into 2 µm sections. The sections were incubated with 667 ng/mL PLOD2-specific polyclonal antibodies (Proteintech Europe, Manchester, UK) or 500 ng/mL anti-human CD66b antibodies (BioLegend, San Diego, CA, USA) at 4 °C overnight. Secondary and colorimetric reactions were performed using the UltraVision TM Detection System according to the manufacturer’s instructions (Thermo Scientific, Freemont, CA, USA). Nuclei were counterstained with Haematoxylin (Carl Roth, Karlsruhe, Germany) and the sections were covered with Mountex ® embedding medium (Medite, Burgdorf, Germany). All stained TMAs were digitalized with an Aperio VERSA 8 high-resolution whole-slide scanner and the digital images were viewed with the Aperio ImageScope software (Leica Biosystems, Nussloch, Germany). Authors C.A.D., H.S. and N.K. independently performed blinded histological analysis. PLOD2 exhibited mainly a cytoplasmic subcellular localization. The expression intensity of the marker was categorized as “weak”, “medium” or “strong” and assigned 1, 2, or 3 points, respectively . As a number of samples exhibited heterogeneous staining, the expression was subsequently graded using the H-Score according to the formula: (1 × X) + (2 × Y) + (3 × Z), where X + Y + Z = 100% of the total tumor area (1) CD66b (neutrophilic marker) was assessed by counting the number of positive cells at 20× magnification in at least two different fields per TMA spot. Samples with an average of ≤5 cells/field were considered as “CD66b low ” and samples with >5 cells/field as “CD66b high ” .
Both H4 and U251 cell lines were a kind gift from Prof A. Temme (University Hospital Dresden) but are also commercially available. The main characteristics of these cells are shown in . The cells were cultured in Dulbecco’s Modified Eagle Medium (Gibco ® DMEM; Thermo Fisher Scientific, Dreieich, Germany) supplemented with 10% fetal calf serum (FCS; Pan Biotech, Aidenbach, Germany), and 1% Penicillin-Streptomycin (Gibco ® DMEM; Thermo Fisher Scientific). The cells were transfected with the OmicsLink TM shRNA clone HSH013271-nH1-b against PLOD2 or with CSHCTR001-nH1 as a control (both from GeneCopoeia, Rockville, MD, USA) using Panfect A-plus transfection reagent (Pan Biotech) according to the manufacturer’s protocol. Transfected cells were selected with 1 µg/mL Puromycin (InvivoGen, Toulouse, France) and then maintained in cell culture medium containing 0.3 µg/mL Puromycin. The efficiency of PLOD2 knock-down (sh-PLOD2) compared to control transfection (sh-control) was assessed both at protein and mRNA levels . Based on these results, clone 6 (C6) for H4 and clone 3 (C3) for U251 cells were used in all subsequent experiments. To exclude potential clonal effects, selected experiments were performed using clone 4 (C4) for H4 and clone 1 (C1) for U251 cells . In all figures depicting in vitro studies, “+” indicates the sample where the respective cells (either sh-PLOD2 or sh-control) were used.
GBM cells were lysed with a commercially available buffer containing Triton X-100 and protease/phosphatase inhibitors (both from Cell Signaling Technology, Frankfurt am Main, Germany). Cell debris was removed by centrifugation and the lysates were incubated with an SDS-Loading buffer containing 4% glycerin, 0.8% SDS, 1.6% beta-mercaptoethanol and 0.04% bromophenol blue (all from Carl Roth). Samples were separated by SDS-PAGE followed by transfer to Immobilon-P (Merck Millipore) or Roti ® -Fluoro (Carl Roth) PVDF membranes. The membranes were incubated with the following primary antibodies: anti-Catenin D1, anti-CD44, anti-CD99 (from Cell Signaling Technology), anti-MT1-MMP and anti-PLOD2 (from Proteintech) overnight at 4 °C. Secondary reactions were performed for 1h at room temperature using HRP-, AlexaFluor ® 488- or AlexaFluor ® 647-coupled antibodies (all from Cell Signaling Technology). All antibodies were diluted as recommended by the respective manufacturer using the SignalBoost™ Immunoreaction Enhancer Kit (Merck Millipore). Signal detection was performed on a ChemoStar imaging system (Intas Science Imaging, Göttingen, Germany).
The mRNA from sh-PLOD2 and sh-control GBM cells was isolated with the InnuPREP RNA Mini Kit 2.0 (Analytik Jena AG, Jena, Germany) according to the manufacturer’s instructions. Subsequently, reverse transcription was performed with the LunaScript RT SuperMix Kit (New England Biolabs, Frankfurt am Main, Germany). The samples were incubated with primers against PLOD2 or GAPDH in the presence of Luna Universal qPCR Mix (New England Biolabs). The following primers were used: PLOD2 forward 5′- CATGGACACAGGATAATGGCTG-3′ PLOD2 reverse 5′-AGGGGTTGGTTGCTCAATAAAA-3′ GAPDH forward 5′-AGGGCTGCTTTTAACTCTGGT-3′ GAPDH reverse 5′-CCCCACTTGATTTTGGAGGGA-3′
GBM cells were seeded at a density of 2000 cells/well and 4000 cells/well in 96-well plates. At the indicated time-points, fresh medium containing 10% MTT (3-(4, 5-dimethylthiazol-2-yl)-2, 5-diphenyltetrazolium bromide) (Carl Roth) was added and the samples were incubated for 4 h at 37 °C to allow for the formation of formazan crystals. After lysis with a solution containing isopropanol and hydrochloric acid (both from Carl Roth), colorimetric detection was performed at OD 540 -OD 690 on a TECAN plate reader (Tecan, Männedorf, Switzerland).
Ninety-six-well plates were coated with 1% high-gelling agarose (Carl Roth). GBM cells (1000 cells/well) were mixed with low-gelling agarose (Carl Roth) at a final concentration of 0.3% and were added on top of the first layer. The low-gelling agarose was allowed to solidify for 1 h at 4 °C and culture medium was added to each well. The samples were incubated at 37 °C for 10 days with medium change every 3–4 days. The samples were subsequently stained with a solution containing 0.05% Crystal Violet (Carl Roth). Colonies with a diameter of at least 50 µm were counted using a BZ-X810 microscope (Keyence, Neu-Isenburg, Germany).
The invasion of GBM cells was assessed with the ORIS TM cell invasion system (Platypus Technologies LLC, Madison, WI, USA) according to the manufacturer’s instructions. The GBM cells were allowed to invade for 72 h in a matrix containing 1 mg/mL collagen I. The degree of “gap-closure” was quantified with the ImageJ 1.48v software.
The release of matrix metalloproteases (MMPs) by GBM cells was analyzed by gelatin zymography, as described previously . Briefly, 10 5 cells/mL were incubated at 37 °C in DMEM medium, supplemented as above. As serum-supplemented cell culture medium also contains MMPs, medium without cells was used as control. The supernatants were collected at 48 h and mixed with Zymogram sample buffer at a final concentration of 80 mM Tris pH 6.8, 1% SDS, 4% glycerol and 0.006% bromophenol blue. Proteins were separated by SDS-PAGE containing 0.2% gelatin 180 Bloom and then renatured in 2.5% Triton-X-100 for 1 h at room temperature. The enzymatic reaction was performed overnight at 37 °C in a buffer containing 50 mM Tris pH 7.5, 200 mM NaCl, 5 mM CaCl 2 and 1% Triton-X-100. The gels were stained with a solution containing 0.5% Coomassie blue, 30% methanol and 10% acetic acid for 1 h at room temperature. Finally, the gels were de-stained with 30% methanol and 10% acetic acid until the digested bands became visible. All chemicals were from Carl Roth (Karlsruhe, Germany). The gelatinolytic bands were quantified with ImageJ 1.48v software. The release of MMPs by neutrophils was assessed as above, except for using a different cell number (10 6 cells/mL) and duration of stimulation (1 h).
Diluted blood (1:1, v / v in phosphate buffered saline (PBS)) was subjected to density gradient centrifugation using Pancoll (Pan Biotech). The mononuclear cell fraction was discarded, and the neutrophil fraction was collected in a fresh test tube. Erythrocytes were removed by sedimentation with a solution containing 1% polyvinyl alcohol (Sigma-Aldrich, Burlington, MA, USA) and, subsequently, by lysis with pre-warmed Aqua Braun (B. Braun, Melsungen, Germany). The resulting neutrophils were cultured in DMEM medium supplemented as above. The purity of the neutrophil population after isolation was routinely >98%.
Neutrophils (10 6 cells/mL) were stimulated as indicated and were stained 24 h later with FITC Annexin V/propidium iodide according to the manufacturer’s instructions (BioLegend). Quantification was performed with a BD FACSCanto II flow cytometer (BD Biosciences, Heidelberg, Germany).
Clinical data were analyzed with the SPSS statistical software version 26 (IBM Corporation). Survival curves (5-year, 3-year or 1-year cut-off) were plotted according to the Kaplan–Meier method. Significance was initially tested by univariate analysis using the log-rank test. Multivariate analysis was subsequently used to determine the prognostic value of selected variables using Cox’s proportional hazard linear regression models adjusted for age, Karnofsky Performance Scale (KPS), therapy, extent of surgical resection and MGMT methylation status. The in vitro data were analyzed with the paired student’s t -test. In all studies, the level of significance was set at p ≤ 0.05.
|
Taylor Swift does not boost face recognition in reaction time-based Concealed Information Test: investigating target-familiarity effects | 3cf4384e-69b5-46f9-bdbc-e865390ec85f | 11522165 | Forensic Medicine[mh] | We pre-registered both experiments on the open science framework (Experiment 1: https://osf.io/7ny5g ; Experiment 2: https://osf.io/zdu5x . The Ethics Review Committee of the Faculty approved both experiments (approval codes: OZL_231_140_12_2020_S7 and 231_140_12_2020_S8). The Inquisit scripts, pilot data, and data are available here: https://osf.io/j6vuf/?view_only= . We cannot share all stimuli because we do not have permission of the depicted individuals and due to copyright reasons (familiar targets). Deviations from pre-registration There are several deviations from the pre-registration: first, in addition to the analyses across the two probes, we conducted separate analyses for the thief and the victim, in line with earlier work (Sauerland et al., , ). Second, we pre-registered Bayesian analyses for Experiment 1 but by oversight did not include it in the pre-registration for Experiment 2. We conducted the Bayesian analyses for both experiments for consistency. Participants Power analyses For Experiment 1, we conducted a power analysis using MorePower 6.0.4 (Campbell & Thompson, ). We based the estimated effect size on Suchotzki et al. ( ; Experiment 2) who reported an interaction effect between stimulus type and target familiarity of η p 2 = .25. We set the alpha error probability to .05, and power to .95. This led to a required sample size of 44 participants (22 per condition), or 52 participants (26 per condition) when accounting for dropouts and exclusions. However, because initial studies typically overestimate the true effect size (Camerer et al., ) and particularly interaction effects have been difficult to replicate, we aimed to test at least double this amount (i.e., N ≥ 104). For Experiment 2, we conducted a power analysis with G*Power 3.1.9.7 (Faul et al., ). Based on the null results of Experiment 1, we considered a lower effect size (i.e., a medium effect size of η p 2 = .06), set power to .95 and the α-error probability to .05. The power analysis returned a required sample size of 54 participants. Considering a likely dropout and exclusion rate of +/- 15% we aimed to test at least 64 participants or as many as we could within two months. Samples In Experiment 1, we tested 109 participants. We excluded 12 participants, because they had an error rate of more than 50% (Kleinberg & Verschuere, ), four participants because they answered more than one attention check question wrong (Sauerland et al., , ), and three participants dropped out during the experiment. The final sample consisted of 90 participants ( M age = 23.41, SD age = 3.88, range = 19–46). The unfamiliar-targets and the familiar-targets condition included 41 and 49 participants, respectively. Participants (75.6% female, 23.3% male, 1.1% non-binary) were bachelor students (51.1%), master students (45.6%), or indicated no university track (3.3%). Most participants were students at the Faculty of Psychology and Neuroscience (51.1%) or not a student at this University (36.7%). Participants’ most common native languages included German (31.1%), English (21.1%), Luxembourgish (14.4%), and Dutch (13.3%). In Experiment 2, we tested 97 participants. Ten participants dropped out and we excluded four participants for one of the following reasons: an error rate of more than 50% in response to the targets, completing the study twice, recognizing one of the actresses in the stimulus film from real life, or failing two of the three attention check questions. The final sample consisted of N = 83 participants ( M Age = 21.4 years, SD Age = 5.2, range = 17–54 years). Participants (75.9% females, 21.7% males, 2.4% non-binary) were mostly Bachelor students (90.4%) and Master students (4.8%). Most participants were students at the Faculty of Psychology and Neuroscience (86.7%) or not a student at this University (9.6%). The most common native languages were German (50.6%), Dutch (22.9%) and Greek (13.8%). Design We used a 2 (target familiarity: familiar vs. unfamiliar) x 2 (stimulus type: probe vs. irrelevants) mixed design and manipulated target familiarity (familiar vs. unfamiliar) between subjects and stimulus type (probe vs. irrelevants ) within subjects in both experiments. Materials Stimulus films As a base for the two stimulus films, we used the four films from Sauerland et al. . In these films, one of two protagonists was visible primarily from the front and one primarily from the side. For the current study, we cut the four films to receive two films that included frontal close-ups and some distant profile views of both protagonists. The two films show the same action with two women: on a square in a pedestrian zone, the future thief walks up to the future victim to ask for directions. In the next scene, the victim is sitting down and is busy on her phone. Sneaking up from behind, the thief steals the victim’s purse. The actors switched between roles of thief and victim in the two versions to avoid confounding effects due to the actor or role. Table contains an overview of the facial viewing times of the protagonists in each stimulus film, split into close-ups and distant shots of the face in a frontal and profile view. Facial stimuli Fourteen photographs included the two probes, four targets, and eight irrelevants. Two targets and four irrelevants each matched the general description of one of the probes, as determined in earlier research (Sauerland et al., ). The photographs showed faces from a frontal view from the collarbone up, with loose, open hair, without jewelry, accessories, or glasses and exhibited a neutral expression. All clothing was edited to be black. In the unfamiliar target condition, we randomly selected two targets from the non-probe photographs per probe. In the familiar target condition, the targets consisted of four celebrities. Celebrity faces To select the celebrities, we conducted two pilot studies. Our aim was to identify four celebrities that matched the general description of the probes (cf. Sauerland et al., ; Wells et al., ) and were generally well-known in a student population. For the first pilot, we selected 12 celebrities that matched the general description of the two probes (six for each probe). Thirty participants (18 women, 9 men, 2 non-binary, 1 prefer not to say; M age = 21.63, SD age = 2.48) sequentially viewed the celebrity faces on Qualtrics and indicated whether they knew them. Participants could either indicate celebrities’ names or where they knew them from. More than 90% of the participants were familiar with Selena Gomez and Emma Watson. The other celebrities reached only 66% familiarity or less or were found too old to match the 21-year old probes in hindsight (i.e., Jennifer Aniston, Scarlet Johansson). We therefore conducted a second pilot with another set of 12 celebrities. From this pilot, we selected Taylor Swift and Katy Perry as familiar targets, based on familiarity scores of 93% and 83%, respectively. Thus, the four familiar targets were Selena Gomez, Katy Perry, Emma Watson, and Taylor Swift. Once we had established the identity of the familiar targets, we searched the internet with the celebrities’ names as the search term on Google Images to obtain high quality photographs with an unobstructed facial frontal view. In Experiment 1, we edited the chosen images using Adobe Photoshop (2022) to match the appearance of the unfamiliar target photos. We adjusted the background to a solid white and cropped the photos from the shoulders be the same size as the unfamiliar stimuli. We edited out all jewelry and accessories and adjusted the clothes to black shirts. In Experiment 2, we further edited the familiar targets to look less distinctive. For example, celebrities wore heavy make-up and lighting was distinctive to photographs taken in a regulated room. We adjusted the colors and levels using the Camera Raw Filter in Photoshop to match the color grading to the pictures. Additionally, we reduced make-up intensity. Finally, we reduced the picture quality by lowering the resolution and adding noise to match the lower quality of the other pictures. Celebrity information (Experiment 2) To improve familiar target recognition in Experiment 2, participants received information about and photos of each of the celebrities at the beginning of the testing session, prior to the CIT protocol. For example, participants saw three photos of Taylor Swift alongside her occupation, her most known songs and the number of Grammy awards she won. Reaction time-based Concealed Information Test The RT-CIT of was programmed using Inquisit 6.5.2 Lab (Experiment 1) and Inquisit 6.6.1 Lab (Experiment 2). The data were recorded in milliseconds, using one joint CIT protocol for the thief and victim. All photos were 388 × 462 pixels in size. We followed a similar protocol as other RT-CIT studies with facial stimuli (e.g., Georgiadou et al., ; Sauerland et al., ). Participants received instructions to press the ‘L’ key on the keyboard as fast as possible in response to a facial stimulus, with the exception of the two targets. For these stimuli, they should press the ‘A’ key. We included the names of the celebrities in the instructions of the familiar targets condition. Participants then viewed the four targets for 30 s, accompanied by instructions to encode these faces. A practice block followed, showing each photo (2 × 2 targets, 2 × 4 irrelevants, and 2 × 1 probes) once. Above each facial stimuli, the question ‘Do you recognize this person?’ appeared. The labels ‘Yes’ and ‘No’ appeared to the left and right sides of each stimulus, respectively. Participants had 1500 ms to respond. After 800 ms, they received the warning ‘Too slow!’ Participants received the feedback ‘Wrong’ when they responded different than instructed. We randomly set the inter-stimulus interval between 250, 500, or 750 ms. A second, identical practice block followed if participants made more than one mistake or if they had responded too slow. Preceding the second practice block, participants received reminder instructions on how to respond. After this second practice block, participants continued to the actual task regardless of performance. Before the task started, the instructions appeared again, and participants viewed the targets for five more seconds. Similar to previous experiments, participants viewed the targets for a total of 35 s (Sauerland et al., ). During the actual task, every stimulus appeared 18 times, at a random sequence. The labels and feedback were identical to the practice blocks. In total, the actual task consisted of 252 trials (2 × 7 × 18). The RT-CIT of Experiment 2 was identical to the first experiment except that the participants had to complete three practice phases rather than one or two. During the first practice phase, the stimulus only disappeared once participants pressed the A or L key. This way, participants could control the pace of the test. They already received the ‘Wrong’ feedback if applicable. In the second practice phase, the stimuli disappeared after 1500 ms, if no key was pressed. The last practice trial was identical to the second practice trial except that participants additionally received feedback on their speed, with the words ‘Too slow!’ appearing on the screen after 800 ms. Participants viewed each of the 14 stimuli twice during each training phase. If a participant made too many mistakes during one of the practice phases, they had to repeat it once again. To make sure that the target photos were memorized, they were shown for five seconds and the instructions were repeated after each practice phase and before the actual task began. Follow-up photo display After completion of the RT-CIT task, participants viewed a photo recognition display that showed all 14 stimuli used, arranged in four rows, alternating between three and four photos. The 14 photographs included the two probes, four targets, and eight irrelevants. Two targets and four irrelevants each matched the general description of one of the probes, as determined in earlier research (Sauerland et al., ). The images were numbered (1, 2, 3, etc.). Participants indicated which individuals they recognized from the video. This allowed us to roughly determine if participants in the CIT conditions had explicit memory of the probes. Attention check Participants answered three forced-choice attention check questions with four or five answer options. The forced-choice questions were ‘Where did the actors first meet?’; ‘What was the victim doing when the handbag was stolen?’, and ‘What color was the stolen handbag?’ We excluded participants if they did not answer at least two of the three questions correctly. Procedure Recruitment occurred online, using the University’s research participation system Sona, social media sites such as Facebook, Whatsapp and Instagram, and other research platforms (e.g., SurveyCircle, ). Participants received instructions to complete the study on a laptop or PC and that they may have to download a plug-in. Participants could not take part if they had participated in one of the pilot studies or previous RT-CIT experiments in the department. After reading an information letter and providing consent, the software randomly allocated participants to one of two conditions (familiar vs. unfamiliar). Participants received instructions to watch the film closely and to pay close attention because they would answer questions about the film later. After the film, participants answered the attention check questions and in Experiment 2, participants in the familiar condition additionally read the information about the celebrities. Subsequently, participants completed the RT-CIT task. Participants then indicated the probes in the follow-up photo display and indicated whether they recognized any of the individuals in the study except for the celebrities from ‘real life’. Finally, participants were debriefed and thanked for their participation. The studies lasted approximately 20 to 30 min. SONA participants received credit for their participation. There are several deviations from the pre-registration: first, in addition to the analyses across the two probes, we conducted separate analyses for the thief and the victim, in line with earlier work (Sauerland et al., , ). Second, we pre-registered Bayesian analyses for Experiment 1 but by oversight did not include it in the pre-registration for Experiment 2. We conducted the Bayesian analyses for both experiments for consistency. Power analyses For Experiment 1, we conducted a power analysis using MorePower 6.0.4 (Campbell & Thompson, ). We based the estimated effect size on Suchotzki et al. ( ; Experiment 2) who reported an interaction effect between stimulus type and target familiarity of η p 2 = .25. We set the alpha error probability to .05, and power to .95. This led to a required sample size of 44 participants (22 per condition), or 52 participants (26 per condition) when accounting for dropouts and exclusions. However, because initial studies typically overestimate the true effect size (Camerer et al., ) and particularly interaction effects have been difficult to replicate, we aimed to test at least double this amount (i.e., N ≥ 104). For Experiment 2, we conducted a power analysis with G*Power 3.1.9.7 (Faul et al., ). Based on the null results of Experiment 1, we considered a lower effect size (i.e., a medium effect size of η p 2 = .06), set power to .95 and the α-error probability to .05. The power analysis returned a required sample size of 54 participants. Considering a likely dropout and exclusion rate of +/- 15% we aimed to test at least 64 participants or as many as we could within two months. Samples In Experiment 1, we tested 109 participants. We excluded 12 participants, because they had an error rate of more than 50% (Kleinberg & Verschuere, ), four participants because they answered more than one attention check question wrong (Sauerland et al., , ), and three participants dropped out during the experiment. The final sample consisted of 90 participants ( M age = 23.41, SD age = 3.88, range = 19–46). The unfamiliar-targets and the familiar-targets condition included 41 and 49 participants, respectively. Participants (75.6% female, 23.3% male, 1.1% non-binary) were bachelor students (51.1%), master students (45.6%), or indicated no university track (3.3%). Most participants were students at the Faculty of Psychology and Neuroscience (51.1%) or not a student at this University (36.7%). Participants’ most common native languages included German (31.1%), English (21.1%), Luxembourgish (14.4%), and Dutch (13.3%). In Experiment 2, we tested 97 participants. Ten participants dropped out and we excluded four participants for one of the following reasons: an error rate of more than 50% in response to the targets, completing the study twice, recognizing one of the actresses in the stimulus film from real life, or failing two of the three attention check questions. The final sample consisted of N = 83 participants ( M Age = 21.4 years, SD Age = 5.2, range = 17–54 years). Participants (75.9% females, 21.7% males, 2.4% non-binary) were mostly Bachelor students (90.4%) and Master students (4.8%). Most participants were students at the Faculty of Psychology and Neuroscience (86.7%) or not a student at this University (9.6%). The most common native languages were German (50.6%), Dutch (22.9%) and Greek (13.8%). Design We used a 2 (target familiarity: familiar vs. unfamiliar) x 2 (stimulus type: probe vs. irrelevants) mixed design and manipulated target familiarity (familiar vs. unfamiliar) between subjects and stimulus type (probe vs. irrelevants ) within subjects in both experiments. For Experiment 1, we conducted a power analysis using MorePower 6.0.4 (Campbell & Thompson, ). We based the estimated effect size on Suchotzki et al. ( ; Experiment 2) who reported an interaction effect between stimulus type and target familiarity of η p 2 = .25. We set the alpha error probability to .05, and power to .95. This led to a required sample size of 44 participants (22 per condition), or 52 participants (26 per condition) when accounting for dropouts and exclusions. However, because initial studies typically overestimate the true effect size (Camerer et al., ) and particularly interaction effects have been difficult to replicate, we aimed to test at least double this amount (i.e., N ≥ 104). For Experiment 2, we conducted a power analysis with G*Power 3.1.9.7 (Faul et al., ). Based on the null results of Experiment 1, we considered a lower effect size (i.e., a medium effect size of η p 2 = .06), set power to .95 and the α-error probability to .05. The power analysis returned a required sample size of 54 participants. Considering a likely dropout and exclusion rate of +/- 15% we aimed to test at least 64 participants or as many as we could within two months. In Experiment 1, we tested 109 participants. We excluded 12 participants, because they had an error rate of more than 50% (Kleinberg & Verschuere, ), four participants because they answered more than one attention check question wrong (Sauerland et al., , ), and three participants dropped out during the experiment. The final sample consisted of 90 participants ( M age = 23.41, SD age = 3.88, range = 19–46). The unfamiliar-targets and the familiar-targets condition included 41 and 49 participants, respectively. Participants (75.6% female, 23.3% male, 1.1% non-binary) were bachelor students (51.1%), master students (45.6%), or indicated no university track (3.3%). Most participants were students at the Faculty of Psychology and Neuroscience (51.1%) or not a student at this University (36.7%). Participants’ most common native languages included German (31.1%), English (21.1%), Luxembourgish (14.4%), and Dutch (13.3%). In Experiment 2, we tested 97 participants. Ten participants dropped out and we excluded four participants for one of the following reasons: an error rate of more than 50% in response to the targets, completing the study twice, recognizing one of the actresses in the stimulus film from real life, or failing two of the three attention check questions. The final sample consisted of N = 83 participants ( M Age = 21.4 years, SD Age = 5.2, range = 17–54 years). Participants (75.9% females, 21.7% males, 2.4% non-binary) were mostly Bachelor students (90.4%) and Master students (4.8%). Most participants were students at the Faculty of Psychology and Neuroscience (86.7%) or not a student at this University (9.6%). The most common native languages were German (50.6%), Dutch (22.9%) and Greek (13.8%). We used a 2 (target familiarity: familiar vs. unfamiliar) x 2 (stimulus type: probe vs. irrelevants) mixed design and manipulated target familiarity (familiar vs. unfamiliar) between subjects and stimulus type (probe vs. irrelevants ) within subjects in both experiments. Stimulus films As a base for the two stimulus films, we used the four films from Sauerland et al. . In these films, one of two protagonists was visible primarily from the front and one primarily from the side. For the current study, we cut the four films to receive two films that included frontal close-ups and some distant profile views of both protagonists. The two films show the same action with two women: on a square in a pedestrian zone, the future thief walks up to the future victim to ask for directions. In the next scene, the victim is sitting down and is busy on her phone. Sneaking up from behind, the thief steals the victim’s purse. The actors switched between roles of thief and victim in the two versions to avoid confounding effects due to the actor or role. Table contains an overview of the facial viewing times of the protagonists in each stimulus film, split into close-ups and distant shots of the face in a frontal and profile view. Facial stimuli Fourteen photographs included the two probes, four targets, and eight irrelevants. Two targets and four irrelevants each matched the general description of one of the probes, as determined in earlier research (Sauerland et al., ). The photographs showed faces from a frontal view from the collarbone up, with loose, open hair, without jewelry, accessories, or glasses and exhibited a neutral expression. All clothing was edited to be black. In the unfamiliar target condition, we randomly selected two targets from the non-probe photographs per probe. In the familiar target condition, the targets consisted of four celebrities. Celebrity faces To select the celebrities, we conducted two pilot studies. Our aim was to identify four celebrities that matched the general description of the probes (cf. Sauerland et al., ; Wells et al., ) and were generally well-known in a student population. For the first pilot, we selected 12 celebrities that matched the general description of the two probes (six for each probe). Thirty participants (18 women, 9 men, 2 non-binary, 1 prefer not to say; M age = 21.63, SD age = 2.48) sequentially viewed the celebrity faces on Qualtrics and indicated whether they knew them. Participants could either indicate celebrities’ names or where they knew them from. More than 90% of the participants were familiar with Selena Gomez and Emma Watson. The other celebrities reached only 66% familiarity or less or were found too old to match the 21-year old probes in hindsight (i.e., Jennifer Aniston, Scarlet Johansson). We therefore conducted a second pilot with another set of 12 celebrities. From this pilot, we selected Taylor Swift and Katy Perry as familiar targets, based on familiarity scores of 93% and 83%, respectively. Thus, the four familiar targets were Selena Gomez, Katy Perry, Emma Watson, and Taylor Swift. Once we had established the identity of the familiar targets, we searched the internet with the celebrities’ names as the search term on Google Images to obtain high quality photographs with an unobstructed facial frontal view. In Experiment 1, we edited the chosen images using Adobe Photoshop (2022) to match the appearance of the unfamiliar target photos. We adjusted the background to a solid white and cropped the photos from the shoulders be the same size as the unfamiliar stimuli. We edited out all jewelry and accessories and adjusted the clothes to black shirts. In Experiment 2, we further edited the familiar targets to look less distinctive. For example, celebrities wore heavy make-up and lighting was distinctive to photographs taken in a regulated room. We adjusted the colors and levels using the Camera Raw Filter in Photoshop to match the color grading to the pictures. Additionally, we reduced make-up intensity. Finally, we reduced the picture quality by lowering the resolution and adding noise to match the lower quality of the other pictures. Celebrity information (Experiment 2) To improve familiar target recognition in Experiment 2, participants received information about and photos of each of the celebrities at the beginning of the testing session, prior to the CIT protocol. For example, participants saw three photos of Taylor Swift alongside her occupation, her most known songs and the number of Grammy awards she won. Reaction time-based Concealed Information Test The RT-CIT of was programmed using Inquisit 6.5.2 Lab (Experiment 1) and Inquisit 6.6.1 Lab (Experiment 2). The data were recorded in milliseconds, using one joint CIT protocol for the thief and victim. All photos were 388 × 462 pixels in size. We followed a similar protocol as other RT-CIT studies with facial stimuli (e.g., Georgiadou et al., ; Sauerland et al., ). Participants received instructions to press the ‘L’ key on the keyboard as fast as possible in response to a facial stimulus, with the exception of the two targets. For these stimuli, they should press the ‘A’ key. We included the names of the celebrities in the instructions of the familiar targets condition. Participants then viewed the four targets for 30 s, accompanied by instructions to encode these faces. A practice block followed, showing each photo (2 × 2 targets, 2 × 4 irrelevants, and 2 × 1 probes) once. Above each facial stimuli, the question ‘Do you recognize this person?’ appeared. The labels ‘Yes’ and ‘No’ appeared to the left and right sides of each stimulus, respectively. Participants had 1500 ms to respond. After 800 ms, they received the warning ‘Too slow!’ Participants received the feedback ‘Wrong’ when they responded different than instructed. We randomly set the inter-stimulus interval between 250, 500, or 750 ms. A second, identical practice block followed if participants made more than one mistake or if they had responded too slow. Preceding the second practice block, participants received reminder instructions on how to respond. After this second practice block, participants continued to the actual task regardless of performance. Before the task started, the instructions appeared again, and participants viewed the targets for five more seconds. Similar to previous experiments, participants viewed the targets for a total of 35 s (Sauerland et al., ). During the actual task, every stimulus appeared 18 times, at a random sequence. The labels and feedback were identical to the practice blocks. In total, the actual task consisted of 252 trials (2 × 7 × 18). The RT-CIT of Experiment 2 was identical to the first experiment except that the participants had to complete three practice phases rather than one or two. During the first practice phase, the stimulus only disappeared once participants pressed the A or L key. This way, participants could control the pace of the test. They already received the ‘Wrong’ feedback if applicable. In the second practice phase, the stimuli disappeared after 1500 ms, if no key was pressed. The last practice trial was identical to the second practice trial except that participants additionally received feedback on their speed, with the words ‘Too slow!’ appearing on the screen after 800 ms. Participants viewed each of the 14 stimuli twice during each training phase. If a participant made too many mistakes during one of the practice phases, they had to repeat it once again. To make sure that the target photos were memorized, they were shown for five seconds and the instructions were repeated after each practice phase and before the actual task began. Follow-up photo display After completion of the RT-CIT task, participants viewed a photo recognition display that showed all 14 stimuli used, arranged in four rows, alternating between three and four photos. The 14 photographs included the two probes, four targets, and eight irrelevants. Two targets and four irrelevants each matched the general description of one of the probes, as determined in earlier research (Sauerland et al., ). The images were numbered (1, 2, 3, etc.). Participants indicated which individuals they recognized from the video. This allowed us to roughly determine if participants in the CIT conditions had explicit memory of the probes. Attention check Participants answered three forced-choice attention check questions with four or five answer options. The forced-choice questions were ‘Where did the actors first meet?’; ‘What was the victim doing when the handbag was stolen?’, and ‘What color was the stolen handbag?’ We excluded participants if they did not answer at least two of the three questions correctly. Procedure Recruitment occurred online, using the University’s research participation system Sona, social media sites such as Facebook, Whatsapp and Instagram, and other research platforms (e.g., SurveyCircle, ). Participants received instructions to complete the study on a laptop or PC and that they may have to download a plug-in. Participants could not take part if they had participated in one of the pilot studies or previous RT-CIT experiments in the department. After reading an information letter and providing consent, the software randomly allocated participants to one of two conditions (familiar vs. unfamiliar). Participants received instructions to watch the film closely and to pay close attention because they would answer questions about the film later. After the film, participants answered the attention check questions and in Experiment 2, participants in the familiar condition additionally read the information about the celebrities. Subsequently, participants completed the RT-CIT task. Participants then indicated the probes in the follow-up photo display and indicated whether they recognized any of the individuals in the study except for the celebrities from ‘real life’. Finally, participants were debriefed and thanked for their participation. The studies lasted approximately 20 to 30 min. SONA participants received credit for their participation. As a base for the two stimulus films, we used the four films from Sauerland et al. . In these films, one of two protagonists was visible primarily from the front and one primarily from the side. For the current study, we cut the four films to receive two films that included frontal close-ups and some distant profile views of both protagonists. The two films show the same action with two women: on a square in a pedestrian zone, the future thief walks up to the future victim to ask for directions. In the next scene, the victim is sitting down and is busy on her phone. Sneaking up from behind, the thief steals the victim’s purse. The actors switched between roles of thief and victim in the two versions to avoid confounding effects due to the actor or role. Table contains an overview of the facial viewing times of the protagonists in each stimulus film, split into close-ups and distant shots of the face in a frontal and profile view. Fourteen photographs included the two probes, four targets, and eight irrelevants. Two targets and four irrelevants each matched the general description of one of the probes, as determined in earlier research (Sauerland et al., ). The photographs showed faces from a frontal view from the collarbone up, with loose, open hair, without jewelry, accessories, or glasses and exhibited a neutral expression. All clothing was edited to be black. In the unfamiliar target condition, we randomly selected two targets from the non-probe photographs per probe. In the familiar target condition, the targets consisted of four celebrities. To select the celebrities, we conducted two pilot studies. Our aim was to identify four celebrities that matched the general description of the probes (cf. Sauerland et al., ; Wells et al., ) and were generally well-known in a student population. For the first pilot, we selected 12 celebrities that matched the general description of the two probes (six for each probe). Thirty participants (18 women, 9 men, 2 non-binary, 1 prefer not to say; M age = 21.63, SD age = 2.48) sequentially viewed the celebrity faces on Qualtrics and indicated whether they knew them. Participants could either indicate celebrities’ names or where they knew them from. More than 90% of the participants were familiar with Selena Gomez and Emma Watson. The other celebrities reached only 66% familiarity or less or were found too old to match the 21-year old probes in hindsight (i.e., Jennifer Aniston, Scarlet Johansson). We therefore conducted a second pilot with another set of 12 celebrities. From this pilot, we selected Taylor Swift and Katy Perry as familiar targets, based on familiarity scores of 93% and 83%, respectively. Thus, the four familiar targets were Selena Gomez, Katy Perry, Emma Watson, and Taylor Swift. Once we had established the identity of the familiar targets, we searched the internet with the celebrities’ names as the search term on Google Images to obtain high quality photographs with an unobstructed facial frontal view. In Experiment 1, we edited the chosen images using Adobe Photoshop (2022) to match the appearance of the unfamiliar target photos. We adjusted the background to a solid white and cropped the photos from the shoulders be the same size as the unfamiliar stimuli. We edited out all jewelry and accessories and adjusted the clothes to black shirts. In Experiment 2, we further edited the familiar targets to look less distinctive. For example, celebrities wore heavy make-up and lighting was distinctive to photographs taken in a regulated room. We adjusted the colors and levels using the Camera Raw Filter in Photoshop to match the color grading to the pictures. Additionally, we reduced make-up intensity. Finally, we reduced the picture quality by lowering the resolution and adding noise to match the lower quality of the other pictures. To improve familiar target recognition in Experiment 2, participants received information about and photos of each of the celebrities at the beginning of the testing session, prior to the CIT protocol. For example, participants saw three photos of Taylor Swift alongside her occupation, her most known songs and the number of Grammy awards she won. The RT-CIT of was programmed using Inquisit 6.5.2 Lab (Experiment 1) and Inquisit 6.6.1 Lab (Experiment 2). The data were recorded in milliseconds, using one joint CIT protocol for the thief and victim. All photos were 388 × 462 pixels in size. We followed a similar protocol as other RT-CIT studies with facial stimuli (e.g., Georgiadou et al., ; Sauerland et al., ). Participants received instructions to press the ‘L’ key on the keyboard as fast as possible in response to a facial stimulus, with the exception of the two targets. For these stimuli, they should press the ‘A’ key. We included the names of the celebrities in the instructions of the familiar targets condition. Participants then viewed the four targets for 30 s, accompanied by instructions to encode these faces. A practice block followed, showing each photo (2 × 2 targets, 2 × 4 irrelevants, and 2 × 1 probes) once. Above each facial stimuli, the question ‘Do you recognize this person?’ appeared. The labels ‘Yes’ and ‘No’ appeared to the left and right sides of each stimulus, respectively. Participants had 1500 ms to respond. After 800 ms, they received the warning ‘Too slow!’ Participants received the feedback ‘Wrong’ when they responded different than instructed. We randomly set the inter-stimulus interval between 250, 500, or 750 ms. A second, identical practice block followed if participants made more than one mistake or if they had responded too slow. Preceding the second practice block, participants received reminder instructions on how to respond. After this second practice block, participants continued to the actual task regardless of performance. Before the task started, the instructions appeared again, and participants viewed the targets for five more seconds. Similar to previous experiments, participants viewed the targets for a total of 35 s (Sauerland et al., ). During the actual task, every stimulus appeared 18 times, at a random sequence. The labels and feedback were identical to the practice blocks. In total, the actual task consisted of 252 trials (2 × 7 × 18). The RT-CIT of Experiment 2 was identical to the first experiment except that the participants had to complete three practice phases rather than one or two. During the first practice phase, the stimulus only disappeared once participants pressed the A or L key. This way, participants could control the pace of the test. They already received the ‘Wrong’ feedback if applicable. In the second practice phase, the stimuli disappeared after 1500 ms, if no key was pressed. The last practice trial was identical to the second practice trial except that participants additionally received feedback on their speed, with the words ‘Too slow!’ appearing on the screen after 800 ms. Participants viewed each of the 14 stimuli twice during each training phase. If a participant made too many mistakes during one of the practice phases, they had to repeat it once again. To make sure that the target photos were memorized, they were shown for five seconds and the instructions were repeated after each practice phase and before the actual task began. After completion of the RT-CIT task, participants viewed a photo recognition display that showed all 14 stimuli used, arranged in four rows, alternating between three and four photos. The 14 photographs included the two probes, four targets, and eight irrelevants. Two targets and four irrelevants each matched the general description of one of the probes, as determined in earlier research (Sauerland et al., ). The images were numbered (1, 2, 3, etc.). Participants indicated which individuals they recognized from the video. This allowed us to roughly determine if participants in the CIT conditions had explicit memory of the probes. Participants answered three forced-choice attention check questions with four or five answer options. The forced-choice questions were ‘Where did the actors first meet?’; ‘What was the victim doing when the handbag was stolen?’, and ‘What color was the stolen handbag?’ We excluded participants if they did not answer at least two of the three questions correctly. Recruitment occurred online, using the University’s research participation system Sona, social media sites such as Facebook, Whatsapp and Instagram, and other research platforms (e.g., SurveyCircle, ). Participants received instructions to complete the study on a laptop or PC and that they may have to download a plug-in. Participants could not take part if they had participated in one of the pilot studies or previous RT-CIT experiments in the department. After reading an information letter and providing consent, the software randomly allocated participants to one of two conditions (familiar vs. unfamiliar). Participants received instructions to watch the film closely and to pay close attention because they would answer questions about the film later. After the film, participants answered the attention check questions and in Experiment 2, participants in the familiar condition additionally read the information about the celebrities. Subsequently, participants completed the RT-CIT task. Participants then indicated the probes in the follow-up photo display and indicated whether they recognized any of the individuals in the study except for the celebrities from ‘real life’. Finally, participants were debriefed and thanked for their participation. The studies lasted approximately 20 to 30 min. SONA participants received credit for their participation. Data preparation Exclusion criteria in both experiments included answering less than 2 out of 3 attention checks correctly (Sauerland et al., , ), and displaying a high error rate or non-completed trials (i.e., ≥ 50%; cf. Kleinberg & Verschuere, ). Based on earlier RT-CIT experiments with facial stimuli, we only included correct trials with reaction times between 150 ms and 1500 ms (Georgiadou et al., ; Sauerland et al., ). Correct trials consisted of pressing ‘A’ for targets and pressing ‘L’ for all other facial stimuli. We aggregated reaction times into mean reaction times per stimulus type (probes, targets, and irrelevants). In addition to the pre-registered analyses, we also present separate results for the thief and the victim, in addition to the analyses collapsed across both probes. Pre-registered analyses: RT-CIT effect and moderation by familiarity Table shows the mean reaction times and inferential statistics for the comparison of reaction times for probes and irrelevants for Experiment 1 and 2. We conducted a 2 (stimulus type: probe vs. irrelevants) x 2 (target familiarity: familiar vs. unfamiliar) mixed measures ANOVA to test the effect of familiarity on the CIT effect. For Experiment 1, the main effect of stimulus type was significant, F (1,88) = 22.73, p < .001, n p 2 = .205, as expected. Supporting hypothesis 1, the reaction times for probes ( M = 515 ms, SD = 70) were slower than for irrelevants ( M = 502 ms, SD = 56), d = 0.47, 95% CI [0.25; 0.69], revealing a moderate CIT effect. Additionally, the main effect of familiarity was significant, F (1,88) = 51.99, p < .001, n p 2 = .371. Specifically, reaction times were slower in the unfamiliar condition ( M Probe = 560 ms, SD Probe = 61; M Irrelevant = 538 ms, SD Irrelevant = 43) than in the familiar condition ( M Probe = 478 ms, SD Probe = 54; M Irrelevant = 471 ms, SD Irrelevant = 45). The two main effects were moderated by a significant interaction between stimulus type and target-familiarity, F (1,88) = 6.38, p = .013, n p 2 = .068. Yet, as Fig. illustrates, the nature of the interaction was opposite to hypothesis 2: The CIT effect was smaller rather than larger for familiar targets compared to unfamiliar targets. For familiar targets, the difference between probes ( M = 478 ms, SD = 54) and irrelevants ( M = 471 ms, SD = 45) was unexpectedly non-significant, t (48) = 1.66, p = .103, d = 0.24, 95% CI [-0.05 0.52]. For unfamiliar targets, reactions to probes were significantly slower than to irrelevants (probes: M = 560 ms, SD = 61; irrelevants: M = 538 ms, SD = 43), with a moderate to large effect size, t (40) = 4.94, p ˂ .001, d = 0.77, 95% CI [0.42; 1.12]. To evaluate the strength of the evidence, we also computed a Bayesian mixed measures ANOVA using JASP 0.18.3.0. The model that fit the data best was the model that included the interaction of stimulus type and familiarity and the main effects (BF M = 13.41). The model with interaction fit the data 2.99 times better than the model that included only the main effects (the JASP outputs are available on https://osf.io/j6vuf/?view_only= ). This indicates that celebrity targets reduced the CIT effect in Experiment 1. In Experiment 2, the main effect of stimulus type was again significant, F (1,81) = 70.92, p < .001, n p 2 = .476. Supporting hypothesis 1, reaction times to probes ( M = 525 ms, SD = 53) were significantly slower than to irrelevants ( M = 503 ms, SD = 49), d = 0.95, 95% CI [0.69; 1.21], with a large effect size. Additionally, the main effect of familiarity was significant, F (1,81) = 53.33, p < .001, n p 2 = .397 Specifically, reaction times were slower in the unfamiliar condition ( M Probe = 554 ms, SD Probe = 41; M Irrelevant = 528 ms, SD Irrelevant = 42) than in the familiar condition ( M Probe = 486 ms, SD Probe = 42; M Irrelevant = 469 ms, SD Irrelevant = 38). The expected stimulus type x familiarity interaction was non-significant, F (1,81) = 3.05, p = .085, n p 2 = .036. Similar to Experiment 1 and contrary to hypothesis 2, the CIT effect was not larger in the familiar condition, d = 0.79, 95% CI [0.40; 1.16], than in the unfamiliar condition, d = 1.09, 95% CI [0.73; 1.44]. This finding was confirmed by an additional, non-pre-registered analysis of the Bayesian ANOVA. The model that fit the data best was the model with the two main effects (BF M = 4.31) and the data were about (BF = 0.90) as likely under the model with the interaction as under the model without the interaction (the outputs are available at https://osf.io/j6vuf/?view_only= ). Exclusion criteria in both experiments included answering less than 2 out of 3 attention checks correctly (Sauerland et al., , ), and displaying a high error rate or non-completed trials (i.e., ≥ 50%; cf. Kleinberg & Verschuere, ). Based on earlier RT-CIT experiments with facial stimuli, we only included correct trials with reaction times between 150 ms and 1500 ms (Georgiadou et al., ; Sauerland et al., ). Correct trials consisted of pressing ‘A’ for targets and pressing ‘L’ for all other facial stimuli. We aggregated reaction times into mean reaction times per stimulus type (probes, targets, and irrelevants). In addition to the pre-registered analyses, we also present separate results for the thief and the victim, in addition to the analyses collapsed across both probes. Table shows the mean reaction times and inferential statistics for the comparison of reaction times for probes and irrelevants for Experiment 1 and 2. We conducted a 2 (stimulus type: probe vs. irrelevants) x 2 (target familiarity: familiar vs. unfamiliar) mixed measures ANOVA to test the effect of familiarity on the CIT effect. For Experiment 1, the main effect of stimulus type was significant, F (1,88) = 22.73, p < .001, n p 2 = .205, as expected. Supporting hypothesis 1, the reaction times for probes ( M = 515 ms, SD = 70) were slower than for irrelevants ( M = 502 ms, SD = 56), d = 0.47, 95% CI [0.25; 0.69], revealing a moderate CIT effect. Additionally, the main effect of familiarity was significant, F (1,88) = 51.99, p < .001, n p 2 = .371. Specifically, reaction times were slower in the unfamiliar condition ( M Probe = 560 ms, SD Probe = 61; M Irrelevant = 538 ms, SD Irrelevant = 43) than in the familiar condition ( M Probe = 478 ms, SD Probe = 54; M Irrelevant = 471 ms, SD Irrelevant = 45). The two main effects were moderated by a significant interaction between stimulus type and target-familiarity, F (1,88) = 6.38, p = .013, n p 2 = .068. Yet, as Fig. illustrates, the nature of the interaction was opposite to hypothesis 2: The CIT effect was smaller rather than larger for familiar targets compared to unfamiliar targets. For familiar targets, the difference between probes ( M = 478 ms, SD = 54) and irrelevants ( M = 471 ms, SD = 45) was unexpectedly non-significant, t (48) = 1.66, p = .103, d = 0.24, 95% CI [-0.05 0.52]. For unfamiliar targets, reactions to probes were significantly slower than to irrelevants (probes: M = 560 ms, SD = 61; irrelevants: M = 538 ms, SD = 43), with a moderate to large effect size, t (40) = 4.94, p ˂ .001, d = 0.77, 95% CI [0.42; 1.12]. To evaluate the strength of the evidence, we also computed a Bayesian mixed measures ANOVA using JASP 0.18.3.0. The model that fit the data best was the model that included the interaction of stimulus type and familiarity and the main effects (BF M = 13.41). The model with interaction fit the data 2.99 times better than the model that included only the main effects (the JASP outputs are available on https://osf.io/j6vuf/?view_only= ). This indicates that celebrity targets reduced the CIT effect in Experiment 1. In Experiment 2, the main effect of stimulus type was again significant, F (1,81) = 70.92, p < .001, n p 2 = .476. Supporting hypothesis 1, reaction times to probes ( M = 525 ms, SD = 53) were significantly slower than to irrelevants ( M = 503 ms, SD = 49), d = 0.95, 95% CI [0.69; 1.21], with a large effect size. Additionally, the main effect of familiarity was significant, F (1,81) = 53.33, p < .001, n p 2 = .397 Specifically, reaction times were slower in the unfamiliar condition ( M Probe = 554 ms, SD Probe = 41; M Irrelevant = 528 ms, SD Irrelevant = 42) than in the familiar condition ( M Probe = 486 ms, SD Probe = 42; M Irrelevant = 469 ms, SD Irrelevant = 38). The expected stimulus type x familiarity interaction was non-significant, F (1,81) = 3.05, p = .085, n p 2 = .036. Similar to Experiment 1 and contrary to hypothesis 2, the CIT effect was not larger in the familiar condition, d = 0.79, 95% CI [0.40; 1.16], than in the unfamiliar condition, d = 1.09, 95% CI [0.73; 1.44]. This finding was confirmed by an additional, non-pre-registered analysis of the Bayesian ANOVA. The model that fit the data best was the model with the two main effects (BF M = 4.31) and the data were about (BF = 0.90) as likely under the model with the interaction as under the model without the interaction (the outputs are available at https://osf.io/j6vuf/?view_only= ). CIT effect for thief and victim Consistent with earlier studies (Georgiadou et al., ; Sauerland et al., ) but deviating from the pre-registration, we report the results across probes but also separately for thief and victim. As illustrated in Table , the results for the separate analyses for the thief and victim were analogous throughout. Error rates Because error rates are typically very low and lead to less reliable results (Kleinberg & Verschuere, ), in line with previous work, we tested our hypotheses exclusively on reaction times. For sake of completeness: in Experiment 1, error rates for probes and irrelevants were very low ( M Probes = 5.80, SD Probes = 7.98, and M Irrelevants = 3.16, SD Irrelevants = 4.80), with higher error rates for targets, M = 21.22 ( SD Targets = 10.65). A a 2 (stimulus type: probe vs. irrelevants) x 2 (target familiarity: familiar vs. unfamiliar) ANOVA on the error rates revealed significant main effects of stimulus type, F (1,88) = 29.33, p < .001, n p 2 = .250, and familiarity, F (1,88) = 11.30, p = .001, n p 2 = .114. The interaction effect was also significant, F (1,88) = 10.66, p = .002, n p 2 = .108. Error rates were significantly higher for probes than irrelevants for both unfamiliar, M Probes = 8.94, SD Probes = 9.14, and M Irrelevants = 4.47, SD Irrelevants = 4.74, and familiar targets, M Probes = 3.18, SD Probes = 5.75, and M Irrelevants = 2.07, SD Irrelevants = 4.61. In line with the RT findings and contrary to expectations, this CIT effect was again stronger for unfamiliar targets, t (40) = 4.63, p < .001, d = 0.72, 95% CI [0.38; 1.07], than for familiar targets, t (48) = 2.27, p = .027, d = 0.33, 95% CI [0.04; 0.61]. In Experiment 2, error rates for probes and irrelevants were very low ( M Probes = 3.15, SD Probes = 5.77, and M Irrelevants = 1.70, SD Irrelevants = 2.65), with higher error rates for targets, M = 15.48 ( SD Targets = 8.51). The 2 × 2 ANOVA on the error rates revealed significant main effects of stimulus type, F (1,81) = 5.48, p = .022, n p 2 = .063, and familiarity, F (1,81) = 6.84, p = .011, n p 2 = .078. The interaction effect was non-significant, F (1,81) = 2.54, p = .115, n p 2 = .030. Error rates were significantly higher for probes than irrelevants for unfamiliar targets ( M Probes = 4.40, SD Probes = 7.06, and M Irrelevants = 2.20, SD Irrelevants = 3.24), t (47) = 2.45, p = .018, d = 0.35, 95% CI [0.06; 0.64], but not for familiar targets, M Probes = 1.43, SD Probes = 2.46, and M Irrelevants = 1.01, SD Irrelevants = 1.25, t (34) = 0.96, p = .344, d = 0.16, 95% CI [-0.17; 0.49]. This is in line with the RT findings indicating that the CIT effect was stronger for unfamiliar than familiar targets. Follow-up photo display We conducted a nonparametric binominal test against 1/7 odds, with a chance level of 0.14 (Georgiadou et al., ). In Experiment 1, participants recognized both the thief ( M = 0.60 [0.50; 0.70]) and the victim ( M = 0.64, [0.54; 0.75]) above chance level, ps < .001. Recognition accuracy in this task did not differ systematically as a function of target familiarity, with BF 10 = 1.77 for the thief and BF 10 = 0.26 for the victim. Likewise, in Experiment 2, participants recognized the thief ( M = 0.67, [0.57; 0.78]) and the victim ( M = 0.70, [0.60; 0.80]) above chance level, ps < .001. Recognition accuracy in this task did not systematically differ as a function of target familiarity, with BF 10 = 0.89 for the thief and BF 10 = 0.26 for the victim. Consistent with earlier studies (Georgiadou et al., ; Sauerland et al., ) but deviating from the pre-registration, we report the results across probes but also separately for thief and victim. As illustrated in Table , the results for the separate analyses for the thief and victim were analogous throughout. Because error rates are typically very low and lead to less reliable results (Kleinberg & Verschuere, ), in line with previous work, we tested our hypotheses exclusively on reaction times. For sake of completeness: in Experiment 1, error rates for probes and irrelevants were very low ( M Probes = 5.80, SD Probes = 7.98, and M Irrelevants = 3.16, SD Irrelevants = 4.80), with higher error rates for targets, M = 21.22 ( SD Targets = 10.65). A a 2 (stimulus type: probe vs. irrelevants) x 2 (target familiarity: familiar vs. unfamiliar) ANOVA on the error rates revealed significant main effects of stimulus type, F (1,88) = 29.33, p < .001, n p 2 = .250, and familiarity, F (1,88) = 11.30, p = .001, n p 2 = .114. The interaction effect was also significant, F (1,88) = 10.66, p = .002, n p 2 = .108. Error rates were significantly higher for probes than irrelevants for both unfamiliar, M Probes = 8.94, SD Probes = 9.14, and M Irrelevants = 4.47, SD Irrelevants = 4.74, and familiar targets, M Probes = 3.18, SD Probes = 5.75, and M Irrelevants = 2.07, SD Irrelevants = 4.61. In line with the RT findings and contrary to expectations, this CIT effect was again stronger for unfamiliar targets, t (40) = 4.63, p < .001, d = 0.72, 95% CI [0.38; 1.07], than for familiar targets, t (48) = 2.27, p = .027, d = 0.33, 95% CI [0.04; 0.61]. In Experiment 2, error rates for probes and irrelevants were very low ( M Probes = 3.15, SD Probes = 5.77, and M Irrelevants = 1.70, SD Irrelevants = 2.65), with higher error rates for targets, M = 15.48 ( SD Targets = 8.51). The 2 × 2 ANOVA on the error rates revealed significant main effects of stimulus type, F (1,81) = 5.48, p = .022, n p 2 = .063, and familiarity, F (1,81) = 6.84, p = .011, n p 2 = .078. The interaction effect was non-significant, F (1,81) = 2.54, p = .115, n p 2 = .030. Error rates were significantly higher for probes than irrelevants for unfamiliar targets ( M Probes = 4.40, SD Probes = 7.06, and M Irrelevants = 2.20, SD Irrelevants = 3.24), t (47) = 2.45, p = .018, d = 0.35, 95% CI [0.06; 0.64], but not for familiar targets, M Probes = 1.43, SD Probes = 2.46, and M Irrelevants = 1.01, SD Irrelevants = 1.25, t (34) = 0.96, p = .344, d = 0.16, 95% CI [-0.17; 0.49]. This is in line with the RT findings indicating that the CIT effect was stronger for unfamiliar than familiar targets. We conducted a nonparametric binominal test against 1/7 odds, with a chance level of 0.14 (Georgiadou et al., ). In Experiment 1, participants recognized both the thief ( M = 0.60 [0.50; 0.70]) and the victim ( M = 0.64, [0.54; 0.75]) above chance level, ps < .001. Recognition accuracy in this task did not differ systematically as a function of target familiarity, with BF 10 = 1.77 for the thief and BF 10 = 0.26 for the victim. Likewise, in Experiment 2, participants recognized the thief ( M = 0.67, [0.57; 0.78]) and the victim ( M = 0.70, [0.60; 0.80]) above chance level, ps < .001. Recognition accuracy in this task did not systematically differ as a function of target familiarity, with BF 10 = 0.89 for the thief and BF 10 = 0.26 for the victim. In two pre-registered experiments, we examined the effect of target-familiarity on the usefulness of the reaction time-based Concealed Information Test for diagnosing face recognition. In both experiments, reaction times were longer to probes than to irrelevants in all conditions, indicating the expected CIT effect. Effect sizes were moderate to large (Experiment 1: d = 0.47; Experiment 2: d = 0.95 across conditions). Consistent with previous findings (e.g., Georgiadou et al., ; Sauerland et al., ), the RT-CIT was an effective tool for face recognition. Contrary to expectations, however, the familiar targets did not increase the validity of the test. In Experiment 1, celebrity targets unexpectedly even reduced the size of the CIT effect (cf. hypothesis 2). In Experiment 2, the expected interaction between familiarity and CIT effect did not emerge. We were unable to replicate previous findings on the moderating effect of target-familiarity (Suchotzki et al., ). Of note, another recent study failed to replicate the target familiarity effect. Koller et al. used names, birthdates of friends, and the address of their former residing city as familiar targets. Using familiar targets did not increase the RT-CIT effect. One possible explanation for our findings could be that familiar targets made the task easier, as evidenced by faster reaction times and lower error rates in the familiar vs. unfamiliar conditions in both experiments. Familiar targets do not need to be learned after all (Koller, ). Due to the familiarity, focusing on a single feature of the target may have been enough for participants to recognize the celebrities. Consequently, the stimulus may not have been processed during decision making. Reliance on multiple features to increase response conflict, and hence the CIT effect, is known as the target focus hypothesis (Koller et al., ). Another possible explanation for our findings could be that the celebrity targets stood out too much from the other facial stimuli, especially in Experiment 1. Indeed, the familiar celebrity targets in Experiment 1 differed from unfamiliar targets in terms of lighting, amount of make-up, and photo quality and this made them stand out. The RT-CIT effect decreases when targets are too different from the other stimuli. For instance, one study used a joker card as a target while the other stimuli consisted of regular playing cards (Kubo & Nittono, ). Reaction times for targets were significantly longer than for the other stimuli and the effect size was moderate in the control condition ( d = 0.31), but large in the condition participants concealed knowing the probe ( d = 0.77). Here, the joker might have stood out due to the associated special significance of the card in card games. In our experiments, participants who learnt familiar targets were faster and made significantly fewer mistakes than participants who learnt unfamiliar targets. This was true for probes, targets, and irrelevants. In Suchotzki et al. , however, participants made more mistakes in response to probes when targets were familiar to them, compared to unfamiliar. This indicates that our familiar targets condition may have been too easy, reducing the task to a target detection task similar to Kubo and Nittono . Unprompted feedback from participants that the task was rather easy, and the targets were easily distinguishable supports this conclusion. Evidence that our task may have become a target detection task emerges from the responses to targets in both experiments. Participants responded significantly slower to unfamiliar targets than to familiar targets, with large effect sizes in both experiments ( d E1 = 1.31; d E2 = 1.24). Furthermore, target error rates were significantly lower for familiar than unfamiliar targets, but only in Experiment 1 ( d E1 = 0.63; d E2 < 0.01). This suggests that our attempts to decrease the distinctiveness of familiar targets in Experiment 2 was only partially successful. Participants may have detected the familiar targets rather effortlessly, consequently, targets may not have been processed entirely and features that could have increased response conflict did not get incorporated in the decision making process. Although we were unable to replicate the familiar targets effect, the observed RT-CIT effect sizes are noteworthy in themselves. Consistent with earlier findings (Sauerland et al., ; d = 0.65 to 0.87; Sauerland et al., ; d = 0.74 and 0.97), we obtained moderate to large CIT effects, with the exception of the familiar targets condition in Experiment 1, where targets arguably stood out the most. The findings from these six experiments combined (the current two experiments; Sauerland et al., , ) suggest that the RT-CIT may have the potential to for diagnosing facial recognition. This is crucial for the prevention of false identifications. Limitations and future directions This research has several limitations. First, we assumed familiarity based on recognition and knowledge of celebrities’ names and faces using two pilot studies. Knowledge and recognition of a celebrity may however not be sufficient for increasing response conflict. Rather, familiar targets that convey emotional significance and salience to the participants could be imperative for eliciting the expected familiarity effect (Kleinberg & Verschuere, ; Suchotzki et al., ). To address this issue, future studies may use friends and family members as targets. This may come with privacy concerns and logistical challenges. To protect people’s privacy, researchers would be obliged to obtain consent for several targets per participant, most likely lowering participation rates. Furthermore, participants may not have close ties to people that match the person descriptions of the people in the stimulus events. A second limitation is that the period between the presentation of the stimulus videos and the administration of the RT-CIT in this study was relatively short. The interval time between the event in question (i.e., the crime) and the CIT in real life would be significantly longer. Detection-accuracy stays relatively strong over time for well-encoded verbal stimuli (Seymour & Fraynt, ). Although this has not been tested for facial recognition, this suggests that the current results may generalize to longer retention intervals. Yet, witnesses may not encode perpetrator faces well when events are relatively short. Although observation time did not moderate the CIT effect (Sauerland et al., ), the validity of the RT-CIT may be affected when short observation is combined with a longer retention interval. Longer retention intervals are associated with decreased facial recognition, eyewitness identification (Deffenbacher et al., ), and overconfidence in recognition accuracy (Palmer et al., ). Future research should investigate whether the RT-CIT is diagnostic of face recognition even with longer time delays between encoding of the event and testing. This research has several limitations. First, we assumed familiarity based on recognition and knowledge of celebrities’ names and faces using two pilot studies. Knowledge and recognition of a celebrity may however not be sufficient for increasing response conflict. Rather, familiar targets that convey emotional significance and salience to the participants could be imperative for eliciting the expected familiarity effect (Kleinberg & Verschuere, ; Suchotzki et al., ). To address this issue, future studies may use friends and family members as targets. This may come with privacy concerns and logistical challenges. To protect people’s privacy, researchers would be obliged to obtain consent for several targets per participant, most likely lowering participation rates. Furthermore, participants may not have close ties to people that match the person descriptions of the people in the stimulus events. A second limitation is that the period between the presentation of the stimulus videos and the administration of the RT-CIT in this study was relatively short. The interval time between the event in question (i.e., the crime) and the CIT in real life would be significantly longer. Detection-accuracy stays relatively strong over time for well-encoded verbal stimuli (Seymour & Fraynt, ). Although this has not been tested for facial recognition, this suggests that the current results may generalize to longer retention intervals. Yet, witnesses may not encode perpetrator faces well when events are relatively short. Although observation time did not moderate the CIT effect (Sauerland et al., ), the validity of the RT-CIT may be affected when short observation is combined with a longer retention interval. Longer retention intervals are associated with decreased facial recognition, eyewitness identification (Deffenbacher et al., ), and overconfidence in recognition accuracy (Palmer et al., ). Future research should investigate whether the RT-CIT is diagnostic of face recognition even with longer time delays between encoding of the event and testing. We tested the usefulness of familiar targets as a means to increase the validity of the RT-CIT. While our findings did not support the hypothesis that familiar targets can increase the capacity of the RT-CIT to diagnose face recognition, our results showed moderate to strong capacity to diagnose face recognition in general. These findings add to the accumulating body of research indicating that the RT-CIT can be successful in diagnosing face recognition. Future studies may test whether an emotional significance of target-familiarity or whether other modalities such as target’s names can increase the diagnosticity of the RT-CIT (Koller et al., ; Lukács et al., ; Suchotzki et al., ). An effective RT-CIT could have important implications in the legal field. It could possibly lead to a lower percentage of misidentification due to its reliance on automatic processes and indirect recognition procedure (Sauerland et al., ). |
A white paper from the | a2352090-9cc9-4463-92bc-a377b194bf54 | 11705400 | Biochemistry[mh] | The conference explored a wide range of critical themes each emphasizing the complex and evolving nature of modern education (Fig. ). Central to these discussions was the transformative potential of the digital revolution, which emerged as a recurring and dominant theme throughout the event. The integration of technology into teaching and learning practices was lauded for its potential to enhance engagement, accessibility, and efficiency. However, the ‘power and perils of AI’ were also acknowledged, prompting discussions on the ethical implications, potential biases, and the need for critical thinking in the digital age. The discussion also highlighted topics such as the digital divide and teaching on open‐source platforms while recognizing artificial intelligence (AI) biases, particularly from Eurocentric data. The conference underscored the importance of equipping educators with the skills and knowledge to navigate the complexities of the digital landscape, ensuring that technology serves as an enabler rather than a disruptor of effective learning. The traditional teaching‐learning model, focused on facts and linear teaching, was deemed inadequate for the future. Emphasis should be on active learning, multidisciplinary teaching, authentic and experimental learning, and the cultivation of personal development traits like resilience and grit. The future of education, as outlined in the OECD Education Framework 2030 (The Future of Education and Skills—Education 2030), emphasizes the interconnectedness of knowledge, skills, attitudes, and values. In alignment with this framework, it is also strategic to develop the assessment of noncognitive competencies as they are essential to both the individual and collective well‐being. More examples such as the innovative approach of Aalto University's Design Factory, with franchises in many countries, to solve real‐world problems are needed. Future skills, as outlined by the VALUE (Valid Assessment of Learning in Undergraduate Education) rubrics, are emphasized. Additionally, new learning opportunities, such as Manchester University's emphasis on social responsibility and making a real‐world impact, are central to this evolving educational paradigm (Fig. ). In today's rapidly evolving educational landscape, the human–human interaction remains crucial, necessitating significant changes in assessment and monitoring practices. The conference also addressed the profound transformation over the past decade which expanded the role of educators beyond traditional boundaries. A modern, effective teacher was proposed to be the one who challenges conventional notions, making learning as dynamic and engaging as our ever‐changing world. Today's educators must possess expanded knowledge structures that encompass social, international, and cultural dimensions. Their skill set should now include the ability to navigate disciplinary boundaries, foster digital literacy, and appreciate cultural diversity. While the fundamental role of an educator—to guide and inspire learning—remains unchallenged, there is an increased emphasis on adaptability and a commitment to evolving educational values. To become effective change agents, educators need to move away from a ‘banking education system’, that is, an educational system that focuses solely on the acquisition of knowledge and skills. Instead, they need to embrace an educational approach that cultivates a professional identity aligned with the values of good research practice. This paradigm requires curricula that views change as a continuum and a process of continuous improvement, encouraging self‐reflection and critical thinking. Educators must prepare for the future by engaging in collaborative learning with students, peers, and society, while collectively adapting to new challenges. The transformative power of education lies in its ability to inspire joy, foster to foster self‐criticism, and to celebrate diversity. It was agreed that only by embodying these principles, education can empower learners to become agents of change in an ever‐evolving world. The purpose of education extends beyond mere qualification; it encompasses socialization and the development of an individual professional identity. This holistic approach aims to foster student development so they can navigate complex societal challenges, to contribute meaningfully to their communities, and to continuously adapt to a changing global landscape. By focusing on these multifaceted goals, education can truly become a catalyst for personal growth and societal progress. More details about these critical themes are given below. The digital revolution in education Discussions at the conference emphasized the profound impact of technological advancements on the educational landscape. The rapid changes in higher education, particularly those resulting from the COVID‐19 pandemic and advancements in AI, were key focal points. The integration of technology into teaching and learning practices has the ability to enhance engagement, accessibility, and efficiency. Transformative digital learning tools enable the next generation of scientists to develop an experimental mindset, build scientific skills, competencies and attitudes, and be prepared for the future workplace. Technologies support the creation of interactive, problem‐solving laboratory simulations and advanced digital worksheets that enable self‐led learning and improvement. These digital tools enable students to practice, to apply feedback, and to master the advanced skills needed for excellent scientific literacy. The conference showcased these innovative approaches to leveraging technology, from interactive laboratory simulations and digital worksheets to AI‐driven data analysis and assessment tools. The event highlighted the importance of interdisciplinary collaboration and the integration of diverse subjects to create comprehensive and relevant curricula. The rise of AI and its impact on various fields, including drug discovery, molecular modeling, and precision medicine, necessitates a shift toward an interdisciplinary approach that equips students with the skills and knowledge to navigate the complexities of the modern scientific landscape, necessitating the fusion of diverse subjects to create comprehensive curricula. However, the ‘power and perils of AI’ were also discussed, sparking thoughtful discussions on the ethical implications, potential biases, and the need for critical thinking in the digital age. Discussions highlighted concerns about digital footprints and the environmental impact, the importance of critical evaluation, and responsible use of technology. The need for tailored approaches in teaching and learning were emphasized ensuring that technology serves as a complement to, rather than a replacement for, traditional teaching methods. The conference underscored the importance of equipping educators with the skills and knowledge to navigate the complexities of the digital landscape, ensuring that technology serves as an enabler rather than a disruptor of effective learning. Active learning and student engagement A key theme that emerged from the conference was the critical importance of active learning, particularly in addressing the engagement challenges often faced in large classes. Presenters and participants alike emphasized that active learning is not just a trendy concept, but a crucial strategy for improving student engagement and learning, deepening understanding, and fostering critical thinking skills. Several innovative approaches to active learning were showcased: Interactive Activities: Presenters demonstrated how incorporating hands‐on, problem‐solving tasks can transform traditionally passive lectures into dynamic learning experiences. These activities ranged from short, in‐class exercises to more complex, technology‐enabled simulations. Group Discussions: The power of peer‐to‐peer learning was highlighted through structured group discussions. Techniques such as think‐pair‐share and jigsaw discussions were presented as effective ways to promote collaborative learning and expose students to diverse perspectives. Peer Teaching: Conference participants shared success stories of implementing peer teaching strategies, where students take turns explaining concepts to their classmates. This approach not only reinforces learning for the student–teacher but also provides relatable instruction for the listeners. Technology‐Enhanced Learning: Various digital tools and platforms were discussed as means to facilitate active learning in large classes. These included real‐time polling systems, collaborative online workspaces, and gamified learning applications. Flipped Classroom Model: Several presenters advocated for the flipped classroom approach, where lecture content is accessed online before class, allowing in‐person time to be dedicated to active problem‐solving and discussions. The conference also addressed the challenges of implementing active learning strategies, particularly in large classes. Presenters discussed methods for managing logistics, ensuring equitable participation, and assessing learning outcomes in active learning environments. The importance of faculty development programs to support educators in adopting these methods was also emphasized. As the conference concluded, there was a clear consensus that the future of education lies in moving away from passive, lecture‐based instruction toward more engaging, student‐centered approaches. The challenge now lies in scaling these practices across institutions and disciplines, ensuring that active learning becomes the norm rather than the exception in higher education. Transitions and inclusivity in education The transition from high school to university is a pivotal juncture in a student's academic journey. Acknowledging this significant leap from school to university and recognizing the significant challenges faced by students during this transition are critically important. Identifying obstacles that hinder effective learning during the transition and addressing ‘dislocation, loss of belonging, and unfamiliarity’ experienced by students necessitate proactive measures to ensure a smooth and successful passage. The conference explored strategies to empower students during this critical phase, emphasizing the importance of bridging the gap between secondary and tertiary education. Strategies to support students during this transition were discussed, including the use of technology to create flexible learning opportunities and the importance of fostering a sense of belonging, adaptability, and self‐regulated learning. Strategies such as mentorship programs, orientation initiatives, and the cultivation of a supportive learning environment were proposed to empower students during this critical phase. Formative assessment and self‐regulated learning The role and importance of formative feedback in reinforcing self‐regulated learning strategies, enhancing motivation, and improving academic outcomes were emphasized. Techniques to enhance students' self‐assessment abilities and the use of feedback to drive lifelong learning were key points of discussion. The concepts of ‘self‐assessment ability’ and ‘self‐regulation promoting students’ self‐evaluation skills' were explored and the Equity, Agency and Transparency (EAT) Assessment Framework was highlighted as a research‐informed approach and tool for educators to implement effective inclusive assessment practices. Interdisciplinary and reflective teaching and learning (including professional development) The cultivation of transferable skills, such as communication, collaboration, and problem‐solving, was deemed essential to prepare students for the complexities of the 21st‐century workforce. The conference explored innovative approaches to integrating these skills into the science curriculum, emphasizing the importance of self‐reflection and mentorship in fostering their development. Using arts to nurture reflective attitudes and the importance of interdisciplinary approaches in education were highlighted. Encouraging learners to engage with art as a means of self‐discovery and critical thinking and to reflect on their experiences can lead to deeper learning and personal growth. The professional development of all those who teach was recognized as a cornerstone of educational excellence. The conference emphasized the need for continuous learning and growth among educators, fostering a culture of reflection, innovation, and collaboration. The National Professional Development Framework of Ireland, which supports all who teach in higher education by providing a structured outline of activities aligned with teaching and learning goals was emphasized. Launched in 2016, this framework aims to empower staff to create, discover, and engage in meaningful personal and professional development, to engage in peer dialog and support and to enhance and develop the pedagogy of individual disciplines and to enable learning from other disciplines. The importance of mentorship and peer support was highlighted, along with the value of structured professional development programs that cater to the diverse needs and aspirations of educators. The FEBS Education and Training Academy, launched during the conference, exemplifies this commitment to empowering educators through high‐quality training and fostering a community dedicated to excellence in science education (Fig. ). The need for tailored approaches to enhance disciplinary excellence was also underscored, with The Disciplinary Excellence in Teaching, Learning, and Assessment (DELTA) Framework in Ireland serving as a model for a strategic path for disciplines to enhance teaching. The framework is recognized through the DELTA National Award, which aims to support staff across disciplines to work collaboratively to articulate, evidence, and plan their engagement in and commitment to teaching and learning enhancement, toward student success. Such a model provides a capacity building and planning tool for forward‐looking discipline groups, enabling the sharing of good practices for enhancing teaching and learning within and across disciplines.
Discussions at the conference emphasized the profound impact of technological advancements on the educational landscape. The rapid changes in higher education, particularly those resulting from the COVID‐19 pandemic and advancements in AI, were key focal points. The integration of technology into teaching and learning practices has the ability to enhance engagement, accessibility, and efficiency. Transformative digital learning tools enable the next generation of scientists to develop an experimental mindset, build scientific skills, competencies and attitudes, and be prepared for the future workplace. Technologies support the creation of interactive, problem‐solving laboratory simulations and advanced digital worksheets that enable self‐led learning and improvement. These digital tools enable students to practice, to apply feedback, and to master the advanced skills needed for excellent scientific literacy. The conference showcased these innovative approaches to leveraging technology, from interactive laboratory simulations and digital worksheets to AI‐driven data analysis and assessment tools. The event highlighted the importance of interdisciplinary collaboration and the integration of diverse subjects to create comprehensive and relevant curricula. The rise of AI and its impact on various fields, including drug discovery, molecular modeling, and precision medicine, necessitates a shift toward an interdisciplinary approach that equips students with the skills and knowledge to navigate the complexities of the modern scientific landscape, necessitating the fusion of diverse subjects to create comprehensive curricula. However, the ‘power and perils of AI’ were also discussed, sparking thoughtful discussions on the ethical implications, potential biases, and the need for critical thinking in the digital age. Discussions highlighted concerns about digital footprints and the environmental impact, the importance of critical evaluation, and responsible use of technology. The need for tailored approaches in teaching and learning were emphasized ensuring that technology serves as a complement to, rather than a replacement for, traditional teaching methods. The conference underscored the importance of equipping educators with the skills and knowledge to navigate the complexities of the digital landscape, ensuring that technology serves as an enabler rather than a disruptor of effective learning.
A key theme that emerged from the conference was the critical importance of active learning, particularly in addressing the engagement challenges often faced in large classes. Presenters and participants alike emphasized that active learning is not just a trendy concept, but a crucial strategy for improving student engagement and learning, deepening understanding, and fostering critical thinking skills. Several innovative approaches to active learning were showcased: Interactive Activities: Presenters demonstrated how incorporating hands‐on, problem‐solving tasks can transform traditionally passive lectures into dynamic learning experiences. These activities ranged from short, in‐class exercises to more complex, technology‐enabled simulations. Group Discussions: The power of peer‐to‐peer learning was highlighted through structured group discussions. Techniques such as think‐pair‐share and jigsaw discussions were presented as effective ways to promote collaborative learning and expose students to diverse perspectives. Peer Teaching: Conference participants shared success stories of implementing peer teaching strategies, where students take turns explaining concepts to their classmates. This approach not only reinforces learning for the student–teacher but also provides relatable instruction for the listeners. Technology‐Enhanced Learning: Various digital tools and platforms were discussed as means to facilitate active learning in large classes. These included real‐time polling systems, collaborative online workspaces, and gamified learning applications. Flipped Classroom Model: Several presenters advocated for the flipped classroom approach, where lecture content is accessed online before class, allowing in‐person time to be dedicated to active problem‐solving and discussions. The conference also addressed the challenges of implementing active learning strategies, particularly in large classes. Presenters discussed methods for managing logistics, ensuring equitable participation, and assessing learning outcomes in active learning environments. The importance of faculty development programs to support educators in adopting these methods was also emphasized. As the conference concluded, there was a clear consensus that the future of education lies in moving away from passive, lecture‐based instruction toward more engaging, student‐centered approaches. The challenge now lies in scaling these practices across institutions and disciplines, ensuring that active learning becomes the norm rather than the exception in higher education.
The transition from high school to university is a pivotal juncture in a student's academic journey. Acknowledging this significant leap from school to university and recognizing the significant challenges faced by students during this transition are critically important. Identifying obstacles that hinder effective learning during the transition and addressing ‘dislocation, loss of belonging, and unfamiliarity’ experienced by students necessitate proactive measures to ensure a smooth and successful passage. The conference explored strategies to empower students during this critical phase, emphasizing the importance of bridging the gap between secondary and tertiary education. Strategies to support students during this transition were discussed, including the use of technology to create flexible learning opportunities and the importance of fostering a sense of belonging, adaptability, and self‐regulated learning. Strategies such as mentorship programs, orientation initiatives, and the cultivation of a supportive learning environment were proposed to empower students during this critical phase.
The role and importance of formative feedback in reinforcing self‐regulated learning strategies, enhancing motivation, and improving academic outcomes were emphasized. Techniques to enhance students' self‐assessment abilities and the use of feedback to drive lifelong learning were key points of discussion. The concepts of ‘self‐assessment ability’ and ‘self‐regulation promoting students’ self‐evaluation skills' were explored and the Equity, Agency and Transparency (EAT) Assessment Framework was highlighted as a research‐informed approach and tool for educators to implement effective inclusive assessment practices.
The cultivation of transferable skills, such as communication, collaboration, and problem‐solving, was deemed essential to prepare students for the complexities of the 21st‐century workforce. The conference explored innovative approaches to integrating these skills into the science curriculum, emphasizing the importance of self‐reflection and mentorship in fostering their development. Using arts to nurture reflective attitudes and the importance of interdisciplinary approaches in education were highlighted. Encouraging learners to engage with art as a means of self‐discovery and critical thinking and to reflect on their experiences can lead to deeper learning and personal growth. The professional development of all those who teach was recognized as a cornerstone of educational excellence. The conference emphasized the need for continuous learning and growth among educators, fostering a culture of reflection, innovation, and collaboration. The National Professional Development Framework of Ireland, which supports all who teach in higher education by providing a structured outline of activities aligned with teaching and learning goals was emphasized. Launched in 2016, this framework aims to empower staff to create, discover, and engage in meaningful personal and professional development, to engage in peer dialog and support and to enhance and develop the pedagogy of individual disciplines and to enable learning from other disciplines. The importance of mentorship and peer support was highlighted, along with the value of structured professional development programs that cater to the diverse needs and aspirations of educators. The FEBS Education and Training Academy, launched during the conference, exemplifies this commitment to empowering educators through high‐quality training and fostering a community dedicated to excellence in science education (Fig. ). The need for tailored approaches to enhance disciplinary excellence was also underscored, with The Disciplinary Excellence in Teaching, Learning, and Assessment (DELTA) Framework in Ireland serving as a model for a strategic path for disciplines to enhance teaching. The framework is recognized through the DELTA National Award, which aims to support staff across disciplines to work collaboratively to articulate, evidence, and plan their engagement in and commitment to teaching and learning enhancement, toward student success. Such a model provides a capacity building and planning tool for forward‐looking discipline groups, enabling the sharing of good practices for enhancing teaching and learning within and across disciplines.
The insights gained from the FEBS Education and Training Conference highlight the need for transformative changes in molecular life sciences education. To address the identified challenges and opportunities identified, we propose the following actionable strategies that combine our recommendations with a call to action for all stakeholders in the field (Fig. ): Embrace technological integration and digital literacy Actions for Institutions: ○Develop and implement policies that support digital and open teaching and learning, including the promotion of Open Educational Resources (OER) to ensure equity and accessibility in education, allowing resources to be freely accessible, shared, and adapted. ○Provide training for educators on the effective use of AI and digital tools to enhance teaching and learning experiences. ○Recognize teaching excellence with similar esteem to research excellence, in career evaluations. Actions for Educators: ○Thoughtfully integrate technology as a powerful tool to enhance learning experiences and to streamline administrative tasks. ○Engage in training programs to understand how to embed active learning in teaching, and how to effectively use AI and digital tools to enhance teaching and learning experiences. Actions for Students: Actively participate in digital learning opportunities and develop critical digital literacy skills. Foster active learning and student engagement Action for Institutions: Invest in redesigning learning spaces to facilitate active learning strategies, even with large classes. Action for Educators: Embed diverse active learning techniques such as group discussions, problem‐solving activities, and peer teaching. Action for Students: Embrace opportunities for collaborative learning and take ownership of the personal educational journey. Support inclusive transitions and diversity Action for Institutions: Establish comprehensive support systems to ease the transition from school to university, addressing challenges like dislocation and loss of a sense of belonging. Action for Educators: ○Create inclusive learning environments that cater to diverse student needs and promote emotional resilience. ○Utilize technology to offer flexible and asynchronous learning opportunities that promote self‐regulated learning and adaptability. Action for Students: Engage with support services and contribute to creating a welcoming community for all learners. Implement formative assessment and self‐regulated learning Action for Institutions: Adopt frameworks like the EAT framework to guide assessment practices. Action for Educators: Develop assessment literacy and integrate formative assessment techniques to enhance self‐regulated learning and lifelong learning competencies. Action for Students: Actively engage in self‐assessment and use feedback to drive continuous improvement. Promote interdisciplinary and reflective learning Action for Institutions: Encourage cross‐departmental collaboration to develop interdisciplinary curricula. Action for Educators: Integrate arts and interdisciplinary approaches in the curriculum to foster reflective attitudes and critical thinking and to promote a sense of belonging. Action for Students: Embrace opportunities for interdisciplinary learning and engage in reflective practices to deepen your understanding. Cultivate professional development and learning communities Action for Institutions: Establish structured professional development programs that cater to the diverse needs of educators and foster the creation of communities of practice for sharing practices of teaching and learning. Action for Educators: Actively participate in mentorship programs, peer support networks, communities of practice, and other continuous learning opportunities. Action for Students: Seek mentorship opportunities and contribute to peer learning communities. Foster innovation and evidence‐based teaching Action for Institutions: Create ‘mistake‐friendly’ environments that encourage experimentation and innovation in teaching methods; cultivate and celebrate excellence in teaching. Action for Educators: Adopt evidence‐based teaching practices and consider contributing to educational research in molecular life sciences. Action for Students: Provide constructive feedback on teaching methods and participate in educational research studies. We call upon all stakeholders—policymakers, institutional leaders, educators, and students—to embrace these strategies and to take concrete steps toward their implementation. Together, we can create a dynamic, inclusive, and forward‐thinking educational ecosystem that prepares molecular life scientists to shape the future of our field and to contribute meaningfully to society. This concerted effort will ensure that the next generation of scientists is equipped with the knowledge, skills, and values necessary to tackle complex global challenges and to drive innovation in the field.
Actions for Institutions: ○Develop and implement policies that support digital and open teaching and learning, including the promotion of Open Educational Resources (OER) to ensure equity and accessibility in education, allowing resources to be freely accessible, shared, and adapted. ○Provide training for educators on the effective use of AI and digital tools to enhance teaching and learning experiences. ○Recognize teaching excellence with similar esteem to research excellence, in career evaluations. Actions for Educators: ○Thoughtfully integrate technology as a powerful tool to enhance learning experiences and to streamline administrative tasks. ○Engage in training programs to understand how to embed active learning in teaching, and how to effectively use AI and digital tools to enhance teaching and learning experiences. Actions for Students: Actively participate in digital learning opportunities and develop critical digital literacy skills.
Action for Institutions: Invest in redesigning learning spaces to facilitate active learning strategies, even with large classes. Action for Educators: Embed diverse active learning techniques such as group discussions, problem‐solving activities, and peer teaching. Action for Students: Embrace opportunities for collaborative learning and take ownership of the personal educational journey.
Action for Institutions: Establish comprehensive support systems to ease the transition from school to university, addressing challenges like dislocation and loss of a sense of belonging. Action for Educators: ○Create inclusive learning environments that cater to diverse student needs and promote emotional resilience. ○Utilize technology to offer flexible and asynchronous learning opportunities that promote self‐regulated learning and adaptability. Action for Students: Engage with support services and contribute to creating a welcoming community for all learners.
Action for Institutions: Adopt frameworks like the EAT framework to guide assessment practices. Action for Educators: Develop assessment literacy and integrate formative assessment techniques to enhance self‐regulated learning and lifelong learning competencies. Action for Students: Actively engage in self‐assessment and use feedback to drive continuous improvement.
Action for Institutions: Encourage cross‐departmental collaboration to develop interdisciplinary curricula. Action for Educators: Integrate arts and interdisciplinary approaches in the curriculum to foster reflective attitudes and critical thinking and to promote a sense of belonging. Action for Students: Embrace opportunities for interdisciplinary learning and engage in reflective practices to deepen your understanding.
Action for Institutions: Establish structured professional development programs that cater to the diverse needs of educators and foster the creation of communities of practice for sharing practices of teaching and learning. Action for Educators: Actively participate in mentorship programs, peer support networks, communities of practice, and other continuous learning opportunities. Action for Students: Seek mentorship opportunities and contribute to peer learning communities.
Action for Institutions: Create ‘mistake‐friendly’ environments that encourage experimentation and innovation in teaching methods; cultivate and celebrate excellence in teaching. Action for Educators: Adopt evidence‐based teaching practices and consider contributing to educational research in molecular life sciences. Action for Students: Provide constructive feedback on teaching methods and participate in educational research studies. We call upon all stakeholders—policymakers, institutional leaders, educators, and students—to embrace these strategies and to take concrete steps toward their implementation. Together, we can create a dynamic, inclusive, and forward‐thinking educational ecosystem that prepares molecular life scientists to shape the future of our field and to contribute meaningfully to society. This concerted effort will ensure that the next generation of scientists is equipped with the knowledge, skills, and values necessary to tackle complex global challenges and to drive innovation in the field.
The Molecular Life Sciences Education Conference highlighted the urgent need for innovation and collaboration in education. This document serves as a guiding framework for future actions, emphasizing the importance of education, and training in shaping the next generation of scientists. The insights, recommendations, and collaborative spirit that emerged from the conference can undoubtedly shape the future of education, ensuring that the next generation of scientists is equipped with the knowledge, skills, and values to tackle the complex challenges that lie ahead. As we move forward, it is imperative that we continue to build upon the momentum generated by the FEBS ETC. By embracing innovation, fostering collaboration, and prioritizing the needs of our students, we can create a brighter future for molecular life sciences education, one that empowers learners to become agents of change in a world that desperately needs their expertise and passion. By working together, we can overcome challenges and ensure a bright future for molecular life sciences education.
The authors declare no conflict of interest.
LV: participated in writing the manuscript, providing initial material; NS: participated in writing the manuscript, design of illustrative material; LVM: participated in writing and correcting the manuscript; MJC: participated in writing and correcting the manuscript; DP: participated in writing and correcting the manuscript; FM: participated in writing the manuscript, providing initial material; JD: participated in writing the manuscript; FGS: participated in writing the manuscript, preparing overall structure of the manuscript.
|
Liver Transplantation for Acute Liver Failure due to Mushroom Poisoning | 32571509-b30e-427e-bd39-233dbc3e3a49 | 11843272 | Surgical Procedures, Operative[mh] | Acute liver failure is defined as severe acute liver injury of shorter than twenty-six weeks’ duration with encephalopathy and impaired synthetic function (INR ≥ 1.5) in a patient without cirrhosis or preexisting liver disease. Liver transplantation is a life-saving, curative treatment for patients exposed to mushroom poisoning. There is no clear consensus on which patients and when liver transplantation should be performed. Timely referral of patients to experienced centers is very important in terms of patient management and decision-making. The most common causes of acute liver failure, which is characterized by liver dysfunction, neurological dysfunction, and coagulopathy, are drug-induced hepatotoxicity, viral hepatitis, Wilson’s disease, Budd-Chiari Syndrome, and mushroom poisoning. The Amanita phalloides is considered one of the most perilous mushrooms, as it is responsible for the majority of fatal cases of mushroom poisoning in humans worldwide. Among approximately 100 000 types of mushrooms worldwide, about 100 are poisonous to humans and can develop a clinical syndrome of intoxication. In most instances, the clinical scenario starts with gastrointestinal tract symptoms, followed by acute hepatic decompensation and multiple organ failure, which could ensue, with a recorded mortality rate ranging between 4.8% and 34.5%. , Acute liver injury secondary to mushroom poisoning has been managed by detoxifying agents like N-acetylcysteine, silymarin, and penicillin. Supportive treatments with activated charcoal have been reported as well. In more severe situations, plasmapheresis or extracorporeal liver support has been adopted. , When the clinical situation ends up with acute liver failure, manifested by progressive liver dysfunction, neurological impairment, and coagulopathy emergency liver transplantation (LT) may be a life-saving necessity. , However, LT is an ultra-major procedure with numerous logistics and officials that are not readily available in all regions or medical centers, as well as its overwhelming expenses. In addition, the possible complications and the need for life-long immunosuppression are among the drawbacks of LT in this clinical setting. In essence, it has been admitted that emergent LT is associated with lower survival rates compared to LT in the elective setting practiced for various chronic liver diseases. In this retrospective study, we presented the outcome of acute liver failure cases secondary to mushroom poisoning that were subjected to emergency LT in Liver Transplant Institutions in Türkiye. In this descriptive study, the demographic and clinical data of 26 patients who presented to emergency department throughout Türkiye with the clinical features of acute hepatic failure secondary to mushroom poisoning between October 2008 and November 2023 and who underwent emergency LT after fulfilling national criteria (mentioned later) were retrospectively reviewed. Patients with lacking medical records, patients <18 years old age, and patients who presented to emergency units with hepatic injury secondary to causes other than mushroom ingestion were excluded from the study. Age (years), gender (male and female), body mass index (BMI; kg/m 2 ), patient blood group (O, A, B, AB), presence and level of hepatic encephalopathy (grade I-II-III-IV), type of transplantation (deceased or living donor LT), duration of stay in the intensive care unit, graft type of the patients registered in the Turkish Ministry of Health data system, graft loss, cold ischemia time (hours), warm ischemia time (minutes), post-operative complications and follow-up (days) period after emergent LT were reviewed. In addition, partial thromboplastin time (PTT; seconds), international normalized ratio (INR), albumin (g/dL), creatinine (mg/dL), aspartate aminotransferase (AST; U/L), alanine transaminase (ALT; U/L), gamma glutamyl transpeptidase (GGT; IU/L), alkaline phosphatase (ALP), ammonia (ug/dL), white blood cells (WBC; 103/uL), platelet (103/uL) and hemoglobin (HGB; g/dL) values, which were obtained during their first admission to the hospital and before discharge or death after emergency LT were examined. The model for end-stage liver disease (MELD) score was calculated for each case using pocket program ( https://www.mdcalc.com/calc/10437/model-end-stage-liver-disease-meld ). Since we did not have any patients with a MELD score between 1 and 10 points, they were not included in the grouping. Definition of Acute Liver Failure Acute liver failure is defined as severe acute liver injury of shorter than twenty-six weeks’ duration with encephalopathy and impaired synthetic function (INR ≥ 1.5) in a patient without cirrhosis or preexisting liver disease. The diagnosis of acute liver failure induced by mushroom poisoning was based on: (i) the recent ingestion of wild mushrooms associated with watery diarrhea, vomiting, and/or abdominal pain within 24 hours after ingestion; (ii) the clinical and laboratory criteria of acute liver failure; and (iii) the absence of any other cause for acute liver failure. Criteria for Emergent LT The decision for emergent LT was based on King’s College criteria for non-acetaminophen causes. These criteria are as follows: INR > 6.5 or presence of three of the following five following criteria: (i) age (≤10 or >40 years), (ii) certain etiology (non-A, non-B viral hepatitis, and other drugs), (iii) interval between jaundice and encephalopathy (>7 days), (iv) INR > 3.5, (v) serum total bilirubin level (>17.4 mg/dL). Study Protocol and Ethics Committee Approval This study involved human participants and abided by the ethical standards of the institutional and national research committee, as well as the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. Ethical approval was obtained from the institutional review board (IRB) of Ankara City Hospital for Non-Interventional Clinical Research (approval no: 2022/ E2-22-18844; date: May 27, 2022). Statistical Analysis IBM SPSS Statistics software version 25.0 (IBM SPSS Corp.; Armonk, NY, USA) was used for this study. Qualitative variables were given as numbers (percentage), while quantitative variables were given as median and interquartile range (IQR = Q3-Q1). Mann–Whitney U-test was used to compare quantitative data, and Chi-square test was used to compare qualitative data. P < .05 was considered statistically significant. Acute liver failure is defined as severe acute liver injury of shorter than twenty-six weeks’ duration with encephalopathy and impaired synthetic function (INR ≥ 1.5) in a patient without cirrhosis or preexisting liver disease. The diagnosis of acute liver failure induced by mushroom poisoning was based on: (i) the recent ingestion of wild mushrooms associated with watery diarrhea, vomiting, and/or abdominal pain within 24 hours after ingestion; (ii) the clinical and laboratory criteria of acute liver failure; and (iii) the absence of any other cause for acute liver failure. The decision for emergent LT was based on King’s College criteria for non-acetaminophen causes. These criteria are as follows: INR > 6.5 or presence of three of the following five following criteria: (i) age (≤10 or >40 years), (ii) certain etiology (non-A, non-B viral hepatitis, and other drugs), (iii) interval between jaundice and encephalopathy (>7 days), (iv) INR > 3.5, (v) serum total bilirubin level (>17.4 mg/dL). This study involved human participants and abided by the ethical standards of the institutional and national research committee, as well as the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. Ethical approval was obtained from the institutional review board (IRB) of Ankara City Hospital for Non-Interventional Clinical Research (approval no: 2022/ E2-22-18844; date: May 27, 2022). IBM SPSS Statistics software version 25.0 (IBM SPSS Corp.; Armonk, NY, USA) was used for this study. Qualitative variables were given as numbers (percentage), while quantitative variables were given as median and interquartile range (IQR = Q3-Q1). Mann–Whitney U-test was used to compare quantitative data, and Chi-square test was used to compare qualitative data. P < .05 was considered statistically significant. General Evaluation of the Study Group Twenty-six patients diagnosed with acute liver failure secondary to mushroom poisoning who underwent emergency LT were included in the study. Twelve (46.2%) of the patients were male and the remaining 14 (53.8%) were female. The median (IQR) age, BMI, MELD, from poisoning to hospital admission, ICU stay, hospital stay, CIT, and WIT were 39 (36) years, 25 (8) kg/m 2 , 25 (14) points, 6 (21) days, 4 (4) days, 18 (23) days, 3 (6) hours, and 48 (15) minutes, respectively. Subgroup Analysis Patients were divided into two groups: alive (alive group; n = 18; 69.2%) and dead (dead group; n = 8; 30.8%). Due to the nature of mushroom consumption, there were no cases of mushroom poisoning during the winter season. The median (IQR) age of the dead group was 53 (31) and in alive group , it was 38 (28) years ( P = .196). The mortality rate was 16.7% in males and 42.8% in females but there were no statistically significant differences between groups in terms of gender ( P = .216). The median (IQR) BMI was 24 (5) for the alive group and 26 (10) for the dead group ( P = .701). The median (IQR) MELD score was 23 (11) points for the alive group and 34 (8) points for the dead group and statistically significant differences were found between groups in terms of MELD scores ( P = .016). Deceased donor liver transplantation was performed in 14 (53.8%) patients, and LDLT was performed in 12 (46.2%) patients. Twelve DDLT patients (85.7%) received a whole liver graft, and remaining two received a split liver graft. Regarding blood group, 17 (65.4%) were blood group A, 5 (19.2%) were blood group O, 3 (11.5%) were blood group B, and 1 (3.8%) was group AB. The rate of patients with Rh was found to be 84.6%, and the rate of patients with Rh (−) was 15.4%. The median (IQR) CIT was 3 (6) hours for the alive group and 5 (8) hours for the dead group ( P = .528). The median (IQR) WIT was 45 (16) minutes for the alive group and 50 (15) minutes for the dead group ( P = .964). The median (IQR) ICU stay was 4 (4) days for the alive group and 5 (6) days for the dead group ( P = .834). The median (IQR) hospital stay was 24 (22) days for the alive group and 6 (6) days for the dead group ( P = .004). Since patients who develop mortality in the early postoperative period, the hospital stay is very low as expected in the dead group. While none of the patients in the alive group were re-transplanted, three of the patients in the dead group were re-transplanted ( P = .022). The indication for re-transplantation in all these patients was postoperative primary graft dysfunction. There was no difference in the degree of hepatic encephalopathy between the alive and dead groups ( P = .401). There were no statistically significant differences between the alive and dead groups in terms of AST ( P = 1.000), ALT ( P = .928), total bilirubin ( P = .052), Creatinine ( P = .383), Ammonia ( P = .278), PTT ( P = .928), INR ( P = .591), Albumin ( P = .976), Platelets ( P = .888), and HGB ( P = .541) levels on the first days of hospital admission. Details were summarized in . Twenty-six patients diagnosed with acute liver failure secondary to mushroom poisoning who underwent emergency LT were included in the study. Twelve (46.2%) of the patients were male and the remaining 14 (53.8%) were female. The median (IQR) age, BMI, MELD, from poisoning to hospital admission, ICU stay, hospital stay, CIT, and WIT were 39 (36) years, 25 (8) kg/m 2 , 25 (14) points, 6 (21) days, 4 (4) days, 18 (23) days, 3 (6) hours, and 48 (15) minutes, respectively. Patients were divided into two groups: alive (alive group; n = 18; 69.2%) and dead (dead group; n = 8; 30.8%). Due to the nature of mushroom consumption, there were no cases of mushroom poisoning during the winter season. The median (IQR) age of the dead group was 53 (31) and in alive group , it was 38 (28) years ( P = .196). The mortality rate was 16.7% in males and 42.8% in females but there were no statistically significant differences between groups in terms of gender ( P = .216). The median (IQR) BMI was 24 (5) for the alive group and 26 (10) for the dead group ( P = .701). The median (IQR) MELD score was 23 (11) points for the alive group and 34 (8) points for the dead group and statistically significant differences were found between groups in terms of MELD scores ( P = .016). Deceased donor liver transplantation was performed in 14 (53.8%) patients, and LDLT was performed in 12 (46.2%) patients. Twelve DDLT patients (85.7%) received a whole liver graft, and remaining two received a split liver graft. Regarding blood group, 17 (65.4%) were blood group A, 5 (19.2%) were blood group O, 3 (11.5%) were blood group B, and 1 (3.8%) was group AB. The rate of patients with Rh was found to be 84.6%, and the rate of patients with Rh (−) was 15.4%. The median (IQR) CIT was 3 (6) hours for the alive group and 5 (8) hours for the dead group ( P = .528). The median (IQR) WIT was 45 (16) minutes for the alive group and 50 (15) minutes for the dead group ( P = .964). The median (IQR) ICU stay was 4 (4) days for the alive group and 5 (6) days for the dead group ( P = .834). The median (IQR) hospital stay was 24 (22) days for the alive group and 6 (6) days for the dead group ( P = .004). Since patients who develop mortality in the early postoperative period, the hospital stay is very low as expected in the dead group. While none of the patients in the alive group were re-transplanted, three of the patients in the dead group were re-transplanted ( P = .022). The indication for re-transplantation in all these patients was postoperative primary graft dysfunction. There was no difference in the degree of hepatic encephalopathy between the alive and dead groups ( P = .401). There were no statistically significant differences between the alive and dead groups in terms of AST ( P = 1.000), ALT ( P = .928), total bilirubin ( P = .052), Creatinine ( P = .383), Ammonia ( P = .278), PTT ( P = .928), INR ( P = .591), Albumin ( P = .976), Platelets ( P = .888), and HGB ( P = .541) levels on the first days of hospital admission. Details were summarized in . Liver transplantation is considered the biggest development in the management of numerous etiologies of acute liver failure in the last years. Though emergent LT has inferior survival rates compared to elective LT, a 1-year survival rate of as much as 80% can still be achieved in the emergency setting LT. Early identification of acute liver failure that is not responding to medical treatment and could otherwise benefit from LT is crucial in the decision-making process of LT in the emergency setting. In this study, we managed to identify those early determinants of the indication of emergency LT. We found that the mortality rate among female patients undergoing emergent LT was double the rate among male patients, though the rate of emergent LT procedures was equal among both genders. This is somewhat different from a recent study that found a similar mortality rate among both genders, though male patients were more often subjected to emergency LT. Another 10-years study also reported a higher rate of emergent LT among male patients (from 59.4% to 62.6%) presenting with acute liver failure. It is obvious that this situation, which does not show statistical significance but shows a significant difference proportionally, needs to be supported by other clinical studies. In our opinion, many factors such as the retrospective nature of the study, the small sample size, and the absence of a control group without liver transplantation despite having mushroom poisoning may affect this situation. Data extracted from the United network for organ sharing (UNOS) and the European liver transplant registry (ELTR) showed worse outcomes in those recipients aged 50 and over. , Our results were aligned with this finding. Around 60% of mortalities among our recipients were over 50 years old. We concur with the point that old age is associated with worse outcomes in the emergency LT. In our study, it was found that 50% of the cases who underwent emergency LT presented to emergency department with mushroom poisoning during the autumn season. Interestingly, the same finding had been reported in an epidemiological study of mushroom poisoning in USA. It has been mentioned that the 1-year survival rate among emergency LT recipients is lower compared to chronic liver disease recipients who underwent surgery under elective circumstances. , However, it should be noted that the survival rates for acute liver failure patients who underwent emergency LT have improved significantly over the last three decades, with estimated 1- and 5-year survival rates of 79% and 72% reported from a European study. Another study from the USA reported 1- and 5-year survival rates of 85% and 69%, respectively. In a recent study from Türkiye Ertugrul and colleagues, the 1- and 5-year overall survival rates of 13 patients who underwent LDLT due to acute liver failure were 84.7% and 69.3%, respectively. In our cohort, the 1- and 5-year overall survival rates were 69.2% and 69.2%, being lower than those reported in the aforementioned studies. This could be explained by the fact that all of our patients underwent emergency LT while presenting with higher grades of hepatic encephalopathy and poor general condition. The most frequently encountered complication was post-operative disturbed consciousness and delayed neurological recovery. Similar studies showed that the most common cause of death in emergency LT patients was infection. , We didn’t encounter infectious complications among our recipients, though. In fact, the post-operative outcome after emergency LT for acute liver failure is influenced by multiple factors. , From a study that included 75 thousand cases and spanned 13 years, the blood groups of emergency LT recipients were as follows: blood group O (44.01%), blood group A (37.6%), and (5.02%) were blood group AB (14). Another study of a 10-year period showed that blood group O recipients constituted 49.2%, followed by blood group A (36.7%), blood group B (11.5%), and finally 2.6% were blood group AB (12). Though the small sample size of our study, our data is similar to the reported literature. We found 66.6% of our patients were blood group A, 25% were blood group O, and 8.4% were blood group B. Therefore, most acute liver failure cases in need for emergency LT were among blood group A and O. Like the reported literature, it appears that AB blood group patients were less likely to receive an emergency LT in the acute setting. , It is well appreciated that the mere presence of hepatic encephalopathy early in the course of acute liver failure underlies a poor prognosis after emergency LT. If renal failure or another organ dysfunction ensues, the prognosis could be worse. Though the LT candidacy criteria could vary among transplant institutions, the presence of hepatic encephalopathy is one of the main criteria in the decision-making for emergency LT. In patients who progress to grade III-IV hepatic encephalopathy, the mortality rate could reach between 40% and100%. In our study, a relationship was found between advanced-grade hepatic encephalopathy and high mortality rate. Model for end-stage liver disease score is a consistent tool for prioritizing enlisted patients for LT since 2002. It ranges between 6 and 40 with patients with higher scores being of poor prognosis. , , Consistent with the literature, MELD scores were found to be high in the patients who died in our study. There was one mortality in 6 patients with MELD scores between 11 and 20. One out of 8 patients with MELD scores between 21 and 30 died. Again, 6 out of 8 patients whose MELD scores were between 31 and 40 died ( P = .017). In other words, the MELD score, which consists of bilirubin, INR, and creatinine components, is an important factor associated with mortality in acute liver failure. However, the sample size of this study is the most important obstacle to giving a strong message. In a study by Kim and colleagues, patients with acute liver failure and a body mass index of more than 30 kg/m 2 were noted to have worse outcome when they underwent emergency LT. In our study, we didn’t find a significant difference in BMI among recipients who died, and was not considered a risk factor from a statistical standpoint. Coagulation impairment, namely INR and PTT rise, are an integral components of acute liver failure diagnosis. It results from acute liver insults that cause massive liver damage and associated liver dysfunction. Many studies have shown high INR and ammonia levels to be among the most important prognostic factors. , Similarly, median ammonia values were observed to be high in the patients in our study, but this elevation did not reach to the statistical significance. Albumin is an important surrogate of the liver’s synthetic function. Its plasma level decreases in all diseases with liver dysfunction. In mushroom poisoning, the decline in plasma albumin levels will parallel the liver damage and the degree of liver dysfunction. In addition, it is a component of the Child-Pugh score used for decades in many regions to categorize liver performance. Studies have found that a decrease in albumin levels is associated with mortality. In our study, albumin levels were not found to be a significant factor. Alanine aminotransferase and AST are enzymes found extensively in hepatocytes. Elevations in plasma ALT and AST levels are associated with hepatocyte damage, and the degree of rise correlates with the degree of liver damage. Mushroom poisoning is associated with a massive rise in AST and ALT. Different studies have shown a poor prognosis in emergency LT recipients who underwent surgery with increased plasma levels of AST, ALT, INR, BUN, and Creatine. , A recent study by Badsar and colleagues reviewed the laboratory results of patients with mushroom poisoning taken at the time of admission and observed that 28.4% of the patients had leukocytosis as well as a platelet count below 100 000. They also found that 5.9% of patients had elevated plasma AST levels and 9.8% had elevated plasma ALT values, and those patients had a worse prognosis. In our study, there was no statistically significant correlation between the serum level of transaminases or total bilirubin and the emergency LT outcome. It is thought that this result is due to the development of fulminant hepatitis in the patients. Liver transplantation in acute liver failure, especially in mushroom poisoning, is a situation that requires rapid decision-making. In Western countries, the use of LDLT in acute liver failure is still a controversial issue due to many factors such as donor safety, approval process, and recipient outcomes, largely because the use of cadaveric organs is higher in Western countries. In many countries, including Turkey, there is no other option other than using living donors in more than half of the patients in emergent conditions. A systematic review published in recent years has shown that LDLT in acute liver failure provides similar results to DDLT. Another study has shown that donor complications are similar in emergency and elective liver transplantation. This study has shown that there is no difference between DDLT and LDLT procedures in acute liver failure. When the complications in living liver donors used for urgent LDLT are examined, it is seen that they are no different from those performed in elective LDLT. This study has several limitations. Firstly, since the study is retrospective, it is difficult to access all of the patient data. In order to avoid such problems, joint working groups and software systems should be developed across the country, so that it is known which parameters to record in patients applying to health care centers with acute liver disease, thus preventing data loss. Secondly, the number of patients evaluated in the study was insufficient to put forward strong arguments. To overcome this problem, multicentric studies should be organized. Thirdly, there is no analysis regarding who and under what circumstances LT should be performed in acute liver failure. To overcome this problem, a control group should be created among patients who had mushroom poisoning and did not undergo LT in the same period, and both groups should be compared in terms of demographic and clinical parameters. In conclusion, despite the limitations mentioned above, this study is one of the most comprehensive studies in the literature in which acute liver failure due to mushroom poisoning and liver transplantation were examined. This study showed that mortality was higher in patients with high MELD scores or patients who underwent re-transplantation. The data that support the findings of this study are available on request from the corresponding author. |
COL11A1 as an novel biomarker for breast cancer with machine learning and immunohistochemistry validation | b0f7f2be-dc84-437e-b257-53c69b2184ed | 9660229 | Anatomy[mh] | Breast cancer is one of the most commonly diagnosed malignant tumors in the world. As the most frequent malignant tumor in women, more than 2.1 million women have been diagnosed with breast cancer in 2018, and approximately 500,000 women have died from this disease . Although advances in early detection and effective systemic treatment have decreased breast cancer mortality rates in North America and the European Union, breast cancer remains the most common cause of cancer death in less developed countries, second only to lung cancer, and almost all patients in the advanced stage have a poor prognosis . Therefore, new therapeutic approaches and goals need to be developed to reduce disease recurrence and death. With the advances in machine learning, we have achieved great success for disease diagnosis, risk stratification, and the establishment of prognostic models , such as using medical imaging and artificial intelligence for the identification of lesions , the discovery of new biomarkers through data mining, drug discovery, and risk model construction . Traditionally, machine learning approaches are divided into supervised learning, unsupervised learning, and reinforcement learning categories. We can predict and classify huge data using machine learning algorithms based on known training data. As reported by Rahman , Linear Regression (LR), Support Vector Machine (SVM), Multi-Layer Perceptron (MLP), and Vector Auto-Regression have been the most widely used algorithms for tackling the Coronavirus pandemic (COVID-19). Thus, our aim was to identify potential prognosis-related biomarkers in breast cancer by computational approaches to assist clinical decision-making. In recent years, immunotherapy has emerged as a novel option for a variety of solid tumors . Unlike other solid cancers, breast cancer is insensitive to immunotherapy. While the recognition of the importance of the tumor microenvironment (TME) in breast cancer progression, response to treatment, and resistance, the assessment of its immune infiltration and stromal cell infiltration has opened the opportunity for breast cancer immunotherapy. Retrospective studies have shown that patients with breast cancer with higher levels of stromal-infiltrating immune cells generally have longer progression-free survival (PFS) and overall survival (OS) , and the results of immune checkpoint inhibitor (ICI) therapies for TNbreast cancer are encouraging . Studies are ongoing to unravel the immunoediting function of the host immune system in breast cancer to identify patients who will benefit from therapy . Collagen type XI alpha 1 (COL11A1) is a type XI collagen, which belongs to the collagen family. Although it is mainly involved in the biological process of bone development , high levels of COL11A1 are associated with tumor metastasis, treatment resistance, and poor clinical outcome in several solid tumors types such as breast, pancreas, and colorectal cancers . Gu et al. showed that COL11A1 was highly expressed in breast cancer tissues, and COL11A1 variant E was also significantly correlated with lymph nodes involvement and metastasis in breast cancers . As an important component of the structure of the extracellular matrix (ECM), COL11A1 was identified as a correlated predictor of dangerous immune infiltrates in pancreatic adenocarcinoma . However, the role of COL11A1 in the TME of breast cancers remains unclear.
Data sourcing and pre-processing In this study a total of six breast cancer datasets were included. Clinical and expression profile data of patients originating from The Cancer Genome Atlas Program (TCGA) dataset were downloaded using the TCGA Biolink package . GSE42568, GSE109169, GSE138536, GSE173839, and GSE103668 were derived from the GEO database. GSE42568, includes 104 breast cancer and 17 normal breast biopsies, GSE109169, includes 25 paired breast samples. GSE138536 is a single-cell sequencing data containing 8 breast cancer samples. GSE173839 and GSE103668, includes follow-up information of breast cancer patients receiving immunotherapy. Cell line expression and protein level expression data were obtained from the Cancer Cell Line Encyclopedia (CCLE) and the Clinical Proteomic Tumor Analysis Consortium (CPTAC) databases. Immune infiltration scores were evaluated using the R package ESTIMATE. Differential gene analysis of samples We performed a differential gene analysis of the breast cancer expression profile data comparing TCGA datasets of tumor and normal tissues using the edgeR package, and the threshold criteria were |LogFC| >4, and the adjust p-value less than 0.01. Machine learning identifies feature genes We first defined patients with an OS shorter than 3 years as the short-term survival group, while those with a survival time greater than 3 years were defined as the long-term survival group. We used the random survival forest to identify short-term related feature genes . Machine learning algorithm lasso regression, and Support Vector Machine (SVM) were used to select feature genes . Variables of greater importance than 0.3 in random forests were defined as significant. The lambda with the smallest value was defined as significant for lasso regression. For the SVM algorithm, the top 10 feature support vectors were defined as the important variables. The intersection genes of the three machine learning algorithms were defined as the core genes. Validated expression of hub gene Samples from TCGA and GTEX databases were used to validate the expression of the hub gene at the transcriptome level. GSE42568 and GSE109169 were also used to validate the expression differences of COL11A1 in tumor and normal tissues. Data from the CCLE database were used to validate gene expression differences between different cell lines. The protein-level expression differences of COL11A1 were performed through the CPTAC database. Analysis of the prognosis value of the hub gene To further validate the prognostic value of the core gene, we evaluated the association of the expression of the hub gene on OS, DSS and PFS, respectively, using the R survival package. The cut-off values of the patient subgroups were performed using the R package survminer and component differences were obtained by the log-rank test. Role of the hub gene in the TME The level of immune infiltration was evaluated by the ESTIMATE package, which calculated a stromal score and estimated score in each sample, according to gene expression. Additionally, the IOBR package calculated B cell, cancer-associated fibroblasts (CAF), CD4 T cell, CD8 T cell, endothelial cell, macrophage, natural killer cell, and other cell infiltration scores. The Spearman’s test was used to calculate detailed correlations between core genes and B-cell, CD4 and CD8 immune cell markers. Evaluation of prognosis was associated with the level immune infiltration was performed through the TIMER2 website . Hub gene and relationship with cancer-associated fibroblasts CAFs play an important role in tumor recurrence and resistance to therapy, as the main component of the tumor stroma. Therefore, we further evaluated the correlation of the hub gene with tumor-associated fibroblasts. We first validated the differential expression of this gene in different cell clusters in GSE138536, a single-cell data set. We then calculated the correlation between the hub gene and the classical fibroblast-associated markers, and finally, we evaluated the association between the level of infiltration and the clinical prognosis. Hub gene and relationship with immunotherapy Immunotherapy offers a new pathway for patients, but not all patients can benefit from this option, and screening of the potentially benefitting population is necessary. Considering that immune checkpoints play an important role in tumor immunotherapy, we first examined the correlation between core genes and immune checkpoints using the Pearson’s test. Then two breast cancer immunotherapy datasets, GSE173839 and GSE103668, containing follow-up information, were interrogated to verify the differential expression of the hub gene between the immune response and immune tolerance groups. Additionally, we analyzed the correlation of this gene with 21 genes related to m6A methylation. COL11A1-related immune regulation genes We extracted the expression data of COL11A1 and 150 immune regulation genes, including chemokines , receptors , MHC , immunoinhibitory genes , immunosuppressive genes (46). The Pearson’s correlation between COL11A1 and immune regulation genes was further calculated. Prognosis signature construction and validation for OS Immune regulation genes associated with COL11A1 were put into univariate and multivariate Cox regressions with OS. Univariate significance genes were included in multivariate Cox regression. Then a prognostic signature model was constructed based on the multivariate Cox regression coefficients. An area under the ROC curve (AUC) was used to test the predictive efficiency of the model. Nomogram construction and validation To assess whether the signature had an independent prognostic value compared to other clinical variables, we performed univariate and multivariate cox regression analyses and visualized the regression results to construct a Nomogram model, and the C-index and calibration curve were used to evaluate the predictive efficacy and stability of the model, respectively. Immunohistochemistry of COL11A1 between normal and tumor breast tissues The tissues were washed with PBS and then incubated with 3% H2O2 for 10 min. The antibody including against COL11A1 (1:100, 21841-1-AP, proteintech, CA) were incubated at room temperature for 2 h. After incubation with polymer enhancer for 20 min, the tissue was incubated with polymer enhancer and enzyme-labeled rabbit polymers. Slides were washed with PBS and fresh diaminobenzidine, counterstained with hematoxylin, antigen retrieval performed using 0.1% HCl, dehydrated with ethanol, cleaned with xylene, and fixed with neutral balata. The results were observed and photographed using a fluorescence microscope and visualized under a light microscope at 100× and 200x magnification by a blinded observer. Controls without primary antibodies showed no immunolabeling. Light to dark brown staining indicated a positive result.
In this study a total of six breast cancer datasets were included. Clinical and expression profile data of patients originating from The Cancer Genome Atlas Program (TCGA) dataset were downloaded using the TCGA Biolink package . GSE42568, GSE109169, GSE138536, GSE173839, and GSE103668 were derived from the GEO database. GSE42568, includes 104 breast cancer and 17 normal breast biopsies, GSE109169, includes 25 paired breast samples. GSE138536 is a single-cell sequencing data containing 8 breast cancer samples. GSE173839 and GSE103668, includes follow-up information of breast cancer patients receiving immunotherapy. Cell line expression and protein level expression data were obtained from the Cancer Cell Line Encyclopedia (CCLE) and the Clinical Proteomic Tumor Analysis Consortium (CPTAC) databases. Immune infiltration scores were evaluated using the R package ESTIMATE.
We performed a differential gene analysis of the breast cancer expression profile data comparing TCGA datasets of tumor and normal tissues using the edgeR package, and the threshold criteria were |LogFC| >4, and the adjust p-value less than 0.01.
We first defined patients with an OS shorter than 3 years as the short-term survival group, while those with a survival time greater than 3 years were defined as the long-term survival group. We used the random survival forest to identify short-term related feature genes . Machine learning algorithm lasso regression, and Support Vector Machine (SVM) were used to select feature genes . Variables of greater importance than 0.3 in random forests were defined as significant. The lambda with the smallest value was defined as significant for lasso regression. For the SVM algorithm, the top 10 feature support vectors were defined as the important variables. The intersection genes of the three machine learning algorithms were defined as the core genes.
Samples from TCGA and GTEX databases were used to validate the expression of the hub gene at the transcriptome level. GSE42568 and GSE109169 were also used to validate the expression differences of COL11A1 in tumor and normal tissues. Data from the CCLE database were used to validate gene expression differences between different cell lines. The protein-level expression differences of COL11A1 were performed through the CPTAC database.
To further validate the prognostic value of the core gene, we evaluated the association of the expression of the hub gene on OS, DSS and PFS, respectively, using the R survival package. The cut-off values of the patient subgroups were performed using the R package survminer and component differences were obtained by the log-rank test.
The level of immune infiltration was evaluated by the ESTIMATE package, which calculated a stromal score and estimated score in each sample, according to gene expression. Additionally, the IOBR package calculated B cell, cancer-associated fibroblasts (CAF), CD4 T cell, CD8 T cell, endothelial cell, macrophage, natural killer cell, and other cell infiltration scores. The Spearman’s test was used to calculate detailed correlations between core genes and B-cell, CD4 and CD8 immune cell markers. Evaluation of prognosis was associated with the level immune infiltration was performed through the TIMER2 website .
CAFs play an important role in tumor recurrence and resistance to therapy, as the main component of the tumor stroma. Therefore, we further evaluated the correlation of the hub gene with tumor-associated fibroblasts. We first validated the differential expression of this gene in different cell clusters in GSE138536, a single-cell data set. We then calculated the correlation between the hub gene and the classical fibroblast-associated markers, and finally, we evaluated the association between the level of infiltration and the clinical prognosis.
Immunotherapy offers a new pathway for patients, but not all patients can benefit from this option, and screening of the potentially benefitting population is necessary. Considering that immune checkpoints play an important role in tumor immunotherapy, we first examined the correlation between core genes and immune checkpoints using the Pearson’s test. Then two breast cancer immunotherapy datasets, GSE173839 and GSE103668, containing follow-up information, were interrogated to verify the differential expression of the hub gene between the immune response and immune tolerance groups. Additionally, we analyzed the correlation of this gene with 21 genes related to m6A methylation.
We extracted the expression data of COL11A1 and 150 immune regulation genes, including chemokines , receptors , MHC , immunoinhibitory genes , immunosuppressive genes (46). The Pearson’s correlation between COL11A1 and immune regulation genes was further calculated.
Immune regulation genes associated with COL11A1 were put into univariate and multivariate Cox regressions with OS. Univariate significance genes were included in multivariate Cox regression. Then a prognostic signature model was constructed based on the multivariate Cox regression coefficients. An area under the ROC curve (AUC) was used to test the predictive efficiency of the model.
To assess whether the signature had an independent prognostic value compared to other clinical variables, we performed univariate and multivariate cox regression analyses and visualized the regression results to construct a Nomogram model, and the C-index and calibration curve were used to evaluate the predictive efficacy and stability of the model, respectively.
The tissues were washed with PBS and then incubated with 3% H2O2 for 10 min. The antibody including against COL11A1 (1:100, 21841-1-AP, proteintech, CA) were incubated at room temperature for 2 h. After incubation with polymer enhancer for 20 min, the tissue was incubated with polymer enhancer and enzyme-labeled rabbit polymers. Slides were washed with PBS and fresh diaminobenzidine, counterstained with hematoxylin, antigen retrieval performed using 0.1% HCl, dehydrated with ethanol, cleaned with xylene, and fixed with neutral balata. The results were observed and photographed using a fluorescence microscope and visualized under a light microscope at 100× and 200x magnification by a blinded observer. Controls without primary antibodies showed no immunolabeling. Light to dark brown staining indicated a positive result.
Workflow of this study are shown in . Identification of the hub gene COL11A1 A total of 149 up-regulated genes were identified by the differential analysis of tumor and normal tissues in the TCGA dataset . The results are shown in the volcano map . Random survival forest analysis of these differential genes revealed 11 genes with an importance greater than 0.3 . Lasso regression, a machine learning algorithm, was also used for feature variable screening, and a total of 25 candidate genes were selected when the minimum value of lambda was equal to 0.018 . Conversely, the top 10 feature support vectors obtained by the SVM algorithm, were also selected as candidate genes . Inserting the above results in a Venn diagram, we found that only COL11A1 was common to the results of the three algorithms, and thus this gene was identified as the cores gene of the study . COL11A1 is highly expressed in breast cancer samples and is associated with a poor prognosis The heatmap shows the expression of the COL11A1 gene in normal breast cancer tissues and in different cancer tissues. We found that this gene was significantly highly expressed in breast cancer tissues, while it was almost absence in normal tissues . To further verify this result, we performed an expression difference analysis using tumor samples from TCGA and normal samples derived from GTEx, and obtained consistent results . Furthermore, both GEO datasets, GSE42568 and GSE 109169, also confirmed that the expression of COL11A1 was higher in the tumor compared to normal tissues . Furthermore, we also verified that the expression of this gene at the cell line and protein level, and the results suggested that COL11A1 was highly expressed in HCC38, HCC1395, MDAMB157, HCC1954, and ZR751 cell lines and had lower expression in T47D, MCF7, HCC1428, CAMA1, and BT483 cell lines . The results of the protein expression analysis suggested that COL11A1 had higher expression in tumor samples compared to normal tissues . Finally, we performed a survival analysis of this gene and found that high expression of COL11A1 was associated with a poor prognosis in patients, either in terms of OS, disease-specific survival, or progress-free interval . High expression COL10A1 promoted tumor immune infiltration The TME is closely associated with tumor progression and metastasis; thus we explored the correlation between this gene and the TME of breast cancer in the current study. The results suggested that COL11A1 expression was significantly positively correlated with the stromal score (r=0.49, p<0.001) and the ESTIMATE score (r=0.29, p<0.001) in the TME . Furthermore, the results of immune cell infiltration analysis also revealed that the expression of COL11A1 was negatively correlated with the level of B cells, CD4 and CD8 T cells and positively correlated with CAFs . Further analysis of the correlation between this gene and marker genes of B cells, CD4, and CD8 T cells, revealed that COL11A1 was significantly negatively correlated with a marker of B cells, positively correlated with a marker of CD4 T cells, and negatively correlated with a marker of T cells (r=-0.156, r=0.113 and r=-0.160, respectively; p<0.001) . The results of the immune infiltration and survival analysis suggested that in patients with low expression of COL11A1, the degree of B cell infiltration was negatively correlated with patient prognosis. This finding was also applied to the high expression group of COL11A1 expression . However, the level of CD4 and CD8 T cell infiltration was negatively correlated with patient prognosis. These findings further support the association of COL11A1 with tumor immune infiltration and patient prognosis . High expression COL11A1 positively correlated with CAFs CAFs are present in the tumor stroma and contribute to tumor invasion by promoting the epithelial-mesenchymal transition and participating in tumor angiogenesis. We first analyzed the distribution of COL11A1 in different clusters of cells at the single-cell level, and the results suggested that the data were clustered into four clusters, namely myofibroblasts, Mono/macro, epithelial, and fibroblasts . COL11A1 was distributed most significantly in myofibroblasts and fibroblasts. Additionally, the expression of this gene was significantly higher in fibroblasts than in epithelial cells . Further analysis revealed the relationship between COL11A1 and the classical CAF marker gene, and we found that the gene was significantly positively correlated with the CAF marker gene (FAP, PDPN, THY1, ACTA2, COL1A1, PDGFRA, and PDGFRB; p<0.001) . The results of the survival analysis suggested that the deeper the immune infiltration, the worse the prognosis of patients with low expression of COL11A1 and the opposite results of the prognosis analysis of patients with high expression of COL11A1 . COL11A1 predicted the response rate to immunotherapy The results of COL11A1 and immune checkpoints suggest that COL11A1 expression was positively correlated with immune checkpoints (CD276, TIGIT, and ENTPD1; p<0.001) . Further analysis of the results of two immunotherapy data sets revealed that before the analysis of two data sets, 67.39% of the genes overlapped and two data sets had a batch effect, after removal of the effect, new data did not show any batch effect . When we analyzed the differences between COL11A1 expression in the response groups and the absence of response groups, we found that COL11A1 showed high expression in the response group, although this was not significant, while compared to the absence of response samples (p=0.33) . COL11A1 with m6A methylation m6A methylation, as a modification of RNA molecules, has become a hot research topic in the life sciences field in recent years. Studies have shown that genes related to m6A methylation promote tumor progression and may mediate tumor immune tolerance. Therefore, we further analyzed the correlation between this gene and m6A methylation-related genes. The results suggest that this gene is associated with several m6A methylation-related genes . COL11A1 related immune regulation genes and the construction of a five-gene signature A total of 43 related immune regulation genes of COL11A1 were identified from breast cancer samples. To determine the prognostic value of these genes, we constructed predictive models using univariate and multivariate Cox regression. The results of the univariate analysis suggested that a total of 19 immune regulation genes were associated with the prognosis , and the multivariate results demonstrate that 5 immune regulation genes were independent risk factors associated with patient outcomes . Finally, we constructed a 5 gene signature prognostic model based on the above results. We divided patients into the high-risk and low-risk groups based on the median value of the model scores. We found that patients in the high-risk group had a worse prognosis than those in the low-risk group (HR=2.47, 95% CI 1.83-3.33, p<0.001). The area under the model’s ROC curve was 0.658, suggesting that the model had a high predictive value . The internal validation of the model also demonstrated that high-risk patients had a poorer outcome than low-risk patients (HR=2.30, 95% CI 1.47-3.60, p<0.001). The area under the ROC was 0.651 in the validation group, indicating that the model was robust . Nomogram modeling and efficacy evaluation We first evaluated the clinical value of the signature-based on univariate and multivariate COX regression analysis. Univariate results suggested that age, stage, estrogen receptor (ER) positivity, progesterone (PR) positivity, and the risk score could be risk factors that influences the prognosis of patients. The results of the multivariate analysis showed that age, stage, ER status, and risk score were independent prognostic factors for patients . Finally, we visualized the analysis results to construct a nomogram model to predict the OS of the patients . It is noteworthy that when we built the final version of model, PR status was included, although it showed no significance in the multivariate analysis results. Nonetheless, this variable is very important in clinical decision-making. The C-index of this model is 0.776, and the model correction prediction curve had a small bias from the ideal curve, suggesting that the model could predict the OS of patients at 5 and 10 years with more accuracy . COL11A1 expression high in breast cancer with clinical samples Our Immunohistochemistry (IHC) results demonstrate that COL11A1 could express high in breast cancer tissues while compared with normal tissues in clinical samples .
A total of 149 up-regulated genes were identified by the differential analysis of tumor and normal tissues in the TCGA dataset . The results are shown in the volcano map . Random survival forest analysis of these differential genes revealed 11 genes with an importance greater than 0.3 . Lasso regression, a machine learning algorithm, was also used for feature variable screening, and a total of 25 candidate genes were selected when the minimum value of lambda was equal to 0.018 . Conversely, the top 10 feature support vectors obtained by the SVM algorithm, were also selected as candidate genes . Inserting the above results in a Venn diagram, we found that only COL11A1 was common to the results of the three algorithms, and thus this gene was identified as the cores gene of the study .
The heatmap shows the expression of the COL11A1 gene in normal breast cancer tissues and in different cancer tissues. We found that this gene was significantly highly expressed in breast cancer tissues, while it was almost absence in normal tissues . To further verify this result, we performed an expression difference analysis using tumor samples from TCGA and normal samples derived from GTEx, and obtained consistent results . Furthermore, both GEO datasets, GSE42568 and GSE 109169, also confirmed that the expression of COL11A1 was higher in the tumor compared to normal tissues . Furthermore, we also verified that the expression of this gene at the cell line and protein level, and the results suggested that COL11A1 was highly expressed in HCC38, HCC1395, MDAMB157, HCC1954, and ZR751 cell lines and had lower expression in T47D, MCF7, HCC1428, CAMA1, and BT483 cell lines . The results of the protein expression analysis suggested that COL11A1 had higher expression in tumor samples compared to normal tissues . Finally, we performed a survival analysis of this gene and found that high expression of COL11A1 was associated with a poor prognosis in patients, either in terms of OS, disease-specific survival, or progress-free interval .
The TME is closely associated with tumor progression and metastasis; thus we explored the correlation between this gene and the TME of breast cancer in the current study. The results suggested that COL11A1 expression was significantly positively correlated with the stromal score (r=0.49, p<0.001) and the ESTIMATE score (r=0.29, p<0.001) in the TME . Furthermore, the results of immune cell infiltration analysis also revealed that the expression of COL11A1 was negatively correlated with the level of B cells, CD4 and CD8 T cells and positively correlated with CAFs . Further analysis of the correlation between this gene and marker genes of B cells, CD4, and CD8 T cells, revealed that COL11A1 was significantly negatively correlated with a marker of B cells, positively correlated with a marker of CD4 T cells, and negatively correlated with a marker of T cells (r=-0.156, r=0.113 and r=-0.160, respectively; p<0.001) . The results of the immune infiltration and survival analysis suggested that in patients with low expression of COL11A1, the degree of B cell infiltration was negatively correlated with patient prognosis. This finding was also applied to the high expression group of COL11A1 expression . However, the level of CD4 and CD8 T cell infiltration was negatively correlated with patient prognosis. These findings further support the association of COL11A1 with tumor immune infiltration and patient prognosis .
CAFs are present in the tumor stroma and contribute to tumor invasion by promoting the epithelial-mesenchymal transition and participating in tumor angiogenesis. We first analyzed the distribution of COL11A1 in different clusters of cells at the single-cell level, and the results suggested that the data were clustered into four clusters, namely myofibroblasts, Mono/macro, epithelial, and fibroblasts . COL11A1 was distributed most significantly in myofibroblasts and fibroblasts. Additionally, the expression of this gene was significantly higher in fibroblasts than in epithelial cells . Further analysis revealed the relationship between COL11A1 and the classical CAF marker gene, and we found that the gene was significantly positively correlated with the CAF marker gene (FAP, PDPN, THY1, ACTA2, COL1A1, PDGFRA, and PDGFRB; p<0.001) . The results of the survival analysis suggested that the deeper the immune infiltration, the worse the prognosis of patients with low expression of COL11A1 and the opposite results of the prognosis analysis of patients with high expression of COL11A1 .
The results of COL11A1 and immune checkpoints suggest that COL11A1 expression was positively correlated with immune checkpoints (CD276, TIGIT, and ENTPD1; p<0.001) . Further analysis of the results of two immunotherapy data sets revealed that before the analysis of two data sets, 67.39% of the genes overlapped and two data sets had a batch effect, after removal of the effect, new data did not show any batch effect . When we analyzed the differences between COL11A1 expression in the response groups and the absence of response groups, we found that COL11A1 showed high expression in the response group, although this was not significant, while compared to the absence of response samples (p=0.33) .
m6A methylation, as a modification of RNA molecules, has become a hot research topic in the life sciences field in recent years. Studies have shown that genes related to m6A methylation promote tumor progression and may mediate tumor immune tolerance. Therefore, we further analyzed the correlation between this gene and m6A methylation-related genes. The results suggest that this gene is associated with several m6A methylation-related genes .
A total of 43 related immune regulation genes of COL11A1 were identified from breast cancer samples. To determine the prognostic value of these genes, we constructed predictive models using univariate and multivariate Cox regression. The results of the univariate analysis suggested that a total of 19 immune regulation genes were associated with the prognosis , and the multivariate results demonstrate that 5 immune regulation genes were independent risk factors associated with patient outcomes . Finally, we constructed a 5 gene signature prognostic model based on the above results. We divided patients into the high-risk and low-risk groups based on the median value of the model scores. We found that patients in the high-risk group had a worse prognosis than those in the low-risk group (HR=2.47, 95% CI 1.83-3.33, p<0.001). The area under the model’s ROC curve was 0.658, suggesting that the model had a high predictive value . The internal validation of the model also demonstrated that high-risk patients had a poorer outcome than low-risk patients (HR=2.30, 95% CI 1.47-3.60, p<0.001). The area under the ROC was 0.651 in the validation group, indicating that the model was robust .
We first evaluated the clinical value of the signature-based on univariate and multivariate COX regression analysis. Univariate results suggested that age, stage, estrogen receptor (ER) positivity, progesterone (PR) positivity, and the risk score could be risk factors that influences the prognosis of patients. The results of the multivariate analysis showed that age, stage, ER status, and risk score were independent prognostic factors for patients . Finally, we visualized the analysis results to construct a nomogram model to predict the OS of the patients . It is noteworthy that when we built the final version of model, PR status was included, although it showed no significance in the multivariate analysis results. Nonetheless, this variable is very important in clinical decision-making. The C-index of this model is 0.776, and the model correction prediction curve had a small bias from the ideal curve, suggesting that the model could predict the OS of patients at 5 and 10 years with more accuracy .
Our Immunohistochemistry (IHC) results demonstrate that COL11A1 could express high in breast cancer tissues while compared with normal tissues in clinical samples .
Breast cancer, as one of the three most common tumors in the world, and seriously threatens women’s health worldwide. Although early-stage breast cancer can be successfully treated with surgery, chemotherapy, or combined therapy, more than 30% of patients diagnosed in early-stage will eventually progress and develop an advanced stage . Advanced breast cancer is incurable with traditional treatments and has a long-term survival rate of less than 5% . These data reveal the urgent need for innovative treatments to reduce relapse and metastasis of breast cancer. The successful application of immunotherapy in a variety of solid tumors and the results of immunological checkpoint antagonists targeting programmed cell death 1 (PD-1) and programmed death ligand-1 (PD-L1) in metastatic breast cancer have raised interest in the area of immune-based strategies for breast cancer . Therefore, it is of great significance to explore new immune-related biomarkers to predict treatment response and as a predictor of prognosis . We found that COL11A1 was highly expressed at both the transcriptome and protein levels in breast cancer tissues and could serve as a marker of a poor prognosis. Furthermore, we also found that COL11A1 was positively correlated with risk factors in the breast cancer TME. Finally, based on the above results, we identified a COL11A1-associated immunological signature as a predictor in breast cancer. COL11A1 is located on chromosome 1p21.1, which encodes one of the three alpha chains of type XI collagen and plays a role in the development of skeletal development and fibrillogenesis. But the expression and biological function of COL11A1 in cancers are still controversial and tumor-specific. Some studies have reported that COL11A1 is highly expressed and correlated with poor prognosis in breast cancers, while its expression is low and serves as a good prognostic indicator in some hematological tumors . Therefore, more precise mechanisms of COL11A1 should be explored in breast cancers. The composition of immune cells and stromal cells in the TME has been known to play an important role in metastasis, immune escape, and therapeutic resistance in cancers . Pearce et al. showed that the COL11A1-related signature was positively correlated with Treg and TH2 in ovarian cancer specimens, demonstrating a poorer prognosis. In a recent study, the high expression of COL11A1 was positively correlated with CD4+T and CD8+T cells, tumor-associated macrophages (TAM), neutrophils and dendritic cells in colon adenocarcinoma, while the function of these immune cells in colon adenocarcinoma TME has not been identified . These results suggest that, as one of the components of ECM, COL11A1 may be affected by variable TME in different cancer contexts. The early stage of mammary tumorigenesis is characterized as a stage of acute inflammation, which could activate innate immune cells, such as neutrophils, dendritic cells (DC), and tumor-specific T cells, to eliminate breast cancer cells. While transformed cells escape elimination and a chronic inflammation-like TME is established, which is mainly composed of suppressive immune cells, CAFs, and endothelial cells, leading to immune evasion of advanced breast cancers . As shown in our study, COL11A1 had a negative correlation with immune cells (B cells, CD4+ T cells, CD8+ T cells, natural killer cells, and macrophages) in the TME, but showed a positive correlation with CAFs and endothelial cells, which was consistent with the results indicating that overexpression of COL11A1 was only observed in CAF-enriched areas of different cancers and was associated with poorer prognosis . All these results implied that COL11A1 could be involved in the tumor immune evasion process and could act as a poor immune-related biomarker in breast cancers. As representative of ICIs, the Food and Drug Administration has approved treatment with anti-PD-1 and anti-PD-L1 monoclonal antibodies for metastatic triple negative breast cancer (TNbreast cancer) immunotherapy. Faced with these options for breast cancer, it is critical to select potential breast cancer patients populations that could benefit from ICI treatment . In addition to PD-L1 expression and tumor mutational burden (TMB), some studies have proposed that TME characteristics could also be used as an indicator to predict response to ICI treatment . Furthermore, a retrospective study identified Meflin as serve as a predictive marker of CAF, which could increase the sensitivity to ICI treatment . In our study, patients with higher expression of COL11A1 also showed a better response to ICI treatment, which indicates that COL11A1 has candidate potential to predict response to ICIs treatment, in addition to being an immune-related biomarker for prognosis. However, previous studies have shown the opposite predictive role of COL11A1 in response to PD1 checkpoint immunotherapy, reconfirming the heterogeneity and complexity of TME in cancers . Thus, it is of great importance to take advantage of multi-omics methods and computational algorithms to interpret the function of genes at the single cell level in different contexts. With the development of next-generation sequencing technology and computational intelligence techniques, more and more disease markers are being identified, and drugs developed based on these targets will greatly improve patients’ clinical benefits in the future . In the present study, we used machine learning to identify a new breast cancer marker and further confirm its potential to become a new target. However, relying on a single gene to predict the patient’s prognosis presents drawbacks because, due to the heterogeneity of the disease, disease development can be associated with the abnormal expression of multiple genes. Therefore, we screened five genes related to the immune pathway associated with COL11A1 in breast cancer and constructed a signature to assess the prognosis of the patient based on these genes. This signature also independently predicted patient prognosis compared with patient clinical variables, implying that our multigene signature had high predictive efficacy. In addition, we constructed the NOMOGRAM model, a visual predictive tool based on signature and clinical variables, which, compared with single-gene models and models containing only clinical information. This tool, compared with single-gene models and models containing only clinical information, showed richer predictive properties and greatly enhanced the clinical value of the model.
We identified COL11A1 as a potential therapeutic target in breast cancer through machine learning, and the high expression of this gene was generally associated with a poor prognosis. Additionally, this gene was also closely associated with breast cancer tumor immune infiltrating cells and could be involved in the tumor immune infiltration process. However, there are some limitations in our study. First, additional machine learning algorithms need to be attempted and elaborately combined to obtain accurate training results. Second, single-cell sequencing data from breast cancer should be included to further clarify the relationship between COL11A1 and the TME in breast cancer. Third, additional clinical RCTs are needed to confirm the predictivity of COL11A1 in the immunotherapy response of breast cancers. Fourth, the possible role of COL11A1 involved in the TME of breast cancers should be further explored through basic research studies.
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/ .
The studies involving human participants were reviewed and approved by Ethics Committee of Hainan Medical University. The patients/participants provided their written informed consent to participate in this study.
WS, ZC (2 nd author), HL, WP, and CL: Conceptualization, data curation, formal analysis, roles/writing—original draft, writing—review and editing. GW,RF, ZC (8 th author), and GC: Roles/writing—original draft. PF, WP and CL: Funding acquisition, methodology, project administration, resources, supervision. All authors contributed to the article and approved the submitted version.
We thank CM for providing experiment validation of COL11A1 with IHC technique.
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.
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
|
The “Make Surgical Pathology Easy” project: learning Pathology through tailored digital infographics - the case for renovation of an old teaching method | 6acd3e36-c845-484e-af63-d88dd535affd | 8488984 | Pathology[mh] | Pathology is the medical specialty which concerns the examination of tissue specimens for diagnostic purposes. Actually, it is an interdisciplinary field where a plethora of different information, from clinical to molecular, converge and are gathered to reach a diagnosis. In order to begin its speculations, Pathology must fulfil the task of bringing into view its specific object of study, which is not directly accessible to senses: the microscopic world of injured tissues. From gross inspection of surgical specimens to microscopic examination of tissue slides, the diagnostic process is led through the observation and interpretation of images, making Pathology a constitutionally visual discipline. While the observational process is carried on, several background information act as a filter. Namely, the pathologist’s interpretation process is always directed and influenced by additional data – such as medical history of the patient, clinical presentations, imaging features of the lesion and other details which are not an intrinsic part of the Pathology domain. These notions represent the specific context in which the pathologic diagnosis is set. Consequently, diagnosing images in Pathology is based on a complex interplay between perception, cognition and ability to integrate informations. In Pathology, at baseline, education is imparted in both visual and experiential forms . During their residency programs, pathologists are trained to scan/search through histological slides, rapidly process visual stimuli and extract key elements from the scene they are looking at. With time and experience, this sequence becomes automated but the development of a visual expertise and procedural knowledge in Pathology is an intricate process that requires an innate observational aptitude, appropriate training and adequate feedbacks. By referring to the different learning styles in Pathology, “learn by doing” (converging style) prevails among residents; as pathologists mature, “learn by observing” with the application of prior experience to problem-solving (diverging style) and conceptual thinking (assimilating style) become dominant . Pathologists use cognitive processes, such as perception, attention, memory and search, to collect data from the case, including macroscopic and microscopic findings and clinical or radiologic information. Based on these data, hypotheses are elaborated, and then checked against the data . Histopathological images can be considered as multilayered texts with their own internal narrative. The pathologist’s eye must be tuned to detect at a glance the essential features in a histological image. Such features are defined as ‘diagnostic criteria’ and describe salient characteristics of cells, cellular arrangement in specific structures and the relationship of cellular structures with the surrounding tissue environment. A specific set of criteria constitutes the ‘pattern’ of a given disease, a kind of ideal representation of that disease which each pathologist bears in mind as a template. Pattern recognition at microscopic level is a multistep process including visual examination of the slides, the assessment of morphologic criteria and the association of a specific combination of criteria with a pathologic entity. As Pathologists become more familiar with how normal and abnormal, benign and malignant, and how one disease versus another look like (as well as possible variations), they become more efficient at extracting relevant image features required to render accurate diagnoses . Therefore, in pathologists with well-structured competences and a robust visual expertise, the process of pattern recognition happens in a fluid, spontaneous and almost immediate manner, through a well-run feature extraction analysis. However, reaching such a level of integrated consciousness requires struggle and intense learning effort . Pathologists have to process a large quantity of theoretical information, they need to learn about different types of subjects and, most importantly, their knowledge needs to be retained. However, in many domains and sub-domains of Pathology, the list of common and rare diagnostic entities may be rather large so that becoming acquainted with each of them is virtually impossible, resulting in a partial learning. Furthermore, medical education, including education in morphological sciences, is mainly imparted by promoting an analytical learning style, a deliberate rule-based method of solving a problem or reaching at a diagnosis . Sequential learning is excellent for conceptual topics but inefficient for topics, involving pattern recognition , which require integration of different levels of information and a strong anatomo-clinical correlation. Gaining confidence with pattern recognition occurs as a function of time spent on learning and the number of images viewed during a certain period of time . However, other factors deeply influence the efficacy of learning through images. In fact, a gap exists between the ideal representation of a certain disease, its estimated pattern and images pathologists are faced with in their routine practice. Typically, not all diagnostic criteria are simultaneously present in individual cases of a certain disease. Some diagnostic clues may simply be lacking due to the intrinsic biological diversity of single specimens, others may be unrecognizable due to supervening artifacts. In addition, the timing of the disease at the moment the biopsy is taken and other factors can dramatically modify the way the disease looks in a certain specimen, from case to case. In this context, the use of visual methods is of invaluable help in facilitating the process of learning and fill the gap between blurred chaotic reality and coherent scientific models . Visual methods can be broadly defined as all kinds of visual materials which can be employed in education and research to build and disseminate knowledge. They are extremely heterogeneous and include drawings, photography, videos and artworks of various types. Drawing, alongside with photography, has traditionally been the most commonly used visual method in Pathology education. Producing visual representations is constitutive to sciences and a fundamental part of scientific communication and knowledge production. Illustrations are a powerful tool to circumvent the limits of human comprehension when it is challenged with the theoretical and the invisible. They match the need to give shape to what cannot be routinely experienced and constitute perhaps the most effective method scientists rely on to divulge their findings. Illustrations can condense and express in a more accessible form concepts that result from complex elaboration of data. The birth of visually-oriented medical disciplines was strongly linked to the production of illustrations. Historically, before the invention and diffusion of photographic techniques, manual drawings were the only visual tools to document the observations made under experimental conditions, at macroscopic and microscopic level . Drawing firstly emerged as a powerful form of recording new findings and describing the disease processes, superior to any verbal description. For this reason, the great anatomists and pathologists of the past closely collaborated with artists to include detailed illustrations in their treatises. For example, Andreas Vesalius’ revolutionary work De Humani Corporis Fabrica libri septem , which is considered a cornerstone in modern Anatomy, was enriched with more than 150 exquisite woodcut illustrations. Often, however, anatomists and pathologists were artists themselves, like the father of neurosciences and laureate Nobel Santiago Ramón y Cajal. He is best remembered for shedding light on the structure and function of nervous system and the development of his ‘neuron doctrine’ . Using special staining techniques to selectively visualize the neurons, he reconstructed in single sketches the results of multiple microscopic observations made in different focal planes. With this procedure, he could render a visual representation of the neural networks as they could not be directly visualized, but only deduced from a series of independent observations . Among the contemporaries of Ramón y Cajal, Camillo Golgi was the first to describe and draw the “internal reticular apparatus”, while Enrico Sertoli summarized his observations on testes’ morphology in the drawing of a pattern, becoming the forerunner of all subsequent schematic representation of testis in the histology manuals , . The most prolific of medical illustrators was certainly Frank Netter, who created more than 4000 pictures for CIBA pharmaceutical company alone . The extraordinary ability to capture the essential lines of a given anatomical region or physiological/pathological process is central in Netter’s artistic sketches. He created dramatic images, which are superb medical illustrations and powerful teaching aids, and still widely used by both students and experienced practitioners , . A modern example of illustration in Pathology is Donald Gleason’s graphic presentation of prostate cancer progression patterns. Dr. Gleason managed to illustrate in a single image the entire biological spectrum, presented as a continuum, of prostatic acinar adenocarcinoma. A simple hand-drawing could recapitulate the results of years of research and hundreds of observations in a neat, swift and yet very informative manner, still unsurpassed after more than 60 years . These emblematic cases highlight how illustrations are not just a means to passively reproduce forms as we perceive them but, most importantly, they are instruments that go beyond and reveal what is concealed in perceptions. Producing visual representation is an experimental act and integral part of the investigative process so that images can be considered substantial cognitive instruments. Nowadays, despite the ubiquitous and large-scale availability of microscopic photography, which can better accomplish faithful reproductions of an object of interest, drawing and illustration still play an irreplaceable role in Pathology research and education , giving emphasis to details which may not be evident in a photograph. Obviously, with the passing of time, graphic techniques have evolved. Manual drawing has been partially replaced by digital drawing. On the other side, drawing has gradually lost the urge of being representational and strengthened its function of integrating notions and presenting them in a comprehensive and synthetic form. This tendency mirrors the need to simplify and control the extreme and constantly growing complexity of modern medical knowledge. Simple illustrations have given way to more sophisticated forms of visual representations called infographics, including graphs and texts and incorporating a high number of information from different interconnected domains. The term ‘infographics’ is a blend of the two words ‘information’ and ‘graphics’, so can be described as ‘information visualizations’. In that sense, they can be seen as a ‘knowledge assemblage’ including text, numbers, graphs, charts, drawings and so on . Infographics are an innovative and engaging method of visually communicating information in a colorful and concise manner , translating complex data, be it research or medical, into smaller, more relatable amounts for the individual to understand . Information are more likely to be retained if learnt from an infographic than from text or verbal communication alone . Infographics create a new platform where visual materials can be gleaned in a faster way and better fixed in memory, being particularly suitable for medical education . A summary of the many roles of infographics in medical education is provided in . In the field of Pathology, infographics are increasingly shared by free open-access websites and social media platforms such as Twitter and Facebook. Our goal is to describe the conceptual model of ideation, processing and translation to a big audience of the Infographics collected in the educational project called ‘Make Surgical Pathology Easy’. The project was started in 2018 by Dr. Abhijit Das, albeit the ideation and conceptualization of the designs had been initiated long back during his post-graduation period at ‘All India Institute of Medical Science’ (AIIMS) in New Delhi, India. Over the course of 3 years (from January 2018 to May 2021) a total of 123 infographics were developed and shared for free on Facebook and Twitter. The infographics published by Dr. Abhijit Das are hand-drawn schematic images (artistic pencil-sketches) placed on PowerPoint slides and then labelled accordingly. They cover various aspects of Surgical and General Pathology and follow a holistic approach in the understanding of the diagnostic process, providing different levels of information. These include the classification and pathophysiology of neoplastic and non-neoplastic diseases, the definition of the histopathological and immunohistochemical features and the explanation of common molecular pathways. Thus constructed, infographics shed light on diagnostic criteria, differentials and predictive/prognostic markers for a particular disease or a particular group of diseases. The number of infographics for different subspecialties topics is shown in . Infographic design is a simple step-wise process starting with the selection of a particular topic of discussion (e.g. posterior circulating stroke, ). A comprehensive and thorough understanding of the concerned topic is of utmost importance and can be obtained from any valid source. Recent updates and developments related to the topic should be included and supplemented by formulating notes (e.g. percentage of ischemic stroke, etiology, normal anatomical variant of blood supply, most common sites for vertebral dissection etc., ). In the next step, all robust information is selected, critically reviewed and sorted to build a coherent storytelling. At this point, considering proper utilization of space, a tentative sketch of the storytelling is created on a plan paper using colored pencils and markers. The final sketch is then placed on a PowerPoint slide for labelling. This process is often carried out by a professional team of graphic artists for proper alignment and organization of text in relation to the schemes. The PowerPoint slide is then saved as image (.tiff or .jpeg format, ) It can be difficult to find all salient morphological features of a particular disease in histopathology slides. In this respect, an infographic can be of great value by summing them up in a single image. For example, a single histopathology slide might not contain all typical morphological features of classic chordoma (i.e. vacuolated cell nest, cords, myxoid background and physaliphorous cells etc.), which can be instead presented simultaneously in one infographic . Moreover, the same infographic can contain the morphological features of the other types of chordoma (e.g. chondroid background in chondroid chordoma, spindle cell component and the numerous mitoses in dedifferentiated chordoma, ). Lastly, supplemented information (e.g., basic immunohistochemistry panel for chordoma, ) can be added to make the whole infographic a complete resume of current diagnostic knowledge about a topic. Infographics related to a particular group of diseases are intended to provide maximum information in an exhaustive and simplified manner. For example, a single infographic representing ‘pemphigus and its differentials’ can easily illustrate the notable features of subcorneal vesiculobullous diseases (i.e. pemphigus foliaceus, bullous impetigo, subcorneal pustular dermatosis & staphylococcal scalded skin syndrome), as well as bullous diseases with suprabasal cleft formation (i.e. pemphigus vulgaris, Darier disease, Hailey Hailey disease, Grover disease & paraneoplastic pemphigus etc.) This type of infographic will certainly help to summarize one particular topic in an effective and faster way. The ‘Make Surgical Pathology Easy’ infographics can be accessed directly from Facebook page ( https://www.facebook.com/AbhijitSurgPath ) or Twitter channel ( https://twitter.com/AbhijitSurgPath ). Up to May 23, 2021, the Facebook and Twitter pages have respectively 4136 and 2213 followers. The infographics hosted on the Facebook page have generated a total of 3967 likes while those available on Twitter channel have produced a total of 1675 likes. Users from several countries approached the infographics, with India being the most involved. Infographic-based learning process can be a significant learning tool for the several domains and sub-domains of Pathology. Teaching in morphological sciences aims to develop three types of knowledge: declarative (conceptual), procedural (strategic) and conditional , . Declarative (conceptual) knowledge comprises the sum of information and data about a topic; it is basically “what you know about something” , . Within conceptual knowledge, infographics are intended to present large amounts of data in an easy and compact manner . Infographics include technical terminology, specific details and elements which are the basis of conceptual knowledge being an incredibly efficient tool for it. The procedural (strategic) knowledge is the one exercised in the performance of a task, it is basically ‘how you know to do something’ . Within procedural knowledge, infographics increase the ability to interpret histopathological/immunohistochemical patterns and to correlate them with clinical problems. By outlining workflow interconnections in simple and concise graphical pathways, infographics can be of great help in the mental process of generating, formulating and achieving a correct diagnosis. They are oriented towards diagnostic problem solving skills, recapitulating subject-specific techniques and criteria to use appropriate procedures . The conditional knowledge regards the application of conceptual or strategic knowledge, it is basically ‘when, where and why to use something’. In this context, infographics explore the interrelationships among the basic elements within a larger domain or sub-domain of Pathology, allowing the awareness of the rationale behind a diagnostic strategy . In recent years, in the field of Pathology, several profit and not-profit online resources have been developed to supplement traditional textbook materials and in the last months the global pandemic of SARS-CoV-2 has further strengthened this trend. There is a wealth of web-based data for pathologists and different platforms providing them. Histopathology-related posts on Facebook, Twitter or Instagram contain images of histopathology slides with or without annotations, quizzes, detailed explanations and links to publications. Whole slide images can be visualized and navigated at any magnification . Free of charge video-based tutorials and live stream lectures can be found on YouTube, Twitter or dedicated “open-access” online platforms like pathCast , . Pathology education performed by traditional methods is not equal worldwide, and the quality and extent of education may vary from country to country , creating knowledge disparity between advantaged and less-advantaged countries. In this context, online resources and social media tools provide an alternative, free and less time-consuming method for pathologists to share their knowledge and experience with the next generation . Pathologists can share cases and respond to individual questions from other pathologists around the world, proposing a new model of teaching which is similar to that used in collegial discussions of cases at the multiheaded microscope, but on an amplified magnitude . Accordingly, infographics entirely reproduce the traditional way of teaching in Pathology, by facilitating the understanding of complex subjects and translation of what has been learned into the diagnostic practice. Moreover, they can be flexibly adapted to meet the needs of both trainees and practicing pathologists. In this sense, infographics emerge as integral part of continuing education in Pathology, mitigating the time-honored barriers of conventional learning systems. In conclusion, digitalization has given new life to an old teaching method, allowing its adaptation to social media platforms and its diffusion on a global scale. |
Magnolol Ameliorates Behavioral Impairments and Neuropathology in a Transgenic Mouse Model of Alzheimer's Disease | e7ad2bc0-76f7-4f65-b676-c1a1dcea8af2 | 7354664 | Pathology[mh] | Alzheimer's disease (AD), the most common type of dementia in the elderly population, is characterized by progressive memory loss and cognitive decline . AD affected about 26.6 million people worldwide in 2006, and the number is expected to rise to 131.5 million by 2050 . The healthcare and economic burden due to AD to the society is enormous . Pathologically, the extracellular beta-amyloid plaque (A β ) deposits composed of A β peptides and the intracellular neurofibrillary tangles as a result of tau protein accumulation in the brain are the two major hallmarks of AD. Accumulating evidence suggests that neuroinflammation and the loss of neuronal synapses are observed in the early stage of AD and are associated with cognitive decline . The underlying mechanisms of the onset and progression of AD are still unclear. Mutation of the amyloid precursor protein (APP) induces the abnormal production of A β peptides by β - and γ -secretase and is believed to play a critical role in the early onset of familial AD . In the APP processing, beta-site APP-cleaving enzyme 1 (BACE-1), the major β -secretase enzyme, is directly involved in the cleavage of APP. Neurons of BACE-1−/− mice do not generate A β , suggesting that BACE-1 is the neuronal β -secretase . Phosphorylated APP at the Thr668 site (p-APP (T668)) can facilitate the APP processing at the β -cleavage site. Presenilin 1 (PS1) and anterior pharynx-defective 1 (APH-1) are the two major components of γ -secretase . Neprilysin (NEP) and insulin-degrading enzyme (IDE) are the two major A β -degrading enzymes that promote A β degradation . The TgCRND8 transgenic mouse, a well-known aggressive APP mouse model of AD, expresses the human APP gene with double KM670/671NL+V717F Swedish and Indiana familial AD mutations. TgCRND8 mice overexpress mutant human APP at a level approximately 5-fold higher than endogenous murine APP. TgCRND8 mice express extracellular A β 40 and A β 42 at one month of age, and the soluble A β 42/A β 40 ratios are found to be elevated at two months of age . At 3 months of age, TgCRND8 mice develop the phenotypes closely resembling human AD such as A β deposits, astrocytic activation, microglial activation, neuritic dystrophy, inflammation, and learning and memory deficits . Therefore, the TgCRND8 mouse model is a valuable tool for investigating new therapeutic agents for AD and elucidating the underlying anti-AD molecular mechanisms. Currently available drugs for AD can only ameliorate symptoms of AD but fall short of reversing or even slowing down the disease progression. Therefore, therapeutic strategies for thwarting AD progression clearly remain an unmet medical need. Magnolol (MN) (the chemical structure is shown in ) is the essential natural neolignan and the main active ingredient responsible for the therapeutic properties of the bark of Magnolia officinalis , a herb widely used in Chinese medicine to treat inflammatory diseases with low toxicity . The content of MN in the bark of M. officinalis is about 1.0-1.25% . MN has been shown to exert various pharmacological activities such as anti-inflammation , antioxidation , and neuroprotection . MN has recently been reported to possess anti-AD effects in experimental models of AD . MN significantly alleviates the A β -induced neurotoxicity via suppressing the intracellular calcium elevation, the reactive oxygen species production, the caspase-3 activity, and inflammation, as well as promoting the phagocytosis and degradation of A β . In addition, MN has been shown to prevent the cognitive deficits induced by scopolamine in mice via inhibition of the acetylcholinesterase activity and oxidative stress . Moreover, MN has been demonstrated to ameliorate learning and memory impairments by preserving cholinergic function in the forebrain of the SAMP8 mice . Importantly, MN could cross the blood-brain barrier (BBB) and remain relatively stable in the brain after oral administration . Moreover, no troublesome side effects have been reported so far in humans after ingestion of MN . All these observations indicate that MN may be the active principle responsible for the anti-AD activity of M. officinalis . However, the molecular mechanisms underlying the anti-AD actions of MN hitherto remain unexplored. In the present study, we aimed to investigate whether MN could ameliorate the learning and memory impairments in TgCRND8 transgenic mice and illustrate its mechanisms of action. 2.1. Chemicals and Reagents Magnolol (MN, purity ≥ 98%) was purchased from Hong Kong University of Science and Technology Research and Development Corporation Limited. Its identity was confirmed by comparing its 1 H NMR and 13 C NMR spectra with that published in the literature . A β 42 peptide was purchased from GL Biochem Ltd. (Shanghai, China). Donepezil hydrochloride was purchased from Sigma-Aldrich (St. Louis, MO, USA). All other chemicals and reagents used in this study were of analytical grade. 2.2. Animals TgCRND8 mice harbor the genetic background of (C57BL/6J) × (C3H/HeJ × C57BL/6J). Male TgCRND8 mice and female wild-type C57BL/6J were used to breed a colony of experimental animals. Nontransgenic littermates that did not express human APP transgene were identified as wild-type mice and used as a control group. The mice were bred in the Run Run Shaw Science Building, The Chinese University of Hong Kong, and routinely maintained on a 12 h light/dark cycle under controlled humidity (50 ± 10%) and temperature (22 ± 2°C) with access to food and water ad libitum . 2.3. Genotyping of TgCRND8 Mice In order to genotype the APP transgene, DNA was extracted from the ear tissues of all mice. The APP transgene was determined by a transgene-specific PCR reaction using the following primers: forwards: TGTCCAAGATGCAGCAGAACGGCTAC, reverse: GGCCGCGGAGAAATGAAGAAACGCCA. Briefly, a visible amount of ear tissue was digested in the non-SDS tissue digesting buffer (500 mM KCl, 100 mM Tris-HCl, 0.45% NP-40 (Igepal™ CA-630), 0.1 mg/ml gelatin, and 0.45% Tween 20) with proteinase K (Cat. V900887, Sigma) at 55°C overnight. After heated at 98°C for 10 min to inactivate proteinase K, the supernatant was collected after centrifugation and then undergone a PCR reaction with a TaKaRa Taq™ package (Cat. R001A, TaKaRa) and the primers. The reactions were run at 95°C for 5 min, followed by 45 cycles at 95°C for 30 s, 54°C for 40 s, 72°C for 80 s, and 72°C for 10 min. The PCR samples were separated using 1% agarose gel, then observed under UV light for 30 s. The mice with APP transgene were identified as transgenic mice, while those without APP transgene as wild-type mice. 2.4. Grouping of TgCRND8 Mice and Drug Treatment Three-month-old male mice were randomly assigned to five groups of 10 animal each: (a) wild type (WT), (b) TgCRND8 (Tg)+vehicle, (c) Tg+MN (20 mg/kg), (d) Tg+MN (40 mg/kg), and (e) Tg+donepezil (5 mg/kg). The dosages of MN used in this study were chosen based on our pilot study (data not shown). Donepezil was chosen as a positive control based on previous publications . MN was suspended in 0.5% sodium carboxymethyl cellulose (CMC-Na) while donepezil was dissolved in normal saline. Mice were administered orally with MN and donepezil by gavage once daily for 4 months, whereas the mice in the WT group and the Tg+vehicle group received an equal volume of 0.5% CMC-Na. After drug treatment, the spatial learning and memory functions were assessed using the open-field test, the novel objective recognition (NOR) test, and the radial arm maze (RAM) test. shows the experimental design and schedule. 2.5. Open-Field Test (OFT) The locomotor activity of the mice was determined by the OFT. Briefly, the mice were placed in an open field (40 × 60 × 50 cm) with a brown floor divided into 12 equal squares and a frontal glass wall . The mice were subjected to two identical sessions on two consecutive days, with the first session for training and the second one for testing. Each session lasted for 6 min. The number of line crossing with four paws and the number of rearing (number of times the animals stood on their hind legs) were recorded to investigate the exploratory behavior and locomotor activity of mice, respectively, by two observers who were blinded to the grouping information. To avoid perturbation to the animals due to urine and feces, the apparatus was cleaned with 10% ethanol solution and a piece of dry cloth between two tests. 2.6. Novel Object Recognition Test (NORT) The NORT were conducted in an open-field arena (30 × 30 × 30 cm) constructed with polyvinyl chloride, plywood, and acrylic as previously described . The tasks included a training session and a recognition session for two consecutive days. On day 1, the mice were allowed to explore two identical objects (5 × 5 × 5 cm, blue plastic cubes) for 5 min in the training session. On day 2, one of the objects was replaced with a new shape and color (5 × 5 × 7 cm, a white plastic square pyramid), and the mice were acclimatized to the area for 5 min in the recognition session. The fields were decontaminated with a 10% ethanol solution between the tests. The animals were allowed to explore the test area by touching or sniffing the objects with their forepaws and/or noses at a distance of less than 2 cm. The total exploration time was the amount of time devoted to locating the two objects. The time of each mouse spent exploring the objects was recorded by two investigators who were blinded to the experimental design. The cognitive function was determined using a recognition index, which was the exploration time involved with either of the two objects (training session) or the novel object (recognition session) divided by the total exploration time in exploring both objects. 2.7. Radial Arm Maze Test (RAMT) The spatial learning and memory functions of mice were determined using the RAMT. The apparatus for the RAMT was obtained from Xinruan Information Technology Co. Ltd. (Xinruan, Shanghai, China) with a video tracking software of SuperMaze V2.0. The apparatus comprises eight radial arms (10 cm high, 5 cm wide, and 35 cm long) numbered from 1 to 8 and a central platform (22 cm in diameter). The RAMT was conducted as described in our previous studies . During the behavioral test period, to stimulate hunger, the mice were maintained on a restricted diet with only water being available ad libitum. The body weight of mice was kept at 85-90% of free-feeding level. The RAMT lasted for 8 consecutive days: 2 days for habituation trials, 5 days for training trials, and 1 day for task test. At the habituation trial, 3 or 4 mice were simultaneously put in the central platform of RAM, and all arms were baited with several food pellets about 10 mg each. After two days of habituation trial, the mice were trained with 1 trial daily for 5 consecutive days. At the training trial, only 4 constant arms were baited with one food pellet about 10 mg, which was placed in the nontransparent food cup to prevent visual detection, and only one mouse was placed in the central platform. The mice were trained to run to the end of the baited arms and consume all the food pellets within 10 min. The mice were subjected to working and reference memory task tests on the eighth day. In the task test, the same four arms were baited with one food pellet about 10 mg, and an arm entry was counted when all four limbs of the mice were within an arm. After all of the food pellets had been consumed or 10 min had passed, the task test was completed. In the task test, two observers who were blinded to the grouping information recorded the following data: (1) the number of working memory errors (WMEs), which meant the reentries into an already visited baited arm during the period of task test; (2) the number of reference memory errors (RMEs), which meant the entries into the nonbaited arms during the period of task test; and (3) the number of total entries to complete the task test. 2.8. Brain Tissue Collection Twenty-four hours after the NORT, the brain tissues of the mice were harvested quickly under deep anesthesia. After washing with ice-cold normal saline, the brains were bisected in the midsagittal plane. One hemisphere was used for enzyme-linked immunosorbent assay (ELISA) kit analysis, while the opposite hemisphere was used for western blotting analysis. These samples were immediately stored at -80°C until used. On the other hand, for immunofluorescence analysis, 4 mice in each group were deeply anesthetized and perfused intracardially with normal saline, followed by 4% paraformaldehyde (PFA) solution in 0.1 M phosphate buffer (PB, pH 7.4). The brain tissues were collected, postfixed in 4% PFA for 24 h, and then dehydrated in 30% sucrose at 4°C for 2-3 days. Transverse sections of the brain tissue (20 μ m) were obtained using a cryostat (Leica CM1850, Leica Microsystems GmbH, Wetzlar, Germany), then transferred to gelatin-coated slides at 20°C before further processing. 2.9. Measurement of the Levels of A β 40 and A β 42 in the Brains of TgCRND8 Mice The levels of A β 40 and A β 42 in the brain tissues of TgCRND8 mice were measured using commercial mouse A β 40 (Cat. KMB3481, Invitrogen, USA) and mouse A β 42 (Cat. KMB3441, Invitrogen, USA) ELISA kits according to the manufacturer's protocols. Briefly, the brain hemisphere was homogenized in 8 × volume of homogenization buffer (5 M guanidine-HCl diluted in 50 mM Tris (pH 8.0)) with 1 × protease inhibitor cocktail containing AEBSF (Cat. P2714, Sigma, USA). The homogenate was mixed with an orbital shaker for 4 h at room temperature. After centrifugation at 16,000 × g at 4°C for 20 min, the supernatants were collected and diluted with standard diluent buffer to an appropriate concentration. The diluted supernatants were added into the wells that precoated with mAb to NH2 terminus of A β and then incubated for 2 h at room temperature to bind the antigen. The mouse A β detection antibody solution and anti-rabbit IgG HRP solution were sequentially added after washing three times with 1 × wash buffer. The reaction was terminated by adding the stop solution. The absorbance was determined at 450 nm within 10 min using a FLUOstar OPTIMA microplate reader (BMG Labtech, Offenburg, Germany). The levels of A β 40 and A β 42 in the brain tissues were calculated using the standard curves and expressed as pg/mg protein. 2.10. Determination of Cytokines in the Brain Tissues of TgCRND8 Mice The brain tissues (100 mg) of mice were homogenized in lysis buffer (150 mM NaCl, 50 mM Tris-HCl (pH 7.4), 1 mM Na 3 VO 4 , 0.5% NP40, 1 mM NaF, and 1 mM DTT). After incubating for 15 min on ice, the homogenates were centrifuged at 10,000 × g at 4°C for 30 min. The protein levels of interleukin-6 (IL-6, Cat. No: ab100712), interleukin-1beta (IL-1 β , Cat. No: ab100704), tumor necrosis factor alpha (TNF- α , Cat. No: LS-F5192), and interleukin-10 (IL-10, Cat. No: ab100697) in the supernatants were measured using commercially available sandwich ELISA kits (LifeSpan BioSciences, Seattle, USA, and Abcam, Cambridge, UK, respectively) per the manufacturer's protocols. The levels of IL-6, IL-1 β , TNF- α , and IL-10 were expressed as pg/mg protein. 2.11. Immunofluorescence Assay Brain sections containing the cortex and hippocampus were blocked with 5% bovine serum albumin (BSA) at room temperature for 1 h, then incubated with the following primary antibodies against mouse antiglial fibrillary acidic protein (GFAP) (1 : 1000, Cat. C106874, Sigma, USA) and rabbit anti-ionized calcium-binding adaptor molecule 1 (Iba-1) (1 : 1000, Cat. 019-19741, Wako, Japan), which were used to detect astrocyte and microglia, respectively. After washing for 5 min × 3 times with 1 × PBS (0.01 M, pH 7.4), the sections were then incubated with secondary antibodies against Alexa Fluor™ 647 streptavidin (Cat. S21374, Invitrogen, USA) or Alexa Fluor 488 Goat anti-Rabbit IgG (H+L) (Cat. R37116, Invitrogen, USA) for 2 h at room temperature in the dark. After washing with 1 × PBS for 3 times, sections were mounted on gelatin-coated glass slides and cover-slipped with antifade mounting medium (Dako, Glostrup, Denmark) for microscopic examination. BSA rather than the primary antibody was used as a negative control. Fluorescent images were obtained using a Zeiss fluorescence microscope (Zeiss, Gottingen, Germany) equipped with an ORCA-Flash 4.0 v2 digital CMOS camera (Hamamatsu Photonics, Iwata City, Japan). Images were determined using an unbiased computer-assisted Image J software (NIH, Bethesda, MD, USA). The mean areas of the GFAP-positive astrocytes and Iba-1-positive microglia in each restricted area were quantified based on the method described in a previous study . 2.12. Western Blotting Analysis Cytoplasmic and nuclear proteins were isolated from 100 mg of frozen brain tissues using the nuclear and cytosolic protein extraction kit (Chemicon, Temecula, CA, USA). Protein concentrations were determined using the BCA Protein Assay Kit. The protein samples were separated by sodium dodecyl sulfate-polyacrylamide (SDS-PAGE) at 80 V for 2 h. The separated proteins were transferred to polyvinylidene difluoride membrane (PVD) membranes using a transblotting apparatus (Bio-Rad Laboratories, USA) at 15 V for 30 min. The membranes were blocked with 5% ( w / v ) nonfat milk in TBS-T (Tris-buffer saline containing 0.1% Tween-20) at room temperature for 2 h. Subsequently, the membranes were incubated with an appropriate amount of primary antibodies against insulin-degrading enzyme (IDE) and presenilin-1 (PS-1) (Santa Cruz), β -site APP-cleaving enzyme-1 (BACE-1) and anterior pharynx-defective-1 (APH-1) (Sigma), neprilysin (NEP) (R&D Systems), p-APP (Thr688), postsynaptic density protein 93 (PSD93), postsynaptic density protein 95 (PSD95), synapsin 1 (SYN 1), synaptotagmin-1 (SYT 1), synaptophysin (SYN), nuclear factor kappa-B (NF- κ B) p65, p-NF- κ B p65, phosphor-glycogen synthase kinase-3 β (p-GSK-3 β ), GSK-3 β , phosphor-Akt (p-Akt), Akt (Cell Signaling Technology Inc., Beverly, MA, USA), and β -actin (Santa Cruz Biotechnology Inc., USA), respectively, at 4°C overnight. Next, the membranes were washed with TBS-T three times and probed with horseradish peroxidase-conjugated secondary antibodies for 1 h at room temperature. Finally, the membranes were washed with TBS-T three times before the protein bands were determined by the ECL western blotting detection reagents (Amersham Biosciences, Buckinghamshire, UK). The intensity of each band was quantified using Image J software. 2.13. Statistical Analysis All data were presented as the mean ± SEM. Multiple group comparisons were analyzed to detect intergroup differences using one-way ANOVA followed by post hoc Bonferroni's test. GraphPad Prism software (Version 5, GraphPad Software, Inc., CA, USA) was used for the statistical analysis. A difference was considered statistically significant when p < 0.05. Magnolol (MN, purity ≥ 98%) was purchased from Hong Kong University of Science and Technology Research and Development Corporation Limited. Its identity was confirmed by comparing its 1 H NMR and 13 C NMR spectra with that published in the literature . A β 42 peptide was purchased from GL Biochem Ltd. (Shanghai, China). Donepezil hydrochloride was purchased from Sigma-Aldrich (St. Louis, MO, USA). All other chemicals and reagents used in this study were of analytical grade. TgCRND8 mice harbor the genetic background of (C57BL/6J) × (C3H/HeJ × C57BL/6J). Male TgCRND8 mice and female wild-type C57BL/6J were used to breed a colony of experimental animals. Nontransgenic littermates that did not express human APP transgene were identified as wild-type mice and used as a control group. The mice were bred in the Run Run Shaw Science Building, The Chinese University of Hong Kong, and routinely maintained on a 12 h light/dark cycle under controlled humidity (50 ± 10%) and temperature (22 ± 2°C) with access to food and water ad libitum . In order to genotype the APP transgene, DNA was extracted from the ear tissues of all mice. The APP transgene was determined by a transgene-specific PCR reaction using the following primers: forwards: TGTCCAAGATGCAGCAGAACGGCTAC, reverse: GGCCGCGGAGAAATGAAGAAACGCCA. Briefly, a visible amount of ear tissue was digested in the non-SDS tissue digesting buffer (500 mM KCl, 100 mM Tris-HCl, 0.45% NP-40 (Igepal™ CA-630), 0.1 mg/ml gelatin, and 0.45% Tween 20) with proteinase K (Cat. V900887, Sigma) at 55°C overnight. After heated at 98°C for 10 min to inactivate proteinase K, the supernatant was collected after centrifugation and then undergone a PCR reaction with a TaKaRa Taq™ package (Cat. R001A, TaKaRa) and the primers. The reactions were run at 95°C for 5 min, followed by 45 cycles at 95°C for 30 s, 54°C for 40 s, 72°C for 80 s, and 72°C for 10 min. The PCR samples were separated using 1% agarose gel, then observed under UV light for 30 s. The mice with APP transgene were identified as transgenic mice, while those without APP transgene as wild-type mice. Three-month-old male mice were randomly assigned to five groups of 10 animal each: (a) wild type (WT), (b) TgCRND8 (Tg)+vehicle, (c) Tg+MN (20 mg/kg), (d) Tg+MN (40 mg/kg), and (e) Tg+donepezil (5 mg/kg). The dosages of MN used in this study were chosen based on our pilot study (data not shown). Donepezil was chosen as a positive control based on previous publications . MN was suspended in 0.5% sodium carboxymethyl cellulose (CMC-Na) while donepezil was dissolved in normal saline. Mice were administered orally with MN and donepezil by gavage once daily for 4 months, whereas the mice in the WT group and the Tg+vehicle group received an equal volume of 0.5% CMC-Na. After drug treatment, the spatial learning and memory functions were assessed using the open-field test, the novel objective recognition (NOR) test, and the radial arm maze (RAM) test. shows the experimental design and schedule. The locomotor activity of the mice was determined by the OFT. Briefly, the mice were placed in an open field (40 × 60 × 50 cm) with a brown floor divided into 12 equal squares and a frontal glass wall . The mice were subjected to two identical sessions on two consecutive days, with the first session for training and the second one for testing. Each session lasted for 6 min. The number of line crossing with four paws and the number of rearing (number of times the animals stood on their hind legs) were recorded to investigate the exploratory behavior and locomotor activity of mice, respectively, by two observers who were blinded to the grouping information. To avoid perturbation to the animals due to urine and feces, the apparatus was cleaned with 10% ethanol solution and a piece of dry cloth between two tests. The NORT were conducted in an open-field arena (30 × 30 × 30 cm) constructed with polyvinyl chloride, plywood, and acrylic as previously described . The tasks included a training session and a recognition session for two consecutive days. On day 1, the mice were allowed to explore two identical objects (5 × 5 × 5 cm, blue plastic cubes) for 5 min in the training session. On day 2, one of the objects was replaced with a new shape and color (5 × 5 × 7 cm, a white plastic square pyramid), and the mice were acclimatized to the area for 5 min in the recognition session. The fields were decontaminated with a 10% ethanol solution between the tests. The animals were allowed to explore the test area by touching or sniffing the objects with their forepaws and/or noses at a distance of less than 2 cm. The total exploration time was the amount of time devoted to locating the two objects. The time of each mouse spent exploring the objects was recorded by two investigators who were blinded to the experimental design. The cognitive function was determined using a recognition index, which was the exploration time involved with either of the two objects (training session) or the novel object (recognition session) divided by the total exploration time in exploring both objects. The spatial learning and memory functions of mice were determined using the RAMT. The apparatus for the RAMT was obtained from Xinruan Information Technology Co. Ltd. (Xinruan, Shanghai, China) with a video tracking software of SuperMaze V2.0. The apparatus comprises eight radial arms (10 cm high, 5 cm wide, and 35 cm long) numbered from 1 to 8 and a central platform (22 cm in diameter). The RAMT was conducted as described in our previous studies . During the behavioral test period, to stimulate hunger, the mice were maintained on a restricted diet with only water being available ad libitum. The body weight of mice was kept at 85-90% of free-feeding level. The RAMT lasted for 8 consecutive days: 2 days for habituation trials, 5 days for training trials, and 1 day for task test. At the habituation trial, 3 or 4 mice were simultaneously put in the central platform of RAM, and all arms were baited with several food pellets about 10 mg each. After two days of habituation trial, the mice were trained with 1 trial daily for 5 consecutive days. At the training trial, only 4 constant arms were baited with one food pellet about 10 mg, which was placed in the nontransparent food cup to prevent visual detection, and only one mouse was placed in the central platform. The mice were trained to run to the end of the baited arms and consume all the food pellets within 10 min. The mice were subjected to working and reference memory task tests on the eighth day. In the task test, the same four arms were baited with one food pellet about 10 mg, and an arm entry was counted when all four limbs of the mice were within an arm. After all of the food pellets had been consumed or 10 min had passed, the task test was completed. In the task test, two observers who were blinded to the grouping information recorded the following data: (1) the number of working memory errors (WMEs), which meant the reentries into an already visited baited arm during the period of task test; (2) the number of reference memory errors (RMEs), which meant the entries into the nonbaited arms during the period of task test; and (3) the number of total entries to complete the task test. Twenty-four hours after the NORT, the brain tissues of the mice were harvested quickly under deep anesthesia. After washing with ice-cold normal saline, the brains were bisected in the midsagittal plane. One hemisphere was used for enzyme-linked immunosorbent assay (ELISA) kit analysis, while the opposite hemisphere was used for western blotting analysis. These samples were immediately stored at -80°C until used. On the other hand, for immunofluorescence analysis, 4 mice in each group were deeply anesthetized and perfused intracardially with normal saline, followed by 4% paraformaldehyde (PFA) solution in 0.1 M phosphate buffer (PB, pH 7.4). The brain tissues were collected, postfixed in 4% PFA for 24 h, and then dehydrated in 30% sucrose at 4°C for 2-3 days. Transverse sections of the brain tissue (20 μ m) were obtained using a cryostat (Leica CM1850, Leica Microsystems GmbH, Wetzlar, Germany), then transferred to gelatin-coated slides at 20°C before further processing. β 40 and A β 42 in the Brains of TgCRND8 Mice The levels of A β 40 and A β 42 in the brain tissues of TgCRND8 mice were measured using commercial mouse A β 40 (Cat. KMB3481, Invitrogen, USA) and mouse A β 42 (Cat. KMB3441, Invitrogen, USA) ELISA kits according to the manufacturer's protocols. Briefly, the brain hemisphere was homogenized in 8 × volume of homogenization buffer (5 M guanidine-HCl diluted in 50 mM Tris (pH 8.0)) with 1 × protease inhibitor cocktail containing AEBSF (Cat. P2714, Sigma, USA). The homogenate was mixed with an orbital shaker for 4 h at room temperature. After centrifugation at 16,000 × g at 4°C for 20 min, the supernatants were collected and diluted with standard diluent buffer to an appropriate concentration. The diluted supernatants were added into the wells that precoated with mAb to NH2 terminus of A β and then incubated for 2 h at room temperature to bind the antigen. The mouse A β detection antibody solution and anti-rabbit IgG HRP solution were sequentially added after washing three times with 1 × wash buffer. The reaction was terminated by adding the stop solution. The absorbance was determined at 450 nm within 10 min using a FLUOstar OPTIMA microplate reader (BMG Labtech, Offenburg, Germany). The levels of A β 40 and A β 42 in the brain tissues were calculated using the standard curves and expressed as pg/mg protein. The brain tissues (100 mg) of mice were homogenized in lysis buffer (150 mM NaCl, 50 mM Tris-HCl (pH 7.4), 1 mM Na 3 VO 4 , 0.5% NP40, 1 mM NaF, and 1 mM DTT). After incubating for 15 min on ice, the homogenates were centrifuged at 10,000 × g at 4°C for 30 min. The protein levels of interleukin-6 (IL-6, Cat. No: ab100712), interleukin-1beta (IL-1 β , Cat. No: ab100704), tumor necrosis factor alpha (TNF- α , Cat. No: LS-F5192), and interleukin-10 (IL-10, Cat. No: ab100697) in the supernatants were measured using commercially available sandwich ELISA kits (LifeSpan BioSciences, Seattle, USA, and Abcam, Cambridge, UK, respectively) per the manufacturer's protocols. The levels of IL-6, IL-1 β , TNF- α , and IL-10 were expressed as pg/mg protein. Brain sections containing the cortex and hippocampus were blocked with 5% bovine serum albumin (BSA) at room temperature for 1 h, then incubated with the following primary antibodies against mouse antiglial fibrillary acidic protein (GFAP) (1 : 1000, Cat. C106874, Sigma, USA) and rabbit anti-ionized calcium-binding adaptor molecule 1 (Iba-1) (1 : 1000, Cat. 019-19741, Wako, Japan), which were used to detect astrocyte and microglia, respectively. After washing for 5 min × 3 times with 1 × PBS (0.01 M, pH 7.4), the sections were then incubated with secondary antibodies against Alexa Fluor™ 647 streptavidin (Cat. S21374, Invitrogen, USA) or Alexa Fluor 488 Goat anti-Rabbit IgG (H+L) (Cat. R37116, Invitrogen, USA) for 2 h at room temperature in the dark. After washing with 1 × PBS for 3 times, sections were mounted on gelatin-coated glass slides and cover-slipped with antifade mounting medium (Dako, Glostrup, Denmark) for microscopic examination. BSA rather than the primary antibody was used as a negative control. Fluorescent images were obtained using a Zeiss fluorescence microscope (Zeiss, Gottingen, Germany) equipped with an ORCA-Flash 4.0 v2 digital CMOS camera (Hamamatsu Photonics, Iwata City, Japan). Images were determined using an unbiased computer-assisted Image J software (NIH, Bethesda, MD, USA). The mean areas of the GFAP-positive astrocytes and Iba-1-positive microglia in each restricted area were quantified based on the method described in a previous study . Cytoplasmic and nuclear proteins were isolated from 100 mg of frozen brain tissues using the nuclear and cytosolic protein extraction kit (Chemicon, Temecula, CA, USA). Protein concentrations were determined using the BCA Protein Assay Kit. The protein samples were separated by sodium dodecyl sulfate-polyacrylamide (SDS-PAGE) at 80 V for 2 h. The separated proteins were transferred to polyvinylidene difluoride membrane (PVD) membranes using a transblotting apparatus (Bio-Rad Laboratories, USA) at 15 V for 30 min. The membranes were blocked with 5% ( w / v ) nonfat milk in TBS-T (Tris-buffer saline containing 0.1% Tween-20) at room temperature for 2 h. Subsequently, the membranes were incubated with an appropriate amount of primary antibodies against insulin-degrading enzyme (IDE) and presenilin-1 (PS-1) (Santa Cruz), β -site APP-cleaving enzyme-1 (BACE-1) and anterior pharynx-defective-1 (APH-1) (Sigma), neprilysin (NEP) (R&D Systems), p-APP (Thr688), postsynaptic density protein 93 (PSD93), postsynaptic density protein 95 (PSD95), synapsin 1 (SYN 1), synaptotagmin-1 (SYT 1), synaptophysin (SYN), nuclear factor kappa-B (NF- κ B) p65, p-NF- κ B p65, phosphor-glycogen synthase kinase-3 β (p-GSK-3 β ), GSK-3 β , phosphor-Akt (p-Akt), Akt (Cell Signaling Technology Inc., Beverly, MA, USA), and β -actin (Santa Cruz Biotechnology Inc., USA), respectively, at 4°C overnight. Next, the membranes were washed with TBS-T three times and probed with horseradish peroxidase-conjugated secondary antibodies for 1 h at room temperature. Finally, the membranes were washed with TBS-T three times before the protein bands were determined by the ECL western blotting detection reagents (Amersham Biosciences, Buckinghamshire, UK). The intensity of each band was quantified using Image J software. All data were presented as the mean ± SEM. Multiple group comparisons were analyzed to detect intergroup differences using one-way ANOVA followed by post hoc Bonferroni's test. GraphPad Prism software (Version 5, GraphPad Software, Inc., CA, USA) was used for the statistical analysis. A difference was considered statistically significant when p < 0.05. 3.1. Effects of MN on the Locomotor Activity of TgCRND8 Mice Ten mice for each group were used to perform the open-field test. The number of rearings significantly decreased in TgCRND8 mice in the open-field test, while the number of crossings did not change markedly, as compared with the WT control group. Treatment with MN (20 and 40 mg/kg) or donepezil (5 mg/kg) did not significantly affect the number of rearings and crossings of TgCRND8 mice in the open-field test, as compared with the Tg vehicle control group (Figures and ). 3.2. Effects of MN on the Recognition Impairment of TgCRND8 Mice The NORT is designed to measure the spontaneous preference of rodents to explore an unfamiliar object rather than a familiar object and is usually used to explore the recognition potential of the mice. Ten mice for each group were used to conduct the NORT. As shown in Figures and , the recognition index was not significantly changed ( F (4, 45) = 0.009, p > 0.05) in the training session of the NORT . The recognition index was significantly different among various groups ( F (4, 45) = 3.720, p < 0.05) in the recognition session of the NORT . The recognition index in the TgCRND8 mice was significantly lower ( p < 0.01) as compared with the WT control group. The mice treated with MN (20 and 40 mg/kg) exhibited a higher recognition index ( p < 0.05 for both) in the recognition session of the NORT as compared with the Tg vehicle control group. These results suggest that MN could ameliorate the recognition impairments in TgCRND8 mice. 3.3. Effects of MN on Learning and Memory Impairments in TgCRND8 Mice The effects of MN on spatial learning and memory deficits were determined using the RAMT. Ten mice for each group were used to perform the RAMT. As shown in , the number of total entries was significantly elevated ( F (4, 45) = 13.410, p < 0.001) in the Tg vehicle group when compared to the WT group. The mice treated with MN (20 and 40 mg/kg) markedly decreased the number of total entries ( p < 0.05 and p < 0.001, respectively) when compared to the Tg vehicle control group. Donepezil (5 mg/kg) treatment also markedly attenuated the number of total entries ( p < 0.01) when compared to the Tg vehicle control. The effects of MN on the numbers of WMEs and RMEs were shown in Figures and , respectively. The results demonstrated that the numbers of WMEs ( F (4, 45) = 22.870, p < 0.001) and RMEs ( F (4, 45) = 9.041, p < 0.001) were effectively elevated in the Tg vehicle group when compared to the WT group. The mice treated with MN (20 and 40 mg/kg) significantly attenuated the numbers of WMEs ( p < 0.05 and p < 0.001, respectively) and RMEs ( p < 0.05 and p < 0.001, respectively) when compared to the Tg vehicle control group. Donepezil (5 mg/kg) treatment also obviously reduced the number of WMEs ( p < 0.01) and RMEs ( p < 0.05) when compared to the Tg vehicle control. 3.4. Effects of MN on the A β Deposition and APP Processing in the Brain Tissues of TgCRND8 Mice The brain tissues from six mice for each group were used to determine the levels of A β 40 and A β 42 . As shown in Figures and , the protein levels of A β 40 ( F (4, 25) = 101.200, p < 0.001) and A β 42 ( F (4, 25) = 72.800, p < 0.001) were significantly higher in the brain of TgCRND8 mice when compared to the WT control. Treatment with MN (20 and 40 mg/kg) markedly reduced the protein levels of A β 40 ( p < 0.001 for both) and A β 42 ( p < 0.05 and p < 0.001, respectively) in the brain tissues of TgCRND8 mice, as compared to the WT control. Similarly, treatment with donepezil (5 mg/kg) also significantly decreased the levels of A β 42 in the brain of TgCRND8 mice ( p < 0.05). The brain tissues from three mice for each group were used to measure the protein expressions of APP processing. As shown in Figures and , the protein levels of BACE-1 ( F (4, 10) = 28.180, p < 0.001), p-APP (T668) ( F (4, 10) = 20.590, p < 0.001), APH-1 ( F (4, 10) = 25.940, p < 0.001), and PS-1 ( F (4, 10) = 24.920, p < 0.001) in the brain tissues of TgCRND8 mice were significantly increased, while the protein expressions of NEP ( F (4, 10) = 25.630, p < 0.001) and IDE ( F (4, 10) = 28.690, p < 0.001) were markedly decreased, when compared with the WT group. Treatment with MN (20 mg/kg) markedly suppressed the protein expressions of BACE-1 ( p < 0.01), p-APP ( p < 0.001), APH-1 ( p < 0.05), and PS-1 ( p < 0.001) but did not alter the protein expressions of NEP and IDE in the brain tissues of TgCRND8 mice when compared with the Tg vehicle group. On the other hand, treatment with MN (40 mg/kg) significantly inhibited the protein expressions of BACE-1 ( p < 0.001), p-APP ( p < 0.001), APH-1 ( p < 0.001), and PS-1 ( p < 0.001) but significantly enhanced the protein expressions of NEP ( p < 0.001) and IDE ( p < 0.001) in the brain tissues of TgCRND8 mice, as compared to the WT control group. Similarly, treatment with donepezil (5 mg/kg) significantly inhibited the protein expressions of BACE-1 ( p < 0.05), p-APP (T668) ( p < 0.05), and PS-1 ( p < 0.001) but markedly elevated the protein expressions of NEP ( p < 0.05) and IDE ( p < 0.05) in the brain tissues of TgCRND8 mice as compared to the WT control group. 3.5. Effects of MN on the Astrocytes and Microglia in the Hippocampus and Cerebral Cortex of TgCRND8 Mice The brain tissues from four mice for each group were used to evaluate the microglia and astrocytes in the hippocampus and cerebral cortex of the mice. As shown in Figures and , a significant increase of microglial density was observed in the hippocampus ( F (4, 15) = 26.580, p < 0.001) and cerebral cortex ( F (4, 15) = 15.460, p < 0.001) of TgCRND8 mice when compared with the WT group. Treatment with MN (20 and 40 mg/kg) significantly decreased the microglial density in the hippocampus ( p < 0.01 and p < 0.001, respectively) and cerebral cortex ( p < 0.05 and p < 0.001, respectively) of TgCRND8 mice when compared with the Tg vehicle control. Similarly, treatment with donepezil (5 mg/kg) also effectively ameliorated the microglial density in the hippocampus ( p < 0.05) and cerebral cortex ( p < 0.05) of TgCRND8 mice. Figures and revealed a marked increase of the astrocyte density in the hippocampus ( F (4, 15) = 19.770, p < 0.001) and cerebral cortex ( F (4, 15) = 19.240, p < 0.001) of TgCRND8 mice, when compared with the WT group. Treatment with MN (20 and 40 mg/kg) significantly attenuated the astrocyte density in the hippocampus ( p < 0.01 and p < 0.001, respectively) and cerebral cortex ( p < 0.05 and p < 0.001, respectively) of TgCRND8 mice when compared with the Tg vehicle control. Similarly, treatment with donepezil (5 mg/kg) also significantly reduced the astrocyte density in the hippocampus ( p < 0.05) and cerebral cortex ( p < 0.05) of TgCRND8 mice. 3.6. Effects of MN on the Levels of IL-6, IL-1 β , TNF- α , and IL-10 in the Brains of TgCRND8 Mice The brain tissues from six mice for each group were used to assay the protein levels of an inflammatory mediator. As shown in , the protein levels of TNF- α ( F (4, 25) = 12.950, p < 0.001) (a), IL-1 β ( F (4, 25) = 19.790, p < 0.001) (b), and IL-6 ( F (4, 25) = 65.380, p < 0.001) (c) were significantly increased, while the release of IL-10 ( F (4, 25) = 36.560, p < 0.001) (d) was markedly decreased in the brains of TgCRND8 mice, when compared to the WT control group. Treatment with MN (20 and 40 mg/kg) markedly decreased the protein productions of TNF- α ( p < 0.001 for both), IL-1 β ( p < 0.001 for both), and IL-6 ( p < 0.001 for both), while significantly elevated the release of IL-10 ( p < 0.001 for both) in the brains of TgCRND8 mice, as compared to the Tg vehicle control group. Similarly, treatment with donepezil (5 mg/kg) also significantly reversed these changes of cytokines in TgCRND8 mice. 3.7. Effects of MN on the Synaptic Dysfunction in the Brains of TgCRND8 Mice Synaptic dysfunction is an early event in AD patients and correlates well with cognitive impairment during the course of the disease . To evaluate the effects of MN (20 and 40 mg/kg) on synaptic pathology, we examined the levels of PSD93, PSD95, SYN 1, SYT, and SYN in the brains of TgCRND8 mice ( n = 3) . The results demonstrated that the protein levels of PSD93 ( F (4, 10) = 15.890, p < 0.001), PSD95 ( F (4, 10) = 35.750, p < 0.001), SYN 1 ( F (4, 10) = 16.940, p < 0.001), SYT 1 ( F (4, 10) = 27.240, p < 0.001), and SYN ( F (4, 10) = 25.300, p < 0.001) were significantly decreased in the brains of TgCRND8 mice as compared with the WT control group. Treatment with MN (20 mg/kg) markedly elevated the protein expressions of SYN 1 ( p < 0.01) in the brains of TgCRND8 mice. Interestingly, MN (40 mg/kg) significantly enhanced the protein expressions of PSD93 ( p < 0.001), PSD95 ( p < 0.001), SYN 1 ( p < 0.001), SYT 1 ( p < 0.001), and SYN ( p < 0.001) in the brains of TgCRND8 mice as compared with the Tg vehicle control. Similarly, donepezil (5 mg/kg) obviously increased the protein expressions of PSD93 ( p < 0.01), PSD95 ( p < 0.001), SYN 1 ( p < 0.001), SYT 1 ( p < 0.001), and SYN ( p < 0.001) in the brains of TgCRND8 mice, as compared with the Tg vehicle control. 3.8. Effects of MN on the NF- κ B and PI3K/Akt/GSK-3 β Pathways in the Brains of TgCRND8 Mice The NF- κ B pathway plays a central role in regulating inflammatory responses. The brain tissues of three mice for each group were used to determine the protein expressions of the NF- κ B and PI3K/Akt/GSK-3 β pathways. As shown in , the ratio of p-NF- κ B p65/NF- κ B p65 ( F (4, 10) = 41.350, p < 0.001) was significantly accentuated in the brains of TgCRND8 mice when compared with the WT control. However, treatment with MN (20 and 40 mg/kg) markedly reduced the ratio of the protein expression of p-NF- κ B p65/NF- κ B p65 ( p < 0.001 for both) in the brains of TgCRND8 mice, when compared with the Tg vehicle control. Treatment with donepezil (5 mg/kg) also conspicuously suppressed the ratio of the protein expression of p-NF- κ B p65/NF- κ B p65 ( p < 0.001) in the brains of TgCRND8 mice when compared with the Tg vehicle control. The results shown in revealed that there was a significant decrease in the protein expressions of p-GSK-3 β (Ser9) ( F (4, 10) = 35.090, p < 0.001) and p-Akt (Ser473) ( F (4, 10) = 7.257, p < 0.01) in the brain tissues of TgCRND8 mice, when compared with the WT group. Treatment with MN (20 and 40 mg/kg) markedly enhanced the ratios of the protein expressions of p-GSK-3 β (Ser9)/GSK-3 β ( p < 0.05 and p < 0.001, respectively) and p-Akt (Ser473)/Akt ( p < 0.05 for both) in the brain tissues of TgCRND8 mice when compared with the Tg vehicle group. Similarly, treatment with donepezil (5 mg/kg) also obviously increased the ratios of the protein expressions of p-GSK-3 β (Ser9)/GSK-3 β ( p < 0.01) and p-Akt (Ser473)/Akt ( p < 0.05) in the brain tissues of TgCRND8 mice when compared with the Tg vehicle group. Ten mice for each group were used to perform the open-field test. The number of rearings significantly decreased in TgCRND8 mice in the open-field test, while the number of crossings did not change markedly, as compared with the WT control group. Treatment with MN (20 and 40 mg/kg) or donepezil (5 mg/kg) did not significantly affect the number of rearings and crossings of TgCRND8 mice in the open-field test, as compared with the Tg vehicle control group (Figures and ). The NORT is designed to measure the spontaneous preference of rodents to explore an unfamiliar object rather than a familiar object and is usually used to explore the recognition potential of the mice. Ten mice for each group were used to conduct the NORT. As shown in Figures and , the recognition index was not significantly changed ( F (4, 45) = 0.009, p > 0.05) in the training session of the NORT . The recognition index was significantly different among various groups ( F (4, 45) = 3.720, p < 0.05) in the recognition session of the NORT . The recognition index in the TgCRND8 mice was significantly lower ( p < 0.01) as compared with the WT control group. The mice treated with MN (20 and 40 mg/kg) exhibited a higher recognition index ( p < 0.05 for both) in the recognition session of the NORT as compared with the Tg vehicle control group. These results suggest that MN could ameliorate the recognition impairments in TgCRND8 mice. The effects of MN on spatial learning and memory deficits were determined using the RAMT. Ten mice for each group were used to perform the RAMT. As shown in , the number of total entries was significantly elevated ( F (4, 45) = 13.410, p < 0.001) in the Tg vehicle group when compared to the WT group. The mice treated with MN (20 and 40 mg/kg) markedly decreased the number of total entries ( p < 0.05 and p < 0.001, respectively) when compared to the Tg vehicle control group. Donepezil (5 mg/kg) treatment also markedly attenuated the number of total entries ( p < 0.01) when compared to the Tg vehicle control. The effects of MN on the numbers of WMEs and RMEs were shown in Figures and , respectively. The results demonstrated that the numbers of WMEs ( F (4, 45) = 22.870, p < 0.001) and RMEs ( F (4, 45) = 9.041, p < 0.001) were effectively elevated in the Tg vehicle group when compared to the WT group. The mice treated with MN (20 and 40 mg/kg) significantly attenuated the numbers of WMEs ( p < 0.05 and p < 0.001, respectively) and RMEs ( p < 0.05 and p < 0.001, respectively) when compared to the Tg vehicle control group. Donepezil (5 mg/kg) treatment also obviously reduced the number of WMEs ( p < 0.01) and RMEs ( p < 0.05) when compared to the Tg vehicle control. β Deposition and APP Processing in the Brain Tissues of TgCRND8 Mice The brain tissues from six mice for each group were used to determine the levels of A β 40 and A β 42 . As shown in Figures and , the protein levels of A β 40 ( F (4, 25) = 101.200, p < 0.001) and A β 42 ( F (4, 25) = 72.800, p < 0.001) were significantly higher in the brain of TgCRND8 mice when compared to the WT control. Treatment with MN (20 and 40 mg/kg) markedly reduced the protein levels of A β 40 ( p < 0.001 for both) and A β 42 ( p < 0.05 and p < 0.001, respectively) in the brain tissues of TgCRND8 mice, as compared to the WT control. Similarly, treatment with donepezil (5 mg/kg) also significantly decreased the levels of A β 42 in the brain of TgCRND8 mice ( p < 0.05). The brain tissues from three mice for each group were used to measure the protein expressions of APP processing. As shown in Figures and , the protein levels of BACE-1 ( F (4, 10) = 28.180, p < 0.001), p-APP (T668) ( F (4, 10) = 20.590, p < 0.001), APH-1 ( F (4, 10) = 25.940, p < 0.001), and PS-1 ( F (4, 10) = 24.920, p < 0.001) in the brain tissues of TgCRND8 mice were significantly increased, while the protein expressions of NEP ( F (4, 10) = 25.630, p < 0.001) and IDE ( F (4, 10) = 28.690, p < 0.001) were markedly decreased, when compared with the WT group. Treatment with MN (20 mg/kg) markedly suppressed the protein expressions of BACE-1 ( p < 0.01), p-APP ( p < 0.001), APH-1 ( p < 0.05), and PS-1 ( p < 0.001) but did not alter the protein expressions of NEP and IDE in the brain tissues of TgCRND8 mice when compared with the Tg vehicle group. On the other hand, treatment with MN (40 mg/kg) significantly inhibited the protein expressions of BACE-1 ( p < 0.001), p-APP ( p < 0.001), APH-1 ( p < 0.001), and PS-1 ( p < 0.001) but significantly enhanced the protein expressions of NEP ( p < 0.001) and IDE ( p < 0.001) in the brain tissues of TgCRND8 mice, as compared to the WT control group. Similarly, treatment with donepezil (5 mg/kg) significantly inhibited the protein expressions of BACE-1 ( p < 0.05), p-APP (T668) ( p < 0.05), and PS-1 ( p < 0.001) but markedly elevated the protein expressions of NEP ( p < 0.05) and IDE ( p < 0.05) in the brain tissues of TgCRND8 mice as compared to the WT control group. The brain tissues from four mice for each group were used to evaluate the microglia and astrocytes in the hippocampus and cerebral cortex of the mice. As shown in Figures and , a significant increase of microglial density was observed in the hippocampus ( F (4, 15) = 26.580, p < 0.001) and cerebral cortex ( F (4, 15) = 15.460, p < 0.001) of TgCRND8 mice when compared with the WT group. Treatment with MN (20 and 40 mg/kg) significantly decreased the microglial density in the hippocampus ( p < 0.01 and p < 0.001, respectively) and cerebral cortex ( p < 0.05 and p < 0.001, respectively) of TgCRND8 mice when compared with the Tg vehicle control. Similarly, treatment with donepezil (5 mg/kg) also effectively ameliorated the microglial density in the hippocampus ( p < 0.05) and cerebral cortex ( p < 0.05) of TgCRND8 mice. Figures and revealed a marked increase of the astrocyte density in the hippocampus ( F (4, 15) = 19.770, p < 0.001) and cerebral cortex ( F (4, 15) = 19.240, p < 0.001) of TgCRND8 mice, when compared with the WT group. Treatment with MN (20 and 40 mg/kg) significantly attenuated the astrocyte density in the hippocampus ( p < 0.01 and p < 0.001, respectively) and cerebral cortex ( p < 0.05 and p < 0.001, respectively) of TgCRND8 mice when compared with the Tg vehicle control. Similarly, treatment with donepezil (5 mg/kg) also significantly reduced the astrocyte density in the hippocampus ( p < 0.05) and cerebral cortex ( p < 0.05) of TgCRND8 mice. β , TNF- α , and IL-10 in the Brains of TgCRND8 Mice The brain tissues from six mice for each group were used to assay the protein levels of an inflammatory mediator. As shown in , the protein levels of TNF- α ( F (4, 25) = 12.950, p < 0.001) (a), IL-1 β ( F (4, 25) = 19.790, p < 0.001) (b), and IL-6 ( F (4, 25) = 65.380, p < 0.001) (c) were significantly increased, while the release of IL-10 ( F (4, 25) = 36.560, p < 0.001) (d) was markedly decreased in the brains of TgCRND8 mice, when compared to the WT control group. Treatment with MN (20 and 40 mg/kg) markedly decreased the protein productions of TNF- α ( p < 0.001 for both), IL-1 β ( p < 0.001 for both), and IL-6 ( p < 0.001 for both), while significantly elevated the release of IL-10 ( p < 0.001 for both) in the brains of TgCRND8 mice, as compared to the Tg vehicle control group. Similarly, treatment with donepezil (5 mg/kg) also significantly reversed these changes of cytokines in TgCRND8 mice. Synaptic dysfunction is an early event in AD patients and correlates well with cognitive impairment during the course of the disease . To evaluate the effects of MN (20 and 40 mg/kg) on synaptic pathology, we examined the levels of PSD93, PSD95, SYN 1, SYT, and SYN in the brains of TgCRND8 mice ( n = 3) . The results demonstrated that the protein levels of PSD93 ( F (4, 10) = 15.890, p < 0.001), PSD95 ( F (4, 10) = 35.750, p < 0.001), SYN 1 ( F (4, 10) = 16.940, p < 0.001), SYT 1 ( F (4, 10) = 27.240, p < 0.001), and SYN ( F (4, 10) = 25.300, p < 0.001) were significantly decreased in the brains of TgCRND8 mice as compared with the WT control group. Treatment with MN (20 mg/kg) markedly elevated the protein expressions of SYN 1 ( p < 0.01) in the brains of TgCRND8 mice. Interestingly, MN (40 mg/kg) significantly enhanced the protein expressions of PSD93 ( p < 0.001), PSD95 ( p < 0.001), SYN 1 ( p < 0.001), SYT 1 ( p < 0.001), and SYN ( p < 0.001) in the brains of TgCRND8 mice as compared with the Tg vehicle control. Similarly, donepezil (5 mg/kg) obviously increased the protein expressions of PSD93 ( p < 0.01), PSD95 ( p < 0.001), SYN 1 ( p < 0.001), SYT 1 ( p < 0.001), and SYN ( p < 0.001) in the brains of TgCRND8 mice, as compared with the Tg vehicle control. κ B and PI3K/Akt/GSK-3 β Pathways in the Brains of TgCRND8 Mice The NF- κ B pathway plays a central role in regulating inflammatory responses. The brain tissues of three mice for each group were used to determine the protein expressions of the NF- κ B and PI3K/Akt/GSK-3 β pathways. As shown in , the ratio of p-NF- κ B p65/NF- κ B p65 ( F (4, 10) = 41.350, p < 0.001) was significantly accentuated in the brains of TgCRND8 mice when compared with the WT control. However, treatment with MN (20 and 40 mg/kg) markedly reduced the ratio of the protein expression of p-NF- κ B p65/NF- κ B p65 ( p < 0.001 for both) in the brains of TgCRND8 mice, when compared with the Tg vehicle control. Treatment with donepezil (5 mg/kg) also conspicuously suppressed the ratio of the protein expression of p-NF- κ B p65/NF- κ B p65 ( p < 0.001) in the brains of TgCRND8 mice when compared with the Tg vehicle control. The results shown in revealed that there was a significant decrease in the protein expressions of p-GSK-3 β (Ser9) ( F (4, 10) = 35.090, p < 0.001) and p-Akt (Ser473) ( F (4, 10) = 7.257, p < 0.01) in the brain tissues of TgCRND8 mice, when compared with the WT group. Treatment with MN (20 and 40 mg/kg) markedly enhanced the ratios of the protein expressions of p-GSK-3 β (Ser9)/GSK-3 β ( p < 0.05 and p < 0.001, respectively) and p-Akt (Ser473)/Akt ( p < 0.05 for both) in the brain tissues of TgCRND8 mice when compared with the Tg vehicle group. Similarly, treatment with donepezil (5 mg/kg) also obviously increased the ratios of the protein expressions of p-GSK-3 β (Ser9)/GSK-3 β ( p < 0.01) and p-Akt (Ser473)/Akt ( p < 0.05) in the brain tissues of TgCRND8 mice when compared with the Tg vehicle group. Previous studies have demonstrated that MN administration exerts significant therapeutic action on AD . In the present study, our data for the first time revealed that MN could ameliorate the cognitive deficits in TgCRND8 transgenic mice via inhibition of neuroinflammation and synaptic dysfunction through modulating the PI3K/Akt/GSK-3 β and NF- κ B pathways. Donepezil, an acetylcholinesterase inhibitor approved by the United States Food and Drug Administration for the treatment of mild to moderate AD, was used as a positive control in this study. However, it has undesirable side effects in AD patients such as nausea, vomiting, diarrhea, dizziness, drowsiness, and trouble sleeping . In this study, our results indicated that donepezil could suppress the neuroinflammation and synaptic dysfunction to improve cognitive deficits in TgCRND8 transgenic mice via regulating the PI3K/Akt/GSK-3 β and NF- κ B pathways. Interestingly, MN at a dose of 40 mg/kg exerted more potent effects than donepezil in the improvement of cognitive deficits, inhibition of neuroinflammation, A β deposition, and synaptic dysfunction in TgCRND8 mice. Moreover, it has been shown that the 50% lethal dose (LD 50 ) values of MN and donepezil by oral administration in mice are about 2200 mg/kg and 45.2 mg/kg, respectively, , suggesting that the toxicity of MN is about 50 times less than donepezil, thus is a safer plant-derived compound for AD treatment. A β deposition in the brain is one of the major hallmarks of AD pathogenesis. In the amyloidogenic pathway, APP is primarily processed by β - and γ -secretases to produce A β . BACE-1, a key β -secretase, is essential for initiating A β production . γ -Secretase involves a large proteinase complex comprising at least four major protein ingredients such as APH-1 and PS-1 . Both IDE and NEP are the two major A β -degrading enzymes in the APP processing . Thus, inhibition of the activities of β -secretase or γ -secretase or enhancement of the A β -degrading enzymatic activities may help to reduce the A β deposition. The results of our present study revealed that MN treatment significantly inhibited the protein expressions of BACE-1, p-APP (T668), APH-1, and PS-1, as well as the protein levels of A β 40 and A β 42 , while markedly enhanced the protein expression of NEP and IDE in the brains of TgCRND8 mice. These results strongly suggested that MN mitigated the A β deposition via inhibiting the activities of β - and γ -secretases and enhancing the activities of A β -degrading enzymes in the brains of TgCRND8 mice. A growing body of evidence has revealed that neuroinflammation plays an important role in AD pathology . Neuroinflammation in AD is considered to be primarily driven by microglial cells . Importantly, the oligomers and fibrils of A β are capable of priming microglial cells through elevating the production of inflammatory cytokines (TNF- α , IL-6, and IL-1 β ) and suppressing anti-inflammatory mediators (IL-4 and IL-10), thereby promoting the activation of microglia . Overactivation and gliosis of microglia have been found to play critical roles in AD pathology . The activated microglia can be polarized into divergent M2 or M1 phenotypes and secrete corresponding anti-inflammatory or proinflammatory mediators. Overactivation of the M1 microglia phenotype leads to the secretion of proinflammatory cytokines, subsequently inducing neurotoxicity and neurodegeneration . In contrast, the M2 phenotype microglia enhance the release of anti-inflammatory mediators to promote tissue repair and extracellular matrix reconstruction . An abnormal elevation of proinflammatory cytokines has been reported to be closely associated with memory impairment and affective disorders such as depression and anxiety . Consequently, inhibition of neuroinflammation is crucial for the treatment of AD. MN has been reported to inhibit neuroinflammation in mice exhibiting depressive-like behaviors . To confirm the impacts of MN on M1/M2 microglia polarization, we measured the levels of specific proinflammatory cytokines and anti-inflammatory mediators in the brains of TgCRND8 mice. Our results consistently demonstrated that MN treatment attenuated the release of proinflammatory cytokines such as IL-1 β , IL-6, and TNF- α , while augmented the production of an anti-inflammatory mediator such as IL-10, indicating its anti-inflammatory potential. Interestingly, our present experimental results revealed that MN could attenuate the release of proinflammatory cytokines while accentuating the production of anti-inflammatory mediators, indicating that the anti-inflammatory activities of MN also contribute to its cognition-enhancing functions in TgCRND8 mice. AD patients have a series of pathological characteristics, such as neurofibrillary tangles, senile plaques, and a massive loss of brain weight and volume, especially in the hippocampus, a region most closely associated with memory functions . A synapse is a basic unit that transmits information among neurons and is under tight spatiotemporal regulation, and the aberrant function of synapses is strongly implicated in various neurological disorders such as AD . Loss of functional synapses accompanied by learning and memory impairments is also apparent in various AD transgenic animal models . Synaptic plasticity forms the molecular foundation of learning and memory in the central nervous system. As the structural function of cognition, synapse proteins such as PSD93, PSD95, SYN 1, SYT 1, and SYN play crucial roles in signaling conduction and learning and memory function. In the present study, we found that MN administration significantly increased the levels of some synaptic proteins including PSD93, PSD95, SYN 1, SYT 1, and SYN in the brains of TgCRND8 mice. Taken together, these results suggest that the preventive effects of MN treatment on the learning and memory impairments of TgCRND8 mice are at least partially associated with the improvement of the synaptic dysfunction. The phosphoinositide 3-kinase/protein kinase B (PI3K/Akt) signaling pathway is important for sustaining the function of neurons. Increasing evidence reveals that the PI3K/Akt/GSK-3 β pathway can be altered by A β deposit in the brains of AD patients. Dysfunction of the PI3K/Akt signaling pathway can increase the activity of GSK-3 β and regulate the metabolism of A β , leading to the hyperphosphorylation and deposition of tau protein, thus contributing to the formation of neurofibrillary tangles in the AD brains . The “GSK3 hypothesis in AD” suggests that the overactivation of GSK-3 β is closely related to several features of the pathology of AD, including microglia-mediated inflammation, A β production, APP processing, tau phosphorylation, neuronal death, and cognitive deficits . A recent study revealed that GSK-3 β deletion in the dentate gyrus of mice suppressed hippocampal synaptic transmission and decreased the levels of synapse proteins such as PSD93 and SYN . Thus, inhibition of GSK-3 β has been found to be beneficial and provide a potential therapeutic target for neurodegenerative diseases . A previous study demonstrated that MN could modulate the activation of Akt in SAMP8 mice . Our present results indicated that MN markedly activated the Akt activity and inhibited the GSK-3 β activation to maintain the synaptic plasticity and memory function of TgCRND8 mice. NF- κ B is a key nuclear transcription factor playing a cardinal role in the inflammatory response . Under normal physiological conditions, the inactive NF- κ B bound to the inhibitory protein I κ B kinase and is located in the cytosol . After stimulated by cellular stress, the NF- κ B complex is activated by dissociation from I κ B . Activated NF- κ B is then translocated from the cytosol to the nucleus to promote the expression of downstream proinflammatory target genes . More interestingly, the activation of inflammasome plays a critical role in the pathogenesis of AD, such as A β production and cognitive impairments, via modulating the chronic inflammatory response . Therefore, inhibition of the activation of the NF- κ B pathway would be beneficial for the suppression of inflammatory processes. MN is widely used as an effective anti-inflammatory agent. A recent study showed that MN markedly inhibited the inflammatory responses stimulated by the fimbriae of Porphyromonas gingivalis via inhibition of the NF- κ B pathway in RAW264.7 macrophages . MN also attenuated the NLRP3 inflammasome via inhibition of the NF- κ B signaling pathway on lupus nephritis in MRL/lpr mice . Consistent with these previous studies, we also found MN treatment conspicuously alleviated neuroinflammation partially by inhibiting the NF- κ B pathway activation in the brains of TgCRND8 mice. is a schematic drawing depicting the molecular mechanisms associated with the cognitive deficits improving the effects of MN on TgCRND8 mice. In sum, our project has revealed for the first time that MN ameliorates the learning and memory impairments in TgCRND8 mice, and the neuroprotective effects of MN are attributed to (1) the suppression of A β deposition via inhibiting the activities of β - and γ -secretases, and the enhancement of the activities of A β -degrading enzymes, and (2) the inhibition of neuroinflammatory and synaptic dysfunction, partially via regulation of PI3K/Akt/GSK-3 β and NF- κ B signaling pathways. Among the underlying mechanisms, the synaptic dysfunction inhibitory effect may be dominant in the anti-AD effects of MN. Therefore, more investigations are warranted to explore the anti-AD effects of MN targeting synaptic dysfunction modulation. We believe that MN is a promising naturally occurring constituent worthy of further development into anti-AD therapeutics. |
Multidrug-Resistant | 915fc6f6-d1d8-41d8-bbd8-4f16f4c1e3d8 | 11357147 | Microbiology[mh] | Bivalves are filter-feeding animals able to accumulate contaminants and microorganisms. Freshwater bivalves (FBs), such as mussels, are widespread in environments, like the Tua River, where they act as natural filters, concentrating particles, including bacteria . This crucial ecological function not only makes them effective bioindicators of water quality, but also potential reservoirs for microbial pathogens . Among the Portuguese species of freshwater bivalves, the focus of this study is centered on the autochthonous species, Anodonta anatina . This species can be found in Europe and Asia, from the Iberian Peninsula, in the south, to Scandinavia, in the north, and Russia, in the east. In terms of habitat selection and host fish utilization, A. anatina is considered a generalist species, as it colonizes lotic and lentic systems, including streams, large rivers, lakes, and reservoirs . Antimicrobial resistance (AMR) refers to the absence or diminished effectiveness of antimicrobial agents in inhibiting bacterial growth and, in cases involving pathogenic microorganisms, this resistance can lead to treatment failure , being recognized as one of the ten major global public health concerns . Indeed, AMR is responsible for numerous deaths annually worldwide, with a significant portion occurring within the European Union . Antibiotic-resistant bacteria (ARB) can be acquired by humans through various means, including exposure in healthcare facilities, community settings, and contact with animals or contaminated environments . Escherichia coli is an indicator of fecal contamination in food, marine, and freshwater environments, and has also been suggested as a possible indicator to assess the AMR status in environmental settings . While most E. coli strains in the human gut are commensal, certain strains exhibit a pathogenic potential , which are responsible for causing diarrhea diseases and based on virulence traits can be grouped in different pathotypes: enteropathogenic (EPEC), shiga toxin-producing (STEC), enteroaggregative (EAEC), enterotoxigenic (ETEC), enteroinvasive (EIEC), and diffusely adherent E. coli (DAEC) . In addition, there are also the extraintestinal pathogenic (ExPEC) strains that include neonatal meningitis-associated E. coli (NMEC), uropathogenic E. coli (UPEC), sepsis-causing (SPEC) and, avian pathogenic (APEC) . Moreover, the use and misuse of antibiotics have led to the emergence of multidrug-resistant (MDR) E. coli strains, posing significant challenges for clinical treatment and public health management . Integrating indicators of fecal contamination, such as MDR E. coli , into ecological assessments could enhance monitoring efforts, aligning with the ‘One Health’ concept, which emphasizes the interrelation of human, animal, and environmental health . The presence of antibiotics in aquatic environments is an escalating concern due to their pervasive use in agriculture, human medicine, and industrial applications. These substances are often introduced into water bodies through agricultural runoff, sewage effluents, aquaculture, and industrial discharges, leading to the contamination of rivers, lakes, and streams . Antibiotics can disrupt the microbial ecology of water bodies, impacting primary producers, such as algae , posing risks to higher trophic levels, including humans, who consume contaminated fish and shellfish . The mobility of ARB through interconnected water systems facilitates the global dissemination of resistance, complicating international efforts to manage and contain antimicrobial resistance . This contamination has far-reaching implications, affecting ecological health, water quality, and public health. The Clermont phylogenetic method categorizes E. coli into distinct phylogroups, which are associated with varying pathogenic potentials and ecological niches . Different phylogroups, such as B2 and D, are linked to specific pathogenic strains that cause serious infections, while others, like A and B1, are typically less harmful. This classification aids in identifying the sources and risks posed by different E. coli strains, which is essential for public health, epidemiology, and the development of targeted interventions . Although bacteria serve as a feeding source for bivalves, they can also inhabit the body tissues of healthy individuals outside the gut. Mussels possess the ability to establish either mutually beneficial or antagonist symbiotic relations with bacteria . Given the growing consumption of raw or undercooked foods, understanding the potential transmission of MDR E. coli through FBs is essential for mitigating public health risks . The present study aims to evaluate the phylogenetic diversity, pathotypic characterization, and antimicrobial susceptibility of E. coli isolates from FBs sampled within the Tua River region in Portugal.
2.1. Freshwater Bivalve Collection Fifteen A. anatina freshwater bivalves were collected from the Tua River basin (year 2022) at two different locations, location I (Chelas 41°30′45.82′′ N; 7°12′32.92′′ W: 7 FB) and location II (Barcel 41°24′18.69′′ N; 7°9′38.93′′ W: 8 FB), which are, respectively, upstream and downstream of Mirandela city (northeast of Portugal) . The characterization of the study area and sampling processing are fully described in previous studies . The freshwater bivalves were collected and maintained alive in a cooler with moist towels and transported to the Department of Veterinary Sciences, Antimicrobials, Biocides & Biofilms Unit (AB2Lab-DCV, CITAB), of the University of Trás-os-Montes and Alto Douro (UTAD), located in Vila Real, Portugal. The sampling of mussels was carried out with a permit granted by the Institute for the Conservation of Nature and Forestry (ICNF). No ethics committee approval was needed, and no animal experiments were performed in the scope of this research. 2.2. Sample Processing, Isolation, and Identification of Bacteria The fifteen FBs ( A. anatina ) were subjected to measurements of shell dimensions, followed by an aseptic opening to collect and weigh soft tissues. These tissues were then transferred to flasks containing Brain Heart Infusion (BHI) medium and incubated at 37 ± 1 °C for 24 h. Subsequently, the samples were inoculated on Chromocult ® Coliform Agar (CCA ® ) (Oxoid, Basingstoke, UK), a chromogenic medium, and the plates further incubated at 37 ± 1 °C for 24 h. Presumptive E. coli colonies were identified based on their characteristic blue/purple coloration on CCA ® according to the manufacturer’s guidelines. Confirmation of characteristic colonies was carried out using lactose fermentation on MacConkey agar plates and standard biochemical test-IMViC reactions (Indol, Methyl-red, Voges-Proskauer, and Citrate) were inoculated and incubated at 35–37 ± 1 °C for 24 h. Each bacterial isolate was assigned a specific code comprising letters (LI/LII) indicating the sampling location, Chelas (LI) or Barcel (LII), followed by an alphanumeric code representing the individual bivalve, and the designation “Ec” standing for E. coli followed by a number representing the strain (e.g., LIFB1Ec1), as can be observed in . 2.3. Antimicrobial Susceptibility Test Susceptibility assays were performed using the agar disk diffusion method, following the Kirby–Bauer technique, in accordance with the guidelines provided by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) , on Mueller–Hinton (MH) agar plates (OXOID, UK). A total of twenty-one antibiotics, covering seven antibiotic classes, was employed in the susceptibility assays: β-lactams (including penicillins: amoxicillin (AML, 10 μg), amoxicillin/clavulanic acid (AMC, 20/10 μg), piperacillin (PRL, 100 μg), and piperacillin/tazobactam (TZP, 100/10 μg); cephalosporins: cefoxitin (FOX, 30 μg), ceftazidime (CAZ, 30 μg), cefotaxime (CTX, 30 μg), and ceftriaxone (CRO, 30 μg); carbapenems: imipenem (IMP, 10 μg), meropenem (MEN, 10 μg), and ertapenem (ETP, 10 μg); and monobactam: aztreonam (ATM, 30 μg). Additionally, fluoroquinolones: ciprofloxacin (CIP, 5 μg); aminoglycosides: kanamycin (K, 30 μg), tobramycin (TOB, 10 μg), gentamicin (CN, 30 μg), and amikacin (AK, 30 μg); sulphonamides: sulfamethoxazole/trimethoprim (SXT, 23.75/1.25 μg); amphenicols: chloramphenicol (C, 30 μg); tetracyclines: tetracycline (TE, 30 μg); and fosfomycin (FOS, 50 μg) were included. Interpretation of the results was based on the breakpoints guidelines provided by the Clinical and Laboratory Standards Institute . The bacteria were classified as susceptible (S), intermediate (I), or resistant (R). Reference strain E. coli ATCC 25922 was adopted as the control strain. Isolates exhibiting resistance to at least three different antimicrobial groups were categorized as multidrug-resistant (MDR) . 2.4. Phylogenetic Determination of E. coli Isolates To determine the phylogenetic groups (A, B1, B2, D, E, F, and clade I) of E. coli , the multiplex PCR method described by Clermont et al. was employed . This involved the amplification and sequencing of six conserved housekeeping genes (chuA, yjaA, Tsp.E4.C2, arpA, arpA (group E), and trpA (group C)). Specific primers for PCR amplification were synthesized by STAB-Vida (Caparica, Portugal), as listed in . In brief, DNA extraction was performed using the GF-1 Bacterial DNA Extraction Kit (Vivantis, Shah Alam, Malaysia) following the manufacturer’s instructions. Subsequently, the concentration of DNA samples was quantified using a Biotek Powerwave XS2 Microplate Reader (Agilent Technologies, Winooski, VT, USA). by measuring absorbances at A260 nm and A280 nm, with sample purity assessed by the A260/A280 nm ratio. An evaluation of DNA integrity was performed through 0.6% agarose gel electrophoresis, dyed with 2 μL GreenSafe dye. Each reaction was performed in a total volume of 20 µL. The reaction mixture comprised 3 µL of genomic DNA (30 ng/µL), 10× PCR buffer, 2 mM of each deoxyribonucleotide triphosphate (dNTP), 2 units of Taq polymerase (Bioron), 25 mM of MgCl 2 , 10 µL of specific forward and reverse primers , and ddH 2 O to obtain the final volume. The PCR reactions proceeded under the following conditions: initial denaturation at 95 ± 1 °C for 5 min, followed by 34 cycles consisting of denaturation at 94 ± 1 °C for 45 s, annealing at 57 ± 1 °C for 45 s (for group E) or 59 ± 1 °C for 45 s (for quadruplex and group C), extension at 72 ± 1 °C for 1 min, and a final extension step at 72 ± 1 °C for 5 min. The procedure was finalized by electrophoresis on 2% agarose gels in 1× Tris-borate-EDTA (TBE) buffer supplemented with GreenSafe DNA Gel Stain. The phylogenetic groups of each strain were determined after analyzing the electrophoresis gel, through the presence and/or absence of the genes represented in . 2.5. Determination of E. coli Pathotypes DNA extraction was performed using NZYTECH bacterial Cell Lysis Buffer (Ref. MB17801, NZYTECH, Coimbra, Portugal) with heat treatment at 95 °C for 15 min followed by centrifugation at 10,000 rpm for 3 min. The identification of E. coli pathotypes (ETEC, EIEC, EAEC, and EHEC/STEC) was conducted through the real-time multiplex PCR technique using 20 µL of mixture (specific primers and probes for PCR amplification synthesized by Eurofins Genomics, Germany, ultra-pure water and mastermix with hot start temperature from NZYTECH, Portugal, ref. MB23003) and 5 µL of DNA template. Primers and probes were described by the EURL-VTEC_Method 02 for EHEC/VTEC , the EU-RL VTEC_Method_07 for EIEC , the EU-RL VTEC_Method_08 for ETEC , the EURL-VTEC_Method_05 for EAEC , and the ISO/TS 13136:2012 standard and are defined in . Positive controls, including E. coli strains, LMV_E_37 (eae+; bfp+), LMV_E_38 (est+), LMV_E_39 (12 et+), LMV_E_40 (ipaH+), LMV_E_41 (aggr+; cvd432+), and O157:H7 (eae+; stx1+; stx2+), were used. The thermoprofile used for the real-time PCR reaction was the initial denaturation step at 95 °C for 3 min, followed by 40 cycles consisting of denaturation at 95 °C for 15 s, annealing at 52 °C for 25 s, and extension at 72 °C for 30 s.
Fifteen A. anatina freshwater bivalves were collected from the Tua River basin (year 2022) at two different locations, location I (Chelas 41°30′45.82′′ N; 7°12′32.92′′ W: 7 FB) and location II (Barcel 41°24′18.69′′ N; 7°9′38.93′′ W: 8 FB), which are, respectively, upstream and downstream of Mirandela city (northeast of Portugal) . The characterization of the study area and sampling processing are fully described in previous studies . The freshwater bivalves were collected and maintained alive in a cooler with moist towels and transported to the Department of Veterinary Sciences, Antimicrobials, Biocides & Biofilms Unit (AB2Lab-DCV, CITAB), of the University of Trás-os-Montes and Alto Douro (UTAD), located in Vila Real, Portugal. The sampling of mussels was carried out with a permit granted by the Institute for the Conservation of Nature and Forestry (ICNF). No ethics committee approval was needed, and no animal experiments were performed in the scope of this research.
The fifteen FBs ( A. anatina ) were subjected to measurements of shell dimensions, followed by an aseptic opening to collect and weigh soft tissues. These tissues were then transferred to flasks containing Brain Heart Infusion (BHI) medium and incubated at 37 ± 1 °C for 24 h. Subsequently, the samples were inoculated on Chromocult ® Coliform Agar (CCA ® ) (Oxoid, Basingstoke, UK), a chromogenic medium, and the plates further incubated at 37 ± 1 °C for 24 h. Presumptive E. coli colonies were identified based on their characteristic blue/purple coloration on CCA ® according to the manufacturer’s guidelines. Confirmation of characteristic colonies was carried out using lactose fermentation on MacConkey agar plates and standard biochemical test-IMViC reactions (Indol, Methyl-red, Voges-Proskauer, and Citrate) were inoculated and incubated at 35–37 ± 1 °C for 24 h. Each bacterial isolate was assigned a specific code comprising letters (LI/LII) indicating the sampling location, Chelas (LI) or Barcel (LII), followed by an alphanumeric code representing the individual bivalve, and the designation “Ec” standing for E. coli followed by a number representing the strain (e.g., LIFB1Ec1), as can be observed in .
Susceptibility assays were performed using the agar disk diffusion method, following the Kirby–Bauer technique, in accordance with the guidelines provided by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) , on Mueller–Hinton (MH) agar plates (OXOID, UK). A total of twenty-one antibiotics, covering seven antibiotic classes, was employed in the susceptibility assays: β-lactams (including penicillins: amoxicillin (AML, 10 μg), amoxicillin/clavulanic acid (AMC, 20/10 μg), piperacillin (PRL, 100 μg), and piperacillin/tazobactam (TZP, 100/10 μg); cephalosporins: cefoxitin (FOX, 30 μg), ceftazidime (CAZ, 30 μg), cefotaxime (CTX, 30 μg), and ceftriaxone (CRO, 30 μg); carbapenems: imipenem (IMP, 10 μg), meropenem (MEN, 10 μg), and ertapenem (ETP, 10 μg); and monobactam: aztreonam (ATM, 30 μg). Additionally, fluoroquinolones: ciprofloxacin (CIP, 5 μg); aminoglycosides: kanamycin (K, 30 μg), tobramycin (TOB, 10 μg), gentamicin (CN, 30 μg), and amikacin (AK, 30 μg); sulphonamides: sulfamethoxazole/trimethoprim (SXT, 23.75/1.25 μg); amphenicols: chloramphenicol (C, 30 μg); tetracyclines: tetracycline (TE, 30 μg); and fosfomycin (FOS, 50 μg) were included. Interpretation of the results was based on the breakpoints guidelines provided by the Clinical and Laboratory Standards Institute . The bacteria were classified as susceptible (S), intermediate (I), or resistant (R). Reference strain E. coli ATCC 25922 was adopted as the control strain. Isolates exhibiting resistance to at least three different antimicrobial groups were categorized as multidrug-resistant (MDR) .
To determine the phylogenetic groups (A, B1, B2, D, E, F, and clade I) of E. coli , the multiplex PCR method described by Clermont et al. was employed . This involved the amplification and sequencing of six conserved housekeeping genes (chuA, yjaA, Tsp.E4.C2, arpA, arpA (group E), and trpA (group C)). Specific primers for PCR amplification were synthesized by STAB-Vida (Caparica, Portugal), as listed in . In brief, DNA extraction was performed using the GF-1 Bacterial DNA Extraction Kit (Vivantis, Shah Alam, Malaysia) following the manufacturer’s instructions. Subsequently, the concentration of DNA samples was quantified using a Biotek Powerwave XS2 Microplate Reader (Agilent Technologies, Winooski, VT, USA). by measuring absorbances at A260 nm and A280 nm, with sample purity assessed by the A260/A280 nm ratio. An evaluation of DNA integrity was performed through 0.6% agarose gel electrophoresis, dyed with 2 μL GreenSafe dye. Each reaction was performed in a total volume of 20 µL. The reaction mixture comprised 3 µL of genomic DNA (30 ng/µL), 10× PCR buffer, 2 mM of each deoxyribonucleotide triphosphate (dNTP), 2 units of Taq polymerase (Bioron), 25 mM of MgCl 2 , 10 µL of specific forward and reverse primers , and ddH 2 O to obtain the final volume. The PCR reactions proceeded under the following conditions: initial denaturation at 95 ± 1 °C for 5 min, followed by 34 cycles consisting of denaturation at 94 ± 1 °C for 45 s, annealing at 57 ± 1 °C for 45 s (for group E) or 59 ± 1 °C for 45 s (for quadruplex and group C), extension at 72 ± 1 °C for 1 min, and a final extension step at 72 ± 1 °C for 5 min. The procedure was finalized by electrophoresis on 2% agarose gels in 1× Tris-borate-EDTA (TBE) buffer supplemented with GreenSafe DNA Gel Stain. The phylogenetic groups of each strain were determined after analyzing the electrophoresis gel, through the presence and/or absence of the genes represented in .
DNA extraction was performed using NZYTECH bacterial Cell Lysis Buffer (Ref. MB17801, NZYTECH, Coimbra, Portugal) with heat treatment at 95 °C for 15 min followed by centrifugation at 10,000 rpm for 3 min. The identification of E. coli pathotypes (ETEC, EIEC, EAEC, and EHEC/STEC) was conducted through the real-time multiplex PCR technique using 20 µL of mixture (specific primers and probes for PCR amplification synthesized by Eurofins Genomics, Germany, ultra-pure water and mastermix with hot start temperature from NZYTECH, Portugal, ref. MB23003) and 5 µL of DNA template. Primers and probes were described by the EURL-VTEC_Method 02 for EHEC/VTEC , the EU-RL VTEC_Method_07 for EIEC , the EU-RL VTEC_Method_08 for ETEC , the EURL-VTEC_Method_05 for EAEC , and the ISO/TS 13136:2012 standard and are defined in . Positive controls, including E. coli strains, LMV_E_37 (eae+; bfp+), LMV_E_38 (est+), LMV_E_39 (12 et+), LMV_E_40 (ipaH+), LMV_E_41 (aggr+; cvd432+), and O157:H7 (eae+; stx1+; stx2+), were used. The thermoprofile used for the real-time PCR reaction was the initial denaturation step at 95 °C for 3 min, followed by 40 cycles consisting of denaturation at 95 °C for 15 s, annealing at 52 °C for 25 s, and extension at 72 °C for 30 s.
3.1. Bivalve Characterization Despite being listed by the International Union for Conservation of Nature as Least Concern , A. anatina has suffered a strong decline in the last 20 years and, as such, there was a rule to collect only 5% of the total number of individuals existing at the sampling sites. The characteristics of each bivalve are specified in . 3.2. Antimicrobial Susceptibility Tests presents an overview of the antimicrobial susceptibility of the twenty E. coli isolates. All the isolates showed susceptibility to PRL, FOX, CAZ, ATM, CN, STX, C, TE, and FOS. AK showed intermediate resistance for 35.0% (7 isolates) of the 20 isolates, CTX and TOB for 30.0% (6 isolates) each, K for 25.0% (5 isolates), TZP for 15.0% (3 isolates), CRO for 10.0% (2 isolates), and finally IMP for 5.0% (1 isolate). MEN was the antibiotic to which all isolates showed resistance (100%); in decreasing resistance order of the isolates: TOB with 50.0% (10 isolates), ETP and K with 45.0% (9 isolates), AMC and AK with 35.0% (7 isolates), AML with 30.0% (6 isolates), and CIP with 5.0% (1 isolate). 3.3. Multiresistant Isolates The analysis of the resistance profiles for the twenty E. coli isolates showed that none of the isolates under study were susceptible to all the tested antimicrobial groups; however, 8 (40.0%) were resistant to two groups and 12 (60.0%) were MDR (resistant to three or more classes of antimicrobials). The largest number of isolates (10 isolates) exhibited resistance to three different antimicrobial classes; moreover, one isolate exhibited resistance to four classes and one to five classes. reviews the multiple MDR patterns exhibited by the 20 isolates. 3.4. Phylogenetic Analysis and E. coli Pathotype Identification The classification of the E. coli isolates into the phylogroups was proposed by Clermont et al. . Overall, 55% (11/20) belonged to phylogroup B1, 15% (3/20) to phylogroups D or E, 10% (2/20) to phylogroup A, 10% (2/20) to phylogroup E or Clade I, and for 10% (2/20) of the isolates it was not possible to identify to which phylogroup they belonged, thus being termed unknown. The results obtained are summarized in . The multiplex PCR analysis aimed at identifying E. coli pathotypes revealed that none of the twenty isolates harbored virulence genes associated with diarrheagenic E. coli strains. Hence, these isolates do not belong to pathogenic E. coli strains, indicating the predominantly commensal nature of the recovered strains in this study.
Despite being listed by the International Union for Conservation of Nature as Least Concern , A. anatina has suffered a strong decline in the last 20 years and, as such, there was a rule to collect only 5% of the total number of individuals existing at the sampling sites. The characteristics of each bivalve are specified in .
presents an overview of the antimicrobial susceptibility of the twenty E. coli isolates. All the isolates showed susceptibility to PRL, FOX, CAZ, ATM, CN, STX, C, TE, and FOS. AK showed intermediate resistance for 35.0% (7 isolates) of the 20 isolates, CTX and TOB for 30.0% (6 isolates) each, K for 25.0% (5 isolates), TZP for 15.0% (3 isolates), CRO for 10.0% (2 isolates), and finally IMP for 5.0% (1 isolate). MEN was the antibiotic to which all isolates showed resistance (100%); in decreasing resistance order of the isolates: TOB with 50.0% (10 isolates), ETP and K with 45.0% (9 isolates), AMC and AK with 35.0% (7 isolates), AML with 30.0% (6 isolates), and CIP with 5.0% (1 isolate).
The analysis of the resistance profiles for the twenty E. coli isolates showed that none of the isolates under study were susceptible to all the tested antimicrobial groups; however, 8 (40.0%) were resistant to two groups and 12 (60.0%) were MDR (resistant to three or more classes of antimicrobials). The largest number of isolates (10 isolates) exhibited resistance to three different antimicrobial classes; moreover, one isolate exhibited resistance to four classes and one to five classes. reviews the multiple MDR patterns exhibited by the 20 isolates.
The classification of the E. coli isolates into the phylogroups was proposed by Clermont et al. . Overall, 55% (11/20) belonged to phylogroup B1, 15% (3/20) to phylogroups D or E, 10% (2/20) to phylogroup A, 10% (2/20) to phylogroup E or Clade I, and for 10% (2/20) of the isolates it was not possible to identify to which phylogroup they belonged, thus being termed unknown. The results obtained are summarized in . The multiplex PCR analysis aimed at identifying E. coli pathotypes revealed that none of the twenty isolates harbored virulence genes associated with diarrheagenic E. coli strains. Hence, these isolates do not belong to pathogenic E. coli strains, indicating the predominantly commensal nature of the recovered strains in this study.
The present study’s findings highlight the importance of freshwater bivalves as bioindicators for assessing contamination levels and associated risks to public health. Escherichia coli , a well-known indicator of fecal contamination in water sources, poses a significant global concern, particularly with the emergence of multidrug-resistant (MDR) strains . Indeed, the present study aimed to investigate the phylogenetic diversity of E. coli isolated from freshwater bivalves ( Anodonta anatina ) and to characterize their phenotypes and antibiotic resistance profiles. There are no differences in the bivalves’ characteristics between locations, namely in length and weight. Out of the total isolates examined, twelve (60.0%) were classified as multidrug-resistant (MDR), demonstrating resistance to three or more antimicrobial classes. Consistent findings across multiple studies indicate a higher prevalence of antibiotic-resistant bacteria, including MDR strains, in organisms compared to water samples. Consequently, mollusks serve as a more reliable matrix for monitoring MDR bacteria, offering more robust assessment results than water samples. This highlights the importance of incorporating organism-based surveillance approaches into antimicrobial-resistance monitoring programs to better understand and address the spread of multidrug resistance in aquatic environments . Noteworthy, all isolates exhibited resistance to meropenem, and the isolates from location I displayed resistance not only to meropenem, but also ertapenem, belonging to the carbapenem class, which are deemed last-resort antibiotics . Additionally, amoxicillin and amoxicillin/clavulanic acid resistance were observed in Chelas (location I) in contrast to location II (Barcel), further distinguishing the resistance profiles between the two sites . These variances maybe linked to the anthropogenic impact between the two locations. In location I (Chelas), the impacts mainly come from wastewater from hospitals, care facilities, and agriculture, whereas in location II (Barcel), they are primarily due to the agri-food industry present in the area. In line with this result, previous research linked the presence of E. coli strains exhibiting higher levels of multidrug resistance due to anthropogenic influences. Varandas et al. stated that higher resistance rates were observed in locals with higher industrial activity, a larger population density, and pressures from livestock farming . These findings align closely with the results of the present study, particularly in elucidating the marked differences in resistance profiles between location I and location II. This further emphasizes the intricate interplay between environmental factors, human activities, and the emergence and dissemination of antibiotic resistance in natural ecosystems. The emergence of multidrug-resistant E. coli strains is of paramount concern globally, given their rapid dissemination, as highlighted by the WHO . These strains exhibit resistance not only to amoxicillin-based antibiotics, but also to fluoroquinolones, cephalosporins, and carbapenems, rendering conventional therapeutic options ineffective, as noted by a previous study . Various studies have consistently shown that β-lactam antibiotics have been the preferred choice for treating infections caused by pathogenic strains of E. coli . However, the widespread use and sometimes inappropriate administration of these antibiotics have contributed to the emergence of β-lactamase-producing strains. This phenomenon emphasizes how closely patterns of antibiotic use and the emergence of antibiotic resistance in bacterial populations are. The development of innovative therapeutic approaches and prudent antibiotic prescribing practices are two crucial steps for halting the spread of resistance mechanisms and reducing the negative effects of antimicrobial resistance on public health . The efficacy of β-lactam antibiotics is increasingly compromised by the rise in carbapenem-resistant Enterobacteriaceae, as highlighted by previous studies . Alarmingly, existing protocols for assessing the hygiene and safety of harvested and distributed bivalves nationally and internationally do not incorporate evaluations for antimicrobial resistances that could be transmitted through these organisms, as noted by the WHO . Addressing this gap in regulatory frameworks is imperative to safeguard public health and mitigate the spread of antibiotic resistance through foodborne pathways. Concerning the phylogroups, the twenty isolates were allocated to one of the phylogroups delineated by Clermont et al. . Each phylogroup plays a distinct ecological role, emphasizing the importance of categorizing E. coli strains into different groups to comprehend their pathogenicity, host interactions, and ecological impact on aquatic systems, as highlighted by other studies . The author Giacometti et al. noted that phylogroup B1 is among the most prevalent in mollusks, a finding consistent with the results of the present study. Furthermore, research by Bong et al. demonstrated that multidrug-resistant isolates predominantly belong to phylogenetic group B1, further underlining the significance of phylogroup classification in understanding the dynamics of antimicrobial resistance in aquatic environments. Studies conducted on wastewater and surface water samples consistently indicate that most isolates belong to phylogroups A and B1, reflecting the prevalence of these groups in environments impacted by human activities, as demonstrated in various studies . This observation aligns with research highlighting the robust survival capacity of strains in phylogroup B1 in aquatic environments . The prevalence and resistance of phylogroup B1 in the studied bivalves may be attributed to its widespread distribution in this ecosystem. Notably, a significant proportion of the isolates characterized in this study belonged to phylogroups A and B1, suggesting a commensal origin, whereas E. coli strains classified as phylogroups B2 or D are typically associated with extraintestinal infections . The real-time multiplex PCR analysis aimed at identifying E. coli pathotypes revealed that none of the isolates were associated with diarrheagenic E. coli groups. This finding underscores the primarily commensal nature of the isolated E. coli strains in this study, suggesting their potential role as indicators rather than direct pathogens in the aquatic environment. This distinction is crucial for understanding the ecological dynamics of E. coli populations and assessing the associated risks of antimicrobial-resistance transmission within aquatic ecosystems. This study provides a comprehensive overview of freshwater bivalves as reservoirs of multidrug-resistant bacteria, shedding light on their potential implications for broader river ecosystems. Additionally, it underscores the emerging research domain of using freshwater bivalves as indicators for antimicrobial resistance, emphasizing the imperative need to further investigations to validate their effectiveness in this role. However, certain limitations are acknowledged, primarily stemming from the constrained collection of bivalves due to the endangered status of the species A. anatina . Moreover, the study underscores the importance of employing multidisciplinary methodologies to assess the ecological integrity of aquatic systems and advocates for the integration of microbiological analyses into ecosystem monitoring endeavors, guided by the ‘One Health’ concept. This holistic approach is essential for comprehensively understanding and addressing the complex dynamics of antimicrobial resistance in aquatic environments.
This work consisted in the study of E. coli strains recovered from freshwater bivalves in a river in Portugal. It revealed significant antimicrobial resistance, including to carbapenems, with variability in resistance rates between the two different sampling sites, linked to anthropogenic influences. The detection of carbapenem resistance is concerning due to potential food chain transmission. Most isolates belonged to commensal phylogroups, primarily B1, likely due to fecal contamination. These findings underline the importance of the One Health approach for monitoring and preventing the spread of antimicrobial resistance through aquatic environments, highlighting the need for further research and preventive measures to protect public and environmental health. Addressing this issue requires a comprehensive strategy that uses advancements in technology, regulatory frameworks, and public engagement to protect water quality and control the spread of antibiotic resistance.
|
Water, sanitation, and hygiene in selected health facilities in Ethiopia: risks for healthcare acquired antimicrobial-resistant infections | 37e4004a-7c73-4bec-b479-316772890ac8 | 11647025 | Microbiology[mh] | Background Hospital-acquired infections (HAIs) can increase morbidity and mortality, increase health care cost due to prolonged stay, and contribute to increased microbial resistance due to the widespread occurrence of multi-drug resistant (MDR) pathogens in health facilities . Approximately, one-third of neonatal deaths annually (680,000) caused by infections The share of HAIs to this remains uncertain, but earlier studies have shown that rates of neonatal infections among hospital-born children in low-income countries are 3–20 times higher than those in higher income countries . Most of these infections were present soon after birth and were resistant to antibiotics. Inadequate water, sanitation and hygiene (WASH) and the low adherence to infection control protocols, unsafe waste management, exacerbated by the overcrowding of health facilities increase the risk for HAIs . Recent estimates suggest that HAIs affect about 8% of patients in regular wards and more than half of patients admitted in intensive care units (ICU) in low income settings . A recent study from Jimma University Medical Center reported a prevalence of HAIs of 19%, and the risk was significantly higher in those that received surgical procedures . A study conducted in rural health care facilities in Ethiopia, Kenya, Mozambique, Rwanda, Uganda and Zambia, reported that less than 50% of the surveyed facilities had access to: improved water sources on their premises, improved sanitation, hand washing facilities with constant access to water and soap . In Ethiopia, only an estimated 55% of health facilities have access to basic water services . However, such data is scarce for lower level health facilities such as woreda (district) health centers and health posts, where the problem may be even more significant. The global burden of HAIs has increased due to antibiotic-resistant bacteria, raising risk to health, particularly in developing countries . The WHO African region estimated 1.05 million deaths associated with bacterial antimicrobial resistance (AMR) in 2019 . A recent study revealed that 23.5% of the patients had HAIs, with surgical site infections (SSI) being the most common, and primarily acquired for preventive purposes. From this, Ethiopia used 698.2 tons of antibiotics in 2018, according to the country’s most recent national data, with a per capita usage of 5.8grams, where the one antibiotic product that completely explained the 20.8% consumption; level of beta-lactamase-resistant . The Ethiopian policy brief and the regional state reports of Oromia and SNNPs regions primarily indicate that the need of cooperation with the long-term investment of a lasting solution, as well as the necessity of WASH response to avoid cholera outbreaks . Understanding the magnitude of nosocomial pathogens and their AMR would help design interventions that improve WASH and infection prevention control in health care facilities, but will also contribute to improving the quality of health care delivered. Therefore, the present study aimed to assess WASH facilities and practices, and levels of nosocomial pathogens in hand-touch sites in selected health facilities in Oromia and SNNPs Regions. Materials and methods 2.1 Study area and design A facility based observational cross-sectional study design was employed for the WASH compliance and survey of pathogens occurrence from random spots in the health facilities of two regions. This study is reporting on a baseline assessment conducted in health facilities of Bidre town in Bale zone from Oromia Region, Bulle town in Gedeo zone and Doyogena in Kembata-Tembaro zone both from SNNPs region. The WASH assessments included all health facilities that were functional at the time of the survey (i.e., health post, health centers, and hospitals). 2.2 Sample size and sampling procedure To select appropriate sample size, the current study involved all hospitals from 3 districts and 30% of health facilities sample from the WASH program implementation of UNICEF-Ethiopia in 2 specific regions randomly. The study site was selected based on the list provided by UNICEF-Ethiopia and the possibility to transfer microbial samples in time (in 24 h) was considered. From the list of health facilities, we selected a sub-sample, stratified by type of facility. From the 31 health care facilities, we selected a subset of 12 health facilities from which sample was collected. We excluded pharmacies and clinics and focused on health posts, health centers and hospitals. 2.3 Assessment of WASH in health facilities To assess the facilities of WASH and practices, the observational checklist of core questions for infection prevention and control (IPC) and WASH common indicators is developed based on international standards—WHO/UNICEF . All questionnaires and checklists were translated into Amharic/Oromifaa and were pretested prior to the interviews. The checklist allowed the collection of information on the prevailing sanitary conditions, access to water and hand-washing facility, as well as hand-washing and waste disposal practices. The WHO protocol on monitoring fulfilment of opportunities for hand-hygiene was used to assess the health personnel’s adherence to hand- hygiene guidelines from June 2021 to July 2021. 2.4 Surface and water sample collection Sample collection was performed on August 2021 following the United States Center for Disease Control and Prevention (CDC) and Public health England guidelines . Surface and water sample primarily collected from hospitals and health centers. Surface sample collection was performed using sterile cotton swabs. The swabs were first moist in sterile normal saline solution. The samples were collected from surfaces including beds, door handles, walls, gowns, autoclaves, tables, and chairs. The sampling areas included out-patients departments, different wards, pharmacy, laboratories, receptions, toilets and cafeterias in the health facilities. Water samples were collected from sources from which the health facilities obtain water for washing, drinking and other activities in the healthcare settings. A total of 14 water sample is collected and delivered for analysis from delivery wards, medical ward, tanker, and bore-hole and rainwater collection systems. Overall, 59 water samples were collected from all health facilities, including health centers and health posts. 2.5 Sample handling and transportation The collected surface samples were immediately put in Amies transport media and kept in pre-cooled ice box and transported to SNNPs region Public Health Institute laboratory. On arrival at the laboratory, the surface samples were transferred to the nutrient broth and enriched overnight at 37°C. After an overnight incubation, the samples were inoculated on blood agar and MacConkey agar plates and put overnight at 37°C. In case of no growth after an overnight culture, the plates were incubated for an additional 24 h. The water samples were assessed for their safety using modified Method 9,215 to enumerate heterotrophic bacteria and membrane filtration technique for Gram-negative bacteria . To enumerate heterotrophic bacteria, 1 mL of each water sample was pipetted into a sterile petri dish. After thoroughly mixing, the melted MacConkey agar was poured into the dish. The melted medium was mixed thoroughly with the sample and solidified. The plates were incubated for 48 h at 37°C. The Gram-negative bacteria were counted by filtrating 100 mL water samples through 0.45 μm pore size-47 mm, and cellulose nitrate membranes using the modified ISO 9308-1 protocol . The samples were incubated on MacConkey agar for 24 h at 37°C. All results of Gram-negative bacteria were expressed as colony forming units per 100 mL water. The bacterial colonies were collected and put in Trypticase Soy Broth containing 20% glycerol and were transported to the National bacteriology and mycology Reference Laboratory (NRL) at the Ethiopian Public Health Institute, where they were stored in deep-freeze until further analyses. 2.6 Bacterial isolation and identification The bacteria were refreshed by culturing on three different culture media: (i) 5% sheep blood agar plate, (ii) MacConkey agar plate, and (iii) Mannitol salt agar plate. Colony appearance on culture plates, microscopic examination, and biochemical tests were used to identify Gram-positive and Gram-negative bacteria. 2.6.1 Identification of gram-positive cocci The common Gram-positive cocci are Staphylococcus spp. and Streptococcus spp. We used Blood agar and Mannitol salt agar media for isolation of Staphylococcus spp : The culture plates were incubated in air at 37°C for 24 h. Colony morphology on culture plates and microscopic examination for Gram-positive cocci in clusters were used for initial Staphylococcus spp. identification. Catalase and coagulase tests were used to classify Staphylococcus spp. into Staphylococcus aureus and coagulase negative Staphylococcus . All Staphylococci are Catalase positive and only S. aureus is coagulase positive. Streptococcus spp. were identified based on colony morphology on: (i) blood agar plates (beta hemolytic, alpha hemolytic and non-hemolytic), (ii) microscopic examination for Gram-positive in chin, and (iii) different biochemical tests. Negative catalase test differentiated Streptococcus spp. from Staphylococci Bacillus spp. Blood agar with 5% sheep blood media was used for the bacteria isolation. Colony morphology on the culture plates and gram stain were used for the bacterial identification. To differentiate Bacillus cereus from other Bacillus species we used citrate test which is only positive for B. cereus . 2.6.2 Identification of gram-negative bacilli The common gram negative bacteria are generally divided into two major categories: Fermenters and non-fermenters. Fermenters gram-negative bacilli utilize lactose and become pink color colonies on MacConkey agar while non-fermenters cannot utilize lactose and they are colorless colonies on MacConkey agar plate. Biochemical tests such as Triple Sugar Iron Agar (TSI), urea, citrate, Sulfide Indole Motility (SIM) medium, growth in Lysine Iron Agar (LIA), and oxidase were additionally used to identify Gram-negative bacteria. 2.7 Antimicrobial susceptibility testing The antimicrobial Susceptibility Tests (AST) were performed based on the Kirby–Bauer disk diffusion method on Mueller-Hinton agar (MHA) as recommended by clinical and laboratory standard Institute (CLSI) for all Gram-negative bacteria and Staphylococcus species . Well-isolated three to four colonies were emulsified in a tube containing sterile normal saline and the turbidity adjusted to 0.5 McFarland standards. The emulsified bacterial suspension was uniformly streaked on MHA plates using sterile cotton swabs, on which the antibiotic disks were applied and incubated for 18–24 h at 37°C. The antibiotic agents tested in this study were ampicillin (10 μg), amoxicillin-clavulanic acid (20/10 μg), pepracillin/ tazobactum, cefazolin (30 μg), cefuroxime (30 μg), cefotaxime (30 μg), ceftazidime (30 μg), cefepime (30 μg), cefoxitin (30 μg), ciprofloxacin (5 μg), amikacin (30 μg), meropenem (10 μg), chloramphenicol, tetracycline, cotrimoxazole, and penicillin. Penicillin and cefoxitin were tested only for Staphylococcus species and the result of oxacillin was determined from cefoxitin breakpoint. Antibiotic susceptibility results were interpreted according to the CLSI zone size interpretive standards . Intermediate results were considered resistant. Multidrug resistance (MDR) was defined according to guidelines compiled by the European Center for Disease prevention and Control (ECDC) and the Centers for Disease Control and Prevention (CDC) . Accordingly, bacterial isolates that were resistant to at least one agent in three different antimicrobial categories were considered as MDR. 2.8 Quality assurance All media, biochemical reagents, gram stain reagents and antibiotic disks were checked for their quality using standards ATCC strains. Standard ATCC quality strains used for this study were S. aureus ATCC ® 25923, E. coli ATCC ® 2592, P. aeruginosa ATCC 27853. 2.9 Data analysis Epi-info was used for data entry and the data was subsequently exported to Microsoft Excel and SPSS version 26 for data cleaning and further analysis. The frequencies of bacterial isolates and antimicrobial susceptibility were calculated. Mean and frequencies (percentage) were used to present descriptive data. 2.10 Ethics Ethical clearance was obtained from the Institutional Review Board of the College of Natural and Computational Sciences of Addis Ababa University (Ref. No: IRB/04/14/2021). Additionally, the research was ethically approved by letter of support is sent to Oromia and SNNPs regional Health Bureaus with the letter of minute no. (ምሳኒፕ/453/13/21). The Oromia and SNPP’s Regional Health Bureau Ethics review committee also reviewed and approved the research for the implementation. Prior to the collection of data, the informed consent was obtained from staff and the administration of the each health facilities. Every task and procedures was completed in accordance with the WHO guidance, rule and regulations. Study area and design A facility based observational cross-sectional study design was employed for the WASH compliance and survey of pathogens occurrence from random spots in the health facilities of two regions. This study is reporting on a baseline assessment conducted in health facilities of Bidre town in Bale zone from Oromia Region, Bulle town in Gedeo zone and Doyogena in Kembata-Tembaro zone both from SNNPs region. The WASH assessments included all health facilities that were functional at the time of the survey (i.e., health post, health centers, and hospitals). Sample size and sampling procedure To select appropriate sample size, the current study involved all hospitals from 3 districts and 30% of health facilities sample from the WASH program implementation of UNICEF-Ethiopia in 2 specific regions randomly. The study site was selected based on the list provided by UNICEF-Ethiopia and the possibility to transfer microbial samples in time (in 24 h) was considered. From the list of health facilities, we selected a sub-sample, stratified by type of facility. From the 31 health care facilities, we selected a subset of 12 health facilities from which sample was collected. We excluded pharmacies and clinics and focused on health posts, health centers and hospitals. Assessment of WASH in health facilities To assess the facilities of WASH and practices, the observational checklist of core questions for infection prevention and control (IPC) and WASH common indicators is developed based on international standards—WHO/UNICEF . All questionnaires and checklists were translated into Amharic/Oromifaa and were pretested prior to the interviews. The checklist allowed the collection of information on the prevailing sanitary conditions, access to water and hand-washing facility, as well as hand-washing and waste disposal practices. The WHO protocol on monitoring fulfilment of opportunities for hand-hygiene was used to assess the health personnel’s adherence to hand- hygiene guidelines from June 2021 to July 2021. Surface and water sample collection Sample collection was performed on August 2021 following the United States Center for Disease Control and Prevention (CDC) and Public health England guidelines . Surface and water sample primarily collected from hospitals and health centers. Surface sample collection was performed using sterile cotton swabs. The swabs were first moist in sterile normal saline solution. The samples were collected from surfaces including beds, door handles, walls, gowns, autoclaves, tables, and chairs. The sampling areas included out-patients departments, different wards, pharmacy, laboratories, receptions, toilets and cafeterias in the health facilities. Water samples were collected from sources from which the health facilities obtain water for washing, drinking and other activities in the healthcare settings. A total of 14 water sample is collected and delivered for analysis from delivery wards, medical ward, tanker, and bore-hole and rainwater collection systems. Overall, 59 water samples were collected from all health facilities, including health centers and health posts. Sample handling and transportation The collected surface samples were immediately put in Amies transport media and kept in pre-cooled ice box and transported to SNNPs region Public Health Institute laboratory. On arrival at the laboratory, the surface samples were transferred to the nutrient broth and enriched overnight at 37°C. After an overnight incubation, the samples were inoculated on blood agar and MacConkey agar plates and put overnight at 37°C. In case of no growth after an overnight culture, the plates were incubated for an additional 24 h. The water samples were assessed for their safety using modified Method 9,215 to enumerate heterotrophic bacteria and membrane filtration technique for Gram-negative bacteria . To enumerate heterotrophic bacteria, 1 mL of each water sample was pipetted into a sterile petri dish. After thoroughly mixing, the melted MacConkey agar was poured into the dish. The melted medium was mixed thoroughly with the sample and solidified. The plates were incubated for 48 h at 37°C. The Gram-negative bacteria were counted by filtrating 100 mL water samples through 0.45 μm pore size-47 mm, and cellulose nitrate membranes using the modified ISO 9308-1 protocol . The samples were incubated on MacConkey agar for 24 h at 37°C. All results of Gram-negative bacteria were expressed as colony forming units per 100 mL water. The bacterial colonies were collected and put in Trypticase Soy Broth containing 20% glycerol and were transported to the National bacteriology and mycology Reference Laboratory (NRL) at the Ethiopian Public Health Institute, where they were stored in deep-freeze until further analyses. Bacterial isolation and identification The bacteria were refreshed by culturing on three different culture media: (i) 5% sheep blood agar plate, (ii) MacConkey agar plate, and (iii) Mannitol salt agar plate. Colony appearance on culture plates, microscopic examination, and biochemical tests were used to identify Gram-positive and Gram-negative bacteria. 2.6.1 Identification of gram-positive cocci The common Gram-positive cocci are Staphylococcus spp. and Streptococcus spp. We used Blood agar and Mannitol salt agar media for isolation of Staphylococcus spp : The culture plates were incubated in air at 37°C for 24 h. Colony morphology on culture plates and microscopic examination for Gram-positive cocci in clusters were used for initial Staphylococcus spp. identification. Catalase and coagulase tests were used to classify Staphylococcus spp. into Staphylococcus aureus and coagulase negative Staphylococcus . All Staphylococci are Catalase positive and only S. aureus is coagulase positive. Streptococcus spp. were identified based on colony morphology on: (i) blood agar plates (beta hemolytic, alpha hemolytic and non-hemolytic), (ii) microscopic examination for Gram-positive in chin, and (iii) different biochemical tests. Negative catalase test differentiated Streptococcus spp. from Staphylococci Bacillus spp. Blood agar with 5% sheep blood media was used for the bacteria isolation. Colony morphology on the culture plates and gram stain were used for the bacterial identification. To differentiate Bacillus cereus from other Bacillus species we used citrate test which is only positive for B. cereus . 2.6.2 Identification of gram-negative bacilli The common gram negative bacteria are generally divided into two major categories: Fermenters and non-fermenters. Fermenters gram-negative bacilli utilize lactose and become pink color colonies on MacConkey agar while non-fermenters cannot utilize lactose and they are colorless colonies on MacConkey agar plate. Biochemical tests such as Triple Sugar Iron Agar (TSI), urea, citrate, Sulfide Indole Motility (SIM) medium, growth in Lysine Iron Agar (LIA), and oxidase were additionally used to identify Gram-negative bacteria. Identification of gram-positive cocci The common Gram-positive cocci are Staphylococcus spp. and Streptococcus spp. We used Blood agar and Mannitol salt agar media for isolation of Staphylococcus spp : The culture plates were incubated in air at 37°C for 24 h. Colony morphology on culture plates and microscopic examination for Gram-positive cocci in clusters were used for initial Staphylococcus spp. identification. Catalase and coagulase tests were used to classify Staphylococcus spp. into Staphylococcus aureus and coagulase negative Staphylococcus . All Staphylococci are Catalase positive and only S. aureus is coagulase positive. Streptococcus spp. were identified based on colony morphology on: (i) blood agar plates (beta hemolytic, alpha hemolytic and non-hemolytic), (ii) microscopic examination for Gram-positive in chin, and (iii) different biochemical tests. Negative catalase test differentiated Streptococcus spp. from Staphylococci Bacillus spp. Blood agar with 5% sheep blood media was used for the bacteria isolation. Colony morphology on the culture plates and gram stain were used for the bacterial identification. To differentiate Bacillus cereus from other Bacillus species we used citrate test which is only positive for B. cereus . Identification of gram-negative bacilli The common gram negative bacteria are generally divided into two major categories: Fermenters and non-fermenters. Fermenters gram-negative bacilli utilize lactose and become pink color colonies on MacConkey agar while non-fermenters cannot utilize lactose and they are colorless colonies on MacConkey agar plate. Biochemical tests such as Triple Sugar Iron Agar (TSI), urea, citrate, Sulfide Indole Motility (SIM) medium, growth in Lysine Iron Agar (LIA), and oxidase were additionally used to identify Gram-negative bacteria. Antimicrobial susceptibility testing The antimicrobial Susceptibility Tests (AST) were performed based on the Kirby–Bauer disk diffusion method on Mueller-Hinton agar (MHA) as recommended by clinical and laboratory standard Institute (CLSI) for all Gram-negative bacteria and Staphylococcus species . Well-isolated three to four colonies were emulsified in a tube containing sterile normal saline and the turbidity adjusted to 0.5 McFarland standards. The emulsified bacterial suspension was uniformly streaked on MHA plates using sterile cotton swabs, on which the antibiotic disks were applied and incubated for 18–24 h at 37°C. The antibiotic agents tested in this study were ampicillin (10 μg), amoxicillin-clavulanic acid (20/10 μg), pepracillin/ tazobactum, cefazolin (30 μg), cefuroxime (30 μg), cefotaxime (30 μg), ceftazidime (30 μg), cefepime (30 μg), cefoxitin (30 μg), ciprofloxacin (5 μg), amikacin (30 μg), meropenem (10 μg), chloramphenicol, tetracycline, cotrimoxazole, and penicillin. Penicillin and cefoxitin were tested only for Staphylococcus species and the result of oxacillin was determined from cefoxitin breakpoint. Antibiotic susceptibility results were interpreted according to the CLSI zone size interpretive standards . Intermediate results were considered resistant. Multidrug resistance (MDR) was defined according to guidelines compiled by the European Center for Disease prevention and Control (ECDC) and the Centers for Disease Control and Prevention (CDC) . Accordingly, bacterial isolates that were resistant to at least one agent in three different antimicrobial categories were considered as MDR. Quality assurance All media, biochemical reagents, gram stain reagents and antibiotic disks were checked for their quality using standards ATCC strains. Standard ATCC quality strains used for this study were S. aureus ATCC ® 25923, E. coli ATCC ® 2592, P. aeruginosa ATCC 27853. Data analysis Epi-info was used for data entry and the data was subsequently exported to Microsoft Excel and SPSS version 26 for data cleaning and further analysis. The frequencies of bacterial isolates and antimicrobial susceptibility were calculated. Mean and frequencies (percentage) were used to present descriptive data. Ethics Ethical clearance was obtained from the Institutional Review Board of the College of Natural and Computational Sciences of Addis Ababa University (Ref. No: IRB/04/14/2021). Additionally, the research was ethically approved by letter of support is sent to Oromia and SNNPs regional Health Bureaus with the letter of minute no. (ምሳኒፕ/453/13/21). The Oromia and SNPP’s Regional Health Bureau Ethics review committee also reviewed and approved the research for the implementation. Prior to the collection of data, the informed consent was obtained from staff and the administration of the each health facilities. Every task and procedures was completed in accordance with the WHO guidance, rule and regulations. Results WASH assessments were conducted in 26 health facilities in Bulle and Doyogena (SNNPs Region) and in Bidre (Oromia Region). The assessments included hospitals ( n = 3), health posts ( n = 13), clinics ( n = 8), and health centers ( n = 2; ). A great majority of the health facilities relied on tanker trucks for their water supply . At the time of the survey, piped water supply was available in only 11 of the 26 health facilities. Open pit latrines (14/26) were the commonest type of toilet and only in 8 out of the 26 facilities, the toilets were accessible for people with limited mobility. Infectious waste was primarily dumped into an open/protected pit, incinerated, and added to other wastes. Sharp waste was mostly collected for off-site disposal, autoclaved, or incinerated. Only 10 of the 26 assessed health facilities had guidelines on standard precautions for IPC. Only six had cleaning protocols available, and only in one health facility, the staff responsible for cleaning received training. Environmental disinfectant was only available in only 8 of the 26 health facilities. Hand-hygiene opportunities were directly observed (1,194 ± 326 min) and evaluated using the WHO checklist to assess compliance . Hand-hygiene opportunities were: (i) before touching a patient; (ii) before a procedure; (iii) after body fluid exposure/risk; (iv) after touching a patient; (v) after touching a patient’s surrounding. Hand-hygiene compliance was overall low, but varied by site. The lowest compliance was for Bidre (4%), followed by Doyogena (14%), and Bulle (36%). A total of 90 surface swabs and 14 water samples were collected from which a number of bacteria ( n = 224) were identified . Over 70% of the identified bacteria were from four categories: Staphylococcus spp., Bacillus spp., E. coli, and Klebsiella spp. These bacteria were the most widely distributed and were also found in high-risk locations including neonatal intensive care units, delivery and surgical rooms . More details on the identified bacteria by study sites, location and sample source can be found in the and . presents the antimicrobial resistance of the identified bacterial isolates. Antimicrobial susceptibility was detected in 50% or more of the isolates for penicillin, cefazolin, ampicillin, oxacillin, and cotrimoxazole. More than 50% of the isolates displayed multi-drug resistance, defined as resistant to at least one agent in three different antimicrobial categories. Discussion Water supply, availability of clean and accessible toilets, as well as infection prevention measures were found suboptimal. Hand hygiene practice by health workers was very low. Consequently, surface swabs and water samples revealed high bacterial contamination, with some of the identified bacteria known for their pathogenicity. These bacteria were also found in highly sensitive areas like surgical rooms, delivery rooms, and neonatal intensive care units (ICUs). Our findings highlight the need to invest in safely managed water supply, provision of safely managed sanitation services, but also strict hygiene and environmental cleaning in health facilities. Earlier studies assessing 1,318 health facilities in multiple African countries including Ethiopia showed that less than 50% of the facilities had access to improved water sources on premises, improved sanitation, and consistent access to water and soap for handwashing . A recent meta-analyses of studies on health workers’ handwashing practice in Ethiopia also estimated that 57.87% (95% CI: 44.14–71.61) practiced hand-washing , a figure that is higher than estimates from the current study. This difference may be explained by the rather rigorous evaluation of hand-hygiene practice in this study assessed using the more systematic WHO’s protocol of hand-hygiene opportunities. It can as well suggest that the selected sites have more significant WASH constraints, further justifying their selection for WASH and IPC improvements by the planned intervention. The poor WASH and IPC conditions observed in the health facilities can greatly impact the quality of the health care provided. First, satisfaction with WASH and IPC conditions can be associated with lower job satisfaction as reported from a recent multi-country study . Second, health facilities with suboptimal WASH and IPC procedures increase the risk for nosocomial infections. Indeed, a recent meta-analyses pooling results from 18 studies in Ethiopia , estimated the prevalence of nosocomial infections to be as high as 17% (95% CI 14.10–19.82). This prevalence can be even higher when considering vulnerable sub-groups like neonates, infants and young children. Indeed, studies have shown that HAIs contribute significantly to neonatal infections and mortality in low income countries like Ethiopia . A number of pathogenic bacteria associated with nosocomial infections have been identified from highly sensitive locations like surgical rooms, delivery room, and neonatal ICU . Klebsiella spp. , E. coli, Acinetobacter spp. , bacillus spp. and Staphylococcus spp. were identified in high number of samples collected from various locations. Poor hand-hygiene and bacterial contamination with AMR was a common feature of health facilities in all the three sites. More concerning is that a large number of the identified bacteria displayed antibiotic resistance and these same species were reported to be the major pathogens identified in bloodstream isolates ( n = 11,471) of hospital-acquired neonatal infections . A recent global study showed that most of the bacterial isolates identified in our study were responsible for high rates of deaths associated with AMR, particularly in sub-Saharan African (SSA) countries . This study might need further investigation of evidence with the recent finding of an estimation that Ethiopia used the lowest dose of antibiotics (28%) among central SSA (4.2 billion DDD, i.e., 42%) in 2018 . The present study has a number of limitations that need to be considered when interpreting our findings. First, this is a cross-sectional study and thus only provides a snapshot of the situation at the time of the survey. Second, the survey happened during the COVID-19 pandemic that in principle would have increased awareness on hand-hygiene because of the nation-wide campaigns. Third, this is a baseline assessment of health facilities selected for WASH/IPC intervention and thus may not be representative. However, evidence from our WASH data is in line with previous assessments and thus can be indicative of situations in similar settings in Ethiopia. Conclusion The health facilities assessed were confronted with serious problems related to WASH. Compliance to hand-hygiene practice by the health care workers was very low. Analyses of environmental and water samples revealed high levels of bacterial contamination. Most of the identified bacteria displayed AMR. Beyond increasing access to health coverage, emphasis should be put to improving infrastructure and services. This requires safely managed water supply, provision of safely managed sanitation services, but also strict hygiene and environmental cleaning in health facilities. Ensuring the supply chain of critical consumables such as soap, chlorine and decontaminants or disinfectants is key, but this will also need to be accompanied by behavioral change on hand hygiene and environmental cleaning practices. A critical element of strengthening health systems should also focus on antibiotic stewardship. The current study employed a cross-sectional study design to evaluate the present situation or existing facilities, and inadequacy of WASH in health settings and assess the risk factors focusing on antimicrobial resistance (AMR). The reason for the selection of the current study design, is suitable for capturing a snapshot of current WASH conditions and related health risks. However, the study design cannot establish a causal relationship between inadequate WASH facilities and heightened AMR infection rates. Therefore, we suggested that a longitudinal study design would have more comprehensive insights into and establish the long-term impacts of WASH improvements on the AMR, which offers evidence that is more robust over time. In addition, the need for continuous monitoring is required to understand how the WASH improvements might influence health outcomes. The recommendation emphasizes the need for urgent advocacy for policies requiring health facilities to adhere to WHO recommendations of WASH standards and improvement as well as infection control protocols. This might include the provision of safe water, sanitation, and adequate hygiene facilities for clients in order to reduce the risk of HAIs and AMR. The healthcare facilities, particularly the hospitals, should establish routine monitoring of NIs through established protocols and reporting of NIs to identify potential and critical hazards in infection rates over time. Additionally, promoting the regular training programs, particularly continuous professional developments (CPD) on infection control and AMR prevention for healthcare workers is crucial. |
The optional apex sites for quad zygomatic implant placement in edentulous patients with severe alveolar bone resorption: a CBCT anatomical analysis | 3f986ea6-c306-4e4b-9452-411778834b6b | 11569599 | Dentistry[mh] | Edentulism is often a critical yet challenging case in prosthodontics. Numerous clinical studies have confirmed that implant-supported prostheses can provide substantially improved oral function rehabilitation . For patients with severe alveolar bone resorption, zygomatic implants (ZIs) offer a reliable and effective solution for oral rehabilitation. Studies indicate that in such patients, the success rate of zygomatic implants ranges from 94.2–100% . The concept of zygomatic implants was introduced by Branemark in 1990 . Subsequently, numerous enhancements and innovations in ZI techniques have been developed . For patients with severe maxillary alveolar bone loss, a single ZI on each side may not provide adequate stability for full-arch implant restoration and could increase potential risks. The currently recommended approach is to place two ZIs on each side, i.e., quad zygomatic implants. This can offer ample support for full-arch prosthesis in edentulous patients . Quad ZIs occupy a significant portion of the zygomatic bone and require precise placement. It is crucial to evaluate the zygoma’s structure and determine the suitable apex point for ZI insertion before proceeding with quad zygomatic approach. The success and durability of ZIs hinge on the extent of bone-implant contact (BIC) within the zygoma. Moreover, the implant should not harm adjacent maxillofacial structures, thereby preventing intra- and post-operative complications. The zygoma’s inner aspect is close to the infratemporal fossa, home to vital neurovascular elements like the pterygoid plexus, maxillary artery and its offshoots, and the maxillary and mandibular nerve branches. Injury to these structures can result in severe complications, including deep hematomas, sensory and motor impairments, and potentially fatal outcomes. Therefore, careful placement of quad ZIs is imperative to avoid encroaching on the infratemporal fossa and avert serious postoperative complications. Moreover, orbital cavity penetration has been reported as a intra-operative complication with the incidence of 5.9% . As a typical and the most serious ZI therapy complication, orbital penetration by ZI could lead to severe pain in the region of the orbit, persistent anesthesia, physiological abduction and elevation of the involved eye , extraocular muscle injury, diplopia , eye movement limitation , etc. Intranasinus technique may increase the risk of orbit invasion due to lack of vision and control of the drills during implant bed preparation . The alveolar ridge is categorized into six stages of resorption severity according to the Cawood and Howell classification . For patients exhibiting varying degrees of alveolar bone loss, the location for quad zygomatic implants may vary. Currently, few research has been conducted to compare the choices of implant sites among patients with different levels of alveolar bone resorption. Two studies assessed the zygomatic BIC of quad ZIs using Cone Beam Computed Tomography (CBCT) images . Hung et al. evaluated the zygomatic BIC at various implant sites and its association with the infratemporal fossa, proposing the most suitable implant location in the zygoma . This research focused on an Asian demographic. Conversely, Bertos et al. conducted their research on a European cohort, examining the influence of alveolar bone resorption on the BIC of quad ZIs, the volume of implant-engaged zygoma bone, and its correlation with the maxillary sinus. Nevertheless, there is still a lacking of research on how alveolar bone resorption affects the BIC and its relationship with the infratemporal fossa at individual ZI sites. This study aims to investigate the difference in BIC, the infratemporal fossa intrusion, and the relationship between ZIs and orbit among differenct apex sites for quad ZIs placement in edentulous patients with varying degrees of alveolar bone loss.
Patient selection This study received approval from the Institutional Review Board of the University (Ethical Approval No: PKUSSIRB-202162013). The clinical research is registered under the number ChiCTR2100044472 (18/03/2021). Edentulous patients in need of implant treatment who underwent CBCT scans at the University dental school clinic between October 2019 and August 2021, and who met the inclusion/exclusion criteria, were included in the study. All participants read and signed informed consents. Inclusion criteria: (1) Patient age over 18 years old; (2) Maxillary edentulism for a minimum of 3 months; (3) Require implant restoration, and undergo a CBCT examination; (4) Have read and signed the informed consent form; (5) Alveolar bone resorption classified as either Class IV or Class V/VI according to the Cawood & Howell classification. Exclusion criteria : (1) Anatomical abnormalities in the maxilla or zygoma; (2) Severe facial asymmetry; (3) Alveolar bone resorption classified from level I to III according to the Cawood & Howell classification; (4) Edentulism due to maxillofacial trauma or tumor resection surgery. Classification of alveolar bone resorption level The edentulous patients’ alveolar bone resorption was categorized using the Cawood & Howell classification as per Bertos et al. . A knife-edge ridge form of the residual ridge, insufficient in width and exceeding 5 mm in height, was classified as Class IV alveolar bone resorption. Conversely, a flat residual ridge, insufficient in width and less than 5 mm in height, was classified as Class V/VI alveolar bone resorption. The level of alveolar bone resorption was assessed individually on each side, utilizing both two-dimensional and three-dimensional CBCT reconstructed images. Virtual implant planning CBCT data was exported in the Digital Imaging and Communications in Medicine (DICOM) format and imported into the planning software (Nobel Clinician 2.10.1.3, Nobel Biocare). For each side of the maxilla and zygoma, two zygomatic implants (Branemark System Zygoma Tiunite RP) were virtually planned in accordance with protocols from previously published articles , namely one anterior zygomatic implant and one posterior implant, totaling four implants per case (Fig. ). Implant site selection Entery point on the alveolar ridge Points E1 and E2 are designated on the alveolar ridge (Fig. ). Point E1 is located 5 mm palatally from where the lowest point of the maxillary alveolar ridge intersects with a perpendicular line extending from the lateral margin of the nasal incisure . When viewing the maxilla from above, point E2 is situated 5 mm palatally from where The lowermost point of the alveolar crest was identified by taking a line at a tangent to the lateral margin of the infraorbital foramen. The degree of alveolar bone resorption is assessed individually at points E1 and E2 . Apex point in the zygoma Draw a horizontal line, IM, at the lower orbital margin and a vertical line, LM, parallel to the median plane, intersecting the lateral orbital margin. At their intersection lies point C. Construct an angular bisector between IM and LM to locate point O at their intersection with the orbital margin. Connect points C and O with line L1 (Fig. ). Shift L1 medially by 5 mm to create a parallel line L0; shift L1 laterally by 5 mm and 10 mm to form parallel lines L2 and L3, respectively. These lines extend from the orbital margin to the zygoma’s lower edge. The apex of the quad ZIs is positioned in the zygomatic bone between L0 and L3. Segment L0 and L3 into four equal parts to identify quarter points A0, B0, C0, and A3, B3, C3, from superior to inferior. Draw lines LA, LB, and LC by connecting A0 to A3, B0 to B3, and C0 to C3. The intersection points of LA, LB, LC with L1, L2 are labeled A1, B1, C1 and A2, B2, C2, respectively. The areas that the lines LA, LB, LC pass through is namely the upper, middle and lower section of the zygomatic. The 12 points thus identified represent the center of the 12 zygomatic segments . The ZI apex will be positioned at these 12 points as indicated in Fig. . According to the previous studies of Hung and colleagues , the apex of the anterior ZI is positioned at the upper section (points A series), with its entry at point E1. The apex of the posterior ZI is positioned at the middle and lower sections (points B and C series), with its entry at point E2. Measurements and data collection The patient’s CBCT was imported into the planning software (NobelClinician 2.10.1.3, Nobel Biocare, Zurich) for virtual planning of the quad ZI site. Implant simulations (Branemark System Zygoma Tiunite RP) were virtually designed. Measurements of the quad ZIs at 12 apex locations were taken, with the following detailed metrics: Measurements of Bone-Implant contact (BIC) This is a linear measurement protocol. The BIC was the average value of zygomatic BIC length on the facial and the temporal sides in the facial -temporal cross section (Fig. ) . The zygomatic bone-implant contact (zBIC) and alveolar bone-implant contact (aBIC) were assessed. The average contact length between the ZI and the zygomatic or alveolar bone on both the facial and temporal sides was calculated as zBIC or aBIC, respectively. The total bone-implant contact (tBIC) represents the combined measurement of zBIC and aBIC. Evaluation of ZI Intrusion The occurrence of ZI intrusion into the infratemporal fossa was examined, along with the depth and length of such intrusions (Fig. ). Meausrement of the distance between the orbit and the anterior implant The closest distance between the implant and patient’s orbit cavity was measured for anterior ZI at each apex point. Determination of Optional Zygomatic Implant position The possible location for ZI placement within the zygoma was identified by analyzing the implant’s BIC and the frequency of infratemporal fossa intrusion and orbital penetration. This process was then applied to determine the implant sites for patients with alveolar bone resorption Class IV and V/VI. The findings from these two subgroups were compared to the overall participant data to identify any differences. Statistical analysis All analyses were performed with SPSS (SPSS Statistics 27.0, IBM) as follows: A one way ANOVA was used to compare the zygomatic BIC (zBIC) and alveolar BIC (aBIC) among ZIs at various apex locations to determine the most suitable insertion points. Subsequently, BIC was assessed within each subgroup based on the degree of alveolar bone resorption, followed by a Chi-square test to compare BIC across subgroups and to ascertain the possible apex design for zygomatic implants. A Chi-square test was employed to evaluate the frequency of ZI penetration into the infratemporal fossa among subgroups and the entire cohort. Additionally, the extent and magnitude of ZI intrusion were compared using the same statistical method. When P is less than 0.05, it is considered that there is a significant difference.
This study received approval from the Institutional Review Board of the University (Ethical Approval No: PKUSSIRB-202162013). The clinical research is registered under the number ChiCTR2100044472 (18/03/2021). Edentulous patients in need of implant treatment who underwent CBCT scans at the University dental school clinic between October 2019 and August 2021, and who met the inclusion/exclusion criteria, were included in the study. All participants read and signed informed consents. Inclusion criteria: (1) Patient age over 18 years old; (2) Maxillary edentulism for a minimum of 3 months; (3) Require implant restoration, and undergo a CBCT examination; (4) Have read and signed the informed consent form; (5) Alveolar bone resorption classified as either Class IV or Class V/VI according to the Cawood & Howell classification. Exclusion criteria : (1) Anatomical abnormalities in the maxilla or zygoma; (2) Severe facial asymmetry; (3) Alveolar bone resorption classified from level I to III according to the Cawood & Howell classification; (4) Edentulism due to maxillofacial trauma or tumor resection surgery.
The edentulous patients’ alveolar bone resorption was categorized using the Cawood & Howell classification as per Bertos et al. . A knife-edge ridge form of the residual ridge, insufficient in width and exceeding 5 mm in height, was classified as Class IV alveolar bone resorption. Conversely, a flat residual ridge, insufficient in width and less than 5 mm in height, was classified as Class V/VI alveolar bone resorption. The level of alveolar bone resorption was assessed individually on each side, utilizing both two-dimensional and three-dimensional CBCT reconstructed images.
CBCT data was exported in the Digital Imaging and Communications in Medicine (DICOM) format and imported into the planning software (Nobel Clinician 2.10.1.3, Nobel Biocare). For each side of the maxilla and zygoma, two zygomatic implants (Branemark System Zygoma Tiunite RP) were virtually planned in accordance with protocols from previously published articles , namely one anterior zygomatic implant and one posterior implant, totaling four implants per case (Fig. ).
Entery point on the alveolar ridge Points E1 and E2 are designated on the alveolar ridge (Fig. ). Point E1 is located 5 mm palatally from where the lowest point of the maxillary alveolar ridge intersects with a perpendicular line extending from the lateral margin of the nasal incisure . When viewing the maxilla from above, point E2 is situated 5 mm palatally from where The lowermost point of the alveolar crest was identified by taking a line at a tangent to the lateral margin of the infraorbital foramen. The degree of alveolar bone resorption is assessed individually at points E1 and E2 . Apex point in the zygoma Draw a horizontal line, IM, at the lower orbital margin and a vertical line, LM, parallel to the median plane, intersecting the lateral orbital margin. At their intersection lies point C. Construct an angular bisector between IM and LM to locate point O at their intersection with the orbital margin. Connect points C and O with line L1 (Fig. ). Shift L1 medially by 5 mm to create a parallel line L0; shift L1 laterally by 5 mm and 10 mm to form parallel lines L2 and L3, respectively. These lines extend from the orbital margin to the zygoma’s lower edge. The apex of the quad ZIs is positioned in the zygomatic bone between L0 and L3. Segment L0 and L3 into four equal parts to identify quarter points A0, B0, C0, and A3, B3, C3, from superior to inferior. Draw lines LA, LB, and LC by connecting A0 to A3, B0 to B3, and C0 to C3. The intersection points of LA, LB, LC with L1, L2 are labeled A1, B1, C1 and A2, B2, C2, respectively. The areas that the lines LA, LB, LC pass through is namely the upper, middle and lower section of the zygomatic. The 12 points thus identified represent the center of the 12 zygomatic segments . The ZI apex will be positioned at these 12 points as indicated in Fig. . According to the previous studies of Hung and colleagues , the apex of the anterior ZI is positioned at the upper section (points A series), with its entry at point E1. The apex of the posterior ZI is positioned at the middle and lower sections (points B and C series), with its entry at point E2.
Points E1 and E2 are designated on the alveolar ridge (Fig. ). Point E1 is located 5 mm palatally from where the lowest point of the maxillary alveolar ridge intersects with a perpendicular line extending from the lateral margin of the nasal incisure . When viewing the maxilla from above, point E2 is situated 5 mm palatally from where The lowermost point of the alveolar crest was identified by taking a line at a tangent to the lateral margin of the infraorbital foramen. The degree of alveolar bone resorption is assessed individually at points E1 and E2 .
Draw a horizontal line, IM, at the lower orbital margin and a vertical line, LM, parallel to the median plane, intersecting the lateral orbital margin. At their intersection lies point C. Construct an angular bisector between IM and LM to locate point O at their intersection with the orbital margin. Connect points C and O with line L1 (Fig. ). Shift L1 medially by 5 mm to create a parallel line L0; shift L1 laterally by 5 mm and 10 mm to form parallel lines L2 and L3, respectively. These lines extend from the orbital margin to the zygoma’s lower edge. The apex of the quad ZIs is positioned in the zygomatic bone between L0 and L3. Segment L0 and L3 into four equal parts to identify quarter points A0, B0, C0, and A3, B3, C3, from superior to inferior. Draw lines LA, LB, and LC by connecting A0 to A3, B0 to B3, and C0 to C3. The intersection points of LA, LB, LC with L1, L2 are labeled A1, B1, C1 and A2, B2, C2, respectively. The areas that the lines LA, LB, LC pass through is namely the upper, middle and lower section of the zygomatic. The 12 points thus identified represent the center of the 12 zygomatic segments . The ZI apex will be positioned at these 12 points as indicated in Fig. . According to the previous studies of Hung and colleagues , the apex of the anterior ZI is positioned at the upper section (points A series), with its entry at point E1. The apex of the posterior ZI is positioned at the middle and lower sections (points B and C series), with its entry at point E2.
The patient’s CBCT was imported into the planning software (NobelClinician 2.10.1.3, Nobel Biocare, Zurich) for virtual planning of the quad ZI site. Implant simulations (Branemark System Zygoma Tiunite RP) were virtually designed. Measurements of the quad ZIs at 12 apex locations were taken, with the following detailed metrics: Measurements of Bone-Implant contact (BIC) This is a linear measurement protocol. The BIC was the average value of zygomatic BIC length on the facial and the temporal sides in the facial -temporal cross section (Fig. ) . The zygomatic bone-implant contact (zBIC) and alveolar bone-implant contact (aBIC) were assessed. The average contact length between the ZI and the zygomatic or alveolar bone on both the facial and temporal sides was calculated as zBIC or aBIC, respectively. The total bone-implant contact (tBIC) represents the combined measurement of zBIC and aBIC. Evaluation of ZI Intrusion The occurrence of ZI intrusion into the infratemporal fossa was examined, along with the depth and length of such intrusions (Fig. ). Meausrement of the distance between the orbit and the anterior implant The closest distance between the implant and patient’s orbit cavity was measured for anterior ZI at each apex point. Determination of Optional Zygomatic Implant position The possible location for ZI placement within the zygoma was identified by analyzing the implant’s BIC and the frequency of infratemporal fossa intrusion and orbital penetration. This process was then applied to determine the implant sites for patients with alveolar bone resorption Class IV and V/VI. The findings from these two subgroups were compared to the overall participant data to identify any differences.
This is a linear measurement protocol. The BIC was the average value of zygomatic BIC length on the facial and the temporal sides in the facial -temporal cross section (Fig. ) . The zygomatic bone-implant contact (zBIC) and alveolar bone-implant contact (aBIC) were assessed. The average contact length between the ZI and the zygomatic or alveolar bone on both the facial and temporal sides was calculated as zBIC or aBIC, respectively. The total bone-implant contact (tBIC) represents the combined measurement of zBIC and aBIC.
The occurrence of ZI intrusion into the infratemporal fossa was examined, along with the depth and length of such intrusions (Fig. ).
The closest distance between the implant and patient’s orbit cavity was measured for anterior ZI at each apex point.
The possible location for ZI placement within the zygoma was identified by analyzing the implant’s BIC and the frequency of infratemporal fossa intrusion and orbital penetration. This process was then applied to determine the implant sites for patients with alveolar bone resorption Class IV and V/VI. The findings from these two subgroups were compared to the overall participant data to identify any differences.
All analyses were performed with SPSS (SPSS Statistics 27.0, IBM) as follows: A one way ANOVA was used to compare the zygomatic BIC (zBIC) and alveolar BIC (aBIC) among ZIs at various apex locations to determine the most suitable insertion points. Subsequently, BIC was assessed within each subgroup based on the degree of alveolar bone resorption, followed by a Chi-square test to compare BIC across subgroups and to ascertain the possible apex design for zygomatic implants. A Chi-square test was employed to evaluate the frequency of ZI penetration into the infratemporal fossa among subgroups and the entire cohort. Additionally, the extent and magnitude of ZI intrusion were compared using the same statistical method. When P is less than 0.05, it is considered that there is a significant difference.
Demographics Maxillary edentulous patients who visited the University dental school clinic from March 2021 to October 2022 for implant-supported prosthesis treatment and underwent CBCT were evaluated. A total of 48 patients were collected. The edentulous arches were classified using the Cawood and Howell classification. Of these patients, eleven were excluded because their alveolar bone classification was categorized as level III. Nine patients were excluded due to inadequate CBCT images that precluded the virtual design of zygomatic implants. Ultimately, 28 participants were included in the study, comprising 12 males and 16 females, with an average age of 63.8 ± 12.6 years. The 28 edentulous maxillae were divided into 56 hemi-maxillae, with 39 of the 56 edentulous posterior residual ridges classified as Class IV and the remaining 17 as Class V/VI. In total, 112 implants were virtually planned. BIC of zygomatic implants BIC of the anterior zygomatic implant The apex points of the anterior ZIs were at A0, A1, A2, and A3. The total BIC (tBIC) and the zygomatic BIC (zBIC) of implants decreased significantly from A3 to A0 ( P < 0.01, Table ) while the alveolar BIC (aBIC) showed no significant change ( P = 0.769). The average zBIC at A3, A2, A1, and A0 was 18.3 ± 3.9 mm, 13.4 ± 3.7 mm, 7.4 ± 2.8 mm, and 4.2 ± 1.7 mm respectively. In the Class IV subgroup, tBIC and zBIC decreased from A3 to A0, with a significant difference observed among ZIs at all apex points ( P < 0.01), However, no significant difference was found in aBIC among implants at any apex points ( P = 0.849). In the Class V/VI subgroups, tBIC and zBIC also decreased from A3 to A0. A significant difference in zBIC was noted among implants at all apex points ( P < 0.01), while no significant difference was observed in tBIC between implants at A3 and A2 ( P = 0.217), or between those at A1 and A0 ( P = 0.132). The tBIC of implants at A3 and A2 was significantly higher than that of implants at A1 and A0 ( P < 0.05). No significant difference in aBIC was found among implants at any apex points ( P = 0.939). For implants at the four apex points of the upper zygoma, no significant difference in tBIC, zBIC, or aBIC was detected between the two subgroups ( P > 0.05). BIC of the posterior zygomatic implant The apex points of the posterior implants comprised eight locations: B0, B1, B2, and B3 at the middle zygoma, and C0, C1, C2, and C3 at the lower zygoma. Within the B series, the highest zBIC and tBIC were recorded at the B2 point (16.3 ± 5.3 mm; 22.1 ± 7.5 mm), while the C series showed the highest values at the C1 point (13.8 ± 5.0 mm; 19.5 ± 5.5 mm) as indicated in Table . The B2 point demonstrated the highest zBIC and tBIC among all eight apex sites, with a significant difference from the other points. No significant difference in zBIC and tBIC were observed among the B1, B3, C1, and C2 points (zBIC at B3: 13.4 ± 4.8 mm, at B1: 13.5 ± 5.5 mm, at C1: 13.8 ± 5.0 mm, at C2: 12.3 ± 5.8 mm), though these were significantly higher than those at the B0, C0, and C3 points (zBIC at B0: 8.9 ± 4.9 mm, at C0: 10.9 ± 4.1 mm, at C3: 8.5 ± 4.5 mm). Across all apex points, aBIC did not show significant variation ( P > 0.05) (Table ). In the Class IV and Class V/VI subgroups, the highest BIC was noted at the B2 point in the middle zygoma and at the C1 point in the lower zygoma. For Class IV subgroup, zBIC and tBIC at the B2 point were significantly higher compared to other apex points. Between the C1 and B1 apex points, no significant difference in zBIC and tBIC was detected. In the Class V/VI subgroup, zBIC and tBIC did not significantly differ between the B2 and B1 points, but a significant difference was found between the B2 point and other points, excluding B1 (Table ). No significant difference was found in aBIC and tBIC of implants at the same apex points between Class IV and V/VI subgroups. Relationship between zygomatic implants and the Infratemporal Fossa (Table ) Anterior zygomatic implants and infratemporal fossa For the anterior zygomatic implants, only one of the 56 implants (1.8%) at the A3 point intruded into the infratemporal fossa. The depth of the intrusion was 1.2 mm. No other anterior implants showed entry into the fossa. There was no significant difference in the rate of ZI intrusion among the four apex points (A series points) ( P = 0.390). The single implant that intruded into the infratemporal fossa was from the subgroup of class IV, although there was no significant difference in the rate of anterior ZI intrusion between the two subgroups. Posterior zygomatic implants and infratemporal fossa For the B series apex points (mid-zygoma), no implant intrusion into the infratemporal fossa was observed at B0. However, intrusion rates at B1, B2, and B3 were 1.8%, 42.9%, and 92.9%, respectively, with the average depth and length of intrusion into the infratemporal fossa increasing progressively. The average intrusion depth was 2.4 mm, 2.2 ± 1.0 mm, and 3.9 ± 1.6 mm at B1, B2, and B3 respectively. Significant differences in intrusion rates were noted between B3 and B0, and B2 and B0, but not between B1 and B0. For the C series apex points (lower zygoma), varying rates of ZI intrusion into the infratemporal fossa were recorded across the four sites. From C0 to C3, the intrusion rate, depth and length of ZI exposure in the infratemporal fossa increased (Table ). At the C3 apex point, all implants intruded into the infratemporal fossa, with an average intrusion depth of 7.8 ± 2.6 mm. Significant differences in intrusion rates were present among all C series sites. In Class IV subgroup, no ZI intrusion into infratemporal fossa occurred at B0 and B1 site, while 35.9% of the ZI at B2 site and 89.7% of the ZI at B3 site showed intrusion, with intrusion depth of 2.4 ± 1.0 mm (B2) and 3.9 ± 1.7 mm (B3). For the C series apex points, 15.4% of ZI at C0, 56.4% of ZI at C1 site, and over 90% of ZI at C2 and C3 sites entered the infratemporal fossa (Table ), with intrusion depths increasing from C0 to C3. Within the Class V/VI subgroup, there was no ZI intrusion into the infratemporal fossa at the B0 site, and only one ZI at the B1 site (5.9%) penetrated the infratemporal fossa with an intrusion depth of 2.4 mm. In contrast, 58.8% of ZIs at the B2 site and all ZIs at the B3 site (100%) intruded into the infratemporal fossa (Table ). Regarding the C series sites, 23.5% of ZIs at the C0 site, 64.7% of ZIs at the C1 site, and over 90% of ZIs at both the C2 and C3 sites entered the infratemporal fossa (Table ). Although the exact values for ZI intrusion rate, as well as the average depth and length of intrusion, were higher in the Class V/VI subgroup compared to Class IVsubgroup, there was no statistically significant difference in the rate, depth, or length of ZI intrusion at any of the middle and lower zygoma apex sites between the two subgroups ( P > 0.05). The risk of orbital cavity penetration No orbital cavity penetration was detected in all the cases at all apex points. The distance between the anterior ZI and the orbital cavity was shown in Table . The average distance between the anterior ZI and the orbit was 2.5 ± 1.0 mm, 3.2 ± 1.0 mm, 3.8 ± 1.0 mm, and 4.3 ± 0.9 mm at A3, A2, A1, and A0 apex points respectively. There was significant difference in ZI-orbital distance among the four apex groups ( P < 0.001). No significant difference in ZI-orbital distance was found between the Class IV and Class V/VI subgroups at all the A series points. Three anterior ZI at A3 point showed less than 1 mm distance from the orbital cavity.
Maxillary edentulous patients who visited the University dental school clinic from March 2021 to October 2022 for implant-supported prosthesis treatment and underwent CBCT were evaluated. A total of 48 patients were collected. The edentulous arches were classified using the Cawood and Howell classification. Of these patients, eleven were excluded because their alveolar bone classification was categorized as level III. Nine patients were excluded due to inadequate CBCT images that precluded the virtual design of zygomatic implants. Ultimately, 28 participants were included in the study, comprising 12 males and 16 females, with an average age of 63.8 ± 12.6 years. The 28 edentulous maxillae were divided into 56 hemi-maxillae, with 39 of the 56 edentulous posterior residual ridges classified as Class IV and the remaining 17 as Class V/VI. In total, 112 implants were virtually planned.
BIC of the anterior zygomatic implant The apex points of the anterior ZIs were at A0, A1, A2, and A3. The total BIC (tBIC) and the zygomatic BIC (zBIC) of implants decreased significantly from A3 to A0 ( P < 0.01, Table ) while the alveolar BIC (aBIC) showed no significant change ( P = 0.769). The average zBIC at A3, A2, A1, and A0 was 18.3 ± 3.9 mm, 13.4 ± 3.7 mm, 7.4 ± 2.8 mm, and 4.2 ± 1.7 mm respectively. In the Class IV subgroup, tBIC and zBIC decreased from A3 to A0, with a significant difference observed among ZIs at all apex points ( P < 0.01), However, no significant difference was found in aBIC among implants at any apex points ( P = 0.849). In the Class V/VI subgroups, tBIC and zBIC also decreased from A3 to A0. A significant difference in zBIC was noted among implants at all apex points ( P < 0.01), while no significant difference was observed in tBIC between implants at A3 and A2 ( P = 0.217), or between those at A1 and A0 ( P = 0.132). The tBIC of implants at A3 and A2 was significantly higher than that of implants at A1 and A0 ( P < 0.05). No significant difference in aBIC was found among implants at any apex points ( P = 0.939). For implants at the four apex points of the upper zygoma, no significant difference in tBIC, zBIC, or aBIC was detected between the two subgroups ( P > 0.05). BIC of the posterior zygomatic implant The apex points of the posterior implants comprised eight locations: B0, B1, B2, and B3 at the middle zygoma, and C0, C1, C2, and C3 at the lower zygoma. Within the B series, the highest zBIC and tBIC were recorded at the B2 point (16.3 ± 5.3 mm; 22.1 ± 7.5 mm), while the C series showed the highest values at the C1 point (13.8 ± 5.0 mm; 19.5 ± 5.5 mm) as indicated in Table . The B2 point demonstrated the highest zBIC and tBIC among all eight apex sites, with a significant difference from the other points. No significant difference in zBIC and tBIC were observed among the B1, B3, C1, and C2 points (zBIC at B3: 13.4 ± 4.8 mm, at B1: 13.5 ± 5.5 mm, at C1: 13.8 ± 5.0 mm, at C2: 12.3 ± 5.8 mm), though these were significantly higher than those at the B0, C0, and C3 points (zBIC at B0: 8.9 ± 4.9 mm, at C0: 10.9 ± 4.1 mm, at C3: 8.5 ± 4.5 mm). Across all apex points, aBIC did not show significant variation ( P > 0.05) (Table ). In the Class IV and Class V/VI subgroups, the highest BIC was noted at the B2 point in the middle zygoma and at the C1 point in the lower zygoma. For Class IV subgroup, zBIC and tBIC at the B2 point were significantly higher compared to other apex points. Between the C1 and B1 apex points, no significant difference in zBIC and tBIC was detected. In the Class V/VI subgroup, zBIC and tBIC did not significantly differ between the B2 and B1 points, but a significant difference was found between the B2 point and other points, excluding B1 (Table ). No significant difference was found in aBIC and tBIC of implants at the same apex points between Class IV and V/VI subgroups. Relationship between zygomatic implants and the Infratemporal Fossa (Table ) Anterior zygomatic implants and infratemporal fossa For the anterior zygomatic implants, only one of the 56 implants (1.8%) at the A3 point intruded into the infratemporal fossa. The depth of the intrusion was 1.2 mm. No other anterior implants showed entry into the fossa. There was no significant difference in the rate of ZI intrusion among the four apex points (A series points) ( P = 0.390). The single implant that intruded into the infratemporal fossa was from the subgroup of class IV, although there was no significant difference in the rate of anterior ZI intrusion between the two subgroups. Posterior zygomatic implants and infratemporal fossa For the B series apex points (mid-zygoma), no implant intrusion into the infratemporal fossa was observed at B0. However, intrusion rates at B1, B2, and B3 were 1.8%, 42.9%, and 92.9%, respectively, with the average depth and length of intrusion into the infratemporal fossa increasing progressively. The average intrusion depth was 2.4 mm, 2.2 ± 1.0 mm, and 3.9 ± 1.6 mm at B1, B2, and B3 respectively. Significant differences in intrusion rates were noted between B3 and B0, and B2 and B0, but not between B1 and B0. For the C series apex points (lower zygoma), varying rates of ZI intrusion into the infratemporal fossa were recorded across the four sites. From C0 to C3, the intrusion rate, depth and length of ZI exposure in the infratemporal fossa increased (Table ). At the C3 apex point, all implants intruded into the infratemporal fossa, with an average intrusion depth of 7.8 ± 2.6 mm. Significant differences in intrusion rates were present among all C series sites. In Class IV subgroup, no ZI intrusion into infratemporal fossa occurred at B0 and B1 site, while 35.9% of the ZI at B2 site and 89.7% of the ZI at B3 site showed intrusion, with intrusion depth of 2.4 ± 1.0 mm (B2) and 3.9 ± 1.7 mm (B3). For the C series apex points, 15.4% of ZI at C0, 56.4% of ZI at C1 site, and over 90% of ZI at C2 and C3 sites entered the infratemporal fossa (Table ), with intrusion depths increasing from C0 to C3. Within the Class V/VI subgroup, there was no ZI intrusion into the infratemporal fossa at the B0 site, and only one ZI at the B1 site (5.9%) penetrated the infratemporal fossa with an intrusion depth of 2.4 mm. In contrast, 58.8% of ZIs at the B2 site and all ZIs at the B3 site (100%) intruded into the infratemporal fossa (Table ). Regarding the C series sites, 23.5% of ZIs at the C0 site, 64.7% of ZIs at the C1 site, and over 90% of ZIs at both the C2 and C3 sites entered the infratemporal fossa (Table ). Although the exact values for ZI intrusion rate, as well as the average depth and length of intrusion, were higher in the Class V/VI subgroup compared to Class IVsubgroup, there was no statistically significant difference in the rate, depth, or length of ZI intrusion at any of the middle and lower zygoma apex sites between the two subgroups ( P > 0.05). The risk of orbital cavity penetration No orbital cavity penetration was detected in all the cases at all apex points. The distance between the anterior ZI and the orbital cavity was shown in Table . The average distance between the anterior ZI and the orbit was 2.5 ± 1.0 mm, 3.2 ± 1.0 mm, 3.8 ± 1.0 mm, and 4.3 ± 0.9 mm at A3, A2, A1, and A0 apex points respectively. There was significant difference in ZI-orbital distance among the four apex groups ( P < 0.001). No significant difference in ZI-orbital distance was found between the Class IV and Class V/VI subgroups at all the A series points. Three anterior ZI at A3 point showed less than 1 mm distance from the orbital cavity.
The apex points of the anterior ZIs were at A0, A1, A2, and A3. The total BIC (tBIC) and the zygomatic BIC (zBIC) of implants decreased significantly from A3 to A0 ( P < 0.01, Table ) while the alveolar BIC (aBIC) showed no significant change ( P = 0.769). The average zBIC at A3, A2, A1, and A0 was 18.3 ± 3.9 mm, 13.4 ± 3.7 mm, 7.4 ± 2.8 mm, and 4.2 ± 1.7 mm respectively. In the Class IV subgroup, tBIC and zBIC decreased from A3 to A0, with a significant difference observed among ZIs at all apex points ( P < 0.01), However, no significant difference was found in aBIC among implants at any apex points ( P = 0.849). In the Class V/VI subgroups, tBIC and zBIC also decreased from A3 to A0. A significant difference in zBIC was noted among implants at all apex points ( P < 0.01), while no significant difference was observed in tBIC between implants at A3 and A2 ( P = 0.217), or between those at A1 and A0 ( P = 0.132). The tBIC of implants at A3 and A2 was significantly higher than that of implants at A1 and A0 ( P < 0.05). No significant difference in aBIC was found among implants at any apex points ( P = 0.939). For implants at the four apex points of the upper zygoma, no significant difference in tBIC, zBIC, or aBIC was detected between the two subgroups ( P > 0.05).
The apex points of the posterior implants comprised eight locations: B0, B1, B2, and B3 at the middle zygoma, and C0, C1, C2, and C3 at the lower zygoma. Within the B series, the highest zBIC and tBIC were recorded at the B2 point (16.3 ± 5.3 mm; 22.1 ± 7.5 mm), while the C series showed the highest values at the C1 point (13.8 ± 5.0 mm; 19.5 ± 5.5 mm) as indicated in Table . The B2 point demonstrated the highest zBIC and tBIC among all eight apex sites, with a significant difference from the other points. No significant difference in zBIC and tBIC were observed among the B1, B3, C1, and C2 points (zBIC at B3: 13.4 ± 4.8 mm, at B1: 13.5 ± 5.5 mm, at C1: 13.8 ± 5.0 mm, at C2: 12.3 ± 5.8 mm), though these were significantly higher than those at the B0, C0, and C3 points (zBIC at B0: 8.9 ± 4.9 mm, at C0: 10.9 ± 4.1 mm, at C3: 8.5 ± 4.5 mm). Across all apex points, aBIC did not show significant variation ( P > 0.05) (Table ). In the Class IV and Class V/VI subgroups, the highest BIC was noted at the B2 point in the middle zygoma and at the C1 point in the lower zygoma. For Class IV subgroup, zBIC and tBIC at the B2 point were significantly higher compared to other apex points. Between the C1 and B1 apex points, no significant difference in zBIC and tBIC was detected. In the Class V/VI subgroup, zBIC and tBIC did not significantly differ between the B2 and B1 points, but a significant difference was found between the B2 point and other points, excluding B1 (Table ). No significant difference was found in aBIC and tBIC of implants at the same apex points between Class IV and V/VI subgroups.
) Anterior zygomatic implants and infratemporal fossa For the anterior zygomatic implants, only one of the 56 implants (1.8%) at the A3 point intruded into the infratemporal fossa. The depth of the intrusion was 1.2 mm. No other anterior implants showed entry into the fossa. There was no significant difference in the rate of ZI intrusion among the four apex points (A series points) ( P = 0.390). The single implant that intruded into the infratemporal fossa was from the subgroup of class IV, although there was no significant difference in the rate of anterior ZI intrusion between the two subgroups. Posterior zygomatic implants and infratemporal fossa For the B series apex points (mid-zygoma), no implant intrusion into the infratemporal fossa was observed at B0. However, intrusion rates at B1, B2, and B3 were 1.8%, 42.9%, and 92.9%, respectively, with the average depth and length of intrusion into the infratemporal fossa increasing progressively. The average intrusion depth was 2.4 mm, 2.2 ± 1.0 mm, and 3.9 ± 1.6 mm at B1, B2, and B3 respectively. Significant differences in intrusion rates were noted between B3 and B0, and B2 and B0, but not between B1 and B0. For the C series apex points (lower zygoma), varying rates of ZI intrusion into the infratemporal fossa were recorded across the four sites. From C0 to C3, the intrusion rate, depth and length of ZI exposure in the infratemporal fossa increased (Table ). At the C3 apex point, all implants intruded into the infratemporal fossa, with an average intrusion depth of 7.8 ± 2.6 mm. Significant differences in intrusion rates were present among all C series sites. In Class IV subgroup, no ZI intrusion into infratemporal fossa occurred at B0 and B1 site, while 35.9% of the ZI at B2 site and 89.7% of the ZI at B3 site showed intrusion, with intrusion depth of 2.4 ± 1.0 mm (B2) and 3.9 ± 1.7 mm (B3). For the C series apex points, 15.4% of ZI at C0, 56.4% of ZI at C1 site, and over 90% of ZI at C2 and C3 sites entered the infratemporal fossa (Table ), with intrusion depths increasing from C0 to C3. Within the Class V/VI subgroup, there was no ZI intrusion into the infratemporal fossa at the B0 site, and only one ZI at the B1 site (5.9%) penetrated the infratemporal fossa with an intrusion depth of 2.4 mm. In contrast, 58.8% of ZIs at the B2 site and all ZIs at the B3 site (100%) intruded into the infratemporal fossa (Table ). Regarding the C series sites, 23.5% of ZIs at the C0 site, 64.7% of ZIs at the C1 site, and over 90% of ZIs at both the C2 and C3 sites entered the infratemporal fossa (Table ). Although the exact values for ZI intrusion rate, as well as the average depth and length of intrusion, were higher in the Class V/VI subgroup compared to Class IVsubgroup, there was no statistically significant difference in the rate, depth, or length of ZI intrusion at any of the middle and lower zygoma apex sites between the two subgroups ( P > 0.05).
For the anterior zygomatic implants, only one of the 56 implants (1.8%) at the A3 point intruded into the infratemporal fossa. The depth of the intrusion was 1.2 mm. No other anterior implants showed entry into the fossa. There was no significant difference in the rate of ZI intrusion among the four apex points (A series points) ( P = 0.390). The single implant that intruded into the infratemporal fossa was from the subgroup of class IV, although there was no significant difference in the rate of anterior ZI intrusion between the two subgroups.
For the B series apex points (mid-zygoma), no implant intrusion into the infratemporal fossa was observed at B0. However, intrusion rates at B1, B2, and B3 were 1.8%, 42.9%, and 92.9%, respectively, with the average depth and length of intrusion into the infratemporal fossa increasing progressively. The average intrusion depth was 2.4 mm, 2.2 ± 1.0 mm, and 3.9 ± 1.6 mm at B1, B2, and B3 respectively. Significant differences in intrusion rates were noted between B3 and B0, and B2 and B0, but not between B1 and B0. For the C series apex points (lower zygoma), varying rates of ZI intrusion into the infratemporal fossa were recorded across the four sites. From C0 to C3, the intrusion rate, depth and length of ZI exposure in the infratemporal fossa increased (Table ). At the C3 apex point, all implants intruded into the infratemporal fossa, with an average intrusion depth of 7.8 ± 2.6 mm. Significant differences in intrusion rates were present among all C series sites. In Class IV subgroup, no ZI intrusion into infratemporal fossa occurred at B0 and B1 site, while 35.9% of the ZI at B2 site and 89.7% of the ZI at B3 site showed intrusion, with intrusion depth of 2.4 ± 1.0 mm (B2) and 3.9 ± 1.7 mm (B3). For the C series apex points, 15.4% of ZI at C0, 56.4% of ZI at C1 site, and over 90% of ZI at C2 and C3 sites entered the infratemporal fossa (Table ), with intrusion depths increasing from C0 to C3. Within the Class V/VI subgroup, there was no ZI intrusion into the infratemporal fossa at the B0 site, and only one ZI at the B1 site (5.9%) penetrated the infratemporal fossa with an intrusion depth of 2.4 mm. In contrast, 58.8% of ZIs at the B2 site and all ZIs at the B3 site (100%) intruded into the infratemporal fossa (Table ). Regarding the C series sites, 23.5% of ZIs at the C0 site, 64.7% of ZIs at the C1 site, and over 90% of ZIs at both the C2 and C3 sites entered the infratemporal fossa (Table ). Although the exact values for ZI intrusion rate, as well as the average depth and length of intrusion, were higher in the Class V/VI subgroup compared to Class IVsubgroup, there was no statistically significant difference in the rate, depth, or length of ZI intrusion at any of the middle and lower zygoma apex sites between the two subgroups ( P > 0.05).
No orbital cavity penetration was detected in all the cases at all apex points. The distance between the anterior ZI and the orbital cavity was shown in Table . The average distance between the anterior ZI and the orbit was 2.5 ± 1.0 mm, 3.2 ± 1.0 mm, 3.8 ± 1.0 mm, and 4.3 ± 0.9 mm at A3, A2, A1, and A0 apex points respectively. There was significant difference in ZI-orbital distance among the four apex groups ( P < 0.001). No significant difference in ZI-orbital distance was found between the Class IV and Class V/VI subgroups at all the A series points. Three anterior ZI at A3 point showed less than 1 mm distance from the orbital cavity.
Choice of apex site for quad zygomatic implants The anterior ZI of quad zygomatic implant exhibited the highest zygomatic BIC when the A3 apex point at the upper zygoma was selected. The posterior ZI demonstrated the highest zygomatic BIC at the B2 apex point in the middle zygoma and the C1 apex point in the lower zygoma. These findings align with the clinical study by Wu and colleagues 。. The zygomatic BIC of anterior ZI increased progressively as the apex point of ZI moved from A0 to A3, due to the elongating distance the ZI traversed through the zygoma. For the posterior ZI, the zygomatic BIC rose from B0 to B2, then diminished from B2 to B3, and similarly increased from C0 to C1 before decreasing from C1 to C3. This pattern is attributed to the portion of the ZI that is exposed in the infratemporal fossa. The more distal the apex point, the greater the percentage of ZI exposure in the infratemporal fossa. Among all middle and lower zygoma apex points, the highest average zBIC and tBIC can be achieved when ZI ended at the B2 point, The B1, B3, C1, and C2 apex points can also provide relatively high BIC for the ZI. Infratemporal fossa serves as the passage way of many important neurovascular structure, including the maxillary artery and its branches, the pterygoid venous piexus, mandibular nerve, and more. Given the zygomatic bone’s curvature and the invariably straight path of the ZI, there’s a significant risk of implants entering the infratemporal fossa when selecting distal apex points. Rossi et al. have recommended adjusting the ZI insertion angle to prevent its encroachment into the infratemporal fossa and to protect the vital neurovascular structures . For the anterior quad ZI, the selection of the A3 apex point resulted in only one instance of infratemporal fossa intrusion out of 56 ZIs. Statistical analysis indicated no significant difference in the rate of infratemporal fossa intrusion across all A series apex points, implying that A3 remains a secure and preferred choice. For the posterior quad ZI, despite the highest zBIC and tBIC at the B2 point, the rate of infratemporal fossa intrusion was 42.9%, with an average intrusion depth of 2.2 ± 1.0 mm, potentially harming the contents in infratemporal fossa. All apex points in the lower zygoma (C series) demonstrated a risk of ZI intrusion into the infratemporal fossa, particularly at C2 and C3. However, at the B1 site, only one of the 56 posterior quad ZIs intruded into the infratemporal fossa, mirroring the anterior ZI at the A3 apex point. Therefore, placing ZI at the B1 site is considered safe for achieving high BIC while avoiding damage to the content in infratemporal fossa. In summary, maximizing the zygomatic bone volume for optimal BIC, while ensuring patient safety, A3 and B1 emerged as the superior apex points for anterior and posterior quad ZI, respectively. The optimal sites for the apex points of quad ZIs were identified as the upper posterior and the anterior middle portions of the zygoma. Risk of ZI intrusion into the infratemporal fossa and difference between the two subgroups In this study the edentulous maxilla residual ridges were classified in to Class IV and Class V/VI according to the Cawood and Howell classification . Simillar to the entire cohort, the highest zygomatic and total BIC for the anterior ZI in both subgroups was achieved at the A3 apex point, with no significant difference in the rate of intrusion into the infratemporal fossa between the Class IV and Class V/VI subgroups at the A3 point. For the posterior ZI, the highest zBIC and tBIC were achieved when the ZI was placed at the B2 apex point. The second highest zBIC and tBIC point in the Class IV subgroup were observed at the C1, B3, and B1 points. In the Class V/VI subgroup, the B1 point was also a favorable choice for the apex since there was no significant difference in zBIC and tBIC between the B1 and B2 points. Considering the risk of intrusion into the infratemporal fossa, it was found that the rate of posterior ZI penetration into the infratemporal fossa was high in ZIs at B2, B3, and all C series points. Consequently, the B1 and B0 points were safer choices. Taking into account both BIC and infratemporal fossa risk factors, the optimal apex point for the posterior ZI was B1. One case of anterior ZI intrusion into the infratemporal fossa at the A3 apex point in the Class IV subgroup was noted, along with a posterior ZI intrusion at the B1 apex point in the Class V/VI subgroup. The zygomas in these cases exhibited greater curvature and were thinner than those in other cases, resulting in a shallower infratemporal fossa and an increased likelihood of ZI penetration into infratemporal fossa. No significant differences were found between Class IV and Class V/VI subgroups at any apex point sites, indicating that the risk of ZI intrusion is influenced by zygomatic anatomy rather than the classification of residual ridge resorption. Preoperative analysis of the zygomatic anatomy is crucial, particularly for patients with prominent zygomatic bones. Should ZI intrusion be detected, the anterior ZI apex point could be moved forward to the A2 point, and the posterior ZI apex point could be adjusted toward the B0 point. For the same apex point, no significant differences were observed in BIC or the rate, length, and depth of ZI intrusion into infratemporal fossa between Class IV and Class V/IV subgroups. These findings suggest that the classification of the edentulous residual ridge has minimal impact on the primary stability and safety of quad ZI, which can be safely applied in patients with severe maxillary residual ridge resorption. It should be noted that the infratemporal intrusion in this Chinese population may be different from those of the western demographic groups since the facial features of asian population are characterized by a relatively flatter facial profile and more prominent zygomatic bones. Risk of orbital cavity penetration Orbital cavity penetration is a severe complication with the incidence of 5.9%[21] in ZI placement. Care must be taken to avoid the bony orbit during ZI placement. In this study, three anterior ZI at A3 apex point showed less than 1 mm distance from the orbital cavity. Even with the computer assisted implant surgery (CAIS), the deviation of ZI placement remains above 2 mm for static and dynamic CAIS . The orbital penetration risk should be taken into account when A3 point was selected as the apex point for anterior ZI. The zygomatic anatomy-guided approach (ZAGA) concept that focus on interindividual anatomic differences can be considered to avoid this complication . Possible risk of interference between the anterior and posterior ZIs There are potential risk of interference between the anterior and posterior ZIs when this zygoma segementation protocol was used. When B series points were taken as the apex points for the posterior ZI, there are chances that the anterior ZI pass by in a distance of less than 1.6 mm at some point. There is a risk for the two ZIs coming into contact considering the deviations in ZI placement. Causions should be taken when the B series points were taken as the apex point for the posterior ZI. This study has some limitations. First, the study design and outcome are based solely on CBCT anatomical analysis and virtual implant planning, and the BIC measurement was base on linear instead of area measurement, and these may be different from actual clinical practice. Second, the classification of the edentulous residual ridge was based on ridge height measurements and clinical examination, suggesting that a more precise classification system is needed for future research. Third, the choice of apex points for the anterior ZI was set at the A series point, in actual clinical practice, points B series can also be selected as apical points for anterior ZIs. Risk of orbital injury need to be considered when placing ZIs at A series points, especially A3 apex point. More refined zygomatic bone segmentation is needed to avoid orbital risk. These needs further investigation in future studies.
The anterior ZI of quad zygomatic implant exhibited the highest zygomatic BIC when the A3 apex point at the upper zygoma was selected. The posterior ZI demonstrated the highest zygomatic BIC at the B2 apex point in the middle zygoma and the C1 apex point in the lower zygoma. These findings align with the clinical study by Wu and colleagues 。. The zygomatic BIC of anterior ZI increased progressively as the apex point of ZI moved from A0 to A3, due to the elongating distance the ZI traversed through the zygoma. For the posterior ZI, the zygomatic BIC rose from B0 to B2, then diminished from B2 to B3, and similarly increased from C0 to C1 before decreasing from C1 to C3. This pattern is attributed to the portion of the ZI that is exposed in the infratemporal fossa. The more distal the apex point, the greater the percentage of ZI exposure in the infratemporal fossa. Among all middle and lower zygoma apex points, the highest average zBIC and tBIC can be achieved when ZI ended at the B2 point, The B1, B3, C1, and C2 apex points can also provide relatively high BIC for the ZI. Infratemporal fossa serves as the passage way of many important neurovascular structure, including the maxillary artery and its branches, the pterygoid venous piexus, mandibular nerve, and more. Given the zygomatic bone’s curvature and the invariably straight path of the ZI, there’s a significant risk of implants entering the infratemporal fossa when selecting distal apex points. Rossi et al. have recommended adjusting the ZI insertion angle to prevent its encroachment into the infratemporal fossa and to protect the vital neurovascular structures . For the anterior quad ZI, the selection of the A3 apex point resulted in only one instance of infratemporal fossa intrusion out of 56 ZIs. Statistical analysis indicated no significant difference in the rate of infratemporal fossa intrusion across all A series apex points, implying that A3 remains a secure and preferred choice. For the posterior quad ZI, despite the highest zBIC and tBIC at the B2 point, the rate of infratemporal fossa intrusion was 42.9%, with an average intrusion depth of 2.2 ± 1.0 mm, potentially harming the contents in infratemporal fossa. All apex points in the lower zygoma (C series) demonstrated a risk of ZI intrusion into the infratemporal fossa, particularly at C2 and C3. However, at the B1 site, only one of the 56 posterior quad ZIs intruded into the infratemporal fossa, mirroring the anterior ZI at the A3 apex point. Therefore, placing ZI at the B1 site is considered safe for achieving high BIC while avoiding damage to the content in infratemporal fossa. In summary, maximizing the zygomatic bone volume for optimal BIC, while ensuring patient safety, A3 and B1 emerged as the superior apex points for anterior and posterior quad ZI, respectively. The optimal sites for the apex points of quad ZIs were identified as the upper posterior and the anterior middle portions of the zygoma.
In this study the edentulous maxilla residual ridges were classified in to Class IV and Class V/VI according to the Cawood and Howell classification . Simillar to the entire cohort, the highest zygomatic and total BIC for the anterior ZI in both subgroups was achieved at the A3 apex point, with no significant difference in the rate of intrusion into the infratemporal fossa between the Class IV and Class V/VI subgroups at the A3 point. For the posterior ZI, the highest zBIC and tBIC were achieved when the ZI was placed at the B2 apex point. The second highest zBIC and tBIC point in the Class IV subgroup were observed at the C1, B3, and B1 points. In the Class V/VI subgroup, the B1 point was also a favorable choice for the apex since there was no significant difference in zBIC and tBIC between the B1 and B2 points. Considering the risk of intrusion into the infratemporal fossa, it was found that the rate of posterior ZI penetration into the infratemporal fossa was high in ZIs at B2, B3, and all C series points. Consequently, the B1 and B0 points were safer choices. Taking into account both BIC and infratemporal fossa risk factors, the optimal apex point for the posterior ZI was B1. One case of anterior ZI intrusion into the infratemporal fossa at the A3 apex point in the Class IV subgroup was noted, along with a posterior ZI intrusion at the B1 apex point in the Class V/VI subgroup. The zygomas in these cases exhibited greater curvature and were thinner than those in other cases, resulting in a shallower infratemporal fossa and an increased likelihood of ZI penetration into infratemporal fossa. No significant differences were found between Class IV and Class V/VI subgroups at any apex point sites, indicating that the risk of ZI intrusion is influenced by zygomatic anatomy rather than the classification of residual ridge resorption. Preoperative analysis of the zygomatic anatomy is crucial, particularly for patients with prominent zygomatic bones. Should ZI intrusion be detected, the anterior ZI apex point could be moved forward to the A2 point, and the posterior ZI apex point could be adjusted toward the B0 point. For the same apex point, no significant differences were observed in BIC or the rate, length, and depth of ZI intrusion into infratemporal fossa between Class IV and Class V/IV subgroups. These findings suggest that the classification of the edentulous residual ridge has minimal impact on the primary stability and safety of quad ZI, which can be safely applied in patients with severe maxillary residual ridge resorption. It should be noted that the infratemporal intrusion in this Chinese population may be different from those of the western demographic groups since the facial features of asian population are characterized by a relatively flatter facial profile and more prominent zygomatic bones.
Orbital cavity penetration is a severe complication with the incidence of 5.9%[21] in ZI placement. Care must be taken to avoid the bony orbit during ZI placement. In this study, three anterior ZI at A3 apex point showed less than 1 mm distance from the orbital cavity. Even with the computer assisted implant surgery (CAIS), the deviation of ZI placement remains above 2 mm for static and dynamic CAIS . The orbital penetration risk should be taken into account when A3 point was selected as the apex point for anterior ZI. The zygomatic anatomy-guided approach (ZAGA) concept that focus on interindividual anatomic differences can be considered to avoid this complication .
There are potential risk of interference between the anterior and posterior ZIs when this zygoma segementation protocol was used. When B series points were taken as the apex points for the posterior ZI, there are chances that the anterior ZI pass by in a distance of less than 1.6 mm at some point. There is a risk for the two ZIs coming into contact considering the deviations in ZI placement. Causions should be taken when the B series points were taken as the apex point for the posterior ZI. This study has some limitations. First, the study design and outcome are based solely on CBCT anatomical analysis and virtual implant planning, and the BIC measurement was base on linear instead of area measurement, and these may be different from actual clinical practice. Second, the classification of the edentulous residual ridge was based on ridge height measurements and clinical examination, suggesting that a more precise classification system is needed for future research. Third, the choice of apex points for the anterior ZI was set at the A series point, in actual clinical practice, points B series can also be selected as apical points for anterior ZIs. Risk of orbital injury need to be considered when placing ZIs at A series points, especially A3 apex point. More refined zygomatic bone segmentation is needed to avoid orbital risk. These needs further investigation in future studies.
For the placement of quad zygomatic implants, the optional apex location for the anterior ZI is A3, while for the posterior ZI, it is B1. These apex points ensure favorable BIC and a reduced risk of infratemporal fossa invasion. Anterior ZI positioned at A3 point may present high risk for orbital penetration and may not be reccomended in a Quad ZI approach. The degree of residual alveolar bone resorption does not affect the BIC of quad zygomatic implants.
|
Finding a needle in a haystack: The identification of clinical practice guidelines for psychosocial oncology through an environmental scan of the academic and gray literature | 14b9bcab-2905-47cb-95cf-d868b8594be2 | 10891449 | Internal Medicine[mh] | INTRODUCTION Cancer is one of the leading causes of death across the world, with as many as 19.3 million new cases diagnosed worldwide in 2020 alone. As incidence rates are expected to rise by nearly 60% over the next 20 years, the burden of this disease will also increase, with an unprecedented number of individuals, families, and communities facing not only the associated physical ailments and escalating costs incurred by healthcare systems, but also a wide range of psychological, social, and occupational challenges. , Common psychosocial symptoms reported by patients with cancer include overall psychological distress, , , symptoms of depression , and anxiety, , demoralization, interpersonal difficulties, , and reduced quality of life. In 2007, the US Institute of Medicine recommended that treating psychosocial symptoms in patients with cancer is best practice for comprehensive cancer care. , This assertion has been widely endorsed, as evidenced by the rapid advances in the availability of new psychosocial oncology services, upsurges in relevant peer‐reviewed articles, care standards, and reports, and the development of new interventions. , The rapid expansion of psychosocial oncology poses a challenge for clinicians providing day‐to‐day clinical care. Psychosocial oncology clinicians experience large caseloads, have limited time for selecting and reviewing relevant scientific literature, and might have limited capacity to translate empirical knowledge into practical interventions to be delivered at the point of care. , , Clinical practice guidelines (CPGs) are well‐suited to bridge the gap between scholarly knowledge and service delivery. CPGs are systematically developed tools designed to summarize the latest scientific evidence and provide clinically applicable recommendations to inform treatment decisions for various health conditions. CPGs targeting psychosocial symptoms for patients with cancer provide recommendations based on the latest scientific findings for clinical care at all stages of the cancer care continuum, from screening to treatment and follow‐up, and also help standardize service offerings and optimize the use of often limited resources. , , Yet, little is known about the current scope and accessibility of CPGs in psychosocial oncology. Currently, there is no universally accessible, comprehensive repository for CPGs, which makes it difficult for clinician to find and access such evidence‐based documents. While some CPGs may be accessible through “traditional” academic sources (e.g., PubMed database), clinicians may lack institutional access to databases depending on their work settings. Further, some CPGs are not indexed in academic databases and can only be accessed through non‐commercial sources. For example, certain special interest groups and government‐funded organizations only publish their CPGs on their organizational websites. 1.1 Study objective To bridge this knowledge gap, we conducted an environmental scan to identify the number and scope of clinical practice guidelines that target the psychosocial symptoms of individuals diagnosed with cancer across the cancer care trajectory and examine their accessibility through different search methods. Study objective To bridge this knowledge gap, we conducted an environmental scan to identify the number and scope of clinical practice guidelines that target the psychosocial symptoms of individuals diagnosed with cancer across the cancer care trajectory and examine their accessibility through different search methods. METHODS 2.1 Procedures An environmental scan is a type of review that encompasses both published literature (e.g., peer‐reviewed articles) and unpublished or informally published documents. We selected this review method to increase the likelihood of identifying existing psychosocial oncology CPGs, which are often uploaded onto institutional or organizational websites by their developers. Our environmental scan was informed by the recommendations for systematic reviews for CPGs outlined by Johnston et al. We included two separate systematic searches to review (a) the academic literature and (b) the “gray” literature. Initial searches were conducted from January 2023 to February 2023 and updated in July 2023. 2.2 Inclusion and exclusion criteria Eligibility criteria were established in two steps. First, we identified relevant CPGs based on methodological characteristics, as per the National Guidelines Clearinghouse 2013 revised criteria and the Institute of Medicine's 2011 definition of clinical practice guidelines, which included: (1) Systematically developed treatment decision‐making recommendations for healthcare professionals; (2) Recommendations developed or endorsed by a professional, medical, public, or private agency/organization; and (3) Full CPG text developed and/or available in English. Second, we used the P opulation, I nterventions, C ontent, A ttributes of eligible CPGs, and R ecommendation characteristics (PICAR) framework, which adapted the PICOTS framework to CPGs : (1) Population included patients diagnosed with cancer at any disease stage; (2) Recommendations targeted at least one psychosocial aspect of cancer; (3) Recommendations for the treatment of psychosocial symptoms focused on non‐pharmacological interventions; and (4) Recommendations were published, revised, and/or updated within the past 10 years. Supplemental guidelines developed by the same organization complementary to pre‐existing CPGs were also included. Exclusion criteria included: (1) Reports exclusively documenting updated methodology and outcomes; (2) Consensus statements synthesizing expert and stakeholder opinions about care standards for use in contexts with minimal evidence ; (3) Recommendations not relevant to psychosocial oncology; (4) Over 10 years since date of publication or update; and (5) Full text and/or recommendations not available in English. 2.3 Search strategy 2.3.1 Academic literature search This search consisted of a systematic search of five peer‐reviewed academic databases (PubMed, Medline, JSTOR, PsycINFO, and Clinical Key) and was developed in consultation with a research librarian at McGill University. Relevant keywords for each applicable PICAR component were identified through a review of the literature on psychosocial oncology care and in consultation with experts and our research team. Keywords were utilized as either free‐text or MeSH terms (if available) and paired with Boolean operators. To maximize the retrieval of relevant results and ensure a comprehensive search, spelling variations were included through database‐specific truncation and search settings were set to identify keywords in all text (i.e., titles, abstracts, subject heading, full text, etc.). Results were filtered by language to include only English results and by publication type (i.e., guideline) when applicable. See Appendix, Table for the electronic search strategy for each database. 2.3.2 Gray literature search The gray literature search included a multistep approach. Step 1 consisted of a generic Internet search using the Google search engine through the Google Chrome browser. Based on recommendations for limiting method‐inherent selection bias, cookies and search history were cleared from the browser prior to conducting the search. , Keywords and search terms from the academic search were adapted for the generic Internet search. Step 2 consisted of a Google Scholar search, which indexes both peer‐reviewed and gray literature resources. Keywords and search terms from the academic search were used to generate relevant search strings and searched within both “full text” and “title” using the advanced search functions. Step 3 consisted of a targeted Internet search for CPGs through the official websites of organizations, groups, agencies, and government bodies involved in psychosocial oncology or guideline development. The identification of relevant websites was informed by our literature search and consultation with our research team as well as domain experts. We consulted the websites of the following organizations: the National Comprehensive Cancer Network (NCCN; https://www.nccn.org/ ), the Canadian Association of Psychosocial Oncology (CAPO; https://www.capo.ca/ ), Cancer Care Ontario (CCO; https://www.cancercareontario.ca/en ), the European Society for Medical Oncology (ESMO; https://www.esmo.org/ ), the American Society of Clinical Oncology (ASCO; https://www.asco.org/ ), the American Psychosocial Oncology Society (APOS; https://apos‐society.org/ ), and the International Psycho‐Oncology Society (IPOS; https://www.ipos‐society.org/ ). 2.4 Identification and data extraction All results generated from the academic and targeted Internet search were retained. As both the generic Internet search and Google Scholar search were expected to produce more results than was feasible to screen, we followed the recommendation by Haddaway et al. to limit screening to the first 300 results for each search string (i.e., 30 pages). Results were imported into Zotero 6.0.22, deduplicated, and screened for eligibility. An initial screening of titles and abstracts was conducted, followed by a full‐text screening. If multiple versions of a CPG were identified, only the latest version of the CPG was retained for review. Following identification, data extracted from guidelines included: developer, location, psychosocial symptoms/concerns addressed, target population (age and cancer type), and stage of care. CPG identification and data extraction was independently conducted by two reviewers. All disagreements between the reviewers were recorded and discussed until consensus was reached. Procedures An environmental scan is a type of review that encompasses both published literature (e.g., peer‐reviewed articles) and unpublished or informally published documents. We selected this review method to increase the likelihood of identifying existing psychosocial oncology CPGs, which are often uploaded onto institutional or organizational websites by their developers. Our environmental scan was informed by the recommendations for systematic reviews for CPGs outlined by Johnston et al. We included two separate systematic searches to review (a) the academic literature and (b) the “gray” literature. Initial searches were conducted from January 2023 to February 2023 and updated in July 2023. Inclusion and exclusion criteria Eligibility criteria were established in two steps. First, we identified relevant CPGs based on methodological characteristics, as per the National Guidelines Clearinghouse 2013 revised criteria and the Institute of Medicine's 2011 definition of clinical practice guidelines, which included: (1) Systematically developed treatment decision‐making recommendations for healthcare professionals; (2) Recommendations developed or endorsed by a professional, medical, public, or private agency/organization; and (3) Full CPG text developed and/or available in English. Second, we used the P opulation, I nterventions, C ontent, A ttributes of eligible CPGs, and R ecommendation characteristics (PICAR) framework, which adapted the PICOTS framework to CPGs : (1) Population included patients diagnosed with cancer at any disease stage; (2) Recommendations targeted at least one psychosocial aspect of cancer; (3) Recommendations for the treatment of psychosocial symptoms focused on non‐pharmacological interventions; and (4) Recommendations were published, revised, and/or updated within the past 10 years. Supplemental guidelines developed by the same organization complementary to pre‐existing CPGs were also included. Exclusion criteria included: (1) Reports exclusively documenting updated methodology and outcomes; (2) Consensus statements synthesizing expert and stakeholder opinions about care standards for use in contexts with minimal evidence ; (3) Recommendations not relevant to psychosocial oncology; (4) Over 10 years since date of publication or update; and (5) Full text and/or recommendations not available in English. Search strategy 2.3.1 Academic literature search This search consisted of a systematic search of five peer‐reviewed academic databases (PubMed, Medline, JSTOR, PsycINFO, and Clinical Key) and was developed in consultation with a research librarian at McGill University. Relevant keywords for each applicable PICAR component were identified through a review of the literature on psychosocial oncology care and in consultation with experts and our research team. Keywords were utilized as either free‐text or MeSH terms (if available) and paired with Boolean operators. To maximize the retrieval of relevant results and ensure a comprehensive search, spelling variations were included through database‐specific truncation and search settings were set to identify keywords in all text (i.e., titles, abstracts, subject heading, full text, etc.). Results were filtered by language to include only English results and by publication type (i.e., guideline) when applicable. See Appendix, Table for the electronic search strategy for each database. 2.3.2 Gray literature search The gray literature search included a multistep approach. Step 1 consisted of a generic Internet search using the Google search engine through the Google Chrome browser. Based on recommendations for limiting method‐inherent selection bias, cookies and search history were cleared from the browser prior to conducting the search. , Keywords and search terms from the academic search were adapted for the generic Internet search. Step 2 consisted of a Google Scholar search, which indexes both peer‐reviewed and gray literature resources. Keywords and search terms from the academic search were used to generate relevant search strings and searched within both “full text” and “title” using the advanced search functions. Step 3 consisted of a targeted Internet search for CPGs through the official websites of organizations, groups, agencies, and government bodies involved in psychosocial oncology or guideline development. The identification of relevant websites was informed by our literature search and consultation with our research team as well as domain experts. We consulted the websites of the following organizations: the National Comprehensive Cancer Network (NCCN; https://www.nccn.org/ ), the Canadian Association of Psychosocial Oncology (CAPO; https://www.capo.ca/ ), Cancer Care Ontario (CCO; https://www.cancercareontario.ca/en ), the European Society for Medical Oncology (ESMO; https://www.esmo.org/ ), the American Society of Clinical Oncology (ASCO; https://www.asco.org/ ), the American Psychosocial Oncology Society (APOS; https://apos‐society.org/ ), and the International Psycho‐Oncology Society (IPOS; https://www.ipos‐society.org/ ). Academic literature search This search consisted of a systematic search of five peer‐reviewed academic databases (PubMed, Medline, JSTOR, PsycINFO, and Clinical Key) and was developed in consultation with a research librarian at McGill University. Relevant keywords for each applicable PICAR component were identified through a review of the literature on psychosocial oncology care and in consultation with experts and our research team. Keywords were utilized as either free‐text or MeSH terms (if available) and paired with Boolean operators. To maximize the retrieval of relevant results and ensure a comprehensive search, spelling variations were included through database‐specific truncation and search settings were set to identify keywords in all text (i.e., titles, abstracts, subject heading, full text, etc.). Results were filtered by language to include only English results and by publication type (i.e., guideline) when applicable. See Appendix, Table for the electronic search strategy for each database. Gray literature search The gray literature search included a multistep approach. Step 1 consisted of a generic Internet search using the Google search engine through the Google Chrome browser. Based on recommendations for limiting method‐inherent selection bias, cookies and search history were cleared from the browser prior to conducting the search. , Keywords and search terms from the academic search were adapted for the generic Internet search. Step 2 consisted of a Google Scholar search, which indexes both peer‐reviewed and gray literature resources. Keywords and search terms from the academic search were used to generate relevant search strings and searched within both “full text” and “title” using the advanced search functions. Step 3 consisted of a targeted Internet search for CPGs through the official websites of organizations, groups, agencies, and government bodies involved in psychosocial oncology or guideline development. The identification of relevant websites was informed by our literature search and consultation with our research team as well as domain experts. We consulted the websites of the following organizations: the National Comprehensive Cancer Network (NCCN; https://www.nccn.org/ ), the Canadian Association of Psychosocial Oncology (CAPO; https://www.capo.ca/ ), Cancer Care Ontario (CCO; https://www.cancercareontario.ca/en ), the European Society for Medical Oncology (ESMO; https://www.esmo.org/ ), the American Society of Clinical Oncology (ASCO; https://www.asco.org/ ), the American Psychosocial Oncology Society (APOS; https://apos‐society.org/ ), and the International Psycho‐Oncology Society (IPOS; https://www.ipos‐society.org/ ). Identification and data extraction All results generated from the academic and targeted Internet search were retained. As both the generic Internet search and Google Scholar search were expected to produce more results than was feasible to screen, we followed the recommendation by Haddaway et al. to limit screening to the first 300 results for each search string (i.e., 30 pages). Results were imported into Zotero 6.0.22, deduplicated, and screened for eligibility. An initial screening of titles and abstracts was conducted, followed by a full‐text screening. If multiple versions of a CPG were identified, only the latest version of the CPG was retained for review. Following identification, data extracted from guidelines included: developer, location, psychosocial symptoms/concerns addressed, target population (age and cancer type), and stage of care. CPG identification and data extraction was independently conducted by two reviewers. All disagreements between the reviewers were recorded and discussed until consensus was reached. RESULTS 3.1 Screening and reliability Two reviewers independently screened the results from all search methods based on the eligibility criteria described above and identified a total of 35 psychosocial oncology CPGs. Following a consensus meeting, 14 disagreements in identification resulted in 12 excluded records and two determined to meet full eligibility criteria. Reasons for exclusion were insufficient relevance to psychosocial oncology ( n = 7), no evidence for the use of systematic methods to develop treatment recommendations ( n = 4), and availability of a more up‐to‐date version ( n = 1). Reliability at the full‐text review stage was excellent between the two reviewers (Cohen's κ = 0.88). 3.2 Clinical practice guideline identification A total of 10,423 initial records were generated by the various search methods employed in this study. Prior to screening and deduplication, five records were flagged by Retraction Watch and removed. The listed reasons for retraction included: copyright claims ( n = 1), self‐plagiarism ( n = 1), errors in the data ( n = 1), problems with the results ( n = 2), objections by a third party ( n = 1), and withdrawal ( n = 1). A total of 218 records were retained for a full‐text review. Based on both the methodological and PICAR eligibility criteria, 195 records were excluded following a full‐text review and reviewer consensus meetings. In total, 23 records were deemed eligible and identified as psychosocial oncology CPGs. Refer to Table for a complete list of all identified CPGs. Additionally, a PRISMA flow diagram detailing the CPG selection process can be found in Figure . The academic literature search yielded a total of 7498 relevant hits and identified 9 CPGs that met full eligibility criteria, only 1 of which were unique to this search (i.e., not identified by any other search method). In contrast, the gray literature search generated 2925 records and identified a total of 22 CPGs, 14 of which were unique to this search. Differences emerged between the three different types of gray literature searches. First, the targeted Internet search yielded 860 records and identified 21 CPGs in total, 1 of which was unique to this search. Next, the naïve Google search yielded 1098 records and identified 16 CPGs overall, 1 of which was unique to the search. Lastly, the Google Scholar search generated 900 records and identified 6 CPGs, all of which were also identified by another search. Refer to Table for a visual representation of the identification of each CPG by the different search methods. 3.3 Characteristics of identified clinical practice guidelines The 23 psychosocial oncology CPGs identified through the environmental scan were published and/or updated between 2014 and 2023 and were developed by multiple organizations, including the American Cancer Society, the American Society of Clinical Oncology (ASCO), the Canadian Association of Psychosocial Oncology (CAPO), Cancer Care Ontario, the European Society for Medical Oncology (ESMO), the National Comprehensive Cancer Network (NCCN), and the Italian Medical Oncology Association. The majority of the CPGs were developed in the USA ( n = 14, 60.9%), followed by Canada ( n = 5, 21.7%) and Europe ( n = 4, 17.4%). The psychosocial symptoms or concerns addressed by the CPGs included distress ( n = 15, 65.2%), depression ( n = 15, 65.2%), fatigue ( n = 11, 47.8%), anxiety ( n = 10, 43.5%), social support ( n = 8, 34.8%), sexual dysfunction ( n = 7, 30.4%), existential/spiritual concerns ( n = 6, 26.1%), quality of life ( n = 5, 21.7%), sleep disturbances ( n = 4, 17.4%), and body image preoccupations ( n = 4, 17.4%). Approximately half of the CPGs targeted a single specific or limited range of symptoms ( n = 11, 47.8%) whereas the other half addressed global psychosocial care, addressing a wider scope of potential symptoms and concerns ( n = 12, 52.2%). See Table for more detailed characteristics of the 23 relevant CPGs. Screening and reliability Two reviewers independently screened the results from all search methods based on the eligibility criteria described above and identified a total of 35 psychosocial oncology CPGs. Following a consensus meeting, 14 disagreements in identification resulted in 12 excluded records and two determined to meet full eligibility criteria. Reasons for exclusion were insufficient relevance to psychosocial oncology ( n = 7), no evidence for the use of systematic methods to develop treatment recommendations ( n = 4), and availability of a more up‐to‐date version ( n = 1). Reliability at the full‐text review stage was excellent between the two reviewers (Cohen's κ = 0.88). Clinical practice guideline identification A total of 10,423 initial records were generated by the various search methods employed in this study. Prior to screening and deduplication, five records were flagged by Retraction Watch and removed. The listed reasons for retraction included: copyright claims ( n = 1), self‐plagiarism ( n = 1), errors in the data ( n = 1), problems with the results ( n = 2), objections by a third party ( n = 1), and withdrawal ( n = 1). A total of 218 records were retained for a full‐text review. Based on both the methodological and PICAR eligibility criteria, 195 records were excluded following a full‐text review and reviewer consensus meetings. In total, 23 records were deemed eligible and identified as psychosocial oncology CPGs. Refer to Table for a complete list of all identified CPGs. Additionally, a PRISMA flow diagram detailing the CPG selection process can be found in Figure . The academic literature search yielded a total of 7498 relevant hits and identified 9 CPGs that met full eligibility criteria, only 1 of which were unique to this search (i.e., not identified by any other search method). In contrast, the gray literature search generated 2925 records and identified a total of 22 CPGs, 14 of which were unique to this search. Differences emerged between the three different types of gray literature searches. First, the targeted Internet search yielded 860 records and identified 21 CPGs in total, 1 of which was unique to this search. Next, the naïve Google search yielded 1098 records and identified 16 CPGs overall, 1 of which was unique to the search. Lastly, the Google Scholar search generated 900 records and identified 6 CPGs, all of which were also identified by another search. Refer to Table for a visual representation of the identification of each CPG by the different search methods. Characteristics of identified clinical practice guidelines The 23 psychosocial oncology CPGs identified through the environmental scan were published and/or updated between 2014 and 2023 and were developed by multiple organizations, including the American Cancer Society, the American Society of Clinical Oncology (ASCO), the Canadian Association of Psychosocial Oncology (CAPO), Cancer Care Ontario, the European Society for Medical Oncology (ESMO), the National Comprehensive Cancer Network (NCCN), and the Italian Medical Oncology Association. The majority of the CPGs were developed in the USA ( n = 14, 60.9%), followed by Canada ( n = 5, 21.7%) and Europe ( n = 4, 17.4%). The psychosocial symptoms or concerns addressed by the CPGs included distress ( n = 15, 65.2%), depression ( n = 15, 65.2%), fatigue ( n = 11, 47.8%), anxiety ( n = 10, 43.5%), social support ( n = 8, 34.8%), sexual dysfunction ( n = 7, 30.4%), existential/spiritual concerns ( n = 6, 26.1%), quality of life ( n = 5, 21.7%), sleep disturbances ( n = 4, 17.4%), and body image preoccupations ( n = 4, 17.4%). Approximately half of the CPGs targeted a single specific or limited range of symptoms ( n = 11, 47.8%) whereas the other half addressed global psychosocial care, addressing a wider scope of potential symptoms and concerns ( n = 12, 52.2%). See Table for more detailed characteristics of the 23 relevant CPGs. DISCUSSION Clinical practice guidelines are essential tools for quality healthcare provision and clinical practice. They are meant to support a wide range of purposes, including informing professionals about new pharmaceutical and non‐pharmaceutical interventions, reducing the variability in clinical practice, improving patient‐reported outcomes, and establishing widely applicable clinical standards. , , The current study identified 23 up‐to‐date CPGs that provide recommendations for non‐pharmacological interventions in the treatment of the psychosocial symptoms of individuals diagnosed with cancer. Results revealed the academic literature search to be less efficient than the gray literature search at identifying CPGs, as evidenced by yielding fewer CPGs despite generating the highest number of relevant hits prior to screening. In contrast, the gray literature search yielded a greater number of CPGs missed by the academic search and identified all but one of the psychosocial oncology CPGs. In terms of the different gray literature search methods, the targeted Internet search was the most effective, identifying the greatest number of overall and unique CPGs. This suggests that clinicians looking for psychosocial oncology CPGs do not need to conduct a thorough, time‐consuming academic search; rather, they will be best served by conducting a targeted search through the websites of key organizations and special interest groups. The efficiency of this search method is especially important given its widespread accessibility and time effectiveness. However, our findings also highlight some key concerns regarding the current state of psychosocial oncology CPGs as evidence‐based tools. The accessibility of CPGs through the gray literature search is encouraging when considering that clinicians report more using informal search methods to find information and tools more frequently than academic databases. However, an Internet search alone may not be sufficient. Our academic search identified several CPGs not detected by the gray literature search, meaning that even effective informal search methods may fail to identify relevant CPGs identifiable through a complementary database search. Yet, there are several barriers to conducting this type of search. Due to the high number of records generated, it would require a significant investment of time on the part of clinicians, who often face high work demands and lack time to conduct such a search. Further, even if clinicians were able and willing to conduct a database search, they are more likely to run the risk of identifying outdated versions of existing CPGs. Previously, clinicians could refer to the National Guidelines Clearinghouse (NGC)—a federally funded resource – which served as a repository for all CPGs. However, NGC is no longer operative due to loss funding in 2018. Since then, no new online repository for CPGs has been established. An online hub for all currently available CPGs, including psychosocial oncology CPGs, would play an essential role in addressing these barriers and ensuring ease of access for clinicians to a variety of CPGs ranging in focus (e.g., targeting a specific symptom or providing more global care recommendations), organization of origin, country of development, and methods of assessing evidence. The limited availability of up‐to‐date CPGs through peer‐reviewed sources also raises concerns about their quality. The peer‐review component of journal publication allows for additional independent appraisal and oversight into CPG development, providing an opportunity for improvements to the guideline content and recommendations prior to publication. Without an independent, external assessment of the quality and methodological rigor, it is unclear whether encouraging the use of these CPGs represents a step towards evidence‐based practice and optimal patient care. As such, even if many of these CPGs are easily accessible, clinicians may face a dilemma where they wish to be more evidence‐based in their practice but are unsure about the quality and/or applicability of the CPGs they find. Thus, implementation and adherence of CPGs may be well‐served by publication in a peer‐reviewed journal, given their function as a form of “quality control” for many institutions, clinicians, or other stakeholders. Yet, guideline developers may neglect to publish newly developed or updated CPGs in peer‐reviewed journals for a variety of reasons. For example, the formulation of recommendations by the guideline development team emerged from an extensive review process by content experts; peer‐reviewers and editors may be unaware of the full range of considerations and, consequently, suggested changes and edits to recommendations may fail to account for the full scope of relevant information. In addition, engaging in the peer‐review process is time‐consuming and significant time delays between each stage of submission and revision may result in the findings and recommendations of the CPG already being outdated by the time it is published. The risk of outdated guidelines is a more overarching concern. The most common reason for exclusion in our study was due to guidelines failing to be updated or revised within the past 10 years, meaning that recommendations will often fail to translate the latest empirical findings. This is coherent with previous findings that guidelines often fail to be updated, despite overarching recommendations to review CPGs for updates every 3 years. , , This limits their applicability as guides for bridging the science‐practice gap. Even a high quality and helpful CPG is no longer well‐suited for use as an evidence‐based tool if its findings are outdated, despite the large amounts of financial and non‐financial resources invested into its conception and development. The lack of updating also increases the risk that resources will be invested into the creation of new CPGs by different organizations, thus duplicating the previous work of guidelines development teams. While some government‐funded organizations have enough resources to adhere to more rigorous and structured updating procedures (e.g., the National Comprehensive Cancer Network (NCCN)), it is important to acknowledge many institutions face important barriers to the regular appraisal and updating of existing CPGs. 4.1 Limitations This CPG environmental scan includes several limitations. All identified CPGs were developed by Western organizations in high‐income countries (North America and Europe). Although there were no geographical constraints on the search, this might be a consequence of restricting search results to English language results only. Consequently, the recommendations of the CPGs identified most likely reflect the cultural norms of these countries and are designed for application within their healthcare systems and cultural contexts, which could limit their usefulness for clinicians practicing in different countries. Although various organizations have spearheaded efforts to produce CPG adaptations to address the specific needs of their regions, , the identification of such adaptation was beyond the scope of this study. There may also be additional psychosocial oncology CPGs not accessible through the methodology used herein. Clinicians working in psychosocial oncology settings may have exclusive access to CPGs through their own professional networks and resources (e.g., CPGs developed for exclusive use by employees of a particular hospital network). As such, the current findings may not reflect the psychosocial oncology CPGs most directly accessed or used by clinicians. 4.2 Future directions and conclusions Our findings suggest that psychosocial oncology CPGs address a wide range of symptoms and concerns, are accessible to clinicians, and can be found through even informal search methods. Several concerns about the utility of these CPGs as evidence‐based tools arose. Most notably, many existing psychosocial oncology CPGs fail to be updated following their initial publication and little information is available about their quality. Future research should seek to evaluate the methodological quality of currently available psychosocial oncology CPGs. In addition, our findings do not inform us on the current state of CPG use by clinicians providing psychosocial care to individuals diagnosed with cancer. Previous research found that clinicians often do not integrate evidence‐based interventions into their service delivery and report limited knowledge about CPGs as a whole. , A more in‐depth understanding of the current clinical use and implementation of psychosocial oncology CPGs would more effectively address barriers to their use and further support efforts to mobilize resources to keep CPGs up‐to‐date. Limitations This CPG environmental scan includes several limitations. All identified CPGs were developed by Western organizations in high‐income countries (North America and Europe). Although there were no geographical constraints on the search, this might be a consequence of restricting search results to English language results only. Consequently, the recommendations of the CPGs identified most likely reflect the cultural norms of these countries and are designed for application within their healthcare systems and cultural contexts, which could limit their usefulness for clinicians practicing in different countries. Although various organizations have spearheaded efforts to produce CPG adaptations to address the specific needs of their regions, , the identification of such adaptation was beyond the scope of this study. There may also be additional psychosocial oncology CPGs not accessible through the methodology used herein. Clinicians working in psychosocial oncology settings may have exclusive access to CPGs through their own professional networks and resources (e.g., CPGs developed for exclusive use by employees of a particular hospital network). As such, the current findings may not reflect the psychosocial oncology CPGs most directly accessed or used by clinicians. Future directions and conclusions Our findings suggest that psychosocial oncology CPGs address a wide range of symptoms and concerns, are accessible to clinicians, and can be found through even informal search methods. Several concerns about the utility of these CPGs as evidence‐based tools arose. Most notably, many existing psychosocial oncology CPGs fail to be updated following their initial publication and little information is available about their quality. Future research should seek to evaluate the methodological quality of currently available psychosocial oncology CPGs. In addition, our findings do not inform us on the current state of CPG use by clinicians providing psychosocial care to individuals diagnosed with cancer. Previous research found that clinicians often do not integrate evidence‐based interventions into their service delivery and report limited knowledge about CPGs as a whole. , A more in‐depth understanding of the current clinical use and implementation of psychosocial oncology CPGs would more effectively address barriers to their use and further support efforts to mobilize resources to keep CPGs up‐to‐date. Catherine Bergeron: Conceptualization (lead); data curation (lead); formal analysis (lead); methodology (lead); project administration (lead); writing – original draft (lead); writing – review and editing (lead). Michelle Azzi: Data curation (supporting); formal analysis (supporting); project administration (supporting); writing – review and editing (supporting). Adina Coroiu: Formal analysis (supporting); methodology (supporting); writing – original draft (supporting); writing – review and editing (supporting). Carmen G. Loiselle: Investigation (supporting); methodology (supporting); supervision (supporting); writing – review and editing (supporting). Martin Drapeau: Investigation (supporting); supervision (supporting); writing – review and editing (supporting). Annett Körner: Conceptualization (equal); formal analysis (supporting); funding acquisition (lead); investigation (supporting); methodology (supporting); resources (supporting); supervision (supporting); writing – original draft (supporting); writing – review and editing (supporting). The authors declare they have no conflict of interests. |
An Introduction to Artificial Intelligence in Developmental and Behavioral Pediatrics | cc465f95-ad2a-45f1-98ef-b31720003914 | 9907689 | Pediatrics[mh] | First coined – in 1956 by John McCarthy, artificial intelligence (AI) is an interdisciplinary field of computer science that involves the use of computers to develop systems able to perform tasks that are generally associated with intelligence in the intuitive sense. The rise of “big data,” alongside the development of increasingly complex algorithms and enhanced computational power and storage capabilities, has contributed to the recent surge in AI-based technologies. AI operates on a continuum, variously assisting, augmenting, or autonomizing task performance. Automating repetitive tasks, for example, may only require an “assisted” form of intelligence. On the other end of the human-machine continuum, however, an “autonomous” form of intelligence is required for machines to independently make decisions in adaptive intelligent systems. Within the health care context, AI is not envisioned as a technology that would supersede the need for skilled human clinicians. Rather, AI-based technologies will likely play pivotal roles in augmenting existing diagnostic and therapeutic toolkits to improve outcomes. While pediatricians may have limited familiarity with AI in a health care context, they are likely already making use of AI-powered technologies in their daily lives. Email spam filters, e-commerce platforms, and entertainment recommendation systems, for example, all rely on AI. Machine Learning: Underlying Principles Machine learning (ML) refers specifically to AI methodologies that incorporate an adaptive element wherein systems have the ability to “learn” using data to improve overall accuracy. At a basic level, all ML involves an input, a function (or some mathematical calculations), and an output. In ML models, the independent variables are termed inputs or features (e.g., age, gender, medical history, clinical symptom) and the dependent variable is referred to as the output label or target variable (e.g., diagnostic label, disease level, survival time). Both structured and unstructured data can be used to train ML models (see Fig. ). While statistical and ML techniques overlap, they are distinguishable by their underlying goals. Statistical learning is usually hypothesis-driven with the goal of inferring relationships between variables. ML is method-driven with the goal of building a model that makes actionable predictions. Machine Learning Workflow Clinicians familiar with the development and validation of existing behavioral screening and diagnostic tools will already have a general sense of product development workflow. Specific to the ML workflow, however, is an ability to learn from exposure to data, without the level of prespecified instructions or prior assumptions required in traditional product development. The likelihood of discovering new features and associations is therefore higher because workflow does not require the product designer to determine in advance which variables may be important. When building predictive ML models, the full data set must first be split into parts. Typically, the largest of these parts is the “training set” used for initial model training. A smaller portion of the data, the “validation set,” is then used to support hyperparameter tuning and model selection. Ultimately, the data sets help “train” the system to learn from similar patients, clinical features, or outcomes, which helps the algorithm become more accurate over time as data increase and more tests are conducted. At the end of this iterative process, a “test” set may be used to test the model's generalizability to data that was not previously seen during any prior aspect of the model's development. Prospective clinical validation studies may then be conducted to test the real-world performance of the model on data that were not previously available to model developers. Figure illustrates a typical ML workflow. Machine Learning Approaches Data type, structure, and number of features, along with the nature of the clinical questions being explored, all inform the type of ML approach that may be taken. Classical ML, which includes supervised learning and unsupervised learning , is generally applied to less complex data sets and clinical scenarios with a small number of features. Table provides a brief descriptive summary of these approaches. Neural Networks and Deep Learning While the ML techniques described above are suitable for many clinical problems, in cases in which nonlinear and complex relationships need to be mapped, networks and deep learning techniques may be more appropriate. An artificial neural network is a complex form of ML model designed to mimic how the neurons in the brain work. This is achieved through multiple layers of aggregation nodes known as neurons because they simulate the function of biological neurons with mathematical functions guiding when each node neuron would fire a signal to another. Features can be multiplied or added together repeatedly. The mathematical formulation of neural networks is such that complex nonlinear relationships can be modeled efficiently before an output layer that depends on the prediction task. Given the complexity and volume of health care data, this technique is becoming increasingly popular. Figure illustrates both simple neural network and deep learning neural network. Evaluating Model Performance The ability to explain the output of a model and assess its performance, both in the clinical context and in relation to its intended purpose, will increasingly become part of the future clinician's role. Key performance metrics relevant to ML are summarized in Table . We should note that these metrics do not always provide a straightforward interpretation of the performance of a model because they can be biased by the nature of the data; the end result should be determined by whether clinical value can be obtained from the model. The model's accuracy and threshold, along with the disease prevalence and the model's performance compared with existing “gold-standard” non-AI–informed approaches, should all be considered. Machine learning (ML) refers specifically to AI methodologies that incorporate an adaptive element wherein systems have the ability to “learn” using data to improve overall accuracy. At a basic level, all ML involves an input, a function (or some mathematical calculations), and an output. In ML models, the independent variables are termed inputs or features (e.g., age, gender, medical history, clinical symptom) and the dependent variable is referred to as the output label or target variable (e.g., diagnostic label, disease level, survival time). Both structured and unstructured data can be used to train ML models (see Fig. ). While statistical and ML techniques overlap, they are distinguishable by their underlying goals. Statistical learning is usually hypothesis-driven with the goal of inferring relationships between variables. ML is method-driven with the goal of building a model that makes actionable predictions. Clinicians familiar with the development and validation of existing behavioral screening and diagnostic tools will already have a general sense of product development workflow. Specific to the ML workflow, however, is an ability to learn from exposure to data, without the level of prespecified instructions or prior assumptions required in traditional product development. The likelihood of discovering new features and associations is therefore higher because workflow does not require the product designer to determine in advance which variables may be important. When building predictive ML models, the full data set must first be split into parts. Typically, the largest of these parts is the “training set” used for initial model training. A smaller portion of the data, the “validation set,” is then used to support hyperparameter tuning and model selection. Ultimately, the data sets help “train” the system to learn from similar patients, clinical features, or outcomes, which helps the algorithm become more accurate over time as data increase and more tests are conducted. At the end of this iterative process, a “test” set may be used to test the model's generalizability to data that was not previously seen during any prior aspect of the model's development. Prospective clinical validation studies may then be conducted to test the real-world performance of the model on data that were not previously available to model developers. Figure illustrates a typical ML workflow. Data type, structure, and number of features, along with the nature of the clinical questions being explored, all inform the type of ML approach that may be taken. Classical ML, which includes supervised learning and unsupervised learning , is generally applied to less complex data sets and clinical scenarios with a small number of features. Table provides a brief descriptive summary of these approaches. Neural Networks and Deep Learning While the ML techniques described above are suitable for many clinical problems, in cases in which nonlinear and complex relationships need to be mapped, networks and deep learning techniques may be more appropriate. An artificial neural network is a complex form of ML model designed to mimic how the neurons in the brain work. This is achieved through multiple layers of aggregation nodes known as neurons because they simulate the function of biological neurons with mathematical functions guiding when each node neuron would fire a signal to another. Features can be multiplied or added together repeatedly. The mathematical formulation of neural networks is such that complex nonlinear relationships can be modeled efficiently before an output layer that depends on the prediction task. Given the complexity and volume of health care data, this technique is becoming increasingly popular. Figure illustrates both simple neural network and deep learning neural network. While the ML techniques described above are suitable for many clinical problems, in cases in which nonlinear and complex relationships need to be mapped, networks and deep learning techniques may be more appropriate. An artificial neural network is a complex form of ML model designed to mimic how the neurons in the brain work. This is achieved through multiple layers of aggregation nodes known as neurons because they simulate the function of biological neurons with mathematical functions guiding when each node neuron would fire a signal to another. Features can be multiplied or added together repeatedly. The mathematical formulation of neural networks is such that complex nonlinear relationships can be modeled efficiently before an output layer that depends on the prediction task. Given the complexity and volume of health care data, this technique is becoming increasingly popular. Figure illustrates both simple neural network and deep learning neural network. The ability to explain the output of a model and assess its performance, both in the clinical context and in relation to its intended purpose, will increasingly become part of the future clinician's role. Key performance metrics relevant to ML are summarized in Table . We should note that these metrics do not always provide a straightforward interpretation of the performance of a model because they can be biased by the nature of the data; the end result should be determined by whether clinical value can be obtained from the model. The model's accuracy and threshold, along with the disease prevalence and the model's performance compared with existing “gold-standard” non-AI–informed approaches, should all be considered. Opportunities In the field of developmental and behavioral pediatrics, artificial intelligence (AI) can assist with a broad array of tasks including diagnosis, risk prediction and stratification, treatment, administration, and regulation. Core analytic tasks that may fall under the AI umbrella are depicted in Figure . Enhancing Clinical Decision-Making, Risk Prediction, and Diagnosis Massive and constantly expanding quantities of medical data including electronic medical records (EMRs), high-resolution medical images, public health data sets, genomics, and wearables have exceeded the limits of human analysis. AI offers opportunities to harness and derive clinically meaningful insights from this ever-growing volume of health care data in ways that traditional analytic techniques cannot. Neural networks, for example, trained on much larger quantities of data than any single clinician could possibly be exposed to in the course of their career, can support the identification of subtle nonlinear data patterns. Such approaches promise to significantly enhance interpretation of medical scans, pathology slides, and other imaging data that rely on pattern recognition. By observing subtle nonlinear correlations in the data, , machine learning (ML) approaches have potential to augment risk prediction and diagnostic processes and ultimately provide an enhanced quality of care. A number of studies within the field of developmental and behavioral pediatrics demonstrate the potential for AI to enhance risk prediction practices. ML techniques were used, for example, to analyze the EMRs of over a million individuals to identify risk for Fragile X based on associations with comorbid medical conditions. The resulting predictive model was able to flag Fragile X cases 5 years earlier than current practice without relying on genetic or familial data. A similar approach was taken to predict autism spectrum disorder (ASD) risk based on high-prevalence comorbidity clusters detected in EMRs. Digital biomarkers inferred from deep comorbidity patterns have also been leveraged to develop an ASD comorbid risk score with a superior predictive performance than some questionnaire-based screening tools. Automated speech analysis was combined with ML in another study to accurately predict risk for psychosis onset in clinically vulnerable youths. In this proof-of-principle study, speech features from transcripts of interviews with at-risk youth were fed into a classification algorithm to assess their predictive value for psychosis. Research has also highlighted the potential for AI to streamline diagnostic pathways for conditions such as ASD and attention-deficit/hyperactivity disorder (ADHD). Such research is promising given that streamlining diagnosis could allow for earlier treatment initiation during the critical neurodevelopmental window. For example, researchers have used ML to examine behavioral phenotypes of children with ASD with a high rate of accuracy and to shorten the time for observation-based screening and diagnosis. – In addition, ML has been used to differentiate between ASD and ADHD with high accuracy using a small number of measured behaviors. Research has also explored facial expression analysis based on dynamic deep learning and 3D behavior analysis to detect and distinguish between ADHD, ASD, and comorbid ADHD/ASD presentations. Expanding Treatment Options Along with risk prediction and diagnostics, AI holds potential to enhance treatment delivery in the field of developmental and behavioral pediatrics. AI robots have been used in a number of intervention studies to enhance social skill acquisition and spontaneous language development in children with ASD. , The potential utility of AI-enabled technologies to treat disruptive behaviors and mood and anxiety disorders in children is also being explored. In the field of ADHD, a number of emerging technologies show promise to infer behaviors that can then be used to tailor feedback to enhance self-regulation. For example, in 1 study, a neural network was used to learn behavioral intervention delivery techniques based on human demonstrations. The trained network then enabled a robot to autonomously deliver a similar intervention to children with ADHD. As health data infrastructure expands in size and sophistication, future AI technologies could potentially support increasingly personalized treatment options. Once comprehensive and integrated biologic, anatomic, physiologic, environmental, socioeconomic, behavioral, and pharmacogenomic patient data become routinely available, AI-based nearest-neighbor analysis could be used to identify “digital twins.” Patients with similar genomic and clinical features could be identified and clustered, for example, to allow for highly targeted treatments. Such approaches, while currently largely theoretical, could help predict therapeutic and adverse medication responses more accurately and also form an evidence base for personalized treatment pathways. Streamlining Care and Enhancing Workplace Efficiency Artificial intelligence algorithms have potential to automate many arduous administrative tasks, thereby streamlining care pathways and freeing clinicians to spend more time with patients. Natural language processing solutions, for example, are being developed to decrease reliance on human scribes in clinical encounters. Such approaches may be used to process and transform clinical-free text into meaningful structured outcomes, automate some documentation practices through text summarization, and scan text-based reports to support accurate and rapid diagnostic recommendations. Natural language processing solutions may prove particularly valuable in fields such as child and adolescent psychiatry and developmental and behavioral pediatrics that are extremely text-heavy. Developmental behavioral assessments often involve text-heavy tasks such as documentation of complex patient histories, results of extensive testing, and collateral history from multiple informants. Given the current national US shortage of child and adolescent psychiatrists with a median of 11 psychiatrists per 100,000 children, AI-based approaches that streamline and automate administrative tasks seem particularly promising. AI-assisted image interpretation has also shown potential to increase workplace productivity and provide considerable cost savings over current practice. , Techniques to streamline medical research and drug discovery by using natural language processing to rapidly scan biomedical literature and data mine molecular structures are also being developed. Promoting Equity and Access Artificial intelligence has potential to address several bias and access disparities apparent in existing care models. While access to developmental and behavioral specialists is extremely limited in much of the world, it is estimated that over 50% of the global population has access to a smartphone. Digital AI–based diagnostic and treatment platforms could thus potentially expand access to underserved and geographically remote populations. Thoughtful use of AI may also help to address racial, socioeconomic, and gender biases. In the field of developmental and behavioral pediatrics, for example, it has been noted that despite ASD prevalence rates being roughly equal across racial/ethnic and socioeconomic groups, human clinicians are more likely to diagnose Black, Latinx, and Asian children, as well as children from low-income families, at a later date than White children and children with a higher socioeconomic status. By integrating and training on large racial-conscious and gender-conscious data sets, AI algorithms can assess thousands of traits and features and build on the findings to assist clinicians in making more accurate, timely, and less biased ASD diagnoses. Challenges While AI presents multiple opportunities to the field of developmental and behavioral pediatrics, to date, a very few AI-based technologies have been broadly integrated into clinical practice. Challenges to the widespread deployment of AI in health care settings, along with potential solutions, are outlined below. Data Bias Any AI algorithm is only as good as the data from which it was derived. Simply put, the performance of the algorithm and the quality of its prediction are dependent on the quality of the data supplied. If the data are biased, imbalanced, or otherwise an incomplete representation of the target group, the derivative model's generalizability will be limited. A class imbalance problem can occur, for example, in cases in which the total number of 1 class of data (i.e., “girls” or “disease positive status”) is far less than the total number of another class of data (i.e., “boys” or “disease negative status”). These biases can sometimes be identified and addressed through techniques such as over- or undersampling, but at other times with “black-box” learning, it is harder to detect and fix the bias in the algorithm(s). Representative populations for most conditions, including in pediatrics, are not a homogeneous group. Thus, it is essential that the data sets used to train these models include balanced data with diverse representations of the clinical symptoms across gender, age, and race in order for the model to be generalizable. Without such safeguards, models may, in fact, perpetuate or amplify stereotypes or biases. , Data Sharing and Privacy Concerns To robustly train an algorithm, sufficient data are required, yet pediatric data sets can be limited by small sample sizes, especially when split by age group. For the case of rare pediatric diseases, extremely low prevalence rates mean the amount and type of data available to train a model on very limited. Creative use of deep learning techniques such as generative adversarial networks may be required in such cases to counteract the lack of data. Generative adversarial networks pit one neural network against another for the purpose of generating synthetic (yet realistic) data to support a variety of tasks such as image and voice generation. While such techniques can be extremely useful when appropriately applied, overreliance on synthetic data also comes with its own set of risks. Multiple data sets may also be combined to produce a sufficient volume of data for model training. However, combining data sets presents its own set of challenges including data privacy and ownership issues and difficulties integrating data with heterogeneous features. Federated learning is an emerging ML technique with potential to address some of these data sharing and privacy concerns. Federated learning allows algorithms to train across many decentralized servers or edge devices, exchanging parameters (i.e., the models' weights and biases) without explicitly exchanging the data samples themselves. This technique obviates many of the privacy issues engendered by uploading highly sensitive health data from different sources onto a single server. Algorithmic Transparency and Explainability A lack of transparency in certain types of ML algorithms such as deep neural networks has raised concern about their clinical trustworthiness. Many models used to analyze images and text, for example, include levels of complexity and multidimensionality that exceed intuitive understanding or interpretation. In cases in which a clinician is unable to understand how the algorithm produces an output, should the algorithm be relied on as part of their clinical decision-making process? Such concerns have led to calls for algorithmic deconvolution before use in health care settings. Other researchers argue, however, that current approaches to explainability disregard the reality that local explanations can be unreliable or too superficial to be meaningful and that rigorous model validation before deployment may be a more important marker of trustworthiness. Consumer and Clinician Preparedness If patients and clinicians mistrust AI-based technologies, and/or lack sufficient training to understand, in broad terms, how they function, clinical adoption may be delayed. As with all new tools, implementation matters, and discipline is required to ensure safe deployment of AI-based devices without loss of clinician skill. Overreliance on AI-based imaging at the expense of history and physical examination, for example, should be avoided. Patient reservations that will require consideration include safety, cost, choice, data bias, and data security concerns. While the American Medical Association has called for research into how AI should be addressed in medical education, current medical training lacks a consistent approach to AI education, and key licensing examinations do not test on this content. Clinicians and medical students alike have identified knowledge gaps and reservations regarding the use of AI in health care. – A number of preliminary frameworks for integrating AI curriculum into medical training have been proposed , , ; however, additional research is urgently required to develop and then integrate standardized AI content into medical training pathways. Ethical Ambiguities From an ethical standpoint, users of this technology must consider the direct impact and unintended consequences of AI implementation in general as well as specific implications within the clinical context. A number of well-publicized non–health-related cases have illustrated AI data privacy concerns, along with the ethically problematic potential for AI to amplify social, racial, and gender biases. There are also ethical concerns around the magnitude of harm that could occur if an ML algorithm, deployed clinically at scale, were to malfunction; associated impacts could far exceed the harm caused by a single clinician's malpractice. The use of AI for clinical decision-making also raises questions of accountability, such as who is liable if unintended consequences result from use of the technology (e.g., missed diagnosis), or what course of action an autonomous therapy chatbot might take if it detects speech patterns indicative of risk for self-harm. Ethical AI frameworks addressing such concerns are under development, , and researchers are calling for AI technologies to undergo robust simulation, validation, and prospective scrutiny before clinical adoption. Regulatory and Payment Barriers Notwithstanding data, provider adoption, and ethical safeguards, AI technologies face several systematic challenges to be readily implemented into clinical practice, including regulatory, interoperability with EMRs and data exchange, and payment barriers. Given that AI devices can learn from data and alter their algorithms accordingly, traditional medical device regulatory frameworks might not be sufficient. As a result, the Food and Drug Administration has developed a proposed regulatory framework that includes a potential “Predetermined Change Control Plan” for premarket submissions, including “Software Pre-Specifications” and an “Algorithm Change Protocol,” to address the iterative nature of AI/ML-based Software as a Medical Device technologies. Health care organizations and practices will also need to establish a data infrastructure and privacy policy for data that are stored across multiple servers and sources (e.g., medical records, health sensors, medical devices, etc). Development of new digital medical software and devices that use AI are likely to outpace the current health care payment structure. New billing codes associated with new treatments and procedures require formal approvals by national organizations with subsequent adoption by insurances, both public and private. This process can take many months to years. To facilitate provider and patient adoption of new AI technologies which may improve quality of care, streamlined development of billing codes for technologies using AI should be developed. In the field of developmental and behavioral pediatrics, artificial intelligence (AI) can assist with a broad array of tasks including diagnosis, risk prediction and stratification, treatment, administration, and regulation. Core analytic tasks that may fall under the AI umbrella are depicted in Figure . Enhancing Clinical Decision-Making, Risk Prediction, and Diagnosis Massive and constantly expanding quantities of medical data including electronic medical records (EMRs), high-resolution medical images, public health data sets, genomics, and wearables have exceeded the limits of human analysis. AI offers opportunities to harness and derive clinically meaningful insights from this ever-growing volume of health care data in ways that traditional analytic techniques cannot. Neural networks, for example, trained on much larger quantities of data than any single clinician could possibly be exposed to in the course of their career, can support the identification of subtle nonlinear data patterns. Such approaches promise to significantly enhance interpretation of medical scans, pathology slides, and other imaging data that rely on pattern recognition. By observing subtle nonlinear correlations in the data, , machine learning (ML) approaches have potential to augment risk prediction and diagnostic processes and ultimately provide an enhanced quality of care. A number of studies within the field of developmental and behavioral pediatrics demonstrate the potential for AI to enhance risk prediction practices. ML techniques were used, for example, to analyze the EMRs of over a million individuals to identify risk for Fragile X based on associations with comorbid medical conditions. The resulting predictive model was able to flag Fragile X cases 5 years earlier than current practice without relying on genetic or familial data. A similar approach was taken to predict autism spectrum disorder (ASD) risk based on high-prevalence comorbidity clusters detected in EMRs. Digital biomarkers inferred from deep comorbidity patterns have also been leveraged to develop an ASD comorbid risk score with a superior predictive performance than some questionnaire-based screening tools. Automated speech analysis was combined with ML in another study to accurately predict risk for psychosis onset in clinically vulnerable youths. In this proof-of-principle study, speech features from transcripts of interviews with at-risk youth were fed into a classification algorithm to assess their predictive value for psychosis. Research has also highlighted the potential for AI to streamline diagnostic pathways for conditions such as ASD and attention-deficit/hyperactivity disorder (ADHD). Such research is promising given that streamlining diagnosis could allow for earlier treatment initiation during the critical neurodevelopmental window. For example, researchers have used ML to examine behavioral phenotypes of children with ASD with a high rate of accuracy and to shorten the time for observation-based screening and diagnosis. – In addition, ML has been used to differentiate between ASD and ADHD with high accuracy using a small number of measured behaviors. Research has also explored facial expression analysis based on dynamic deep learning and 3D behavior analysis to detect and distinguish between ADHD, ASD, and comorbid ADHD/ASD presentations. Expanding Treatment Options Along with risk prediction and diagnostics, AI holds potential to enhance treatment delivery in the field of developmental and behavioral pediatrics. AI robots have been used in a number of intervention studies to enhance social skill acquisition and spontaneous language development in children with ASD. , The potential utility of AI-enabled technologies to treat disruptive behaviors and mood and anxiety disorders in children is also being explored. In the field of ADHD, a number of emerging technologies show promise to infer behaviors that can then be used to tailor feedback to enhance self-regulation. For example, in 1 study, a neural network was used to learn behavioral intervention delivery techniques based on human demonstrations. The trained network then enabled a robot to autonomously deliver a similar intervention to children with ADHD. As health data infrastructure expands in size and sophistication, future AI technologies could potentially support increasingly personalized treatment options. Once comprehensive and integrated biologic, anatomic, physiologic, environmental, socioeconomic, behavioral, and pharmacogenomic patient data become routinely available, AI-based nearest-neighbor analysis could be used to identify “digital twins.” Patients with similar genomic and clinical features could be identified and clustered, for example, to allow for highly targeted treatments. Such approaches, while currently largely theoretical, could help predict therapeutic and adverse medication responses more accurately and also form an evidence base for personalized treatment pathways. Streamlining Care and Enhancing Workplace Efficiency Artificial intelligence algorithms have potential to automate many arduous administrative tasks, thereby streamlining care pathways and freeing clinicians to spend more time with patients. Natural language processing solutions, for example, are being developed to decrease reliance on human scribes in clinical encounters. Such approaches may be used to process and transform clinical-free text into meaningful structured outcomes, automate some documentation practices through text summarization, and scan text-based reports to support accurate and rapid diagnostic recommendations. Natural language processing solutions may prove particularly valuable in fields such as child and adolescent psychiatry and developmental and behavioral pediatrics that are extremely text-heavy. Developmental behavioral assessments often involve text-heavy tasks such as documentation of complex patient histories, results of extensive testing, and collateral history from multiple informants. Given the current national US shortage of child and adolescent psychiatrists with a median of 11 psychiatrists per 100,000 children, AI-based approaches that streamline and automate administrative tasks seem particularly promising. AI-assisted image interpretation has also shown potential to increase workplace productivity and provide considerable cost savings over current practice. , Techniques to streamline medical research and drug discovery by using natural language processing to rapidly scan biomedical literature and data mine molecular structures are also being developed. Promoting Equity and Access Artificial intelligence has potential to address several bias and access disparities apparent in existing care models. While access to developmental and behavioral specialists is extremely limited in much of the world, it is estimated that over 50% of the global population has access to a smartphone. Digital AI–based diagnostic and treatment platforms could thus potentially expand access to underserved and geographically remote populations. Thoughtful use of AI may also help to address racial, socioeconomic, and gender biases. In the field of developmental and behavioral pediatrics, for example, it has been noted that despite ASD prevalence rates being roughly equal across racial/ethnic and socioeconomic groups, human clinicians are more likely to diagnose Black, Latinx, and Asian children, as well as children from low-income families, at a later date than White children and children with a higher socioeconomic status. By integrating and training on large racial-conscious and gender-conscious data sets, AI algorithms can assess thousands of traits and features and build on the findings to assist clinicians in making more accurate, timely, and less biased ASD diagnoses. Massive and constantly expanding quantities of medical data including electronic medical records (EMRs), high-resolution medical images, public health data sets, genomics, and wearables have exceeded the limits of human analysis. AI offers opportunities to harness and derive clinically meaningful insights from this ever-growing volume of health care data in ways that traditional analytic techniques cannot. Neural networks, for example, trained on much larger quantities of data than any single clinician could possibly be exposed to in the course of their career, can support the identification of subtle nonlinear data patterns. Such approaches promise to significantly enhance interpretation of medical scans, pathology slides, and other imaging data that rely on pattern recognition. By observing subtle nonlinear correlations in the data, , machine learning (ML) approaches have potential to augment risk prediction and diagnostic processes and ultimately provide an enhanced quality of care. A number of studies within the field of developmental and behavioral pediatrics demonstrate the potential for AI to enhance risk prediction practices. ML techniques were used, for example, to analyze the EMRs of over a million individuals to identify risk for Fragile X based on associations with comorbid medical conditions. The resulting predictive model was able to flag Fragile X cases 5 years earlier than current practice without relying on genetic or familial data. A similar approach was taken to predict autism spectrum disorder (ASD) risk based on high-prevalence comorbidity clusters detected in EMRs. Digital biomarkers inferred from deep comorbidity patterns have also been leveraged to develop an ASD comorbid risk score with a superior predictive performance than some questionnaire-based screening tools. Automated speech analysis was combined with ML in another study to accurately predict risk for psychosis onset in clinically vulnerable youths. In this proof-of-principle study, speech features from transcripts of interviews with at-risk youth were fed into a classification algorithm to assess their predictive value for psychosis. Research has also highlighted the potential for AI to streamline diagnostic pathways for conditions such as ASD and attention-deficit/hyperactivity disorder (ADHD). Such research is promising given that streamlining diagnosis could allow for earlier treatment initiation during the critical neurodevelopmental window. For example, researchers have used ML to examine behavioral phenotypes of children with ASD with a high rate of accuracy and to shorten the time for observation-based screening and diagnosis. – In addition, ML has been used to differentiate between ASD and ADHD with high accuracy using a small number of measured behaviors. Research has also explored facial expression analysis based on dynamic deep learning and 3D behavior analysis to detect and distinguish between ADHD, ASD, and comorbid ADHD/ASD presentations. Along with risk prediction and diagnostics, AI holds potential to enhance treatment delivery in the field of developmental and behavioral pediatrics. AI robots have been used in a number of intervention studies to enhance social skill acquisition and spontaneous language development in children with ASD. , The potential utility of AI-enabled technologies to treat disruptive behaviors and mood and anxiety disorders in children is also being explored. In the field of ADHD, a number of emerging technologies show promise to infer behaviors that can then be used to tailor feedback to enhance self-regulation. For example, in 1 study, a neural network was used to learn behavioral intervention delivery techniques based on human demonstrations. The trained network then enabled a robot to autonomously deliver a similar intervention to children with ADHD. As health data infrastructure expands in size and sophistication, future AI technologies could potentially support increasingly personalized treatment options. Once comprehensive and integrated biologic, anatomic, physiologic, environmental, socioeconomic, behavioral, and pharmacogenomic patient data become routinely available, AI-based nearest-neighbor analysis could be used to identify “digital twins.” Patients with similar genomic and clinical features could be identified and clustered, for example, to allow for highly targeted treatments. Such approaches, while currently largely theoretical, could help predict therapeutic and adverse medication responses more accurately and also form an evidence base for personalized treatment pathways. Artificial intelligence algorithms have potential to automate many arduous administrative tasks, thereby streamlining care pathways and freeing clinicians to spend more time with patients. Natural language processing solutions, for example, are being developed to decrease reliance on human scribes in clinical encounters. Such approaches may be used to process and transform clinical-free text into meaningful structured outcomes, automate some documentation practices through text summarization, and scan text-based reports to support accurate and rapid diagnostic recommendations. Natural language processing solutions may prove particularly valuable in fields such as child and adolescent psychiatry and developmental and behavioral pediatrics that are extremely text-heavy. Developmental behavioral assessments often involve text-heavy tasks such as documentation of complex patient histories, results of extensive testing, and collateral history from multiple informants. Given the current national US shortage of child and adolescent psychiatrists with a median of 11 psychiatrists per 100,000 children, AI-based approaches that streamline and automate administrative tasks seem particularly promising. AI-assisted image interpretation has also shown potential to increase workplace productivity and provide considerable cost savings over current practice. , Techniques to streamline medical research and drug discovery by using natural language processing to rapidly scan biomedical literature and data mine molecular structures are also being developed. Artificial intelligence has potential to address several bias and access disparities apparent in existing care models. While access to developmental and behavioral specialists is extremely limited in much of the world, it is estimated that over 50% of the global population has access to a smartphone. Digital AI–based diagnostic and treatment platforms could thus potentially expand access to underserved and geographically remote populations. Thoughtful use of AI may also help to address racial, socioeconomic, and gender biases. In the field of developmental and behavioral pediatrics, for example, it has been noted that despite ASD prevalence rates being roughly equal across racial/ethnic and socioeconomic groups, human clinicians are more likely to diagnose Black, Latinx, and Asian children, as well as children from low-income families, at a later date than White children and children with a higher socioeconomic status. By integrating and training on large racial-conscious and gender-conscious data sets, AI algorithms can assess thousands of traits and features and build on the findings to assist clinicians in making more accurate, timely, and less biased ASD diagnoses. While AI presents multiple opportunities to the field of developmental and behavioral pediatrics, to date, a very few AI-based technologies have been broadly integrated into clinical practice. Challenges to the widespread deployment of AI in health care settings, along with potential solutions, are outlined below. Data Bias Any AI algorithm is only as good as the data from which it was derived. Simply put, the performance of the algorithm and the quality of its prediction are dependent on the quality of the data supplied. If the data are biased, imbalanced, or otherwise an incomplete representation of the target group, the derivative model's generalizability will be limited. A class imbalance problem can occur, for example, in cases in which the total number of 1 class of data (i.e., “girls” or “disease positive status”) is far less than the total number of another class of data (i.e., “boys” or “disease negative status”). These biases can sometimes be identified and addressed through techniques such as over- or undersampling, but at other times with “black-box” learning, it is harder to detect and fix the bias in the algorithm(s). Representative populations for most conditions, including in pediatrics, are not a homogeneous group. Thus, it is essential that the data sets used to train these models include balanced data with diverse representations of the clinical symptoms across gender, age, and race in order for the model to be generalizable. Without such safeguards, models may, in fact, perpetuate or amplify stereotypes or biases. , Data Sharing and Privacy Concerns To robustly train an algorithm, sufficient data are required, yet pediatric data sets can be limited by small sample sizes, especially when split by age group. For the case of rare pediatric diseases, extremely low prevalence rates mean the amount and type of data available to train a model on very limited. Creative use of deep learning techniques such as generative adversarial networks may be required in such cases to counteract the lack of data. Generative adversarial networks pit one neural network against another for the purpose of generating synthetic (yet realistic) data to support a variety of tasks such as image and voice generation. While such techniques can be extremely useful when appropriately applied, overreliance on synthetic data also comes with its own set of risks. Multiple data sets may also be combined to produce a sufficient volume of data for model training. However, combining data sets presents its own set of challenges including data privacy and ownership issues and difficulties integrating data with heterogeneous features. Federated learning is an emerging ML technique with potential to address some of these data sharing and privacy concerns. Federated learning allows algorithms to train across many decentralized servers or edge devices, exchanging parameters (i.e., the models' weights and biases) without explicitly exchanging the data samples themselves. This technique obviates many of the privacy issues engendered by uploading highly sensitive health data from different sources onto a single server. Algorithmic Transparency and Explainability A lack of transparency in certain types of ML algorithms such as deep neural networks has raised concern about their clinical trustworthiness. Many models used to analyze images and text, for example, include levels of complexity and multidimensionality that exceed intuitive understanding or interpretation. In cases in which a clinician is unable to understand how the algorithm produces an output, should the algorithm be relied on as part of their clinical decision-making process? Such concerns have led to calls for algorithmic deconvolution before use in health care settings. Other researchers argue, however, that current approaches to explainability disregard the reality that local explanations can be unreliable or too superficial to be meaningful and that rigorous model validation before deployment may be a more important marker of trustworthiness. Consumer and Clinician Preparedness If patients and clinicians mistrust AI-based technologies, and/or lack sufficient training to understand, in broad terms, how they function, clinical adoption may be delayed. As with all new tools, implementation matters, and discipline is required to ensure safe deployment of AI-based devices without loss of clinician skill. Overreliance on AI-based imaging at the expense of history and physical examination, for example, should be avoided. Patient reservations that will require consideration include safety, cost, choice, data bias, and data security concerns. While the American Medical Association has called for research into how AI should be addressed in medical education, current medical training lacks a consistent approach to AI education, and key licensing examinations do not test on this content. Clinicians and medical students alike have identified knowledge gaps and reservations regarding the use of AI in health care. – A number of preliminary frameworks for integrating AI curriculum into medical training have been proposed , , ; however, additional research is urgently required to develop and then integrate standardized AI content into medical training pathways. Ethical Ambiguities From an ethical standpoint, users of this technology must consider the direct impact and unintended consequences of AI implementation in general as well as specific implications within the clinical context. A number of well-publicized non–health-related cases have illustrated AI data privacy concerns, along with the ethically problematic potential for AI to amplify social, racial, and gender biases. There are also ethical concerns around the magnitude of harm that could occur if an ML algorithm, deployed clinically at scale, were to malfunction; associated impacts could far exceed the harm caused by a single clinician's malpractice. The use of AI for clinical decision-making also raises questions of accountability, such as who is liable if unintended consequences result from use of the technology (e.g., missed diagnosis), or what course of action an autonomous therapy chatbot might take if it detects speech patterns indicative of risk for self-harm. Ethical AI frameworks addressing such concerns are under development, , and researchers are calling for AI technologies to undergo robust simulation, validation, and prospective scrutiny before clinical adoption. Regulatory and Payment Barriers Notwithstanding data, provider adoption, and ethical safeguards, AI technologies face several systematic challenges to be readily implemented into clinical practice, including regulatory, interoperability with EMRs and data exchange, and payment barriers. Given that AI devices can learn from data and alter their algorithms accordingly, traditional medical device regulatory frameworks might not be sufficient. As a result, the Food and Drug Administration has developed a proposed regulatory framework that includes a potential “Predetermined Change Control Plan” for premarket submissions, including “Software Pre-Specifications” and an “Algorithm Change Protocol,” to address the iterative nature of AI/ML-based Software as a Medical Device technologies. Health care organizations and practices will also need to establish a data infrastructure and privacy policy for data that are stored across multiple servers and sources (e.g., medical records, health sensors, medical devices, etc). Development of new digital medical software and devices that use AI are likely to outpace the current health care payment structure. New billing codes associated with new treatments and procedures require formal approvals by national organizations with subsequent adoption by insurances, both public and private. This process can take many months to years. To facilitate provider and patient adoption of new AI technologies which may improve quality of care, streamlined development of billing codes for technologies using AI should be developed. Any AI algorithm is only as good as the data from which it was derived. Simply put, the performance of the algorithm and the quality of its prediction are dependent on the quality of the data supplied. If the data are biased, imbalanced, or otherwise an incomplete representation of the target group, the derivative model's generalizability will be limited. A class imbalance problem can occur, for example, in cases in which the total number of 1 class of data (i.e., “girls” or “disease positive status”) is far less than the total number of another class of data (i.e., “boys” or “disease negative status”). These biases can sometimes be identified and addressed through techniques such as over- or undersampling, but at other times with “black-box” learning, it is harder to detect and fix the bias in the algorithm(s). Representative populations for most conditions, including in pediatrics, are not a homogeneous group. Thus, it is essential that the data sets used to train these models include balanced data with diverse representations of the clinical symptoms across gender, age, and race in order for the model to be generalizable. Without such safeguards, models may, in fact, perpetuate or amplify stereotypes or biases. , To robustly train an algorithm, sufficient data are required, yet pediatric data sets can be limited by small sample sizes, especially when split by age group. For the case of rare pediatric diseases, extremely low prevalence rates mean the amount and type of data available to train a model on very limited. Creative use of deep learning techniques such as generative adversarial networks may be required in such cases to counteract the lack of data. Generative adversarial networks pit one neural network against another for the purpose of generating synthetic (yet realistic) data to support a variety of tasks such as image and voice generation. While such techniques can be extremely useful when appropriately applied, overreliance on synthetic data also comes with its own set of risks. Multiple data sets may also be combined to produce a sufficient volume of data for model training. However, combining data sets presents its own set of challenges including data privacy and ownership issues and difficulties integrating data with heterogeneous features. Federated learning is an emerging ML technique with potential to address some of these data sharing and privacy concerns. Federated learning allows algorithms to train across many decentralized servers or edge devices, exchanging parameters (i.e., the models' weights and biases) without explicitly exchanging the data samples themselves. This technique obviates many of the privacy issues engendered by uploading highly sensitive health data from different sources onto a single server. A lack of transparency in certain types of ML algorithms such as deep neural networks has raised concern about their clinical trustworthiness. Many models used to analyze images and text, for example, include levels of complexity and multidimensionality that exceed intuitive understanding or interpretation. In cases in which a clinician is unable to understand how the algorithm produces an output, should the algorithm be relied on as part of their clinical decision-making process? Such concerns have led to calls for algorithmic deconvolution before use in health care settings. Other researchers argue, however, that current approaches to explainability disregard the reality that local explanations can be unreliable or too superficial to be meaningful and that rigorous model validation before deployment may be a more important marker of trustworthiness. If patients and clinicians mistrust AI-based technologies, and/or lack sufficient training to understand, in broad terms, how they function, clinical adoption may be delayed. As with all new tools, implementation matters, and discipline is required to ensure safe deployment of AI-based devices without loss of clinician skill. Overreliance on AI-based imaging at the expense of history and physical examination, for example, should be avoided. Patient reservations that will require consideration include safety, cost, choice, data bias, and data security concerns. While the American Medical Association has called for research into how AI should be addressed in medical education, current medical training lacks a consistent approach to AI education, and key licensing examinations do not test on this content. Clinicians and medical students alike have identified knowledge gaps and reservations regarding the use of AI in health care. – A number of preliminary frameworks for integrating AI curriculum into medical training have been proposed , , ; however, additional research is urgently required to develop and then integrate standardized AI content into medical training pathways. From an ethical standpoint, users of this technology must consider the direct impact and unintended consequences of AI implementation in general as well as specific implications within the clinical context. A number of well-publicized non–health-related cases have illustrated AI data privacy concerns, along with the ethically problematic potential for AI to amplify social, racial, and gender biases. There are also ethical concerns around the magnitude of harm that could occur if an ML algorithm, deployed clinically at scale, were to malfunction; associated impacts could far exceed the harm caused by a single clinician's malpractice. The use of AI for clinical decision-making also raises questions of accountability, such as who is liable if unintended consequences result from use of the technology (e.g., missed diagnosis), or what course of action an autonomous therapy chatbot might take if it detects speech patterns indicative of risk for self-harm. Ethical AI frameworks addressing such concerns are under development, , and researchers are calling for AI technologies to undergo robust simulation, validation, and prospective scrutiny before clinical adoption. Notwithstanding data, provider adoption, and ethical safeguards, AI technologies face several systematic challenges to be readily implemented into clinical practice, including regulatory, interoperability with EMRs and data exchange, and payment barriers. Given that AI devices can learn from data and alter their algorithms accordingly, traditional medical device regulatory frameworks might not be sufficient. As a result, the Food and Drug Administration has developed a proposed regulatory framework that includes a potential “Predetermined Change Control Plan” for premarket submissions, including “Software Pre-Specifications” and an “Algorithm Change Protocol,” to address the iterative nature of AI/ML-based Software as a Medical Device technologies. Health care organizations and practices will also need to establish a data infrastructure and privacy policy for data that are stored across multiple servers and sources (e.g., medical records, health sensors, medical devices, etc). Development of new digital medical software and devices that use AI are likely to outpace the current health care payment structure. New billing codes associated with new treatments and procedures require formal approvals by national organizations with subsequent adoption by insurances, both public and private. This process can take many months to years. To facilitate provider and patient adoption of new AI technologies which may improve quality of care, streamlined development of billing codes for technologies using AI should be developed. Artificial intelligence (AI) in health care is not just a futuristic premise, and adoption has shifted from the “early adopter” fringe to a mainstream concept. The convergence of enhanced computational power and cloud storage solutions, increasingly sophisticated machine learning (ML) approaches and rapidly expanding volumes of digitized health care data, has ushered in this new wave of AI-based technologies. Strong economic investment in the AI health care sector, together with the growing number of AI-driven devices being granted regulatory approval, underscores the increasing role of AI in the future health care landscape. In the field of developmental and behavioral pediatrics, we are at an inflection point at which AI-driven technologies show potential to augment clinical decision-making, risk prediction, diagnostics, and treatment delivery. In addition, AI may be leveraged to automate certain time-intensive and arduous clinical tasks and to streamline workflows. Future research is still needed to address impediments to widespread clinical adoption. These include data bias, privacy, ownership and integration issues, disquietude over a perceived lack of algorithmic transparency, regulatory and payer bottlenecks, ethical ambiguities, and lack of rigorous and standardized AI-focused clinician training. AI technologies are not meant to replace the practicing physician or his/her clinical judgment, nor will they serve as a panacea to all the shortcomings of modern health care. However, we are optimistic about the future of AI in health care, including developmental and behavioral pediatrics. By enhancing the efficiency and impact of health care processes, AI approaches promise to reduce barriers to care and maximize the time clinicians are able to spend with their patients. |
Clinical symptoms and molecular epidemiologic characteristics of varicella patients among children and adults in Ganzhou, China | dbf9a77b-59a7-4b19-a2c1-1dfcbf8ba85f | 11844084 | Biochemistry[mh] | Varicella-zoster virus (VZV) is a highly transmissible double-stranded linear DNA virus, approximately 125 kb in length, belonging to the Alphaherpesvirinae subfamily of the Orthoherpesviridae family. Upon initial infection of the human body, VZV causes viremia and spreads throughout the body, leading to chickenpox, which is characterized by symptoms such as fever, headache, and a distinctive rash . Subsequently, the virus spreads through the bloodstream to sensory ganglia, such as the dorsal root ganglia and trigeminal ganglia, entering a long-term latent state where it ceases replication and causes no symptoms . When the body’s immune system is weakened, the virus can reactivate, migrating along sensory nerves to the skin and causing shingles . The symptoms of shingles include severe pain and a vesicular rash that follows the distribution of the affected nerves. Additionally, postherpetic neuralgia may occur as a major complication . Although most cases of varicella are relatively mild at presentation, they can become severe or fatal by causing serious complications, such as pneumonia and encephalitis . Several studies have shown that the severity of disease in VZV-infected individuals tends to correlate with age at the time of infection, with younger and older people generally being more susceptible and at greater risk of severe disease. Immunocompromised or deficient individuals, such as HIV-infected individuals, are also at high risk of developing severe disease or even dying after infection . There are significant differences, with primary VZV infection usually occurring in childhood in temperate regions, whereas in tropical regions, the first infection is delayed until adulthood . Differences in immune function and response mechanisms to the virus between children and adults may lead to differences in disease symptoms, disease progression, genotypes of infection, and treatment of primary VZV infection. Since the 1950s, researchers began isolating and characterizing VZV. In 1986, Davison and Scott successfully deciphered the complete nucleotide sequence of VZV (Dumas strain) . Subsequently, VZV was found to have different genotypes, and according to the new nomenclature of VZV proposed by Breuer et al. in 2010, VZV can be categorized into five established evolutionary Clades (Clades 1–5) and two provisional evolutionary Clades (Clades VI and VII) . There were certain genetic variations and geographic distribution differences among these Clades: Clades 1 and 3 were the predominant evolutionary branches in Europe and the Americas; Clades 4 and 5 were found mainly in the tropical regions of Africa and Central America; Clade VI was found in France and Italy located in southern Europe; and Clade VII was isolated only in the United States . Many studies have shown that VZV Clade 2 was overwhelmingly prevalent in Asia, and a live attenuated vaccine used to prevent infections in China was developed on the basis of the Oka strain of Clade 2 . Genotyping of VZV could help reveal the genetic diversity, evolutionary patterns, and transmission pathways of the virus . Since genetic variation among different viral strains of VZV consists of almost unevenly distributed single-nucleotide changes, rapid genotyping of isolates could be performed by single-nucleotide polymorphism (SNP) mapping of highly variable regions of the VZV genome . SNP sites based on a short fragment (447 bp) of the open reading frame (ORF) 22 had been reported to distinguish three major genotypes of VZV: E (European, Clade 1 or 3), J (Japanese, Clade 2), and M (Mosaic, Clade 4 or 5) . Furthermore, a comparison of the differences in gene sequences between the Oka vaccine strain and the parental Oka strain revealed that over one-third of the nucleotide differences were in ORF62 . Therefore, SNP analysis of the ORF62 fragment could distinguish the parental Oka strain from the vaccine strain, both of which belonged to Clade 2. Located inland approximately 400 km north of Hong Kong, China, Ganzhou is the largest city in Jiangxi Province, with a population of nearly 9 million. In this study, the clinical symptoms, disease progression, and laboratory test results of varicella patients among children and adults in Ganzhou from August 2021 to December 2022 were retrospectively analysed, and the genetic diversity of the isolates was characterized. The disease differences between different age groups infected with VZV were investigated, and the molecular epidemiological characteristics of VZV were clarified to provide a reference for prevention and treatment. Study population and sample collection Varicella patients, including those with varicella primary cases and varicella breakthrough cases, were diagnosed at the Department of Infection of Ganzhou Fifth People’s Hospital from August 2021 to December 2022. Information on all patients’ medical records, including age, sex, infection time, patient source, clinical symptoms, complications, and laboratory test results, was collected. Whole blood was collected from patients, and the plasma was centrifuged and stored at -80 °C until subsequent experiments. Amplification of VZV fragments Since viremia usually occurs during varicella, viral DNA was directly extracted from the plasma samples using the TaKaRa MiniBEST Viral RNA/DNA Extraction Kit Version 5.0 (TaKaRa, Japan). None of the samples had undergone virus isolation. The VZV ORF22 and ORF62 fragments were amplified by nested polymerase chain reaction using TransTaq ® DNA Polymerase High Fidelity (TransGen, China); the amplification primers were shown in Table . For the ORF22 fragment, the first amplification conditions were as follows: 96 °C for 10 min; 33 cycles of 95 °C for 30 s, 58 °C for 30 s, and 72 °C for 40 s; and 72 °C for 10 min. For the second amplification conditions, the cycling conditions were the same, except that the annealing temperature was 60 °C and the incubation at 72 °C in each cycle was 50 s. For the ORF62 fragment, the first amplification conditions were as follows: 96 °C for 10 min; 34 cycles of 95 °C for 30 s, 55 °C for 30 s, and 72 °C for 2.5 min; and 72 °C for 10 min. For the second amplification conditions, the cycling conditions were the same, except that the annealing temperature was 60 °C, and the incubation at 72 °C in each cycle was 2 min. The amplified products were tested with a 1% agarose gel. The possible positive samples were purified and sequenced by Tsing Ke Biotech Co., Ltd. (Kunming, China). Phylogenetic analyses of VZV sequences The sequences of ORF22 and ORF62 amplified in the present study were spliced using seqMan software in the DNAStar package, and the maximum likelihood tree was constructed using MEGA 7 with 1000 bootstrap replications. SNP analysis was also performed on 10 base loci of the ORF22 and ORF62 fragments in previous studies , which were used to identify the isolated VZV strain genotypes. The reference sequences for the phylogenetic analyses and SNP analyses, including Clades 1 to 5, were downloaded from the GenBank database ( https://www.ncbi.nlm.nih.gov/ ). The Clade 1 reference strains included Dumas (X04370), Kel (DQ479954), MSP (AY548170), BC (AY548171), SD (DQ479953), 36 (DQ479958), 49 (DQ479959), 32P5 (DQ479961) and 32P22 (DQ479962); the Clade 2 reference strains included the wild-type strain pOka (AB097933), pOka-derived vaccine strain vOka (AB097932), VariVax (DQ008355) and Varilrix (DQ008354); the Clade 3 reference strains included 11 (DQ479955), 22 (DQ479956), and HJO (AJ871403); the Clade 4 reference strains included 8 (DQ479960) and DR (DQ452050); and the Clade 5 reference strains included CA123 (DQ457052). For the sequences with mutations, reference sequences of ORF22 (MW545808) and ORF62 (AB097933) were downloaded from the GenBank database. Amino acids were translated and compared using DNAMAN 9 to explore whether these mutations lead to changes in amino acids. Statistical analysis Statistical analysis was performed using the software Statistical Package for Social Sciences (SPSS, version 26; IBM Corporation, Armonk, USA). The normality of the quantitative data was tested by the Shapiro-Wilk test. Data with a normal distribution were statistically described by the mean ± standard deviation, and those without a normal distribution were statistically described by the median (interquartile spacing); qualitative data were described by a combination of numbers and percentages . Comparisons between two or more groups of rates were made using Fisher’s exact probability method, and a two-sided P value of less than 0.05 was considered statistically significant. Varicella patients, including those with varicella primary cases and varicella breakthrough cases, were diagnosed at the Department of Infection of Ganzhou Fifth People’s Hospital from August 2021 to December 2022. Information on all patients’ medical records, including age, sex, infection time, patient source, clinical symptoms, complications, and laboratory test results, was collected. Whole blood was collected from patients, and the plasma was centrifuged and stored at -80 °C until subsequent experiments. Since viremia usually occurs during varicella, viral DNA was directly extracted from the plasma samples using the TaKaRa MiniBEST Viral RNA/DNA Extraction Kit Version 5.0 (TaKaRa, Japan). None of the samples had undergone virus isolation. The VZV ORF22 and ORF62 fragments were amplified by nested polymerase chain reaction using TransTaq ® DNA Polymerase High Fidelity (TransGen, China); the amplification primers were shown in Table . For the ORF22 fragment, the first amplification conditions were as follows: 96 °C for 10 min; 33 cycles of 95 °C for 30 s, 58 °C for 30 s, and 72 °C for 40 s; and 72 °C for 10 min. For the second amplification conditions, the cycling conditions were the same, except that the annealing temperature was 60 °C and the incubation at 72 °C in each cycle was 50 s. For the ORF62 fragment, the first amplification conditions were as follows: 96 °C for 10 min; 34 cycles of 95 °C for 30 s, 55 °C for 30 s, and 72 °C for 2.5 min; and 72 °C for 10 min. For the second amplification conditions, the cycling conditions were the same, except that the annealing temperature was 60 °C, and the incubation at 72 °C in each cycle was 2 min. The amplified products were tested with a 1% agarose gel. The possible positive samples were purified and sequenced by Tsing Ke Biotech Co., Ltd. (Kunming, China). The sequences of ORF22 and ORF62 amplified in the present study were spliced using seqMan software in the DNAStar package, and the maximum likelihood tree was constructed using MEGA 7 with 1000 bootstrap replications. SNP analysis was also performed on 10 base loci of the ORF22 and ORF62 fragments in previous studies , which were used to identify the isolated VZV strain genotypes. The reference sequences for the phylogenetic analyses and SNP analyses, including Clades 1 to 5, were downloaded from the GenBank database ( https://www.ncbi.nlm.nih.gov/ ). The Clade 1 reference strains included Dumas (X04370), Kel (DQ479954), MSP (AY548170), BC (AY548171), SD (DQ479953), 36 (DQ479958), 49 (DQ479959), 32P5 (DQ479961) and 32P22 (DQ479962); the Clade 2 reference strains included the wild-type strain pOka (AB097933), pOka-derived vaccine strain vOka (AB097932), VariVax (DQ008355) and Varilrix (DQ008354); the Clade 3 reference strains included 11 (DQ479955), 22 (DQ479956), and HJO (AJ871403); the Clade 4 reference strains included 8 (DQ479960) and DR (DQ452050); and the Clade 5 reference strains included CA123 (DQ457052). For the sequences with mutations, reference sequences of ORF22 (MW545808) and ORF62 (AB097933) were downloaded from the GenBank database. Amino acids were translated and compared using DNAMAN 9 to explore whether these mutations lead to changes in amino acids. Statistical analysis was performed using the software Statistical Package for Social Sciences (SPSS, version 26; IBM Corporation, Armonk, USA). The normality of the quantitative data was tested by the Shapiro-Wilk test. Data with a normal distribution were statistically described by the mean ± standard deviation, and those without a normal distribution were statistically described by the median (interquartile spacing); qualitative data were described by a combination of numbers and percentages . Comparisons between two or more groups of rates were made using Fisher’s exact probability method, and a two-sided P value of less than 0.05 was considered statistically significant. The nucleotide sequences of all VZV isolates in this study have been submitted to GenBank with the accession numbers OR972777-OR972808 for ORF22 and OR972809-OR972840 for ORF62. Characteristics of the study population A total of 39 patients with varicella were admitted to Ganzhou Fifth People’s Hospital from August 2021 to December 2022. The median age was 13.0 (8.0, 27.0) years, with a male-to-female ratio of 22:17. In the present study, patients with varicella were categorized into two groups, children and adults, according to their ages: 0–17 years and 18–55 years. There were 26 (66.7%) children and 13 (33.3%) adults. There was a difference in the source of varicella patients between children and adults ( P = 0.006), with 92.3% of adult VZV infections from inpatients (Table ). In contrast, children were almost equally represented by outpatient and inpatient sources. The rate of discomfort, such as fever, sore throat, dizziness/headache, fatigue, or cough, was significantly greater in adults than in children (100.0% versus 57.7%, P = 0.007). Among them, the proportion of adults presenting with sore throat and fatigue was significantly greater than that of children ( P < 0.05). Moreover, the incidence of complications after VZV infection was significantly greater in adults than in children (69.2% versus 7.7%, P < 0.001). There were three complications in children, including acute upper respiratory infection, acute tonsillitis, and skin infection, and nine complications in adults, including hepatitis, acute upper respiratory infection, pneumonia, thrombocytopenia, myocarditis, capillary leak syndrome, candida stomatitis, acute tonsillitis, and skin infection. Among these patients, adults were more likely to have hepatitis complications than children were (0.0% versus 46.2%, P = 0.001). There were no differences in the year of infection, sex distribution, or site of first varicella presentation between children and adults ( P > 0.05). Four (30.8%) adults with a mean age of 34.3 ± 5.9 years developed severe disease after VZV infection, and one died. All of them had concurrent hepatitis, and the results of relevant laboratory tests were well above the upper limit of the normal range: aspartate aminotransferase, 142.60 (2890.00) U/L; alanine aminotransferase, 835.73 ± 869.30 U/L; γ-glutamyl transferase, 160.15 ± 89.93 U/L; lactic dehydrogenase, 2077.50 ± 2018.58 U/L; and C-reactive protein, 33.10 ± 10.32 mg/L. Changes in laboratory tests in the study population Nearly half (45.8%) of the children developed leukopenia after VZV infection. Accordingly, their median WBC count of 4.86 (3.71) × 10 9 /L was slightly below the lower limit of the normal reference range (Table ). In contrast, adults had abnormal results for several laboratory tests: a mean eosinophil percentage of 0.25 (1.27)% and a mean Ca 2+ of 2.16 ± 0.13 mmol/L were below the lower limit of the normal reference range, whereas a mean basophil percentage of 1.22 ± 0.63%, a median alanine aminotransferase of 65.40 (125.60) U/L, a median aspartate aminotransferase of 49.80 (54.50) U/L, a median γ-glutamyl transferase of 55.10 (121.15) U/L, a median lactic dehydrogenase of 308.80 (127.90) U/L, a mean C-reactive protein of 21.76 ± 16.55 mg/L and a mean glucose of 6.50 ± 1.73 mmol/L were above the upper limit of the normal reference range. These results were consistent with the laboratory test results, which revealed that 53.8% of adults with VZV infection had abnormal liver function and that 23.1% had hypocalcemia. Genotyping and phylogenetic analysis of Ganzhou strains In this study, the ORF22 and ORF62 fragments of VZV were successfully amplified and sequenced in 32 plasma samples, with a success rate of 94.1% (32 out of 34). VZV genotyping was performed on the basis of the SNP sites in the amplified fragments (Table ). Among them, four SNP sites were identified in the ORF22 fragment, which was consistent with those of pOka and vOka, indicating that these sequences belonged to Clade 2. In addition, six SNP sites were identified in the ORF62 fragment, which were consistent with those of pOka (Clade 2), CA123 (Clade 5), and DR (Clade 4). Taken together, these results indicate that these isolates belonged to Clade 2, as all of their SNP sites were consistent with those of pOka. A phylogenetic tree was further constructed by the maximum likelihood method to determine the genotypes and evolutionary relationships of the isolates (Fig. ). Consistent with the results of the SNP locus analysis, the maximum likelihood tree revealed that the isolated virulent strains clustered with pOka but not with vOka, further confirming that all the clinical isolates in this study belonged to Clade 2. Mutational analysis of the Ganzhou strain The maximum likelihood tree revealed that six VZV strains (GZ5M03, GZ8M08, GZ15M21, GZ15F24, GZ21M104, and GZ37M110) were genetically distant from the reference strain pOka and the other isolates, indicating that these isolates had mutations in their gene sequences (Fig. ). Among these isolates, four (66.7%) were found in children, two (33.3%) were found in adults, and 83.3% of the variant patients were male. Further analysis revealed that the ORF22 and ORF62 fragments of the six mutant strains had a total of 12 base mutations, of which the ORF22 fragment contained 9 (75.0%) mutated bases and the ORF62 fragment contained 3 (25.0%) mutated bases (Table ). Among the 12 mutated bases, the highest mutation prevalence from base A to base G (A→G) (41.7%), T→C incidence was the next highest (25.0%), followed by the incidence of mutations in G→A, C→Y, A→T and T→A (8.3% each). Most (50.0%) mutants had 2-base mutations, whereas the GZ15F24 mutant had 4-base mutations. Among the mutated sequences, 75.0% were missense mutations, and 25.0% were synonymous mutations (Table ). Eight of the nine bases mutated in the ORF22 fragment were missense mutations (88.9%), and one of the three bases mutated in the ORF62 fragment (33.3%) was a missense mutation. In the isolates from children, base mutations at sites 37,940 and 37,990 in GZ8M08 resulted in changes in amino acids 1286 and 1303 encoded by the ORF22 fragment, with the former changing from methionine to isoleucine and the latter changing from glutamine to leucine; base mutations at sites 37,891 and 37,939 in GZ15F24 resulted in changes in amino acids 1270 and 1286, with the former changing from glutamic acid to glycine and the latter changing from methionine to threonine; a base mutation at site 107,122 of isolate GZ15M21 resulted in a change in amino acid 671, which is encoded by the ORF62 fragment from valine to alanine. In the isolates from adults, the mutations at sites 37,986 and 38,143 of GZ21M104 resulted in changes in amino acids 1302 and 1354 encoded by the ORF22 fragment, the former from arginine to arginine/cysteine and the latter from valine to glutamic acid; base mutations at positions 37,974 and 38,133 in isolate GZ37M110 resulted in changes in amino acids 1298 and 1351 of the ORF22 fragment, the former from tyrosine to histidine and the latter from threonine to alanine. A total of 39 patients with varicella were admitted to Ganzhou Fifth People’s Hospital from August 2021 to December 2022. The median age was 13.0 (8.0, 27.0) years, with a male-to-female ratio of 22:17. In the present study, patients with varicella were categorized into two groups, children and adults, according to their ages: 0–17 years and 18–55 years. There were 26 (66.7%) children and 13 (33.3%) adults. There was a difference in the source of varicella patients between children and adults ( P = 0.006), with 92.3% of adult VZV infections from inpatients (Table ). In contrast, children were almost equally represented by outpatient and inpatient sources. The rate of discomfort, such as fever, sore throat, dizziness/headache, fatigue, or cough, was significantly greater in adults than in children (100.0% versus 57.7%, P = 0.007). Among them, the proportion of adults presenting with sore throat and fatigue was significantly greater than that of children ( P < 0.05). Moreover, the incidence of complications after VZV infection was significantly greater in adults than in children (69.2% versus 7.7%, P < 0.001). There were three complications in children, including acute upper respiratory infection, acute tonsillitis, and skin infection, and nine complications in adults, including hepatitis, acute upper respiratory infection, pneumonia, thrombocytopenia, myocarditis, capillary leak syndrome, candida stomatitis, acute tonsillitis, and skin infection. Among these patients, adults were more likely to have hepatitis complications than children were (0.0% versus 46.2%, P = 0.001). There were no differences in the year of infection, sex distribution, or site of first varicella presentation between children and adults ( P > 0.05). Four (30.8%) adults with a mean age of 34.3 ± 5.9 years developed severe disease after VZV infection, and one died. All of them had concurrent hepatitis, and the results of relevant laboratory tests were well above the upper limit of the normal range: aspartate aminotransferase, 142.60 (2890.00) U/L; alanine aminotransferase, 835.73 ± 869.30 U/L; γ-glutamyl transferase, 160.15 ± 89.93 U/L; lactic dehydrogenase, 2077.50 ± 2018.58 U/L; and C-reactive protein, 33.10 ± 10.32 mg/L. Nearly half (45.8%) of the children developed leukopenia after VZV infection. Accordingly, their median WBC count of 4.86 (3.71) × 10 9 /L was slightly below the lower limit of the normal reference range (Table ). In contrast, adults had abnormal results for several laboratory tests: a mean eosinophil percentage of 0.25 (1.27)% and a mean Ca 2+ of 2.16 ± 0.13 mmol/L were below the lower limit of the normal reference range, whereas a mean basophil percentage of 1.22 ± 0.63%, a median alanine aminotransferase of 65.40 (125.60) U/L, a median aspartate aminotransferase of 49.80 (54.50) U/L, a median γ-glutamyl transferase of 55.10 (121.15) U/L, a median lactic dehydrogenase of 308.80 (127.90) U/L, a mean C-reactive protein of 21.76 ± 16.55 mg/L and a mean glucose of 6.50 ± 1.73 mmol/L were above the upper limit of the normal reference range. These results were consistent with the laboratory test results, which revealed that 53.8% of adults with VZV infection had abnormal liver function and that 23.1% had hypocalcemia. In this study, the ORF22 and ORF62 fragments of VZV were successfully amplified and sequenced in 32 plasma samples, with a success rate of 94.1% (32 out of 34). VZV genotyping was performed on the basis of the SNP sites in the amplified fragments (Table ). Among them, four SNP sites were identified in the ORF22 fragment, which was consistent with those of pOka and vOka, indicating that these sequences belonged to Clade 2. In addition, six SNP sites were identified in the ORF62 fragment, which were consistent with those of pOka (Clade 2), CA123 (Clade 5), and DR (Clade 4). Taken together, these results indicate that these isolates belonged to Clade 2, as all of their SNP sites were consistent with those of pOka. A phylogenetic tree was further constructed by the maximum likelihood method to determine the genotypes and evolutionary relationships of the isolates (Fig. ). Consistent with the results of the SNP locus analysis, the maximum likelihood tree revealed that the isolated virulent strains clustered with pOka but not with vOka, further confirming that all the clinical isolates in this study belonged to Clade 2. The maximum likelihood tree revealed that six VZV strains (GZ5M03, GZ8M08, GZ15M21, GZ15F24, GZ21M104, and GZ37M110) were genetically distant from the reference strain pOka and the other isolates, indicating that these isolates had mutations in their gene sequences (Fig. ). Among these isolates, four (66.7%) were found in children, two (33.3%) were found in adults, and 83.3% of the variant patients were male. Further analysis revealed that the ORF22 and ORF62 fragments of the six mutant strains had a total of 12 base mutations, of which the ORF22 fragment contained 9 (75.0%) mutated bases and the ORF62 fragment contained 3 (25.0%) mutated bases (Table ). Among the 12 mutated bases, the highest mutation prevalence from base A to base G (A→G) (41.7%), T→C incidence was the next highest (25.0%), followed by the incidence of mutations in G→A, C→Y, A→T and T→A (8.3% each). Most (50.0%) mutants had 2-base mutations, whereas the GZ15F24 mutant had 4-base mutations. Among the mutated sequences, 75.0% were missense mutations, and 25.0% were synonymous mutations (Table ). Eight of the nine bases mutated in the ORF22 fragment were missense mutations (88.9%), and one of the three bases mutated in the ORF62 fragment (33.3%) was a missense mutation. In the isolates from children, base mutations at sites 37,940 and 37,990 in GZ8M08 resulted in changes in amino acids 1286 and 1303 encoded by the ORF22 fragment, with the former changing from methionine to isoleucine and the latter changing from glutamine to leucine; base mutations at sites 37,891 and 37,939 in GZ15F24 resulted in changes in amino acids 1270 and 1286, with the former changing from glutamic acid to glycine and the latter changing from methionine to threonine; a base mutation at site 107,122 of isolate GZ15M21 resulted in a change in amino acid 671, which is encoded by the ORF62 fragment from valine to alanine. In the isolates from adults, the mutations at sites 37,986 and 38,143 of GZ21M104 resulted in changes in amino acids 1302 and 1354 encoded by the ORF22 fragment, the former from arginine to arginine/cysteine and the latter from valine to glutamic acid; base mutations at positions 37,974 and 38,133 in isolate GZ37M110 resulted in changes in amino acids 1298 and 1351 of the ORF22 fragment, the former from tyrosine to histidine and the latter from threonine to alanine. VZV was highly contagious in individuals without a history of varicella and could be reactivated in human ganglion cells after initial infection . Primary VZV infection resulting in varicella was usually considered a mild childhood disease; however, data from past studies had shown that varicella in adults was 25 times more likely to be severe (and possibly even fatal) than those in children . The present study revealed that varicella in children often presented as a self-limiting disease, with the main symptom being fever. However, four adults developed severe diseases, one of whom died. These results were consistent with those of previous studies. Moreover, the present study also compared the results of laboratory markers between children and adults and revealed that test results regarding liver function, such as alanine aminotransferase, aspartate aminotransferase, γ-glutamyl transferase, and lactic dehydrogenase, were higher than the upper limit of the normal reference range; however, this phenomenon was not found among children. Therefore, more attention should be given to adult varicella patients because they might be at greater risk of developing serious illnesses, especially liver dysfunction. Molecular epidemiological information on VZV was crucial for characterizing endemic strains, tracing transmission routes, distinguishing genotypes, and exploring evolutionary patterns . Previous studies have shown that Clade 2 was predominant in China, and the live attenuated vaccine used to prevent varicella in China was designed on the basis of the Oka strain of Clade 2 . In a small number of vaccinated children, the varicella vaccine virus reactivates and causes herpes zoster, sometimes with meningitis or meningoencephalitis. Occasionally the herpes zoster rash can be extensive enough to be confused with an early varicella rash. Cases have been reported in the United States, Germany, Greece, Switzerland and Japan . In the present study, all clinical VZV isolates belonged to the highly homogeneous wild-type Clade 2, and no vaccine strain infections were detected. This indicates that the current varicella vaccine in Ganzhou should continue to be promoted to reduce the incidence of varicella, as it remains effective. Moreover, the genomic sequencing technology developed in Ganzhou and described in this report will enable physicians to diagnose any vaccine-related adverse events that occur in Chinese children. Several studies have shown that the epidemiological subtypes of VZV in peripheral regions of China, such as Xinjiang, Xizang and Guangdong Provinces, were more complex than those in internal regions . Three clades, including Clade 2, Clade 4 and Clade 5, were found to be prevalent in Guangdong Province . Ganzhou, the sample site of the present study, borders Guangdong Province, with a border of 1400 km. Although only Clade 2 had been found to be prevalent in Ganzhou, continuous monitoring of the molecular characteristics of VZV was urgently needed to avoid the complication of the VZV genotype, as cross-border activities were quite common in this region. It was hypothesized that VZV appeared in early primitive humans in Africa approximately 7 million years ago and evolved with primates . The VZV genome was considered highly conserved, but it was still evolving, with mutations occurring occasionally, some of which were fixed and some of which were random . In this study, twelve previously unidentified mutated bases were present in the ORF22 and ORF62 fragments of the six isolates, with the ORF22 fragment having a more significant percentage of mutated bases than the ORF62 fragment (75.0% versus 25.0%). Moreover, the incidence of amino acid changes resulting from mutated bases was greater in the ORF22 fragment than in the ORF62 fragment (88.9% versus 33.3%). Overall, the frequency of prevalent genetic variation in VZV in the population is low, possibly due to its low reproduction rate in infected hosts, thus limiting the frequency of introducing new mutations. Two limitations should be considered when interpreting the results of the present study. First, there were no tests to determine whether the differences in laboratory test results between children and adults were statistically significant due to the small sample sizes of these subgroups. Second, the ORF22 and ORF62 fragments of VZV among the two patients were failed to be amplified, probably due to the low viral load in the plasma samples. In future studies, the simultaneous use of vesicular fluid and blood samples may enhance the amplification efficiency of VZV DNA. In this study, clinical symptoms, disease progression, and laboratory test results of varicella patients were retrospectively analyzed to explore the differences in disease between children and adults, and molecular epidemiology was used to characterize the genetic variation of VZV isolates in Ganzhou. All isolates were Clade 2 wildtype with high homogeneity. Adults were more severely infected with VZV than children, with a greater tendency to involve the liver and a higher risk of developing severe disease. These results reveal that preventive and therapeutic measures should be intensified for adult VZV-susceptible populations and those at high risk of severe disease after infection. Below is the link to the electronic supplementary material. Supplementary Material 1 |
Comparative Physical Study of Three Pharmaceutically
Active Benzodiazepine Derivatives: Crystalline versus Amorphous State
and Crystallization Tendency | c8539220-7eaa-478e-a0c3-ec0c4e058b5b | 8594866 | Pharmacology[mh] | Introduction The chemical modification of active pharmaceutical ingredients
(APIs) is one of the main strategies to identify better drugs with
reduced side effects and increased efficacy or bioavailability. A
historical example is that of the active ingredient of aspirin: derivatization
of salicylic acid, the active principle present in willow barks, into
acetylsalicylic acid leads to substantial reduction of the side effects
of the naturally occurring drug. Given
that low solubility in water and thus low oral bioavailability is
one of the main issues in current drug research, chemical derivatization
of APIs in the form, e.g., of hydrochloride salts with enhanced solubility
is often pursued. , Another related strategy for efficient
drug administration is the development of a prodrug, i.e., an inactive
compound (usually a derivative of an active drug) that undergoes in vivo transformation, through enzymes or metabolic processes,
into the active parent drug. This strategy has been applied successfully
to improve the pharmacokinetic properties of drugs since the middle
of the last century, when the term prodrug was first introduced. Nowadays, prodrugs make almost 10% of the administered
drugs, reaching a peak of 20% of the market between 2000 and 2008. , While chemical derivatization is mainly aimed at identifying
drugs
with better biochemical properties, it also obviously affects the
physical properties of the parent API. In the vast majority of cases,
the induced changes in physical properties stem from relatively minor
chemical changes, as the derivative (prodrug, salt, etc.) is usually
one or two metabolic steps away from the active parent drug. The chemical modification may, for example, determine
a modified crystal structure of the resulting drug and have an impact
also on the possible polymorphism and relative stability of different
crystalline forms, which is of relevance for API storage prior to
industrial processing. These aspects are extremely important for the
pharmaceutical industry, as polymorphism or the possible stability
of an amorphous (glass and supercooled liquid) phase can have a strong
impact on the viable protocols for the preparation of suitable formulations
for the administration of APIs. , Drug derivatization
also affects the glass transition temperature
and the kinetic stability of the amorphous form of the drug. It is
well-known that amorphous pharmaceuticals have better dissolution
and thus better bioavailability properties than their crystalline
counterparts, , and a few amorphous drugs have
appeared on the market in recent years. , The amorphous
form of a drug may be present in a formulation as a result of industrial
processing via, e.g., milling and spray or freeze drying. − Despite their advantage in terms of solubility, however, amorphous
drugs are not thermodynamically stable and are thus prone to recrystallization
into the lower-solubility crystalline form. , − A better understanding of the amorphous state is
needed to advance in the formulation of amorphous drugs. In the context
of drug modification strategies, it would be extremely useful to be
able to predict how different drug derivatives behave in terms of
kinetic stability and tendency toward recrystallization of the amorphous
form, both in the case of amorphous API phases formed spontaneously
or purposefully during formulation of a medicament. The present paper
takes a step in this direction by comparing the physical properties
of the amorphous and crystalline forms of three distinct pharmacologically
active benzodiazepines, with the aim of exploring possible routes
to increase the kinetic stability of amorphous derivatives. The common molecular structure of the benzodiazepine drugs consists
of a rigid benzene ring and a flexible diazepine ring fused together.
Several benzodiazepines also display a third six-membered ring covalently
attached to a carbon atom of the diazepine ring (see, e.g., the molecular
structures displayed as insets to ). These drugs work by enhancing the effect
of the gamma - aminobutyric acid neurotransmitter,
and they have sedative, hypnotic, anxiolytic, anticonvulsant, and
muscle relaxant properties. According to a WHO report of 2017, 322
million people suffer from depression as of 2015 and almost as many
suffer from other anxiety disorders and
it is estimated that 40% of patients with depressive and anxiety disorders
are prescribed benzodiazepines. Oral
administration is the most common route of administration of benzodiazepines
(although injectable, inhalation, and rectal forms are also available),
but, given that they are lipophilic drugs, problems of low solubility
and bioavailability may arise in the gastrointestinal tract. , Low bioavailability may result in the need of a higher dose administered
to the patient, to account for the percentage that is not absorbed
and metabolized. This may lead to undesirable adverse side effects,
which are already pretty severe with high doses of this type of drugs. Here, we study three related benzodiazepine derivatives: Diazepam,
Nordazepam (also known as Nordiazepam or desmethyldiazepam), and Tetrazepam.
Diazepam (see inset to a) is one of the best known benzodiazepines and was first
marketed as Valium. It is used as a treatment for various mental diseases,
but its primary use is for anxiety, states of agitation, or panic
attacks. Diazepam has been studied extensively in both crystalline
and amorphous states, sometimes in comparative studies with other
benzodiazepines. − Its main active metabolite is Nordazepam, whose chemical structure
differs from that of Diazepam only by the substitution of the methyl
group linked to the nitrogen 1 of the diazepine by a hydrogen atom
(see the inset to b). This difference, however, is highly significant in that
it confers the Nordazepam derivative the possibility of self-aggregation
via hydrogen bonding via the H-functionalization of the nitrogen atom.
Tetrazepam (inset to c) differs from Diazepam in that the benzene ring attached
to the carbon 5 of the diazepine ring is substituted by a cyclohexene
ring. It was marketed principally as a treatment for muscle spasms
and panic attacks but was suspended from the market across the European
Union in 2013, due to cutaneous toxicity. Our comparative study
of these three pharmaceutically active ingredients
encompasses both their crystalline and amorphous forms (supercooled
liquid and glass), as well as the transition between the supercooled
liquid phase to the crystalline one. We focus in particular on the
molecular conformations and intermolecular interactions in the crystal
phase, Hirshfeld surfaces, calorimetric properties, dynamic relaxations,
and recrystallization kinetics, the latter two measured by dielectric
spectroscopy. Our aim is to understand how the modifications in molecular
structure and the resulting intermolecular interactions affect the
crystal structure and molecular dynamics in the amorphous phase, as
well as the melting point, glass transition temperature, and tendency
toward recrystallization of the various derivatives, with the aim
of identifying possible structure–property correlations. The
study of molecular relaxation processes in diazepines is particularly
interesting due to the inherent flexibility of the seven-membered
diazepine ring, which leads to conformational diversity of the molecules
and therefore to the possible existence of a relaxational inter-conformer
conversion dynamics. To the best of our knowledge, only a few very
recent studies have focused on the interpretation of the dielectric
relaxation of flexible heterocyclic molecules. A further outcome of this work is therefore to expand the
current experimental knowledge of the conformational dynamics of flexible
cyclic or ring-containing molecules.
Materials
and Methods Tetrazepam (TETRA, hereinafter) is a powder of
medicinal grade
kindly supplied by Daiichi Sankyo France SAS. Samples of medicinal
grade Nordazepam (NOR) were kindly provided by Bouchara-Recordati
(France) and medicinal grade Diazepam (DIA) was kindly supplied by
Neuraxpharm (Spain). The powders of the three diazepines, with purities
higher than 99.5%, were used as received without further purification.
Differential scanning calorimetry (DSC) experiments were carried out
under a nitrogen atmosphere on samples loaded in pierced aluminum
pans, by means of a Q100 calorimeter from TA Instruments. Measurements
were performed using heating/cooling rates of 10 K min –1 and sample masses of the order of 5 mg, as determined with a microbalance
with 0.01 mg sensitivity. Powder X-ray diffraction patterns
have been acquired by means of
a vertically mounted INEL cylindrical position-sensitive detector
(CPS-120) using the Debye–Scherrer geometry and transmission
mode. Monochromatic Cu Kα 1 (λ = 1.54056 Å)
radiation was selected by means of a quartz monochromator. Cubic phase
Na 2 Ca 3 Al 2 F 4 was used for
external calibration. The analysis of the diffraction patterns (fitting
of diffraction peaks by means of the Materials Studio software ) was carried out using the published monoclinic
(P2 1 /c) structures of TETRA, DIA, and NOR. Hirshfeld surface analyses were performed by means of the CrystalExplorer
software ( https://crystalexplorer.scb.uwa.edu.au/ ). Broadband dielectric spectroscopy (BDS) measurements were
carried
on the amorphous form (supercooled liquid and glass states) of the
drugs, by means of a Novocontrol Alpha analyzer. The samples were
placed in a stainless steel parallel-plate capacitor specially designed
for the analysis of liquid samples, with the two electrodes kept at
a fixed distance by means of cylindrical silica spacers of 50 μm
diameter. Temperature control of the capacitor and thus of the sample
was achieved with a nitrogen-gas flow cryostat with a precision of
0.1 K. To obtain the amorphous form, the powders were initially melted
in the capacitor outside the cryostat, cooled at room temperature,
and melted again inside the cryostat. Each sample was then cooled
with a cooling rate of 10 K min –1 to 123 K to avoid
recrystallization, and isothermal spectra were then acquired every
2 or 5 K, waiting each time 5 min for temperature stabilization. Dielectric
spectra were measured in the frequency range between 10 –2 and 10 7 Hz, from 123 K up to the melting temperature
of each compound (404.1, 415.6, and 487 K, for Diazepam, Tetrazepam,
and Nordazepam, respectively). To obtain relaxation times and
quantify the changes in relaxation
dynamics, we employed the Grafity software to fit the dielectric spectra
as the sum of a power law representing the dc conductivity contribution,
modeled as a term of the form in the complex
permittivity, where s is an exponent close to unity,
and a Havriliak–Negami
(HN) function for each relaxation component. Overall, the spectra contained four different relaxation components
(referred to as α, β, γ, and γ′ in
the text), and the total complex permittivity was modeled as follows: 1 Here, ω
= 2πν
is the angular frequency, ε ∞ is the permittivity
in the high frequency limit, Δε i is the dielectric intensity (or relaxation strength) of relaxation i ( i = α, β, γ or γ′), a i and b i are parameters
describing the shape of the corresponding loss curves, and τ HN, i is a time parameter connected to the
characteristic relaxation time τ max, i , corresponding to the maximum loss of relaxation i . In terms of the fit parameters, τ max, i (which we will refer to as τ i in the following, for simplicity) is given by the following: 2 The shape parameters a and b can
vary between 0 and 1. Specific cases of the HN function are the Cole–Cole and Cole–Davidson functions, which are obtained for b =
1 and a = 1, respectively. In the case of the Cole–Cole
function, reduces
to τ i = τ HN, i . Throughout the text, we refer to τ max, i simply as the relaxation time, and use for it the
symbols τ or τ i to simplify
the notation. Most dielectric spectra displayed only two relaxations
in the accessible frequency window, namely, either the α and
β relaxations (near and above T g ) or else the intramolecular γ and γ′ relaxations
(well below T g , see ), so that our fit procedure
only involved at most two HN functions at the time. The (primary)
α relaxation turned out to be well described by a Cole–Davidson
function, while all secondary relaxations could be fitted with Cole–Cole
functions. This reduced significantly the actual number of free fit
parameters that had to be employed in each fit.
Results 3.1 Differential Scanning Calorimetry Results shows the
DSC traces obtained for the three diazepines DIA, NOR, and TETRA.
In all three cases, the as-received powders were completely crystalline,
as the first heating ramp only displayed a melting endotherm with
onsets at 404.1, 487.0, and 415.6 K for DIA, NOR, and TETRA, respectively.
Values coincide within the experimental error with those available
in the scientific literature. − , , The melting point of NOR and the enthalpy of melting are both significantly
higher than that of the other two derivatives, likely due to the presence
of N-H···O=C hydrogen bonds, which can only form in
demethylated derivative (see the next section). The subsequent
cooling ramp leads to a glassy phase for all three pharmaceuticals,
and on reheating, a step-like transition can be observed in the DSC
traces, corresponding to the glass transition temperature ( T g ). In most cases, though not in all DSC runs,
TETRA and NOR displayed (at least partial) recrystallization of the
supercooled liquid in the heat up run, followed again by the melting
peak (see inset to b). The recrystallized phase is the same as the initial one,
as the melting temperature is the same on heating the recrystallized
sample. The supercooled TETRA and NOR liquids were observed to crystallize
also in dielectric spectroscopy experiments (see ), while recrystallization
of DIA was absent also in this case. The sample geometry and the vessel
are quite different in DSC (droplet in aluminum pan) and dielectric
(film in stainless steel cylinder with silica spacers) experiments.
The fact that the three samples displayed the same tendency toward
recrystallization under such different experimental conditions indicates
that the recrystallization of TETRA and NOR probably took place by
homogeneous (rather than heterogeneous) nucleation of the crystal
phase. The characteristic onset temperatures of the glass transition,
recrystallization, and melting points are listed in for all three pharmaceutically
active compounds, together with the melting enthalpies. The recrystallization
temperature is only listed for completeness, as it did not always
occur in all DSC scans at the same temperature. This is not surprising,
as nucleation is a stochastic event that depends on the characteristics
of the sample (heterogeneous vs homogeneous nucleation) and its history
(e.g., cooling rate from the liquid phase, temperature at which it
is then kept). It may be seen that T m and T g roughly scale with one another:
the T g / T m ratio
is 0.78
for DIA, 0.71 for NOR, and 0.75 for TETRA. The values for TETRA and
DIA are quite similar, albeit T m is slightly
higher for TETRA than that for DIA, while T g is somewhat lower for TETRA than that for DIA. The glass transition
temperature is often found to display a correlation with the molecular
weight M w . In particular, the empirical
rule T g ≈ M w 1/2 appears to be fulfilled in the case of van
der Waals molecular liquids. Such correlation
probably reflects the fact that the extent of van der Waals interactions
increases with the molecular mass (due to the increase of molecular
polarizability and of the closest intermolecular contacts), and the
fact that, at a given fixed temperature, a massive molecule has lower
mobility, but it does not take into account hydrogen bonding or any
other type of directional intermolecular bonds. In fact, the glass
transition temperature of the studied diazepines does not correlate
with the molecular weight: NOR, which has the lowest weight, has the
highest glass transition temperature. The origin of the higher T g is likely the same as that of the higher T m , namely, the presence of intermolecular H-bonds
in the liquid phase of NOR. Indeed, in the absence of any H bonding
the aforementioned correlation of molecular weight and glass transition
temperature would result in a T g value
of NOR closer to those of DIA and TETRA, which is not observed. 3.2 X-ray Diffraction Results and Analysis All three compounds display, in the crystalline phase, the same monoclinic
space group (P 2 1 /c). The diazepine ring of all molecules
adopts a bent boat-like conformation, with two possible isoenergetic
conformers, which are mirror images of one another. The two conformers
have opposite chirality and are named P (plus) or M (minus) according
to the sign of the (O=)C–C(H 2 )–N=C
torsion angle (see the inset to ). All three crystals contain a 1:1 ratio of P and
M conformers. The geometry of the conformers is similar in all three
compounds. For example, the angle formed by the C=N bond with
the plane of the fused benzene ring is equal to 41.6, 38.5, and 48.6°
in crystalline DIA, NOR, and TETRA, respectively. The analysis
of the X-ray structures at room temperature shows unambiguously that
NOR is the only compound of the three related drugs studied that forms
strong hydrogen bonds in the crystalline state, namely, intermolecular
N–H···O bonds involving the amine nitrogen of
the diazepine ring and the carbonyl oxygen of the same group of a
nearest-neighbor molecule in the crystal structure (see ). This is in agreement with
the higher melting point and enthalpy of fusion of NOR compared with
the other two compounds . It is interesting to point out in this respect that while
in both crystalline DIA and TETRA the carbonyl group and the adjacent
methyl group are basically coplanar, with a H 3 C–N–C=O
torsion angle smaller than 2°, in the case of NOR, which is a priori the only compound where the corresponding (peptide)
moiety is expected to be planar due to the amide electronic resonance,
the H–N–C=O torsion angle is instead approximately
10°. Non-planar peptide bonds are not uncommon in H-bonded structures
such as proteins in their native state. In the case of crystalline NOR, the lack of planarity of the amide
group is likely a consequence of H-bond formation. A recent work by some of us has shown that DIA and
TETRA, while
not forming N–H···O bonds, display weak but
extensive C–H···O interactions between the electron-rich
carbonyl group and the weakly polar C–H bonds of CH 2 groups. While intermolecular N–H···O
bonds are at least partially present also in the amorphous state of
NOR, as testified by its much higher glass transition temperature
(see ),
it is unlikely that the C–H···O interactions
play any role in the amorphous state of the three compounds, as we
argue further in . A straightforward comparison of the hydrogen
bond scheme in the
solid state of the three compounds can be carried out based on the
analysis of the Hirshfeld surface areas (see ). This surface represents a particular way
of partitioning the overall electron density in a molecular crystal
into individual molecular units, which
provides a three-dimensional image of the close contacts in the crystal
by guaranteeing maximum proximity of the corresponding Hirshfeld volumes
of nearest-neighbor molecules. − The color code employed by convention
is that a yellow or red color indicates points of short intermolecular
contact, while blue indicates regions of the Hirshfeld surface corresponding
to directions in which the intermolecular distance is comparatively
longer. ,
adapted
from ref , shows
the key intermolecular contacts derived from the Hirshfeld surface
area analysis at room temperature in the crystalline state. It evidences
the relevance of the hydrogen bond scheme for these compounds and,
in particular, that of the O···H for NOR compared to
DIA and TETRA, in agreement with the role of the strong N–H···O
H–bond interaction in NOR. It is interesting to note that there is a correlation
between melting
point, density, and Hirshfeld surface and volume parameters . In particular,
the Hirshfeld molecular volume and surface and the Hirshfeld volume
normalized to molecular weight are the largest for TETRA, which has
the smallest density and the lowest T m of the three derivatives, and they are the smallest for NOR, which
has the largest density and highest T m . This correlation evidences the influence on the melting temperature
of the hydrogen bonds in crystalline NOR. We point out that
the correlation is instead not strictly verified
when considering the glass transition temperature of all derivatives,
as T g,DIA > T g,TETRA . However, as mentioned, the T g of NOR
is significantly higher than that of the other two compounds, which
is indicative of the presence of some H bonding also in the liquid
phase of this compound. Instead of tightly bound stable H-bonded dimers
in the liquid phase, only short-lived H bonds are expected to occur,
and it is likely that a given NOR molecule only takes part, at most,
in one H-bond at a time. 3.3 Broadband Dielectric Spectroscopy
Results In order to see in detail how the small difference
in molecular
formula as well as the relevance of the hydrogen-bond network between
the three studied benzodiazepines affects the molecular mobility and
conformational dynamics in the amorphous state, we carried out dielectric
spectroscopy experiments on all three compounds in their amorphous
states. shows
the dielectric loss function of the three compounds at few selected
temperatures, plotted against the frequency of the applied electric
field. 3.3.1 Structural Relaxation For all three
diazepines, the most intense loss peak is observed at high temperatures
, and corresponds
to the structural relaxation (or α relaxation) of the supercooled
liquid phase. Below the calorimetric glass transition temperature T g (at which τ α = 10 2 s), the peak frequency of the α relaxation lies outside
the experimental frequency window, and only the tail of the α
peak is observed. When the temperature is increased above T g , the onset of the cooperative relaxation dynamics
of the liquid phase is signaled by the appearance in the experimental
frequency window of the α peak maximum, which then shifts to
higher frequencies as the temperature is further increased. The intensity of the α loss feature of both DIA and NOR is
roughly constant above T g . Instead, recrystallization
upon heating can be clearly discerned in the series of loss spectra
in the case of TETRA. Indeed, at temperatures higher than 335 K the
dielectric intensity of the α peak of TETRA is observed to decrease
further and further as the amorphous fraction in the sample decreases
(the dielectric loss intensity is proportional to the number density
of molecules in the amorphous supercooled liquid state ). To analyze the relaxation dynamics of
the cooperative α relaxation
in detail, we fitted all dielectric spectra as the sum of several
Havriliak–Negami components (see ), each corresponding to a distinct relaxation,
in order to extract the temperature-dependent relaxation times ( , see the section). The fits are shown in along with experimental
data. We found in particular that the fit with Havriliak–Negami
curves resulted in a Cole–Davidson function for the structural
relaxation. It can be observed in that the α peak of each compound has
exactly
the same shape regardless of temperature: the isothermal spectra at
various temperatures could be superposed onto one another by rescaling
the frequency scale and the signal intensity to those of the loss
maximum. This master-curve scaling was employed in the fitting procedure,
by imposing the same Cole–Davidson (CD) exponent in all high-temperature
spectra of a given compound, as indicated for selected temperatures
in the three panels of . The CD exponent that best described the α peaks was
found to be b = 0.59 ± 0.03 for DIA and TETRA,
and b = 0.50 ± 0.02 for NOR. This result indicates
a slightly greater cooperativity for NOR with respect to DIA and TETRA, , possibly related to the presence of intermolecular H-bonds in NOR. shows the
α relaxation times of all three studied diazepines versus the
inverse temperature (Arrhenius plot). The α relaxation time
follows the Vogel–Fulcher–Tamman temperature-dependence
typical of cooperative structural relaxations: − 3 Here, τ 0 is
the characteristic time at infinite temperature, D is the fragility strength coefficient, and T 0 is the Vogel–Fulcher temperature. The so-called “kinetic”
or “dielectric” glass transition temperature T g of the sample is defined as the temperature
at which relaxation times reaches 100 s, i.e., where log 10 (τ α /[s]) = 2 (horizontal yellow line in a). The kinetic glass
transition temperatures are 312.6, 309.0, and 347.2 K for DIA, TETRA,
and NOR, respectively . These values are very similar to the ones found in DSC
(see ), as expected. It is interesting
to compare the dependence of the relaxation times
with the inverse temperature rescaled to T g (the so-called Angell plot), as shown in b. The reduced temperature T / T g is a measure of how far above or
deep into the glass state is a sample. Remarkably, we find that the
structural relaxation times of the three pharmaceuticals coincide
in the Angell plot, which means that despite the structural differences
and the almost 40 K of difference in T g (and even more in T m ), the supercooled
liquid of these pharmaceuticals behaves cooperatively in the same
way when the distance from T g is the same.
This result is reflected in the VFT parameters listed in (in particular, in
the similar value of the fragility strength coefficient D ), and it can also be seen in the values of the so-called fragility
index ( m p ) of the amorphous samples, which
is defined as: 4 The fragility
index is virtually the same, within the error, for
DIA, NOR, and TETRA. The fragility index has often been related to
the capacity of a sample to recrystallize when heated from the amorphous
to the liquid state. − This, however, is only an empirical generalization,
and the present case confirms that such empirical rule fails, given
the identical fragility of the three samples and their noticeable
difference in recrystallization behavior. Also, the apparent activation
energy at T g , i.e., the slope of the tangent
to the Arrhenius plot of the structural relaxation at the glass transition,
cannot be taken as a reliable predictor of the tendency toward nucleation:
in fact, this parameter is again virtually identical in the case of
DIA and TETRA (see ), which exhibit instead very distinct nucleation tendency. 3.3.2 Secondary Relaxations Besides the
α relaxation, three more secondary peaks were observed in the
loss spectra at higher frequency (or lower temperature) than the cooperative
loss , both
in the supercooled liquid and the glass states. One of the secondary
relaxations, which we label as β, can be observed in all three
cases as a high-frequency shoulder to the structural peak. Another
secondary peak (γ) is observed in the glass state of all three
compounds, i.e., at low temperatures. Finally, at the lowest temperatures
studied a third secondary peak (γ′) could be discerned
in DIA and TETRA. In the case of NOR, the loss intensity at frequencies
higher than that of the γ peak was very low, so that it would
appear that the γ′ relaxation was almost absent in this
compound. We have nonetheless performed a fit of this spectral region
for completeness. All secondary relaxations could be fitted with symmetric
Cole–Cole functions (see section). a displays the full Arrhenius relaxation
maps of DIA (half points), NOR (open points), and TETRA (solid points).
As visible in this figure, all secondary relaxations displayed a simply
activated dependence on temperature, described by the Arrhenius law: 5 where τ ∞ is the characteristic
time at very high (infinite) temperature (it
plays the same role as τ 0 in the VFT ), E a is the activation energy, and R is the universal
gas constant. The β relaxation of all
three compounds displayed a kink
at T ≈ T g ( b), where its activation
energy E a, β (proportional
to the slope in the Arrhenius or Angell plots) was found to change
discontinuously (it cannot be excluded that above T g , the activation energy of the β process is actually
slightly dependent on T ). This cross-over in the
temperature dependence is typical of the so-called Johari–Goldstein
(JG) secondary relaxation, a local whole-molecule relaxation that
is strongly correlated with the structural one and that is a feature
common to most glass formers. − It can be easily seen in a and a that the difference
in glass transition temperature is reflected both in the α and
β relaxations. In fact, at the same given temperature, both
α and β relaxation times are much longer for NOR than
for DIA or TETRA, corresponding to much slower molecular dynamics.
The analysis shown in b provides a means to further verify the JG character of the
β relaxation. In fact, the β relaxations of DIA, NOR,
and TETRA are observed to be virtually superposed in the Angell plot,
where the three compounds all display a kink at T g / T ≈ 1, and the β activation
energy below T g is virtually the same
(within the error) for all three compounds (see ). The fact that the (secondary) β
relaxation time scales with T g (which
as discussed in is actually related to the kinetic arrest of the α
relaxation) is typical of JG relaxations. The study of this type of relaxation is particularly relevant
for
amorphous drugs because several studies have brought forth the idea
that the kinetic stability of a molecular glass is correlated with
the secondary β relaxation. In particular, it has been argued
experimentally that a small-molecule glass is kinetically stable only
below the onset temperature of the JG relaxation, typically few tens
of degrees below T g . In the case of the diazepines, the relaxation time of the
β JG relaxation reaches the standard value of 100 s between
30 and 40 K below the T g of the compound.
In our experiments, NOR and TETRA displayed a tendency to recrystallize
above T g , while DIA did not. It should
be noted that the onset of the β relaxation is likely a minimal
requirement for recrystallization: in our experiments, supercooled
DIA was not observed to recrystallize during a period of few days
even above the onset of the α relaxation, i.e., above T g . − The main theoretical model
concerning the JG relaxation is the
Coupling Model (hereafter, CM). , The CM interprets the
JG relaxations as a local, non-cooperative whole-molecule process,
which acts as the “precursor” at shorter times of the
α relaxation. , The characteristic CM relaxation
times in the supercooled liquid state are given by the following approximated
equation, which should approximately equal the experimental JG relaxation
times: 6 Here, t c is the correlation time (usually
of the order of 2 ps) and n , called the coupling
parameter, is related to the Havriliak–Negami exponents of
the α relaxation by the approximate relation 1 – n = ( ab ) 1/1.23 . In the case of the studied diazepines, the Havriliak–Negami
function reduces to a Cole–Davidson equation with a single
exponent b , which is found to be independent of temperature,
so that the coupling parameter is constant and equal to n = 1 – ( b ) 1/1.23 . then predicts that the β
relaxation time is perfectly correlated with the structural relaxation
time and thus scales with T g , as indeed
observed. Despite this, the relaxation times calculated with the CM
theory do not coincide with the experimental JG ones. This might be
due to the fact that the β relaxation is observed only as a
shoulder of the α peak, in which case it has been shown that
the fitting procedure that we employed does not reproduce the precursor
frequency predicted by the CM. It is nevertheless worth pointing out
that the difference at T g between the
theoretical times and the experimental ones can be off by as many
as two orders of magnitude (see b). We finally discuss the fastest secondary
relaxations observed in
our samples. These relaxations must stem from intramolecular degrees
of freedom. In the case of the benzodiazepine ring, the only degree
of freedom corresponds to the chirality inversion between P and M
conformers discussed in the previous section. Apart from this, all
three molecules possess a torsional degree of freedom corresponding
to the single covalent bond linking the fused benzodiazepine ring
with the six-membered carbon ring. There are two more degrees of freedom
in some of the derivatives, namely, the internal rotation of the methyl
group in DIA and TETRA, and a possible conformational interconversion
dynamics of the non-planar cyclohexene ring of TETRA. Neither of these
processes is expected to give rise to a dielectric relaxation feature,
due to the lack of dipole moment of either moiety, so that there are
only two possible candidates for the experimentally observed γ
and γ′ relaxations. As visible in the Angell plot
of b, neither
the γ nor the γ′
relaxation scales with the α relaxation or with the glass transition
temperature, which indicates that they correspond to local relaxation
processes of very low cooperativity. Looking at the relaxation maps
of a, it can
be seen that the three γ relaxations have very similar relaxation
times at a given fixed temperature in all three compounds and also
that the corresponding activation energies E a,γ are close for all studied diazepines . Instead, the α and
β relaxations have very different relaxation times between NOR
on one hand and DIA and TETRA on the other, as stated previously,
and the γ′ relaxation is quite separated in DIA and TETRA.
The similarity of the γ relaxation times and activation energy,
and the fact that this relaxation is unaffected by the distance from
the glass transition temperature suggest that the γ relaxation
is an intramolecular relaxation process common to all three diazepines. As mentioned in , all three studied benzodiazepines exist
in two possible
equivalent conformations of opposite chirality. Both conformers, P
and M, are present in the crystal phase of each compound. In the gas
phase and in solution, benzodiazepines are known to be relatively
flexible and to display inter-conversion dynamics between the two
equivalent conformations, accompanied by a reorientation by 60°
of the CH 2 moiety attached to the carbonyl group, as discussed,
e.g., by Mielcarek et al . The conformational dynamics of DIA and NOR was reported in previous
studies for molecules in solution, and it was found that the activation
energy was not significantly dependent on the solvent. The conformational
activation energies were found experimentally to be 74 and 52 kJ/mol
for DIA and NOR, respectively. , Because the conformational
transition is accompanied also by a
change in position of the polar carbonyl group and of the nitrogen
atoms and thus of the direction of the
molecular dipole moment, such conformational change should be observable
in dielectric spectroscopy. The fact that the γ relaxation is
observed in all three compounds at very similar relaxation times leads
us to assign this process to the inter-conversion dynamics between
P and M conformations (see inset to ). It can instead be ruled out that the γ′
relaxation can correspond to such dynamics, considering that the DIA
and NOR derivatives, which have identical fused benzodiazepine rings,
have γ′ relaxation times differing by more than two orders
of magnitude. It may seem surprising that the M–P interconversion
takes
place also in the liquid phase of NOR due to the presence of hydrogen
bonds. It must however be considered that the H-bond network in a
liquid phase is dynamic and in general only involves a fraction of
the molecules at a given time. The dielectric signal of the P–M
interconversion dynamics of NOR, namely, the γ relaxation of
this compound, likely stems from the fraction of molecules that are
not involved in H-bonding at a given time. It is worth pointing out,
in this respect, that the relaxation time and activation energies
are similar but not identical in the three compounds. We also remark
that the experimental values of the corresponding activation energy
in solution are roughly twice those of the γ relaxations reported
in . It should
however be kept in mind that the extent of H bonding will differ depending
on the liquid phase, and, more importantly, our measurements of the
γ dynamics are all in the glass state of the pure compound.
It is well-known that the temperature dependence of the structural
and JG relaxations displays an abrupt change at T g due to the loss of ergodic equilibrium when going from
the supercooled liquid to the glass phase. This is clearly visible
for the case of the β JG relaxation of benzodiazepine
in b, as discussed
earlier. The same effect is expected to be visible for any relaxation
process whose characteristic time is affected by the viscosity, and
it could be that the interconversion rate between P and M conformers
(γ relaxation time) is partially affected by changes of macroscopic
properties of the sample such as its viscosity (although it cannot
depend only on it, as b shows). Dielectric relaxation studies of flexible heterocyclic
molecules are relatively uncommon, and, to the best of our knowledge,
ours is one of the few dielectric spectroscopy studies that have provided
a clear identification of the ring conformational dynamics in polycyclic
molecules. , , Finally, concerning the γ′ relaxation, both the
range
of temperature in which it is observed and its characteristic relaxation
time are very different between DIA and TETRA, as mentioned, albeit
that its activation energy is of the same order of magnitude in both
compounds. Given that this relaxation is virtually absent in NOR,
it is likely that it is suppressed or at least strongly hindered by
the presence of intermolecular hydrogen bonds. All three studied benzodiazepines
have, as mentioned, a further degree of freedom, corresponding to
the torsional rotation around the covalent bond linking the fused
double ring with the six-membered carbon ring. , While the latter has basically no dipole moment, a rotation of the
double ring about this covalent bond could lead to a rigid rotation
of the molecular dipole moment, which would contribute a dielectric
loss signal. Therefore, we tentatively assign the γ′
relaxation to the rigid rotation, likely by a small angle, of the
double ring about its bond with the six-membered carbon ring. Such
rotation might be partially hindered, in the case of NOR, by the presence
of a network of intermolecular hydrogen bonds, which rationalizes
the extremely weak signal of the γ′ relaxation in this
compound. The difference between the γ′ activation energy
and relaxation times of DIA and TETRA might then be attributed to
the different steric hindrance of the two distinct six-member rings,
namely, a bulkier phenyl ring in the case of DIA and a non-planar
cyclohexene ring in the case of TETRA. This tentative interpretation
is consistent with the much faster γ′ relaxation dynamics
in TETRA. 3.3.3 Crystallization Kinetics Dielectric
spectroscopy was employed to determine the kinetics of isothermal
recrystallization from the supercooled liquid state of NOR and TETRA
(as mentioned, DIA was not observed to recrystallize in short times).
To this purpose, we acquired series of dielectric spectra at fixed
temperature and analyzed the variation in time of the static dielectric
constant, which is related to the dielectric intensity of the structural
relaxation process. Since NOR has significantly higher glass transition
temperature than TETRA, at temperatures at which the latter compound
showed recrystallization at detectable rates, NOR is close to being
in the glass state, where the recrystallization onset time and recrystallization
rates are too long to allow a dielectric measurement. Therefore, because
such “isothermal comparison” of the recrystallization
process cannot be carried out, we have chosen different temperatures
to study recrystallization at roughly the same reduced temperature T / T g . displays the series of isothermal
permittivity spectra (real and imaginary part) during recrystallization
of TETRA at T = 331 K (corresponding to T / T g,TETRA = 1.07) and of NOR at T = 375 K (corresponding to T / T g,NOR = 1.08). The effect of recrystallization is visible
as a decrease over time of the dielectric intensity of the α
loss feature, or equivalently a decrease of the static permittivity
value ε s , defined as the value of ε′( f ) at the lowest frequency displayed in the figure ( f = 1 Hz for TETRA and f = 2 Hz for NOR,
respectively). The onset time t o of the
recrystallization process was determined as the time at which the
initially constant value of ε s in the supercooled
liquid phase was observed to start decreasing. The evolution of ε s with time elapsed from the start of the recrystallization
is displayed in e. It is clear that the recrystallization of NOR at T / T g,NOR = 1.08 is slower than that of
TETRA at T / T g,TETRA =
1.07, despite the fact that the structural (α) relaxation frequency
and thus the cooperative mobility are, under such conditions, higher
by a factor of four in NOR than in TETRA, as testified by the position
of the loss maxima in panels (c) and (d) of . In order to study the kinetics of recrystallization, we define
as customary , a normalized static permittivity
value as: 7 Here, ε s (SL)
and ε s (C) are the
static permittivity of the supercooled liquid and the crystal phase,
as measured before the onset of nucleation of the crystal phase and
at the end of the crystal growth, respectively, while ε s ( t ) is the static permittivity of the partially
recrystallized sample as function of time. The global kinetics of
crystallization can be modeled with the help of the Avrami equation, , which is based on the nucleation-and-growth model of the transition
from the liquid to the crystal phase. According to this model, the
renormalized static permittivity should vary in time as: , 8 Here, n is
the Avrami exponent and Z is a constant from which
the recrystallization rate in s –1 can be obtained , as k = Z 1/ n . According to , the quantity ln(−ln(1 – ε n )) should
be linearly proportional to the logarithm of the time elapsed since
the onset of recrystallization, t – t o . This is indeed observed in the Avrami plot
displayed in f. The values of the obtained fit parameters are n = 1.01 ± 0.05, k = (7 ± 3)·10 –5 s –1 for TETRA and n = 1.1 ± 0.1, k = (4 ± 2)·10 –5 s –1 for NOR. The fact that the
value of the Avrami exponent is close to unity for both derivatives
indicates a strongly anisotropic (one-dimensional) growth of the crystal
phase after a sporadic nucleation event. , , A value of n = 1 also allows
direct estimation of the crystal growth rate, that is, separation
of the nucleation and crystal growth phases of the recrystallization. The vertical separation in f, in which assuming an identical
value of n can be related to the difference in recrystallization
rate k between the two samples (see the discussion
of of ref , confirms the slower crystal
growth kinetics directly visible in e, and is consistent with the experimental
ranges of values of the recrystallization rate k of
TETRA and NOR under these conditions. We also studied the recrystallization
of NOR at T = 368 K ( T/T g = 1.06). The latter temperature
was chosen so that the structural relaxation frequency was the same
for both compounds (a condition usually referred to as “isochronal
condition” in the scientific literature). Because the two compounds
have similar fragility indexes, this condition is very similar to
that of same reduced temperature, T / T g . The crystal growth rate of NOR was so slow under these
conditions (at a temperature only 5 K below the crystallization temperature
of ) that we
could not complete it during three full days of continuous measurements.
The crystallization (growth) rate k for NOR at 368
K ( k = (7 ± 3)·10 –6 s –1 ) was one order of magnitude smaller than that for
TETRA at 331 K, and our experiments show that the (homogeneous) nucleation
time is very different in DIA with respect to its derivatives.
Differential Scanning Calorimetry Results shows the
DSC traces obtained for the three diazepines DIA, NOR, and TETRA.
In all three cases, the as-received powders were completely crystalline,
as the first heating ramp only displayed a melting endotherm with
onsets at 404.1, 487.0, and 415.6 K for DIA, NOR, and TETRA, respectively.
Values coincide within the experimental error with those available
in the scientific literature. − , , The melting point of NOR and the enthalpy of melting are both significantly
higher than that of the other two derivatives, likely due to the presence
of N-H···O=C hydrogen bonds, which can only form in
demethylated derivative (see the next section). The subsequent
cooling ramp leads to a glassy phase for all three pharmaceuticals,
and on reheating, a step-like transition can be observed in the DSC
traces, corresponding to the glass transition temperature ( T g ). In most cases, though not in all DSC runs,
TETRA and NOR displayed (at least partial) recrystallization of the
supercooled liquid in the heat up run, followed again by the melting
peak (see inset to b). The recrystallized phase is the same as the initial one,
as the melting temperature is the same on heating the recrystallized
sample. The supercooled TETRA and NOR liquids were observed to crystallize
also in dielectric spectroscopy experiments (see ), while recrystallization
of DIA was absent also in this case. The sample geometry and the vessel
are quite different in DSC (droplet in aluminum pan) and dielectric
(film in stainless steel cylinder with silica spacers) experiments.
The fact that the three samples displayed the same tendency toward
recrystallization under such different experimental conditions indicates
that the recrystallization of TETRA and NOR probably took place by
homogeneous (rather than heterogeneous) nucleation of the crystal
phase. The characteristic onset temperatures of the glass transition,
recrystallization, and melting points are listed in for all three pharmaceutically
active compounds, together with the melting enthalpies. The recrystallization
temperature is only listed for completeness, as it did not always
occur in all DSC scans at the same temperature. This is not surprising,
as nucleation is a stochastic event that depends on the characteristics
of the sample (heterogeneous vs homogeneous nucleation) and its history
(e.g., cooling rate from the liquid phase, temperature at which it
is then kept). It may be seen that T m and T g roughly scale with one another:
the T g / T m ratio
is 0.78
for DIA, 0.71 for NOR, and 0.75 for TETRA. The values for TETRA and
DIA are quite similar, albeit T m is slightly
higher for TETRA than that for DIA, while T g is somewhat lower for TETRA than that for DIA. The glass transition
temperature is often found to display a correlation with the molecular
weight M w . In particular, the empirical
rule T g ≈ M w 1/2 appears to be fulfilled in the case of van
der Waals molecular liquids. Such correlation
probably reflects the fact that the extent of van der Waals interactions
increases with the molecular mass (due to the increase of molecular
polarizability and of the closest intermolecular contacts), and the
fact that, at a given fixed temperature, a massive molecule has lower
mobility, but it does not take into account hydrogen bonding or any
other type of directional intermolecular bonds. In fact, the glass
transition temperature of the studied diazepines does not correlate
with the molecular weight: NOR, which has the lowest weight, has the
highest glass transition temperature. The origin of the higher T g is likely the same as that of the higher T m , namely, the presence of intermolecular H-bonds
in the liquid phase of NOR. Indeed, in the absence of any H bonding
the aforementioned correlation of molecular weight and glass transition
temperature would result in a T g value
of NOR closer to those of DIA and TETRA, which is not observed.
X-ray Diffraction Results and Analysis All three compounds display, in the crystalline phase, the same monoclinic
space group (P 2 1 /c). The diazepine ring of all molecules
adopts a bent boat-like conformation, with two possible isoenergetic
conformers, which are mirror images of one another. The two conformers
have opposite chirality and are named P (plus) or M (minus) according
to the sign of the (O=)C–C(H 2 )–N=C
torsion angle (see the inset to ). All three crystals contain a 1:1 ratio of P and
M conformers. The geometry of the conformers is similar in all three
compounds. For example, the angle formed by the C=N bond with
the plane of the fused benzene ring is equal to 41.6, 38.5, and 48.6°
in crystalline DIA, NOR, and TETRA, respectively. The analysis
of the X-ray structures at room temperature shows unambiguously that
NOR is the only compound of the three related drugs studied that forms
strong hydrogen bonds in the crystalline state, namely, intermolecular
N–H···O bonds involving the amine nitrogen of
the diazepine ring and the carbonyl oxygen of the same group of a
nearest-neighbor molecule in the crystal structure (see ). This is in agreement with
the higher melting point and enthalpy of fusion of NOR compared with
the other two compounds . It is interesting to point out in this respect that while
in both crystalline DIA and TETRA the carbonyl group and the adjacent
methyl group are basically coplanar, with a H 3 C–N–C=O
torsion angle smaller than 2°, in the case of NOR, which is a priori the only compound where the corresponding (peptide)
moiety is expected to be planar due to the amide electronic resonance,
the H–N–C=O torsion angle is instead approximately
10°. Non-planar peptide bonds are not uncommon in H-bonded structures
such as proteins in their native state. In the case of crystalline NOR, the lack of planarity of the amide
group is likely a consequence of H-bond formation. A recent work by some of us has shown that DIA and
TETRA, while
not forming N–H···O bonds, display weak but
extensive C–H···O interactions between the electron-rich
carbonyl group and the weakly polar C–H bonds of CH 2 groups. While intermolecular N–H···O
bonds are at least partially present also in the amorphous state of
NOR, as testified by its much higher glass transition temperature
(see ),
it is unlikely that the C–H···O interactions
play any role in the amorphous state of the three compounds, as we
argue further in . A straightforward comparison of the hydrogen
bond scheme in the
solid state of the three compounds can be carried out based on the
analysis of the Hirshfeld surface areas (see ). This surface represents a particular way
of partitioning the overall electron density in a molecular crystal
into individual molecular units, which
provides a three-dimensional image of the close contacts in the crystal
by guaranteeing maximum proximity of the corresponding Hirshfeld volumes
of nearest-neighbor molecules. − The color code employed by convention
is that a yellow or red color indicates points of short intermolecular
contact, while blue indicates regions of the Hirshfeld surface corresponding
to directions in which the intermolecular distance is comparatively
longer. ,
adapted
from ref , shows
the key intermolecular contacts derived from the Hirshfeld surface
area analysis at room temperature in the crystalline state. It evidences
the relevance of the hydrogen bond scheme for these compounds and,
in particular, that of the O···H for NOR compared to
DIA and TETRA, in agreement with the role of the strong N–H···O
H–bond interaction in NOR. It is interesting to note that there is a correlation
between melting
point, density, and Hirshfeld surface and volume parameters . In particular,
the Hirshfeld molecular volume and surface and the Hirshfeld volume
normalized to molecular weight are the largest for TETRA, which has
the smallest density and the lowest T m of the three derivatives, and they are the smallest for NOR, which
has the largest density and highest T m . This correlation evidences the influence on the melting temperature
of the hydrogen bonds in crystalline NOR. We point out that
the correlation is instead not strictly verified
when considering the glass transition temperature of all derivatives,
as T g,DIA > T g,TETRA . However, as mentioned, the T g of NOR
is significantly higher than that of the other two compounds, which
is indicative of the presence of some H bonding also in the liquid
phase of this compound. Instead of tightly bound stable H-bonded dimers
in the liquid phase, only short-lived H bonds are expected to occur,
and it is likely that a given NOR molecule only takes part, at most,
in one H-bond at a time.
Broadband Dielectric Spectroscopy
Results In order to see in detail how the small difference
in molecular
formula as well as the relevance of the hydrogen-bond network between
the three studied benzodiazepines affects the molecular mobility and
conformational dynamics in the amorphous state, we carried out dielectric
spectroscopy experiments on all three compounds in their amorphous
states. shows
the dielectric loss function of the three compounds at few selected
temperatures, plotted against the frequency of the applied electric
field. 3.3.1 Structural Relaxation For all three
diazepines, the most intense loss peak is observed at high temperatures
, and corresponds
to the structural relaxation (or α relaxation) of the supercooled
liquid phase. Below the calorimetric glass transition temperature T g (at which τ α = 10 2 s), the peak frequency of the α relaxation lies outside
the experimental frequency window, and only the tail of the α
peak is observed. When the temperature is increased above T g , the onset of the cooperative relaxation dynamics
of the liquid phase is signaled by the appearance in the experimental
frequency window of the α peak maximum, which then shifts to
higher frequencies as the temperature is further increased. The intensity of the α loss feature of both DIA and NOR is
roughly constant above T g . Instead, recrystallization
upon heating can be clearly discerned in the series of loss spectra
in the case of TETRA. Indeed, at temperatures higher than 335 K the
dielectric intensity of the α peak of TETRA is observed to decrease
further and further as the amorphous fraction in the sample decreases
(the dielectric loss intensity is proportional to the number density
of molecules in the amorphous supercooled liquid state ). To analyze the relaxation dynamics of
the cooperative α relaxation
in detail, we fitted all dielectric spectra as the sum of several
Havriliak–Negami components (see ), each corresponding to a distinct relaxation,
in order to extract the temperature-dependent relaxation times ( , see the section). The fits are shown in along with experimental
data. We found in particular that the fit with Havriliak–Negami
curves resulted in a Cole–Davidson function for the structural
relaxation. It can be observed in that the α peak of each compound has
exactly
the same shape regardless of temperature: the isothermal spectra at
various temperatures could be superposed onto one another by rescaling
the frequency scale and the signal intensity to those of the loss
maximum. This master-curve scaling was employed in the fitting procedure,
by imposing the same Cole–Davidson (CD) exponent in all high-temperature
spectra of a given compound, as indicated for selected temperatures
in the three panels of . The CD exponent that best described the α peaks was
found to be b = 0.59 ± 0.03 for DIA and TETRA,
and b = 0.50 ± 0.02 for NOR. This result indicates
a slightly greater cooperativity for NOR with respect to DIA and TETRA, , possibly related to the presence of intermolecular H-bonds in NOR. shows the
α relaxation times of all three studied diazepines versus the
inverse temperature (Arrhenius plot). The α relaxation time
follows the Vogel–Fulcher–Tamman temperature-dependence
typical of cooperative structural relaxations: − 3 Here, τ 0 is
the characteristic time at infinite temperature, D is the fragility strength coefficient, and T 0 is the Vogel–Fulcher temperature. The so-called “kinetic”
or “dielectric” glass transition temperature T g of the sample is defined as the temperature
at which relaxation times reaches 100 s, i.e., where log 10 (τ α /[s]) = 2 (horizontal yellow line in a). The kinetic glass
transition temperatures are 312.6, 309.0, and 347.2 K for DIA, TETRA,
and NOR, respectively . These values are very similar to the ones found in DSC
(see ), as expected. It is interesting
to compare the dependence of the relaxation times
with the inverse temperature rescaled to T g (the so-called Angell plot), as shown in b. The reduced temperature T / T g is a measure of how far above or
deep into the glass state is a sample. Remarkably, we find that the
structural relaxation times of the three pharmaceuticals coincide
in the Angell plot, which means that despite the structural differences
and the almost 40 K of difference in T g (and even more in T m ), the supercooled
liquid of these pharmaceuticals behaves cooperatively in the same
way when the distance from T g is the same.
This result is reflected in the VFT parameters listed in (in particular, in
the similar value of the fragility strength coefficient D ), and it can also be seen in the values of the so-called fragility
index ( m p ) of the amorphous samples, which
is defined as: 4 The fragility
index is virtually the same, within the error, for
DIA, NOR, and TETRA. The fragility index has often been related to
the capacity of a sample to recrystallize when heated from the amorphous
to the liquid state. − This, however, is only an empirical generalization,
and the present case confirms that such empirical rule fails, given
the identical fragility of the three samples and their noticeable
difference in recrystallization behavior. Also, the apparent activation
energy at T g , i.e., the slope of the tangent
to the Arrhenius plot of the structural relaxation at the glass transition,
cannot be taken as a reliable predictor of the tendency toward nucleation:
in fact, this parameter is again virtually identical in the case of
DIA and TETRA (see ), which exhibit instead very distinct nucleation tendency. 3.3.2 Secondary Relaxations Besides the
α relaxation, three more secondary peaks were observed in the
loss spectra at higher frequency (or lower temperature) than the cooperative
loss , both
in the supercooled liquid and the glass states. One of the secondary
relaxations, which we label as β, can be observed in all three
cases as a high-frequency shoulder to the structural peak. Another
secondary peak (γ) is observed in the glass state of all three
compounds, i.e., at low temperatures. Finally, at the lowest temperatures
studied a third secondary peak (γ′) could be discerned
in DIA and TETRA. In the case of NOR, the loss intensity at frequencies
higher than that of the γ peak was very low, so that it would
appear that the γ′ relaxation was almost absent in this
compound. We have nonetheless performed a fit of this spectral region
for completeness. All secondary relaxations could be fitted with symmetric
Cole–Cole functions (see section). a displays the full Arrhenius relaxation
maps of DIA (half points), NOR (open points), and TETRA (solid points).
As visible in this figure, all secondary relaxations displayed a simply
activated dependence on temperature, described by the Arrhenius law: 5 where τ ∞ is the characteristic
time at very high (infinite) temperature (it
plays the same role as τ 0 in the VFT ), E a is the activation energy, and R is the universal
gas constant. The β relaxation of all
three compounds displayed a kink
at T ≈ T g ( b), where its activation
energy E a, β (proportional
to the slope in the Arrhenius or Angell plots) was found to change
discontinuously (it cannot be excluded that above T g , the activation energy of the β process is actually
slightly dependent on T ). This cross-over in the
temperature dependence is typical of the so-called Johari–Goldstein
(JG) secondary relaxation, a local whole-molecule relaxation that
is strongly correlated with the structural one and that is a feature
common to most glass formers. − It can be easily seen in a and a that the difference
in glass transition temperature is reflected both in the α and
β relaxations. In fact, at the same given temperature, both
α and β relaxation times are much longer for NOR than
for DIA or TETRA, corresponding to much slower molecular dynamics.
The analysis shown in b provides a means to further verify the JG character of the
β relaxation. In fact, the β relaxations of DIA, NOR,
and TETRA are observed to be virtually superposed in the Angell plot,
where the three compounds all display a kink at T g / T ≈ 1, and the β activation
energy below T g is virtually the same
(within the error) for all three compounds (see ). The fact that the (secondary) β
relaxation time scales with T g (which
as discussed in is actually related to the kinetic arrest of the α
relaxation) is typical of JG relaxations. The study of this type of relaxation is particularly relevant
for
amorphous drugs because several studies have brought forth the idea
that the kinetic stability of a molecular glass is correlated with
the secondary β relaxation. In particular, it has been argued
experimentally that a small-molecule glass is kinetically stable only
below the onset temperature of the JG relaxation, typically few tens
of degrees below T g . In the case of the diazepines, the relaxation time of the
β JG relaxation reaches the standard value of 100 s between
30 and 40 K below the T g of the compound.
In our experiments, NOR and TETRA displayed a tendency to recrystallize
above T g , while DIA did not. It should
be noted that the onset of the β relaxation is likely a minimal
requirement for recrystallization: in our experiments, supercooled
DIA was not observed to recrystallize during a period of few days
even above the onset of the α relaxation, i.e., above T g . − The main theoretical model
concerning the JG relaxation is the
Coupling Model (hereafter, CM). , The CM interprets the
JG relaxations as a local, non-cooperative whole-molecule process,
which acts as the “precursor” at shorter times of the
α relaxation. , The characteristic CM relaxation
times in the supercooled liquid state are given by the following approximated
equation, which should approximately equal the experimental JG relaxation
times: 6 Here, t c is the correlation time (usually
of the order of 2 ps) and n , called the coupling
parameter, is related to the Havriliak–Negami exponents of
the α relaxation by the approximate relation 1 – n = ( ab ) 1/1.23 . In the case of the studied diazepines, the Havriliak–Negami
function reduces to a Cole–Davidson equation with a single
exponent b , which is found to be independent of temperature,
so that the coupling parameter is constant and equal to n = 1 – ( b ) 1/1.23 . then predicts that the β
relaxation time is perfectly correlated with the structural relaxation
time and thus scales with T g , as indeed
observed. Despite this, the relaxation times calculated with the CM
theory do not coincide with the experimental JG ones. This might be
due to the fact that the β relaxation is observed only as a
shoulder of the α peak, in which case it has been shown that
the fitting procedure that we employed does not reproduce the precursor
frequency predicted by the CM. It is nevertheless worth pointing out
that the difference at T g between the
theoretical times and the experimental ones can be off by as many
as two orders of magnitude (see b). We finally discuss the fastest secondary
relaxations observed in
our samples. These relaxations must stem from intramolecular degrees
of freedom. In the case of the benzodiazepine ring, the only degree
of freedom corresponds to the chirality inversion between P and M
conformers discussed in the previous section. Apart from this, all
three molecules possess a torsional degree of freedom corresponding
to the single covalent bond linking the fused benzodiazepine ring
with the six-membered carbon ring. There are two more degrees of freedom
in some of the derivatives, namely, the internal rotation of the methyl
group in DIA and TETRA, and a possible conformational interconversion
dynamics of the non-planar cyclohexene ring of TETRA. Neither of these
processes is expected to give rise to a dielectric relaxation feature,
due to the lack of dipole moment of either moiety, so that there are
only two possible candidates for the experimentally observed γ
and γ′ relaxations. As visible in the Angell plot
of b, neither
the γ nor the γ′
relaxation scales with the α relaxation or with the glass transition
temperature, which indicates that they correspond to local relaxation
processes of very low cooperativity. Looking at the relaxation maps
of a, it can
be seen that the three γ relaxations have very similar relaxation
times at a given fixed temperature in all three compounds and also
that the corresponding activation energies E a,γ are close for all studied diazepines . Instead, the α and
β relaxations have very different relaxation times between NOR
on one hand and DIA and TETRA on the other, as stated previously,
and the γ′ relaxation is quite separated in DIA and TETRA.
The similarity of the γ relaxation times and activation energy,
and the fact that this relaxation is unaffected by the distance from
the glass transition temperature suggest that the γ relaxation
is an intramolecular relaxation process common to all three diazepines. As mentioned in , all three studied benzodiazepines exist
in two possible
equivalent conformations of opposite chirality. Both conformers, P
and M, are present in the crystal phase of each compound. In the gas
phase and in solution, benzodiazepines are known to be relatively
flexible and to display inter-conversion dynamics between the two
equivalent conformations, accompanied by a reorientation by 60°
of the CH 2 moiety attached to the carbonyl group, as discussed,
e.g., by Mielcarek et al . The conformational dynamics of DIA and NOR was reported in previous
studies for molecules in solution, and it was found that the activation
energy was not significantly dependent on the solvent. The conformational
activation energies were found experimentally to be 74 and 52 kJ/mol
for DIA and NOR, respectively. , Because the conformational
transition is accompanied also by a
change in position of the polar carbonyl group and of the nitrogen
atoms and thus of the direction of the
molecular dipole moment, such conformational change should be observable
in dielectric spectroscopy. The fact that the γ relaxation is
observed in all three compounds at very similar relaxation times leads
us to assign this process to the inter-conversion dynamics between
P and M conformations (see inset to ). It can instead be ruled out that the γ′
relaxation can correspond to such dynamics, considering that the DIA
and NOR derivatives, which have identical fused benzodiazepine rings,
have γ′ relaxation times differing by more than two orders
of magnitude. It may seem surprising that the M–P interconversion
takes
place also in the liquid phase of NOR due to the presence of hydrogen
bonds. It must however be considered that the H-bond network in a
liquid phase is dynamic and in general only involves a fraction of
the molecules at a given time. The dielectric signal of the P–M
interconversion dynamics of NOR, namely, the γ relaxation of
this compound, likely stems from the fraction of molecules that are
not involved in H-bonding at a given time. It is worth pointing out,
in this respect, that the relaxation time and activation energies
are similar but not identical in the three compounds. We also remark
that the experimental values of the corresponding activation energy
in solution are roughly twice those of the γ relaxations reported
in . It should
however be kept in mind that the extent of H bonding will differ depending
on the liquid phase, and, more importantly, our measurements of the
γ dynamics are all in the glass state of the pure compound.
It is well-known that the temperature dependence of the structural
and JG relaxations displays an abrupt change at T g due to the loss of ergodic equilibrium when going from
the supercooled liquid to the glass phase. This is clearly visible
for the case of the β JG relaxation of benzodiazepine
in b, as discussed
earlier. The same effect is expected to be visible for any relaxation
process whose characteristic time is affected by the viscosity, and
it could be that the interconversion rate between P and M conformers
(γ relaxation time) is partially affected by changes of macroscopic
properties of the sample such as its viscosity (although it cannot
depend only on it, as b shows). Dielectric relaxation studies of flexible heterocyclic
molecules are relatively uncommon, and, to the best of our knowledge,
ours is one of the few dielectric spectroscopy studies that have provided
a clear identification of the ring conformational dynamics in polycyclic
molecules. , , Finally, concerning the γ′ relaxation, both the
range
of temperature in which it is observed and its characteristic relaxation
time are very different between DIA and TETRA, as mentioned, albeit
that its activation energy is of the same order of magnitude in both
compounds. Given that this relaxation is virtually absent in NOR,
it is likely that it is suppressed or at least strongly hindered by
the presence of intermolecular hydrogen bonds. All three studied benzodiazepines
have, as mentioned, a further degree of freedom, corresponding to
the torsional rotation around the covalent bond linking the fused
double ring with the six-membered carbon ring. , While the latter has basically no dipole moment, a rotation of the
double ring about this covalent bond could lead to a rigid rotation
of the molecular dipole moment, which would contribute a dielectric
loss signal. Therefore, we tentatively assign the γ′
relaxation to the rigid rotation, likely by a small angle, of the
double ring about its bond with the six-membered carbon ring. Such
rotation might be partially hindered, in the case of NOR, by the presence
of a network of intermolecular hydrogen bonds, which rationalizes
the extremely weak signal of the γ′ relaxation in this
compound. The difference between the γ′ activation energy
and relaxation times of DIA and TETRA might then be attributed to
the different steric hindrance of the two distinct six-member rings,
namely, a bulkier phenyl ring in the case of DIA and a non-planar
cyclohexene ring in the case of TETRA. This tentative interpretation
is consistent with the much faster γ′ relaxation dynamics
in TETRA. 3.3.3 Crystallization Kinetics Dielectric
spectroscopy was employed to determine the kinetics of isothermal
recrystallization from the supercooled liquid state of NOR and TETRA
(as mentioned, DIA was not observed to recrystallize in short times).
To this purpose, we acquired series of dielectric spectra at fixed
temperature and analyzed the variation in time of the static dielectric
constant, which is related to the dielectric intensity of the structural
relaxation process. Since NOR has significantly higher glass transition
temperature than TETRA, at temperatures at which the latter compound
showed recrystallization at detectable rates, NOR is close to being
in the glass state, where the recrystallization onset time and recrystallization
rates are too long to allow a dielectric measurement. Therefore, because
such “isothermal comparison” of the recrystallization
process cannot be carried out, we have chosen different temperatures
to study recrystallization at roughly the same reduced temperature T / T g . displays the series of isothermal
permittivity spectra (real and imaginary part) during recrystallization
of TETRA at T = 331 K (corresponding to T / T g,TETRA = 1.07) and of NOR at T = 375 K (corresponding to T / T g,NOR = 1.08). The effect of recrystallization is visible
as a decrease over time of the dielectric intensity of the α
loss feature, or equivalently a decrease of the static permittivity
value ε s , defined as the value of ε′( f ) at the lowest frequency displayed in the figure ( f = 1 Hz for TETRA and f = 2 Hz for NOR,
respectively). The onset time t o of the
recrystallization process was determined as the time at which the
initially constant value of ε s in the supercooled
liquid phase was observed to start decreasing. The evolution of ε s with time elapsed from the start of the recrystallization
is displayed in e. It is clear that the recrystallization of NOR at T / T g,NOR = 1.08 is slower than that of
TETRA at T / T g,TETRA =
1.07, despite the fact that the structural (α) relaxation frequency
and thus the cooperative mobility are, under such conditions, higher
by a factor of four in NOR than in TETRA, as testified by the position
of the loss maxima in panels (c) and (d) of . In order to study the kinetics of recrystallization, we define
as customary , a normalized static permittivity
value as: 7 Here, ε s (SL)
and ε s (C) are the
static permittivity of the supercooled liquid and the crystal phase,
as measured before the onset of nucleation of the crystal phase and
at the end of the crystal growth, respectively, while ε s ( t ) is the static permittivity of the partially
recrystallized sample as function of time. The global kinetics of
crystallization can be modeled with the help of the Avrami equation, , which is based on the nucleation-and-growth model of the transition
from the liquid to the crystal phase. According to this model, the
renormalized static permittivity should vary in time as: , 8 Here, n is
the Avrami exponent and Z is a constant from which
the recrystallization rate in s –1 can be obtained , as k = Z 1/ n . According to , the quantity ln(−ln(1 – ε n )) should
be linearly proportional to the logarithm of the time elapsed since
the onset of recrystallization, t – t o . This is indeed observed in the Avrami plot
displayed in f. The values of the obtained fit parameters are n = 1.01 ± 0.05, k = (7 ± 3)·10 –5 s –1 for TETRA and n = 1.1 ± 0.1, k = (4 ± 2)·10 –5 s –1 for NOR. The fact that the
value of the Avrami exponent is close to unity for both derivatives
indicates a strongly anisotropic (one-dimensional) growth of the crystal
phase after a sporadic nucleation event. , , A value of n = 1 also allows
direct estimation of the crystal growth rate, that is, separation
of the nucleation and crystal growth phases of the recrystallization. The vertical separation in f, in which assuming an identical
value of n can be related to the difference in recrystallization
rate k between the two samples (see the discussion
of of ref , confirms the slower crystal
growth kinetics directly visible in e, and is consistent with the experimental
ranges of values of the recrystallization rate k of
TETRA and NOR under these conditions. We also studied the recrystallization
of NOR at T = 368 K ( T/T g = 1.06). The latter temperature
was chosen so that the structural relaxation frequency was the same
for both compounds (a condition usually referred to as “isochronal
condition” in the scientific literature). Because the two compounds
have similar fragility indexes, this condition is very similar to
that of same reduced temperature, T / T g . The crystal growth rate of NOR was so slow under these
conditions (at a temperature only 5 K below the crystallization temperature
of ) that we
could not complete it during three full days of continuous measurements.
The crystallization (growth) rate k for NOR at 368
K ( k = (7 ± 3)·10 –6 s –1 ) was one order of magnitude smaller than that for
TETRA at 331 K, and our experiments show that the (homogeneous) nucleation
time is very different in DIA with respect to its derivatives.
Structural Relaxation For all three
diazepines, the most intense loss peak is observed at high temperatures
, and corresponds
to the structural relaxation (or α relaxation) of the supercooled
liquid phase. Below the calorimetric glass transition temperature T g (at which τ α = 10 2 s), the peak frequency of the α relaxation lies outside
the experimental frequency window, and only the tail of the α
peak is observed. When the temperature is increased above T g , the onset of the cooperative relaxation dynamics
of the liquid phase is signaled by the appearance in the experimental
frequency window of the α peak maximum, which then shifts to
higher frequencies as the temperature is further increased. The intensity of the α loss feature of both DIA and NOR is
roughly constant above T g . Instead, recrystallization
upon heating can be clearly discerned in the series of loss spectra
in the case of TETRA. Indeed, at temperatures higher than 335 K the
dielectric intensity of the α peak of TETRA is observed to decrease
further and further as the amorphous fraction in the sample decreases
(the dielectric loss intensity is proportional to the number density
of molecules in the amorphous supercooled liquid state ). To analyze the relaxation dynamics of
the cooperative α relaxation
in detail, we fitted all dielectric spectra as the sum of several
Havriliak–Negami components (see ), each corresponding to a distinct relaxation,
in order to extract the temperature-dependent relaxation times ( , see the section). The fits are shown in along with experimental
data. We found in particular that the fit with Havriliak–Negami
curves resulted in a Cole–Davidson function for the structural
relaxation. It can be observed in that the α peak of each compound has
exactly
the same shape regardless of temperature: the isothermal spectra at
various temperatures could be superposed onto one another by rescaling
the frequency scale and the signal intensity to those of the loss
maximum. This master-curve scaling was employed in the fitting procedure,
by imposing the same Cole–Davidson (CD) exponent in all high-temperature
spectra of a given compound, as indicated for selected temperatures
in the three panels of . The CD exponent that best described the α peaks was
found to be b = 0.59 ± 0.03 for DIA and TETRA,
and b = 0.50 ± 0.02 for NOR. This result indicates
a slightly greater cooperativity for NOR with respect to DIA and TETRA, , possibly related to the presence of intermolecular H-bonds in NOR. shows the
α relaxation times of all three studied diazepines versus the
inverse temperature (Arrhenius plot). The α relaxation time
follows the Vogel–Fulcher–Tamman temperature-dependence
typical of cooperative structural relaxations: − 3 Here, τ 0 is
the characteristic time at infinite temperature, D is the fragility strength coefficient, and T 0 is the Vogel–Fulcher temperature. The so-called “kinetic”
or “dielectric” glass transition temperature T g of the sample is defined as the temperature
at which relaxation times reaches 100 s, i.e., where log 10 (τ α /[s]) = 2 (horizontal yellow line in a). The kinetic glass
transition temperatures are 312.6, 309.0, and 347.2 K for DIA, TETRA,
and NOR, respectively . These values are very similar to the ones found in DSC
(see ), as expected. It is interesting
to compare the dependence of the relaxation times
with the inverse temperature rescaled to T g (the so-called Angell plot), as shown in b. The reduced temperature T / T g is a measure of how far above or
deep into the glass state is a sample. Remarkably, we find that the
structural relaxation times of the three pharmaceuticals coincide
in the Angell plot, which means that despite the structural differences
and the almost 40 K of difference in T g (and even more in T m ), the supercooled
liquid of these pharmaceuticals behaves cooperatively in the same
way when the distance from T g is the same.
This result is reflected in the VFT parameters listed in (in particular, in
the similar value of the fragility strength coefficient D ), and it can also be seen in the values of the so-called fragility
index ( m p ) of the amorphous samples, which
is defined as: 4 The fragility
index is virtually the same, within the error, for
DIA, NOR, and TETRA. The fragility index has often been related to
the capacity of a sample to recrystallize when heated from the amorphous
to the liquid state. − This, however, is only an empirical generalization,
and the present case confirms that such empirical rule fails, given
the identical fragility of the three samples and their noticeable
difference in recrystallization behavior. Also, the apparent activation
energy at T g , i.e., the slope of the tangent
to the Arrhenius plot of the structural relaxation at the glass transition,
cannot be taken as a reliable predictor of the tendency toward nucleation:
in fact, this parameter is again virtually identical in the case of
DIA and TETRA (see ), which exhibit instead very distinct nucleation tendency.
Secondary Relaxations Besides the
α relaxation, three more secondary peaks were observed in the
loss spectra at higher frequency (or lower temperature) than the cooperative
loss , both
in the supercooled liquid and the glass states. One of the secondary
relaxations, which we label as β, can be observed in all three
cases as a high-frequency shoulder to the structural peak. Another
secondary peak (γ) is observed in the glass state of all three
compounds, i.e., at low temperatures. Finally, at the lowest temperatures
studied a third secondary peak (γ′) could be discerned
in DIA and TETRA. In the case of NOR, the loss intensity at frequencies
higher than that of the γ peak was very low, so that it would
appear that the γ′ relaxation was almost absent in this
compound. We have nonetheless performed a fit of this spectral region
for completeness. All secondary relaxations could be fitted with symmetric
Cole–Cole functions (see section). a displays the full Arrhenius relaxation
maps of DIA (half points), NOR (open points), and TETRA (solid points).
As visible in this figure, all secondary relaxations displayed a simply
activated dependence on temperature, described by the Arrhenius law: 5 where τ ∞ is the characteristic
time at very high (infinite) temperature (it
plays the same role as τ 0 in the VFT ), E a is the activation energy, and R is the universal
gas constant. The β relaxation of all
three compounds displayed a kink
at T ≈ T g ( b), where its activation
energy E a, β (proportional
to the slope in the Arrhenius or Angell plots) was found to change
discontinuously (it cannot be excluded that above T g , the activation energy of the β process is actually
slightly dependent on T ). This cross-over in the
temperature dependence is typical of the so-called Johari–Goldstein
(JG) secondary relaxation, a local whole-molecule relaxation that
is strongly correlated with the structural one and that is a feature
common to most glass formers. − It can be easily seen in a and a that the difference
in glass transition temperature is reflected both in the α and
β relaxations. In fact, at the same given temperature, both
α and β relaxation times are much longer for NOR than
for DIA or TETRA, corresponding to much slower molecular dynamics.
The analysis shown in b provides a means to further verify the JG character of the
β relaxation. In fact, the β relaxations of DIA, NOR,
and TETRA are observed to be virtually superposed in the Angell plot,
where the three compounds all display a kink at T g / T ≈ 1, and the β activation
energy below T g is virtually the same
(within the error) for all three compounds (see ). The fact that the (secondary) β
relaxation time scales with T g (which
as discussed in is actually related to the kinetic arrest of the α
relaxation) is typical of JG relaxations. The study of this type of relaxation is particularly relevant
for
amorphous drugs because several studies have brought forth the idea
that the kinetic stability of a molecular glass is correlated with
the secondary β relaxation. In particular, it has been argued
experimentally that a small-molecule glass is kinetically stable only
below the onset temperature of the JG relaxation, typically few tens
of degrees below T g . In the case of the diazepines, the relaxation time of the
β JG relaxation reaches the standard value of 100 s between
30 and 40 K below the T g of the compound.
In our experiments, NOR and TETRA displayed a tendency to recrystallize
above T g , while DIA did not. It should
be noted that the onset of the β relaxation is likely a minimal
requirement for recrystallization: in our experiments, supercooled
DIA was not observed to recrystallize during a period of few days
even above the onset of the α relaxation, i.e., above T g . − The main theoretical model
concerning the JG relaxation is the
Coupling Model (hereafter, CM). , The CM interprets the
JG relaxations as a local, non-cooperative whole-molecule process,
which acts as the “precursor” at shorter times of the
α relaxation. , The characteristic CM relaxation
times in the supercooled liquid state are given by the following approximated
equation, which should approximately equal the experimental JG relaxation
times: 6 Here, t c is the correlation time (usually
of the order of 2 ps) and n , called the coupling
parameter, is related to the Havriliak–Negami exponents of
the α relaxation by the approximate relation 1 – n = ( ab ) 1/1.23 . In the case of the studied diazepines, the Havriliak–Negami
function reduces to a Cole–Davidson equation with a single
exponent b , which is found to be independent of temperature,
so that the coupling parameter is constant and equal to n = 1 – ( b ) 1/1.23 . then predicts that the β
relaxation time is perfectly correlated with the structural relaxation
time and thus scales with T g , as indeed
observed. Despite this, the relaxation times calculated with the CM
theory do not coincide with the experimental JG ones. This might be
due to the fact that the β relaxation is observed only as a
shoulder of the α peak, in which case it has been shown that
the fitting procedure that we employed does not reproduce the precursor
frequency predicted by the CM. It is nevertheless worth pointing out
that the difference at T g between the
theoretical times and the experimental ones can be off by as many
as two orders of magnitude (see b). We finally discuss the fastest secondary
relaxations observed in
our samples. These relaxations must stem from intramolecular degrees
of freedom. In the case of the benzodiazepine ring, the only degree
of freedom corresponds to the chirality inversion between P and M
conformers discussed in the previous section. Apart from this, all
three molecules possess a torsional degree of freedom corresponding
to the single covalent bond linking the fused benzodiazepine ring
with the six-membered carbon ring. There are two more degrees of freedom
in some of the derivatives, namely, the internal rotation of the methyl
group in DIA and TETRA, and a possible conformational interconversion
dynamics of the non-planar cyclohexene ring of TETRA. Neither of these
processes is expected to give rise to a dielectric relaxation feature,
due to the lack of dipole moment of either moiety, so that there are
only two possible candidates for the experimentally observed γ
and γ′ relaxations. As visible in the Angell plot
of b, neither
the γ nor the γ′
relaxation scales with the α relaxation or with the glass transition
temperature, which indicates that they correspond to local relaxation
processes of very low cooperativity. Looking at the relaxation maps
of a, it can
be seen that the three γ relaxations have very similar relaxation
times at a given fixed temperature in all three compounds and also
that the corresponding activation energies E a,γ are close for all studied diazepines . Instead, the α and
β relaxations have very different relaxation times between NOR
on one hand and DIA and TETRA on the other, as stated previously,
and the γ′ relaxation is quite separated in DIA and TETRA.
The similarity of the γ relaxation times and activation energy,
and the fact that this relaxation is unaffected by the distance from
the glass transition temperature suggest that the γ relaxation
is an intramolecular relaxation process common to all three diazepines. As mentioned in , all three studied benzodiazepines exist
in two possible
equivalent conformations of opposite chirality. Both conformers, P
and M, are present in the crystal phase of each compound. In the gas
phase and in solution, benzodiazepines are known to be relatively
flexible and to display inter-conversion dynamics between the two
equivalent conformations, accompanied by a reorientation by 60°
of the CH 2 moiety attached to the carbonyl group, as discussed,
e.g., by Mielcarek et al . The conformational dynamics of DIA and NOR was reported in previous
studies for molecules in solution, and it was found that the activation
energy was not significantly dependent on the solvent. The conformational
activation energies were found experimentally to be 74 and 52 kJ/mol
for DIA and NOR, respectively. , Because the conformational
transition is accompanied also by a
change in position of the polar carbonyl group and of the nitrogen
atoms and thus of the direction of the
molecular dipole moment, such conformational change should be observable
in dielectric spectroscopy. The fact that the γ relaxation is
observed in all three compounds at very similar relaxation times leads
us to assign this process to the inter-conversion dynamics between
P and M conformations (see inset to ). It can instead be ruled out that the γ′
relaxation can correspond to such dynamics, considering that the DIA
and NOR derivatives, which have identical fused benzodiazepine rings,
have γ′ relaxation times differing by more than two orders
of magnitude. It may seem surprising that the M–P interconversion
takes
place also in the liquid phase of NOR due to the presence of hydrogen
bonds. It must however be considered that the H-bond network in a
liquid phase is dynamic and in general only involves a fraction of
the molecules at a given time. The dielectric signal of the P–M
interconversion dynamics of NOR, namely, the γ relaxation of
this compound, likely stems from the fraction of molecules that are
not involved in H-bonding at a given time. It is worth pointing out,
in this respect, that the relaxation time and activation energies
are similar but not identical in the three compounds. We also remark
that the experimental values of the corresponding activation energy
in solution are roughly twice those of the γ relaxations reported
in . It should
however be kept in mind that the extent of H bonding will differ depending
on the liquid phase, and, more importantly, our measurements of the
γ dynamics are all in the glass state of the pure compound.
It is well-known that the temperature dependence of the structural
and JG relaxations displays an abrupt change at T g due to the loss of ergodic equilibrium when going from
the supercooled liquid to the glass phase. This is clearly visible
for the case of the β JG relaxation of benzodiazepine
in b, as discussed
earlier. The same effect is expected to be visible for any relaxation
process whose characteristic time is affected by the viscosity, and
it could be that the interconversion rate between P and M conformers
(γ relaxation time) is partially affected by changes of macroscopic
properties of the sample such as its viscosity (although it cannot
depend only on it, as b shows). Dielectric relaxation studies of flexible heterocyclic
molecules are relatively uncommon, and, to the best of our knowledge,
ours is one of the few dielectric spectroscopy studies that have provided
a clear identification of the ring conformational dynamics in polycyclic
molecules. , , Finally, concerning the γ′ relaxation, both the
range
of temperature in which it is observed and its characteristic relaxation
time are very different between DIA and TETRA, as mentioned, albeit
that its activation energy is of the same order of magnitude in both
compounds. Given that this relaxation is virtually absent in NOR,
it is likely that it is suppressed or at least strongly hindered by
the presence of intermolecular hydrogen bonds. All three studied benzodiazepines
have, as mentioned, a further degree of freedom, corresponding to
the torsional rotation around the covalent bond linking the fused
double ring with the six-membered carbon ring. , While the latter has basically no dipole moment, a rotation of the
double ring about this covalent bond could lead to a rigid rotation
of the molecular dipole moment, which would contribute a dielectric
loss signal. Therefore, we tentatively assign the γ′
relaxation to the rigid rotation, likely by a small angle, of the
double ring about its bond with the six-membered carbon ring. Such
rotation might be partially hindered, in the case of NOR, by the presence
of a network of intermolecular hydrogen bonds, which rationalizes
the extremely weak signal of the γ′ relaxation in this
compound. The difference between the γ′ activation energy
and relaxation times of DIA and TETRA might then be attributed to
the different steric hindrance of the two distinct six-member rings,
namely, a bulkier phenyl ring in the case of DIA and a non-planar
cyclohexene ring in the case of TETRA. This tentative interpretation
is consistent with the much faster γ′ relaxation dynamics
in TETRA.
Crystallization Kinetics Dielectric
spectroscopy was employed to determine the kinetics of isothermal
recrystallization from the supercooled liquid state of NOR and TETRA
(as mentioned, DIA was not observed to recrystallize in short times).
To this purpose, we acquired series of dielectric spectra at fixed
temperature and analyzed the variation in time of the static dielectric
constant, which is related to the dielectric intensity of the structural
relaxation process. Since NOR has significantly higher glass transition
temperature than TETRA, at temperatures at which the latter compound
showed recrystallization at detectable rates, NOR is close to being
in the glass state, where the recrystallization onset time and recrystallization
rates are too long to allow a dielectric measurement. Therefore, because
such “isothermal comparison” of the recrystallization
process cannot be carried out, we have chosen different temperatures
to study recrystallization at roughly the same reduced temperature T / T g . displays the series of isothermal
permittivity spectra (real and imaginary part) during recrystallization
of TETRA at T = 331 K (corresponding to T / T g,TETRA = 1.07) and of NOR at T = 375 K (corresponding to T / T g,NOR = 1.08). The effect of recrystallization is visible
as a decrease over time of the dielectric intensity of the α
loss feature, or equivalently a decrease of the static permittivity
value ε s , defined as the value of ε′( f ) at the lowest frequency displayed in the figure ( f = 1 Hz for TETRA and f = 2 Hz for NOR,
respectively). The onset time t o of the
recrystallization process was determined as the time at which the
initially constant value of ε s in the supercooled
liquid phase was observed to start decreasing. The evolution of ε s with time elapsed from the start of the recrystallization
is displayed in e. It is clear that the recrystallization of NOR at T / T g,NOR = 1.08 is slower than that of
TETRA at T / T g,TETRA =
1.07, despite the fact that the structural (α) relaxation frequency
and thus the cooperative mobility are, under such conditions, higher
by a factor of four in NOR than in TETRA, as testified by the position
of the loss maxima in panels (c) and (d) of . In order to study the kinetics of recrystallization, we define
as customary , a normalized static permittivity
value as: 7 Here, ε s (SL)
and ε s (C) are the
static permittivity of the supercooled liquid and the crystal phase,
as measured before the onset of nucleation of the crystal phase and
at the end of the crystal growth, respectively, while ε s ( t ) is the static permittivity of the partially
recrystallized sample as function of time. The global kinetics of
crystallization can be modeled with the help of the Avrami equation, , which is based on the nucleation-and-growth model of the transition
from the liquid to the crystal phase. According to this model, the
renormalized static permittivity should vary in time as: , 8 Here, n is
the Avrami exponent and Z is a constant from which
the recrystallization rate in s –1 can be obtained , as k = Z 1/ n . According to , the quantity ln(−ln(1 – ε n )) should
be linearly proportional to the logarithm of the time elapsed since
the onset of recrystallization, t – t o . This is indeed observed in the Avrami plot
displayed in f. The values of the obtained fit parameters are n = 1.01 ± 0.05, k = (7 ± 3)·10 –5 s –1 for TETRA and n = 1.1 ± 0.1, k = (4 ± 2)·10 –5 s –1 for NOR. The fact that the
value of the Avrami exponent is close to unity for both derivatives
indicates a strongly anisotropic (one-dimensional) growth of the crystal
phase after a sporadic nucleation event. , , A value of n = 1 also allows
direct estimation of the crystal growth rate, that is, separation
of the nucleation and crystal growth phases of the recrystallization. The vertical separation in f, in which assuming an identical
value of n can be related to the difference in recrystallization
rate k between the two samples (see the discussion
of of ref , confirms the slower crystal
growth kinetics directly visible in e, and is consistent with the experimental
ranges of values of the recrystallization rate k of
TETRA and NOR under these conditions. We also studied the recrystallization
of NOR at T = 368 K ( T/T g = 1.06). The latter temperature
was chosen so that the structural relaxation frequency was the same
for both compounds (a condition usually referred to as “isochronal
condition” in the scientific literature). Because the two compounds
have similar fragility indexes, this condition is very similar to
that of same reduced temperature, T / T g . The crystal growth rate of NOR was so slow under these
conditions (at a temperature only 5 K below the crystallization temperature
of ) that we
could not complete it during three full days of continuous measurements.
The crystallization (growth) rate k for NOR at 368
K ( k = (7 ± 3)·10 –6 s –1 ) was one order of magnitude smaller than that for
TETRA at 331 K, and our experiments show that the (homogeneous) nucleation
time is very different in DIA with respect to its derivatives.
Discussion These results on three very similar
molecules have important implications.
Several recent studies on different glass former compounds have reported
that the crystallization time (or equivalently the inverse crystallization
rate) and the structural relaxation time are correlated with one another. , , These studies have shown that
there is a power-law correlation between the recrystallization time
and τ α . Our study of very similar molecular
derivatives shows, in a very direct way, that there cannot be a general
quantitative relation between the absolute numerical values of these two quantities in different samples. This is not surprising
in view of the fact that different compounds have, in general, different
power law exponents; , our study further shows that
even related molecular derivatives have different correlation laws.
Hence, the correlation between τ α and the crystallization
growth rate is not only limited to a temperature interval, as implied
by the standard model of crystallization by nucleation and growth
and as shown experimentally in a recent study of ours but also it cannot be used as an a priori predictor of crystallization tendency or rate. Indeed, our study
confirms that supercooled liquids of very similar glass-former molecules
have, at the same value of τ α , not only very
different nucleation times but also quite distinct crystal growth
rates, depending, in the present case, on the extent of hydrogen bonding.
These results are in agreement with the standard model of crystallization
by nucleation and growth: in fact, the nucleation step is mainly determined
by the difference between bulk free energy and by the interfacial
tension of the liquid and crystalline phases, rather than the molecular
mobility; and similarly, the growth kinetics of crystalline nuclei
is not uniquely determined by the molecular mobility alone. Our findings
imply that, to further improve our experimental understanding of the
kinetic stability of amorphous pharmaceutics, correlations with other
(possibly macroscopic) quantities, related to the local structure
in the liquid and crystal states, should be investigated, beyond that
with the structural mobility or viscosity. To summarize, we
have studied three diazepine derivatives of very
similar mass and molecular structure (Diazepam, Nordazepam and Tetrazepam),
to determine how the differences in the molecular structure and thus
intermolecular interactions affect the properties of the crystalline
and amorphous states of these pharmaceutical compounds. Nordazepam
is the only compound that displays N–H···O hydrogen
bonds, leading to the formation of H-bonded dimers in the crystalline
phase, which as a consequence exhibits significantly higher melting
point and melting enthalpy compared to the other two compounds, which
display similar melting temperatures and enthalpies. Nordazepam has
the highest density in the crystalline state and the smallest Hirshfeld
surface and volume of the three. The diazepine ring has a non-planar
structure, and all three benzodiazepine crystalline structures consist
of two isoenergetic P and M conformers, which are mirror images of
one another and occur in a 1:1 ratio. The characteristic angles of
these conformations are similar in the three compounds. The
liquid phase of Nordazepam displays significantly higher glass
transition temperature than the other two compounds, and the dielectric
signature of the structural α relaxation is broader in this
compound than in the other two, indicative of a more cooperative structural
relaxation dynamics. These two experimental observations indicate
at least partial hydrogen bonding also in the liquid phase of Nordazepam.
The presence of different possible molecular conformations, as well
as the torsional degree of freedom between the fused double ring and
the six-membered carbon ring, further enrich the relaxation map in
the amorphous (supercooled liquid and glass) state. All three compounds
display a Johari–Goldstein β relaxation, visible as a
shoulder to the main α loss feature. The relaxation time of
both α and β relaxations scales with the temperature normalized
to the glass transition temperature ( T / T g ). The curvature of the structural relaxation is the
same in all three compounds leading to a virtually identical kinetic
fragility index ( m p ≈ 32). The three compounds display intramolecular relaxations in the glass
state, one of which is common to all of them, and corresponds to the
P-M inter-conformer conversion dynamics of the diazepine heterocycle.
This relaxation does not scale with the cooperative molecular mobility
(α relaxation time), although comparison with liquid-phase studies
indicates that its activation energy is slightly lower in the glass
state compared to the liquid. A fourth, high-frequency secondary relaxation
is present only in Diazepam and Tetrazepam, likely associated with
the rigid rotation of the fused double ring relative to the apolar
six-membered ring. Its almost complete absence in Nordazepam can be
rationalized by the existence of strong hydrogen bonds between the
double rings of neighboring molecules, which prevents such rotation. While supercooled liquid Tetrazepam and Nordazepam are observed
to recrystallize upon heating, with Avrami exponents close to unity
in both cases, Diazepam does not display any tendency toward recrystallization
at least over short periods of time. The crystallization rates of
Tetrazepam and Nordazepam differ, under isochronal conditions of the
structural α relaxation, by more than a decade. We conclude
that the kinetic stability of amorphous diazepines, and especially
the nucleation tendency, does not display any correlation with the
density, kinetic fragility index, or structural or secondary Johari–Goldstein
relaxation time. Only the crystal growth rate, and not the tendency
toward nucleation, is affected by the presence of a hydrogen-bond
network. Our comparison between very similar molecular derivatives
provides a direct confirmation that the search for microscopic criteria
for the kinetic stability of amorphous pharmaceuticals must include,
besides molecular interactions and relaxation dynamics, other parameters
related to the difference in the (local) structure between the liquid
and crystal phases.
|
null | 70f460b4-3e29-44a1-bc3d-2796af13b321 | 8293955 | Pharmacology[mh] | Traditional medicinal plants have been an essential source of remedy for various illnesses since ancient times. Preparations of plant materials such as infusion, decoction, powder, or paste have been used in various traditional practices in different parts of the world. People living in Africa and Asia make use of herbal medications to supplement the conventional medicine practice (Ekor ). There has been an increasing interest in the usage of herbal medicines in recent years. About 80% of the world’s population is using phytotherapeutic medicines (Balekundri and Mannur ). The WHO estimated that the size of the global market for herbal products was USD 60 billion in the year 2000 and this is expected to grow 7% per annum towards USD 5 trillion by the year 2050 (Tan et al. ). Several analyses have clearly verified traditional claims of numerous medicinal plants leading to the commercialisation of the many herbal products and their nomination as leads within the development of pharmaceutical medication (Williams ). Many clinically useful drugs have been discovered based on the knowledge derived from the ethnomedicinal applications of various herbal materials (Balunas and Kinghorn ). Pseudocedrela kotschyi (Schweinf.) Harms (Meliaceae) is an important medicinal plant found in the tropical and subtropical countries of Africa. This plant has been extensively used in the African traditional medicine system for the treatment of a variety of diseases, particularly as analgesic, antimicrobial, antimalarial, anthelminthic, and antidiarrheal agents. The main focus of this review was to establish the ethnopharmacological uses and medicinal characteristics of P. kotschyi and highlight its potential as future drug for the treatment of various tropical diseases.
Scientific manuscripts on P. kotschyi were retrieved from different scientific search engines. Literature search was carried out on PUBMED using Pseudocedrela kotschyi as key words. Additional literature searches on Medline, EMBASE, Science Direct and Google scholar databases were done using pharmacological activity, chemical constituents and traditional uses of Pseudocedrela kotschyi as search terms. Literature published on the topic in English language from inception until September, 2020 were collected, analysed and an up-to-date review on the medicinal potential of P. kotschyi was compiled. Geographical distribution P. kotschyi (common name: dry zone cedar) is a medicinal plant. Other common names of P. kotschyi are Tuna (Hausa) and Emi gbegi in Yoruba. It is found in tropical and subtropical countries of Africa which include Nigeria, Cote d’Ivoire, Senegal, Ghana, Democratic Republic of Congo, and Uganda. The plant often grows as a medium sized tree of about 12–20 ft high (Ayo et al. ; Alain et al. ; Alhassan et al. ). Below is the taxonomical classification of P. kotschyi (Hassler ). Classification Kingdom Plantae Phylum Tracheophyta Class Magnoliopsida Order Sapindales Family Meliaceae Genus Pseudocedrela Species kotschyi Ethnomedicinal uses Different parts of P. kotschyi are used in the traditional treatment of various diseases. The root is used in the treatment of leprosy (Pedersen et al. ), epilepsy, dementia (Kantati et al. ), diabetes (Salihu Shinkafi et al. ), malaria, abdominal pain, diarrhoea (Ahua et al. ), toothache and gingivitis (Tapsoba and Deschamps ). The root is also used as a chewing stick for tooth cleaning and enhancement of oral health (Wolinsky and Sote ; Olabanji et al. ; Adeniyi et al. ). The leaf is used in the treatment of female infertility (Olabanji et al. ), intestinal worms (Koné et al. ) and malaria (Asase et al. ). The stem bark is used in the treatment of cancer (Saidu et al. ), infantile dermatitis (Erinoso et al. ), stomach ache (Asase et al. ), toothache (Kayode and Sanni ), high blood-pressure, skin diseases, and haemorrhoids (Nadembega et al. ). Phytochemistry Phytochemical investigations revealed that P. kotschyi contains a variety of pharmacological active secondary metabolites. A total of 32 compounds have so far reported to have been isolated from the plant which mainly include limonoids, triterpenes, and flavonoids. Limonoids are modified triterpenes which are highly oxygenated and have a typical furanylsteroid as their core structure (Roy and Saraf ). They are also known as tetraterpenoids. Limonoids are rare natural products which occur mainly in plants of Meliaceae and Rutaceae families and less frequently in the Cneoraceae family (Tan and Luo ). Several phragmalin-type limonoid orthoacetates have reportedly been isolated from the of roots of this plant, namely, kotschyins A – H ( 1–8 ) (Hay et al. ; Dal Piaz et al. ). These compounds are complex with a very high degree of oxidation and rearrangement as compared to the parent limonoid structure. Other limonoid derivatives found in the roots and stem bark of P. kotschyi are 7-deacetylgedunin (9) , 7-deacetyl-7-oxogedunin (10) (Hay et al. ), 1α,7α epoxy-gedunin ( 11 ), gedunin ( 12 ) (Dal Piaz et al. ), kostchyienones A ( 13 ) and B ( 14 ), andirobin ( 15 ) methylangolensate ( 16) (Sidjui et al. ). Additional limonoids derivatives that were isolated from the P. kotschyi bark include pseudrelones A – C ( 17–19 ) (Taylor ). The pseudrelones also have a phragmalin nucleus with orthoacetate function but they have a lesser degree of oxidation than the kotschyins. The steroids isolated from this plant include odoratone ( 20 ), spicatin ( 21 ), 11-acetil-odoratol ( 22 ) (Dal Piaz et al. ), β-sitosterol ( 23 ), 3- O -β- d -glucopyranosyl β-sitosterol ( 24 ) stigmasterol ( 25 ), 3- O -β- d -glucopyranosyl stigmasterol ( 26 ) betulinic acid ( 27 ) (Sidjui et al. ). Three secotirucallane triterpenes were also isolated from the stem bark of P. kotschyi . These include, 4-hydroxy-3,4-secotirucalla-7,24-dien-3,21-dioic acid ( 28 ), 3,4-secotirucalla-4(29),7,24-trien-3,21-dioic acid ( 29 ) and 3-methyl ester 3,4-secotirucalla-4(28),7,24-trien-3,21-dioic ( 30 ) (Mambou et al. ). Two flavonoids, namely, 3,6,8-trihydroxy-2-(3,4-dihydroxylphenyl)-4H-chrom-4-one ( 31 ) and quercetin, 3,4′,7-trimethyl ether ( 32 ) have also been isolated from the roots of this plant (Sidjui et al. ). The GCMS analysis of essential oils from root and stem of P. kotschyi indicated that both oils contain mainly sesquiterpenoids. These include, α-cubebene, α-copaene, β-elemene, β-caryophyllene, trans -α-bergamotene, aromadendrene, ( E )-β-farnesene, α-humulene, allo-aromadendrene, γ-muurolene, farnesene, germacrene D, β-selinene, α-selinene, α-muurolene, γ-cadinene, calamenene, δ-cadinene, cadina-1,4-diene, α-calacorene, α-cadinene, β-calacorene, germacrene B, cadalene, epi -cubebol, cubebol, spathulenol, globulol, humulene oxide II, epi -α-cadinol, epi -α-muurolol, α-muurolol, selin-11-en-4-α-ol, α-cadinol and juniper camphor. The stem bark oil was found to comprise largely of sesquiterpene hydrocarbons (79.6%), with δ-cadinene (31.3%) as the major constituents. While the oxygenated sesquiterpenes were found to be abundant in the root with cubebols (32.5%) and cadinols (17.9%) as the major constituents (Boyom et al. ). Pharmacological activity The ethnomedicinal claims for the efficacy of P. kotschyi in the treatment of various diseases have been confirmed by numerous relevant scientific studies. Several pharmacological investigations have been carried out to confirm the traditional medicinal uses of the roots, stem bark, and leaves of P. kotschyi . A wide range of pharmacological activities such as analgesic, antipyretic, anthelminthic, antimalaria, anti-leishmaniasis, hepatoprotective, antioxidant, and antimicrobial, have been reported by researchers so far. Anti-inflammatory, analgesic and antipyretic activities Inflammation is an adaptive response that is triggered by noxious stimuli or conditions, such as tissue injury and infection (Medzhitov ; Ahmed ). Inflammatory response involves the secretion of several chemical mediators and signalling molecules such as nitric oxide (NO), and proinflammatory cytokines, including tumour necrosis, factor-α (TNF-α), interferon-γ (IFNγ), lipopolysaccharides (LPS) and interleukins (Medzhitov ). Even though inflammatory response is meant to be a beneficial process of restoring homeostasis, it is often associated with some disorders like pain and pyrexia due to the secretion of the chemical mediators (Bielefeldt et al. ; Garami et al. ). Chronic secretion of proinflammatory cytokines is also associated with development of diseases such as cancer and diabetes. Hence, anti-inflammatory agents represent an important class of medicines. Extracts and phytoconstituents of P. kotschyi have been reported to possess anti-inflammatory, analgesic and antipyretic properties. The administration of methanol crude extract of P. kotschyi stem bark at a dose of 200 mg/kg/day and its butanol and chloroform fractions have been shown to produce significant analgesic activity when evaluated with a mice model. The extracts and fractions decreased the number of writhes by 88–92% during the acetic acid induced writhing assay (Abubakar et al. ). Akuodor et al. investigated the antipyretic activity of ethanol extract of P. kotschyi leaves on yeast and amphetamine induced hyperpyrexia in rats. They reported that the leaf extract (50, 100 and 150 mg/kg i.p.) displayed a significant ( p < 0.05) dose-dependent decrease in pyrexia. Scientific investigations have shown that 7-deacetylgedunin ( 9) had significant anti-inflammatory activity. Compound 9 was reported to significantly inhibit lipopolysaccharide induced nitric oxide in murine macrophage RAW 264.7 cells with an IC 50 of 4.9 ± 0.1 μM. It also produced the downregulation of mRNA and protein expression of inducible nitric oxide synthase (iNOS) at a dose of 10 µM (Sarigaputi et al. ). These findings suggest that compound 9 produces its anti-inflammatory effect through the modulation of NO production. Chen et al. investigated the anti-inflammatory activity of compound 9 in C57BL/6 mice. Their result showed that its intraperitoneal administration at a dose of 5 mg/kg body weight for two consecutive days significantly decreased LPS-induced mice mortality by 40%. The above findings demonstrate that compound 9 is a promising anti-inflammatory agent from this plant. The anti-inflammatory effect of this compound perhaps accounts for the analgesic and antipyretic properties of the P. kotschyi extracts. Antiparasitic activity Parasitic diseases are among the foremost health problems today, especially in tropical countries of Africa and Asia. Diseases such as malaria, leishmaniasis, trypanosomiasis and helminthiasis affect millions of people each year causing high morbidity and mortality, particularly, in developing countries (Hotez and Kamath ). Hence, there is an urgent need for new drugs to treat and control of these diseases. Extracts obtained from different parts of P. kotschyi have been reported to possess activity against several human parasites. Ahua et al. investigated the anti-leishmaniasis activity of P. kotschyi including several other plants against Leishmania major . The dichloromethane extract of P. kotschyi roots (at a dose of 75 µg/mL) exhibited a marked activity (>90% mortality) against the intracellular form of the parasite which is pathogenically significant for humans. In another study, the anthelminthic activity of an ethanol extract of P. kotschyi roots against Haemonchus contortus (a pathogenic nematode found in small ruminants) was evaluated. The researchers discovered that the ethanol extract possessed larvicidal activity against the helminth with a LC 100 of 0.02 µg/mL (Koné et al. ). The aqueous stem bark extract of this plant (50 mg/mL) has also been demonstrated to exert anthelminthic activity against Lumbricus terrestris with 25.4 min death time (Ukwubile et al. ). The antimalarial effect of P. kotschyi on malaria parasite has been reported in several research manuscripts. Christian et al. investigated the suppressive and curative effect of ethanol extract of P. kotschyi leaves against malaria in Plasmodium berghei berghei infected mice. The results obtained showed that oral administration of the extract (100–400 mg/kg/day) exhibited a significant antimalarial effect which is evident by the suppression of parasitemia and prolong life of infected animals. In an another study, methanol extract of the P. kotschyi leaves at an oral dose of 200 mg/kg/day was found to reduce parasitemia by 90.70% in P. berghei berghei infected mice after four consecutive days of treatment (Dawet and Stephen ). However, the ethanol and aqueous extracts of the P. kotschyi stem bark exhibited lower activity against the malaria parasite (39.43% and 28.36% reduction in parasitemia, respectively) (Dawet and Yakubu ). The limonoid derivatives 9 and 10 were reported to display significant in vitro activity against chloroquine-resistant Plasmodium falciparum with IC 50 values of 1.36 and 1.77 µg/mL, respectively (Hay et al. ). The two compounds also displayed significant antiparasitic activity against Leishmania donovani , Trypanosoma brucei rhodesiense with a low-range IC 50 of 0.99–3.4 µg/mL. In contrast, the orthoacetatate kotschyin A was found to be inactive against all the tested parasites (Hay et al. ). In related work, Sidjui et al. evaluated the in vitro antiplasmodial activity of 14 compounds isolated from P. kotschyi . Their findings showed that the limonoid derivatives 9 , 10 , 13 , 14 and 15 exhibited very significant activity against both chloroquine-sensitive ( Pf3D7 ) and chloroquine-resistant ( PfINDO ) strains of genus Plasmodium with IC 50 values ranging from 0.75 to 9.05 µg/mL. Steverding et al. investigated the trypanocidal and leishmanicidal activities of six limonoids, namely, 9 , 10 , 13 , 14, 15 and 16 , against bloodstream forms of Trypanosoma bruce i and promastigotes of Leishmania major . All the six compounds showed anti-trypanosomal activity with IC 50 values ranging from 3.18 to 14.5 µM. Compounds 9 , 10 , 13 and 14 also displayed leishmanicidal activity with IC 50 of 11.60, 7.63, 2.86 and 14.90 µM, respectively, while 15 and 16 were inactive. The antiplasmodial, trypanocidal, and leishmanicidal activities of these compounds provide justification for the use of crude extract of P. kotschyi in the traditional treatment of malaria and other parasitic infectious diseases. Antimicrobial activity Antimicrobial agents are among the most commonly used medications. The prevalence of antimicrobial resistance in recent years has led to a renewed effort to discover newer antimicrobial agents for the treatment of infectious diseases (Hobson et al. ). Extracts of P. kotschyi were reported to display appreciable activity against some pathogenic microorganisms. Ayo et al. investigated the antimicrobial activity of petroleum ether, ethyl acetate and methanol extracts of the P. kotschyi leaves against Staphylococcus aureus , Salmonella typhi , Streptococcus pyogenes , Candida albicans , and Escherichia coli . The results of the study showed that the ethyl acetate extract exhibited antibacterial activity against all the tested organisms with MIC values of 10–20 mg/mL. In an another similar study, the crude methanol extract of the stem bark of this plant was also shown to exhibit good activity against a panel of pathogenic bacteria and fungi which include methicillin-resistant S. aureus (MRSA), S. aureus , S. pyogenes , Corynebacterium ulcerans , Bacillus subtilis , E. coli, S. typhi , Shigella dysenteriae , Klebsiella pneumoniae , Neisseria gonorrhoeae , Pseudomonas aeruginosa , C. albicans , C. krusei , and C. tropicalis with MIC values of 3.75–10.0 mg/mL (Alhassan et al. ). The methanol extract of the woody stem was also found to possess antifungal activity against C. krusei ATCC 6825 with an MIC of 6.25 mg/mL (Adeniyi et al. ). The secotirucallane triterpenes (compounds 28 , 29 and 30 ) isolated from the bark of P. kotschyi have been reported to possess significant antibacterial activity against Staphylococcus aureus ATCC 25923), Escherichia coli S2(1) and Pseudomonas aeruginosa with MIC ranging from 6 to 64 µg/mL. Compound 29 exhibited the highest antibacterial activity while 30 had the lowest (Mambou et al. ). The presence of these compounds is likely responsible for the antimicrobial property of P. kotschyi extracts and justify the ethnomedicinal use of this plant as a chewing stick for tooth cleaning and enhancement of oral health. Antioxidant and hepatoprotective activities The ethanol extract of P. kotschyi stem bark has been reported to possess DPPH radical scavenging activity with an IC 50 of 4 µg/mL (Alain et al. ). A study on the hepatoprotective activity of methanol and aqueous extracts of the P. kotschyi leaves revealed that both extracts (at a dose of 750 mg/kg/day) were able to protect the liver against paracetamol induced oxidative damage (Eleha et al. ). A similar study conducted by Nchouwet et al. showed that 2 weeks pre-treatment with aqueous and methanol extracts of P. kotschyi stem bark (150 mg/kg/day) significantly suppressed the development of paracetamol induced hepatotoxicity in experimental rats. Hypoglycaemic and digestive enzyme inhibitory activities Diabetes mellitus is disorder associated with abnormal glucose metabolism resulting from insulin insufficiency or dysfunction. It is one of the major non-communicable diseases that affect millions of people globally. Scientific investigation has revealed that P. kotschyi extracts possess some antidiabetic properties. Georgewill and Georgewill investigated the hypoglycaemic effect of aqueous extract of P. kotschyi leaves on alloxan induced diabetic rats. Results of their investigation revealed that oral administration of the extract (200 mg/kg/day for 14 days) caused significant hypoglycaemic effect in the experimental animals. The ethanol extract of roots of this plant was also reported to exhibit inhibitory activity against α-glucosidase (IC 50 = 5.0 ± 0.2 μg/mL), an important digestive enzyme targeted in diabetes treatment (Bothon et al. ). Antiproliferative activity Cancer is an important disease that is characterised by the abnormal rapid proliferation of cells that invade and destroy other tissues (Alhassan et al. ). It is a major public health problem throughout the world. Pharmacological studies have shown that P. kotschyi possesses anticancer potential. Kassim et al. investigated the antiproliferative activity and apoptosis induction effect of aqueous extract of P. kotschyi roots against a panel of prostate cancer cell lines, namely, PC3, DU-145, LNCaP and CWR-22 cell lines. Results from the 3-[4,5-dimethylthiazol-2yl]-2,5-diphenyltetrazolium bromide (MTT) assay showed that all four cancer cell lines exhibited a dose-dependent decrease in cell proliferation and viability after treatment with the aqueous extract with IC 50 values ranging from 12 to 42 µg/mL. The results obtained also showed that LNCaP, PC3, DU-145, and CWR-22 cell lines had 42, 35, 33 and 24% induced apoptotic cells, respectively, after treatment with the same extract. The results of both the antiproliferative and apoptosis assay indicated that the LNCaP cells were the most sensitive to the P. kotschyi extract. Heat shock protein 90 (Hsp90) is a molecular chaperone that is involved in the folding, activation and assembly of several proteins including oncoproteins such as HER2, Survivin, EGFR, Akt, Raf-1, mutant p53 (Calderwood et al. ; Dal Piaz et al. ). Hsp90 is often overexpressed in cancer cells. It has been demonstrated to play a vital role in tumour progression, malignancy and resistance to chemotherapeutic agents (Zhang et al. ). Hence, Hsp90 is recently considered as a viable molecular target for development of new anticancer drugs (Gupta et al. ). Phytoconstituents of P. kotschyi have been shown to possess significant Hsp90 inhibitory activity. Dal Piaz et al. investigated the Hsp90 binding capability of several compounds using a surface plasmon resonance (SPR) approach. They found that the limonoid orthoacetates ( 1–6 ) displayed good binding capability to the protein with compound 4 being the most effective. Compound 4 also exhibited significant anti-proliferative activity against three cancer cell lines, namely, PC-3 (human prostate cancer cells), A2780 (human ovarian carcinoma cells), and MCF-7 (human breast adenocarcinoma cells) with IC 50 values of 62 ± 0.4, 38 ± 0.7 and 25 ± 1.2 µM, respectively. These findings suggest that Hsp90 inhibition is a mechanism of action for anti-proliferative effects of the limonoids orthoacetates from P. kotschyi . These findings provide scientific bases for the future development of new anticancer agents from P. kotschyi in the form of a standardised herbal preparation or as a pure chemical entity. Antidiarrheal activity Treatment of diarrhoea comes under one of the common ethnomedicinal uses of P. kotschyi . To further verify this claim, Essiet et al. investigated the antidiarrheal property of ethanol extract of P kotschyi leaves in Wistar albino rats. Diarrhoea was induced in the animals using castor oil. Results of the investigation revealed that oral administration of the extract (100, 200 and 400 mg/kg) produced significant ( p < 0.05) dose-dependent inhibition of induced diarrhoea (67–91%). The results also showed that the administered doses of the extract decreased intestinal transit time by 57–66% while intestinal fluid accumulation was decreased by 68–82%. This finding undoubtedly supports the traditional use of this plant in the treatment of diarrhoea. Toxicity There is a general perception that plant-based medicinal products are natural and thus, very safe for human consumption. However, this notion is wrong because several plants have been shown to produce wide range of adverse reactions some of which are capable of causing serious injuries, life-threatening conditions, and even death (Ekor ). Hence, it is of paramount importance to investigate the toxicity profile of traditional medicinal plants as well as their phytoconstituents in order to establish their safety. The study of toxicity is an essential component for new drug development process (Hornberg et al. ). Some toxicological studies have been carried out on the extracts of P. kotschyi . Nchouwet et al. investigated the acute and sub-chronic toxicity of P. kotschyi stem bark aqueous extract in albino rats. For the acute toxicity study, the LD 50 was found to be greater than 2000 mg/kg body weight. The sub-chronic administration of the aqueous extract at a dose of 400 mg/kg body weight/day for 28 days, caused significant increase in total protein and HDL-Cholesterol with concomitant decrease in LDL-cholesterol while other biochemical and hematological parameters were found to be within the normal range. However, histological examination revealed the presence of inflammation and necrosis in the kidney and liver tissues of animals treated with 400 mg/kg body weight-/day of extract while tissue samples from animals treated at lower doses remained normal. This implies that the extract may have exhibited some toxic effect on the kidney and liver tissues at 400 mg/kg body weight while it is relatively safe at lower doses. Kabiru et al. conducted a sub-chronic toxicity evaluation of a crude methanol extract of leaves in Sprague-Dawley rats at doses of 40, 200 and 1000 mg/kg body weight/day for 4 weeks. They found that the extract did not produce any significant alteration in both hematological and biochemical parameters when compared with standard controls. This implied that the extract was relatively non-toxic at the tested doses. Ezeokpo et al. also carried out a similar study with an ethanol extract of P. kotschyi leaves in Wistar rats. Their results revealed that the extract (400 mg/kg body weight/day) did not produce any significant derangement in hematological and biochemical parameters after 28 days of treatment. The above findings indicated that methanol and ethanol extracts of P. kotschyi leaves are relatively non-toxic at higher doses compared to the aqueous stem bark extract. However, more detailed research is still required to corroborate this finding. Albeit most of the pharmacological activities and chemical constituents reported on this plant have been obtained from the leaf’s extracts, the toxicity evaluation of ethanol, methanol and chloroform extracts of roots and stem bark of P. kotschy is yet to be carried out and reported. Hence, further toxicity studies on different extracts, fractions and chemical constituents of the root and stem bark of P. kotschyi are still required to ascertain the thorough safety of the plant.
P. kotschyi (common name: dry zone cedar) is a medicinal plant. Other common names of P. kotschyi are Tuna (Hausa) and Emi gbegi in Yoruba. It is found in tropical and subtropical countries of Africa which include Nigeria, Cote d’Ivoire, Senegal, Ghana, Democratic Republic of Congo, and Uganda. The plant often grows as a medium sized tree of about 12–20 ft high (Ayo et al. ; Alain et al. ; Alhassan et al. ). Below is the taxonomical classification of P. kotschyi (Hassler ). Classification Kingdom Plantae Phylum Tracheophyta Class Magnoliopsida Order Sapindales Family Meliaceae Genus Pseudocedrela Species kotschyi
Different parts of P. kotschyi are used in the traditional treatment of various diseases. The root is used in the treatment of leprosy (Pedersen et al. ), epilepsy, dementia (Kantati et al. ), diabetes (Salihu Shinkafi et al. ), malaria, abdominal pain, diarrhoea (Ahua et al. ), toothache and gingivitis (Tapsoba and Deschamps ). The root is also used as a chewing stick for tooth cleaning and enhancement of oral health (Wolinsky and Sote ; Olabanji et al. ; Adeniyi et al. ). The leaf is used in the treatment of female infertility (Olabanji et al. ), intestinal worms (Koné et al. ) and malaria (Asase et al. ). The stem bark is used in the treatment of cancer (Saidu et al. ), infantile dermatitis (Erinoso et al. ), stomach ache (Asase et al. ), toothache (Kayode and Sanni ), high blood-pressure, skin diseases, and haemorrhoids (Nadembega et al. ).
Phytochemical investigations revealed that P. kotschyi contains a variety of pharmacological active secondary metabolites. A total of 32 compounds have so far reported to have been isolated from the plant which mainly include limonoids, triterpenes, and flavonoids. Limonoids are modified triterpenes which are highly oxygenated and have a typical furanylsteroid as their core structure (Roy and Saraf ). They are also known as tetraterpenoids. Limonoids are rare natural products which occur mainly in plants of Meliaceae and Rutaceae families and less frequently in the Cneoraceae family (Tan and Luo ). Several phragmalin-type limonoid orthoacetates have reportedly been isolated from the of roots of this plant, namely, kotschyins A – H ( 1–8 ) (Hay et al. ; Dal Piaz et al. ). These compounds are complex with a very high degree of oxidation and rearrangement as compared to the parent limonoid structure. Other limonoid derivatives found in the roots and stem bark of P. kotschyi are 7-deacetylgedunin (9) , 7-deacetyl-7-oxogedunin (10) (Hay et al. ), 1α,7α epoxy-gedunin ( 11 ), gedunin ( 12 ) (Dal Piaz et al. ), kostchyienones A ( 13 ) and B ( 14 ), andirobin ( 15 ) methylangolensate ( 16) (Sidjui et al. ). Additional limonoids derivatives that were isolated from the P. kotschyi bark include pseudrelones A – C ( 17–19 ) (Taylor ). The pseudrelones also have a phragmalin nucleus with orthoacetate function but they have a lesser degree of oxidation than the kotschyins. The steroids isolated from this plant include odoratone ( 20 ), spicatin ( 21 ), 11-acetil-odoratol ( 22 ) (Dal Piaz et al. ), β-sitosterol ( 23 ), 3- O -β- d -glucopyranosyl β-sitosterol ( 24 ) stigmasterol ( 25 ), 3- O -β- d -glucopyranosyl stigmasterol ( 26 ) betulinic acid ( 27 ) (Sidjui et al. ). Three secotirucallane triterpenes were also isolated from the stem bark of P. kotschyi . These include, 4-hydroxy-3,4-secotirucalla-7,24-dien-3,21-dioic acid ( 28 ), 3,4-secotirucalla-4(29),7,24-trien-3,21-dioic acid ( 29 ) and 3-methyl ester 3,4-secotirucalla-4(28),7,24-trien-3,21-dioic ( 30 ) (Mambou et al. ). Two flavonoids, namely, 3,6,8-trihydroxy-2-(3,4-dihydroxylphenyl)-4H-chrom-4-one ( 31 ) and quercetin, 3,4′,7-trimethyl ether ( 32 ) have also been isolated from the roots of this plant (Sidjui et al. ). The GCMS analysis of essential oils from root and stem of P. kotschyi indicated that both oils contain mainly sesquiterpenoids. These include, α-cubebene, α-copaene, β-elemene, β-caryophyllene, trans -α-bergamotene, aromadendrene, ( E )-β-farnesene, α-humulene, allo-aromadendrene, γ-muurolene, farnesene, germacrene D, β-selinene, α-selinene, α-muurolene, γ-cadinene, calamenene, δ-cadinene, cadina-1,4-diene, α-calacorene, α-cadinene, β-calacorene, germacrene B, cadalene, epi -cubebol, cubebol, spathulenol, globulol, humulene oxide II, epi -α-cadinol, epi -α-muurolol, α-muurolol, selin-11-en-4-α-ol, α-cadinol and juniper camphor. The stem bark oil was found to comprise largely of sesquiterpene hydrocarbons (79.6%), with δ-cadinene (31.3%) as the major constituents. While the oxygenated sesquiterpenes were found to be abundant in the root with cubebols (32.5%) and cadinols (17.9%) as the major constituents (Boyom et al. ).
The ethnomedicinal claims for the efficacy of P. kotschyi in the treatment of various diseases have been confirmed by numerous relevant scientific studies. Several pharmacological investigations have been carried out to confirm the traditional medicinal uses of the roots, stem bark, and leaves of P. kotschyi . A wide range of pharmacological activities such as analgesic, antipyretic, anthelminthic, antimalaria, anti-leishmaniasis, hepatoprotective, antioxidant, and antimicrobial, have been reported by researchers so far.
Inflammation is an adaptive response that is triggered by noxious stimuli or conditions, such as tissue injury and infection (Medzhitov ; Ahmed ). Inflammatory response involves the secretion of several chemical mediators and signalling molecules such as nitric oxide (NO), and proinflammatory cytokines, including tumour necrosis, factor-α (TNF-α), interferon-γ (IFNγ), lipopolysaccharides (LPS) and interleukins (Medzhitov ). Even though inflammatory response is meant to be a beneficial process of restoring homeostasis, it is often associated with some disorders like pain and pyrexia due to the secretion of the chemical mediators (Bielefeldt et al. ; Garami et al. ). Chronic secretion of proinflammatory cytokines is also associated with development of diseases such as cancer and diabetes. Hence, anti-inflammatory agents represent an important class of medicines. Extracts and phytoconstituents of P. kotschyi have been reported to possess anti-inflammatory, analgesic and antipyretic properties. The administration of methanol crude extract of P. kotschyi stem bark at a dose of 200 mg/kg/day and its butanol and chloroform fractions have been shown to produce significant analgesic activity when evaluated with a mice model. The extracts and fractions decreased the number of writhes by 88–92% during the acetic acid induced writhing assay (Abubakar et al. ). Akuodor et al. investigated the antipyretic activity of ethanol extract of P. kotschyi leaves on yeast and amphetamine induced hyperpyrexia in rats. They reported that the leaf extract (50, 100 and 150 mg/kg i.p.) displayed a significant ( p < 0.05) dose-dependent decrease in pyrexia. Scientific investigations have shown that 7-deacetylgedunin ( 9) had significant anti-inflammatory activity. Compound 9 was reported to significantly inhibit lipopolysaccharide induced nitric oxide in murine macrophage RAW 264.7 cells with an IC 50 of 4.9 ± 0.1 μM. It also produced the downregulation of mRNA and protein expression of inducible nitric oxide synthase (iNOS) at a dose of 10 µM (Sarigaputi et al. ). These findings suggest that compound 9 produces its anti-inflammatory effect through the modulation of NO production. Chen et al. investigated the anti-inflammatory activity of compound 9 in C57BL/6 mice. Their result showed that its intraperitoneal administration at a dose of 5 mg/kg body weight for two consecutive days significantly decreased LPS-induced mice mortality by 40%. The above findings demonstrate that compound 9 is a promising anti-inflammatory agent from this plant. The anti-inflammatory effect of this compound perhaps accounts for the analgesic and antipyretic properties of the P. kotschyi extracts.
Parasitic diseases are among the foremost health problems today, especially in tropical countries of Africa and Asia. Diseases such as malaria, leishmaniasis, trypanosomiasis and helminthiasis affect millions of people each year causing high morbidity and mortality, particularly, in developing countries (Hotez and Kamath ). Hence, there is an urgent need for new drugs to treat and control of these diseases. Extracts obtained from different parts of P. kotschyi have been reported to possess activity against several human parasites. Ahua et al. investigated the anti-leishmaniasis activity of P. kotschyi including several other plants against Leishmania major . The dichloromethane extract of P. kotschyi roots (at a dose of 75 µg/mL) exhibited a marked activity (>90% mortality) against the intracellular form of the parasite which is pathogenically significant for humans. In another study, the anthelminthic activity of an ethanol extract of P. kotschyi roots against Haemonchus contortus (a pathogenic nematode found in small ruminants) was evaluated. The researchers discovered that the ethanol extract possessed larvicidal activity against the helminth with a LC 100 of 0.02 µg/mL (Koné et al. ). The aqueous stem bark extract of this plant (50 mg/mL) has also been demonstrated to exert anthelminthic activity against Lumbricus terrestris with 25.4 min death time (Ukwubile et al. ). The antimalarial effect of P. kotschyi on malaria parasite has been reported in several research manuscripts. Christian et al. investigated the suppressive and curative effect of ethanol extract of P. kotschyi leaves against malaria in Plasmodium berghei berghei infected mice. The results obtained showed that oral administration of the extract (100–400 mg/kg/day) exhibited a significant antimalarial effect which is evident by the suppression of parasitemia and prolong life of infected animals. In an another study, methanol extract of the P. kotschyi leaves at an oral dose of 200 mg/kg/day was found to reduce parasitemia by 90.70% in P. berghei berghei infected mice after four consecutive days of treatment (Dawet and Stephen ). However, the ethanol and aqueous extracts of the P. kotschyi stem bark exhibited lower activity against the malaria parasite (39.43% and 28.36% reduction in parasitemia, respectively) (Dawet and Yakubu ). The limonoid derivatives 9 and 10 were reported to display significant in vitro activity against chloroquine-resistant Plasmodium falciparum with IC 50 values of 1.36 and 1.77 µg/mL, respectively (Hay et al. ). The two compounds also displayed significant antiparasitic activity against Leishmania donovani , Trypanosoma brucei rhodesiense with a low-range IC 50 of 0.99–3.4 µg/mL. In contrast, the orthoacetatate kotschyin A was found to be inactive against all the tested parasites (Hay et al. ). In related work, Sidjui et al. evaluated the in vitro antiplasmodial activity of 14 compounds isolated from P. kotschyi . Their findings showed that the limonoid derivatives 9 , 10 , 13 , 14 and 15 exhibited very significant activity against both chloroquine-sensitive ( Pf3D7 ) and chloroquine-resistant ( PfINDO ) strains of genus Plasmodium with IC 50 values ranging from 0.75 to 9.05 µg/mL. Steverding et al. investigated the trypanocidal and leishmanicidal activities of six limonoids, namely, 9 , 10 , 13 , 14, 15 and 16 , against bloodstream forms of Trypanosoma bruce i and promastigotes of Leishmania major . All the six compounds showed anti-trypanosomal activity with IC 50 values ranging from 3.18 to 14.5 µM. Compounds 9 , 10 , 13 and 14 also displayed leishmanicidal activity with IC 50 of 11.60, 7.63, 2.86 and 14.90 µM, respectively, while 15 and 16 were inactive. The antiplasmodial, trypanocidal, and leishmanicidal activities of these compounds provide justification for the use of crude extract of P. kotschyi in the traditional treatment of malaria and other parasitic infectious diseases.
Antimicrobial agents are among the most commonly used medications. The prevalence of antimicrobial resistance in recent years has led to a renewed effort to discover newer antimicrobial agents for the treatment of infectious diseases (Hobson et al. ). Extracts of P. kotschyi were reported to display appreciable activity against some pathogenic microorganisms. Ayo et al. investigated the antimicrobial activity of petroleum ether, ethyl acetate and methanol extracts of the P. kotschyi leaves against Staphylococcus aureus , Salmonella typhi , Streptococcus pyogenes , Candida albicans , and Escherichia coli . The results of the study showed that the ethyl acetate extract exhibited antibacterial activity against all the tested organisms with MIC values of 10–20 mg/mL. In an another similar study, the crude methanol extract of the stem bark of this plant was also shown to exhibit good activity against a panel of pathogenic bacteria and fungi which include methicillin-resistant S. aureus (MRSA), S. aureus , S. pyogenes , Corynebacterium ulcerans , Bacillus subtilis , E. coli, S. typhi , Shigella dysenteriae , Klebsiella pneumoniae , Neisseria gonorrhoeae , Pseudomonas aeruginosa , C. albicans , C. krusei , and C. tropicalis with MIC values of 3.75–10.0 mg/mL (Alhassan et al. ). The methanol extract of the woody stem was also found to possess antifungal activity against C. krusei ATCC 6825 with an MIC of 6.25 mg/mL (Adeniyi et al. ). The secotirucallane triterpenes (compounds 28 , 29 and 30 ) isolated from the bark of P. kotschyi have been reported to possess significant antibacterial activity against Staphylococcus aureus ATCC 25923), Escherichia coli S2(1) and Pseudomonas aeruginosa with MIC ranging from 6 to 64 µg/mL. Compound 29 exhibited the highest antibacterial activity while 30 had the lowest (Mambou et al. ). The presence of these compounds is likely responsible for the antimicrobial property of P. kotschyi extracts and justify the ethnomedicinal use of this plant as a chewing stick for tooth cleaning and enhancement of oral health.
The ethanol extract of P. kotschyi stem bark has been reported to possess DPPH radical scavenging activity with an IC 50 of 4 µg/mL (Alain et al. ). A study on the hepatoprotective activity of methanol and aqueous extracts of the P. kotschyi leaves revealed that both extracts (at a dose of 750 mg/kg/day) were able to protect the liver against paracetamol induced oxidative damage (Eleha et al. ). A similar study conducted by Nchouwet et al. showed that 2 weeks pre-treatment with aqueous and methanol extracts of P. kotschyi stem bark (150 mg/kg/day) significantly suppressed the development of paracetamol induced hepatotoxicity in experimental rats.
Diabetes mellitus is disorder associated with abnormal glucose metabolism resulting from insulin insufficiency or dysfunction. It is one of the major non-communicable diseases that affect millions of people globally. Scientific investigation has revealed that P. kotschyi extracts possess some antidiabetic properties. Georgewill and Georgewill investigated the hypoglycaemic effect of aqueous extract of P. kotschyi leaves on alloxan induced diabetic rats. Results of their investigation revealed that oral administration of the extract (200 mg/kg/day for 14 days) caused significant hypoglycaemic effect in the experimental animals. The ethanol extract of roots of this plant was also reported to exhibit inhibitory activity against α-glucosidase (IC 50 = 5.0 ± 0.2 μg/mL), an important digestive enzyme targeted in diabetes treatment (Bothon et al. ).
Cancer is an important disease that is characterised by the abnormal rapid proliferation of cells that invade and destroy other tissues (Alhassan et al. ). It is a major public health problem throughout the world. Pharmacological studies have shown that P. kotschyi possesses anticancer potential. Kassim et al. investigated the antiproliferative activity and apoptosis induction effect of aqueous extract of P. kotschyi roots against a panel of prostate cancer cell lines, namely, PC3, DU-145, LNCaP and CWR-22 cell lines. Results from the 3-[4,5-dimethylthiazol-2yl]-2,5-diphenyltetrazolium bromide (MTT) assay showed that all four cancer cell lines exhibited a dose-dependent decrease in cell proliferation and viability after treatment with the aqueous extract with IC 50 values ranging from 12 to 42 µg/mL. The results obtained also showed that LNCaP, PC3, DU-145, and CWR-22 cell lines had 42, 35, 33 and 24% induced apoptotic cells, respectively, after treatment with the same extract. The results of both the antiproliferative and apoptosis assay indicated that the LNCaP cells were the most sensitive to the P. kotschyi extract. Heat shock protein 90 (Hsp90) is a molecular chaperone that is involved in the folding, activation and assembly of several proteins including oncoproteins such as HER2, Survivin, EGFR, Akt, Raf-1, mutant p53 (Calderwood et al. ; Dal Piaz et al. ). Hsp90 is often overexpressed in cancer cells. It has been demonstrated to play a vital role in tumour progression, malignancy and resistance to chemotherapeutic agents (Zhang et al. ). Hence, Hsp90 is recently considered as a viable molecular target for development of new anticancer drugs (Gupta et al. ). Phytoconstituents of P. kotschyi have been shown to possess significant Hsp90 inhibitory activity. Dal Piaz et al. investigated the Hsp90 binding capability of several compounds using a surface plasmon resonance (SPR) approach. They found that the limonoid orthoacetates ( 1–6 ) displayed good binding capability to the protein with compound 4 being the most effective. Compound 4 also exhibited significant anti-proliferative activity against three cancer cell lines, namely, PC-3 (human prostate cancer cells), A2780 (human ovarian carcinoma cells), and MCF-7 (human breast adenocarcinoma cells) with IC 50 values of 62 ± 0.4, 38 ± 0.7 and 25 ± 1.2 µM, respectively. These findings suggest that Hsp90 inhibition is a mechanism of action for anti-proliferative effects of the limonoids orthoacetates from P. kotschyi . These findings provide scientific bases for the future development of new anticancer agents from P. kotschyi in the form of a standardised herbal preparation or as a pure chemical entity.
Treatment of diarrhoea comes under one of the common ethnomedicinal uses of P. kotschyi . To further verify this claim, Essiet et al. investigated the antidiarrheal property of ethanol extract of P kotschyi leaves in Wistar albino rats. Diarrhoea was induced in the animals using castor oil. Results of the investigation revealed that oral administration of the extract (100, 200 and 400 mg/kg) produced significant ( p < 0.05) dose-dependent inhibition of induced diarrhoea (67–91%). The results also showed that the administered doses of the extract decreased intestinal transit time by 57–66% while intestinal fluid accumulation was decreased by 68–82%. This finding undoubtedly supports the traditional use of this plant in the treatment of diarrhoea.
There is a general perception that plant-based medicinal products are natural and thus, very safe for human consumption. However, this notion is wrong because several plants have been shown to produce wide range of adverse reactions some of which are capable of causing serious injuries, life-threatening conditions, and even death (Ekor ). Hence, it is of paramount importance to investigate the toxicity profile of traditional medicinal plants as well as their phytoconstituents in order to establish their safety. The study of toxicity is an essential component for new drug development process (Hornberg et al. ). Some toxicological studies have been carried out on the extracts of P. kotschyi . Nchouwet et al. investigated the acute and sub-chronic toxicity of P. kotschyi stem bark aqueous extract in albino rats. For the acute toxicity study, the LD 50 was found to be greater than 2000 mg/kg body weight. The sub-chronic administration of the aqueous extract at a dose of 400 mg/kg body weight/day for 28 days, caused significant increase in total protein and HDL-Cholesterol with concomitant decrease in LDL-cholesterol while other biochemical and hematological parameters were found to be within the normal range. However, histological examination revealed the presence of inflammation and necrosis in the kidney and liver tissues of animals treated with 400 mg/kg body weight-/day of extract while tissue samples from animals treated at lower doses remained normal. This implies that the extract may have exhibited some toxic effect on the kidney and liver tissues at 400 mg/kg body weight while it is relatively safe at lower doses. Kabiru et al. conducted a sub-chronic toxicity evaluation of a crude methanol extract of leaves in Sprague-Dawley rats at doses of 40, 200 and 1000 mg/kg body weight/day for 4 weeks. They found that the extract did not produce any significant alteration in both hematological and biochemical parameters when compared with standard controls. This implied that the extract was relatively non-toxic at the tested doses. Ezeokpo et al. also carried out a similar study with an ethanol extract of P. kotschyi leaves in Wistar rats. Their results revealed that the extract (400 mg/kg body weight/day) did not produce any significant derangement in hematological and biochemical parameters after 28 days of treatment. The above findings indicated that methanol and ethanol extracts of P. kotschyi leaves are relatively non-toxic at higher doses compared to the aqueous stem bark extract. However, more detailed research is still required to corroborate this finding. Albeit most of the pharmacological activities and chemical constituents reported on this plant have been obtained from the leaf’s extracts, the toxicity evaluation of ethanol, methanol and chloroform extracts of roots and stem bark of P. kotschy is yet to be carried out and reported. Hence, further toxicity studies on different extracts, fractions and chemical constituents of the root and stem bark of P. kotschyi are still required to ascertain the thorough safety of the plant.
P. kotschyi is an important medicinal plant which is used in the traditional treatment of different ailments. Based on its ethnomedicinal claims, extensive pharmacological and phytochemical investigations have been carried out which led to the isolation and characterisation of several bioactive constituents. Results from the pharmacological investigations on this plant and its phytoconstituents have demonstrated its high therapeutic potential in the treatment of cancer and tropical diseases, particularly malaria, leishmaniasis and trypanosomiasis. Although, experimental data support the beneficial medicinal properties of P. kotschyi , there were no sufficient data on the toxicity and safety profile of the plant. Nonetheless, this review provides the foundation for future work. Considering the amount of knowledge so far obtained on the medicinal properties of P. kotschyi , further studies on this plant should be directed towards establishing its safety profile as well as design and development of drug product either as single chemical entity or as a standardised herbal preparation. Tropical diseases are among the most neglected health problems in the world. The pharmaceutical industries show little research and development interest in this area albeit the devastating effect of such diseases. Therefore, research finding of this nature should be advanced towards the development of useful medicinal products.
|
Persistent cross-species transmission systems dominate Shiga toxin-producing | ca8ce3f7-0144-4646-b2ed-1cc04547e08b | 11778926 | Biochemistry[mh] | Several areas around the globe experience exceptionally high incidence of Shiga toxin-producing Escherichia coli (STEC), including the virulent serotype E. coli O157:H7. These include Scotland, Ireland, Argentina, and the Canadian province of Alberta . All are home to large populations of agricultural ruminants, STEC’s primary reservoir. However, there are many regions with similar ruminant populations where STEC incidence is unremarkable. What differentiates high-risk regions is unclear. Moreover, with systematic STEC surveillance only conducted in limited parts of the world, there may be unidentified regions with exceptionally high disease burden. STEC infections can arise from local reservoirs, transmitted through food, water, direct animal contact, or contact with contaminated environmental matrices. The most common reservoirs include domesticated ruminants such as cattle, sheep, and goats. Animal contact and consumption of contaminated meat and dairy products are significant risk factors for STEC, as are consumption of leafy greens, tomatoes, and herbs and recreational swimming that have been contaminated by feces from domestic ruminants While STEC has been isolated from a variety of other animal species and outbreaks have been linked to species such as deer and swine, it is unclear what roles they play as maintenance or intermediate hosts. STEC infections can be imported through food items traded nationally and internationally, as has been seen with E. coli O157:H7 outbreaks in romaine lettuce from the United States . Secondary transmission is believed to cause approximately 15% of cases, but transmission of the pathogen is not believed to be sustained through person-to-person transmission over the long term . The mix of STEC infection sources in a region directly influences public health measures needed to control disease burden. Living near cattle and other domesticated ruminants has been linked to STEC incidence, particularly for E. coli O157:H7 These studies suggest an important role for local reservoirs in STEC epidemiology. A comprehensive understanding of STEC’s disease ecology would enable more effective investigations into potential local transmission systems and ultimately their control. Here, we take a phylodynamic, genomic epidemiology approach to more precisely discern the role of the cattle reservoir in the dynamics of E. coli O157:H7 human infections. We focus on the high incidence region of Alberta, Canada to provide insight into characteristics that make the pathogen particularly prominent in such regions. Description of isolates Across the 1215 isolates included in the analyses, we identified 12,273 core genome SNPs. Clade G(vi) constituted 73.6% (n=894) of all isolates . Clade A, which is the most distinct of the E. coli O157:H7 clades, included non-Alberta isolates, two human isolates from Alberta, and no Alberta cattle isolates. The majority of all Alberta isolates belonged to the G(vi) clade (582 of 659; 88.3%), compared to 281 of the 1560 (18.0%) randomly sampled U.S. PulseNet isolates that were successfully assembled and QCed. Among the 62 non-randomly sampled global isolates, only 2 (3.2%) were clade G(vi) . There were 682 (76.3%) clade G(vi) isolates with the stx1a/stx2a profile and 210 (23.5%) with the stx2a -only profile, compared to 2 (0.6%) and 58 (18.1%), respectively, among the 321 isolates outside the G(vi) clade . The majority of clinical cases evolved from local cattle lineages In our primary sample of 121 human and 108 cattle isolates from Alberta from 2007 to 2015, SNP distances were comparable between species . Among sampled human cases, 19 (15.7%; 95% CI 9.7%, 23.4%) were within five SNPs of a sampled cattle strain. The median SNP distance between cattle sequences was 45 (IQR 36–56), compared to 54 (IQR 43–229) SNPs between human sequences from cases in Alberta during the same years. The phylogeny generated by our primary structured coalescent analysis indicated cattle were the primary reservoir, with a high probability that the hosts at nodes along the backbone of the tree were cattle . The root was estimated at 1802 (95% HPD 1731, 1861). The most recent common ancestor (MRCA) of clade G(vi) strains in Alberta was inferred to be a cattle strain, dated to 1969 (95% HPD 1959, 1979). With our assumption of a relaxed molecular clock, the mean clock rate for the core genome was estimated at 9.65×10 –5 (95% HPD 8.13×10 –5 , 1.13×10 –4 ) substitutions/site/year. The effective population size, N e , of the human E. coli O157:H7 population, was estimated as 1060 (95% HPD 698, 1477), and for cattle as 73 (95% HPD 50, 98). We estimated 108 (95% HPD 104, 112) human lineages arose from cattle lineages, and 14 (95% HPD 5, 23) arose from other human lineages . In other words, 88.5% of human lineages seen in Alberta from 2007 to 2015 arose from cattle lineages. We observed minimal influence of our choice of priors . Our sensitivity analysis of equal numbers of isolates from cattle and humans was largely consistent with our primary results, estimating that 94.3% of human lineages arose from cattle lineages . Locally persistent lineages account for the majority of ongoing human disease In our primary analysis, we identified 11 locally persistent lineages (LPLs) . After reincorporating down-sampled isolates, LPLs included a range of 5 (G(vi)-AB LPL 9)–26 isolates (G(vi)-AB LPL 1), with an average of 10. LPL assignment was based on the MCC tree of the combination of four independent chains. LPLs persisted for 5–9 y, with the average LPL spanning 8 y. By definition, MRCAs of each LPL were required to have a posterior probability ≥95% on the MCC tree, and in practice, all had posterior probabilities of 99.7–100%. Additionally, examining all trees sampled from the four chains supported the same major lineages . Our sensitivity analysis of equal numbers of isolates from cattle and humans identified 10 of the same 11 LPLs . G(vi)-AB LPL 9 was no longer identified as an LPL, because it fell below the five-isolate threshold after subsampling. Additionally, G(vi)-AB LPL 8 expanded to include a neighboring branch. LPLs tended to be clustered on the MCC tree. G(vi)-AB LPLs 1–4, 6–8, and 9 and 10 were clustered with MRCAs inferred in 1996 (95% HPD 1992, 1999), 1998 (95% HPD 1995, 2000), and 1993 (95% HPD 1989, 1996), respectively . Cattle were the inferred host of all three ancestral nodes. LPLs were assigned using a threshold of 30 SNPs. In sensitivity analysis testing alternate SNP thresholds, we observed LPLs mimicking the larger clusters of the LPLs from our primary analysis . LPLs included 71 of 108 (65.7%; 95% CI 56.0%, 74.6%) cattle and 46 of 121 (38.0%; 95% CI 29.3%, 47.3%) human isolates. Of the remaining human isolates, 33 (27.3%) were associated with imported infections and 42 (34.7%) with infections from transient local strains. Of the remaining cattle isolates, 11 (10.2%) were imported and 26 (24.1%) were associated with transmission from transient strains. Of the 117 isolates in LPLs, 7 (6.0%) carried only stx2a , and the rest stx1a/stx2a . Among the 112 non-LPL isolates, 1 (0.9%) was stx1a -only, 27 (24.1%) were stx2a -only, 5 (4.5%) were stx2c -only, 68 (60.7%) were stx1a/stx2a , 6 (5.4%) were stx1a/stx2c , and 5 (4.5%) were stx2a/stx2c . To understand long-term persistence, we expanded the phylogeny with additional Alberta Health isolates from 2009 to 2019. Six of the 11 LPLs identified in our primary analysis, G(vi)-AB LPLs 1, 2, 4, 7, 10, and 11, continued to cause disease during the 2016–2019 period . With most cases reported during 2018 and 2019 sequenced, we were able to estimate the proportion of reported E. coli O157:H7 associated with LPLs. Of 217 sequenced cases reported during these 2 y, 162 (74.7%; 95% CI 68.3%, 80.3%) arose from Alberta LPLs. The stx profile of LPL isolates shifted as compared to the primary analysis, with 83 (51.2%; 95% CI 43.3%, 59.2%) of the LPL isolates encoding only stx2a and the rest stx1a/stx2a . Among the 55 non-LPL isolates during 2018–2019, the stx2c -only profile emerged with 16 (29.1%; 95% CI 17.6%, 42.9%) isolates, stx2a -only was found in six (10.9%; 95% CI 4.1%, 22.2%) isolates, and five (9.1%; 95% CI 3.0%, 20.0%) isolates carried both stx2a and stx2c . All five large (≥10 cases) sequenced outbreaks in Alberta during the study period were within clade G(vi). G(vi)-AB LPLs 2 and 7 gave rise to three large outbreaks, accounting for 117 cases (both sequenced and unsequenced), including 83 from an extended outbreak by a single strain in 2018 and 2019, defined as isolates within five SNPs of one another. The two large outbreaks that did not arise from LPLs both occurred in 2014 and were responsible for 164 cases. Locally persistent lineages were not imported Of the 494 U.S. isolates analyzed, nine (1.8%; 95% CI 0.8%, 3.4%) occurred within Alberta LPLs after re-incorporating down-sampled isolates . None of the 62 global isolates were associated with Alberta LPLs. The 9 U.S. isolates were part of G(vi)-AB LPLs 2 (n=3), 4 (n=4), 7 (n=1), and 11 (n=1), all of which had Alberta isolates that spanned 9–13 y and predated the U.S. isolates. There was no evidence of U.S. or global ancestors of LPLs. Based on migration events calculated from the structured tree, we estimated that 11.0% of combined human and cattle Alberta lineages were imported . Alberta sequences were separated from U.S. and global sequences by a median of 63 (IQR 45–236) and 225 (IQR 209–249) SNPs, respectively. Including U.S. and global isolates in the phylogeny did not change which LPLs we identified . The minimum SNPs that LPL isolates differed by was lower than in the Alberta-only analyses, because the core genome shared by all Alberta, U.S., and global isolates was smaller than that of only the Alberta isolates. Alberta sequences included in some LPLs changed slightly. G(vi)-AB LPL 4 lost three Alberta isolates from clinical cases, and G(vi)-AB LPLs 6 and 11 both lost one cattle and one human isolate from Alberta. In these LPLs, the isolates no longer included were the most outlying isolates in the LPLs defined using only Alberta isolates . Of the 217 Alberta human isolates from 2018 and 2019, 160 (73.7%) were still associated with LPLs after the addition of U.S. and global isolates, demonstrating the stability of the extended analysis results. Across the 1215 isolates included in the analyses, we identified 12,273 core genome SNPs. Clade G(vi) constituted 73.6% (n=894) of all isolates . Clade A, which is the most distinct of the E. coli O157:H7 clades, included non-Alberta isolates, two human isolates from Alberta, and no Alberta cattle isolates. The majority of all Alberta isolates belonged to the G(vi) clade (582 of 659; 88.3%), compared to 281 of the 1560 (18.0%) randomly sampled U.S. PulseNet isolates that were successfully assembled and QCed. Among the 62 non-randomly sampled global isolates, only 2 (3.2%) were clade G(vi) . There were 682 (76.3%) clade G(vi) isolates with the stx1a/stx2a profile and 210 (23.5%) with the stx2a -only profile, compared to 2 (0.6%) and 58 (18.1%), respectively, among the 321 isolates outside the G(vi) clade . In our primary sample of 121 human and 108 cattle isolates from Alberta from 2007 to 2015, SNP distances were comparable between species . Among sampled human cases, 19 (15.7%; 95% CI 9.7%, 23.4%) were within five SNPs of a sampled cattle strain. The median SNP distance between cattle sequences was 45 (IQR 36–56), compared to 54 (IQR 43–229) SNPs between human sequences from cases in Alberta during the same years. The phylogeny generated by our primary structured coalescent analysis indicated cattle were the primary reservoir, with a high probability that the hosts at nodes along the backbone of the tree were cattle . The root was estimated at 1802 (95% HPD 1731, 1861). The most recent common ancestor (MRCA) of clade G(vi) strains in Alberta was inferred to be a cattle strain, dated to 1969 (95% HPD 1959, 1979). With our assumption of a relaxed molecular clock, the mean clock rate for the core genome was estimated at 9.65×10 –5 (95% HPD 8.13×10 –5 , 1.13×10 –4 ) substitutions/site/year. The effective population size, N e , of the human E. coli O157:H7 population, was estimated as 1060 (95% HPD 698, 1477), and for cattle as 73 (95% HPD 50, 98). We estimated 108 (95% HPD 104, 112) human lineages arose from cattle lineages, and 14 (95% HPD 5, 23) arose from other human lineages . In other words, 88.5% of human lineages seen in Alberta from 2007 to 2015 arose from cattle lineages. We observed minimal influence of our choice of priors . Our sensitivity analysis of equal numbers of isolates from cattle and humans was largely consistent with our primary results, estimating that 94.3% of human lineages arose from cattle lineages . In our primary analysis, we identified 11 locally persistent lineages (LPLs) . After reincorporating down-sampled isolates, LPLs included a range of 5 (G(vi)-AB LPL 9)–26 isolates (G(vi)-AB LPL 1), with an average of 10. LPL assignment was based on the MCC tree of the combination of four independent chains. LPLs persisted for 5–9 y, with the average LPL spanning 8 y. By definition, MRCAs of each LPL were required to have a posterior probability ≥95% on the MCC tree, and in practice, all had posterior probabilities of 99.7–100%. Additionally, examining all trees sampled from the four chains supported the same major lineages . Our sensitivity analysis of equal numbers of isolates from cattle and humans identified 10 of the same 11 LPLs . G(vi)-AB LPL 9 was no longer identified as an LPL, because it fell below the five-isolate threshold after subsampling. Additionally, G(vi)-AB LPL 8 expanded to include a neighboring branch. LPLs tended to be clustered on the MCC tree. G(vi)-AB LPLs 1–4, 6–8, and 9 and 10 were clustered with MRCAs inferred in 1996 (95% HPD 1992, 1999), 1998 (95% HPD 1995, 2000), and 1993 (95% HPD 1989, 1996), respectively . Cattle were the inferred host of all three ancestral nodes. LPLs were assigned using a threshold of 30 SNPs. In sensitivity analysis testing alternate SNP thresholds, we observed LPLs mimicking the larger clusters of the LPLs from our primary analysis . LPLs included 71 of 108 (65.7%; 95% CI 56.0%, 74.6%) cattle and 46 of 121 (38.0%; 95% CI 29.3%, 47.3%) human isolates. Of the remaining human isolates, 33 (27.3%) were associated with imported infections and 42 (34.7%) with infections from transient local strains. Of the remaining cattle isolates, 11 (10.2%) were imported and 26 (24.1%) were associated with transmission from transient strains. Of the 117 isolates in LPLs, 7 (6.0%) carried only stx2a , and the rest stx1a/stx2a . Among the 112 non-LPL isolates, 1 (0.9%) was stx1a -only, 27 (24.1%) were stx2a -only, 5 (4.5%) were stx2c -only, 68 (60.7%) were stx1a/stx2a , 6 (5.4%) were stx1a/stx2c , and 5 (4.5%) were stx2a/stx2c . To understand long-term persistence, we expanded the phylogeny with additional Alberta Health isolates from 2009 to 2019. Six of the 11 LPLs identified in our primary analysis, G(vi)-AB LPLs 1, 2, 4, 7, 10, and 11, continued to cause disease during the 2016–2019 period . With most cases reported during 2018 and 2019 sequenced, we were able to estimate the proportion of reported E. coli O157:H7 associated with LPLs. Of 217 sequenced cases reported during these 2 y, 162 (74.7%; 95% CI 68.3%, 80.3%) arose from Alberta LPLs. The stx profile of LPL isolates shifted as compared to the primary analysis, with 83 (51.2%; 95% CI 43.3%, 59.2%) of the LPL isolates encoding only stx2a and the rest stx1a/stx2a . Among the 55 non-LPL isolates during 2018–2019, the stx2c -only profile emerged with 16 (29.1%; 95% CI 17.6%, 42.9%) isolates, stx2a -only was found in six (10.9%; 95% CI 4.1%, 22.2%) isolates, and five (9.1%; 95% CI 3.0%, 20.0%) isolates carried both stx2a and stx2c . All five large (≥10 cases) sequenced outbreaks in Alberta during the study period were within clade G(vi). G(vi)-AB LPLs 2 and 7 gave rise to three large outbreaks, accounting for 117 cases (both sequenced and unsequenced), including 83 from an extended outbreak by a single strain in 2018 and 2019, defined as isolates within five SNPs of one another. The two large outbreaks that did not arise from LPLs both occurred in 2014 and were responsible for 164 cases. Of the 494 U.S. isolates analyzed, nine (1.8%; 95% CI 0.8%, 3.4%) occurred within Alberta LPLs after re-incorporating down-sampled isolates . None of the 62 global isolates were associated with Alberta LPLs. The 9 U.S. isolates were part of G(vi)-AB LPLs 2 (n=3), 4 (n=4), 7 (n=1), and 11 (n=1), all of which had Alberta isolates that spanned 9–13 y and predated the U.S. isolates. There was no evidence of U.S. or global ancestors of LPLs. Based on migration events calculated from the structured tree, we estimated that 11.0% of combined human and cattle Alberta lineages were imported . Alberta sequences were separated from U.S. and global sequences by a median of 63 (IQR 45–236) and 225 (IQR 209–249) SNPs, respectively. Including U.S. and global isolates in the phylogeny did not change which LPLs we identified . The minimum SNPs that LPL isolates differed by was lower than in the Alberta-only analyses, because the core genome shared by all Alberta, U.S., and global isolates was smaller than that of only the Alberta isolates. Alberta sequences included in some LPLs changed slightly. G(vi)-AB LPL 4 lost three Alberta isolates from clinical cases, and G(vi)-AB LPLs 6 and 11 both lost one cattle and one human isolate from Alberta. In these LPLs, the isolates no longer included were the most outlying isolates in the LPLs defined using only Alberta isolates . Of the 217 Alberta human isolates from 2018 and 2019, 160 (73.7%) were still associated with LPLs after the addition of U.S. and global isolates, demonstrating the stability of the extended analysis results. Focusing on a region that experiences an especially high incidence of STEC, we conducted a deep genomic epidemiologic analysis of E. coli O157:H7’s multi-host disease dynamics. Our study identified multiple locally evolving lineages transmitted between cattle and humans. These were persistently associated with E. coli O157:H7 illnesses over periods of up to 13 y, the length of our study. Of clinical importance, there was a dramatic shift in the stx profile of the strains arising from locally persistent lineages toward strains carrying only stx2a , which has been associated with increased progression to hemolytic uremic syndrome (HUS) . Our study has provided quantitative estimates of cattle-to-human migration in a high-incidence region, the first such estimates of which we are aware. Our estimates are consistent with prior work that established an increased risk of STEC associated with living near cattle . We showed that 88.5% of strains infecting humans arose from cattle lineages. These transitions can be seen as a combination of the infection of humans from local cattle or cattle-related reservoirs in clade G(vi) and the historic evolution of E. coli O157:H7 from cattle in the rare clades. While our findings indicate the majority of human cases arose from cattle lineages, transmission may involve intermediate hosts or environmental sources several steps removed from the cattle reservoir. Small ruminants (e.g. sheep, goats) have been identified as important STEC reservoirs, and Alberta has experienced outbreaks linked to swine . Exchange of strains between cattle and other animals may occur if co-located, if surface water sources near farms are contaminated, and through wildlife, including deer, birds, and flies Humans can also become infected from environmental sources, such as through swimming in contaminated water. Although transmission systems may be multi-faceted, our analysis demonstrates that local cattle remain an integral part of the transmission system for the vast majority of cases, even when they may not be the immediate source of infection. Indeed, despite our small sample of E. coli O157:H7 isolates from cattle, 15.7% of our human cases were within five SNPs of a cattle isolate, suggesting that cattle were a recent source of transmission, either through direct contact with the animal or their environments or consumption of contaminated food products. The cattle-human transitions we estimated were based on structured coalescent theory, which we used throughout our analyses. This approach is similar to other phylogeographic methods that have previously been applied to E. coli O157:H7 . We inferred the full backbone of the Alberta E. coli O157:H7 phylogeny as arising from cattle, consistent with the postulated global spread of the pathogen via ruminants . Our estimate of the origin of the serotype, at 1802 (95% HPD 1731, 1861), was somewhat earlier than previous estimates, but consistent with global (1890; 95% HPD 1845, 1925) and the United Kingdom (1840; 95% HPD 1817, 1855) studies that used comparable methods. Our dating of the G(vi) clade in Alberta to 1969 (95% HPD 1959, 1979) also corresponds to proposed migrations of clade G into Canada from the U.S. in 1965–1977 . Our study thus adds to the growing body of work on the larger history of E. coli O157:H7, providing an in-depth examination of the G(vi) clade. Our identification of the 11 locally persistent lineages (LPLs) is significant in demonstrating that the majority of Alberta’s reported E. coli O157:H7 illnesses are of local origin. Our definition ensured that every LPL had an Alberta cattle strain and at least five isolates separated by at least 1 y, making the importation of the isolates in a lineage highly unlikely. For an LPL to be fully imported, cattle and human isolates would need to be repeatedly imported from a non-Alberta reservoir where the lineage persisted over several years. Further supporting the evolution of the LPLs within Alberta, all 11 LPLs were in clade G(vi), several were phylogenetically related with MRCAs dating to the late 1990 s, and few non-Alberta isolates fell within LPLs. The nineU.S. isolates associated with Alberta LPLs may reflect Alberta cattle that were slaughtered in the U.S. or infections in travelers from the U.S. Thus, we are confident that the identified LPLs represent locally evolving lineages and potential persistent sources of disease. We also showed that the identification of these LPLs was robust to the sampling strategy, with only the smallest LPL failing to be identified after subsampling left it with <5 isolates. We estimated the proportion of E. coli O157:H7 that were imported into Alberta in two ways. Based on our LPL analysis, we estimated only 27% of human and 10% of cattle E. coli O157:H7 isolates were imported. This was slightly higher than the overall importation estimate of 11% for all Alberta lineages from our global structured coalescent analysis. Our global structured coalescent analysis also estimated that 3% of lineages in the U.S. and 2% of lineages outside the U.S. and Canada had been exported from Alberta, suggesting that Alberta is not a significant contributor to the global E. coli O157:H7 burden beyond its borders. These results place the E. coli O157:H7 population in Alberta within a larger context, indicating that the majority of diseases can be considered local. At least one study has attempted to differentiate local vs. non-local lineages based on travel status, of which may be appropriate in some locations but can miss cases imported through food products, such as produce imported from other countries. To our knowledge, our study provides the first comprehensive determination of local vs. imported status for E. coli O157:H7 cases using external reference cases. Similar studies in regions of both high and moderate incidence would provide further insight into the role of localization on E. coli O157:H7 incidence. Of the 11 lineages we identified as LPLs during the 2007–2015 period, six were also associated with cases that occurred during the 2016–2019 period. During the initial period, 38% of human cases were linked to an LPL, and 6% carried only stx2a . The risk of HUS increases in strains of STEC carrying only stx2a , relative to stx1a/stx2a , meaning the earlier LPL population had fewer high-virulence strains. In 2018 and 2019, the six long-term LPLs were associated with both greater incidence and greater virulence, encompassing 75% of human cases with more than half of LPL isolates carrying only stx2a . The cause of this shift remains unclear, though shifts toward greater virulence in E. coli O157:H7 populations have been seen elsewhere . The growth and diversity of G(vi)-AB LPLs 2, 4, and 7 in the later period suggest these lineages were in stable reservoirs or adapted easily to new niches. Identifying these reservoirs could yield substantial insights into the disease ecology that supports LPL maintenance and opportunities for disease prevention, given the significant portion of illnesses caused by persistent strains. The high proportion of cases associated with cattle-linked local lineages is consistent with what is known about the role of cattle in STEC transmission. Among sporadic STEC infections, 26% have been attributed to animal contact and the farm environment, with a further 19% to pink or raw meat . Similarly, 24% of E. coli O157 outbreaks in the U.S. have been attributed to beef, animal contact, water, or other environmental sources . In Alberta, these are all inherently local exposures, given that 90% of beef consumed in Alberta is produced and/or processed there. Even person-to-person transmission, responsible for 15% of sporadic cases and 16% of outbreaks, includes secondary transmission from cases infected from local sources, which may explain our estimate of 11.5% of human lineages arising from other human lineages. We developed a novel measure of persistence for use in this study, specifically for the purposes of identifying lineages that pose an ongoing threat to public health in a specific region. Persistence has been variably defined in the literature, for example, as shedding of the same strain for at least 4 mo . Most recently, the U.S. CDC identified the first Recurring, Emergent, and Persistent (REP) STEC strain, REPEXH01, an E. coli O157:H7 strain detected since 2017 in over 600 cases. REPEXH01 strains are within 21 allele differences of one another ( https://www.cdc.gov/ncezid/dfwed/outbreak-response/rep-strains/repexh01.html ), and REP strains from similar enteric pathogens are defined based on allele differences of 13–104. Given that we used high-resolution SNP analysis rather than cgMLST, we used a difference of ≤30 SNPs to define persistent lineages. While both our study and the REPEXH01 strain identified by the CDC indicate that persistent strains of E. coli O157:H7 exist, the O157:H7 serotype was defined as sporadic in a German study using the 4 mo shedding definition . This may be due to strain differences between the two locations, but it might also indicate that persistence occurs at the host community level, rather than the individual host level. Understanding microbial drivers of persistence is an active field of research, with early findings suggesting a correlation of STEC persistence to the accessory genome and traits such as biofilm formation and nutrient metabolism . Our approach to studying persistence was specifically designed for longitudinal sampling in high-incidence regions and may be useful for others attempting to identify sources that disproportionately contribute to disease burden. Although we used data from the reservoir species to help define the LPLs in this study, we are testing alternate approaches that rely on only routinely collected public health data. We limited our analysis to E. coli O157:H7 despite the growing importance of non-O157 STEC, as historical multi-species collections of non-O157 isolates are lacking. As serogroups differ meaningfully in exposures, our results may not be generalizable beyond the O157 serogroup. However, cattle are still believed to be a primary reservoir for non-O157 STEC, and cattle density is associated with the risk of several non-O157 serogroups . Person-to-person transmission remains a minor contributor to the STEC burden. For all of these reasons, if we were to conduct this analysis in non-O157 STEC, we expect the majority of human lineages would arise from cattle lineages. Additionally, persistence within the cattle reservoir has been observed for a range of serogroups, suggesting that LPLs also likely exist among non-O157 STEC. Our findings may have implications beyond STEC, as well. Other zoonotic enteric pathogens such as Salmonella and Campylobacter can persist, and outbreaks are regularly linked to localized animal populations and produce-growing operations contaminated by animal feces. The U.S. CDC has also defined REP strains for these pathogens. LPLs could shed light on how and where persistent strains are proliferating, and thus where they can be controlled. The identification of LPLs serves multiple purposes, because they suggest the existence of local reservoir communities that maintain specific strains for long periods. First, they further our understanding of the complex systems that allow STEC to persist. In this study, the LPLs we identified persisted for 5–13 y. The reservoir communities that enable persistence could involve other domestic and wild animals previously found to carry STEC . The feedlot environment also likely plays an important role in persistence, as water troughs and pen floors have been identified as important sources of STEC for cattle . Identifying LPLs is a first step in identifying these reservoir communities and determining what factors enable strains to persist, so as to identify them for targeted interventions. Second, the identification of these LPLs in cattle could identify the specific local reservoirs of STEC. Similar to source tracing in response to outbreaks, LPLs provide an opportunity for cattle growers to identify cattle carrying the specific strains that are associated with a large share of human disease in Alberta. While routinely vaccinating against STEC has not been shown to be efficacious or cost-effective, a ring-type vaccination strategy in response to an identified LPL isolate could overcome the limitations of previous vaccination strategies. Third, the identification of new clinical cases infected with LPL strains could help direct public health investigations toward local sources of transmission. Finally, the disease burden associated with LPLs could be compared across locations and may help explain how high-incidence regions differ from regions with lower incidence. Our analysis was limited to only cattle and humans. Had isolates from a wider range of potential reservoirs been available, we would have been able to elucidate more clearly the roles that various hosts and common sources of infection play in local transmission. Additional hosts may help explain the 1 human-to-cattle predicted transmission, which could be erroneous. As with all studies utilizing public health data, sampling from only severe cases of disease is biased toward clinical isolates. In theory, this could limit the genetic variation among human isolates if virulence is associated with specific lineages. However, clinical isolates were more variable than cattle isolates, dominating the most divergent clade A, so the overrepresentation of severe cases does not appear to have appreciably biased the current study. Similarly, in initially selecting an equal number of human and cattle isolates, we sampled a larger proportion of the human-infecting E. coli O157:H7 population compared to the population that colonizes cattle. As cattle are the primary reservoir of E. coli O157:H7, the pathogen is more prevalent in cattle than in humans, who appear to play a limited role in sustained transmission. In sampling a larger proportion of the strains that infect humans, we likely sampled a wider diversity of these strains compared to those in cattle, which could have biased the analysis toward finding humans as the ancestral host. Thus, the proportion of human lineages arising from cattle lineages (88.5%) might be underestimated, which is also suggested by our sensitivity analysis of equal numbers of cattle and clinical isolates. Finally, we were not able to estimate the impact of strain migration between Alberta and the rest of Canada, because locational metadata for publicly available E. coli O157:H7 sequences from Canada was limited. E. coli O157:H7 infections are a pressing public health problem in many high-incidence regions around the world including Alberta, where a recent childcare outbreak caused >300 illnesses. In the majority of sporadic cases, and even many outbreaks, the source of infection is unknown, making it critical to understand the disease ecology of E. coli O157:H7 at a system level. Here, we have identified a high proportion of human cases arising from cattle lineages and a low proportion of imported cases. Local transmission systems, including intermediate hosts and environmental reservoirs, need to be elucidated to develop management strategies that reduce the risk of STEC infection. In Alberta, local transmission is dominated by a single clade, and over the extended study period, persistent lineages caused an increasing proportion of disease. The local lineages with long-term persistence are of particular concern because of their increasing virulence, yet they also present opportunities as larger, more stable targets for reservoir identification and control. Study design and population We conducted a multi-host genomic epidemiology study in Alberta, Canada. Our primary analysis focused on 2007–2015 due to the availability of isolates from intensive provincial cattle studies . These studies rectally sampled feces from individual animals, hide swabs, fecal pats from the floors of pens of commercial feedlot cattle, or feces from the floors of transport trailers. In studies of pens of cattle, samples were collected from the same cattle at least twice over a 4 to 6 mo period. A one-time composite sample was collected from cattle in transport trailers, which originated from feedlots or auction markets in Alberta. To select both cattle and human isolates, we block randomized by year to ensure representation across the period. We define isolates as single bacterial species obtained from culture. We sampled 123 E. coli O157 cattle isolates from 4660 available. Selected cattle isolates represented 7 of 12 cattle study sites and 56 of 89 sampling occasions from the source studies . We sampled 123 of 1148 E. coli O157 isolates collected from cases reported to the provincial health authority (Alberta Health) during the corresponding time period (Appendix 1). In addition to the 246 isolates for the primary analysis, we contextualized our findings with two additional sets of E. coli O157:H7 isolates : 445 from Alberta Health from 2009 to 2019 and already sequenced as part of other public health activities and 1970 from the U.S. and elsewhere around the world between 1999 and 2019. The additional Alberta Health isolates were sequenced by the National Microbiology Laboratory (NML)-Public Health Agency of Canada (Winnipeg, Manitoba, Canada) as part of PulseNet Canada activities. Isolates sequenced by the NML for 2018 and 2019 constituted the majority of reported E. coli O157:H7 cases for those years (217 of 247; 87.9%). U.S. and global isolates from both cattle and humans were identified from previous literature (n=104) and BV-BRC (n=193). As both processed beef and live cattle are frequently imported into Alberta from the U.S., we selected additional E. coli O157:H7 sequences available through the U.S. CDC’s PulseNet BioProject PRJNA218110. From 2010–2019, 6,791 O157:H7 whole genome sequences were available from the U.S. PulseNet project, 1673 (25%) of which we randomly selected for assembly and clade typing. This study was approved by the University of Calgary Conjoint Health Research Ethics Board, #REB19-0510. A waiver of consent was granted, and all case data were deidentified. Whole genome sequencing, assembly, and initial phylogeny The 246 isolates for the primary analysis were sequenced using Illumina NovaSeq 6000 and assembled into contigs using the Unicycler v04.9 pipeline, as described previously (BioProject PRJNA870153) . Raw read FASTQ files were obtained from Alberta Health for the additional 445 isolates sequenced by the NML and from NCBI for the 152 U.S. and 54 global sequences. We used the SRA Toolkit v3.0.0 to download sequences for U.S. and global isolates using their BioSample (i.e. SAMN) numbers. The corresponding FASTQ files could not be obtained for the six U.S. and seven global isolates we had selected . PopPUNK v2.5.0 was used to cluster Alberta isolates and identify any outside the O157:H7 genomic cluster . For assembling and quality checking (QC) all sequences, we used the Bactopia v3.0.0 pipeline . This pipeline performed an initial QC step on the reads using FastQC v0.12.1, which evaluated read count, sequence coverage, and sequence depth, with failed reads excluded from subsequent assembly. None of the isolates were eliminated during this step for low read quality. We used the Shovill v1.1.0 assembler within the Bactopia pipeline to de novo assemble the Unicycler contigs for the primary analysis and raw reads from the supplementary datasets. Trimmomatic was run as part of Shovill to trim adapters and read ends with quality lower than six and discard reads after trimming with overall quality scores lower than 10. Bactopia generated a quality report on the assemblies, which we assessed based on number of contigs (<500), genome size (≥5.1 Mb), N50 (>30,000), and L50 (≤50). Low-quality assemblies were removed. This included one U.S. sequence, for which two FASTQ files had been attached to a single BioSample identifier; the other sequence for the isolate passed all quality checks and remained in the analysis. Additionally, 16 sequences from the primary analysis dataset and four from the extended Alberta data had a total length of <5.1 Mb. These sequences corresponded exactly to those identified by the PopPUNK analysis to be outside the primary E. coli O157:H7 genomic cluster . Finally, although all isolates were believed to be of cattle or clinical origin during the initial selection, a detailed metadata review identified one isolate of environmental origin in the primary analysis dataset and eight that had been isolated from food items in the extended Alberta data. These were excluded. We used STECFinder v1.1.0 to determine the Shiga toxin gene ( stx ) profile and confirm the E. coli O157:H7 serotype using the wzy or wzx O157 O-antigen genes and detection of the H7 H-antigen. Bactopia’s Snippy workflow, which incorporates Snippy v4.6.0, Gubbins v3.3.0, and IQTree v2.2.2.7, followed by SNP-Sites v2.5.1, was used to generate a core genome SNP alignment with recombinant blocks removed. The maximum likelihood phylogeny of the core genome SNP alignment generated by IQTree was visualized in Microreact v251. The number of core SNPs between isolates was calculated using PairSNP v0.3.1. Clade was determined based on the presence of at least one defining SNP for the clade as published previously Isolates were identified to the clade level, except for clade G where we separated out subclade G(vi). After processing, we had 229 isolates (121 human, 108 cattle) in our primary sample and 430 additional Alberta Health isolates . We had 178 U.S. or global isolates from previous literature (n=88; U.S. n=41, global n=47) and BV-BRC (n=90; U.S. n=75, global n=15). Of the 1673 isolates randomly sampled from the U.S. PulseNet project, 1560 were successfully assembled and passed QC. These included 309 clade G isolates, all of which we included in the analysis; we also randomly sampled and included 69 non-clade G isolates from this sample. Phylodynamic and statistical analyses For our primary analysis, we created a timed phylogeny, a phylogenetic tree on the scale of time, in BEAST2 v2.6.7 using the structured coalescent model in the Mascot v3.0.0 package with demes for cattle and humans . Sequences were down-sampled prior to analysis if within 0–2 SNPs and <3 m from another sequence from the same host type, leaving 115 human and 84 cattle isolates in the primary analysis . The analysis was run using four different seeds to confirm that all converged to the same solution, and tree files were combined before generating a maximum clade credibility (MCC) tree. State transitions between cattle and human isolates over the entirety of the tree, with their 95% highest posterior density (HPD) intervals, were also calculated from the combined tree files. We determined the influence of the prior assumptions on the analysis with a run that sampled from the prior distribution (Appendix 1). We conducted a sensitivity analysis in which we randomly subsampled 84 of the human isolates so that both species had the same number of isolates in the analysis. LPLs were identified based on following criteria: (1) a single lineage of the MCC tree with a most recent common ancestor (MRCA) with ≥95% posterior probability; (2) all isolates ≤30 core SNPs from one another; (3) contained at least 1 cattle isolate; (4) contained ≥5 isolates; and (5) the isolates were collected at sampling events (for cattle) or reported (for humans) over a period of at least 1 y. We counted the number of isolates associated with LPLs, including those down-sampled prior to the phylodynamic analysis. We conducted sensitivity analyses examining different SNP thresholds for the LPL definition. From non-LPL isolates, we estimated the number of local transient isolates vs. imported isolates. For the 121 human E. coli O157:H7 isolates in the primary sample prior to down-sampling, we determined what portion belonged to locally persistent lineages and what portion was likely to be from local transient E. coli O157:H7 populations vs. imported. Human isolates within the LPLs were enumerated (n=46). The 75 human isolates outside LPLs included 56 clade G(vi) isolates and 19 non-G(vi) isolates. Based on the MCC tree from the primary analysis, none of the non-G(vi) human isolates were likely to have been closely related to an isolate from the Alberta cattle population, suggesting that all 19 were imported. As a proportion of all non-LPL human isolates, these 19 constituted 25.3%. While it may be possible that all clade G(vi) isolates were part of a local evolving lineage, it is also possible that the exchange of both cattle and food from other locations was causing the regular importation of clade G(vi) strains and infections. Thus, we used the proportion of non-LPL human isolates outside the G(vi) clade to estimate the proportion of non-LPL human isolates within the G(vi) clade that were imported; i.e., 56 × 25.3 % = 14 . We then conducted a similar exercise for cattle isolates. To contextualize our results in terms of the ongoing human disease burden, we created a timed phylogeny using a constant, unstructured coalescent model of the 199 Alberta isolates from the primary analysis and the additional Alberta Health isolates . The two sets of sequences were combined and down-sampled, leaving 272 human and 84 cattle isolates . We identified LPLs as above, and leveraged the near-complete sequencing of isolates from 2018 and 2019 to calculate the proportion of reported human cases associated with LPLs. Finally, we created a timed phylogeny of Alberta, U.S., and global from 1996 to 2019 to test whether the LPLs were linked to ancestors from locations outside Canada . Due to the size of this tree, we created both unstructured and structured versions. Clade A isolates were excluded due to their small number in Alberta and high level of divergence from other E. coli O157:H7 clades. Down-sampling was conducted separately by species and location. The phylogeny included 358 Alberta, 350 U.S., and 61 global isolates after down-sampling. All BEAST2 analyses were run for 100,000,000 Markov chain Monte Carlo iterations or until all parameters converged with effective sample sizes >200, whichever was longer. Exact binomial 95% confidence intervals (CIs) were computed for proportions. We conducted a multi-host genomic epidemiology study in Alberta, Canada. Our primary analysis focused on 2007–2015 due to the availability of isolates from intensive provincial cattle studies . These studies rectally sampled feces from individual animals, hide swabs, fecal pats from the floors of pens of commercial feedlot cattle, or feces from the floors of transport trailers. In studies of pens of cattle, samples were collected from the same cattle at least twice over a 4 to 6 mo period. A one-time composite sample was collected from cattle in transport trailers, which originated from feedlots or auction markets in Alberta. To select both cattle and human isolates, we block randomized by year to ensure representation across the period. We define isolates as single bacterial species obtained from culture. We sampled 123 E. coli O157 cattle isolates from 4660 available. Selected cattle isolates represented 7 of 12 cattle study sites and 56 of 89 sampling occasions from the source studies . We sampled 123 of 1148 E. coli O157 isolates collected from cases reported to the provincial health authority (Alberta Health) during the corresponding time period (Appendix 1). In addition to the 246 isolates for the primary analysis, we contextualized our findings with two additional sets of E. coli O157:H7 isolates : 445 from Alberta Health from 2009 to 2019 and already sequenced as part of other public health activities and 1970 from the U.S. and elsewhere around the world between 1999 and 2019. The additional Alberta Health isolates were sequenced by the National Microbiology Laboratory (NML)-Public Health Agency of Canada (Winnipeg, Manitoba, Canada) as part of PulseNet Canada activities. Isolates sequenced by the NML for 2018 and 2019 constituted the majority of reported E. coli O157:H7 cases for those years (217 of 247; 87.9%). U.S. and global isolates from both cattle and humans were identified from previous literature (n=104) and BV-BRC (n=193). As both processed beef and live cattle are frequently imported into Alberta from the U.S., we selected additional E. coli O157:H7 sequences available through the U.S. CDC’s PulseNet BioProject PRJNA218110. From 2010–2019, 6,791 O157:H7 whole genome sequences were available from the U.S. PulseNet project, 1673 (25%) of which we randomly selected for assembly and clade typing. This study was approved by the University of Calgary Conjoint Health Research Ethics Board, #REB19-0510. A waiver of consent was granted, and all case data were deidentified. The 246 isolates for the primary analysis were sequenced using Illumina NovaSeq 6000 and assembled into contigs using the Unicycler v04.9 pipeline, as described previously (BioProject PRJNA870153) . Raw read FASTQ files were obtained from Alberta Health for the additional 445 isolates sequenced by the NML and from NCBI for the 152 U.S. and 54 global sequences. We used the SRA Toolkit v3.0.0 to download sequences for U.S. and global isolates using their BioSample (i.e. SAMN) numbers. The corresponding FASTQ files could not be obtained for the six U.S. and seven global isolates we had selected . PopPUNK v2.5.0 was used to cluster Alberta isolates and identify any outside the O157:H7 genomic cluster . For assembling and quality checking (QC) all sequences, we used the Bactopia v3.0.0 pipeline . This pipeline performed an initial QC step on the reads using FastQC v0.12.1, which evaluated read count, sequence coverage, and sequence depth, with failed reads excluded from subsequent assembly. None of the isolates were eliminated during this step for low read quality. We used the Shovill v1.1.0 assembler within the Bactopia pipeline to de novo assemble the Unicycler contigs for the primary analysis and raw reads from the supplementary datasets. Trimmomatic was run as part of Shovill to trim adapters and read ends with quality lower than six and discard reads after trimming with overall quality scores lower than 10. Bactopia generated a quality report on the assemblies, which we assessed based on number of contigs (<500), genome size (≥5.1 Mb), N50 (>30,000), and L50 (≤50). Low-quality assemblies were removed. This included one U.S. sequence, for which two FASTQ files had been attached to a single BioSample identifier; the other sequence for the isolate passed all quality checks and remained in the analysis. Additionally, 16 sequences from the primary analysis dataset and four from the extended Alberta data had a total length of <5.1 Mb. These sequences corresponded exactly to those identified by the PopPUNK analysis to be outside the primary E. coli O157:H7 genomic cluster . Finally, although all isolates were believed to be of cattle or clinical origin during the initial selection, a detailed metadata review identified one isolate of environmental origin in the primary analysis dataset and eight that had been isolated from food items in the extended Alberta data. These were excluded. We used STECFinder v1.1.0 to determine the Shiga toxin gene ( stx ) profile and confirm the E. coli O157:H7 serotype using the wzy or wzx O157 O-antigen genes and detection of the H7 H-antigen. Bactopia’s Snippy workflow, which incorporates Snippy v4.6.0, Gubbins v3.3.0, and IQTree v2.2.2.7, followed by SNP-Sites v2.5.1, was used to generate a core genome SNP alignment with recombinant blocks removed. The maximum likelihood phylogeny of the core genome SNP alignment generated by IQTree was visualized in Microreact v251. The number of core SNPs between isolates was calculated using PairSNP v0.3.1. Clade was determined based on the presence of at least one defining SNP for the clade as published previously Isolates were identified to the clade level, except for clade G where we separated out subclade G(vi). After processing, we had 229 isolates (121 human, 108 cattle) in our primary sample and 430 additional Alberta Health isolates . We had 178 U.S. or global isolates from previous literature (n=88; U.S. n=41, global n=47) and BV-BRC (n=90; U.S. n=75, global n=15). Of the 1673 isolates randomly sampled from the U.S. PulseNet project, 1560 were successfully assembled and passed QC. These included 309 clade G isolates, all of which we included in the analysis; we also randomly sampled and included 69 non-clade G isolates from this sample. For our primary analysis, we created a timed phylogeny, a phylogenetic tree on the scale of time, in BEAST2 v2.6.7 using the structured coalescent model in the Mascot v3.0.0 package with demes for cattle and humans . Sequences were down-sampled prior to analysis if within 0–2 SNPs and <3 m from another sequence from the same host type, leaving 115 human and 84 cattle isolates in the primary analysis . The analysis was run using four different seeds to confirm that all converged to the same solution, and tree files were combined before generating a maximum clade credibility (MCC) tree. State transitions between cattle and human isolates over the entirety of the tree, with their 95% highest posterior density (HPD) intervals, were also calculated from the combined tree files. We determined the influence of the prior assumptions on the analysis with a run that sampled from the prior distribution (Appendix 1). We conducted a sensitivity analysis in which we randomly subsampled 84 of the human isolates so that both species had the same number of isolates in the analysis. LPLs were identified based on following criteria: (1) a single lineage of the MCC tree with a most recent common ancestor (MRCA) with ≥95% posterior probability; (2) all isolates ≤30 core SNPs from one another; (3) contained at least 1 cattle isolate; (4) contained ≥5 isolates; and (5) the isolates were collected at sampling events (for cattle) or reported (for humans) over a period of at least 1 y. We counted the number of isolates associated with LPLs, including those down-sampled prior to the phylodynamic analysis. We conducted sensitivity analyses examining different SNP thresholds for the LPL definition. From non-LPL isolates, we estimated the number of local transient isolates vs. imported isolates. For the 121 human E. coli O157:H7 isolates in the primary sample prior to down-sampling, we determined what portion belonged to locally persistent lineages and what portion was likely to be from local transient E. coli O157:H7 populations vs. imported. Human isolates within the LPLs were enumerated (n=46). The 75 human isolates outside LPLs included 56 clade G(vi) isolates and 19 non-G(vi) isolates. Based on the MCC tree from the primary analysis, none of the non-G(vi) human isolates were likely to have been closely related to an isolate from the Alberta cattle population, suggesting that all 19 were imported. As a proportion of all non-LPL human isolates, these 19 constituted 25.3%. While it may be possible that all clade G(vi) isolates were part of a local evolving lineage, it is also possible that the exchange of both cattle and food from other locations was causing the regular importation of clade G(vi) strains and infections. Thus, we used the proportion of non-LPL human isolates outside the G(vi) clade to estimate the proportion of non-LPL human isolates within the G(vi) clade that were imported; i.e., 56 × 25.3 % = 14 . We then conducted a similar exercise for cattle isolates. To contextualize our results in terms of the ongoing human disease burden, we created a timed phylogeny using a constant, unstructured coalescent model of the 199 Alberta isolates from the primary analysis and the additional Alberta Health isolates . The two sets of sequences were combined and down-sampled, leaving 272 human and 84 cattle isolates . We identified LPLs as above, and leveraged the near-complete sequencing of isolates from 2018 and 2019 to calculate the proportion of reported human cases associated with LPLs. Finally, we created a timed phylogeny of Alberta, U.S., and global from 1996 to 2019 to test whether the LPLs were linked to ancestors from locations outside Canada . Due to the size of this tree, we created both unstructured and structured versions. Clade A isolates were excluded due to their small number in Alberta and high level of divergence from other E. coli O157:H7 clades. Down-sampling was conducted separately by species and location. The phylogeny included 358 Alberta, 350 U.S., and 61 global isolates after down-sampling. All BEAST2 analyses were run for 100,000,000 Markov chain Monte Carlo iterations or until all parameters converged with effective sample sizes >200, whichever was longer. Exact binomial 95% confidence intervals (CIs) were computed for proportions. |
Round-robin testing for LMO2 and MYC as immunohistochemical markers to screen | d9135c03-7170-4aa5-8e8b-c52635b30056 | 11329383 | Anatomy[mh] | Aggressive large B-cell lymphomas (aLBCL), including transformed B-cell lymphomas from low-grade non-Hodgkin lymphomas and Burkitt lymphoma (BL), are the most common lymphomas causing tissue involvement in western countries . Although chronic lymphocytic leukemia/small lymphocytic lymphoma has higher incidence than aLBCL, the disease is largely limited to the peripheral blood, and the diagnostic approach of the disease is not based on tissue examination . It is known that the status of MYC gene is prognostically relevant in aLBCL. The rearrangements involving MYC ( MYC -R) are a defining genetic alteration of high-grade B-cell lymphomas (HGBL) carrying BCL2 and/or BCL6 rearrangements, as well as Burkitt lymphoma (BL) . Furthermore, MYC -R have a prevalence of 5–15% in diffuse large B-cell lymphoma, not otherwise specified (DLBCL-NOS), which is the most common subtype of aLBCL. This lymphoma represents a morphologically, genetically, and clinically heterogeneous entity and the detection of MYC- R associates with a poorer outcome after standard chemoimmunotherapy, as HGBL carrying MYC and BCL2 rearrangements . In addition, 10 to 26% transformed DLBCL (tDLBCL) carry MYC -R . Thus, all these data indicate the need of the identification of MYC status in aLBCL. There has been an extraordinary increase in the knowledge of hematological neoplasms since the publication of the unified REAL classification . New genetic tools, gene expression profiling (GEP), and next-generation sequencing (NGS) have expanded the understanding of the biology of aLBCL. Progress in the understanding of aLBCL points to a more refined classification including the combination of molecular and genetic data that ideally should also include suitable information obtained from morphology and immunohistochemistry (IHC) . However, the current strategy to diagnose aLBCL in most laboratories relies on the use of IHC combined with cytogenetics, where available. Genetic testing is mandatory for the classification of aLBCL . Since the overall incidence of MYC -R in LBCL is low, and cytogenetics is not available elsewhere, it is necessary to identify useful markers to screen MYC -R in routine practice. In previous studies, we observed the utility of the association between LMO2 loss of expression by IHC with the presence of MYC -R in aLBCL . LMO2 is a cysteine-rich protein which is a critical regulator of hematopoiesis, initially described as a recurrent chromosomal translocation partner of the TCR genes associated with T-cell acute lymphoblastic leukemia . GEP studies included LMO2 among the genes defining the GCB-like profile signature . It is currently known that LMO2 is expressed in aLBCL and that the immunohistochemical expression of LMO2 has an impact in the survival of patients treated with immunochemotherapy . The favorable prognosis has been related to mechanisms of genomic instability associated with DNA damage . Our previous studies showing the utility of LMO2 as a marker to identify MYC -R included two independent series of 330 and 365 samples, shared methods, and obtained similar results, unveiling intralaboratory reproducibility . Two studies published later including 90 and 180 aLBCL, respectively, showed similar results to ours . In the present study, we aimed to evaluate the interobserver and interlaboratory reproducibility for LMO2 and MYC detected by IHC in aLBCL. We proceeded in two phases. In the first phase, 50 aLBCL cases from one center, collected retrospectively, were circulated to evaluate the interobserver concordance of IHC. The second phase of the study was conducted prospectively, aiming to evaluate the performance of each laboratory. Thus, each enrolled hospital collected their in-house aLBCL, adding LMO2 antibody to their diagnostic panel. The results of the immunohistochemical panel were correlated with MYC FISH results obtained from each laboratory. At the same time, as we were collecting such prospective data, we also pretended to identify the incidence of MYC -R in the centers involved in the study.
To analyze interobserver reproducibility we performed a round-robin test. Fifty aLBCL diagnosed between 2016 and 2021 were selected from the files of the Pathology laboratory of the Hospital del Mar, Barcelona, based on available material. All cases were diagnosed according to the 4th revised WHO classification . Primary mediastinal large B-cell lymphoma, primary central nervous system lymphoma and HIV-associated lymphomas were excluded. The series included whole tissue sections of 28 excisional biopsies (EB) and 22 core needle biopsies (CNB). Each case comprised a set of slides including hematoxylin and eosin, CD10 (clone SP67), BCL6 (clone GI191E-A8), MUM-1/IRF4 (clone MRQ-43), BCL2 (clone 124), LMO2 (clone 1A9-1), and MYC (clone Y69), all from Ventana, Roche, Tucson, AZ, USA. The immunohistochemical studies were performed, as previously described . During 2020–2021, all cases were circulated and evaluated by 7 hematopathologists (FC, GT, IV, CL-M, LL, NP, and LC) from 5 tertiary hospitals located in the health area of Barcelona, Spain (Hospital de Bellvitge, center 1; Hospital Germans Trias i Pujol, center 2; Hospital del Mar, center 3; Hospital Parc Tauli, center 4; Hospital Mutua Terrassa, center 5), in 2 to 4 individual sessions. The evaluation and assessment for all the antibodies were the same, as previously described . The cutoff used for CD10, BCL6, MUM1/IRF4, and LMO2 was 30%, and for MYC and BCL2 was 40% and 50%, respectively. Lymphoma diagnoses and FISH results of MYC , BCL2 , and BCL6 were blinded for all observers. Split probes for MYC and BCL6 and dual fusion probes for BCL2/IGH and MYC/IGH were all provided by Vysis, Abbott Molecular, and Des Plainescity, IL, USA. FISH was performed and evaluated, as described following the criteria of Ventura . The second phase of the study corresponded to the interlaboratory reproducibility phase. A prospective study was performed from January 2021 to June 2022 by each center. Samples of daily practice with a diagnostic suspicion of aLBCL as per the 4th revised WHO classification were stained with CD10, BCL2, BCL6, MUM1/IRF4, and MYC, according to the protocols of each laboratory. Same entities as in phase one were excluded. Clones and sources are described in supplementary Table . LMO2 was included in the immunohistochemical panel for the diagnostic workout in all cases. MYC , BCL2 , and BCL6 FISH probes were performed and interpreted according to the protocols and probes of each laboratory (supplementary Table ). Each center was asked to fill in an Excel template including blinded ID number, patient data (age, sex, and relevant medical history), IHC and MYC FISH results, and diagnosis. This series include patients diagnosed and treated in each institution corresponding to their health areas of influence. Some differences in terms of healthcare services between the centers exist: centers 1 and 2 receive patients needing complex treatments such as allogenic transplant and CAR-T cell therapy, and their health area covers a population of 1.3 million and 800.000 inhabitants, respectively. Centers 3 to 5 cover similar health areas in terms of the number of population that includes approximately 400,000 inhabitants. Complex treatments are referred to other centers, different to centers 1 and 2. Center 3, in addition, centralize cases for diagnosis-genetic testing. The approach to FISH testing was also variable, since centers 2 and 3 used MYC/IGH probes to determine the partner of MYC -R cases, and center 2 only tested BCL2 and BCL6 FISH for MYC -R cases. Patient samples were collected in accordance with the Declaration of Helsinki and approved by ethics committee. In the present study, we decided to keep the nomenclature of the revised 4th edition of the WHO classification, as it was developed between January 2020 and June 2022. We have only modified Burkitt-like lymphoma with 11q aberration included in the revised 4th edition of the WHO classification and used the mixed term high grade/large B-cell lymphoma with 11q aberrations (HGBL/LBCL-11q), as handled in the 21st EAHP-SH meeting in Florence, 2022. Statistical analysis To quantify the agreement between observers in the phase 1 of the study we used the Fleiss’ kappa index. χ 2 test, unpaired t tests, or non-parametric tests, were used when necessary. For the second phase, accuracy, sensitivity, specificity, and positive/negative predictive ratios were calculated for MYC and LMO2. Likelihood positive and negative ratios were calculated to evaluate the diagnostic accuracy of the results obtained. P values < 0.05 were considered statistically significant for all tests. Data were analyzed using Microsoft Excel 2019 (Microsoft Corporation, Redmond, WA, USA) and the v28.0.1.0 of the software package IBM SPSS Statistics (Armonk, NY, USA).
To quantify the agreement between observers in the phase 1 of the study we used the Fleiss’ kappa index. χ 2 test, unpaired t tests, or non-parametric tests, were used when necessary. For the second phase, accuracy, sensitivity, specificity, and positive/negative predictive ratios were calculated for MYC and LMO2. Likelihood positive and negative ratios were calculated to evaluate the diagnostic accuracy of the results obtained. P values < 0.05 were considered statistically significant for all tests. Data were analyzed using Microsoft Excel 2019 (Microsoft Corporation, Redmond, WA, USA) and the v28.0.1.0 of the software package IBM SPSS Statistics (Armonk, NY, USA).
Interobserver reproducibility In this phase, we analyzed 50 cases including 2 (4%) HGBCL-NOS, 8 (16%) high-grade B-cell lymphoma with MYC and BLC2/BCL6 rearrangement (HGBCL with MYC-BLC2/BCL6 R), 31 (62%) DLBCL-NOS, 7 (14%) tDLBCL (6 from follicular lymphoma-FL; 1 from marginal zone lymphoma, MZL), 1 (2%) BL and 1 (2%) HGBCL/LBCL-11q. The overall incidence of MYC -R in the series was 28% (14/50 cases): 1 HGBCL-NOS, 8 HGBCL with MYC-BLC2/BCL6 R, 4 DLBCL-NOS and 1 BL. Any tDLBCL included in the series presented MYC -R. The patients were 32 males and 18 females, with a median age of 62 years (range 34–91). Thirty-three (66%) cases were nodal and 17 (34%) extranodal. The samples were obtained by excisional biopsies (EB) in 28 (56%) cases and 22 (44%) were core needle biopsies (CNB). We first analyzed how was the concordance of CD10, BCL6, and MUM1/IRF4 to assess the COO of all cases included in this series and in DLBCL-NOS, to compare our results with the previously published by other groups. Fleiss’ Kappa index for COO concordance was 0.84, considering all cases included in the series, and 0.77 for DLBCL-NOS only ( P < 0.001 each). The concordance analysis for the markers using the same approach (total cases and DLBCL-NOS only) was CD10, 0.86/0.79; BCL6, 0.83/0.80; and MUM1/IRF4, 0.88/0.83. For LMO2 and MYC the results were: LMO2, 0.87/0.89; and MYC: 0.70/0.64 ( P < 0.001 for each marker). CD10, BCL6, and MUM1/IRF4 and LMO2 staining obtained high agreement values, whereas the lowest concordance rate was obtained for MYC staining, particularly when the COO analysis was considered (Fig. ). Table summarizes the causes of disagreement in the IHC evaluation between the observers. We classified discrepancies as major when 3 observers disagreed; intermediate when 2 observers disagreed; and minor when only 1 observer disagreed. The discrepancies for LMO2 occurred in 7/50 (14%) cases and were 2 major, 2 intermediate, and 3 minor. The causes of LMO2 discrepancies were primarily attributed to the differences in the interpretation of the cutoff for LMO2 between observers. Interestingly, 6 of 7 cases were CD10 negative and, as published, the level of expression of LMO2 is variable in such cases . Three of 7 cases were non-GCB-like according to Hans’ algorithm. Among the 4 GCB-like, there were 2 DLBCL-NOS, 1 HGBCL with MYC-BLC2/BCL6 , and 1 HGBCL-NOS. Additional variability associated with major and intermediate discrepancies were attributed to the size of the samples and the quality of the tissue in 3 cases. All were CNB, 2 with necrotic areas, and 1 with areas of bad fixation. Minor discrepancies occurred in CD10 negative cases (1 GCB and 2 non-GCB-like) (Fig. a–e). Discrepancies for MYC occurred in 16/50 (32%) cases and were 6 major, 5 intermediate, and 5 minor. In 13 of 16 (81%) discrepant samples, MYC expression varied between 25 and 50%, and the observers agreed with that it was difficult to determine whether the tumor cells were over or not the cutoff defined for MYC (Fig. f–h). The size of the sample was also important, since 11/16 (69%) discrepancies occurred in CNB. Additional discrepancies were attributed to the quality of the samples in 3 cases. Two EB had fixation issues that caused irregular staining of MYC, and 1 CNB had a crush artifact. Only one case, an EB CD10 negative and non-GCB-like, presented simultaneous discrepancies for LMO2 and MYC that were minor and intermediate, respectively. Interlaboratory reproducibility The second phase of the study included 213 cases which were collected during a period of 18 months. Briefly, centers 1 to 5 included 55, 35, 69, 36, and 18 cases, respectively. Table shows the results of all aLBCL included. Overall, the whole series comprised 4 HGBCL-NOS, 19 HGBCL with MYC-BLC2/BCL6 R, 150 DLBCL-NOS, 33 tDLBCL (28 transformed FL and 5 transformed MZL), and 7 BL. The median age of the patients was 68 years (range 19–92 years). About 141 cases were nodal and 72 extranodal. After excluding BL, and following the Hans algorithm, 118 (57%) cases were GCB-like (69 DLBCL-NOS, 4 HGBCL-NOS, 18 HGBCL with MYC-BLC2/BCL6 R, and 27 tDLBCL), and 88 (43%) cases were non-GCB-like (81 DLBCL-NOS, 1 HGBCL with MYC-BLC2/BCL6 R, and 6 tDLBCL). Considering the whole series, 34/213 cases harbored MYC -R, with an overall incidence of 16%: 7 BL, 7 DLBCL-NOS, 19 HGBCL with MYC-BLC2/BCL6 R, and 1 tDLBCL. CD10 was expressed in 101 cases (47%), LMO2 in 132 cases (62%), and MYC in 78 cases (37%). The statistic measures of the performance of LMO2 and MYC compared with the presence of MYC -R as the gold standard of all cases included in the series and CD10 positive cases are shown in Table . Center 5 is not included, since no MYC -R were detected. Supplementary Tables and show the results by center, the Hans algorithm, and double expression of BCL2 and MYC proteins. As expected, the results obtained for LMO2 in CD10 positive cases ameliorated the results of the whole series, except for the negative predictive value (NPV). Comparing LMO2 with MYC, the group of CD10 positive cases showed higher values for the specificity (86% vs 79%), positive predictive value (66% vs 58%), likelihood positive value (5.47 vs 3.78), and accuracy (83% vs 79%), whereas the NPV remained similar (90% vs 91%). Remarkably, taking into account the variability of sources and approaches used for the diagnosis of MYC -R in each laboratory, the overall results were similar to those obtained in our two previous studies (Table ). Specially, high similar values for the specificity and NPV were obtained in the three studies. Overall, we identified 16 dissociated cases that were as follows: 7 cases carrying MYC -R showed double positive expression of CD10 and LMO2 (CD10+/LMO2+/ MYC -R); and 9 cases with CD10+/LMO2- phenotype in which we did not identify MYC -R (CD10+/LMO2-/no- MYC -R). Among the 7 CD10+/LMO2+/ MYC -R cases, 4 had MYC protein expression over 40%. On the contrary, 6 out of 9 cases showing CD10+/LMO2-/no- MYC -R profile, had expression of MYC by IHC below 40%. Finally, the incidence of MYC -R varied among centers (center 1: 25; center 2: 26, center 3: 13; center 4: 6; and center 5: 0%). As centers 1 to 3 receive external patients and consultation cases, we wanted to clarify the real incidence of MYC -R in our series. After excluding the referred cases, centers 1 to 3 had an incidence for MYC -R of 23%, 19%, and 7%, respectively.
In this phase, we analyzed 50 cases including 2 (4%) HGBCL-NOS, 8 (16%) high-grade B-cell lymphoma with MYC and BLC2/BCL6 rearrangement (HGBCL with MYC-BLC2/BCL6 R), 31 (62%) DLBCL-NOS, 7 (14%) tDLBCL (6 from follicular lymphoma-FL; 1 from marginal zone lymphoma, MZL), 1 (2%) BL and 1 (2%) HGBCL/LBCL-11q. The overall incidence of MYC -R in the series was 28% (14/50 cases): 1 HGBCL-NOS, 8 HGBCL with MYC-BLC2/BCL6 R, 4 DLBCL-NOS and 1 BL. Any tDLBCL included in the series presented MYC -R. The patients were 32 males and 18 females, with a median age of 62 years (range 34–91). Thirty-three (66%) cases were nodal and 17 (34%) extranodal. The samples were obtained by excisional biopsies (EB) in 28 (56%) cases and 22 (44%) were core needle biopsies (CNB). We first analyzed how was the concordance of CD10, BCL6, and MUM1/IRF4 to assess the COO of all cases included in this series and in DLBCL-NOS, to compare our results with the previously published by other groups. Fleiss’ Kappa index for COO concordance was 0.84, considering all cases included in the series, and 0.77 for DLBCL-NOS only ( P < 0.001 each). The concordance analysis for the markers using the same approach (total cases and DLBCL-NOS only) was CD10, 0.86/0.79; BCL6, 0.83/0.80; and MUM1/IRF4, 0.88/0.83. For LMO2 and MYC the results were: LMO2, 0.87/0.89; and MYC: 0.70/0.64 ( P < 0.001 for each marker). CD10, BCL6, and MUM1/IRF4 and LMO2 staining obtained high agreement values, whereas the lowest concordance rate was obtained for MYC staining, particularly when the COO analysis was considered (Fig. ). Table summarizes the causes of disagreement in the IHC evaluation between the observers. We classified discrepancies as major when 3 observers disagreed; intermediate when 2 observers disagreed; and minor when only 1 observer disagreed. The discrepancies for LMO2 occurred in 7/50 (14%) cases and were 2 major, 2 intermediate, and 3 minor. The causes of LMO2 discrepancies were primarily attributed to the differences in the interpretation of the cutoff for LMO2 between observers. Interestingly, 6 of 7 cases were CD10 negative and, as published, the level of expression of LMO2 is variable in such cases . Three of 7 cases were non-GCB-like according to Hans’ algorithm. Among the 4 GCB-like, there were 2 DLBCL-NOS, 1 HGBCL with MYC-BLC2/BCL6 , and 1 HGBCL-NOS. Additional variability associated with major and intermediate discrepancies were attributed to the size of the samples and the quality of the tissue in 3 cases. All were CNB, 2 with necrotic areas, and 1 with areas of bad fixation. Minor discrepancies occurred in CD10 negative cases (1 GCB and 2 non-GCB-like) (Fig. a–e). Discrepancies for MYC occurred in 16/50 (32%) cases and were 6 major, 5 intermediate, and 5 minor. In 13 of 16 (81%) discrepant samples, MYC expression varied between 25 and 50%, and the observers agreed with that it was difficult to determine whether the tumor cells were over or not the cutoff defined for MYC (Fig. f–h). The size of the sample was also important, since 11/16 (69%) discrepancies occurred in CNB. Additional discrepancies were attributed to the quality of the samples in 3 cases. Two EB had fixation issues that caused irregular staining of MYC, and 1 CNB had a crush artifact. Only one case, an EB CD10 negative and non-GCB-like, presented simultaneous discrepancies for LMO2 and MYC that were minor and intermediate, respectively.
The second phase of the study included 213 cases which were collected during a period of 18 months. Briefly, centers 1 to 5 included 55, 35, 69, 36, and 18 cases, respectively. Table shows the results of all aLBCL included. Overall, the whole series comprised 4 HGBCL-NOS, 19 HGBCL with MYC-BLC2/BCL6 R, 150 DLBCL-NOS, 33 tDLBCL (28 transformed FL and 5 transformed MZL), and 7 BL. The median age of the patients was 68 years (range 19–92 years). About 141 cases were nodal and 72 extranodal. After excluding BL, and following the Hans algorithm, 118 (57%) cases were GCB-like (69 DLBCL-NOS, 4 HGBCL-NOS, 18 HGBCL with MYC-BLC2/BCL6 R, and 27 tDLBCL), and 88 (43%) cases were non-GCB-like (81 DLBCL-NOS, 1 HGBCL with MYC-BLC2/BCL6 R, and 6 tDLBCL). Considering the whole series, 34/213 cases harbored MYC -R, with an overall incidence of 16%: 7 BL, 7 DLBCL-NOS, 19 HGBCL with MYC-BLC2/BCL6 R, and 1 tDLBCL. CD10 was expressed in 101 cases (47%), LMO2 in 132 cases (62%), and MYC in 78 cases (37%). The statistic measures of the performance of LMO2 and MYC compared with the presence of MYC -R as the gold standard of all cases included in the series and CD10 positive cases are shown in Table . Center 5 is not included, since no MYC -R were detected. Supplementary Tables and show the results by center, the Hans algorithm, and double expression of BCL2 and MYC proteins. As expected, the results obtained for LMO2 in CD10 positive cases ameliorated the results of the whole series, except for the negative predictive value (NPV). Comparing LMO2 with MYC, the group of CD10 positive cases showed higher values for the specificity (86% vs 79%), positive predictive value (66% vs 58%), likelihood positive value (5.47 vs 3.78), and accuracy (83% vs 79%), whereas the NPV remained similar (90% vs 91%). Remarkably, taking into account the variability of sources and approaches used for the diagnosis of MYC -R in each laboratory, the overall results were similar to those obtained in our two previous studies (Table ). Specially, high similar values for the specificity and NPV were obtained in the three studies. Overall, we identified 16 dissociated cases that were as follows: 7 cases carrying MYC -R showed double positive expression of CD10 and LMO2 (CD10+/LMO2+/ MYC -R); and 9 cases with CD10+/LMO2- phenotype in which we did not identify MYC -R (CD10+/LMO2-/no- MYC -R). Among the 7 CD10+/LMO2+/ MYC -R cases, 4 had MYC protein expression over 40%. On the contrary, 6 out of 9 cases showing CD10+/LMO2-/no- MYC -R profile, had expression of MYC by IHC below 40%. Finally, the incidence of MYC -R varied among centers (center 1: 25; center 2: 26, center 3: 13; center 4: 6; and center 5: 0%). As centers 1 to 3 receive external patients and consultation cases, we wanted to clarify the real incidence of MYC -R in our series. After excluding the referred cases, centers 1 to 3 had an incidence for MYC -R of 23%, 19%, and 7%, respectively.
In this study, we aimed to evaluate the clinical reproducibility of LMO2 identified by IHC in aLBCL. To evaluate the interobserver reproducibility, we used a similar strategy to other studies and selected a set of cases that were independently evaluated by 7 hematopathologists. We realized that we agreed in the interpretation of the markers included in the Hans algorithm, as other authors described previously , and these results encouraged us to analyze LMO2 and MYC. We observed fewer discrepancies for LMO2 than MYC and attributed primarily the disagreement to the interpretation of the cutoff used. For LMO2, most discrepancies occurred in CD10 negative cases, and these are the aLBCL showing higher variability of LMO2 expression. In our previous studies, we observed that LMO2 protein expression was very high in CD10 positive cases and mostly negative in MYC -R cases, showing very low variability. However, CD10 negative and non-GCB-like tumors showed more fluctuating expression of LMO2, ranging from 10 to 40% . Such variability was already observed in GEP studies, where ABC and unclassifiable aLBCL had high levels of mRNA LMO2, particularly among tumors with no MYC -R. When MYC -R where present in such cases, LMO2 was low in unclassified but higher values persisted in the ABC subtype . MYC disagreement occurred mostly in CNB with values of MYC expression ranging from 25 to 50%. Our results are similar to those obtained in the study of Mahmoud that analyzed two independent sets of cases and evaluated whole tissue slides, including a total number of 35 aLBCL (5 BL and 30 DLBCL). In this study, the authors obtained an overall Kappa score of 0.68 and attributed such moderate concordance to the inherent heterogeneity of MYC expression in DLBCL. They concluded the need to be cautious when interpreting cases with MYC staining close to 40%. Moreover, the authors showed that the preselection of fields of 1 mm, as used in TMA concordance studies, improved the agreement between observers, but did not eliminate discrepancy at all. In summary, our results indicate higher agreement between observers for LMO2, compared to MYC. In the second phase of the study, we wanted to know how useful was the inclusion of LMO2 in the immunohistochemical panels used for the work up of aLBCL. Then, all centers used the same clone and conditions for LMO2, but did not add changes to their protocols routinely used for the additional markers. Notably, analyzing the total number of cases diagnosed by the five centers, we obtained similar results to our previous studies. In comparison with the results of such series, we observed a slight decrease in the sensitivity, PPV, and positive likelihood ratio in the multicenter study, values concerning the ability to identify the association between LMO2 loss and MYC -R presence. We realized that the approaches to the detection of MYC -R were variable among centers in terms of sources of the probes used, usage of the probes, and interpreters of the FISH technique. It is known that the approach to the diagnosis of MYC -R may influence the ability to detect such genetic alteration . Then, since the methods to study MYC -R may be quite heterogeneous in the real world, the identification of additional markers should help to evaluate the cytogenetic results after FISH testing. Ancillary markers may also help to suspect the presence of MYC cryptic insertions that may occur in aLBCL and decrease the number of false negative cases carrying MYC -R not detected by FISH . In the present study, we have also compared the utility of classifiers such as the Hans algorithm and double expression of MYC and BCL2 proteins to detect MYC -R. The results are shown in supplementary Table and do not improve the CD10/LMO2 approach. Our cohort of CD10 positive cases lacking LMO2 expression predicted the presence of MYC -R with high levels of specificity, accuracy, positive and negative predictive values, and good positive and negative likelihood ratios. We decided to analyze the multicenter results as a unique series assuming the variability of the diagnostic approaches to avoid the bias induced in the screening tests when the number of cases studied is low. With this approach, the specificity, NPV and accuracy were 86%, 90%, and 83%, respectively. When we analyzed the same parameters per center, we observed higher variability due to the lower number of cases included in each center. However, considering the individual results, one center obtained a value of NPV around 80%, one around 90%, and two obtained values of 100%. By using the profile CD10+/LMO2-, it is desirable to obtain very high NPV to avoid false negative cases and therefore miss cases carrying MYC -R. In this series, 4 of 7 false negative cases had high expression of MYC by IHC, suggesting that the combination of CD10, LMO2, and MYC may be useful to screen MYC -R. Likewise, MYC low expression may be useful to clarify false positive cases, as observed in 6 of 9 cases in the series. Nevertheless, the group of cases that we designated as dissociated CD10/LMO2 deserves further analyses to clarify their clinicopathological characteristics and weather combined with additional markers may help to identify MYC -R in aLBCL. Finally, we wanted to know the approximate incidence of MYC -R among centers. Considering the characteristics of each center, we tried to clean external cases received in each center, assuming the hypothesis that incidences by centers should be similar. Then, before excluding cases outside the health area of influence of each, the incidences of MYC -R in aLBCL ranged from 0 to 26%. After exclusion, the incidence varied from 0 to 23%, with centers 1 to 3 showing a decrease of their incidence. These results may be related to the heterogeneity of aLBCL but raises the questions about how to approach to FISH testing and whether epidemiological differences exist among such health areas. To the best of our knowledge, studies evaluating the agreement of MYC interpretation by FISH concordance in aLBCL are lacking. In conclusion, in this study we pretended to evaluate the clinical reproducibility of LMO2 immunohistochemical expression to screen MYC -R in aLBCL. In the first phase of the study, we observed high agreement between the observers interpreting LMO2, higher than the results obtained for MYC. In the second phase, we realized the variable approaches used to diagnose MYC -R, and we conclude that combining the profile CD10, LMO2, and MYC may be a useful method to screen the presence of MYC -R in aLBCL. As a result, all centers enrolled in the study included LMO2 in their diagnostic work up for aLBCL.
ESM 1 (DOCX 23.9 kb)
|
Cesarean section and the gestational duration of subsequent pregnancies: A nationwide register-based cohort study | 3305cac8-2713-4e0b-b755-5c9c217ce86f | 11798495 | Surgical Procedures, Operative[mh] | A pregnancy’s gestational duration is essential in determining the prognosis of a newborn . The World Health Organization (WHO) defines a pregnancy as full-term at 37+0 to 41+6 completed gestational weeks. Delivery outside this interval increases adverse outcomes, and preterm birth is the worldwide leading cause of perinatal mortality and lifelong neurological sequelae . Even within the full-term interval, the outcome still depends on the exact gestational duration. Higher perinatal mortality is observed among newborns born outside the 39+0 to 40+6 interval . A constantly increasing cesarean section (CS) rate is reported worldwide , which can be explained by a number of obstetrical and sociological factors. In Sweden, the CS rate was 17.9% during 2008–2017 . Compared to vaginal delivery, CS is associated with increased incidence of several adverse maternal and neonatal outcomes . While some CSs are necessary, this creates an incentive to limit birth through CS when the indication is relative. A meta-analysis found an adjusted relative risk of 1.12 (95% CI 1.01–1.24) for preterm birth after CS . Reports have since shown an association between late-stage CS and preterm birth . There has been a discussion regarding the association between CS and preterm birth, partly because they share risk factors but also due to the inclusion of both spontaneous and iatrogenic onsets of births in the studied cohorts . Little focus has been directed towards the links between the previous delivery mode and the duration of subsequent pregnancies, missing the opportunity to study other time-points of pregnancy that are also crucial for maternal and fetal health. The type of onset of birth is affected by previous obstetrical history. Women with a previous CS are more likely to have a repeat CS . As an iatrogenic onset of birth, such as a planned CS, has a direct impact on the gestational duration of the pregnancy, the type of onset of birth has to be adequately accounted for when studying relationships between previous delivery modes and gestational duration of subsequent pregnancies. This study aims to determine the impact of previous delivery modes on the incidence of deviant gestational duration of subsequent pregnancies. To do this we have compared women with previous CS(‘s) versus women with previous vaginal delivery(ies), in mothers who had two or more pregnancies. We give specific attention to the order of the previous delivery modes when analysing a woman’s first three pregnancies, covering a larger part of the woman’s child-bearing years. To account for the type of onset of birth, we use survival models as a new approach to evaluate the net effect of the previous delivery modes on the gestational duration of the subsequent pregnancies.
This is a register-based cohort study using pregnancies in the Swedish Medical Birth Register (MBR) . The Swedish National Board of Health and Welfare maintains the register and covers 98–99% of all deliveries nationwide. It provides information about the mother, obstetric history, current pregnancy, delivery, and data concerning the baby. The data are recorded by hospital staff . Individual-level data were pseudonymized for research use and provided under a permit from the Swedish National Board of Health and Welfare. The exposure in this study was the previous delivery mode: CS, vaginal delivery, or a combination of the two. The record of a CS is generated as the surgery is documented in the patient’s medical record. The MBR does not provide information about the stage of labor when the CS is performed, or reliable information about whether the CS was planned or acute, therefore both planned and acute CSs are included in this study. The onset of birth is classified as either spontaneous or iatrogenic (planned CS or induction of labor). The included cohorts were mothers with their first two pregnancies recorded in the MBR, and another analogous, largely overlapping cohort with mothers with their first three pregnancies recorded. We had access to data between 1973–2012. The type of onset of birth was not recorded before 1990, excluding earlier pregnancies. Registry issues leading to exclusion were absent or duplicated maternal ID, missing information of the gestational duration of the pregnancy, or unreliable registration of the gestational duration based on a comparison with the child’s birth weight. Further exclusions were made for stillbirth, multiple gestation, previous preterm birth and use of assisted reproduction techniques in the outcome pregnancy. The inclusion and exclusion criteria are presented in Figs and . The outcome was the gestational duration of the last pregnancy of the series. In the MBR, the best estimate of the gestational duration is based on ultrasound or the last menstrual period, according to a hierarchical set of rules . The majority of the pregnancies included in this study (86.9–88.4%) had their gestational duration estimated by ultrasound. The gestational duration was stratified based on the WHO classification of preterm birth: preterm <37+0, late preterm 34+0–36+6, moderate preterm 32+0–33+6, very preterm 28+0–31+6 and extremely preterm <28+0 gestational weeks . Additional strata were created for gestational weeks 37+0–37+6, 38+0–38+6, 41+0–41+6 and >41+6 . This allowed analyses of the association between previous delivery modes and delivery within a specific and clinically relevant gestational interval, compared to the reference interval of 39+0–40+6 weeks used in this study, and to explore the possibility of non-linear effects. Multivariable logistic regression analyses were conducted with different gestational duration intervals as outcome when analysing pairs of pregnancies, and preterm birth as outcome when analysing trios of pregnancies. For the regression analyses, only series of pregnancies where the last pregnancy had a spontaneous onset of birth were included. Statistical significance was set at p < 0.05, with a Bonferroni-corrected threshold of 0.0056 (0.05/9) for multiple testing of different gestational duration intervals. The regression models were adjusted for: maternal age, body mass index in early pregnancy, mother born outside Sweden, smoking, diabetes (including gestational diabetes), hypertensive disorder (including preeclampsia), baby’s sex, congenital malformation and small or large for gestational age. For the extremely preterm group, diabetes, hypertensive disorders and large for gestational age were not included due to too few cases. Small and large for gestational age were defined as birth weight deviating by more than two standard deviations from the mean for the gestational age according to Marsal, the classification system used in clinical practice in Sweden . Missing body mass index data (12% of records) was handled through mean imputation. The ICD codes used to define the diagnosis-based variables are provided in Supporting Information . Statistical analyses were performed with R version 4.1.2. The code is available on https://github.com/PerinatalLab/prevCS-PTD . The Swedish obstetrical policy is to recommend a planned CS in the third pregnancy after two previous CSs, while recommending a vaginal birth after one previous CS. This clinical practice has consequences when studying this research topic. Excluding women with an iatrogenic onset of birth in the outcome pregnancy would lead to an overrepresentation of women with delivery in early gestational weeks, since they are the ones who had a spontaneous onset of birth before their scheduled CS date. This inflates the odds ratios for preterm delivery after one previous CS, and in particular after multiple previous CSs. To reduce this bias, survival analyses were performed (Kaplan-Meier curves calculated and Cox regression analyses to adjust for the same covariates as in the logistic regression models) where mothers with an iatrogenic onset of the last delivery of the series were included, but censored at the time of birth. This allowed including these mothers, despite lacking knowledge about how long the gestational duration of those pregnancies would have been had there not been an obstetrical intervention ending the pregnancy. The censoring time marks the shortest possible gestational duration of the pregnancy if it would have had a spontaneous onset of birth. This design extracts the most possible information out of each pregnancy while allowing us to keep the original group sizes intact, thereby handling the problem of inflated odds ratios of spontaneous preterm birth for women with a previous CS. Ethical approval for this study was granted by the Regional Ethical Review Board in Gothenburg on July 29, 2013 (Dnr 576–13) and by the Swedish Ethical Review Authority on October 12, 2022 (Dnr 2022-05062-02). Participant consent was waived for this study.
Descriptive characteristics of the cohorts The cohort with two pregnancies consisted of 612 935 women (full cohort), of which 512 515 had a spontaneous onset of the second birth (subcohort) . The first delivery was by CS in 79 965 pregnancies (13.0%) in the full cohort and in 44 264 pregnancies (8.6%) in the subcohort of only spontaneous onsets of the second birth. The distribution of the outcome variable, the gestational duration of the subsequent pregnancy categorised, is presented in the lowermost part of . Analysing the subcohort, the proportions of deliveries within the different gestational age intervals differ between the groups: the previous-CS group has higher birth rates in earlier and later weeks . Other differences between the groups are higher prevalence of diabetes, hypertensive disorders, fetal malformations, growth alterations and baby boys in the previous-CS group. Moreover, the birth rate through CS in the second pregnancy after a CS is 26.4% compared to 1.9% after a vaginal delivery. The cohort with three pregnancies consisted of 157 581 women . The onset of the third birth has a big impact on the observed proportion of preterm births: for women with two previous CSs, the rate of preterm birth is 5.1% for the group as a whole, but 18.1% if only women with a spontaneous onset of the third birth are included. The Swedish obstetrical practice of recommending vaginal delivery after one previous CS is illustrated by the birth rates through CS being 26.4% for the second delivery after one CS, and 98.7% for the third delivery after two CSs. For two consecutive pregnancies, a previous CS shifted more weight to the extremes of the distribution of the gestational duration of the second pregnancy ( . For three consecutive pregnancies ( , the seemingly shorter gestational duration of the third pregnancy for the group with two previous CSs stands out. However, the graphs only include women with a spontaneous onset of the last birth and illustrate the bias discussed before (overinclusion of women with delivery in early gestational weeks), which happens when excluding mothers with an iatrogenic onset of the outcome pregnancy. This illustrates the difficulty in separating the true impacts of previous CSs on the duration of future pregnancies from technical artifacts. Logistic regression shows an increased risk of spontaneous preterm and postterm birth after previous cesarean section For the cohort with two consecutive pregnancies , previous CS was associated with a higher risk of spontaneous preterm and postterm birth, with larger effect sizes for intervals further away from term (logistic regression, all p -values highly significant, <0.001). The three-pregnancy cohort also shows an association between CS and subsequent spontaneous preterm birth. The risk of spontaneous preterm birth is however not higher in the third pregnancy if a mother had a CS in the first pregnancy followed by a vaginal birth in the second pregnancy, compared to mothers with two previous vaginal deliveries (aOR 1.15, 95% CI 0.96–1.36). This specific result stands out in relation to the rest of this study, where a CS consistently has an association with spontaneous preterm birth in a subsequent pregnancy. As a sensitivity analysis, regression models were created which only included pregnancies where the gestational duration was estimated by ultrasound. The results (Supporting Information and Figs) show no meaningful differences compared to the models relying on the best estimate of the gestational duration. Survival analyses show the link between cesarean section and prolonged pregnancy The exclusion of women with an iatrogenic onset of birth of the outcome pregnancy in a logistic regression model leads to falsely high odds ratios for delivery before a scheduled CS date. This problem is clearly seen with the unreasonably high odds ratios in (birth distribution if . The results in have the same problem (birth distribution in . Our solution to the bias generated by the unfortunate combination of the clinical management policy together with logistic regression is to use survival models. The Kaplan-Meier curves of show a prolonged pregnancy for the previous-CS group compared to the previous-vaginal delivery group. In , the analogous survival analysis for the cohort with three consecutive pregnancies shows the scenario of two previous CSs giving the longest prolongation of the third pregnancy, while the scenario of two previous vaginal deliveries corresponds to the shortest gestational duration of the third pregnancy. The scenario of a CS in the first delivery followed by vaginal delivery in the second pregnancy yet again shows a very slight difference versus two previous vaginal deliveries, consistent with the results of the logistic regression model that showed no statistically significant association with spontaneous preterm birth for that group. The Cox regression models calculate the adjusted hazard ratios (aHR) between the different scenarios of previous delivery modes using the same covariates as the previous logistic regression models. For two pregnancies , the aHR for previous CS versus previous vaginal delivery is 0.72 (95% CI 0.71–0.73), i.e., a net effect of a prolongation of the pregnancy following a CS. For three pregnancies , the largest prolongation of the third pregnancy is seen when comparing two previous CSs versus two previous vaginal deliveries (aHR 0.64, 95% CI 0.59–0.69). A very slight prolongation is seen when comparing the scenario of a CS followed by a vaginal delivery versus two previous vaginal deliveries (aHR 0.95, 95% CI 0.93–0.98).
The cohort with two pregnancies consisted of 612 935 women (full cohort), of which 512 515 had a spontaneous onset of the second birth (subcohort) . The first delivery was by CS in 79 965 pregnancies (13.0%) in the full cohort and in 44 264 pregnancies (8.6%) in the subcohort of only spontaneous onsets of the second birth. The distribution of the outcome variable, the gestational duration of the subsequent pregnancy categorised, is presented in the lowermost part of . Analysing the subcohort, the proportions of deliveries within the different gestational age intervals differ between the groups: the previous-CS group has higher birth rates in earlier and later weeks . Other differences between the groups are higher prevalence of diabetes, hypertensive disorders, fetal malformations, growth alterations and baby boys in the previous-CS group. Moreover, the birth rate through CS in the second pregnancy after a CS is 26.4% compared to 1.9% after a vaginal delivery. The cohort with three pregnancies consisted of 157 581 women . The onset of the third birth has a big impact on the observed proportion of preterm births: for women with two previous CSs, the rate of preterm birth is 5.1% for the group as a whole, but 18.1% if only women with a spontaneous onset of the third birth are included. The Swedish obstetrical practice of recommending vaginal delivery after one previous CS is illustrated by the birth rates through CS being 26.4% for the second delivery after one CS, and 98.7% for the third delivery after two CSs. For two consecutive pregnancies, a previous CS shifted more weight to the extremes of the distribution of the gestational duration of the second pregnancy ( . For three consecutive pregnancies ( , the seemingly shorter gestational duration of the third pregnancy for the group with two previous CSs stands out. However, the graphs only include women with a spontaneous onset of the last birth and illustrate the bias discussed before (overinclusion of women with delivery in early gestational weeks), which happens when excluding mothers with an iatrogenic onset of the outcome pregnancy. This illustrates the difficulty in separating the true impacts of previous CSs on the duration of future pregnancies from technical artifacts.
For the cohort with two consecutive pregnancies , previous CS was associated with a higher risk of spontaneous preterm and postterm birth, with larger effect sizes for intervals further away from term (logistic regression, all p -values highly significant, <0.001). The three-pregnancy cohort also shows an association between CS and subsequent spontaneous preterm birth. The risk of spontaneous preterm birth is however not higher in the third pregnancy if a mother had a CS in the first pregnancy followed by a vaginal birth in the second pregnancy, compared to mothers with two previous vaginal deliveries (aOR 1.15, 95% CI 0.96–1.36). This specific result stands out in relation to the rest of this study, where a CS consistently has an association with spontaneous preterm birth in a subsequent pregnancy. As a sensitivity analysis, regression models were created which only included pregnancies where the gestational duration was estimated by ultrasound. The results (Supporting Information and Figs) show no meaningful differences compared to the models relying on the best estimate of the gestational duration.
The exclusion of women with an iatrogenic onset of birth of the outcome pregnancy in a logistic regression model leads to falsely high odds ratios for delivery before a scheduled CS date. This problem is clearly seen with the unreasonably high odds ratios in (birth distribution if . The results in have the same problem (birth distribution in . Our solution to the bias generated by the unfortunate combination of the clinical management policy together with logistic regression is to use survival models. The Kaplan-Meier curves of show a prolonged pregnancy for the previous-CS group compared to the previous-vaginal delivery group. In , the analogous survival analysis for the cohort with three consecutive pregnancies shows the scenario of two previous CSs giving the longest prolongation of the third pregnancy, while the scenario of two previous vaginal deliveries corresponds to the shortest gestational duration of the third pregnancy. The scenario of a CS in the first delivery followed by vaginal delivery in the second pregnancy yet again shows a very slight difference versus two previous vaginal deliveries, consistent with the results of the logistic regression model that showed no statistically significant association with spontaneous preterm birth for that group. The Cox regression models calculate the adjusted hazard ratios (aHR) between the different scenarios of previous delivery modes using the same covariates as the previous logistic regression models. For two pregnancies , the aHR for previous CS versus previous vaginal delivery is 0.72 (95% CI 0.71–0.73), i.e., a net effect of a prolongation of the pregnancy following a CS. For three pregnancies , the largest prolongation of the third pregnancy is seen when comparing two previous CSs versus two previous vaginal deliveries (aHR 0.64, 95% CI 0.59–0.69). A very slight prolongation is seen when comparing the scenario of a CS followed by a vaginal delivery versus two previous vaginal deliveries (aHR 0.95, 95% CI 0.93–0.98).
Main findings Using survival models, the main result of this study is that there is an association between a previous CS and a prolongation of the subsequent pregnancy. While logistic regression shows an increased risk for both spontaneous preterm birth (aOR 1.67, 95% CI 1.57–1.77) and spontaneous postterm birth (aOR 1.55, 95% CI 1.49–1.62) in the second pregnancy after a CS in the first pregnancy, this is not the preferred way of analysing this research question since it overestimates the risk of preterm birth in the second pregnancy. Further, a CS in the first pregnancy followed by a vaginal birth in the second pregnancy is not associated with an increased risk for spontaneous preterm birth in the third pregnancy, compared to two previous vaginal deliveries (aOR 1.15, 95% CI 0.96–1.36). Previous studies have described the association between CS and subsequent preterm birth, but the main novel finding of this study is the association between CS and subsequent prolonged pregnancy, with an extensive methodological elaboration on the importance of taking the type of onset of birth in consideration when studying this subject. Strengths and limitations A strength of this study is the reliability and the level of detail of the included data . A benefit of performing this study on Swedish patient data is the relatively high rate of vaginal birth after CS in Sweden (73.6%) compared to other countries . Even if we have used a survival model to adjust for the bias due to a larger proportion of iatrogenic onsets of birth in the exposure group, the robustness of the results is helped by the very high rate of vaginal birth after one CS in Sweden, since it reduces the uncertainty of the results due to a lesser need for the model and its impacts on the observed data. The study covers a large time period and is nationwide, yet the clinical practice was similar over the entire country during the whole time frame of the study. The benefit of performing this study before the CS rate has increased even further that today is that it keeps problems generated by management differences between the exposure and reference groups as small as possible. A reasonable assumption is that Sweden will follow the global trend with an increased CS rate, and therefore it is likely that the use of advanced statistical maneuvers will become even more important, and unfortunately create even more interpretation problems than seen today. There is, despite our best efforts to minimize it, still a risk of residual confounding that might have affected the observed results. The most significant risk we have identified is the possibility of confounding by indication, because even if we have tried to identify a reasonably homogenous group of women for inclusion and then adjust statistically for covariates differing between the groups, there will on the individual level always be a reason why the specific delivery mode was chosen, generating a difference between the exposure and reference groups that is unlikely to be fully resolved in any observational study design. A large portion of the potential bias that could be related to this phenomenon is, in our opinion, avoided through the inclusion of relatively healthy previous pregnancies, with the most critical step being the decision to only include previous-term pregnancies. Another limitation is the possibility that the outcome of previous pregnancies may affect how likely the woman is to have another pregnancy, which would alter the results of studies such as this, a limitation that cannot be resolved within a birth cohort. Other limitations are the lack of information in the MBR about the indication for the CS and at which stage of labor it was performed. Interpretation The worldwide CS rate is expected to keep increasing, leading inevitably to an increase in the negative consequences of CSs. The aim of this study was to investigate the link between previous delivery modes and the incidence of deviant duration of subsequent pregnancies. While at first glance, our results seem coherent with previous studies, the survival analysis of this study changes that interpretation. Including only mothers with a spontaneous onset of the subsequent birth seems reasonable since we are mainly interested in any biological effects of a previous CS on its gestational duration. However, such selection generates a bias because the group of women with a previous CS are more likely to have an iatrogenic onset of the next pregnancy (45% vs 12%, ), and therefore, spontaneous births will per definition occur before the date of an iatrogenic onset. Excluding women with an iatrogenic onset of birth, while keeping those with a spontaneous onset of birth, will exclude a relatively larger fraction of women from the previous-CS group, making that group seem smaller than it really is, and with a heavy burden of preterm birth, overestimating the risk of preterm birth for women with a previous CS. The apparent effect of a previous CS leading to increased preterm birth risk will largely be due to the study design, as the increased risk of spontaneous preterm birth after previous CS is no longer observed when applying survival models. In these, mothers with an iatrogenic onset of birth in the outcome pregnancy are kept; in effect, they are “followed-up” until the time of their CS. While we do not know their “natural” delivery date, we know that most of these mothers delivered around term, i.e. they had a low risk of spontaneous preterm birth. By including these mothers, the survival analysis shows that previous CS does not increase the risk of preterm birth. Thus, retaining the full group, i.e. not excluding the mothers with an iatrogenic onset of the next delivery, is essential. This result implies that the increased risk observed with logistic regression in this and other studies is likely attributable to a technical artefact rather than a true consequence of previous CS. In fact, with appropriately chosen analysis models, we instead see an association in the other direction–between a previous CS and a prolonged subsequent pregnancy. To our knowledge, this association has not been reported before, as previous research has focused on the increased risk of preterm birth subsequent to CS . A possible explanation to this phenomenon could be that the physiological alteration of uterine function resulting from a CS may exacerbate uterine intolerance or, in the case of prolonged pregnancies, diminish the sensitivity of uterine load . One speculation is that the CS might cause a stiffness in the uterus during the healing process, with an ensuing problem in failing to recognize the signals of onset of labor. This might be due to either the scar itself having poor contractile properties, or the failure of the area to allow effective signaling between the body and the cervix of the uterus, altering the maturation of the cervix and the onset of birth mechanism. Another interesting finding seen in the cohort with three pregnancies was that a vaginal delivery in the second pregnancy after a CS in the first pregnancy proved to be no different in terms of association with subsequent spontaneous preterm birth in the third pregnancy compared to having had two previous vaginal deliveries. It seems that a vaginal birth after a CS could possibly be considered “a stress test” of uterine function in terms of achieving vaginal delivery, selecting a subgroup of women with lower risk of developing complications due to their previous CS, perhaps due to a smaller or more fortunately placed incision. Regarding differences between the previous CS- and previous vaginal delivery groups, the overrepresentation of baby boys in the CS group is noteworthy. This discrepancy might possibly be explained by the fact that baby boys seem to be at generally higher risk of fetal distress . The higher prevalence of SGA, LGA, preeclampsia, and diabetes in the CS group are expected, as these factors are known to increase the likelihood of developing fetal distress . Having taken all this into account, the differences between the cohorts should be considered to be minor and adjusted for. Since previous studies have seen increased risk for subsequent spontaneous preterm births after late-stage CSs , a subanalysis on late-stage CSs in our cohort might have added valuable information, particularly if late-stage CS also increases the risk of subsequent delayed onset of birth. For future studies, investigating if there is any evidence of alterations in uterine function after previous surgery unrelated to the scar itself, would be of interest. That could possibly be grounds for re-evaluating the results of this study.
Using survival models, the main result of this study is that there is an association between a previous CS and a prolongation of the subsequent pregnancy. While logistic regression shows an increased risk for both spontaneous preterm birth (aOR 1.67, 95% CI 1.57–1.77) and spontaneous postterm birth (aOR 1.55, 95% CI 1.49–1.62) in the second pregnancy after a CS in the first pregnancy, this is not the preferred way of analysing this research question since it overestimates the risk of preterm birth in the second pregnancy. Further, a CS in the first pregnancy followed by a vaginal birth in the second pregnancy is not associated with an increased risk for spontaneous preterm birth in the third pregnancy, compared to two previous vaginal deliveries (aOR 1.15, 95% CI 0.96–1.36). Previous studies have described the association between CS and subsequent preterm birth, but the main novel finding of this study is the association between CS and subsequent prolonged pregnancy, with an extensive methodological elaboration on the importance of taking the type of onset of birth in consideration when studying this subject.
A strength of this study is the reliability and the level of detail of the included data . A benefit of performing this study on Swedish patient data is the relatively high rate of vaginal birth after CS in Sweden (73.6%) compared to other countries . Even if we have used a survival model to adjust for the bias due to a larger proportion of iatrogenic onsets of birth in the exposure group, the robustness of the results is helped by the very high rate of vaginal birth after one CS in Sweden, since it reduces the uncertainty of the results due to a lesser need for the model and its impacts on the observed data. The study covers a large time period and is nationwide, yet the clinical practice was similar over the entire country during the whole time frame of the study. The benefit of performing this study before the CS rate has increased even further that today is that it keeps problems generated by management differences between the exposure and reference groups as small as possible. A reasonable assumption is that Sweden will follow the global trend with an increased CS rate, and therefore it is likely that the use of advanced statistical maneuvers will become even more important, and unfortunately create even more interpretation problems than seen today. There is, despite our best efforts to minimize it, still a risk of residual confounding that might have affected the observed results. The most significant risk we have identified is the possibility of confounding by indication, because even if we have tried to identify a reasonably homogenous group of women for inclusion and then adjust statistically for covariates differing between the groups, there will on the individual level always be a reason why the specific delivery mode was chosen, generating a difference between the exposure and reference groups that is unlikely to be fully resolved in any observational study design. A large portion of the potential bias that could be related to this phenomenon is, in our opinion, avoided through the inclusion of relatively healthy previous pregnancies, with the most critical step being the decision to only include previous-term pregnancies. Another limitation is the possibility that the outcome of previous pregnancies may affect how likely the woman is to have another pregnancy, which would alter the results of studies such as this, a limitation that cannot be resolved within a birth cohort. Other limitations are the lack of information in the MBR about the indication for the CS and at which stage of labor it was performed.
The worldwide CS rate is expected to keep increasing, leading inevitably to an increase in the negative consequences of CSs. The aim of this study was to investigate the link between previous delivery modes and the incidence of deviant duration of subsequent pregnancies. While at first glance, our results seem coherent with previous studies, the survival analysis of this study changes that interpretation. Including only mothers with a spontaneous onset of the subsequent birth seems reasonable since we are mainly interested in any biological effects of a previous CS on its gestational duration. However, such selection generates a bias because the group of women with a previous CS are more likely to have an iatrogenic onset of the next pregnancy (45% vs 12%, ), and therefore, spontaneous births will per definition occur before the date of an iatrogenic onset. Excluding women with an iatrogenic onset of birth, while keeping those with a spontaneous onset of birth, will exclude a relatively larger fraction of women from the previous-CS group, making that group seem smaller than it really is, and with a heavy burden of preterm birth, overestimating the risk of preterm birth for women with a previous CS. The apparent effect of a previous CS leading to increased preterm birth risk will largely be due to the study design, as the increased risk of spontaneous preterm birth after previous CS is no longer observed when applying survival models. In these, mothers with an iatrogenic onset of birth in the outcome pregnancy are kept; in effect, they are “followed-up” until the time of their CS. While we do not know their “natural” delivery date, we know that most of these mothers delivered around term, i.e. they had a low risk of spontaneous preterm birth. By including these mothers, the survival analysis shows that previous CS does not increase the risk of preterm birth. Thus, retaining the full group, i.e. not excluding the mothers with an iatrogenic onset of the next delivery, is essential. This result implies that the increased risk observed with logistic regression in this and other studies is likely attributable to a technical artefact rather than a true consequence of previous CS. In fact, with appropriately chosen analysis models, we instead see an association in the other direction–between a previous CS and a prolonged subsequent pregnancy. To our knowledge, this association has not been reported before, as previous research has focused on the increased risk of preterm birth subsequent to CS . A possible explanation to this phenomenon could be that the physiological alteration of uterine function resulting from a CS may exacerbate uterine intolerance or, in the case of prolonged pregnancies, diminish the sensitivity of uterine load . One speculation is that the CS might cause a stiffness in the uterus during the healing process, with an ensuing problem in failing to recognize the signals of onset of labor. This might be due to either the scar itself having poor contractile properties, or the failure of the area to allow effective signaling between the body and the cervix of the uterus, altering the maturation of the cervix and the onset of birth mechanism. Another interesting finding seen in the cohort with three pregnancies was that a vaginal delivery in the second pregnancy after a CS in the first pregnancy proved to be no different in terms of association with subsequent spontaneous preterm birth in the third pregnancy compared to having had two previous vaginal deliveries. It seems that a vaginal birth after a CS could possibly be considered “a stress test” of uterine function in terms of achieving vaginal delivery, selecting a subgroup of women with lower risk of developing complications due to their previous CS, perhaps due to a smaller or more fortunately placed incision. Regarding differences between the previous CS- and previous vaginal delivery groups, the overrepresentation of baby boys in the CS group is noteworthy. This discrepancy might possibly be explained by the fact that baby boys seem to be at generally higher risk of fetal distress . The higher prevalence of SGA, LGA, preeclampsia, and diabetes in the CS group are expected, as these factors are known to increase the likelihood of developing fetal distress . Having taken all this into account, the differences between the cohorts should be considered to be minor and adjusted for. Since previous studies have seen increased risk for subsequent spontaneous preterm births after late-stage CSs , a subanalysis on late-stage CSs in our cohort might have added valuable information, particularly if late-stage CS also increases the risk of subsequent delayed onset of birth. For future studies, investigating if there is any evidence of alterations in uterine function after previous surgery unrelated to the scar itself, would be of interest. That could possibly be grounds for re-evaluating the results of this study.
The issue if there is a relationship between delivery through CS and an increased risk of subsequent preterm delivery is important, since that would be a severe consequence of a medical intervention that is potentially avoidable. Through a new methodological approach, we suggest survival models as the preferred way of addressing this research question, as this provides a tool for handling the large bias generated by a high number of iatrogenic onset of births within the group of women having undergone a previous CS. This approach shows that the dominant association between a CS and the duration of a subsequent pregnancy is the prolonging of it, which contrasts with previous research on this subject. This highlights the importance of choosing appropriate models for complex perinatal outcomes. While the new findings add insights about the relationship between CS and the gestational duration of subsequent pregnancies, the question if a causal relationship between the two exists is yet to be proved.
S1 Table ICD-codes for variable definitions. ICD: International Classification of Diseases. (DOCX) S1 Fig Multivariable logistic regression analyses, first two pregnancies. Adjusted odds ratio for gestational age at birth within the specified interval, in the second pregnancy after cesarean section in the first pregnancy. Only mothers with spontaneous onset of birth and gestational duration of the second pregnancy estimated by ultrasound. (TIF) S2 Fig Multivariable logistic regression analyses, first three pregnancies. Adjusted odds ratio for preterm birth in the third pregnancy for the specified comparison between different sequences of previous delivery modes. Only mothers with spontaneous onset of birth and gestational duration of the third pregnancy estimated by ultrasound. CS : cesarean section, Vag : vaginal delivery. (TIF)
|
Process failure mode and effects analysis for external beam radiotherapy: Introducing a literature-based template and a novel action priority | f2a01c2c-83de-471b-ac96-2b1cdf83ec88 | 11384953 | Internal Medicine[mh] | Introduction Process failure mode and effects analysis (PFMEA) is a systematic method for identifying and managing failure chains. To this end, experienced professionals study processes and their potential failures, independent of their actual occurrence . As an easy-to-implement and cost-efficient method, PFMEA has been widely adopted across radiotherapy departments, e. g., in the U. S. , Brazil , Germany , Italy and Spain . However, practical challenges somewhat impede the acceptance of risk assessments. For instance, a recent German survey disclosed national deficits in risk management knowledge: 80% of the responding institutions evaluated their knowledge as ‘satisfying’ or worse . In order to increase the acceptance of risk assessments, this study proposes a PFMEA template. In particular, the template was compiled for a general external beam radiation therapy (EBRT) process. Various ways of performing FMEA have been described in literature to suit the needs of different industries. Here, we used the PFMEA approach as described in the FMEA handbook of the Automotive Industry Action Group (AIAG) which has, to the best of our knowledge, not yet been applied to radiotherapy. This approach brings about a crucial innovation in failure mode prioritization, namely the substitution of the traditional risk priority number (RPN) with the action priority (AIAG AP). Both prioritization methods utilize the severity S of the failure effect, the occurrence O of the failure causes and the detection D of the failure causes or failure modes. Whereas the RPN is the simple product of S , O , and D , the AIAG AP is a three-dimensional look-up table that evaluates these parameters individually. This way the AIAG AP gives more weight to severity first, then occurrence, then detection . For each of the 1000 possible S-O-D combinations (given 10 steps per parameter), individual action levels can be looked up. In total, there are three different action levels—high (H), medium (M), and low (L)—which imply whether actions must, should , or can be implemented, respectively . Its main intent is failure prevention, as measures reducing the severity or the occurrence lead to a greater reduction of the AP. On the other hand, RPNs give equal weight to S , O , and D and therefore, there is no preferred mitigation strategy (reducing severity and increasing detection are of the same value). Since the AIAG AP table was designed to work with the S - O - D rating systems provided in the FMEA handbook, it should be reviewed before using different rating systems . An example rating system which is used in our department is given in . Instead of reviewing all 1000 cells of the AIAG AP table, we propose another method that mimics the AIAG AP but which can more easily be adapted to our or any other rating system. In the following, this method is called AP for radiation oncology (RO AP). The purpose of this study was thus twofold: Firstly, an EBRT-PFMEA template was created that can be used as the basis for risk assessments across different radiation oncology institutions. The second purpose was to introduce the RO AP as an alternative to the AIAG AP, for example, to rate the failure modes of the EBRT-PFMEA template. The RO AP was benchmarked against the AIAG AP and also against the RPN to investigate the alleged superiority of the AP over the RPN. This benchmark employed Monte Carlo simulations. These two purposes were followed independently, i.e., the template did not affect the simulations and the RO AP was not used to rate the failure modes of the template. Methods and materials 2.1 EBRT-PFMEA template A Microsoft Excel PFMEA work sheet was used to compile the template PFMEA (Form F in the FMEA handbook ). The AIAG approach involved seven steps: 1. Planning and preparation to set the project scope, 2. structure analysis to describe the process flow, 3. function analysis to describe the functions of the structures, 4. failure analysis to deduce failure chains consisting of a failure effect, failure mode, and failure cause, 5. risk analysis to evaluate failure chains considering prevention or detection controls, 6. optimization to identify further controls, 7. results documentation to communicate the conclusions of the PFMEA . We conducted steps 2 to 6 as described hereafter. 2.1.1 Structure analysis During the structure analysis, the EBRT process was described by breaking it down into so-called process steps, process items, and process work elements . The process step is the focus of the analysis and, here, referred to an operation that the patient or the patient’s file passes through. The process item is the result of the process step and process work elements are the needed resources. Here, these were identified by means of the 4M (man, machine, method, and material) as commonly known from the Ishikawa approach . Let physical treatment planning be an exemplary process step. A DICOM RT Plan file can be treated as the physical result and therefore as a process item. To conduct the planning, one would typically need dosimetrists/physicists (man), computers (machine), dose algorithms (method), and beam basic data (material). We used the general EBRT process published in the AAPM’s consensus recommendations for incident learning systems as the basis for the structure analysis. Missing information considered standard of care in today’s practice have been added to the best of our knowledge. 2.1.2 Function analysis During the function analysis, functions were given to the process structures identified before. Functions describe intents and/or requirements and form the functional relationships between the process structures (what is being done and how is it achieved). The requirements are quantifiable features and can be measured or judged. Using the physical planning example from above, one typical function is the dosimetrist selecting the correct beam data, dose algorithm, clinical goals, and dose constraints. Another function is the computer calculating the monitor units. Requirements for the plan, for example, are dose-volume-histogram (DVH) parameters. 2.1.3 Failure analysis In the subsequent failure analysis, all previously identified functions of the process items, process steps, and process work elements were negated and thus became the failure effects, failure modes, and failure causes, respectively. Here, failure effects were described on the process item level as well as the patient level. In the above example, if the dosimetrist selects the wrong dose algorithm (cause), then doses to organs-at-risk might be underestimated but, in reality, exceed limits (mode). In effect, DVH parameters might actually be not fulfilled (process item level) which might result in toxicities (patient level). Negating all process step functions in all possible ways would yield all potential failure modes. However, only those associated with failure effects affecting the patient’s safety were relevant here. To evaluate what is potentially wrong with the process , we assumed that all process steps are carried out, meaning that errors due to omission were not considered. In an attempt to identify this subset of failure modes we incorporated an extensive literature study. On PubMed.gov (National Library of Medicine, Bethesda, MD, U.S.), the following keywords were used: ‘FMEA’, ‘risk analysis’, ‘risk assessment’ and ‘radiation therapy’. 126 publications were found at the time of this study, of which 43 were relevant for linac-based EBRT. Actual information on failure modes were provided in 33 publications: , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , . To keep the EBRT-PFMEA template manageable as a practical starting point, up to ten of the highest ranked failure modes per study were reviewed. Thereafter, incident reports were analyzed to deduce failure modes that actually occurred. To this end, the WHO’s Radiotherapy Risk Profile was used that provided a general list of 48 high-impact errors . Furthermore, 16 ASTRO’s/AAPM’s RO-ILS quarterly reports were used that briefly summarized 7968 incidents and reviewed 69 cases of recurring themes , , , , , , , , , , , , , , , . Lastly, two annual reports of the German reporting and information system for significant events related to radiation exposures in medicine , were considered that summarized 100 events relevant for RT. Failure modes retrieved from the literature survey have been rephrased to match the negation of a process step function. The same principle was applied to failure causes and failure effects. Our literature-informed approach directed the focus of the study on high-risk as well as observed failure modes. 2.1.4 Risk analysis In the subsequent risk analysis, prevention and detection controls that were already integrated into the process were identified. This was done by investigating which functions further upstream or downstream in the process were potentially capable of preventing or detecting failure causes or failure modes. Ratings for S , O , and D were not given as they will be, due to being rather subjective, different for each institution and risk assessment team. 2.1.5 Optimization In the optimization section, further potential prevention and detection controls were proposed if the process itself had no adequate control strategies. They were adopted from the literature survey when given. 2.2 Radiation oncology action priority 2.2.1 Definition The RO AP was defined by means of a weighted sum: R O A P y = very high if y ≥ y vh high if y ≥ y h medium if y ≥ y m low if y < y m o r S = 1 o r O = 1 with y = w S S + w O O + w D D , and w S , w O , and w D being weighting factors that satisfied w S ≥ w O ≥ w D as well as w S + w O + w D = 3 . This way, the weighted sum y never exceeded 30, which is the maximum number one can achieve using a ten-step rating system. A fourth action level, very high (VH), was introduced to distinguish actions that must be implemented immediately. 2.2.2 Benchmark Monte Carlo simulations were performed in order to compare the number of needed actions as required by the AIAG AP, RO AP, and the RPN. We also compared whether the three methods would lead to comparable optimized states. This comparison required establishing the near-same conditions for the RO AP and RPN which is described hereafter. 2.2.2.1 Conditioning the RO AP The threshold values y VH , y H , and y M had to be determined first in order to create RO AP tables that have nearly the same number of high, medium and low combinations as the AIAG AP table. Of the 1000 S-O-D combinations in the AIAG AP table, 318, 214 and 468 combinations refer to H, M, and L, respectively. We identified general threshold values iteratively by using 10,000 random sets of weighting factors { w S , w O , w D } 1 , …, { w S , w O , w D } 10,000 to generate 10,000 random RO AP tables. The threshold values were adjusted until nearly the same numbers of H, M, and L combinations were obtained on average. For that purpose, very high was treated the same as high. 2.2.2.2 Conditioning the RPN To be able to compare the RPN with the AIAG AP, RPN action levels must be established that mimic the purpose of the AP action levels. We referred to the RPN action levels provided by the BFS, DEGRO, DGMP, and DGN . To set up the analogy, RPN values greater than 125 were treated the same as H ratings, and RPN values between 125 and 30 were treated as M ratings. However, these values were derived from a different S-O-D rating system as the one used in the FMEA handbook. Therefore, S-O-D numbers of the FMEA handbook were mapped to the respective criteria of the joint recommendations. For example, the criterion ‘more than once per day’ is O = 9 or O = 10 in the joint recommendations and O = 7 in the FMEA handbook. After the mapping process (see s for details), the new RPN action levels were 110 (H) and 27 (M). 2.2.2.3 Monte Carlo simulation The Monte Carlo simulations were performed for the AIAG AP as illustrated in . The RO AP and RPN simulations were processed analogously with their respective action levels (see , ). The start of the simulations was to draw 10,000 random failure mode ratings, i.e., S-O-D combinations. To obtain more realistic ratings, the S-O-D combinations were drawn from the probability density functions (PDFs) shown in . These PDFs were calculated from 216 S-O-D ratings provided by AAPM TG-100 using a Gaussian kernel density estimator and Silverman’s rule of thumb as bandwidth estimator. In the AIAG AP track, the AIAG APs of the 10,000 random S-O-D ratings were deduced. This determined the sequence of optimization. Then, during the actual optimization, random prevention and detection ‘actions’ were drawn uniformly. This means that either the O or D rating was randomly improved but never both ratings at once. Actions were continuously drawn until the optimized AIAG AP was either M or L. In case of an (optimized) M rating, (further) actions were drawn until a maximum number of four (this number was considered as sufficient defense in depth by IAEA-TECDOC-1685 ). Failure mode ratings were considered acceptable once there was either sufficient defense in depth or an L rating. In the RO AP track, the simulation was repeatedly conducted for 40 random sets of weighting factors { w S , w O , w D } 1 , …, { w S , w O , w D } 40 . This would reveal whether there was a relationship between these and the number of needed actions. Moreover, if a set of weighting factors existed that could replicate the AIAG AP optimization, it could then be determined by interpolation. Once all failure modes were accepted in each track, the average numbers of actions per failure mode could be determined. Additionally, we quantified the achieved optimized states by means of the residual total criticality number ΣSO as a surrogate. The criticality number SO is the product of severity and occurrence and directs the focus on reducing the occurrence of failure modes . Finally, we parametrized the RO AP with the found sets of threshold values and weighting factors to reproduce the AIAG AP. Then, the median RPN values of the four action levels were determined and compared, using the Kruskal-Wallis H test, a non-parametric analysis of variance. 2.2.3 Adaption The general threshold values found in were substituted with appropriate threshold values to calibrate the RO AP for the S-O-D rating system given in . These were chosen manually so that the following intentions were satisfied: If failure modes resulted in mild side effects at worst ( S ≤ 4) and occurred once a month or less ( O ≤ 4), then an acceptable state (L) should easily be achievable because the benefit of the treatment far surpasses the risks. If S ≤ 6 (up to mild toxicity), then depending on the occurrence, actions should or must be taken (M/H) because toxicities, in general, should be avoided. If S ≥ 8 (up to premature death), actions must immediately be taken (VH) because the benefit of the treatment is greatly reduced. EBRT-PFMEA template A Microsoft Excel PFMEA work sheet was used to compile the template PFMEA (Form F in the FMEA handbook ). The AIAG approach involved seven steps: 1. Planning and preparation to set the project scope, 2. structure analysis to describe the process flow, 3. function analysis to describe the functions of the structures, 4. failure analysis to deduce failure chains consisting of a failure effect, failure mode, and failure cause, 5. risk analysis to evaluate failure chains considering prevention or detection controls, 6. optimization to identify further controls, 7. results documentation to communicate the conclusions of the PFMEA . We conducted steps 2 to 6 as described hereafter. 2.1.1 Structure analysis During the structure analysis, the EBRT process was described by breaking it down into so-called process steps, process items, and process work elements . The process step is the focus of the analysis and, here, referred to an operation that the patient or the patient’s file passes through. The process item is the result of the process step and process work elements are the needed resources. Here, these were identified by means of the 4M (man, machine, method, and material) as commonly known from the Ishikawa approach . Let physical treatment planning be an exemplary process step. A DICOM RT Plan file can be treated as the physical result and therefore as a process item. To conduct the planning, one would typically need dosimetrists/physicists (man), computers (machine), dose algorithms (method), and beam basic data (material). We used the general EBRT process published in the AAPM’s consensus recommendations for incident learning systems as the basis for the structure analysis. Missing information considered standard of care in today’s practice have been added to the best of our knowledge. 2.1.2 Function analysis During the function analysis, functions were given to the process structures identified before. Functions describe intents and/or requirements and form the functional relationships between the process structures (what is being done and how is it achieved). The requirements are quantifiable features and can be measured or judged. Using the physical planning example from above, one typical function is the dosimetrist selecting the correct beam data, dose algorithm, clinical goals, and dose constraints. Another function is the computer calculating the monitor units. Requirements for the plan, for example, are dose-volume-histogram (DVH) parameters. 2.1.3 Failure analysis In the subsequent failure analysis, all previously identified functions of the process items, process steps, and process work elements were negated and thus became the failure effects, failure modes, and failure causes, respectively. Here, failure effects were described on the process item level as well as the patient level. In the above example, if the dosimetrist selects the wrong dose algorithm (cause), then doses to organs-at-risk might be underestimated but, in reality, exceed limits (mode). In effect, DVH parameters might actually be not fulfilled (process item level) which might result in toxicities (patient level). Negating all process step functions in all possible ways would yield all potential failure modes. However, only those associated with failure effects affecting the patient’s safety were relevant here. To evaluate what is potentially wrong with the process , we assumed that all process steps are carried out, meaning that errors due to omission were not considered. In an attempt to identify this subset of failure modes we incorporated an extensive literature study. On PubMed.gov (National Library of Medicine, Bethesda, MD, U.S.), the following keywords were used: ‘FMEA’, ‘risk analysis’, ‘risk assessment’ and ‘radiation therapy’. 126 publications were found at the time of this study, of which 43 were relevant for linac-based EBRT. Actual information on failure modes were provided in 33 publications: , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , . To keep the EBRT-PFMEA template manageable as a practical starting point, up to ten of the highest ranked failure modes per study were reviewed. Thereafter, incident reports were analyzed to deduce failure modes that actually occurred. To this end, the WHO’s Radiotherapy Risk Profile was used that provided a general list of 48 high-impact errors . Furthermore, 16 ASTRO’s/AAPM’s RO-ILS quarterly reports were used that briefly summarized 7968 incidents and reviewed 69 cases of recurring themes , , , , , , , , , , , , , , , . Lastly, two annual reports of the German reporting and information system for significant events related to radiation exposures in medicine , were considered that summarized 100 events relevant for RT. Failure modes retrieved from the literature survey have been rephrased to match the negation of a process step function. The same principle was applied to failure causes and failure effects. Our literature-informed approach directed the focus of the study on high-risk as well as observed failure modes. 2.1.4 Risk analysis In the subsequent risk analysis, prevention and detection controls that were already integrated into the process were identified. This was done by investigating which functions further upstream or downstream in the process were potentially capable of preventing or detecting failure causes or failure modes. Ratings for S , O , and D were not given as they will be, due to being rather subjective, different for each institution and risk assessment team. 2.1.5 Optimization In the optimization section, further potential prevention and detection controls were proposed if the process itself had no adequate control strategies. They were adopted from the literature survey when given. Structure analysis During the structure analysis, the EBRT process was described by breaking it down into so-called process steps, process items, and process work elements . The process step is the focus of the analysis and, here, referred to an operation that the patient or the patient’s file passes through. The process item is the result of the process step and process work elements are the needed resources. Here, these were identified by means of the 4M (man, machine, method, and material) as commonly known from the Ishikawa approach . Let physical treatment planning be an exemplary process step. A DICOM RT Plan file can be treated as the physical result and therefore as a process item. To conduct the planning, one would typically need dosimetrists/physicists (man), computers (machine), dose algorithms (method), and beam basic data (material). We used the general EBRT process published in the AAPM’s consensus recommendations for incident learning systems as the basis for the structure analysis. Missing information considered standard of care in today’s practice have been added to the best of our knowledge. Function analysis During the function analysis, functions were given to the process structures identified before. Functions describe intents and/or requirements and form the functional relationships between the process structures (what is being done and how is it achieved). The requirements are quantifiable features and can be measured or judged. Using the physical planning example from above, one typical function is the dosimetrist selecting the correct beam data, dose algorithm, clinical goals, and dose constraints. Another function is the computer calculating the monitor units. Requirements for the plan, for example, are dose-volume-histogram (DVH) parameters. Failure analysis In the subsequent failure analysis, all previously identified functions of the process items, process steps, and process work elements were negated and thus became the failure effects, failure modes, and failure causes, respectively. Here, failure effects were described on the process item level as well as the patient level. In the above example, if the dosimetrist selects the wrong dose algorithm (cause), then doses to organs-at-risk might be underestimated but, in reality, exceed limits (mode). In effect, DVH parameters might actually be not fulfilled (process item level) which might result in toxicities (patient level). Negating all process step functions in all possible ways would yield all potential failure modes. However, only those associated with failure effects affecting the patient’s safety were relevant here. To evaluate what is potentially wrong with the process , we assumed that all process steps are carried out, meaning that errors due to omission were not considered. In an attempt to identify this subset of failure modes we incorporated an extensive literature study. On PubMed.gov (National Library of Medicine, Bethesda, MD, U.S.), the following keywords were used: ‘FMEA’, ‘risk analysis’, ‘risk assessment’ and ‘radiation therapy’. 126 publications were found at the time of this study, of which 43 were relevant for linac-based EBRT. Actual information on failure modes were provided in 33 publications: , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , . To keep the EBRT-PFMEA template manageable as a practical starting point, up to ten of the highest ranked failure modes per study were reviewed. Thereafter, incident reports were analyzed to deduce failure modes that actually occurred. To this end, the WHO’s Radiotherapy Risk Profile was used that provided a general list of 48 high-impact errors . Furthermore, 16 ASTRO’s/AAPM’s RO-ILS quarterly reports were used that briefly summarized 7968 incidents and reviewed 69 cases of recurring themes , , , , , , , , , , , , , , , . Lastly, two annual reports of the German reporting and information system for significant events related to radiation exposures in medicine , were considered that summarized 100 events relevant for RT. Failure modes retrieved from the literature survey have been rephrased to match the negation of a process step function. The same principle was applied to failure causes and failure effects. Our literature-informed approach directed the focus of the study on high-risk as well as observed failure modes. Risk analysis In the subsequent risk analysis, prevention and detection controls that were already integrated into the process were identified. This was done by investigating which functions further upstream or downstream in the process were potentially capable of preventing or detecting failure causes or failure modes. Ratings for S , O , and D were not given as they will be, due to being rather subjective, different for each institution and risk assessment team. Optimization In the optimization section, further potential prevention and detection controls were proposed if the process itself had no adequate control strategies. They were adopted from the literature survey when given. Radiation oncology action priority 2.2.1 Definition The RO AP was defined by means of a weighted sum: R O A P y = very high if y ≥ y vh high if y ≥ y h medium if y ≥ y m low if y < y m o r S = 1 o r O = 1 with y = w S S + w O O + w D D , and w S , w O , and w D being weighting factors that satisfied w S ≥ w O ≥ w D as well as w S + w O + w D = 3 . This way, the weighted sum y never exceeded 30, which is the maximum number one can achieve using a ten-step rating system. A fourth action level, very high (VH), was introduced to distinguish actions that must be implemented immediately. 2.2.2 Benchmark Monte Carlo simulations were performed in order to compare the number of needed actions as required by the AIAG AP, RO AP, and the RPN. We also compared whether the three methods would lead to comparable optimized states. This comparison required establishing the near-same conditions for the RO AP and RPN which is described hereafter. 2.2.2.1 Conditioning the RO AP The threshold values y VH , y H , and y M had to be determined first in order to create RO AP tables that have nearly the same number of high, medium and low combinations as the AIAG AP table. Of the 1000 S-O-D combinations in the AIAG AP table, 318, 214 and 468 combinations refer to H, M, and L, respectively. We identified general threshold values iteratively by using 10,000 random sets of weighting factors { w S , w O , w D } 1 , …, { w S , w O , w D } 10,000 to generate 10,000 random RO AP tables. The threshold values were adjusted until nearly the same numbers of H, M, and L combinations were obtained on average. For that purpose, very high was treated the same as high. 2.2.2.2 Conditioning the RPN To be able to compare the RPN with the AIAG AP, RPN action levels must be established that mimic the purpose of the AP action levels. We referred to the RPN action levels provided by the BFS, DEGRO, DGMP, and DGN . To set up the analogy, RPN values greater than 125 were treated the same as H ratings, and RPN values between 125 and 30 were treated as M ratings. However, these values were derived from a different S-O-D rating system as the one used in the FMEA handbook. Therefore, S-O-D numbers of the FMEA handbook were mapped to the respective criteria of the joint recommendations. For example, the criterion ‘more than once per day’ is O = 9 or O = 10 in the joint recommendations and O = 7 in the FMEA handbook. After the mapping process (see s for details), the new RPN action levels were 110 (H) and 27 (M). 2.2.2.3 Monte Carlo simulation The Monte Carlo simulations were performed for the AIAG AP as illustrated in . The RO AP and RPN simulations were processed analogously with their respective action levels (see , ). The start of the simulations was to draw 10,000 random failure mode ratings, i.e., S-O-D combinations. To obtain more realistic ratings, the S-O-D combinations were drawn from the probability density functions (PDFs) shown in . These PDFs were calculated from 216 S-O-D ratings provided by AAPM TG-100 using a Gaussian kernel density estimator and Silverman’s rule of thumb as bandwidth estimator. In the AIAG AP track, the AIAG APs of the 10,000 random S-O-D ratings were deduced. This determined the sequence of optimization. Then, during the actual optimization, random prevention and detection ‘actions’ were drawn uniformly. This means that either the O or D rating was randomly improved but never both ratings at once. Actions were continuously drawn until the optimized AIAG AP was either M or L. In case of an (optimized) M rating, (further) actions were drawn until a maximum number of four (this number was considered as sufficient defense in depth by IAEA-TECDOC-1685 ). Failure mode ratings were considered acceptable once there was either sufficient defense in depth or an L rating. In the RO AP track, the simulation was repeatedly conducted for 40 random sets of weighting factors { w S , w O , w D } 1 , …, { w S , w O , w D } 40 . This would reveal whether there was a relationship between these and the number of needed actions. Moreover, if a set of weighting factors existed that could replicate the AIAG AP optimization, it could then be determined by interpolation. Once all failure modes were accepted in each track, the average numbers of actions per failure mode could be determined. Additionally, we quantified the achieved optimized states by means of the residual total criticality number ΣSO as a surrogate. The criticality number SO is the product of severity and occurrence and directs the focus on reducing the occurrence of failure modes . Finally, we parametrized the RO AP with the found sets of threshold values and weighting factors to reproduce the AIAG AP. Then, the median RPN values of the four action levels were determined and compared, using the Kruskal-Wallis H test, a non-parametric analysis of variance. 2.2.3 Adaption The general threshold values found in were substituted with appropriate threshold values to calibrate the RO AP for the S-O-D rating system given in . These were chosen manually so that the following intentions were satisfied: If failure modes resulted in mild side effects at worst ( S ≤ 4) and occurred once a month or less ( O ≤ 4), then an acceptable state (L) should easily be achievable because the benefit of the treatment far surpasses the risks. If S ≤ 6 (up to mild toxicity), then depending on the occurrence, actions should or must be taken (M/H) because toxicities, in general, should be avoided. If S ≥ 8 (up to premature death), actions must immediately be taken (VH) because the benefit of the treatment is greatly reduced. Definition The RO AP was defined by means of a weighted sum: R O A P y = very high if y ≥ y vh high if y ≥ y h medium if y ≥ y m low if y < y m o r S = 1 o r O = 1 with y = w S S + w O O + w D D , and w S , w O , and w D being weighting factors that satisfied w S ≥ w O ≥ w D as well as w S + w O + w D = 3 . This way, the weighted sum y never exceeded 30, which is the maximum number one can achieve using a ten-step rating system. A fourth action level, very high (VH), was introduced to distinguish actions that must be implemented immediately. Benchmark Monte Carlo simulations were performed in order to compare the number of needed actions as required by the AIAG AP, RO AP, and the RPN. We also compared whether the three methods would lead to comparable optimized states. This comparison required establishing the near-same conditions for the RO AP and RPN which is described hereafter. 2.2.2.1 Conditioning the RO AP The threshold values y VH , y H , and y M had to be determined first in order to create RO AP tables that have nearly the same number of high, medium and low combinations as the AIAG AP table. Of the 1000 S-O-D combinations in the AIAG AP table, 318, 214 and 468 combinations refer to H, M, and L, respectively. We identified general threshold values iteratively by using 10,000 random sets of weighting factors { w S , w O , w D } 1 , …, { w S , w O , w D } 10,000 to generate 10,000 random RO AP tables. The threshold values were adjusted until nearly the same numbers of H, M, and L combinations were obtained on average. For that purpose, very high was treated the same as high. 2.2.2.2 Conditioning the RPN To be able to compare the RPN with the AIAG AP, RPN action levels must be established that mimic the purpose of the AP action levels. We referred to the RPN action levels provided by the BFS, DEGRO, DGMP, and DGN . To set up the analogy, RPN values greater than 125 were treated the same as H ratings, and RPN values between 125 and 30 were treated as M ratings. However, these values were derived from a different S-O-D rating system as the one used in the FMEA handbook. Therefore, S-O-D numbers of the FMEA handbook were mapped to the respective criteria of the joint recommendations. For example, the criterion ‘more than once per day’ is O = 9 or O = 10 in the joint recommendations and O = 7 in the FMEA handbook. After the mapping process (see s for details), the new RPN action levels were 110 (H) and 27 (M). 2.2.2.3 Monte Carlo simulation The Monte Carlo simulations were performed for the AIAG AP as illustrated in . The RO AP and RPN simulations were processed analogously with their respective action levels (see , ). The start of the simulations was to draw 10,000 random failure mode ratings, i.e., S-O-D combinations. To obtain more realistic ratings, the S-O-D combinations were drawn from the probability density functions (PDFs) shown in . These PDFs were calculated from 216 S-O-D ratings provided by AAPM TG-100 using a Gaussian kernel density estimator and Silverman’s rule of thumb as bandwidth estimator. In the AIAG AP track, the AIAG APs of the 10,000 random S-O-D ratings were deduced. This determined the sequence of optimization. Then, during the actual optimization, random prevention and detection ‘actions’ were drawn uniformly. This means that either the O or D rating was randomly improved but never both ratings at once. Actions were continuously drawn until the optimized AIAG AP was either M or L. In case of an (optimized) M rating, (further) actions were drawn until a maximum number of four (this number was considered as sufficient defense in depth by IAEA-TECDOC-1685 ). Failure mode ratings were considered acceptable once there was either sufficient defense in depth or an L rating. In the RO AP track, the simulation was repeatedly conducted for 40 random sets of weighting factors { w S , w O , w D } 1 , …, { w S , w O , w D } 40 . This would reveal whether there was a relationship between these and the number of needed actions. Moreover, if a set of weighting factors existed that could replicate the AIAG AP optimization, it could then be determined by interpolation. Once all failure modes were accepted in each track, the average numbers of actions per failure mode could be determined. Additionally, we quantified the achieved optimized states by means of the residual total criticality number ΣSO as a surrogate. The criticality number SO is the product of severity and occurrence and directs the focus on reducing the occurrence of failure modes . Finally, we parametrized the RO AP with the found sets of threshold values and weighting factors to reproduce the AIAG AP. Then, the median RPN values of the four action levels were determined and compared, using the Kruskal-Wallis H test, a non-parametric analysis of variance. Conditioning the RO AP The threshold values y VH , y H , and y M had to be determined first in order to create RO AP tables that have nearly the same number of high, medium and low combinations as the AIAG AP table. Of the 1000 S-O-D combinations in the AIAG AP table, 318, 214 and 468 combinations refer to H, M, and L, respectively. We identified general threshold values iteratively by using 10,000 random sets of weighting factors { w S , w O , w D } 1 , …, { w S , w O , w D } 10,000 to generate 10,000 random RO AP tables. The threshold values were adjusted until nearly the same numbers of H, M, and L combinations were obtained on average. For that purpose, very high was treated the same as high. Conditioning the RPN To be able to compare the RPN with the AIAG AP, RPN action levels must be established that mimic the purpose of the AP action levels. We referred to the RPN action levels provided by the BFS, DEGRO, DGMP, and DGN . To set up the analogy, RPN values greater than 125 were treated the same as H ratings, and RPN values between 125 and 30 were treated as M ratings. However, these values were derived from a different S-O-D rating system as the one used in the FMEA handbook. Therefore, S-O-D numbers of the FMEA handbook were mapped to the respective criteria of the joint recommendations. For example, the criterion ‘more than once per day’ is O = 9 or O = 10 in the joint recommendations and O = 7 in the FMEA handbook. After the mapping process (see s for details), the new RPN action levels were 110 (H) and 27 (M). Monte Carlo simulation The Monte Carlo simulations were performed for the AIAG AP as illustrated in . The RO AP and RPN simulations were processed analogously with their respective action levels (see , ). The start of the simulations was to draw 10,000 random failure mode ratings, i.e., S-O-D combinations. To obtain more realistic ratings, the S-O-D combinations were drawn from the probability density functions (PDFs) shown in . These PDFs were calculated from 216 S-O-D ratings provided by AAPM TG-100 using a Gaussian kernel density estimator and Silverman’s rule of thumb as bandwidth estimator. In the AIAG AP track, the AIAG APs of the 10,000 random S-O-D ratings were deduced. This determined the sequence of optimization. Then, during the actual optimization, random prevention and detection ‘actions’ were drawn uniformly. This means that either the O or D rating was randomly improved but never both ratings at once. Actions were continuously drawn until the optimized AIAG AP was either M or L. In case of an (optimized) M rating, (further) actions were drawn until a maximum number of four (this number was considered as sufficient defense in depth by IAEA-TECDOC-1685 ). Failure mode ratings were considered acceptable once there was either sufficient defense in depth or an L rating. In the RO AP track, the simulation was repeatedly conducted for 40 random sets of weighting factors { w S , w O , w D } 1 , …, { w S , w O , w D } 40 . This would reveal whether there was a relationship between these and the number of needed actions. Moreover, if a set of weighting factors existed that could replicate the AIAG AP optimization, it could then be determined by interpolation. Once all failure modes were accepted in each track, the average numbers of actions per failure mode could be determined. Additionally, we quantified the achieved optimized states by means of the residual total criticality number ΣSO as a surrogate. The criticality number SO is the product of severity and occurrence and directs the focus on reducing the occurrence of failure modes . Finally, we parametrized the RO AP with the found sets of threshold values and weighting factors to reproduce the AIAG AP. Then, the median RPN values of the four action levels were determined and compared, using the Kruskal-Wallis H test, a non-parametric analysis of variance. Adaption The general threshold values found in were substituted with appropriate threshold values to calibrate the RO AP for the S-O-D rating system given in . These were chosen manually so that the following intentions were satisfied: If failure modes resulted in mild side effects at worst ( S ≤ 4) and occurred once a month or less ( O ≤ 4), then an acceptable state (L) should easily be achievable because the benefit of the treatment far surpasses the risks. If S ≤ 6 (up to mild toxicity), then depending on the occurrence, actions should or must be taken (M/H) because toxicities, in general, should be avoided. If S ≥ 8 (up to premature death), actions must immediately be taken (VH) because the benefit of the treatment is greatly reduced. Results 3.1 EBRT-PFMEA Template On the basis of the AIAG PFMEA approach, an EBRT-PFMA template with 9 process items, 33 process steps, and 112 process functions could be compiled. The template is given to the reader in full detail via the . Here, an excerpt from the template, showing a subset of the structure and function analysis, is shown in . During the failure analysis, over 1400 failure modes could be gathered through the literature survey and 75 distinct, high-risk failure chains could be identified. 70.7% failure modes were preventable or detectable by the process itself which indicated the already high level of inherent safety of the general process map. In addition, 92 prevention controls and 70 detection controls were identified that could be implemented by users of the template and were not already part of the process. Among the prevention controls, standardized policies, documentation and communication strategies as well as key performance indicators (KPI) were most dominant. Checklists and the four-eyes-principle were a common theme for detecting failure modes or causes. shows the number of failure modes in each process step. With 22 (29.3%) failure modes, treatment delivery was most prone to error. Treatment planning, i. e., medical and physical planning, followed with 18 (24%) failure modes in total. 14 (18.7%) failure modes were identified for primary RT imaging. In , process functions acting as detection controls are listed. Being able to detect 16% failure modes, image guided verification was the most effective detection control. Followed by surface guided radiotherapy and physics (treatment planning) consult, 13.9% and 9.7% of identified failure modes were potentially detectable, respectively. 3.2 Radiation Oncology Action Priority 3.2.1 Benchmark Based on random sets of weighting factors, we found y VH = 24.0, y H = 19.2 and y M = 15.5 to be suitable and general threshold values that yielded 318 VH and H, 212 M and 470 L combinations on average. shows the optimization processes obtained through the Monte Carlo simulations. There are two crucial observations: Firstly, the RO AP could theoretically replicate the AIAG AP with regards to the number of necessary actions (left and right). The associated set of weighting factors was found to be w S = 1.34, w O = 1.20, and w D = 0.46. Then, the RO AP would require 1.7 actions per failure mode, as did the AIAG AP. The greater w S , the more actions would be necessary because the algorithm only drew prevention ( O ) and detection ( D ) controls. Secondly, the RPN optimization stopped at a value of the residual total criticality number ΣSO which was approximately 47% lower than the value obtained through the AIAG AP (right). This indicated that the risk acceptance of the BFS, DEGRO, DGMP, and DGN was more restrictive than the AIAG’s risk acceptance. This inequality increased the required number of actions by 2.1 to 3.6 actions per failure mode. Another consequence was the apparently slower optimization. The average slope for the AIAG AP was –33.7 ΣSO per action, whereas it was –24.7 ΣSO per action for the RPN. shows box plots of the total RPN distribution (leftmost) as well as RPN sub-distributions classified as RO AP action levels. Using the Kruskal-Wallis H test, it could be shown that the medians of all RPN sub-distributions were significantly different from each other ( p < 0.00001). 3.2.2 Adaption We could identify an RO AP table reflecting the intentions described in using the threshold values y VH = 23.0, y H = 18.0, and y M = 14.0 (see s). This RO AP parametrization had 395 (+24.2%) VH and H, 229 (+8.0%) M, and 376 (-20.0%) L combinations and required around 2.4 corrective actions per failure mode (+41.4%). Again, the Kruskal-Wallis H test showed significantly different RPN medians of all groups ( p ≈ 0). EBRT-PFMEA Template On the basis of the AIAG PFMEA approach, an EBRT-PFMA template with 9 process items, 33 process steps, and 112 process functions could be compiled. The template is given to the reader in full detail via the . Here, an excerpt from the template, showing a subset of the structure and function analysis, is shown in . During the failure analysis, over 1400 failure modes could be gathered through the literature survey and 75 distinct, high-risk failure chains could be identified. 70.7% failure modes were preventable or detectable by the process itself which indicated the already high level of inherent safety of the general process map. In addition, 92 prevention controls and 70 detection controls were identified that could be implemented by users of the template and were not already part of the process. Among the prevention controls, standardized policies, documentation and communication strategies as well as key performance indicators (KPI) were most dominant. Checklists and the four-eyes-principle were a common theme for detecting failure modes or causes. shows the number of failure modes in each process step. With 22 (29.3%) failure modes, treatment delivery was most prone to error. Treatment planning, i. e., medical and physical planning, followed with 18 (24%) failure modes in total. 14 (18.7%) failure modes were identified for primary RT imaging. In , process functions acting as detection controls are listed. Being able to detect 16% failure modes, image guided verification was the most effective detection control. Followed by surface guided radiotherapy and physics (treatment planning) consult, 13.9% and 9.7% of identified failure modes were potentially detectable, respectively. Radiation Oncology Action Priority 3.2.1 Benchmark Based on random sets of weighting factors, we found y VH = 24.0, y H = 19.2 and y M = 15.5 to be suitable and general threshold values that yielded 318 VH and H, 212 M and 470 L combinations on average. shows the optimization processes obtained through the Monte Carlo simulations. There are two crucial observations: Firstly, the RO AP could theoretically replicate the AIAG AP with regards to the number of necessary actions (left and right). The associated set of weighting factors was found to be w S = 1.34, w O = 1.20, and w D = 0.46. Then, the RO AP would require 1.7 actions per failure mode, as did the AIAG AP. The greater w S , the more actions would be necessary because the algorithm only drew prevention ( O ) and detection ( D ) controls. Secondly, the RPN optimization stopped at a value of the residual total criticality number ΣSO which was approximately 47% lower than the value obtained through the AIAG AP (right). This indicated that the risk acceptance of the BFS, DEGRO, DGMP, and DGN was more restrictive than the AIAG’s risk acceptance. This inequality increased the required number of actions by 2.1 to 3.6 actions per failure mode. Another consequence was the apparently slower optimization. The average slope for the AIAG AP was –33.7 ΣSO per action, whereas it was –24.7 ΣSO per action for the RPN. shows box plots of the total RPN distribution (leftmost) as well as RPN sub-distributions classified as RO AP action levels. Using the Kruskal-Wallis H test, it could be shown that the medians of all RPN sub-distributions were significantly different from each other ( p < 0.00001). 3.2.2 Adaption We could identify an RO AP table reflecting the intentions described in using the threshold values y VH = 23.0, y H = 18.0, and y M = 14.0 (see s). This RO AP parametrization had 395 (+24.2%) VH and H, 229 (+8.0%) M, and 376 (-20.0%) L combinations and required around 2.4 corrective actions per failure mode (+41.4%). Again, the Kruskal-Wallis H test showed significantly different RPN medians of all groups ( p ≈ 0). Benchmark Based on random sets of weighting factors, we found y VH = 24.0, y H = 19.2 and y M = 15.5 to be suitable and general threshold values that yielded 318 VH and H, 212 M and 470 L combinations on average. shows the optimization processes obtained through the Monte Carlo simulations. There are two crucial observations: Firstly, the RO AP could theoretically replicate the AIAG AP with regards to the number of necessary actions (left and right). The associated set of weighting factors was found to be w S = 1.34, w O = 1.20, and w D = 0.46. Then, the RO AP would require 1.7 actions per failure mode, as did the AIAG AP. The greater w S , the more actions would be necessary because the algorithm only drew prevention ( O ) and detection ( D ) controls. Secondly, the RPN optimization stopped at a value of the residual total criticality number ΣSO which was approximately 47% lower than the value obtained through the AIAG AP (right). This indicated that the risk acceptance of the BFS, DEGRO, DGMP, and DGN was more restrictive than the AIAG’s risk acceptance. This inequality increased the required number of actions by 2.1 to 3.6 actions per failure mode. Another consequence was the apparently slower optimization. The average slope for the AIAG AP was –33.7 ΣSO per action, whereas it was –24.7 ΣSO per action for the RPN. shows box plots of the total RPN distribution (leftmost) as well as RPN sub-distributions classified as RO AP action levels. Using the Kruskal-Wallis H test, it could be shown that the medians of all RPN sub-distributions were significantly different from each other ( p < 0.00001). Adaption We could identify an RO AP table reflecting the intentions described in using the threshold values y VH = 23.0, y H = 18.0, and y M = 14.0 (see s). This RO AP parametrization had 395 (+24.2%) VH and H, 229 (+8.0%) M, and 376 (-20.0%) L combinations and required around 2.4 corrective actions per failure mode (+41.4%). Again, the Kruskal-Wallis H test showed significantly different RPN medians of all groups ( p ≈ 0). Discussion 4.1 EBRT-PFMEA Template In this study, we compiled a practical EBRT-PFMEA template. The advantages of templates are multifold: For example, templates may serve as input for identifying failure modes that have been overlooked in an FMEA started from scratch . In automotive industry, templates are already a common and recommended practice that reduces resources and accumulates past experiences and knowledge from lessons learnt . Templates could also contribute to standardizing the quality of risk assessments across institutions . If mundane causes such as little knowledge or lack of staff or time lower risk assessment efforts, then templates could improve and speed up these efforts. Certainly, little knowledge was true for our department in the phase just after risk assessments were made obligatory. However, as radiation oncology departments gain experience and future medical physics experts in the EU should acquire risk management skills , these causes might diminish over time. In this study, the AIAG PFMEA approach has been applied which differs from other established approaches, such as the ones recommended by AAPM TG-100 or by the BFS, DEGRO, DGMP, and DGN . All three approaches have in common that the process is mapped out, however, the AIAG approach goes beyond that, inter alia, by performing a structure and function analysis. These steps are immensely crucial because they establish what is being done by whom and what needs to be achieved. Defining these structures and functions makes deducing failure modes a purely logical task. This is because the failure modes are simply the negated process step functions. For instance, a function executed incorrectly, too early, too late, in excess or insufficiently, etc. will represent a failure mode. In the other two approaches, identifying failure modes was immediately being done after the process had been mapped out. There, identifying failure modes rather resembled a brainstorming exercise (what could possibly go wrong?). Because of that, the quality of the identified failure modes will necessarily be dependent of the experience of the analysts (obtained, for instance, through past undesirable events). Even though the literature survey provided over 1400 failure modes, we could not include all of them due to practical reasons. Repeated and specialized failure modes have been combined and abstracted. All failure modes that could be eliminated due to the EBRT process design used here and those representing errors due to omission (e. g., ‘peer review not performed’, ‘IGRT not performed’, etc.) have been sorted out. Obviously, patient safety is compromised if crucial functions are not performed. For instance, vertebral bodies were mixed-up in the past because the patient position had not been verified . However, PFMEA is a tool for analyzing what is wrong with the process , thereby identifying safety-improving process functions. To make sure these functions are not omitted, other tools such as checklists should be employed. The process of compiling FMEA-based checklists is illustrated in detail in AAPM TG-275 . According to , there were no failure modes of the pre-treatment verification process, which was unexpected. However, the pre-treatment verification itself serves as an extensive detection method of failure modes that occurred in previous steps. For instance, if an incorrect dose distribution was calculated during physical planning, this should be detected during verification but the respective failure mode is belonging to the planning process. Actual failure modes of the verification (e.g., incorrect execution of measurements) were not included as the patient safety was not considered to be directly compromised through these modes. 4.2 Usage of the EBRT-PFMEA Template In a first step, the reader modifies the template provided in the s in order to create a local template valid for their institution. Because of that, our EBRT-PFMEA template was kept rather general so that differences to the reader’s specific EBRT process can be quickly identified and either be added or removed. The provided process structures and functions should be adapted to the institution’s situation first. Some failure modes might be eliminated or new site-specific failure modes emerge as a consequence. We do not recommend removing crucial process functions since these ensure a high level of inherent safety, as can be seen in . In a second step, the reader’s local template is then used for the actual risk assessment of, e.g., the launch of a new treatment unit or a new treatment technique. 4.3 Limitations of the EBRT-PFMEA Template The EBRT-PFMEA template has not yet been clinically validated. Moreover, we restricted the literature review to the ten highest ranked failure modes per article which all have been ranked by the RPN. Since RPNs are ambiguous in terms of level of risk , some high-risk failure modes might have not appeared in the top ten of each article. This could result in an incomplete template. 4.4 Radiation Oncology Action Priority In this study, we also investigated how the novel action priority concept could be transferred to radiation oncology. The dominating concept in radiation oncology is, however, not the AIAG AP but the RPN, due to the task report TG-100 , the joint recommendations of the BFS, DEGRO, DGMP, and DGN and many other publications. For the AP to be accepted, a proof showing its superiority over the RPN seemed necessary, while at the same time it should easily be adaptable to any rating system, i.e., without the need for reviewing all 1000 cells. In other studies, it has already been shown that the AIAG AP increased the chances of reaching a team consensus (i. e., the same action level) because it only offers three action levels . Using a ten-step rating system, the variance can be as high as 7 steps between individuals, so in the case of the RPN, there could be an uncertainty factor of up to 12 . Therefore, the chances of reaching the same consensus are lower. Furthermore, Barsalou recently compared high APs versus RPN values greater than 100 and showed a significant reduction in the number of corrective actions . Here, we propose the RO AP, a transparent method for calibrating the AP table for any rating system, as an alternative to the fixed AIAG AP table. Building on Braband’s sum rule for the RPN , we added weighting factors and threshold values to their formula. We could show that the RO AP and AIAG AP were practically equivalent in terms of optimization . Moreover, we could verify that the two AP approaches indeed reduced the number of corrective actions compared to the RPN, underlining Barsalou’s findings . However, results of this kind are rather arbitrary because AP and RPN action levels themselves are arbitrary . For example, Noel et al. considered RPN ≥ 200 as critical, implying presumably less corrective actions. We could observe that adjusting the RO AP threshold values to our needs increased the number of actions per failure mode from 1.7 to 2.4. Thus, comparing the AP and the RPN concepts should always be done with caution. As can be seen in , RPNs between 90 and 162 could be any of the four RO AP action levels. As reference, about 38.9% TG-100 failure modes lie in this interval. Our results imply that a considerable amount of failure modes in radiation oncology could be optimized in a more effective way by utilizing the RO AP. 4.5 Limitations of the RO AP The simulation simplified reality. It was assumed that there are unlimited resources so that all failure modes could be optimized. In reality, teams might not find suitable measures or justify that current measures are already sufficient, for example because risks are considered to be already as low as reasonably achievable (ALARA). Furthermore, the failure-prevention intent of the AP was neglected as severity reducing actions were not simulated, even though they would reduce the AP most drastically. In reality, the actual number of actions per failure mode will likely be lower due to aforementioned reasons. EBRT-PFMEA Template In this study, we compiled a practical EBRT-PFMEA template. The advantages of templates are multifold: For example, templates may serve as input for identifying failure modes that have been overlooked in an FMEA started from scratch . In automotive industry, templates are already a common and recommended practice that reduces resources and accumulates past experiences and knowledge from lessons learnt . Templates could also contribute to standardizing the quality of risk assessments across institutions . If mundane causes such as little knowledge or lack of staff or time lower risk assessment efforts, then templates could improve and speed up these efforts. Certainly, little knowledge was true for our department in the phase just after risk assessments were made obligatory. However, as radiation oncology departments gain experience and future medical physics experts in the EU should acquire risk management skills , these causes might diminish over time. In this study, the AIAG PFMEA approach has been applied which differs from other established approaches, such as the ones recommended by AAPM TG-100 or by the BFS, DEGRO, DGMP, and DGN . All three approaches have in common that the process is mapped out, however, the AIAG approach goes beyond that, inter alia, by performing a structure and function analysis. These steps are immensely crucial because they establish what is being done by whom and what needs to be achieved. Defining these structures and functions makes deducing failure modes a purely logical task. This is because the failure modes are simply the negated process step functions. For instance, a function executed incorrectly, too early, too late, in excess or insufficiently, etc. will represent a failure mode. In the other two approaches, identifying failure modes was immediately being done after the process had been mapped out. There, identifying failure modes rather resembled a brainstorming exercise (what could possibly go wrong?). Because of that, the quality of the identified failure modes will necessarily be dependent of the experience of the analysts (obtained, for instance, through past undesirable events). Even though the literature survey provided over 1400 failure modes, we could not include all of them due to practical reasons. Repeated and specialized failure modes have been combined and abstracted. All failure modes that could be eliminated due to the EBRT process design used here and those representing errors due to omission (e. g., ‘peer review not performed’, ‘IGRT not performed’, etc.) have been sorted out. Obviously, patient safety is compromised if crucial functions are not performed. For instance, vertebral bodies were mixed-up in the past because the patient position had not been verified . However, PFMEA is a tool for analyzing what is wrong with the process , thereby identifying safety-improving process functions. To make sure these functions are not omitted, other tools such as checklists should be employed. The process of compiling FMEA-based checklists is illustrated in detail in AAPM TG-275 . According to , there were no failure modes of the pre-treatment verification process, which was unexpected. However, the pre-treatment verification itself serves as an extensive detection method of failure modes that occurred in previous steps. For instance, if an incorrect dose distribution was calculated during physical planning, this should be detected during verification but the respective failure mode is belonging to the planning process. Actual failure modes of the verification (e.g., incorrect execution of measurements) were not included as the patient safety was not considered to be directly compromised through these modes. Usage of the EBRT-PFMEA Template In a first step, the reader modifies the template provided in the s in order to create a local template valid for their institution. Because of that, our EBRT-PFMEA template was kept rather general so that differences to the reader’s specific EBRT process can be quickly identified and either be added or removed. The provided process structures and functions should be adapted to the institution’s situation first. Some failure modes might be eliminated or new site-specific failure modes emerge as a consequence. We do not recommend removing crucial process functions since these ensure a high level of inherent safety, as can be seen in . In a second step, the reader’s local template is then used for the actual risk assessment of, e.g., the launch of a new treatment unit or a new treatment technique. Limitations of the EBRT-PFMEA Template The EBRT-PFMEA template has not yet been clinically validated. Moreover, we restricted the literature review to the ten highest ranked failure modes per article which all have been ranked by the RPN. Since RPNs are ambiguous in terms of level of risk , some high-risk failure modes might have not appeared in the top ten of each article. This could result in an incomplete template. Radiation Oncology Action Priority In this study, we also investigated how the novel action priority concept could be transferred to radiation oncology. The dominating concept in radiation oncology is, however, not the AIAG AP but the RPN, due to the task report TG-100 , the joint recommendations of the BFS, DEGRO, DGMP, and DGN and many other publications. For the AP to be accepted, a proof showing its superiority over the RPN seemed necessary, while at the same time it should easily be adaptable to any rating system, i.e., without the need for reviewing all 1000 cells. In other studies, it has already been shown that the AIAG AP increased the chances of reaching a team consensus (i. e., the same action level) because it only offers three action levels . Using a ten-step rating system, the variance can be as high as 7 steps between individuals, so in the case of the RPN, there could be an uncertainty factor of up to 12 . Therefore, the chances of reaching the same consensus are lower. Furthermore, Barsalou recently compared high APs versus RPN values greater than 100 and showed a significant reduction in the number of corrective actions . Here, we propose the RO AP, a transparent method for calibrating the AP table for any rating system, as an alternative to the fixed AIAG AP table. Building on Braband’s sum rule for the RPN , we added weighting factors and threshold values to their formula. We could show that the RO AP and AIAG AP were practically equivalent in terms of optimization . Moreover, we could verify that the two AP approaches indeed reduced the number of corrective actions compared to the RPN, underlining Barsalou’s findings . However, results of this kind are rather arbitrary because AP and RPN action levels themselves are arbitrary . For example, Noel et al. considered RPN ≥ 200 as critical, implying presumably less corrective actions. We could observe that adjusting the RO AP threshold values to our needs increased the number of actions per failure mode from 1.7 to 2.4. Thus, comparing the AP and the RPN concepts should always be done with caution. As can be seen in , RPNs between 90 and 162 could be any of the four RO AP action levels. As reference, about 38.9% TG-100 failure modes lie in this interval. Our results imply that a considerable amount of failure modes in radiation oncology could be optimized in a more effective way by utilizing the RO AP. Limitations of the RO AP The simulation simplified reality. It was assumed that there are unlimited resources so that all failure modes could be optimized. In reality, teams might not find suitable measures or justify that current measures are already sufficient, for example because risks are considered to be already as low as reasonably achievable (ALARA). Furthermore, the failure-prevention intent of the AP was neglected as severity reducing actions were not simulated, even though they would reduce the AP most drastically. In reality, the actual number of actions per failure mode will likely be lower due to aforementioned reasons. Conclusion A practical necessity for a general FMEA template arises due to legal obligations in the EU to perform risk assessments for unintentional exposures to radiation . Practitioners lacking knowledge in risk management procedures are furnished an adjustable model that helps them in designing a safe process map, identifying failure modes and monitoring effects of control mechanisms over time. In sharing the template, it could become a standardized starting point for all EBRT-based processes. In a collective approach to increase patient safety, we hope to encourage other authors with this initiative to provide FMEA templates for, e. g., brachytherapy or other branches of radiotherapy. The RO AP can reproduce an optimization process similar to the AIAG AP. However, the RO AP has the advantage that it founds on a rather simple weighted sum which makes the action levels of all 1000 S-O-D combinations transparent. In addition, the RO AP table can easily be adapted to any rating system. This replaces the necessity to inspect all cells of the AP table individually when modifying the rating system. Dominik Kornek: Conceptualization, Writing - original draft, Visualization, Investigation, Validation, Formal analysis, Methodology. Christoph Bert: Funding acquisition, Writing - review & editing, Validation, Methodology, Supervision. The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: This project was funded by the Bavarian Ministry of Economic Affairs, Regional Development and Energy (grant number 07 03/686 68/288/21/7/22/8/23/9/24, 07 03/686 68/287/21/4/22/5/23/6/24) and was performed in collaboration with IBA Dosimetry GmbH (Schwarzenbruck, Germany). IBA Dosimetry GmbH had no involvement with the work reported in this paper. |
Examining the relationship between demographic variables and perceived health literacy challenges in Tasmania, Australia | 8f8f53a9-0087-4a94-b079-8ade09e7bec5 | 11730517 | Health Literacy[mh] | INTRODUCTION A crucial part of creating a healthier society is reducing the burden of illness and preventing disease. Health literacy (HL) is essential to self‐efficacy, disease prevention, and attempts to redress inequities in health. Determining how HL differs within a population and identifying the perceived challenges to achieving a high level of HL is crucial to creating a healthy society. The individual elements of HL, the social determinants of health, and the impact on health outcomes have been explored over the last decade. , However, evidence on the interconnected nature of these three factors is lacking. The present paper explores this gap through mixed methods research and attempts to determine how HL, the social determinants of health, and health outcomes interact within a population. Health literacy is an emerging concept, and its definition is continuously expanding and changing. One definition of health literacy, and the one used for this paper, is ‘health literacy represents the personal knowledge and competencies that accumulate through daily activities and social interactions and across generations. Personal knowledge and competencies are mediated by the organizational structures and availability of resources that enable people to access, understand, appraise, and use information and services in ways that promote and maintain good health and well‐being for themselves and those around them’. It is clear from this definition that HL brings together a number of concepts that relate to both the individual‐ and population‐level assets required to make effective decisions about health for themselves, their families, and their overall communities. , , HL is an important element of a person's overall health. International organisations such as the World Health Organisation (WHO) (2015) are currently focused on ways to improve, adapt, and bolster the HL assets of individuals and organisations. In developed nations, including Australia (where this research was conducted) most people will acquire a communicable disease (CD) (e.g., COVID‐19) during their lifetime. However, non‐communicable Diseases (NCDs) are the leading cause of death and disability, accounting for 71% of all deaths globally. In Australia, NCDs are responsible for 89% of all deaths and place a significant burden on the health system. NCDs are also referred to as chronic diseases and tend to be of long duration. They are the result of a combination of genetic, physiological, environmental, and behavioural risk factors and include conditions such as cardiovascular diseases, cancers, chronic respiratory diseases, and diabetes. Their risk factors, such as unhealthy diets, physical inactivity, exposure to tobacco smoke or the harmful use of alcohol, are modifiable. The escalating burden of all diseases continues to contribute to the ‘social gradient’ in health outcomes and is amplifying health inequity and poverty worldwide. Health promotion and health education aim to assist individuals to improve and take control of their own health. Ensuring both translate into the development of health literacy assets is imperative. , Health is seen as a resource for everyday life, and therefore an individual or group must be able to identify and realise aspirations, satisfy needs, and change to cope with their environment. By creating systems and environments that can support people to make healthy choices, we may prevent NCDs from developing, mitigate the spread of CDs, and reduce the impact of both on population health outcomes. In Australia, the National Preventative Health Strategy (2021) has been developed to improve the health and wellbeing at all stages of life. This Strategy uses a systems‐based approach to prevention that addresses the wider determinants of health, reduces health inequities, and decreases the overall burden of disease. However, in isolation, health promotion and disease prevention are not enough to empower individuals and communities to take an active role in improving their health. Health promotion efforts must be tailored and responsive to the existing population's health literacy strengths and challenges to help individuals develop the assets they need to make autonomous decisions regarding their health and wellbeing. In Australia, the Commission of Safety and Quality of Health Care (2014) has created a national action plan to address HL via a coordinated approach to increase the safety, quality, and sustainability of the health care system. Improving HL is one important strategy required to reduce the unsustainable stress on health care systems. HL has been defined as a social determinant of health (SDH) in its own right. In their exploration of HL as a determinant, mediator and/or moderator of health, Pelikan et al. (2018) concluded that comprehensive HL is a critical, direct determinant of health. They highlighted that HL has considerable potential for health promotion and may improve population health to tackle health inequity. The SDHs are described as the non‐medical factors that can influence health outcomes. They are the conditions in which individuals are born, grow, live, work and age, and they can influence the prevalence, distribution, and risk of developing disease. Disadvantaged social and socio‐economic conditions can contribute to low HL levels and poorer health outcomes. Addressing the common risk factors and SDH is an important step in reducing the growing burden of disease in Australia and globally. The SDHs link closely to an individual's demographic characteristics and can thus be predictors of HL challenges. Previous research has identified a range of factors that can influence someone's overall HL, including gender, household make‐up, education level, employment status, identification as an Aboriginal or Torres Strait Islander, and speaking languages other than English at home. , , For research purposes, it is often simpler to collect a person's demographic profile rather than to try to determine their SDHs. In Australia, the National Health Survey (NHS) collects data on the health of Australians, including health conditions, health risk factors, and all domains of the Health Literacy Questionnaire (HLQ). In 2017–2018 the NHS reported that people living in the island state of Tasmania have some of the lowest HL scores nationally on a selection of HLQ domains. The NHS (2017–2018) only looked at the HLQ scores and did not expand to include specific demographic data. This is important for those involved in HL promotion, given that having low HL can result in poor health outcomes. However, low HL may not always be a consequence of the individual's assets and the HLQ is designed to also acknowledge the contributions made by the health services and system. , Extending on this thinking, Ophelia which stands for ‘OPtimising HEalth LIteracy and Access’ provides an innovative process that focuses on a whole‐of‐system approach to developing grounded health literacy interventions that respond to local community strengths and challenges. Therefore, the HLQ serves as an assessment tool within the Ophelia framework, which supports the collection of individual's HL across multiple domains and therefore the health literacy strengths and challenges at a community level. The Ophelia principles aim to enhance HL by focusing on organisation, staff, communication, environment, and access. There are eight Ophelia principles: Outcomes focused, Equity driven, Needs diagnosis, Co‐design, Driven by local wisdom, Sustainable, Responsive and Systematically applied. The Ophelia process involves seven steps: assess the health literacy needs, develop a plan, organise the plan, implement interventions, evaluate outcomes, reflect and sustain, and communicate findings. As noted above, Tasmania experiences some of the worst health outcomes nationally. When comparing the rates of NCDs in the Tasmanian population to those of the nation, there is a greater prevalence of heart disease, diabetes, airway diseases, and cancers. Tasmanians also perform poorly against several NCD risk factors, including fruit and vegetable consumption, cigarette smoking rates, and obesity. Tasmania's growing health concerns have been recognised and government departments are taking steps to address them. The government initiative: Tasmanian Health Literacy Action Plan , has been designed to improve the health of the population. A recent report (2022) which considered the Tasmanian and Australian HLQ data highlights the specific strengths and challenges experienced by people in our communities. Whilst this report was not available when we commenced our research it outlines a set of important recommendations. Fortunately, in response to their findings, the interventions outlined in the report focus mainly on the ‘navigation’ aspect of HL. However, it may be helpful to extend into the other HL domains where the local population reported experiencing difficulties. As this will require continued effort, more research on local strengths, needs, and preferences will be essential to build upon the current recommendations into the future. In recognition of the HL disparities and significant burden of NCDs in the region, this study aimed to use the HLQ to assess how HL scores differ with social demographic characteristics. As described, we anticipated a social gradient of health literacy and health outcomes to exist in Tasmania. However, there is very little existing evidence about the factors (e.g., education, income, social supports) which determine, mediate, and moderate an individual's HL in Tasmania. This exploratory study represents the second study of a six‐stage broader research project and was designed to help identify which sociodemographic characteristics (including income, education status, and physical health) are associated with people's greatest self‐reported HL challenges. This research represents the first phase of an Ophelia study that is specifically focused on informing a future Health Literacy Mediator (HLM) role for communities. This HLM role may be able to address health inequity experienced by some members of our community.
METHOD 2.1 Sampling and participants This study was conducted using a cross‐sectional exploratory design and invited people living in the state of Tasmania, Australia to participate. Following ethics approval from the University of Tasmania Human Research Ethics Committee (H0026170), a convenience sampling approach was used. Participants were recruited between October and December 2021 through email and social media advertising, as well as via discussions with HL networks/organisations and community services (including services supporting migrants and adult literacy/numeracy development). Participants were eligible to participate if they were 18 years of age and over and residing in Tasmania. 2.2 Data collection The Health Literacy Questionnaire (HLQ) was used to collect HL data. The HLQ is a multi‐dimensional assessment tool used to identify HL strengths and challenges. It was developed following a validity‐driven approach utilising data from the general population, patients, health care providers, and policy makers. The HLQ contains 44 questions relating to 9 domains : Domain 1 —Feeling understood and supported by health care providers Domain 2 —Having sufficient information to manage my health Domain 3 —Actively managing my health Domain 4 —Social support for health Domain 5 —Appraisal of health information Domain 6 —Ability to actively engage with health care providers Domain 7 —Navigating the health care system Domain 8 —Ability to find good health information Domain 9 —Understanding health information well enough to know what to do These domains are applicable to all individuals throughout their life course and take account of both the challenges and the needs of society. The first five domains (1 through to 5) have responses ranging from 1 = strongly disagree to 4 = strongly agree , while the remaining domains (6 through to 9) have responses from 1 = cannot do or always difficult to 5 = always easy . It should be noted that the HLQ is not designed to allow the calculation of an overall score. Instead, the scores for each individual domain were examined to identify potential HL challenges and strengths. Each domain is considered to be an independent entity, and therefore researchers can choose to focus on one or more of the nine domains to answer specific research questions and evaluate specific outcomes. The research team discussed all nine domains and carefully selected five domains for this research. The specific domains chosen were intended to enhance our insights into the interactive and social ecological nature of HL, which is important for addressing the research questions. The domains selected for inclusion: 2, 3, 4, 6 and 7, were those that assessed individuals managing their own health, and relationships that individuals develop when interacting with the health care system or their social supports to make health‐related decisions each day. This approach made it possible to explore that interaction point, so that we could go further than just a focus on an individual's HL. This specific focus ensured greater alignment with the Ophelia principles of being equity‐driven, listening to local wisdom and using a needs assessment to inform responsive health and community services. The two qualitative questions in this study asked participants: What challenges/barriers have you experienced in the last 12 months when utilising health information, health care, and community health services? Can you think of anything that would help you overcome each of these barriers? Participants provided their own freeform answers with no limits on the number of discussion points. These barriers and proposed solutions will inform the roles and responsibilities of a HLM. This is out of scope of the current paper, but will be expanded upon in future papers. An online approach was selected due to the ongoing constraints of the COVID‐19 pandemic and subsequent lockdowns at the time of data collection. Participants completed an anonymous survey, which collected quantitative and qualitative data via REDCap (Research Electronic Data Capture), a secure web application. Before participants could progress to the survey questions, a written information leaflet that explained the study, its purpose and the potential benefits was made available and participants were required to provide electronic informed consent. Participants provided demographic information about their age, gender, country of birth, Aboriginal or Torres Strait Islander (ATSI) status, education, employment, and self‐reported number of chronic health condition/s or disability status. 2.3 Data analysis All data gathered from the surveys were analysed using jamovi 1.6.23. Descriptive statistics were calculated, and a series of hierarchical regression analyses were conducted to determine which of the demographic variables could account for the variance in the five individual HLQ domain scores assessed. For this analyses, the general assumptions were tested first. Quantile‐quantile plots revealed that the distribution of errors was acceptable, and the Variance Inflation Factor (VIF) values of less than 10 and tolerance values greater than .02 indicated no significant issues with multivariate collinearity. In each analysis, the variables were broken down into three distinct blocks and a three‐stage hierarchical regression was run. Thematic analysis, using Braun and Clarke's (2006) six‐phase methodology, was used to analyse participants' qualitative comments that identified potential barriers and solutions to their utilisation of health information, health care, and community health services. Thematic analysis was chosen to identify recurring themes within the responses that might reveal meaningful patterns. Some participants provided a single barrier or solution, whilst others listed as many as three. To maintain rigour, after Author 1 completed the analysis, Author 2 independently recoded 5% of the themes, and a comparative discussion occurred, prior to the final themes being decided.
Sampling and participants This study was conducted using a cross‐sectional exploratory design and invited people living in the state of Tasmania, Australia to participate. Following ethics approval from the University of Tasmania Human Research Ethics Committee (H0026170), a convenience sampling approach was used. Participants were recruited between October and December 2021 through email and social media advertising, as well as via discussions with HL networks/organisations and community services (including services supporting migrants and adult literacy/numeracy development). Participants were eligible to participate if they were 18 years of age and over and residing in Tasmania.
Data collection The Health Literacy Questionnaire (HLQ) was used to collect HL data. The HLQ is a multi‐dimensional assessment tool used to identify HL strengths and challenges. It was developed following a validity‐driven approach utilising data from the general population, patients, health care providers, and policy makers. The HLQ contains 44 questions relating to 9 domains : Domain 1 —Feeling understood and supported by health care providers Domain 2 —Having sufficient information to manage my health Domain 3 —Actively managing my health Domain 4 —Social support for health Domain 5 —Appraisal of health information Domain 6 —Ability to actively engage with health care providers Domain 7 —Navigating the health care system Domain 8 —Ability to find good health information Domain 9 —Understanding health information well enough to know what to do These domains are applicable to all individuals throughout their life course and take account of both the challenges and the needs of society. The first five domains (1 through to 5) have responses ranging from 1 = strongly disagree to 4 = strongly agree , while the remaining domains (6 through to 9) have responses from 1 = cannot do or always difficult to 5 = always easy . It should be noted that the HLQ is not designed to allow the calculation of an overall score. Instead, the scores for each individual domain were examined to identify potential HL challenges and strengths. Each domain is considered to be an independent entity, and therefore researchers can choose to focus on one or more of the nine domains to answer specific research questions and evaluate specific outcomes. The research team discussed all nine domains and carefully selected five domains for this research. The specific domains chosen were intended to enhance our insights into the interactive and social ecological nature of HL, which is important for addressing the research questions. The domains selected for inclusion: 2, 3, 4, 6 and 7, were those that assessed individuals managing their own health, and relationships that individuals develop when interacting with the health care system or their social supports to make health‐related decisions each day. This approach made it possible to explore that interaction point, so that we could go further than just a focus on an individual's HL. This specific focus ensured greater alignment with the Ophelia principles of being equity‐driven, listening to local wisdom and using a needs assessment to inform responsive health and community services. The two qualitative questions in this study asked participants: What challenges/barriers have you experienced in the last 12 months when utilising health information, health care, and community health services? Can you think of anything that would help you overcome each of these barriers? Participants provided their own freeform answers with no limits on the number of discussion points. These barriers and proposed solutions will inform the roles and responsibilities of a HLM. This is out of scope of the current paper, but will be expanded upon in future papers. An online approach was selected due to the ongoing constraints of the COVID‐19 pandemic and subsequent lockdowns at the time of data collection. Participants completed an anonymous survey, which collected quantitative and qualitative data via REDCap (Research Electronic Data Capture), a secure web application. Before participants could progress to the survey questions, a written information leaflet that explained the study, its purpose and the potential benefits was made available and participants were required to provide electronic informed consent. Participants provided demographic information about their age, gender, country of birth, Aboriginal or Torres Strait Islander (ATSI) status, education, employment, and self‐reported number of chronic health condition/s or disability status.
Data analysis All data gathered from the surveys were analysed using jamovi 1.6.23. Descriptive statistics were calculated, and a series of hierarchical regression analyses were conducted to determine which of the demographic variables could account for the variance in the five individual HLQ domain scores assessed. For this analyses, the general assumptions were tested first. Quantile‐quantile plots revealed that the distribution of errors was acceptable, and the Variance Inflation Factor (VIF) values of less than 10 and tolerance values greater than .02 indicated no significant issues with multivariate collinearity. In each analysis, the variables were broken down into three distinct blocks and a three‐stage hierarchical regression was run. Thematic analysis, using Braun and Clarke's (2006) six‐phase methodology, was used to analyse participants' qualitative comments that identified potential barriers and solutions to their utilisation of health information, health care, and community health services. Thematic analysis was chosen to identify recurring themes within the responses that might reveal meaningful patterns. Some participants provided a single barrier or solution, whilst others listed as many as three. To maintain rigour, after Author 1 completed the analysis, Author 2 independently recoded 5% of the themes, and a comparative discussion occurred, prior to the final themes being decided.
RESULTS A total of 255 participants returned complete responses to the online survey and were included in the analyses. (A further 15 participants were excluded for completing none ( n = 5) or less than half ( n = 5) of the questions.) The demographic characteristics for the study sample can be found in Table . Approximately 80% of participants identified as female, the majority were highly educated (71%), holding a bachelor's degree or higher, most were from solely English‐speaking households (96%), and Australian citizens (88%). However, some variation in age and education was observed, and over half of participants (54%) reported having at least one chronic health condition. Descriptive statistics were calculated for each of the HLQ domains and can be found in Table . A higher average score implies less challenges, whereas a lower average score implies more challenges. Whilst these findings cannot be compared directly, these results are tabulated alongside data from the Australian Bureau of Statistics (ABS) data collected in the 2017–2018 NHS. It was found that Tasmanians had numerically lower scores and so potentially faced more HL challenges across all assessed domains than the national averages in the NHS. It should be noted, however, that without accessing the ABS complete data set, we cannot analyse this finding further to assess significance between this study and the Tasmanian and national ABS data. Summaries of the regression statistics for the final regression models that included statistically significant demographics are presented in Table . The two demographic characteristics that significantly predicted outcome variables (HLQ scores) were (1) having one or more chronic health conditions and (2) living in an area with a low Index of Relative Socio‐Economic Disadvantage (IRSD). In order to carry out the hierarchical regression, chronic conditions included the categories of 0, 1 or 2 or more chronic conditions, while IRSD included numerical values 1–5 (corresponding with the index of relative socioeconomic disadvantage where a low score indicates relatively greater disadvantage). These predictors were found to be the strongest in the domains of ‘Social support for health’ and ‘Navigating the health care system’, accounting for 15.3% and 13.8% of the variance respectively. Meaning that the more chronic health conditions an individual had were associated with lower HLQ scores and living in a lower IRSD are also associated with lower HLQ scores. No further factors explained a significant amount of variance for our participants. Thematic analysis of the participants' qualitative comments revealed the current challenges that individuals face when utilising health information, health care, and community health services. Overall, the 255 participants identified a total of 276 barriers, and came up with a total of 162 solutions. The main themes identified during analysis of the qualitative data can be found in Table . Themes that emerged in participants' comments regarding barriers were, from most to least commonly mentioned: (1) Availability and access to health care, (2) Lack of perceived support from the health care system, (3) Difficulty understanding and navigating health, and (4) Expense of health care. The themes which arose from the participants' identified solutions to the above barriers were, again in order from most to least commonly identified: (1) Increased availability and access to health care providers/services, (2) Better collaboration between primary health care and other services, (3) Subsidised health care services and funding and (4) Health promotion.
DISCUSSION The aims of this study were to describe current health literacy levels on five of nine domains of the HLQ and to assess how HL scores differ with social demographic characteristics in a Tasmanian sample. The HLQ was used to identify local HL strengths and challenges, while the qualitative questions identified barriers and proposed solutions regarding utilisation of health information, health care, and community health services. Participants with more HL challenges (lower HL scores) were those who had chronic health conditions or who lived in areas of disadvantage. Interventions focusing on the HL of individuals and the HL responsiveness of services may be able to overcome these barriers by incorporating some of these solutions provided by the local community. Other services or organisations may find the Ophelia process (health literacy assessment process) helpful to ensure their services respond to the HL strengths and challenges of their local community. Although it was not a specific aim of this study, the HLQ results in this study can be compared to those described in the NHS for both the Tasmanian and Australian populations. As expected, the mean scores in the current Tasmanian data set were numerically lower on all five domains than the national data. This indicates that Tasmanians may have more challenges than their national counterparts when it comes to actively managing their health, having access to sufficient information, and having enough social support for their health. This finding aligns with the 2018 HLQ data presented in the Optimising Health Care for Tasmanians report. The report outlines several recommendations, which include: Understanding and responding to health literacy diversity; Optimising the reach and impact of health messaging; Improving health literacy responsiveness; Co‐designing health literacy actions; Local Ophelia projects; Evaluation and monitoring. Given the finding that IRSD had an important relationship with the HL strengths and challenges of our participants, our research reinforces the importance of understanding and responding to health literacy diversity, specifically taking into account the IRSD of the individuals and community the health literacy intervention is being designed to support. Further, if co‐design is employed to obtain and include local wisdom in the development of solutions, then it is important to consider how IRSD may impact meaningful contributions. The recommendation from the Optimising Health Care for Tasmania report that health literacy responsiveness be improved is an important one, and our participants reported that they, too, would like to see ‘health care servicing the community’, not the individuals having to work for the care they deserve. Finally, this study provides an example of a local project that follows the Ophelia process to support Tasmanians to understand their health literacy strengths and challenges and generate data‐driven solutions. Whilst the current study's participants were relatively highly educated compared to the ABS data set outlined in the Optimising Health Care for Tasmanians report, our sample is still experiencing certain HL challenges. Those individuals who were less likely to respond and didn't respond to our survey (e.g., minority groups and those of a lower education status) are likely to experience more HL challenges and thus may require significantly more support. , This has important implications for policy makers and health professionals providing services to the Tasmanian community. Specifically, this body of research will inform the future development of a HLM codesigned with local community to address the inequities that exist. The regression models demonstrated that having one or more chronic health conditions accounted for the largest proportion of the variance in terms of HL challenges, as reflected across four out of five of the HLQ domains assessed. There are two ways in which this finding could be interpreted. First, it is possible that greater HL challenges mediate the development of chronic diseases. An alternative explanation is that when people suffer from chronic diseases, they rely on their current HL assets to manage their conditions, and thus challenges are more readily identified. These ideas have been discussed elsewhere in the literature. For example, Friis et al. (2016) found that people with long‐term health conditions reported more difficulties than the general population in understanding health information and actively engaging with health care providers. Other authors have concluded that HL plays a crucial role in chronic disease management and prevention. , Another consideration is that people with chronic health conditions must navigate the health care system relatively often, and thus may encounter barriers and difficulties that are not encountered by those without chronic disease. One study concluded that HL is not just an individual skill but is also highly dependent on the accessibility of the health care system, the communication skills of health care professionals, and the level of complexity of the health information. Therefore, for a HL intervention specific for chronic conditions, it is important that a multidisciplinary team of health professionals work together with community members to provide ongoing support and care. The strength of a team approach and its provision of a range of different skills can meet the varying needs of individuals, particularly in relation to their values and perspectives on their chronic conditions. The regression models demonstrated that greater challenges across HLQ domains were also predicted by living in an area with greater relative socio‐economic disadvantage. It has been identified that disadvantaged social and socio‐economic conditions contribute to low HL levels and poorer health outcomes. , Furthermore, by definition, areas of low IRSD also have poor levels of educational attainment, which is an important determinant of HL. In 2021, Schillinger took this concept one step further and created a framework that describes two primary pathways that generate consequences for health outcomes, based in part, on HL. Tasmanian research on this topic has pointed to the HL challenges (mainly around access) that people are facing in low IRSD areas as a contributing factor in preventable hospital admissions for patients suffering from chronic health conditions. This research has alluded to a connection between HL, chronic health conditions, and socio‐economic disadvantage. This relationship, which has been shown in Figure , is well attested to in the literature and is referred to as the ‘social economic gradient of health’. , Here, there is a tri‐directional relationship between these factors, meaning that they are all related to, and influence, each other. For example, Tasmanians living in areas of low IRSD have more challenges associated with obtaining, processing, and understanding basic health information and services needed to make appropriate health decisions. This in turn increases both their risk of developing a chronic disease, and the severity of that disease. , This is especially problematic because these people are more likely to experience worse health outcomes, inadequate educational attainment and use health services suboptimally. , , It could be argued that these individuals are most in need of services and support. However, the way the health system is currently designed does not specifically target or make services more accessible to such groups. These challenges are mirrored in other regional areas within Australia. Further, this set of problems has also been described in the international literature and raised as an issue that deserves attention by many experts in this field. , Other demographic characteristics have been noted in the literature as predictors of HL challenges. These include level of education, employment status, identification as an Aboriginal or Torres Strait Islander, and speaking languages other than English at home. , , Due to the number of participants and the relative lack of variety in participants' demographics in the current sample, some of these other important demographics were not demonstrated, probably because there was not enough representation to be able to find significant results. Although these characteristics did not contribute significantly to predicting HLQ scores in our participants, it is still critical that they are not ignored. For example, higher educational attainment is associated with higher HL. , It is known that higher educational attainment can then play a significant role in mediating the relationship between HL and health behaviours. Consistent with the Ophelia principles, when seeking to overcome barriers associated with HL it is essential to consider the lived experience and context of the individual. We identified a number of barriers from participants' responses about interacting with health system: availability and access to health care, expense of health care, lack of perceived support from the health care system, and difficulty in understanding and navigating health. Current literature has recognised a similar set of barriers. , We also identified a set of solutions emerging from participants' responses: the need for increased availability and access to health care providers/services, subsidised health care services and funding, better collaboration between primary health care and other services, and an emphasis on health promotion. Each strategy may be utilised to inform change or reform to the health system. This will require a systems approach. The multidimensional nature of HL means that it requires a collaborative approach. As such, those in health, education, and community sectors could benefit from working more collaboratively. The results from this study have highlighted the disparities in HL assets amongst people within one community. Consistent with the Ophelia principles that informed this research, these results could be used to guide future investigations into the most effective interventions to assist in enhancing the HL assets of individuals and communities to improve overall health. A prominent finding of the study is that suffering a chronic health condition or living in an area of low IRSD were predictors of experiencing greater HL challenges across all measured HL domains. This result suggests that interventions aimed at addressing HL and health inequities would do well to be situated in disadvantaged communities and be made available and accessible to those who live with chronic health conditions. Current interventions and strategies at the local, national, and international level have been implemented in various forms. These strategies acknowledge the importance of HL, and its centrality to health inequalities, given that disadvantaged groups are most at risk of poor health behaviours and outcomes. All of these interventions aim to increase people's control over their own health by assisting them to understand and traverse the health care system to decrease health risk factors and improve health outcomes. However, these initiatives do not comprehensively address all the challenges outlined by our participants. As identified in this paper, strategies, interventions, and policy reform should focus on community members living with socio‐economic disadvantage, chronic health conditions or both (as described by the tri‐directional relationship). New initiatives could also address the barriers regarding utilisation of health information, health care, and community health services, which include enhancing the availability of services, providing assertive outreach style support, and developing community understanding of health care and the health care system. A wide range of stakeholders have an important role to play in strengthening HL, as identified by the International Union for Health Promotion and Education (IUHPE) Position Statement on Health Literacy, which emphasises the necessity of a systems approach to HL, underpinned by global, national, regional and local policies. Results from this study do need be interpreted with some degree of caution. Participants were recruited via a convenience sampling approach, which led to limitations with regards to the sampling and demographics of participants. We approached local community groups and organisations including the Migrant Resource Centre, 26TEN (an organisation that supports adult numeracy and literacy development), and the Tasmanian Health Literacy Network to encourage broad participation throughout Tasmania. Despite these efforts, the demographic characteristics were still significantly skewed. Specifically, the majority of participants were Australian‐born, English‐speaking, tertiary‐educated, middle‐aged females. This indicates that the survey was unable to capture the responses of many members of important minority groups, such as culturally and linguistically diverse (CALD) communities. Thus, the results are not representative of the experiences of all members of the Tasmanian community. Future research should aim to recruit a larger proportion of these underrepresented groups (e.g., telephone survey, in‐person survey support, translator assisted survey completion). Future work could also use a longer survey to encompass all domains of the HLQ. These limitations were also exacerbated due to the data collection process being conducted during a time where the impacts of the COVID‐19 pandemic were still being felt across Australia. Consequently, the potential for pandemic‐related impacts on HL assets should not be disregarded. As this research is just one part of a larger project following the Ophelia approach, subsequent papers will use these finding to formulate case‐based discussions to inform a future role of a Health Literacy Mediator for the community.
CONCLUSION In the small island state of Tasmania, individuals from different communities are experiencing ongoing challenges when it comes to the utilisation of health information, health care, and community health services. This alone can contribute to poorer health outcomes for both the individual and their community. This paper identified that those living in areas of greater socio‐economic disadvantage face greater HL challenges than their peers across the HLQ domains of 3, 4, 6 and 7. Whilst those living with one or more chronic health conditions face greater HL challenges for the HLQ Domains 2, 4, 6 and 7. Finally, the results help to identify the participant‐identified key barriers and potential solutions to the local population's health challenges. These findings could be used to develop targeted HL interventions designed to respond to the HL strengths and challenges identified within this population.
This research was made possible by a PhD scholarship from the College of Health and Medicine from the University of Tasmania awarded to M. Spencer. All the other authors have received no financial support for the research, authorship, and/or publication of this article.
Dr R Nash is Editorial Board member of Health Promotion Journal of Australia and co‐author of this article. To minimise bias, they were excluded from all editorial decision‐making related to the acceptance of this article for publication. Other authors have no conflicts of interest declared.
University of Tasmania Human Research Ethics Committee (H0026170).
|
Impact of Socio-demographic Characteristics on Time in Outpatient
Cardiology Clinics: A Retrospective Analysis | 45fcc136-56b2-4e94-9c9d-e89c2d99b02c | 10021097 | Internal Medicine[mh] | Time spent accessing healthcare is a key measure of service quality and
strain. , Elective surgery waiting times are the focus of most analyzes, and have
increased in recent years. However, patients wait in multiple settings—in the community
for primary care, specialist, , and allied health appointments,
and in waiting rooms in emergency and ambulatory
clinics. Compared to elective surgery, these other waiting times are
poorly characterized, providing clinicians and policy makers with an incomplete view
of patient time burden across healthcare systems. This burden is greatest for patients with multiple comorbid conditions, such as
cardiovascular diseases, who require increased healthcare contact. There is
international evidence that elective surgery waiting times are greater for patients
of lower socio-economic status (SES). - This is particularly
concerning in single payer health systems where waiting time should be allocated
according to clinical acuity, rather than ability to pay. However, there are few
studies on patient time burden in other settings. Particularly, there are a lack of data on time spent accessing ambulatory care and in
outpatient clinic waiting rooms. Such time may seem less significant as an absolute,
but cumulates with increasing healthcare contact and has an associated opportunity
cost secondary to missed work hours, estimated at 15 cents per dollar spent on
healthcare. The largest reports on waiting room time are from the USA
and indicate a likely time of 20 to 40 min. , , Some studies suggest patients
from lower socio-economic backgrounds wait longer in this setting as well. An
analysis of 3787 responses to the American Time Use Survey by Ray et al found time
accessing outpatient care was 123 min on average and significantly longer for Black
and Hispanic patients, those with less education, and the unemployed. Oostrom et
al analyzed 21 million outpatient office visits in the USA, finding publicly insured
(Medicaid) patients were 20% more likely than privately insured patients to wait
longer than 20 min. A small 2022 analysis of 423 attendees to a public outpatient
clinic in Ethiopia found those with lower educational attainment were more likely to
have long waiting times than tertiary-educated participants (odds ratio 2.25 [95% CI
1.11, 4.58]). A study of 96 patients in a Nigerian outpatient department
found women were more likely to experience waiting times of ≥180 min than men (31.6%
vs 6.3%, respectively). While these data suggest a relationship may exist, to our
knowledge, there are no studies comparing clinic time with SES in single-payer
healthcare systems such as the UK, Canada, or Australia. In this study, we present data from consecutive patients attending outpatient
cardiology appointments across 3 public hospitals in Sydney, Australia between 2014
and 2019. We aim to describe the “clinic time” (difference between time arrived and
time departed) and assess whether this is impacted by socio-demographic
characteristics including SES, age, gender, number of comorbidities, country of
birth, and language spoken at home.
Setting and Study Population We examined a consecutive patient-level data set of all public outpatient
cardiology encounters across 3 hospitals within Western Sydney Local Health
District (WSLHD) between July 2014 and December 2019. Clinics are consultant-led
and staffed by junior doctors, training cardiologists, and nursing staff.
Patients are referred by general practitioners, emergency departments, or other
doctors and generally do not pay to access these clinics. WSLHD comprises 5
hospitals, 7 community health centers, and serves 946 000 residents in the
western suburbs of Sydney. The population is diverse
with 46.8% of residents born overseas and 50.3% speaking a language other than
English. WSLHD also houses the largest Aboriginal and Torres Strait Islander
population in Australia (approximately 13 000 persons). Inclusion and Exclusion Criteria All adult (>18) patients who accessed outpatient cardiology services in-person
across WSLHD between July 2014 and December 2019 were included in the
analysis. Patients were excluded if their clinic time was not assessed. This was defined if
clinic time data were missing, equal to 0, or if all patients within a clinic
were allocated to a pre-specified time (eg, 30, 45, or 60 min). Extreme values
were excluded with cut-offs of ≤20 min (the presumed time of a consultation
only), or ≥240 min (the entire duration of a morning or afternoon clinic
session) as these times were likely due to data entry error or unreliable
clerical processes. Audio and inpatient consultations were excluded. Data Collection, Handling, and Definitions The data were cleaned, de-identified and processed by the Business Analytics
Service (BAS) at Westmead Hospital and passed to the Westmead Applied Research
Center, University of Sydney, via a secure server. The data contained
patient-level variables on age, gender, Indigenous status, country of birth,
language spoken at home, number of comorbidities, and postcode. Data on country
of birth, Indigenous status, and language spoken is obtained from all patients
via self- report on presentation to hospital. Patient postcode was correlated
with the 2016 socio-economic indexes for areas (SEIFA) Index of relative
socio-economic disadvantage (IRSD) score. This score is derived from 2016
Australian census data and summarizes variables that indicate relative
disadvantage. The lower the score, the higher proportion of disadvantaged people
reside within the postcode of interest. IRSD deciles were applied
to each patient for the final analysis. In addition, the data contained
appointment-level information on time of day, visit type (new or follow up),
referrer (emergency department or other), clinic type (arbitrarily categorized
A-R for consultant and hospital anonymity), arrived time, and departed time. Total clinic time was calculated by measuring the difference between time arrived
and time departed. This is a convenience measure taken by administration staff
as part of the normal clinic workflow. Statistical Analysis Statistical analysis was undertaken using R statistical software (V3.6.1). All
variables of interest were first interrogated visually to assess for normality
of distribution. Means were calculated for normally distributed continuous
variables, and medians for non-normal continuous variables. Categorical
variables were presented as frequencies and percentages. Initially, the proportion of patients waiting longer than the median clinic time
in different demographic groups (Age ≥75 vs <75, IRSD ≤5 vs >5, ≥4
comorbidities vs <4, female vs male, Indigenous vs non-Indigenous, born in
Australia vs born Overseas, and English vs other language spoken at home) was
compared with a chi-squared test. A univariate unadjusted linear regression was
then conducted on the above patient characteristics and clinic process measures
(clinic, visit type (new/follow up), referrer, appointment year, and time of
day) to determine variables associated with increased clinic time. A cox proportional hazard model was then applied to identify patient-level
predictors of increased time in clinic. The model outcome was the time the
patient left clinic. A higher hazard ratio (HR) described greater chance of
leaving clinic earlier and hence shorter total time in clinic. This analytic
approach was selected due to the non-normal distribution of the time data and is
similar to cox proportional hazard models applied to assess time to wound
healing, where a higher HR corresponds to a better outcome. Multivariate models controlled for clinic, visit type, referral source, and the
above demographic characteristics. Results of these models are presented as HRs
with 95% confidence interval (CIs). Further analysis was conducted to identify
interactions between patient and clinic-level variables of interest. Finally,
within-hospital and within-clinic (shorter wait versus longer wait) analysis was
conducted to determine whether discrepancies could be accounted for by
between-hospital and clinic differences.
We examined a consecutive patient-level data set of all public outpatient
cardiology encounters across 3 hospitals within Western Sydney Local Health
District (WSLHD) between July 2014 and December 2019. Clinics are consultant-led
and staffed by junior doctors, training cardiologists, and nursing staff.
Patients are referred by general practitioners, emergency departments, or other
doctors and generally do not pay to access these clinics. WSLHD comprises 5
hospitals, 7 community health centers, and serves 946 000 residents in the
western suburbs of Sydney. The population is diverse
with 46.8% of residents born overseas and 50.3% speaking a language other than
English. WSLHD also houses the largest Aboriginal and Torres Strait Islander
population in Australia (approximately 13 000 persons).
All adult (>18) patients who accessed outpatient cardiology services in-person
across WSLHD between July 2014 and December 2019 were included in the
analysis. Patients were excluded if their clinic time was not assessed. This was defined if
clinic time data were missing, equal to 0, or if all patients within a clinic
were allocated to a pre-specified time (eg, 30, 45, or 60 min). Extreme values
were excluded with cut-offs of ≤20 min (the presumed time of a consultation
only), or ≥240 min (the entire duration of a morning or afternoon clinic
session) as these times were likely due to data entry error or unreliable
clerical processes. Audio and inpatient consultations were excluded.
The data were cleaned, de-identified and processed by the Business Analytics
Service (BAS) at Westmead Hospital and passed to the Westmead Applied Research
Center, University of Sydney, via a secure server. The data contained
patient-level variables on age, gender, Indigenous status, country of birth,
language spoken at home, number of comorbidities, and postcode. Data on country
of birth, Indigenous status, and language spoken is obtained from all patients
via self- report on presentation to hospital. Patient postcode was correlated
with the 2016 socio-economic indexes for areas (SEIFA) Index of relative
socio-economic disadvantage (IRSD) score. This score is derived from 2016
Australian census data and summarizes variables that indicate relative
disadvantage. The lower the score, the higher proportion of disadvantaged people
reside within the postcode of interest. IRSD deciles were applied
to each patient for the final analysis. In addition, the data contained
appointment-level information on time of day, visit type (new or follow up),
referrer (emergency department or other), clinic type (arbitrarily categorized
A-R for consultant and hospital anonymity), arrived time, and departed time. Total clinic time was calculated by measuring the difference between time arrived
and time departed. This is a convenience measure taken by administration staff
as part of the normal clinic workflow.
Statistical analysis was undertaken using R statistical software (V3.6.1). All
variables of interest were first interrogated visually to assess for normality
of distribution. Means were calculated for normally distributed continuous
variables, and medians for non-normal continuous variables. Categorical
variables were presented as frequencies and percentages. Initially, the proportion of patients waiting longer than the median clinic time
in different demographic groups (Age ≥75 vs <75, IRSD ≤5 vs >5, ≥4
comorbidities vs <4, female vs male, Indigenous vs non-Indigenous, born in
Australia vs born Overseas, and English vs other language spoken at home) was
compared with a chi-squared test. A univariate unadjusted linear regression was
then conducted on the above patient characteristics and clinic process measures
(clinic, visit type (new/follow up), referrer, appointment year, and time of
day) to determine variables associated with increased clinic time. A cox proportional hazard model was then applied to identify patient-level
predictors of increased time in clinic. The model outcome was the time the
patient left clinic. A higher hazard ratio (HR) described greater chance of
leaving clinic earlier and hence shorter total time in clinic. This analytic
approach was selected due to the non-normal distribution of the time data and is
similar to cox proportional hazard models applied to assess time to wound
healing, where a higher HR corresponds to a better outcome. Multivariate models controlled for clinic, visit type, referral source, and the
above demographic characteristics. Results of these models are presented as HRs
with 95% confidence interval (CIs). Further analysis was conducted to identify
interactions between patient and clinic-level variables of interest. Finally,
within-hospital and within-clinic (shorter wait versus longer wait) analysis was
conducted to determine whether discrepancies could be accounted for by
between-hospital and clinic differences.
Of 37 456 patients assessed for eligibility, 14 823 were excluded and 22 367 were
included in the final analysis . Of these, 14 925 (65.9%) were male and the mean age was 61.4
(SD 15.2) years. Only 7823 (35.0%) were born in Australia, and 8452 (37.8%) were in
the lowest IRSD decile, indicating they resided in a postcode with a greater
proportion of disadvantaged residents than 90% of postcodes in Australia. A
significant proportion of patients had >4 comorbidities (40.4%). Cardiac risk
factors and comorbid cardiac conditions were also relatively common . Time Spent in Clinic The median total time in clinic was 84 min (interquartile range 58-130). The
distribution was flat across the years of observation, ranging from 69 min in
2014 to 101 min in 2017 . Process Measures as Predictors of Longer Time in Clinic Clinic process measures were analyzed for their association with clinic time. New
patients and those referred from the emergency department were the most likely
to spend longer in clinic (median 120 and 125 min, respectively, ). There was
significant variance between clinics . Linear regression
demonstrated low to moderate association between all process measures and clinic
time besides year of appointment and time of day . Visit type, clinic, and
referral source account for 23.0%, 35.0%, and 20.0% of the variance (R ) in clinic
time, respectively. Patient-Level Predictors of Time in Clinic All patient-level variables were assessed for their correlation with clinic time
in a multivariate cox proportional hazards model controlling for clinic,
referral source and visit type. In the unadjusted model, low (IRSD ≤ 5th decile)
SES patients spent less time in clinic than those of high (IRSD > 5th decile)
SES (median 66 min vs 109 min, ). After adjustment, this was no longer significant (HR 1.02
[0.99-1.06]). Those older than 75 were less likely to leave the clinic (HR 0.94
[0.90-0.97). The relationship between all other sociodemographic characteristics
did not reach significance after adjustment . Interaction Analysis of Demographic, Process Measures, and Socio-Economic
Status Further analysis was performed assessing the interaction between SES, patient
characteristics and clinic process measures. Those of lower SES spent less time
in clinic irrespective of their age, gender, number of comorbidities, country of
birth or language spoken at home. However, after adjustment for visit type,
clinic, and referral source, there was no interaction between SES and any of the
identified demographic variables ( Supplemental Table 1 ). Patients of lower SES were more likely to
attend follow-up appointments (77.2% vs 57.6%), clinics with short clinic time
(66.8% vs 21.1%) and be referred from sources other than the emergency
department, compared to patients of higher SES ( Supplemental Table 1 ). Clinic and Hospital Sub Analysis To assess for discrimination within hospitals and clinics, the association
between socio-economic status and time in clinic was analyzed in a further cox
proportional hazards model adjusted for clinic, referral source and visit type.
Those of lower SES spent slightly less time in clinics in hospital C (57 min vs
60 min, HR 1.24 [1.13-1.37]), though there were no differences within other
hospitals. Within short wait clinics, lower SES spent less time in clinic
(59 min vs 71 min, HR 1.10 [1.05-1.17]). There was no difference according to
SES in longer wait clinics ( Supplemental Table 2 ).
The median total time in clinic was 84 min (interquartile range 58-130). The
distribution was flat across the years of observation, ranging from 69 min in
2014 to 101 min in 2017 .
Clinic process measures were analyzed for their association with clinic time. New
patients and those referred from the emergency department were the most likely
to spend longer in clinic (median 120 and 125 min, respectively, ). There was
significant variance between clinics . Linear regression
demonstrated low to moderate association between all process measures and clinic
time besides year of appointment and time of day . Visit type, clinic, and
referral source account for 23.0%, 35.0%, and 20.0% of the variance (R ) in clinic
time, respectively.
All patient-level variables were assessed for their correlation with clinic time
in a multivariate cox proportional hazards model controlling for clinic,
referral source and visit type. In the unadjusted model, low (IRSD ≤ 5th decile)
SES patients spent less time in clinic than those of high (IRSD > 5th decile)
SES (median 66 min vs 109 min, ). After adjustment, this was no longer significant (HR 1.02
[0.99-1.06]). Those older than 75 were less likely to leave the clinic (HR 0.94
[0.90-0.97). The relationship between all other sociodemographic characteristics
did not reach significance after adjustment .
Further analysis was performed assessing the interaction between SES, patient
characteristics and clinic process measures. Those of lower SES spent less time
in clinic irrespective of their age, gender, number of comorbidities, country of
birth or language spoken at home. However, after adjustment for visit type,
clinic, and referral source, there was no interaction between SES and any of the
identified demographic variables ( Supplemental Table 1 ). Patients of lower SES were more likely to
attend follow-up appointments (77.2% vs 57.6%), clinics with short clinic time
(66.8% vs 21.1%) and be referred from sources other than the emergency
department, compared to patients of higher SES ( Supplemental Table 1 ).
To assess for discrimination within hospitals and clinics, the association
between socio-economic status and time in clinic was analyzed in a further cox
proportional hazards model adjusted for clinic, referral source and visit type.
Those of lower SES spent slightly less time in clinics in hospital C (57 min vs
60 min, HR 1.24 [1.13-1.37]), though there were no differences within other
hospitals. Within short wait clinics, lower SES spent less time in clinic
(59 min vs 71 min, HR 1.10 [1.05-1.17]). There was no difference according to
SES in longer wait clinics ( Supplemental Table 2 ).
This analysis of over 20 000 consecutive outpatient cardiology clinic encounters
aimed to determine whether those of low SES were more likely to spend longer in
clinic. After adjusting for visit type, clinic, and referral source, there was no
difference in clinic time according to SES. Overall, 75% of patients spent at least
1hour in clinic. One quarter spent more than 2 hours. Potential implications of
these findings include consideration of a more productive use of this time in
ambulatory clinics, such as implementing interventions during this time that can
improve health literacy and may improve health outcomes and satisfaction with health
services. , The interaction between SES and time to accessing health services has been debated
for over 20 years. Most data are derived from elective surgery waiting
lists, , and there is some evidence discrimination is reversing as new
policies are introduced. Cooper et al analyzed elective surgery
wait lists in 1997 to 2000, 2001 to 2004, and 2005 to 2007, finding the effect of
SES on waiting time reduced over the period of observation and reversed for knee
replacement and cataract repair in 2005 to 2007, such that the most deprived fifth
waited less than the least deprived fifth. There are less studies of the Australian
system, but most reports suggest discrimination. Johar et al studied
90 162 patients in New South Wales public hospitals, finding that more advantaged
patients waited less for elective surgery at all quintiles of waiting time. Data
from developing countries is also suggestive of discrimination in this setting. A
2017 analysis of 219 surgeries within an Indian teaching hospital found those living
below the poverty line had threefold higher waiting times than those above the
poverty line. However, data are very limited within developing countries,
largely due to a lack of systematic reporting. For example, a recent international
collaboration for systematic reporting of waiting times is limited to organization
for economic co-operation and development (OECD) countries, which are almost
exclusively high-income. The finding of no relation to SES for patients accessing public clinics in our study
is reassuring and may be explained by several reasons. There are likely fewer
opportunities for preferential treatment within waiting rooms (where patients are
seen in the order they arrive) than elective surgery (where waiting time is
determined by clinician priority allocation), which may explain the lack of
association between SES and clinic time in our study. The Australian system is private-public, where patients with insurance that
anticipate a long wait time can opt-in for private hospital care. There is evidence
this preferential service selection model explains elective surgery waiting time
inequity in Australia, though more studies of waiting room time are needed. Many
hospitals in Australia run large public outpatient services where patients generally
do not pay out-of-pocket for services, which are the services analyzed here.
However, higher SES patients are more likely to access privately billed clinics in
the community and findings here may have limited applicability to these care
settings. They do however suggest that the lack of relation to SES of time spent in
public clinics found here may be because of the absence of per-patient payment and
of classification based on public/private status. Patients with cardiovascular disease are more likely to be older, Indigenous, of
lower SES, live in rural areas and have comorbidities than the general
population. Analysis of time in cardiology clinics provides an
opportunity to assess for poorer outcomes among these patient populations. In our
study, we found patients older than 75 were more likely to spend longer in
cardiology clinics. This may be due to these patients having more complex care needs
requiring a longer consultation with additional time to see other health
professional, for example, nurses, allied health workers, social workers. Older
patients may also be more likely to arrive early to clinic appointments, increasing
the overall appointment time. Faiz and Kristoffersen collected data from 1353
outpatient neurology clinic appointments and found older patients were less likely
to arrive late than younger patients (OR 0.74 [0.63-0.88]). In our study, lower SES patients were more likely to attend follow-up appointments
and clinics with shorter waits overall, both strong predictors of reduced total
clinic time. Sub-analysis of these clinics found lower SES patients spent less time
after adjusting for process measures. Importantly, our analysis did not delineate
between consultation and waiting room time. It is possible that lower SES patients
had shorter consult times, which was the primary driver for a shorter total clinic
time. This is supported by an analysis of 70 758 GP consultations in Australia in
2001 to 2002, which found older patients of higher SES had longer consultation
times. A 2020 qualitative analysis of 36 head and neck cancer
appointments found lower SES patients were more passive in their care, engaging in
less agenda setting and information seeking, potentially explaining shorter
consultation times within this group. Further studies are needed to
better define patient time burden while waiting, an indicator of poor care, from
time spent with clinicians, likely an indicator of quality care. The implications of “in-clinic” waiting times are different to those for elective
surgery, specialist and primary care visits, where longer waiting time has been
associated with poorer clinical outcomes. - Increased time in ambulatory
care has been linked to reduced care satisfaction, however the consequences are
primarily economic – the opportunity cost of accessing healthcare. Increasing
workforce casualization, where employees do not have access to sick leave, further
compounds the economic cost of increased clinic time. These implications are
greater for patients that require more contact with healthcare services. Addressing Patient Waiting Time—What Approaches Are Needed? Several methods have been trialed to reduce the time patients spend accessing
healthcare. In the emergency department, the introduction of 4-hour targets in
the UK, Australia and other countries has seen significant reductions in waiting
times. However, there may be diminishing returns from further
reductions. Sullivan et al present an analysis of
12.5 million emergency department episodes of care, finding compliance with
waiting time targets reduced in-hospital mortality. However as compliance
increased past a critical point of 83%, the relationship was lost. Countries
that lack a benchmark likely have even longer waiting times. A 2006 analysis of
675 patients at a public hospital in Barbados revealed a median 377 min length
of stay, over 2 hours longer than targets in Australia and the UK. Despite
some small studies in China, Singapore, and
Korea, there is a paucity of research about interventions to
address in-clinic waiting time. To our knowledge, there are no examples of such
interventions within cardiology outpatient clinics. Irrespective of between-group differences, this study underscores that time spent
accessing healthcare is significant. This time could be better utilized to
deliver health interventions that convert this from wasted to productive time.
There is some literature suggesting waiting room interventions can improve
patient knowledge, but a paucity of robustly designed studies to assess the
efficacy of waiting room interventions on clinical outcomes. , Though a
focus on health outcomes is desirable, waiting room interventions could also
target process outcomes such as patient satisfaction with care, total time in
clinic or consultation time. Integrated delivery of tech-enabled interventions
that begin in the waiting room, continue through the consultation and into the
post-consultation period could contribute to a new paradigm of healthcare that
values patient time whilst also increasing provider efficiency. There are several strengths and weaknesses to this study. We considered a
consecutive sample of patients attending a single specialty within one local
health district. This limited between-hospital and specialty heterogeneity,
however provided limited view on waiting times in rural locations, other cities
and specialties. Data were collected over 5 years, providing insight into
longitudinal waiting time trends within our sample and were convenience based
and likely less prone to bias than data collected by self-report or specifically
measured for the monitoring of waiting time. The convenience nature of these
data also limits generalizability. Approximately 40% of encounters where data
were incomplete or unreliable were excluded to minimize impact on findings
. We did
not have differential data on time spent with clinicians versus in waiting rooms
and could not identify patients that left clinic without being seen by a doctor.
We were unable to characterize the urgency of each patient’s clinic visit and
cannot rule out an effect due to preferential treatment of higher acuity
patients. A sample size calculation was also not performed in this study. All
available data in the sample were analyzed. Finally, data were at the level of
the encounter, not the patient. It is possible there are duplicate patients who
attended clinics multiple times within the data set.
Several methods have been trialed to reduce the time patients spend accessing
healthcare. In the emergency department, the introduction of 4-hour targets in
the UK, Australia and other countries has seen significant reductions in waiting
times. However, there may be diminishing returns from further
reductions. Sullivan et al present an analysis of
12.5 million emergency department episodes of care, finding compliance with
waiting time targets reduced in-hospital mortality. However as compliance
increased past a critical point of 83%, the relationship was lost. Countries
that lack a benchmark likely have even longer waiting times. A 2006 analysis of
675 patients at a public hospital in Barbados revealed a median 377 min length
of stay, over 2 hours longer than targets in Australia and the UK. Despite
some small studies in China, Singapore, and
Korea, there is a paucity of research about interventions to
address in-clinic waiting time. To our knowledge, there are no examples of such
interventions within cardiology outpatient clinics. Irrespective of between-group differences, this study underscores that time spent
accessing healthcare is significant. This time could be better utilized to
deliver health interventions that convert this from wasted to productive time.
There is some literature suggesting waiting room interventions can improve
patient knowledge, but a paucity of robustly designed studies to assess the
efficacy of waiting room interventions on clinical outcomes. , Though a
focus on health outcomes is desirable, waiting room interventions could also
target process outcomes such as patient satisfaction with care, total time in
clinic or consultation time. Integrated delivery of tech-enabled interventions
that begin in the waiting room, continue through the consultation and into the
post-consultation period could contribute to a new paradigm of healthcare that
values patient time whilst also increasing provider efficiency. There are several strengths and weaknesses to this study. We considered a
consecutive sample of patients attending a single specialty within one local
health district. This limited between-hospital and specialty heterogeneity,
however provided limited view on waiting times in rural locations, other cities
and specialties. Data were collected over 5 years, providing insight into
longitudinal waiting time trends within our sample and were convenience based
and likely less prone to bias than data collected by self-report or specifically
measured for the monitoring of waiting time. The convenience nature of these
data also limits generalizability. Approximately 40% of encounters where data
were incomplete or unreliable were excluded to minimize impact on findings
. We did
not have differential data on time spent with clinicians versus in waiting rooms
and could not identify patients that left clinic without being seen by a doctor.
We were unable to characterize the urgency of each patient’s clinic visit and
cannot rule out an effect due to preferential treatment of higher acuity
patients. A sample size calculation was also not performed in this study. All
available data in the sample were analyzed. Finally, data were at the level of
the encounter, not the patient. It is possible there are duplicate patients who
attended clinics multiple times within the data set.
Accessing healthcare presents a significant time burden for patients at all levels of
the health system. In this analysis of 22 367 patients attending publicly funded
outpatient cardiology clinic appointments over 6 years, older patients spent longer
in clinic, but no difference for low SES or other demographically disadvantaged
patients was identified. This is reassuring, however does not exclude the
possibility of disparities. Further studies that are prospective and diverse in
geographical, health service funding, and economic advantage at a country level are
required. Ongoing monitoring of the health system with respect to performance and
inequities is also important. Consideration should be given to the opportunistic
delivery of interventions during this time to improve health engagement and
outcomes.
sj-docx-1-inq-10.1177_00469580231159491 – Supplemental material for
Impact of Socio-demographic Characteristics on Time in Outpatient Cardiology
Clinics: A Retrospective Analysis Click here for additional data file. Supplemental material, sj-docx-1-inq-10.1177_00469580231159491 for Impact of
Socio-demographic Characteristics on Time in Outpatient Cardiology Clinics: A
Retrospective Analysis by Daniel McIntyre, Simone Marschner, Aravinda
Thiagalingam, David Pryce and Clara K. Chow in INQUIRY: The Journal of Health
Care Organization, Provision, and Financing sj-docx-2-inq-10.1177_00469580231159491 – Supplemental material for
Impact of Socio-demographic Characteristics on Time in Outpatient Cardiology
Clinics: A Retrospective Analysis Click here for additional data file. Supplemental material, sj-docx-2-inq-10.1177_00469580231159491 for Impact of
Socio-demographic Characteristics on Time in Outpatient Cardiology Clinics: A
Retrospective Analysis by Daniel McIntyre, Simone Marschner, Aravinda
Thiagalingam, David Pryce and Clara K. Chow in INQUIRY: The Journal of Health
Care Organization, Provision, and Financing
|
Fibrostricturing Crohn's Disease Is Marked by an Increase in Active Eosinophils in the Deeper Layers | 76ec2cf4-2245-47d0-8492-a9584f8ff36e | 11272291 | Anatomy[mh] | Crohn's disease (CD) is characterized by a relapsing-remitting disease course, with transmural intestinal inflammation as a hallmark feature . Therefore, patients with CD are at risk of developing (small) bowel strictures for which surgical intervention is required . Moreover, stricture recurrence is common and cannot always be prevented, resulting in repetitive resections and the potential risk of developing short bowel syndrome . Research has primarily focused on mucosal inflammation, rather than on the resulting fibrostenosis. Due to continuous intestinal injury, growth factors, such as transforming growth factor β (TGF β), are released, which in turn stimulate fibroblast activation and differentiation toward myofibroblasts . Subsequently, excessive extracellular matrix deposition will occur, resulting in fibrosis . In the lung, liver, and intestine, it has already been demonstrated that myofibroblasts and active fibroblasts contribute to the development of fibrosis . However, targeting active fibroblasts and myofibroblasts should be approached with caution, as fibroblasts initiate and propagate wound healing . Of importance, there is a strong crosstalk between fibroblasts and various immune cells. Immune cells produce mediators that are capable of activating fibroblasts, while fibroblasts can reciprocally modulate the activity of immune cells . Although an involvement of several immune cells such as eosinophils, monocytes, dendritic cells, and T cells has been reported in fibrotic processes in other organs, the key immune cells and their mediators involved in the process of stricturing CD remain poorly understood. Colon et al have shown that a deletion of eosinophil peroxidase, a mediator solely produced by eosinophils, can decrease renal fibrosis development by decreasing α-smooth muscle actin levels and collagen I deposition in a murine model. In line, anti–interleukin (IL)-5–mediated eosinophil targeting in mice revealed that eosinophils also play a profibrotic role in hepatic fibrosis . Increased monocyte counts have previously been linked to worse outcomes in both idiopathic pulmonary and colonic fibrosis . Similarly, dendritic cells were revealed to contribute to fibrosis by activating myofibroblasts . Furthermore, the adaptive immune system has previously been described in fibrotic processes as well. In that context, T helper 2 (Th2) and regulatory T cells (Treg) were reported to have profibrotic characteristics, while Th1 cells rather act as anti-fibrotic cells . Although several reports about the involvement of these cells in the development of fibrosis have been published, they largely rely on murine studies and organs outside the gastrointestinal tract. A detailed characterization of the immune cells involved in human fibrostricturing CD is currently lacking. Therefore, we aimed to map the key immune cells contributing to fibrosis in patients with CD. Patient samples Resection specimens were prospectively collected through dermal punch biopsy from 25 patients with CD with fibrostricturing disease requiring ileocolonic resection and from 10 non-inflammatory bowel disease (IBD) controls diagnosed with colorectal cancer and in whom a right hemicolectomy was required. From the CD specimens, macroscopically unaffected, inflamed (ulceration), and fibrostenotic tissue (wall thickening, sampled at site of stricture) was obtained by an IBD specialized pathologist (G.D.H.), while in non-IBD controls, tissue was taken from the macroscopically unaffected resection margin at the terminal ileum. Written informed consent was obtained from each patient under an approved protocol by the University Hospitals of Leuven Ethics Committee Review Board (S53684). Isolation of mucosal and deeper layer intestinal leukocytes Mucosal layers were separated macroscopically from deeper layers, and processed within 1 hour of sample collection (Figure ). The tissue was incubated under magnetic stirring for 10 minutes at 37 °C in RPMI-1640 (Gibco, 21875-034) supplemented with 10% fetal bovine serum (Gibco, 10270-106), 1% penicillin streptomycin 10.000 U/mL (Gibco, 15140-122), 5 mM ethylenediaminetetraacetic acid (Invitrogen, 15575-038), and 2% 1M N-2-hydroxyethylpiperazine-N-2-ethane sulfonic acid (Gibco, 15630-056) to remove mucus and epithelial cells. The epithelial fraction was removed, and this step was repeated once. Afterward, the epithelial fraction was discarded, and the remaining tissue was cut into smaller fractions. The remaining sample was digested using 10 mL Hank's balanced salt solution with Ca 2+ Mg 2+ (Gibco, 24020-117) supplemented with 10 mg/mL collagenase type 4 (Worthington, LS004188), 0.2% Deoxyribonuclease I (Roche, 10104159001), and 2% 1M N-2-hydroxyethylpiperazine-N-2-ethane sulfonic acid at 37 °C for 50 minutes to 1 hour. The isolated single-cell fraction was filtered through a 70-μm cell strainer (Greiner Bio-one, 542070), after which the cells were counted and used for flow cytometry. Single-cell staining and flow cytometry Single cells were stained for viability with either Fixable Viability Dye eFluor 450 (Invitrogen, 65-0863-14) or Fixable Viability Dye eFluor 780 (Invitrogen, 65-0865-14) for 25 minutes at room temperature and protected from light. Cells were washed with phosphate-buffered saline (PBS) (Gibco, 14190-094) + 0.5% Bovine Serum Albumin (BSA) (Sigma-Aldrich, 9048-46-8) and centrifuged for 5 minutes at 400 g. The supernatant was discarded, and the pellet was resuspended in blocking mix (2% heat-inactivated plasma in PBS + 0.5% BSA) for 10 minutes at 4 °C. Cells were washed and centrifuged again and incubated in the antibody mixes (see Supplementary Table 1, http://links.lww.com/CTG/B118 ) for 30 minutes at 4 °C. Afterward, cells were washed again and incubated in 1% PBS-buffered formaldehyde (Merck, 30525-89-4) for 15 minutes at room temperature. Once more, cells were washed with PBS + 0.5% BSA + 2 mM ethylenediaminetetraacetic acid and stored in 4 °C until acquisition. Samples were acquired on a BD Biosciences (BD) LSR Fortessa Special Order Research Product using BD FACSDiva software version 8. The configuration can be found in Supplementary Digital Content (see Supplementary Table 2, http://links.lww.com/CTG/B118 ). Quality control was performed before each acquisition by using FACS-Diva CS&T Research Beads (BD, 655051). For fluorescence compensation settings, anti-rat/anti-hamster Ig,κ CompBeads (BD, 552845), anti-mouse Ig,κ CompBeads (BD, 552843), or MACS Comp Beads anti-REA (Miltenyi, 130-104-693) were used. Fluorescence minus one controls were included. Flow cytometry files were analyzed using the BD FlowJo 10 software. Gating strategies can be found in Supplementary Digital Content (see Supplementary Figures 1–3, http://links.lww.com/CTG/B118 ). Protein isolation and measurement Tissue was mechanically homogenized in PBS + 5% BSA with a Potter-Elvehjem homogenizer after which the homogenized tissue was centrifuged for 5 minutes at 10,000 g . The supernatant was removed and stored at −80 °C until further analysis. Eosinophil cationic protein (ECP) was measured through enzyme-linked immunosorbent assay (ELISA) according to the manufacturer's guidelines in duplicate (Abbexa, abx055137). IL-4, IL-5, IL-10, IL-12p70, IL-13, IL-18, basic fibroblast growth factor (bFGF), vascular endothelial growth factor, IL-1β, interferon (IFN)-γ, eotaxin-1,2,3, and TGF-β1,2,3 protein levels were measured with the mesoscale discovery (MSD) U-plex system according to the manufacturer's guidelines (Mesoscale Discovery). All protein levels were corrected for tissue weight. Tissue processing and immunohistochemistry Tissue was processed (see Supplementary Table 3, http://links.lww.com/CTG/B118 ) and embedded in paraffin after which 5-μm thick slides were cut. Slides were deparaffinized by incubating them 3 minutes in HistoChoice clearing agent (Sigma Aldrich, H2779) twice, 3 minutes in 50% HistoChoice-ethanol, 3 minutes in 100% ethanol (Merck, 64-17-5) twice, 3 minutes in 95% ethanol, 3 minutes in 70% ethanol, 3 minutes in 50% ethanol, and 3 minutes in demineralized water. Antigen retrieval was performed for 20 minutes at 95 °C using a sodium citrate buffer (10 mM sodium citrate [Merck, 61-32-04-3], 0.05% Tween 20 [Sigma-Aldrich, P9416], pH 9.0). Slides were permeabilized for 10 minutes using 0.3% Triton X-100 (Sigma-Aldrich, 9036-19-5) and 0.3M glycine (Merck, 56-40-6) in PBS after which blocking in 1% BSA in PBS-Tween 20 was performed for 1 hour. Primary antibodies were incubated overnight at 4 °C (see Supplementary Table 4, http://links.lww.com/CTG/B118 ). The following day, slides were washed 3 times in PBS-Tween 20 for 5 minutes after which secondary antibodies were added for 1 hour at room temperature and protected from light (see Supplementary Table 4, http://links.lww.com/CTG/B118 ). Slides were washed once for 10 minutes with PBS-Tween 20 after which autofluorescence quenching was performed according to the manufacturer's guidelines (Vector Laboratories, SP-8400-15). 4',6-Diamidino-2-phenylindole was added at room temperature for 15 minutes (see Supplementary Table 4, http://links.lww.com/CTG/B118 ) after which the slides were washed a last time for 10 minutes with PBS-Tween 20. Finally, the slides were mounted and sealed (Vector Laboratories, SP-8400-15). Images were captured with a 25× water immersion objective (0.8 NA; Zeiss) on an LSM 780 confocal microscope (Carl Zeiss Microscopy). Slides stained for immunohistochemistry were analyzed through ImageJ. Evaluation of colocalization Colocalization was assessed based on the immunohistochemical stainings described earlier. On a random selection of slides (3 patients with CD and 3 non-IBD controls), the shortest distance, determined through a straight line, between eosinophils and (active) fibroblasts was measured through ImageJ. On every slide, 20 shortest distances were measured in the mucosa and 20 shortest distances in the deeper layers, after which the median distance was calculated . Evaluation of inflammation and fibrosis Transverse sections (5 μm) were used for a Masson's Trichrome staining, as previously described . Collagen area was calculated through ImageJ. Statistical analysis GraphPad prism 9.4.0 was used to perform the statistical analysis. Normality was determined using a Shapiro-Wilk test, after which an unpaired analysis (unpaired t test or Mann-Whitney analysis) between the non-IBD controls and patients with CD was performed. Comparisons between tissue layers and regions within the patients with CD were conducted using a Friedmann test. Nonparametric spearman correlations were calculated. Data were represented as median (interquartile range). All analyses were corrected for multiple testing by the Bonferroni method. A corrected P value < 0.05 was considered statistically significant. Resection specimens were prospectively collected through dermal punch biopsy from 25 patients with CD with fibrostricturing disease requiring ileocolonic resection and from 10 non-inflammatory bowel disease (IBD) controls diagnosed with colorectal cancer and in whom a right hemicolectomy was required. From the CD specimens, macroscopically unaffected, inflamed (ulceration), and fibrostenotic tissue (wall thickening, sampled at site of stricture) was obtained by an IBD specialized pathologist (G.D.H.), while in non-IBD controls, tissue was taken from the macroscopically unaffected resection margin at the terminal ileum. Written informed consent was obtained from each patient under an approved protocol by the University Hospitals of Leuven Ethics Committee Review Board (S53684). Mucosal layers were separated macroscopically from deeper layers, and processed within 1 hour of sample collection (Figure ). The tissue was incubated under magnetic stirring for 10 minutes at 37 °C in RPMI-1640 (Gibco, 21875-034) supplemented with 10% fetal bovine serum (Gibco, 10270-106), 1% penicillin streptomycin 10.000 U/mL (Gibco, 15140-122), 5 mM ethylenediaminetetraacetic acid (Invitrogen, 15575-038), and 2% 1M N-2-hydroxyethylpiperazine-N-2-ethane sulfonic acid (Gibco, 15630-056) to remove mucus and epithelial cells. The epithelial fraction was removed, and this step was repeated once. Afterward, the epithelial fraction was discarded, and the remaining tissue was cut into smaller fractions. The remaining sample was digested using 10 mL Hank's balanced salt solution with Ca 2+ Mg 2+ (Gibco, 24020-117) supplemented with 10 mg/mL collagenase type 4 (Worthington, LS004188), 0.2% Deoxyribonuclease I (Roche, 10104159001), and 2% 1M N-2-hydroxyethylpiperazine-N-2-ethane sulfonic acid at 37 °C for 50 minutes to 1 hour. The isolated single-cell fraction was filtered through a 70-μm cell strainer (Greiner Bio-one, 542070), after which the cells were counted and used for flow cytometry. Single cells were stained for viability with either Fixable Viability Dye eFluor 450 (Invitrogen, 65-0863-14) or Fixable Viability Dye eFluor 780 (Invitrogen, 65-0865-14) for 25 minutes at room temperature and protected from light. Cells were washed with phosphate-buffered saline (PBS) (Gibco, 14190-094) + 0.5% Bovine Serum Albumin (BSA) (Sigma-Aldrich, 9048-46-8) and centrifuged for 5 minutes at 400 g. The supernatant was discarded, and the pellet was resuspended in blocking mix (2% heat-inactivated plasma in PBS + 0.5% BSA) for 10 minutes at 4 °C. Cells were washed and centrifuged again and incubated in the antibody mixes (see Supplementary Table 1, http://links.lww.com/CTG/B118 ) for 30 minutes at 4 °C. Afterward, cells were washed again and incubated in 1% PBS-buffered formaldehyde (Merck, 30525-89-4) for 15 minutes at room temperature. Once more, cells were washed with PBS + 0.5% BSA + 2 mM ethylenediaminetetraacetic acid and stored in 4 °C until acquisition. Samples were acquired on a BD Biosciences (BD) LSR Fortessa Special Order Research Product using BD FACSDiva software version 8. The configuration can be found in Supplementary Digital Content (see Supplementary Table 2, http://links.lww.com/CTG/B118 ). Quality control was performed before each acquisition by using FACS-Diva CS&T Research Beads (BD, 655051). For fluorescence compensation settings, anti-rat/anti-hamster Ig,κ CompBeads (BD, 552845), anti-mouse Ig,κ CompBeads (BD, 552843), or MACS Comp Beads anti-REA (Miltenyi, 130-104-693) were used. Fluorescence minus one controls were included. Flow cytometry files were analyzed using the BD FlowJo 10 software. Gating strategies can be found in Supplementary Digital Content (see Supplementary Figures 1–3, http://links.lww.com/CTG/B118 ). Tissue was mechanically homogenized in PBS + 5% BSA with a Potter-Elvehjem homogenizer after which the homogenized tissue was centrifuged for 5 minutes at 10,000 g . The supernatant was removed and stored at −80 °C until further analysis. Eosinophil cationic protein (ECP) was measured through enzyme-linked immunosorbent assay (ELISA) according to the manufacturer's guidelines in duplicate (Abbexa, abx055137). IL-4, IL-5, IL-10, IL-12p70, IL-13, IL-18, basic fibroblast growth factor (bFGF), vascular endothelial growth factor, IL-1β, interferon (IFN)-γ, eotaxin-1,2,3, and TGF-β1,2,3 protein levels were measured with the mesoscale discovery (MSD) U-plex system according to the manufacturer's guidelines (Mesoscale Discovery). All protein levels were corrected for tissue weight. Tissue was processed (see Supplementary Table 3, http://links.lww.com/CTG/B118 ) and embedded in paraffin after which 5-μm thick slides were cut. Slides were deparaffinized by incubating them 3 minutes in HistoChoice clearing agent (Sigma Aldrich, H2779) twice, 3 minutes in 50% HistoChoice-ethanol, 3 minutes in 100% ethanol (Merck, 64-17-5) twice, 3 minutes in 95% ethanol, 3 minutes in 70% ethanol, 3 minutes in 50% ethanol, and 3 minutes in demineralized water. Antigen retrieval was performed for 20 minutes at 95 °C using a sodium citrate buffer (10 mM sodium citrate [Merck, 61-32-04-3], 0.05% Tween 20 [Sigma-Aldrich, P9416], pH 9.0). Slides were permeabilized for 10 minutes using 0.3% Triton X-100 (Sigma-Aldrich, 9036-19-5) and 0.3M glycine (Merck, 56-40-6) in PBS after which blocking in 1% BSA in PBS-Tween 20 was performed for 1 hour. Primary antibodies were incubated overnight at 4 °C (see Supplementary Table 4, http://links.lww.com/CTG/B118 ). The following day, slides were washed 3 times in PBS-Tween 20 for 5 minutes after which secondary antibodies were added for 1 hour at room temperature and protected from light (see Supplementary Table 4, http://links.lww.com/CTG/B118 ). Slides were washed once for 10 minutes with PBS-Tween 20 after which autofluorescence quenching was performed according to the manufacturer's guidelines (Vector Laboratories, SP-8400-15). 4',6-Diamidino-2-phenylindole was added at room temperature for 15 minutes (see Supplementary Table 4, http://links.lww.com/CTG/B118 ) after which the slides were washed a last time for 10 minutes with PBS-Tween 20. Finally, the slides were mounted and sealed (Vector Laboratories, SP-8400-15). Images were captured with a 25× water immersion objective (0.8 NA; Zeiss) on an LSM 780 confocal microscope (Carl Zeiss Microscopy). Slides stained for immunohistochemistry were analyzed through ImageJ. Colocalization was assessed based on the immunohistochemical stainings described earlier. On a random selection of slides (3 patients with CD and 3 non-IBD controls), the shortest distance, determined through a straight line, between eosinophils and (active) fibroblasts was measured through ImageJ. On every slide, 20 shortest distances were measured in the mucosa and 20 shortest distances in the deeper layers, after which the median distance was calculated . Transverse sections (5 μm) were used for a Masson's Trichrome staining, as previously described . Collagen area was calculated through ImageJ. GraphPad prism 9.4.0 was used to perform the statistical analysis. Normality was determined using a Shapiro-Wilk test, after which an unpaired analysis (unpaired t test or Mann-Whitney analysis) between the non-IBD controls and patients with CD was performed. Comparisons between tissue layers and regions within the patients with CD were conducted using a Friedmann test. Nonparametric spearman correlations were calculated. Data were represented as median (interquartile range). All analyses were corrected for multiple testing by the Bonferroni method. A corrected P value < 0.05 was considered statistically significant. Patient characteristics Twenty-five patients with CD and 10 non-IBD controls were included. Clinical characteristics are summarized in Table . Assessment of collagen deposition To morphologically characterize tissue and assess fibrosis formation, collagen deposition was assessed in the unaffected tissue of both non-IBD controls and patients with CD and the inflamed and fibrotic regions of patients with CD. Both inflamed and fibrotic regions exhibited increased collagen deposition in comparison with the unaffected region of the non-IBD controls ( P = 0.002 and P < 0.0001, respectively) and patients with CD ( P = 0.002 and P < 0.0001, respectively). Collagen deposition was more pronounced in the fibrotic region compared with that in the inflamed area ( P < 0.0001) (Figure ). Immune cell characterization by flow cytometry To assess the distribution of immune cells in the fibrotic region and in the inflamed tissue of patients with CD, flow cytometry was performed and compared with that in the unaffected tissue of non-IBD controls and the patients with CD. Although no difference in total eosinophil count was observed, active eosinophils, determined through CD69 expression, were enriched in the fibrotic deeper layers when compared with those in the unaffected deeper layers of patients with CD and non-IBD controls ( P = 0.0008 and P = 0.02) (Figure ). Furthermore, the monocyte count was elevated both in the fibrotic ( P = 0.006 for mucosa and P = 0.0004 for deeper layers) and in the inflamed regions ( P = 0.001 for mucosa and P = 0.008 for deeper layers) when compared with the unaffected region in non-IBD controls. In comparison with the unaffected region of patients with CD, monocytes were elevated in the deeper layers only (both P < 0.0001 for inflamed and fibrotic regions). Moreover, the total dendritic cell population was increased in both the fibrotic and in the inflamed deeper layers in comparison with the unaffected deeper layers of the patients with CD (both P < 0.0001) and non-IBD controls ( P = 0.002 and P = 0.009). Focusing on the adaptive immune system, Th2 cells were increased in the fibrotic and in the inflamed deeper layers when compared with the unaffected deeper layers of patients with CD ( P = 0.0002 and P < 0.0001) and non-IBD controls ( P = 0.02 and P = 0.04). Mucosal Th2 cells in the fibrotic and in the inflamed region were only increased in comparison with the unaffected mucosa of patients with CD ( P = 0.008 and P = 0.01). Similarly, Treg were increased in both the fibrotic and inflamed mucosa and deeper layers in comparison with the unaffected ileum ( P < 0.05 for all comparisons). Finally, while we observed a shift toward an increased presence of Th2 cells and Treg, we noticed an opposite trend in Th1 cells, which were decreased in all inflamed and all fibrostenotic layers when compared with paired unaffected regions ( P < 0.05 for all comparisons). No other significant differences in immune cell distribution (Total B cells and T cells, T FH , Th9, Th17, Th22, Th1/17, mast cells, neutrophils, and basophils) were detectable in the inflamed and fibrotic regions when compared with the unaffected region. The median (interquartile range) of all immune cells observed through flow cytometry are available in Supplementary Digital Content (see Supplementary Table 5, http://links.lww.com/CTG/B118 ). Flow cytometry data are summarized in Figure and Supplementary Digital Content (see Supplementary Table 5, http://links.lww.com/CTG/B118 ). Key protein expression levels during fibrosis and inflammation Because a differential presence of several immune cells was identified through flow cytometry, we additionally studied differences at the protein level using ELISA and the MSD platform. Therefore, our investigation focused on cytokines and proteins synthesized by the immune cells that were associated with inflammation and fibrosis based on flow cytometry. IL-5 protein expression was significantly upregulated in both the fibrotic and the inflamed regions compared with non-IBD controls ( P < 0.05 for all comparisons). Likewise, the IL-1β protein expression in inflamed and fibrotic deeper layers was increased in comparison with that in unaffected CD tissue ( P = 0.01 and P = 0.007). Next, IL-4 was significantly increased in the fibrotic and in the inflamed mucosa and deeper layers when compared with the unaffected region in non-IBD controls and patients with CD ( P < 0.05 for all comparisons). Off note, IL-4 in the deeper fibrotic layers was elevated in comparison with the inflamed deeper layers as well ( P = 0.03). However, an increased IL-13 expression could only be found in the fibrotic and inflamed mucosa when compared with the unaffected mucosa of non-IBD controls ( P = 0.007 and P = 0.003, respectively). In addition, TGF-β3 protein levels were increased in the fibrotic and in the inflamed deeper layers when compared with non-IBD controls ( P = 0.007 and P = 0.002). IL-18 protein expression was significantly increased in the fibrotic deeper layers ( P = 0.02), inflamed mucosa ( P = 0.05), and inflamed deeper layers ( P = 0.01) in comparison with the non-IBD controls. When compared with the unaffected region of patients with CD, only fibrotic deeper layers and inflamed mucosa showed significantly elevated IL-18 levels (both P = 0.007). Last, a similar trend for basic FGF could be observed. In comparison with the deeper layers of the non-IBD controls, significantly increased bFGF protein expression in the fibrotic and in the inflamed deeper layers were found ( P = 0.02 and P = 0.0003). Last, only in the inflamed region, a decreased presence of the anti-inflammatory cytokine IL-10 could be identified ( P = 0.04 in the mucosa and P = 0.006 in the deeper layers). Overall, protein levels of the selected cytokines were very similar in the fibrotic and in the inflamed tissues. Based on the increased proportion of active eosinophils in the fibrotic regions, we assessed the expression of eosinophil chemokines and granular proteins. Eotaxin-3 was elevated only in the inflamed deeper layers, when compared with the deeper layers of non-IBD controls ( P = 0.008). ECP expression levels furthermore were increased in both the inflamed mucosa and deeper layers, in comparison with the non-IBD controls ( P = 0.004 and P = 0.0005) and the CD unaffected area ( P = 0.04 and P = 0.02). No such difference could be found between the fibrostenotic and control regions ( P = 0.3 for mucosa and P = 0.2 for deeper layers). Protein expression levels are summarized in Figure and Supplementary Digital Content (see Supplementary Table 6, http://links.lww.com/CTG/B118 ). Assessing eosinophil and (active) fibroblast colocalization To further examine the potential role of eosinophils in fibrosis and given the established role of fibroblasts in fibrosis, we investigated colocalization of eosinophils and (active) fibroblasts in the fibrotic regions on immunohistochemistry . Both cell types colocalized in all examined layers, and no difference in colocalization between the fibrotic deeper layers and the unaffected deeper layers was observed (of both patients with CD and non-IBD controls) (Figure ). Results are summarized in Supplementary Digital Content (see Supplementary Table 7, http://links.lww.com/CTG/B118 ). Twenty-five patients with CD and 10 non-IBD controls were included. Clinical characteristics are summarized in Table . To morphologically characterize tissue and assess fibrosis formation, collagen deposition was assessed in the unaffected tissue of both non-IBD controls and patients with CD and the inflamed and fibrotic regions of patients with CD. Both inflamed and fibrotic regions exhibited increased collagen deposition in comparison with the unaffected region of the non-IBD controls ( P = 0.002 and P < 0.0001, respectively) and patients with CD ( P = 0.002 and P < 0.0001, respectively). Collagen deposition was more pronounced in the fibrotic region compared with that in the inflamed area ( P < 0.0001) (Figure ). To assess the distribution of immune cells in the fibrotic region and in the inflamed tissue of patients with CD, flow cytometry was performed and compared with that in the unaffected tissue of non-IBD controls and the patients with CD. Although no difference in total eosinophil count was observed, active eosinophils, determined through CD69 expression, were enriched in the fibrotic deeper layers when compared with those in the unaffected deeper layers of patients with CD and non-IBD controls ( P = 0.0008 and P = 0.02) (Figure ). Furthermore, the monocyte count was elevated both in the fibrotic ( P = 0.006 for mucosa and P = 0.0004 for deeper layers) and in the inflamed regions ( P = 0.001 for mucosa and P = 0.008 for deeper layers) when compared with the unaffected region in non-IBD controls. In comparison with the unaffected region of patients with CD, monocytes were elevated in the deeper layers only (both P < 0.0001 for inflamed and fibrotic regions). Moreover, the total dendritic cell population was increased in both the fibrotic and in the inflamed deeper layers in comparison with the unaffected deeper layers of the patients with CD (both P < 0.0001) and non-IBD controls ( P = 0.002 and P = 0.009). Focusing on the adaptive immune system, Th2 cells were increased in the fibrotic and in the inflamed deeper layers when compared with the unaffected deeper layers of patients with CD ( P = 0.0002 and P < 0.0001) and non-IBD controls ( P = 0.02 and P = 0.04). Mucosal Th2 cells in the fibrotic and in the inflamed region were only increased in comparison with the unaffected mucosa of patients with CD ( P = 0.008 and P = 0.01). Similarly, Treg were increased in both the fibrotic and inflamed mucosa and deeper layers in comparison with the unaffected ileum ( P < 0.05 for all comparisons). Finally, while we observed a shift toward an increased presence of Th2 cells and Treg, we noticed an opposite trend in Th1 cells, which were decreased in all inflamed and all fibrostenotic layers when compared with paired unaffected regions ( P < 0.05 for all comparisons). No other significant differences in immune cell distribution (Total B cells and T cells, T FH , Th9, Th17, Th22, Th1/17, mast cells, neutrophils, and basophils) were detectable in the inflamed and fibrotic regions when compared with the unaffected region. The median (interquartile range) of all immune cells observed through flow cytometry are available in Supplementary Digital Content (see Supplementary Table 5, http://links.lww.com/CTG/B118 ). Flow cytometry data are summarized in Figure and Supplementary Digital Content (see Supplementary Table 5, http://links.lww.com/CTG/B118 ). Because a differential presence of several immune cells was identified through flow cytometry, we additionally studied differences at the protein level using ELISA and the MSD platform. Therefore, our investigation focused on cytokines and proteins synthesized by the immune cells that were associated with inflammation and fibrosis based on flow cytometry. IL-5 protein expression was significantly upregulated in both the fibrotic and the inflamed regions compared with non-IBD controls ( P < 0.05 for all comparisons). Likewise, the IL-1β protein expression in inflamed and fibrotic deeper layers was increased in comparison with that in unaffected CD tissue ( P = 0.01 and P = 0.007). Next, IL-4 was significantly increased in the fibrotic and in the inflamed mucosa and deeper layers when compared with the unaffected region in non-IBD controls and patients with CD ( P < 0.05 for all comparisons). Off note, IL-4 in the deeper fibrotic layers was elevated in comparison with the inflamed deeper layers as well ( P = 0.03). However, an increased IL-13 expression could only be found in the fibrotic and inflamed mucosa when compared with the unaffected mucosa of non-IBD controls ( P = 0.007 and P = 0.003, respectively). In addition, TGF-β3 protein levels were increased in the fibrotic and in the inflamed deeper layers when compared with non-IBD controls ( P = 0.007 and P = 0.002). IL-18 protein expression was significantly increased in the fibrotic deeper layers ( P = 0.02), inflamed mucosa ( P = 0.05), and inflamed deeper layers ( P = 0.01) in comparison with the non-IBD controls. When compared with the unaffected region of patients with CD, only fibrotic deeper layers and inflamed mucosa showed significantly elevated IL-18 levels (both P = 0.007). Last, a similar trend for basic FGF could be observed. In comparison with the deeper layers of the non-IBD controls, significantly increased bFGF protein expression in the fibrotic and in the inflamed deeper layers were found ( P = 0.02 and P = 0.0003). Last, only in the inflamed region, a decreased presence of the anti-inflammatory cytokine IL-10 could be identified ( P = 0.04 in the mucosa and P = 0.006 in the deeper layers). Overall, protein levels of the selected cytokines were very similar in the fibrotic and in the inflamed tissues. Based on the increased proportion of active eosinophils in the fibrotic regions, we assessed the expression of eosinophil chemokines and granular proteins. Eotaxin-3 was elevated only in the inflamed deeper layers, when compared with the deeper layers of non-IBD controls ( P = 0.008). ECP expression levels furthermore were increased in both the inflamed mucosa and deeper layers, in comparison with the non-IBD controls ( P = 0.004 and P = 0.0005) and the CD unaffected area ( P = 0.04 and P = 0.02). No such difference could be found between the fibrostenotic and control regions ( P = 0.3 for mucosa and P = 0.2 for deeper layers). Protein expression levels are summarized in Figure and Supplementary Digital Content (see Supplementary Table 6, http://links.lww.com/CTG/B118 ). To further examine the potential role of eosinophils in fibrosis and given the established role of fibroblasts in fibrosis, we investigated colocalization of eosinophils and (active) fibroblasts in the fibrotic regions on immunohistochemistry . Both cell types colocalized in all examined layers, and no difference in colocalization between the fibrotic deeper layers and the unaffected deeper layers was observed (of both patients with CD and non-IBD controls) (Figure ). Results are summarized in Supplementary Digital Content (see Supplementary Table 7, http://links.lww.com/CTG/B118 ). The pathophysiology of CD-related fibrosis is still largely unknown. In this study, we demonstrated a differential distribution of active eosinophils, Th1 cells, Th2 cells, Treg, monocytes, and dendritic cells in the different tissue layers of patients with fibrostenotic CD in comparison with that in control tissue. Strikingly, the immunological landscape between fibrotic and inflamed regions looked largely similar, with activated eosinophils as the only differentiator. We observed striking similarities between the inflamed and the fibrotic regions for the immune cell composition, protein expression levels, and distances between eosinophils and (active) fibroblasts. In addition, although collagen deposition was more pronounced in the fibrotic region, confirming the clinical categorization of fibrotic vs inflamed region, the inflamed region also showed significantly increased collagen accumulation in comparison with the unaffected tissue. These findings suggest that inflammation and fibrosis are closely intertwined processes, indicating that remodeling might be spread further than what first meets the eye. Previous literature also suggested that fibrosis is unlikely to occur without the presence of inflammation , although the question of whether inflammation drives fibrosis is a topic of ongoing debate and remains a central point of interest . Nonetheless, in our cohort, we did identify 2 factors that differed between these regions: the presence of active eosinophils and eotaxin-3 and ECP protein expression. Eosinophils have previously been described to be profibrotic in murine models of renal, pulmonary, hepatic, and radiation-induced fibrosis . However, no such data exist for the human gut. In our cohort of patients with CD undergoing surgery for stricturing disease, we observed an increase in activated eosinophils, but not of the total eosinophil count, in the fibrotic deeper layers. However, eosinophils did not colocalize specifically with active fibroblasts in the fibrostenotic tissue. Furthermore, both in the inflamed and the fibrotic regions, we could observe an increased protein expression of IL-1β, IL-4, IL-5, and bFGF. These proteins can all be produced by the eosinophils and have previously been described to stimulate inflammation and fibrosis . Through these proteins, eosinophils could potentially mediate the process of inflammation and fibrosis. Of interest, IL-4 was elevated in the fibrotic deeper layers in comparison with the inflamed deeper layers as well and might contribute to the presence of activated eosinophils solely in the fibrotic deeper layers. In contrast to the increased presence of active eosinophils in the fibrotic deeper layers, the inflamed deeper layers embedded a higher concentration of ECP, which is only produced and secreted by eosinophils. Taken together, these findings might suggest that eosinophils do play a role in both the fibrotic and inflammatory processes, but potentially through different mediators in each process. Hence, eosinophils and their signaling molecules should be further explored as potential new therapeutic targets in fibrostenosing CD. Subsequently, we demonstrated increased Th2 and Treg, but a decrease in the total amount of Th1 cells in both the fibrotic and in the inflamed regions. Th2 cells have previously been proposed as profibrotic and proinflammatory cells through their ability to produce IL-4 and IL-13 . In line with this, we similarly observed an increase in IL-4 and IL-13 protein expression in our cohort. Similar as to the eosinophils, Th2 cells can produce the proinflammatory and profibrotic cytokines IL-1β, IL-4, IL-5, and bFGF . Hence, our findings might point toward a proinflammatory and profibrotic role for the Th2 cells, although no causal or mechanistic relationship has ever been described between fibrosis and Th2 immune responses. Consistent with the current findings, Barrow and Wynn previously showed that mice that developed biliary fibrosis expressed increased Th2 and Treg levels. However, the role of Treg in murine fibrosis models is still being debated. These adaptive immune cells have been described as profibrotic, antifibrotic, or to have no role at all in the development of fibrosis . For example, Treg deletion was shown to reduce cardiac fibrosis and bleomycin-induced lung fibrosis . Moreover, previous studies have also indicated a clear Treg plasticity. It has been shown that Treg can ultimately lose their suppressor ability and adopt a Th2 phenotype through which they can promote fibrosis and remodeling and thereby also exacerbate the existing fibrotic response without initiating the response of excessive healing . Last, both dendritic cells and monocytes have been described to be profibrotic and proinflammatory in a multitude of studies and were increased in the fibrotic and in the inflamed regions of these resection specimens . Increased monocyte counts have previously been linked to worse outcomes in idiopathic pulmonary fibrosis . In addition, a C-C chemokine receptor type 2+ subset contributed to colonic fibrosis in a chronic dextran sodium sulphate colitis model, highlighting a potential profibrotic role . Our findings furthermore revealed an increase in IL-18 in the inflamed and in the fibrotic regions. IL-18 has previously been identified to be upregulated in patients with CD and to have a strong proinflammatory and profibrotic role . In that way, and because monocytes are potent producers of IL-18, these cells could facilitate inflammation and fibrosis. Similarly, dendritic cells have been shown to contribute to fibrosis by activating myofibroblasts . These cell types were found to be elevated in the inflamed and fibrotic regions in our cohort, supporting a role in fibrosis and inflammation. Previously, several studies have investigated the coculturing of eosinophils and fibroblasts. In a first study, IL-33–primed eosinophils were cocultured with intestinal fibroblasts, leading to the release of various profibrotic factors such as IL-1β, IL-6, and periostin . More recently, Kuwabara et al demonstrated that human eosinophils could enhance α-smooth muscle actin expression in a human fetal lung fibroblast cell line when cocultured, indicating a shift toward active fibroblasts. In addition, the impact of IL-3–activated eosinophils on human lung fibroblasts was examined by exposing the fibroblasts to eosinophil degranulating products, resulting in significant changes in gene expression and thereby underscoring the potential importance of eosinophil-fibroblast interactions in tissue remodeling . Although this is the first study profiling the immunological landscape in fibrostricturing CD using flow cytometry, the current design allowed only for association, but cannot establish a causal relationship. Without additional single-cell analyses or a further immunological characterization by flow cytometry, we cannot infer about more specific subpopulations of the various immune cell subtypes. However, identifying eosinophils through single-cell sequencing has shown to be challenging due to their high ribonuclease content, and therefore, flow cytometry still has its place in the immunoprofiling of patients with IBD. Similarly, low sample sizes used in single-cell sequencing assays can be problematic when studying a heterogenous disease such as CD. Nonetheless, further research using novel techniques could lead to a better understanding of our findings through more advanced cellular characterization. In that way, we could obtain better insights on the functionality of these immune cells. In conclusion, we studied transmural intestinal sections of a unique cohort of patients with CD undergoing surgery for a fibrostenotic disease. We were able to identify the differential immune cell distribution and protein expression in unaffected, inflammatory, and fibrotic terminal ileum. Activated eosinophils were more abundant in the deeper layers of the fibrotic tissues, highlighting them and their secreted products as potential treatment targets in fibrostenotic CD. Guarantor of the article: Bram Verstockt, MD, PhD. Specific author contributions: I.J., J.S., M.F., S.V., C.B., T.V., B.V.: study concept and design. A.D., G.B.: intestinal tissue sampling. G.D.H.: histological scoring. I.J., B.J.K., J.C.: data acquisition and analysis. I.J., B.J.K., J.C., G.M., J.S., M.F., S.V., C.B., T.V., B.V.: interpretation of data. I.J.: writing the initial manuscript. All authors approved the final version of the manuscript. Financial support: This research has been funded by a Beligan Inflammatory Bowel Disease Research and Development (BIRD) grant and an internal C1 project (ZKD2906 – C14/17/097) of Katholieke Universiteit Leuven and BIRD grant. T.V., M.F., and J.S. are supported by a senior clinical research fellowship of the Flanders Research Foundation (FWO Vlaanderen; T.V.: 1830517N). B.V. and C.B. are supported by the Clinical Research Fund (KOOR) at the University Hospitals Leuven. B.V. is also supported by the Research Council at the Katholieke Universiteit Leuven. S.V. holds a BOF-FKO from the Katholieke Universiteit Leuven. Potential competing interests: T.V. has received research support and lecture and consultancy fees from Takeda. B.V. reports financial support for research from AbbVie, Biora Therapeutics, Landos, Pfizer, Sossei Heptares, and Takeda; lecture fees from Abbvie, Biogen, Bristol Myers Squibb, Celltrion, Chiesi, Falk, Ferring, Galapagos, Janssen, Lily, MSD, Pfizer, R-Biopharm, Takeda, Truvion, and Viatris; and consultancy fees from Abbvie, Alimentiv, Applied Strategic, Atheneum, Biora Therapeutics, Bristol Myers Squibb, Galapagos, Guidepont, Mylan, Inotrem, Ipsos, Janssen, Lily, Progenity, Sandoz, Santa Ana Bio, Sosei Heptares, Takeda, Tillots Pharma, and Viatris. M.F. receives financial support for research from AbbVie, Amgen, Biogen, EG, Janssen, Pfizer, Takeda, and Viatris; receives speakers' fees from AbbVie, Amgen, Biogen, Boehringer Ingelheim, Falk, Ferring, Janssen-Cilag, Lamepro, MSD, Pfizer, Sandoz, Takeda, Truvion Healthcare, and Viatris; and does consultancy for AbbVie, AgomAb Therapeutics, Boehringer Ingelheim, Celgene, Celltrion, Eli Lilly, Janssen-Cilag, Medtronic, MSD, Pfizer, Regeneron, Samsung Bioepis, Sandoz, Takeda, and ThermoFisher. J.S. receives financial support for research from Galapagos and Viatris; receives speakers' fees from Abbvie, Falk, Takeda, Pfizer, Galapagos, Ferring, Janssen, and Fresenius; and does consultancy for Janssen, Ferring, Fresenius, Abbvie, Galapagos, Celltrion, Pharmacosmos, and Pharmanovia. S.V. receives financial support for research from AbbVie, J&J, Pfizer, Takeda, and Galapagos and receives speakers' and consultancy fees from AbbVie, AbolerIS Pharma, AgomAb, Alimentiv, Arena Pharmaceuticals, AstraZeneca, Avaxia, BMS, Boehringer Ingelheim, Celgene, CVasThera, Cytoki Pharma, Dr Falk Pharma, Ferring, Galapagos, Genentech-Roche, Gilead, GSK, Hospira, Imidomics, Janssen, J&J, Lilly, Materia Prima, MiroBio, Morphic, MrMHealth, Mundipharma, MSD, Pfizer, Prodigest, Progenity, Prometheus, Robarts Clinical Trials, Second Genome, Shire, Surrozen, Takeda, Theravance, Tillots Pharma AG, and Zealand Pharma. G.D.H. receives fees for his activities as central pathology reader from Centocor and receives speakers' fees from Janssen. G.B. receives speakers' fees from Galapagos, Janssen, and Takeda. C.B. has received research fees from Ablynx. Data availability statement: The data underlying this article will be shared on reasonable request. Study Highlights WHAT IS KNOWN ✓ Patients with Crohn's Disease (CD) often develop strictures. ✓ The immune landscape involved in the process of the development of strictures is poorly understood. WHAT IS NEW HERE ✓ We identified the immune landscape in fibrostenotic obstructions of patients with CD. ✓ We primarily identified an increase in active eosinophils in the fibrotic deeper layers. ✓ This immunological characterization can aid to prioritize possible antifibrotic targets for stricture development in patients with CD. ✓ Patients with Crohn's Disease (CD) often develop strictures. ✓ The immune landscape involved in the process of the development of strictures is poorly understood. ✓ We identified the immune landscape in fibrostenotic obstructions of patients with CD. ✓ We primarily identified an increase in active eosinophils in the fibrotic deeper layers. ✓ This immunological characterization can aid to prioritize possible antifibrotic targets for stricture development in patients with CD. |
Enhancing the Resilience of Agroecosystems Through Improved Rhizosphere Processes: A Strategic Review | 1c7a27ca-d94c-4da1-a728-1931ac7597cb | 11720004 | Microbiology[mh] | The swift increase in the global population has markedly heightened the need for energy, food, and other essential goods, leading to the expansion of agricultural land use. Agroecosystems are intricate, evolving environments in which agricultural activities engage with natural ecosystems, affecting soil health, biodiversity, and ecosystem services. Recent studies highlight the significance of agroecosystems in enhancing resistance to climate change, boosting soil health, and fostering biodiversity . For example, soil is a complex ecosystem and is one of our most important resources due to its many vital functions, such as providing food, fuel, and fiber; recycling nutrients that turn into beneficial nutrients for crop and plant necessities; organic matter decomposition, including plants and animal residues; regulation of water supply and quality; and provision of soil microorganisms, which are important for soil health and play vital roles in agroecosystems . However, poor farming techniques, urbanization, and inadequate soil management practices have had negative impacts on agricultural soil, such as reduced soil fertility, the loss of beneficial soil microbiota, and the migration of soil organisms . The top layer of soil surrounding plant roots, known as the rhizosphere, is home to a diverse range of microbes that play vital roles in promoting plant growth, organic matter degradation, and maintaining soil sustainability. However, in the lack of proper soil management, coupled with inappropriate agricultural practices, the quality of rhizosphere soil deteriorates, and its microbial communities weaken . The rhizosphere serves as a microbial hotspot owing to the exudates emitted by plant roots, which supply vital nutrients for microbial activities . Dynamics, for example, root zones, plant growth phases, plant species, and environmental conditions, change the diversity and composition of soil microbial communities (e.g., bacterial and fungi) in the rhizosphere . Multiple species can compose microbial groups, demonstrating comparable mechanisms of action yet differing in their tolerance to plant cultivars and various environmental stresses. The microorganisms found in the rhizosphere include fungi, bacteria, and nematodes . Large numbers of this community have an impartial effect on soil and plants, but some microbes are directly or indirectly connected with the plant–soil systems and provide nutrients from the rhizosphere to the plant. Additionally, rhizosphere microbiota contains microorganisms with both positive and negative effects on plant growth and soil health. Harmful microbes include pathogenic fungi, nematodes, and bacteria, while beneficial microbes, such as PGPM, nitrogen-fixing bacteria, and mycorrhizal fungi, contribute to plant nutrition and soil health . However, beneficial microbes provide essential plant nutrients and contribute to lowering greenhouse gas emissions, improving nutrient availability, suppressing soil-borne plant diseases, and balancing nutrient cycling . Conversely, many authors have pointed out that the use of various agricultural techniques, overuse of mineral fertilizers, agrochemicals, and farming practices could reduce soil fertility and disturb the rhizosphere microbiota, resulting in decreased soil health and crop production . In general, overuse of agrochemicals (i.e., fungicides, herbicides, nematicides, and insecticides) is being used non-judiciously, which has a harmful effect on soil-positive microbiota in the soil as well as the overall health and quality of the soil . This impact is mostly related to changes in microbial parameters such as enzymatic activity and biomass, which are the most important features of soil health and are important tools in monitoring soil quality. Eighty to ninety percent of soil processes are controlled by soil microorganisms and are strongly linked with plant and soil health, in addition to the role of microorganisms in the suppression of soil-borne diseases, nutrient cycling, soil resilience, and maintenance of soil sustainability . Nevertheless, several authors reported that green manure is an alternate way to sustain agriculture and maintain soil health. The integration of the entire plant, containing more nutrients than crop straw, into the soil enhances soil fertility by adding more organic matter. Green manure improves crop yield and growth, reduces pests and insects, suppresses plant diseases, and increases soil enzyme activities, which regulate nutrient fluxes and stimulate plant growth . Therefore, a sustainable solution must be sought by utilizing all available resources to modulate the rhizospheric process for sustainably enhancing agroecosystem productivity. It is important for long-term sustainability in agricultural systems to keep agroecosystems stable and improve rhizosphere processes that make soil more resilient. This includes keeping and expanding the diversity of microorganisms in the rhizosphere. The rhizosphere is crucial for facilitating nutrient cycling, enhancing soil structure, and promoting plant health. The primary goal of this review is to investigate several methods for improving agricultural ecosystem resilience through rhizosphere activity optimization. We can implement these practices by utilizing beneficial microbes, green manures, and intercropping. These strategies jointly enhance synergistic communications between soil microorganisms and plant roots, creating an environment favorable for soil health and plant growth. The findings of this review improve comprehension of sustainable agriculture techniques, highlighting the importance of natural and organic resources in boosting crop production while preserving ecological balance. The rhizosphere refers to the area surrounding the root zone . This zone is considered essential for biological activities because many exudates produced by plant roots provide nutrients for soil microorganisms. Thus, this region is vital for disease resistance, nutrient cycling, and plant growth . The rhizosphere zone has a central role in the food chain of soil microorganisms. The microbial food chain mainly includes dead plant matter and living plant roots, which are the primary source of nutrients , and carbon from living plant tissues and root exudates. These are crucial for plant health and growth as they support disease suppression, protection, and nutrient uptake . The activity of beneficial microorganisms in the rhizosphere not only provides essential nutrients to plants but also influences the quality and quantity of root exudates . Furthermore, various metabolites exuded by plant roots into the soil primarily shape and attract the microbial community composition in the rhizosphere, facilitating interactions with both biotic and abiotic factors. Plants frequently adjust the diversity of these microbes according to the benefits they provide for growth and health . In the rhizosphere, plants engage in a wide range of interactions with soil-dwelling organisms, including exploitation, competition, neutrality, mutualism, and commensalism. These diverse interactions are essential for plants to adapt to varying environmental conditions, which is crucial for their survival and agricultural productivity. In the rhizosphere, plants host a diverse collection of microbes, and they can inhabit the inner tissues (endosphere), and surface tissues (phylloplane and rhizoplane). Collectively, these microorganisms are referred to as the plant microbiome . The plant and soil microbiome is critical for stimulating plant development, improving soil quality and fertility, and supporting agricultural sustainability by strengthening resilience to environmental stressors, nutrient cycling, and disease-fighting . The interplay between roots and their surrounding soil constitutes one of the most prolific and diverse ecosystems in nature. The rhizosphere, the tiny and thin soil layer around the roots, has a vast population of microbes fueled by carbon-rich compounds exuded during photosynthesis and provides essential nourishment for the recruited microorganisms . Overall, the rhizosphere is a kind of microenvironment wherein plant root development and nutrient uptake, soil properties, and microbe activities interact in a corresponding manner . 2.1. Rhizosphere Processes: Nutrient Retention in Agroecosystems The rhizosphere hosts diverse microbial groups that play critical roles in nutrient cycling, promoting plant growth, and protecting plants from pathogens, as well as from biotic and abiotic stresses . Changes in rhizosphere processes directly contribute to nutrient retention in agroecosystems. Through plant–soil interactions, plants and their residues provide organic carbon (C) to the soil. The rate of soil carbon input depends primarily on the rate of plant growth, which is driven by photosynthesis. Root-derived carbon input, in particular, is a key regulatory factor in the plant–soil interaction within the rhizosphere, as it has a more direct impact on soil processes than carbon input from shoots . This root-derived carbon plays a significant role in soil carbon sequestration. The amount of carbon that can be stored as soil organic matter (SOM) depends on the balance between carbon input and the rate at which plant materials decompose. Decomposition is the process by which microorganisms break down organic molecules into their inorganic components, such as nutrients and CO 2 . Carbon dioxide, produced during microbial respiration, is released from the soil back into the atmosphere. However, some decomposition products are protected from further breakdown, playing a critical role in forming stable SOM, which is essential for long-term carbon sequestration in agroecosystems . As a result, the rhizosphere process is particularly crucial for nutrient cycling, stabilizing SOM, making nutrients available for plant growth, and ultimately handling soil microorganisms. On the other hand, maintaining and increasing soil and its process through utilizing carbonized materials such as “Biochar” have long-term stability . Biochars can reduce the emission of N 2 O and CH 4 from the soil . Biochar comprises a stable C compound as a derivative of gasification and pyrolysis and is available either as a pelletized or powder form. Char is ubiquitous in natural soils, containing as much as 35% of total organic C (TOC) in fire-impacted soils. Laird reported that the application of biochar into the soil was found to enhance soil fertility and significantly increase soil organic carbon (SOC) and nutrients, which leads to the improvement in soil health. SOC consists of a complex mix of partially decomposed substances, including polysaccharides, lignin, aliphatic biopolymers, tannins, lipids, proteins, and amino sugars derived from plant litter, as well as microbial and faunal biomass . However, SOC has been classified into slow, active, and inert pools, depending upon the decomposition rate in soil . Furthermore, biochar largely supports the SOC pool for stabilization, while green manure incorporation adds to the more bio-accessible fraction of the soil C reserve. This is because biochar decomposes typically slower than uncharred biomass, e.g., green manure incorporation and fresh crop residues . However, green manure or fresh crop residues in agricultural fields, on the other hand, can be used as a SOC storage/sequestration strategy by supporting cover crops; conservation tillage provides benefits including nutrient cycling, control of surface runoff of water, wind erosion, and crop production . Yazdanpanah reported that applying green manure or plat residues improved the stability and total porosity of soil aggregates and increased the soil C pool, but this depended on the type and application rate of soil amendments. However, Kenney et al. reported that the removal of crop residues (e.g., corn stover and Zea mays L.) decreased the soil C pool. Therefore, the following strategy should be considered to achieve higher soil C sequestration via soil amendments produced from green manure/crop residues or biochar. The indirect effects of using these amendments should be considered in the context of production/control of soil GHG, environmental health, and economic cost benefits of their use. As per our knowledge, the combined effect of biochar and crop residues has rarely been investigated except in one recent study by Nguyen et al. , and it should be considered in the near future. Overall, biochar seems to have a greater positive impact on total soil SC, while the incorporation of crop residues/green manures can increase the microbial biomass C of soil, enhance the rhizosphere process, and reduce carbon dioxide emission. 2.2. Rhizosphere Engineering: Turning to Sustainably Increases the Agroecosystem and Its Productivity The rhizosphere is a hotspot of microbes because many exudates are produced by plant roots, which are the primary source of nutrients for essential microbial processes and sustain the agroecosystems and their functions. The rhizosphere contains diverse microbial groups that perform various functions and affect nutrient cycling and plant growth promotion, and this portion is affected via different strategies such as crop rotation, green manuring, and sustainable use of PGPM and intercropping. However, through the addition of PGPM, Asghar and Kataoka found that plant growth-promoting fungi changed the fungal community composition, enhanced the phosphatase and glycosidase enzyme activities in the soil, and stimulated the lettuce and brassica plant growth. Further, Kataoka et al. reported that incorporating leguminous green manure (Hairy Vetch, Vicia villosa L.) enhances the soil fungal biomass and diversity and provides a healthier environment by producing soil phosphatase enzymes. Furthermore, rhizosphere engineering and its processes are explained with different strategies. In short, these interventions improve the rhizospheric processes and sustainably increase the agroecosystem and its productivity. On the other side of the agroecosystems and maintaining the rhizosphere process, biochar has emerged as a crucial innovation in scientific research, offering multiple benefits for sustainable agriculture, rhizosphere management, and environmental health . Biochar is a type of charcoal produced through the combustion of organic feedstock with limited oxygen. It is considered a valuable soil conditioner and an effective means of carbon sequestration, which helps mitigate global warming and climate change . Biochar decomposes slowly, making it highly durable when applied to soil, where it can remain for the long term . However, generally, biochar is derived from waste materials such as animal manure, forest residues, and agricultural byproducts . Through scientific processes, these waste materials are transformed into valuable products that can directly or indirectly improve soil health and promote plant growth. Biochar enhances crop yield, improves soil quality, and maintains soil health by preserving biochemical properties . In terms of soil biological properties, Uzoma et al. found that biochar positively influences soil microbes and their activity, leading to increased crop yields. However, the effect of biochar depends on the specific soil type and specific crops. Another study by Lu et al. highlighted that biochar’s porous structure creates a favorable niche for soil microbiota, boosting microbial populations. Conversely, biochar derived from rice straw has been shown to reduce certain microbial communities, such as Ascomycota fungi and Actinobacteria , though the abundance and diversity of soil microbes may vary with different types of biochar . The application of biochar not only alters soil microbial communities but also affects nutrient cycling, which can have a direct impact on rhizospheric processes and agroecosystem productivity. The rhizosphere hosts diverse microbial groups that play critical roles in nutrient cycling, promoting plant growth, and protecting plants from pathogens, as well as from biotic and abiotic stresses . Changes in rhizosphere processes directly contribute to nutrient retention in agroecosystems. Through plant–soil interactions, plants and their residues provide organic carbon (C) to the soil. The rate of soil carbon input depends primarily on the rate of plant growth, which is driven by photosynthesis. Root-derived carbon input, in particular, is a key regulatory factor in the plant–soil interaction within the rhizosphere, as it has a more direct impact on soil processes than carbon input from shoots . This root-derived carbon plays a significant role in soil carbon sequestration. The amount of carbon that can be stored as soil organic matter (SOM) depends on the balance between carbon input and the rate at which plant materials decompose. Decomposition is the process by which microorganisms break down organic molecules into their inorganic components, such as nutrients and CO 2 . Carbon dioxide, produced during microbial respiration, is released from the soil back into the atmosphere. However, some decomposition products are protected from further breakdown, playing a critical role in forming stable SOM, which is essential for long-term carbon sequestration in agroecosystems . As a result, the rhizosphere process is particularly crucial for nutrient cycling, stabilizing SOM, making nutrients available for plant growth, and ultimately handling soil microorganisms. On the other hand, maintaining and increasing soil and its process through utilizing carbonized materials such as “Biochar” have long-term stability . Biochars can reduce the emission of N 2 O and CH 4 from the soil . Biochar comprises a stable C compound as a derivative of gasification and pyrolysis and is available either as a pelletized or powder form. Char is ubiquitous in natural soils, containing as much as 35% of total organic C (TOC) in fire-impacted soils. Laird reported that the application of biochar into the soil was found to enhance soil fertility and significantly increase soil organic carbon (SOC) and nutrients, which leads to the improvement in soil health. SOC consists of a complex mix of partially decomposed substances, including polysaccharides, lignin, aliphatic biopolymers, tannins, lipids, proteins, and amino sugars derived from plant litter, as well as microbial and faunal biomass . However, SOC has been classified into slow, active, and inert pools, depending upon the decomposition rate in soil . Furthermore, biochar largely supports the SOC pool for stabilization, while green manure incorporation adds to the more bio-accessible fraction of the soil C reserve. This is because biochar decomposes typically slower than uncharred biomass, e.g., green manure incorporation and fresh crop residues . However, green manure or fresh crop residues in agricultural fields, on the other hand, can be used as a SOC storage/sequestration strategy by supporting cover crops; conservation tillage provides benefits including nutrient cycling, control of surface runoff of water, wind erosion, and crop production . Yazdanpanah reported that applying green manure or plat residues improved the stability and total porosity of soil aggregates and increased the soil C pool, but this depended on the type and application rate of soil amendments. However, Kenney et al. reported that the removal of crop residues (e.g., corn stover and Zea mays L.) decreased the soil C pool. Therefore, the following strategy should be considered to achieve higher soil C sequestration via soil amendments produced from green manure/crop residues or biochar. The indirect effects of using these amendments should be considered in the context of production/control of soil GHG, environmental health, and economic cost benefits of their use. As per our knowledge, the combined effect of biochar and crop residues has rarely been investigated except in one recent study by Nguyen et al. , and it should be considered in the near future. Overall, biochar seems to have a greater positive impact on total soil SC, while the incorporation of crop residues/green manures can increase the microbial biomass C of soil, enhance the rhizosphere process, and reduce carbon dioxide emission. The rhizosphere is a hotspot of microbes because many exudates are produced by plant roots, which are the primary source of nutrients for essential microbial processes and sustain the agroecosystems and their functions. The rhizosphere contains diverse microbial groups that perform various functions and affect nutrient cycling and plant growth promotion, and this portion is affected via different strategies such as crop rotation, green manuring, and sustainable use of PGPM and intercropping. However, through the addition of PGPM, Asghar and Kataoka found that plant growth-promoting fungi changed the fungal community composition, enhanced the phosphatase and glycosidase enzyme activities in the soil, and stimulated the lettuce and brassica plant growth. Further, Kataoka et al. reported that incorporating leguminous green manure (Hairy Vetch, Vicia villosa L.) enhances the soil fungal biomass and diversity and provides a healthier environment by producing soil phosphatase enzymes. Furthermore, rhizosphere engineering and its processes are explained with different strategies. In short, these interventions improve the rhizospheric processes and sustainably increase the agroecosystem and its productivity. On the other side of the agroecosystems and maintaining the rhizosphere process, biochar has emerged as a crucial innovation in scientific research, offering multiple benefits for sustainable agriculture, rhizosphere management, and environmental health . Biochar is a type of charcoal produced through the combustion of organic feedstock with limited oxygen. It is considered a valuable soil conditioner and an effective means of carbon sequestration, which helps mitigate global warming and climate change . Biochar decomposes slowly, making it highly durable when applied to soil, where it can remain for the long term . However, generally, biochar is derived from waste materials such as animal manure, forest residues, and agricultural byproducts . Through scientific processes, these waste materials are transformed into valuable products that can directly or indirectly improve soil health and promote plant growth. Biochar enhances crop yield, improves soil quality, and maintains soil health by preserving biochemical properties . In terms of soil biological properties, Uzoma et al. found that biochar positively influences soil microbes and their activity, leading to increased crop yields. However, the effect of biochar depends on the specific soil type and specific crops. Another study by Lu et al. highlighted that biochar’s porous structure creates a favorable niche for soil microbiota, boosting microbial populations. Conversely, biochar derived from rice straw has been shown to reduce certain microbial communities, such as Ascomycota fungi and Actinobacteria , though the abundance and diversity of soil microbes may vary with different types of biochar . The application of biochar not only alters soil microbial communities but also affects nutrient cycling, which can have a direct impact on rhizospheric processes and agroecosystem productivity. The rhizosphere is home to various soil microbes, including fungi, bacteria, protozoa, and nematodes, which play essential roles in nutrient cycling, plant growth, and maintaining agroecosystems. These microbes facilitate the breakdown of organic matter and nitrogen fixation, making nutrients available to plants . In agriculture, two groups of soil microorganisms are particularly important—rhizosphere fungi and bacteria—due to their critical functions. Fungi and bacteria thrive in soil, adapting to a wide range of environments, including harsh and high-salt conditions . Fungi in the rhizosphere include pathogenic, saprotrophic, and mutualistic species, with mycorrhizal and plant-growth-promoting fungi being especially important for increasing nutrient availability, promoting plant growth and stress tolerance, and improving soil structure and disease resistance . In contrast, the main rhizosphere bacterial phyla are Firmicutes , Actinobacteria , Acidobacteria , and Proteobacteria . Plant growth-promoting rhizobacteria (PGPR) are crucial for enhancing plant growth, improving disease resistance, responding to carbon inputs and nutrient uptake, and assisting in phytoremediation . In recent decades, a substantial body of literature has emerged demonstrating the activities of microbial species in plant growth promotion, biocontrol of phytopathogens, and nutrient cycling, highlighting their potential for development as alternative or supplementary agrochemicals in sustainable agriculture. 3.1. The Role of Microbes in Agroecosystems Rhizosphere microorganisms, such as fungi and bacteria, not only act as pathogens in agricultural plants but also facilitate plant growth, increase nutrient availability, and control plant diseases through their role as biocontrol agents . Recent research has highlighted the utilization of rhizosphere microorganisms to enhance plant development and mitigate diseases, thereby reducing reliance on pesticides and herbicides . These microorganisms establish symbiotic associations with plant roots, exemplified by mycorrhizal fungi, which augment water and nutrient absorption, especially phosphorus. Furthermore, specific bacterial and fungal species might inhibit detrimental infections, thereby reducing dependence on chemical pesticides. Furthermore, these microbes help in the degradation of pollutants, thereby improving soil quality and resilience . In comparison, the excessive application of chemical fertilizers and pesticides has detrimental environmental consequences, and growing reliance on these substances is concerning. Commercially available as biofertilizers and biocontrol agents, a wide range of microorganisms are utilized in agricultural production and the management of agroecosystems through rhizosphere processes that chemical fertilizers often disrupt. The diversity and activity of soil microbial communities are essential for preserving soil structure, enhancing plant growth, and sustaining ecosystem stability, rendering these microorganisms crucial for sustainable agricultural operations . Therefore, enhancing the resilience of agroecosystems through rhizosphere microbes is crucial for promoting sustainable agriculture. 3.2. Plant Growth Promotion via Microbes The application of chemical fertilizers to meet the food demand for a growing human population has led to many unpredicted environmental concerns related to soil–plant health and agroecosystems . To protect soil–plant health and the environment, there is a need to develop and adopt new biological techniques that are not harmful to agroecosystems, soil, and plant health. Many researchers and scientists have already reported that microbes related to biofertilizers are functional and could be an alternative to chemical fertilizers, provide maximum environmental benefits, and improve crop yield and soil microbial biomass. In this context, microbe-related biofertilizers are capable and could be used in an eco-friendly manner to enhance crop production and maintain agroecosystems . Microbe-related biofertilizers are defined as biological inoculants of beneficial fungal and bacterial strain mobilizers that can increase crop productivity and sustain soil biodiversity . Mazid and Khan reported that these inoculants, applicable directly or indirectly to soil and plant roots, are cost-effective and supply important nutrients for plants, enhancing soil nutritional composition, structure, and fertility. Moreover, biofertilizer-related microbes are commonly known as plant growth-promoting microbes (PGPM). There are various fungal and bacterial strains that have already been reported and used commercially to improve plant growth, soil conditions, and the nutritional demand of plants. Such bacterial and fungal strains are Azotobacter , Pseudomonas , Azospirillum , Rhizobium , Streptomyces , Bacillus spp., Trichoderma , Penicillium , Phoma , Fusarium , and Aspergillus spp., and these strains have been gaining global attention due to their potential to control soil diseases, promote plant growth, and maintain agroecosystems . The significant role and potential activities of PGPM are shown in . 3.3. Singling Between Plants and Microbes in the Rhizosphere The rhizosphere contains four main components: soil, microorganisms, roots, and their interactions. These components collectively influence a range of physical, chemical, and biological processes that affect nutrient use, plant development, and overall plant health. The primary goal of rhizosphere signaling is the interaction between plants and the symbiotic microbes in their surrounding area, which promotes the growth of various rhizosphere microbial communities that have a beneficial impact on plant productivity . A robust reciprocal influence marks the interaction between the plant and the rhizosphere microbiome , where they interact through the exchange of signaling molecules produced and recognized by both the plant and its related microbiosystem . Disease-free and asymptomatic plants have complex relationships with their rhizosphere microbes, which improve plant performance and maintain the agroforestry system . Plants change the pH level, the structure of the soil, the amount of oxygen that is available, and the energy source that comes from carbon-rich exudates. These changes have an effect on the microbiome around the plants’ roots. Many of the chemically varying primary and secondary metabolites found in plant root exudates have bioactive effects on microorganisms, influencing their composition and function . The exudation of carbon from plant roots, which accounts for almost one-third of the carbon produced by photosynthesis, significantly influences the results of chemical interactions at both the individual and community levels . Root exudates serve as a principal means of communication between plants and their living environment, enabling several reactions, including nutrient uptake, competition for resources, signaling across species, the attraction of microorganisms, and several other interactions . Furthermore, sugars, amino acids, organic acids, phenolic compounds, and secondary metabolites like coumarins, glucosinolates, benzoxazinoids, camalexin, and triterpenes are the main organic parts that come from exudate . Plant species cultivate a unique microbial community in their rhizosphere by creating a varied carbon-rich environment. With more evidence, primary metabolites, including glucose, organic acids, and amino acids, supply carbon and energy for microbial proliferation, promoting beneficial microorganisms and inhibiting pathogens. Secondary metabolites, like flavonoids, phenolics, and terpenoids, often have bioactive properties, like the ability to kill bacteria or transfer information, that affect only the microbes in the soil and plants, and these changes in microbial community structure can significantly impact soil and agroecosystem . This community provides many adaptive benefits to the plant host by influencing the composition of microorganisms and adjusting their advantageous characteristics . Additionally, chemicals made from root exudates play a key role in creating a stress-resistant microbiome that helps plants deal with abiotic stresses like not getting enough food, being sick, and drought. The identification of stress-derived metabolomics and microbiota is a viable approach to addressing both abiotic and biotic limitations . However, we have not thoroughly investigated the positive impacts of root-enriched microbial species supported by specialized metabolites produced from root exudate . This paragraph of the manuscript on a review of current research on the plant rhizosphere microbiota and the compounds produced by root exudates for nutrient acquisition as depicted in the conceptual . Rhizosphere microorganisms, such as fungi and bacteria, not only act as pathogens in agricultural plants but also facilitate plant growth, increase nutrient availability, and control plant diseases through their role as biocontrol agents . Recent research has highlighted the utilization of rhizosphere microorganisms to enhance plant development and mitigate diseases, thereby reducing reliance on pesticides and herbicides . These microorganisms establish symbiotic associations with plant roots, exemplified by mycorrhizal fungi, which augment water and nutrient absorption, especially phosphorus. Furthermore, specific bacterial and fungal species might inhibit detrimental infections, thereby reducing dependence on chemical pesticides. Furthermore, these microbes help in the degradation of pollutants, thereby improving soil quality and resilience . In comparison, the excessive application of chemical fertilizers and pesticides has detrimental environmental consequences, and growing reliance on these substances is concerning. Commercially available as biofertilizers and biocontrol agents, a wide range of microorganisms are utilized in agricultural production and the management of agroecosystems through rhizosphere processes that chemical fertilizers often disrupt. The diversity and activity of soil microbial communities are essential for preserving soil structure, enhancing plant growth, and sustaining ecosystem stability, rendering these microorganisms crucial for sustainable agricultural operations . Therefore, enhancing the resilience of agroecosystems through rhizosphere microbes is crucial for promoting sustainable agriculture. The application of chemical fertilizers to meet the food demand for a growing human population has led to many unpredicted environmental concerns related to soil–plant health and agroecosystems . To protect soil–plant health and the environment, there is a need to develop and adopt new biological techniques that are not harmful to agroecosystems, soil, and plant health. Many researchers and scientists have already reported that microbes related to biofertilizers are functional and could be an alternative to chemical fertilizers, provide maximum environmental benefits, and improve crop yield and soil microbial biomass. In this context, microbe-related biofertilizers are capable and could be used in an eco-friendly manner to enhance crop production and maintain agroecosystems . Microbe-related biofertilizers are defined as biological inoculants of beneficial fungal and bacterial strain mobilizers that can increase crop productivity and sustain soil biodiversity . Mazid and Khan reported that these inoculants, applicable directly or indirectly to soil and plant roots, are cost-effective and supply important nutrients for plants, enhancing soil nutritional composition, structure, and fertility. Moreover, biofertilizer-related microbes are commonly known as plant growth-promoting microbes (PGPM). There are various fungal and bacterial strains that have already been reported and used commercially to improve plant growth, soil conditions, and the nutritional demand of plants. Such bacterial and fungal strains are Azotobacter , Pseudomonas , Azospirillum , Rhizobium , Streptomyces , Bacillus spp., Trichoderma , Penicillium , Phoma , Fusarium , and Aspergillus spp., and these strains have been gaining global attention due to their potential to control soil diseases, promote plant growth, and maintain agroecosystems . The significant role and potential activities of PGPM are shown in . The rhizosphere contains four main components: soil, microorganisms, roots, and their interactions. These components collectively influence a range of physical, chemical, and biological processes that affect nutrient use, plant development, and overall plant health. The primary goal of rhizosphere signaling is the interaction between plants and the symbiotic microbes in their surrounding area, which promotes the growth of various rhizosphere microbial communities that have a beneficial impact on plant productivity . A robust reciprocal influence marks the interaction between the plant and the rhizosphere microbiome , where they interact through the exchange of signaling molecules produced and recognized by both the plant and its related microbiosystem . Disease-free and asymptomatic plants have complex relationships with their rhizosphere microbes, which improve plant performance and maintain the agroforestry system . Plants change the pH level, the structure of the soil, the amount of oxygen that is available, and the energy source that comes from carbon-rich exudates. These changes have an effect on the microbiome around the plants’ roots. Many of the chemically varying primary and secondary metabolites found in plant root exudates have bioactive effects on microorganisms, influencing their composition and function . The exudation of carbon from plant roots, which accounts for almost one-third of the carbon produced by photosynthesis, significantly influences the results of chemical interactions at both the individual and community levels . Root exudates serve as a principal means of communication between plants and their living environment, enabling several reactions, including nutrient uptake, competition for resources, signaling across species, the attraction of microorganisms, and several other interactions . Furthermore, sugars, amino acids, organic acids, phenolic compounds, and secondary metabolites like coumarins, glucosinolates, benzoxazinoids, camalexin, and triterpenes are the main organic parts that come from exudate . Plant species cultivate a unique microbial community in their rhizosphere by creating a varied carbon-rich environment. With more evidence, primary metabolites, including glucose, organic acids, and amino acids, supply carbon and energy for microbial proliferation, promoting beneficial microorganisms and inhibiting pathogens. Secondary metabolites, like flavonoids, phenolics, and terpenoids, often have bioactive properties, like the ability to kill bacteria or transfer information, that affect only the microbes in the soil and plants, and these changes in microbial community structure can significantly impact soil and agroecosystem . This community provides many adaptive benefits to the plant host by influencing the composition of microorganisms and adjusting their advantageous characteristics . Additionally, chemicals made from root exudates play a key role in creating a stress-resistant microbiome that helps plants deal with abiotic stresses like not getting enough food, being sick, and drought. The identification of stress-derived metabolomics and microbiota is a viable approach to addressing both abiotic and biotic limitations . However, we have not thoroughly investigated the positive impacts of root-enriched microbial species supported by specialized metabolites produced from root exudate . This paragraph of the manuscript on a review of current research on the plant rhizosphere microbiota and the compounds produced by root exudates for nutrient acquisition as depicted in the conceptual . Green manures represent a promising strategy for sustaining agroecosystems and improving the soil health and rhizosphere processes. Green manure can reduce chemical fertilizers in agroecosystems, ultimately improving soil health and quality . Green manure provides various benefits to the agroecosystem and soil, such as soil covering, reduced soil temperature, organic matter improvement, increased water infiltration, and reduced weed infestation . Moreover, green manure has environmental and economic benefits such as food production, maintaining biodiversity, soil carbon sequestration, and soil retention . However, to enhance the resilience of agroecosystems and maintain soil health, planting green manure is an essential and valuable practice and works to reduce soil-borne diseases through biofumigation. Green manure provides better soil health and biodiversity and promotes the beneficial fungal genera that break the soil-borne pathogen’s life cycle that is connected with the crop genotype . Many types of green manures, including leguminous and non-leguminous, have multiple functions and benefits for soil and plant health, such as incorporating leguminous green manure to enhance wheat yield and production due to the mineralization of inorganic nitrogen and stimulating biological fixation . Furthermore, Kataoka et al. also reported that incorporating leguminous green manure (Hairy Vetch, Vicia villosa L.) enhances the soil fungal biomass and diversity and provides a healthier environment by producing soil phosphatase enzymes. Therefore, it could be assumed that planting and incorporating different green manures could maintain the agroecosystems and rhizosphere processes by enhancing biological processes and increasing organic matter. From another point of view, green manure may alter the soil N and C availability. For example, Hairy Vetch (a leguminous plant) fixes N 2 , whereas barley (a non-leguminous plant) has high biomass productivity and thus adds more C, which influences ecosystem functioning and microbial groups in soil . In comparison, Chavarría et al. reported that green manure incorporation can improve agroecosystem services by improving nutrient retention capacity, accelerating soil organic matter, and reducing greenhouse gases in agricultural soils. However, applying alternative management, such as green manuring practices, stimulates the improvement of soil-diverse microbial communities, accelerating the specific enzyme activities related to P and C . Thus, the relationship between microbial community composition and their activity stimulates plant growth and soil quality. 4.1. Green Manure and Their Beneficial Effects on the Rhizosphere Processes The application of green manure is an additional practice for enhancing the resilience of agroecosystems and rhizosphere processes and increasing soil’s productive capacity through nutrient availability. This application is geared towards maintaining agroecosystems and long-term environmental sustainability by enhancing processes such as nutrient cycling and preserving and maintaining soil biodiversity and the physical and chemical stability of the soil. Incorporating green manure can positively affect soil microbial communities by directly adding nutrients after incorporation and indirectly by changing plant and soil properties . Directly, the decomposition of green manure offers an immediate supply of carbon, nitrogen, and other vital elements; hence, it enhances microbial activity and fosters the proliferation of beneficial bacteria . The mineralization of organic matter from green manure produces labile chemicals that act as substrates for soil microorganisms, thereby boosting microbial biomass and diversity . The incorporation of green manure indirectly modifies the soil’s physical and chemical characteristics, including pH, soil structure, and water retention, fostering a conducive environment for microbial colonization and activity . At the same time, Sun et al. reported that the plant could affect soil microbial diversity and community by releasing root exudates. The decomposing residues of brassica green manure release glucosinolates, which assist in controlling parasitic nematodes . Furthermore, the exclusive use of grass species may improve the soil nutrient profile compared to monoculture. The advantages of green manure designate it as a sustainable practice capable of providing various agro-ecological services; however, effective management of green manure is essential to substantially enhance soil carbon stocks and mitigate climate change, particularly in semi-arid Mediterranean regions, by decreasing greenhouse gas emissions . Furthermore, many studies have proven that planting and incorporating green manure have multiple beneficial effects on soil health and the rhizosphere process . The different types of green manure and their beneficial effects on the rhizosphere are shown in . 4.2. Green Manure and Their Effects on Soil Health Soil health is a vital component of sustainable agriculture, signifying the soil’s ability to operate as a living ecosystem that promotes plant productivity, preserving environmental quality, and supporting biodiversity . Healthy soils exhibit an appropriate combination of physical, chemical, and biological attributes, encompassing adequate organic matter, nutrient accessibility, optimal structure, and vibrant microbial communities. Beneficial microorganisms, including nitrogen-fixing bacteria, mycorrhizal fungi, and decomposers, are vital to nutrient cycling, organic waste breakdown, and disease suppression . Soil health is a primary requirement for agriculture and plant growth, but most agricultural soils are poor in organic nutrients and organic matter due to the high applications of inorganic or chemical-based fertilizers to achieve high crop yields. Different agricultural techniques should be adopted; otherwise, it can become very hazardous and lead to a lack of organic matter. Improving soil health through chemical, physical, and biological properties is attributed to the integrated application of green manure and organic application due to better nutrient uptake and preservation of soil health . A variety of green manures, including grains, root crops, legumes, and oil crops, can serve as vegetative covers, each offering different benefits for enhancing soil health. Therefore, growing and incorporating green manure in agricultural lands helps maintain and preserve soil health and soil nutrient availability and ultimately increases soil organic matter content. The incorporation of green manure provides soil nutrients and organic matter to the soil, and organic matter plays an important role in soil biodiversity, soil microbiota, and the arrangement of soil aggregates to improve the soil structure . Therefore, incorporating green manure leads to an increased organic matter content, which has an important role in soil health and quality. Furthermore, green manure suppresses soil-borne diseases through biofumigation. Soil-borne diseases also affect soil health, reduce crop productivity, and negatively affect soil biodiversity. Many soil-borne pathogens have negative effects on plant growth and crop production, such as Phytophthora spp., Fusarium spp., Verticillium spp., and Rhizoctonia spp. Soil-borne diseases reduce agricultural yield by as much as 50–70%, including those of wheat, vegetables, and cotton . However, the practices of green manure in agricultural lands produce significant benefits such as increased soil nutrient availability, enhanced soil organic carbon, reduced soil compaction, improved particle aggregation and soil structure, as well as strengthened microbe activity, diversity, abundance, and suppression of soil-borne diseases. To protect crops from soil-borne diseases, many fumigants and fungicides need to be used regularly. Later, it was noted that the use of fumigants and fungicides or any other chemical-based product causes ecological problems such as human health hazards and reduced beneficial microbes in the soil biodiversity, directly affecting soil health . Consequently, organically, alternative strategies should be adopted to reduce soil-borne diseases, preserve soil health, and improve crop production. Alternatively, incorporation of green manure and cover crops are useful strategies for protecting soil health from soil-borne diseases and improving soil and plant health through biofumigation, providing nutrients and increasing beneficial microbial activity. Hao et al. reported that broccoli ( B. oleracea ) grown as green manure reduces the S. minor sclerotia in the rhizosphere. Whalen also reported that Fusarium wilt is another soil-borne disease that reduces watermelon production in the southeastern United States. The soil-borne disease Fusarium wilts is due to Fusarium oxysporum , which attacks the crop of watermelon and reduces the annual number of fruits, directly affecting the annual yield of watermelon production. For the management of Fusarium wilt disease, Hairy Vetch ( Vicia villosa L.) green manure was incorporated. After incorporating Hairy Vetch, the occurrence of Fusarium wilt disease in the rhizosphere of the watermelon field was reduced, leading to an increased annual watermelon yield by 45% . Papp et al. demonstrated that the cultivation of green manures influences soil microorganism environments, which are essential for sustaining soil functions and ecological system sustainability because they participate in nutrient cycling and organic matter turnover, providing multiple services. Researchers have observed that the incorporation of green manure modifies the composition and dynamics of soil fungus and bacterial communities and encourages beneficial microorganisms. Ntalli and Caboni observed that certain cruciferous green manures, including canola ( Brassica napus L.) and rape ( Brassica rapa L.), upon termination, release isothiocyanates (ITCs) via the hydrolysis of glucosinolates (GSLs) during the decomposition of their tissues, serving as effective biofumigants against various soil-borne pathogens and pests. The utilization of cruciferous species as green manure may facilitate the natural management of possible soil-borne diseases. Furthermore, Waisen et al. demonstrated that brown mustard ( Brassica juncea L.) and radish ( Raphanus raphanistrum subsp. sativus) serve as excellent biofumigant crops that protect against plant–parasitic nematodes while keeping the soil healthy and maintaining the strength of the nematode population structure strong. Aydınlı and Mennan also found that planting arugula ( Eruca sativa ) and radish ( R. sativus ) as winter crops before planting plants that are easily damaged by the root-knot nematode Meloidogyne arenaria may lower the number of eggs laid, the gall index, and the damage that occurs, while also increasing crop yields. Leguminous green manure can substantially enrich the soil with nitrogen throughout its growth and can acidify the rhizosphere by enhancing the absorption of insoluble phosphorus into the soil . Leguminous green manure can fix atmospheric nitrogen via symbiosis with rhizobia in root nodules . Furthermore, the fixed nitrogen can be allocated to intercropped non-leguminous plants within mixed cropping systems, or it may succeed crops in rotational practices. Biological nitrogen-fixing (BNF) systems may decrease the reliance on commercial nitrogen fertilizers . Further, Scavo et al. reported that the existence of Trifolium subterraneum for three straight years resulted in a significant rise in nitric nitrogen, ammoniacal nitrogen, and nitrogen cycle bacteria. Campiglia et al. observed similar wheat yields when they used underground clover as living mulch in intercropping systems. In addition, Guardia et al. showed that the mitigating effect of the legume green manure (vetch) mostly comes from less synthetic nitrogen being added to the next cash crop, less indirect N 2 O emissions from NO 3 − -leaching, and more carbon being stored because photosynthesis is higher. More specifically, green manures are essential for improving soil health via a mix of physical, chemical, and biological processes that promote sustainable agroecosystems. The integration of green manure biomass augments soil organic matter, thereby enhancing soil structure, increasing water retention, and facilitating aeration, which fosters optimal circumstances for root growth . This organic supplement serves as a nutrient reservoir, progressively releasing vital macronutrients such as nitrogen, phosphate, and potassium throughout decomposition and thus diminishing dependence on synthetic fertilizers. Green manures increase the diversity and activity of microbes in the soil. They do this by creating a dynamic food web that supports good microbes and stops soil-borne diseases through competitive exclusion and antibiotic synthesis. Furthermore, green manures facilitate carbon sequestration by integrating plant-derived carbon into stable soil organic components, thus reducing greenhouse gas emissions and enhancing soil resistance to climate change . Adding leguminous green manures also helps fix nitrogen, and the breakdown of bioactive chemicals like glucosinolates in cruciferous green manures acts as natural biofumigants, reducing the number of pests and diseases that can affect the crop. Integrating green manures into cropping systems enhances soil health and bolsters agroecosystem sustainability, establishing it as a fundamental practice for climate-resilient agriculture . Therefore, green manure and cover crops should be adopted to protect soil health, reduce soil-borne diseases as well as refine water use efficiency and many more beneficial impacts, as depicted in the conceptual . The application of green manure is an additional practice for enhancing the resilience of agroecosystems and rhizosphere processes and increasing soil’s productive capacity through nutrient availability. This application is geared towards maintaining agroecosystems and long-term environmental sustainability by enhancing processes such as nutrient cycling and preserving and maintaining soil biodiversity and the physical and chemical stability of the soil. Incorporating green manure can positively affect soil microbial communities by directly adding nutrients after incorporation and indirectly by changing plant and soil properties . Directly, the decomposition of green manure offers an immediate supply of carbon, nitrogen, and other vital elements; hence, it enhances microbial activity and fosters the proliferation of beneficial bacteria . The mineralization of organic matter from green manure produces labile chemicals that act as substrates for soil microorganisms, thereby boosting microbial biomass and diversity . The incorporation of green manure indirectly modifies the soil’s physical and chemical characteristics, including pH, soil structure, and water retention, fostering a conducive environment for microbial colonization and activity . At the same time, Sun et al. reported that the plant could affect soil microbial diversity and community by releasing root exudates. The decomposing residues of brassica green manure release glucosinolates, which assist in controlling parasitic nematodes . Furthermore, the exclusive use of grass species may improve the soil nutrient profile compared to monoculture. The advantages of green manure designate it as a sustainable practice capable of providing various agro-ecological services; however, effective management of green manure is essential to substantially enhance soil carbon stocks and mitigate climate change, particularly in semi-arid Mediterranean regions, by decreasing greenhouse gas emissions . Furthermore, many studies have proven that planting and incorporating green manure have multiple beneficial effects on soil health and the rhizosphere process . The different types of green manure and their beneficial effects on the rhizosphere are shown in . Soil health is a vital component of sustainable agriculture, signifying the soil’s ability to operate as a living ecosystem that promotes plant productivity, preserving environmental quality, and supporting biodiversity . Healthy soils exhibit an appropriate combination of physical, chemical, and biological attributes, encompassing adequate organic matter, nutrient accessibility, optimal structure, and vibrant microbial communities. Beneficial microorganisms, including nitrogen-fixing bacteria, mycorrhizal fungi, and decomposers, are vital to nutrient cycling, organic waste breakdown, and disease suppression . Soil health is a primary requirement for agriculture and plant growth, but most agricultural soils are poor in organic nutrients and organic matter due to the high applications of inorganic or chemical-based fertilizers to achieve high crop yields. Different agricultural techniques should be adopted; otherwise, it can become very hazardous and lead to a lack of organic matter. Improving soil health through chemical, physical, and biological properties is attributed to the integrated application of green manure and organic application due to better nutrient uptake and preservation of soil health . A variety of green manures, including grains, root crops, legumes, and oil crops, can serve as vegetative covers, each offering different benefits for enhancing soil health. Therefore, growing and incorporating green manure in agricultural lands helps maintain and preserve soil health and soil nutrient availability and ultimately increases soil organic matter content. The incorporation of green manure provides soil nutrients and organic matter to the soil, and organic matter plays an important role in soil biodiversity, soil microbiota, and the arrangement of soil aggregates to improve the soil structure . Therefore, incorporating green manure leads to an increased organic matter content, which has an important role in soil health and quality. Furthermore, green manure suppresses soil-borne diseases through biofumigation. Soil-borne diseases also affect soil health, reduce crop productivity, and negatively affect soil biodiversity. Many soil-borne pathogens have negative effects on plant growth and crop production, such as Phytophthora spp., Fusarium spp., Verticillium spp., and Rhizoctonia spp. Soil-borne diseases reduce agricultural yield by as much as 50–70%, including those of wheat, vegetables, and cotton . However, the practices of green manure in agricultural lands produce significant benefits such as increased soil nutrient availability, enhanced soil organic carbon, reduced soil compaction, improved particle aggregation and soil structure, as well as strengthened microbe activity, diversity, abundance, and suppression of soil-borne diseases. To protect crops from soil-borne diseases, many fumigants and fungicides need to be used regularly. Later, it was noted that the use of fumigants and fungicides or any other chemical-based product causes ecological problems such as human health hazards and reduced beneficial microbes in the soil biodiversity, directly affecting soil health . Consequently, organically, alternative strategies should be adopted to reduce soil-borne diseases, preserve soil health, and improve crop production. Alternatively, incorporation of green manure and cover crops are useful strategies for protecting soil health from soil-borne diseases and improving soil and plant health through biofumigation, providing nutrients and increasing beneficial microbial activity. Hao et al. reported that broccoli ( B. oleracea ) grown as green manure reduces the S. minor sclerotia in the rhizosphere. Whalen also reported that Fusarium wilt is another soil-borne disease that reduces watermelon production in the southeastern United States. The soil-borne disease Fusarium wilts is due to Fusarium oxysporum , which attacks the crop of watermelon and reduces the annual number of fruits, directly affecting the annual yield of watermelon production. For the management of Fusarium wilt disease, Hairy Vetch ( Vicia villosa L.) green manure was incorporated. After incorporating Hairy Vetch, the occurrence of Fusarium wilt disease in the rhizosphere of the watermelon field was reduced, leading to an increased annual watermelon yield by 45% . Papp et al. demonstrated that the cultivation of green manures influences soil microorganism environments, which are essential for sustaining soil functions and ecological system sustainability because they participate in nutrient cycling and organic matter turnover, providing multiple services. Researchers have observed that the incorporation of green manure modifies the composition and dynamics of soil fungus and bacterial communities and encourages beneficial microorganisms. Ntalli and Caboni observed that certain cruciferous green manures, including canola ( Brassica napus L.) and rape ( Brassica rapa L.), upon termination, release isothiocyanates (ITCs) via the hydrolysis of glucosinolates (GSLs) during the decomposition of their tissues, serving as effective biofumigants against various soil-borne pathogens and pests. The utilization of cruciferous species as green manure may facilitate the natural management of possible soil-borne diseases. Furthermore, Waisen et al. demonstrated that brown mustard ( Brassica juncea L.) and radish ( Raphanus raphanistrum subsp. sativus) serve as excellent biofumigant crops that protect against plant–parasitic nematodes while keeping the soil healthy and maintaining the strength of the nematode population structure strong. Aydınlı and Mennan also found that planting arugula ( Eruca sativa ) and radish ( R. sativus ) as winter crops before planting plants that are easily damaged by the root-knot nematode Meloidogyne arenaria may lower the number of eggs laid, the gall index, and the damage that occurs, while also increasing crop yields. Leguminous green manure can substantially enrich the soil with nitrogen throughout its growth and can acidify the rhizosphere by enhancing the absorption of insoluble phosphorus into the soil . Leguminous green manure can fix atmospheric nitrogen via symbiosis with rhizobia in root nodules . Furthermore, the fixed nitrogen can be allocated to intercropped non-leguminous plants within mixed cropping systems, or it may succeed crops in rotational practices. Biological nitrogen-fixing (BNF) systems may decrease the reliance on commercial nitrogen fertilizers . Further, Scavo et al. reported that the existence of Trifolium subterraneum for three straight years resulted in a significant rise in nitric nitrogen, ammoniacal nitrogen, and nitrogen cycle bacteria. Campiglia et al. observed similar wheat yields when they used underground clover as living mulch in intercropping systems. In addition, Guardia et al. showed that the mitigating effect of the legume green manure (vetch) mostly comes from less synthetic nitrogen being added to the next cash crop, less indirect N 2 O emissions from NO 3 − -leaching, and more carbon being stored because photosynthesis is higher. More specifically, green manures are essential for improving soil health via a mix of physical, chemical, and biological processes that promote sustainable agroecosystems. The integration of green manure biomass augments soil organic matter, thereby enhancing soil structure, increasing water retention, and facilitating aeration, which fosters optimal circumstances for root growth . This organic supplement serves as a nutrient reservoir, progressively releasing vital macronutrients such as nitrogen, phosphate, and potassium throughout decomposition and thus diminishing dependence on synthetic fertilizers. Green manures increase the diversity and activity of microbes in the soil. They do this by creating a dynamic food web that supports good microbes and stops soil-borne diseases through competitive exclusion and antibiotic synthesis. Furthermore, green manures facilitate carbon sequestration by integrating plant-derived carbon into stable soil organic components, thus reducing greenhouse gas emissions and enhancing soil resistance to climate change . Adding leguminous green manures also helps fix nitrogen, and the breakdown of bioactive chemicals like glucosinolates in cruciferous green manures acts as natural biofumigants, reducing the number of pests and diseases that can affect the crop. Integrating green manures into cropping systems enhances soil health and bolsters agroecosystem sustainability, establishing it as a fundamental practice for climate-resilient agriculture . Therefore, green manure and cover crops should be adopted to protect soil health, reduce soil-borne diseases as well as refine water use efficiency and many more beneficial impacts, as depicted in the conceptual . Understanding the value of agroecosystems requires knowledge of soil, its different properties, and the processes that maintain its health. Unfortunately, the soil is currently being degraded rapidly due to extensive application of chemical-based fertilizers, lack of agricultural techniques, and irregular cropping are rapidly degrading the soil, but what is most concerning is that soil is a non-renewable resource at a human temporal scale . Therefore, scientists and researchers have been trying to adopt alternative strategies to protect soil health, ultimately enhancing agroecosystems and rhizosphere processes. Cropping systems, such as intercropping and crop rotation practices, could serve this purpose in agricultural practices, thereby positively influencing soil health and agroecosystems . Cropping systems are adopted to maximize the crop yields of agroecosystems as well as to maintain agroecosystems through soil health and quality preservation. In contrast, the minimization of chemical-based products directly or indirectly affects agroecosystems and rhizosphere processes . With the primary goal of soil health maintenance to ensure the long-term stability of agroecosystems and the high production of crops, we need to adopt agronomic practices such as intercropping or crop rotations. Intercropping practices can improve agroecosystems by reducing the use of chemical-based fertilizers and soil pollution, preserving soil microbial diversity. 5.1. Intercropping and Their Effects on Rhizosphere Processes and Soil Health Intercropping, which involves the cultivation of two or more crops in a field concurrently, is a useful strategy in agronomic practice. Different crops, such as cover and cash crops, grow at the same time. The practice has gained global attention due to its beneficial effects on the enhancement of agroecosystems and soil health. In brief, intercropping practices can maintain agroecosystems and soil health by the reduced use of chemical-based fertilizers , enhancing soil-plant functions , suppressing the occurrence of soil-borne diseases , increasing soil nutrient and organic matter contents , and promoting the beneficial soil microorganisms, which support the rhizosphere processes . Cong et al. reported that intercropping systems, including that of faba beans, corn, and wheat, had about 11%, 4%, and 23% higher below-ground plant biomass as well as increased organic carbon (C) and nitrogen (N) contents than those in plant species rotations found in Gansu, China. Furthermore, de Medeiros et al. reported that intercropping with beans and pigeon peas considerably minimizes black root rot ( Scytalidium lignicola ) in cassava by up to 50% when matched with cassava in monoculture systems, as well as enhanced the rhizosphere process, that is, increasing the soil enzymes activities, microbial biomass, and soil nutrients related to organic C and N. Increasing the use of intercropping with different plant species have been typically interconnected with the enhancement of rhizosphere processes, providing environmental benefits . For example, leguminous species are advantageous for soil by providing nitrogen-fixing microorganisms and enhancing soil enzyme activities. On the other hand, non-leguminous species usually provide soil cover and nitrogen availability; such intercropping self-regulates soil nitrogen levels to improve soil nutrient use and reduce the C footprint . Furthermore, Graf et al. also reported that compared to sole cropping, intercropping practices on Dactylis glomerata and Medicago sativa have led to increased shoot biomass and nitrous oxide (N 2 O) production rate, which suggests that understanding the intercropping strategies is helpful for the maintenance of agroecosystems and soil health. Intercropping is regaining agronomical and ecological importance because of its ability to self-regulate soil nutrients and preserve soil health. An overview of intercropping in agroecosystems and its beneficial interactions based on Brooker et al. and their outcomes are shown in . Furthermore, legume and cereal intercropping systems have been extensively employed to address limitations on natural resources, promote sustainable agricultural development, and guarantee food security due to their economic and ecological advantages . Faba/wheat intercropping decreased nitrogen fertilizer application by 5–16% while simultaneously enhancing wheat production by 15–25% . However, Luo et al. reported that a 40 percent reduction in synthetic nitrogen input, coupled with soybean intercropping, might sustain sugarcane productivity and promote sustainable soil management. Many research studies have demonstrated that below-ground facilitative interactions among species can improve nutrient availability and utilization. Ref. revealed that the root interactions between maize and faba beans, along with the root exudates from maize, significantly improved the symbiotic N 2 fixation and nodulation in faba beans grown alongside maize. Intercropping can make crop roots release more organic acids and control the mineralization of organic phosphorus, the dissolution of inorganic phosphorus, or both. This makes more phosphorus available in the soil. Recent research shows that growing cereals and legumes together improves metabolic function and changes the structure of the soil’s microbial community through interactions between species . According to Zhang et al. , elevated carbon uptake by maize ( Zea mays L.) promoted nutrient cycling by altering the abundance of functional groups in soil microbes and improving the stability and complexity of microbial networks in soybean/maize intercropping systems. Teshita et al. proposed that, particularly in low-nitrogen soils, intercropping maize with alfalfa ( Medicago sativa L.) could enhance the structural complexity of the soil food web. Moreover, intercropping improves agricultural sustainability by utilizing crop diversity to exploit synergistic interactions across species . This method improves crop yield and nitrogen utilization efficiency through similar and beneficial interactions. The rhizosphere of leguminous plants cultivates distinct microbial communities, including nitrogen-fixing bacterium (nifh) populations, via root exudates in legume-based intercropping systems, enhancing ecological benefits such as nitrogen fixation and phosphorus mobilization . Within these approaches, maize/soybean intercropping is notable for its effective transfer of fixed nitrogen from soybean to maize, hence improving nitrogen efficiency and yields . Multiple factors, such as cropping configurations and fertilization methods, affect how well maize and soybean intercropping systems work. Improving microbes is crucial for nutrient absorption and yield maximization . Understanding how rhizosphere microbial communities react, especially in terms of their ability to fix nitrogen, would help improve intercropping patterns and fertilizer management, which would lead to more sustainable farming and food security. 5.2. Crop Rotation and Their Effects on Rhizosphere Processes and Soil Crop rotation is a conventional and effective method for maintaining agroecosystems, managing biodiversity by improving soil health, suppressing disease and pest occurrences, and hence enhancing crop yields . The efficiency and value of crop rotation are dependent upon various elements, including the types of crops utilized in the rotation, the sequence and frequency of specific crops, the duration of rotation, the agronomic history of the agricultural land, and the properties of the soil . Recent studies reported that farmers had used crop rotation through traditional methods to manage crop pests and diseases and manage soil productivity . Usually, these crop rotations involve growing three or four different types of crops in a sequence. However, with increasing food demand and agricultural production in recent years, many farmers grow just one or two crop species, with pesticides and mineral fertilizers contributing to compensate for the lack of crop rotations . The crop rotations are not always helpful and also decline the soil quality, and crop production has been reported with the number of crops grown linked with short rotation or continuous cropping, including soybean, sugarcane, maize, wheat, and oilseed rape . Interestingly, in crop rotation, oilseed rape is very important for the rotation and even highly profitable when grown as a break crop for cereals. Crop rotations with oilseed rape and maize affect the dynamics and structure of soil plant-associated microbial communities and are known to be beneficial for soil health via suppression of soil-borne diseases and plant growth promotion . Microbial species living nearby or associated with plants are directly influenced by the root’s architecture and the chemical characteristics of root exudates when we rotate with maize or oilseed rape crops . The direct influence of plant roots around soil is known as the rhizosphere. The rhizosphere is the habitat of plant-beneficial microorganisms and plant pathogens . With the research evidence of Benitez, Osborne, and Lehman , crop rotations promote beneficial plant growth-promoting microbes, affect the soil microbial community in the rhizosphere, and influence maize seedling growth characteristics. Furthermore, rotating crops with grain legumes can significantly improve the yield and protein content of subsequent wheat crops, owing to the increased nitrogen availability in the soil from legume biological fixation . Leguminous crops (such as peas and chickpeas) and various chickpea genotypes (cultivars) in rotation can alter soil functional microbial populations and affect the productivity of pulse crops and subsequent wheat crops. Different crops can produce root exudates and distinct residues that enhance soil microbial diversity and activity, as well as microbial biomass, nitrogen, and carbon cycling . Modifications in crop rotation duration and frequency of identical crops throughout time might influence the prevalence of root rot diseases and improve crop production and soil health. Short series rotations exhibit heightened sensitivity to host-specific diseases, resulting in poorer yields compared to lengthier series rotations . In addition, Lupwayi et al. found that the wheat phase of a five-year rotation exhibited greater microbial biomass in both rhizosphere and bulk soil compared to the three-year rotation of the wheat phase, likely due to increased carbon inputs from crop residues. In Canada, particularly in the western part, two phases of pea in a four-year rotation increased soil nitrogen levels, whereas three legume phases markedly altered the function and composition of the rhizosphere bacterial community in comparison to continuous wheat cultivation and growth . Increasing the rate of one crop in rotation can be detrimental to soil health. For example, Bainard et al. observed that a higher pulse phase in rotation caused host-specific fungal pathogens to build up in the soil, which could make rotation less beneficial for crop yield and soil health. Meanwhile, an additional crucial factor in crop rotation design is the ability of soil-borne diseases to utilize alternate crops as hosts or to remain dormant in the soil for extended periods, as well as the response of these crops to disease. Utilizing non-host plants in crop rotations to manage diseases transmitted by the soil is critical for mitigating yield losses, particularly given that many pathogens can persist in the soil for extended periods as spores or other latent forms in the absence of their preferred host plants . For example, Nayyar et al. reported that Fusarium root rot in peas cultivated in rotation on the Canadian prairie was associated with a restricted soil microbial community and reduced populations of arbuscular mycorrhizal fungi and beneficial bacteria. In certain instances, continuous cropping with enhanced crop diversification elevated the population of antagonistic soil microorganisms, hence lowering the populations of soil pathogens and reducing the “take-all” effect in wheat. Typically, incorporating three or more crops within a cropping system can improve soil health and optimize the productivity of crops . However, the benefits of crop rotation are practical; understanding the specific effects of crop systems and how they affect subsequent crops will encourage and adopt a worthy crop rotation, which will promote soil health and increase crop productivity. These findings make it easier to create customized interventions that use microbial populations to maintain soil health, decrease the emissions of greenhouse gases (GHGs) and crop production, and the efficient use of resources in intercropping systems. According to references, we propose intercropping in agroecosystems, as depicted in the conceptual . Intercropping, which involves the cultivation of two or more crops in a field concurrently, is a useful strategy in agronomic practice. Different crops, such as cover and cash crops, grow at the same time. The practice has gained global attention due to its beneficial effects on the enhancement of agroecosystems and soil health. In brief, intercropping practices can maintain agroecosystems and soil health by the reduced use of chemical-based fertilizers , enhancing soil-plant functions , suppressing the occurrence of soil-borne diseases , increasing soil nutrient and organic matter contents , and promoting the beneficial soil microorganisms, which support the rhizosphere processes . Cong et al. reported that intercropping systems, including that of faba beans, corn, and wheat, had about 11%, 4%, and 23% higher below-ground plant biomass as well as increased organic carbon (C) and nitrogen (N) contents than those in plant species rotations found in Gansu, China. Furthermore, de Medeiros et al. reported that intercropping with beans and pigeon peas considerably minimizes black root rot ( Scytalidium lignicola ) in cassava by up to 50% when matched with cassava in monoculture systems, as well as enhanced the rhizosphere process, that is, increasing the soil enzymes activities, microbial biomass, and soil nutrients related to organic C and N. Increasing the use of intercropping with different plant species have been typically interconnected with the enhancement of rhizosphere processes, providing environmental benefits . For example, leguminous species are advantageous for soil by providing nitrogen-fixing microorganisms and enhancing soil enzyme activities. On the other hand, non-leguminous species usually provide soil cover and nitrogen availability; such intercropping self-regulates soil nitrogen levels to improve soil nutrient use and reduce the C footprint . Furthermore, Graf et al. also reported that compared to sole cropping, intercropping practices on Dactylis glomerata and Medicago sativa have led to increased shoot biomass and nitrous oxide (N 2 O) production rate, which suggests that understanding the intercropping strategies is helpful for the maintenance of agroecosystems and soil health. Intercropping is regaining agronomical and ecological importance because of its ability to self-regulate soil nutrients and preserve soil health. An overview of intercropping in agroecosystems and its beneficial interactions based on Brooker et al. and their outcomes are shown in . Furthermore, legume and cereal intercropping systems have been extensively employed to address limitations on natural resources, promote sustainable agricultural development, and guarantee food security due to their economic and ecological advantages . Faba/wheat intercropping decreased nitrogen fertilizer application by 5–16% while simultaneously enhancing wheat production by 15–25% . However, Luo et al. reported that a 40 percent reduction in synthetic nitrogen input, coupled with soybean intercropping, might sustain sugarcane productivity and promote sustainable soil management. Many research studies have demonstrated that below-ground facilitative interactions among species can improve nutrient availability and utilization. Ref. revealed that the root interactions between maize and faba beans, along with the root exudates from maize, significantly improved the symbiotic N 2 fixation and nodulation in faba beans grown alongside maize. Intercropping can make crop roots release more organic acids and control the mineralization of organic phosphorus, the dissolution of inorganic phosphorus, or both. This makes more phosphorus available in the soil. Recent research shows that growing cereals and legumes together improves metabolic function and changes the structure of the soil’s microbial community through interactions between species . According to Zhang et al. , elevated carbon uptake by maize ( Zea mays L.) promoted nutrient cycling by altering the abundance of functional groups in soil microbes and improving the stability and complexity of microbial networks in soybean/maize intercropping systems. Teshita et al. proposed that, particularly in low-nitrogen soils, intercropping maize with alfalfa ( Medicago sativa L.) could enhance the structural complexity of the soil food web. Moreover, intercropping improves agricultural sustainability by utilizing crop diversity to exploit synergistic interactions across species . This method improves crop yield and nitrogen utilization efficiency through similar and beneficial interactions. The rhizosphere of leguminous plants cultivates distinct microbial communities, including nitrogen-fixing bacterium (nifh) populations, via root exudates in legume-based intercropping systems, enhancing ecological benefits such as nitrogen fixation and phosphorus mobilization . Within these approaches, maize/soybean intercropping is notable for its effective transfer of fixed nitrogen from soybean to maize, hence improving nitrogen efficiency and yields . Multiple factors, such as cropping configurations and fertilization methods, affect how well maize and soybean intercropping systems work. Improving microbes is crucial for nutrient absorption and yield maximization . Understanding how rhizosphere microbial communities react, especially in terms of their ability to fix nitrogen, would help improve intercropping patterns and fertilizer management, which would lead to more sustainable farming and food security. Crop rotation is a conventional and effective method for maintaining agroecosystems, managing biodiversity by improving soil health, suppressing disease and pest occurrences, and hence enhancing crop yields . The efficiency and value of crop rotation are dependent upon various elements, including the types of crops utilized in the rotation, the sequence and frequency of specific crops, the duration of rotation, the agronomic history of the agricultural land, and the properties of the soil . Recent studies reported that farmers had used crop rotation through traditional methods to manage crop pests and diseases and manage soil productivity . Usually, these crop rotations involve growing three or four different types of crops in a sequence. However, with increasing food demand and agricultural production in recent years, many farmers grow just one or two crop species, with pesticides and mineral fertilizers contributing to compensate for the lack of crop rotations . The crop rotations are not always helpful and also decline the soil quality, and crop production has been reported with the number of crops grown linked with short rotation or continuous cropping, including soybean, sugarcane, maize, wheat, and oilseed rape . Interestingly, in crop rotation, oilseed rape is very important for the rotation and even highly profitable when grown as a break crop for cereals. Crop rotations with oilseed rape and maize affect the dynamics and structure of soil plant-associated microbial communities and are known to be beneficial for soil health via suppression of soil-borne diseases and plant growth promotion . Microbial species living nearby or associated with plants are directly influenced by the root’s architecture and the chemical characteristics of root exudates when we rotate with maize or oilseed rape crops . The direct influence of plant roots around soil is known as the rhizosphere. The rhizosphere is the habitat of plant-beneficial microorganisms and plant pathogens . With the research evidence of Benitez, Osborne, and Lehman , crop rotations promote beneficial plant growth-promoting microbes, affect the soil microbial community in the rhizosphere, and influence maize seedling growth characteristics. Furthermore, rotating crops with grain legumes can significantly improve the yield and protein content of subsequent wheat crops, owing to the increased nitrogen availability in the soil from legume biological fixation . Leguminous crops (such as peas and chickpeas) and various chickpea genotypes (cultivars) in rotation can alter soil functional microbial populations and affect the productivity of pulse crops and subsequent wheat crops. Different crops can produce root exudates and distinct residues that enhance soil microbial diversity and activity, as well as microbial biomass, nitrogen, and carbon cycling . Modifications in crop rotation duration and frequency of identical crops throughout time might influence the prevalence of root rot diseases and improve crop production and soil health. Short series rotations exhibit heightened sensitivity to host-specific diseases, resulting in poorer yields compared to lengthier series rotations . In addition, Lupwayi et al. found that the wheat phase of a five-year rotation exhibited greater microbial biomass in both rhizosphere and bulk soil compared to the three-year rotation of the wheat phase, likely due to increased carbon inputs from crop residues. In Canada, particularly in the western part, two phases of pea in a four-year rotation increased soil nitrogen levels, whereas three legume phases markedly altered the function and composition of the rhizosphere bacterial community in comparison to continuous wheat cultivation and growth . Increasing the rate of one crop in rotation can be detrimental to soil health. For example, Bainard et al. observed that a higher pulse phase in rotation caused host-specific fungal pathogens to build up in the soil, which could make rotation less beneficial for crop yield and soil health. Meanwhile, an additional crucial factor in crop rotation design is the ability of soil-borne diseases to utilize alternate crops as hosts or to remain dormant in the soil for extended periods, as well as the response of these crops to disease. Utilizing non-host plants in crop rotations to manage diseases transmitted by the soil is critical for mitigating yield losses, particularly given that many pathogens can persist in the soil for extended periods as spores or other latent forms in the absence of their preferred host plants . For example, Nayyar et al. reported that Fusarium root rot in peas cultivated in rotation on the Canadian prairie was associated with a restricted soil microbial community and reduced populations of arbuscular mycorrhizal fungi and beneficial bacteria. In certain instances, continuous cropping with enhanced crop diversification elevated the population of antagonistic soil microorganisms, hence lowering the populations of soil pathogens and reducing the “take-all” effect in wheat. Typically, incorporating three or more crops within a cropping system can improve soil health and optimize the productivity of crops . However, the benefits of crop rotation are practical; understanding the specific effects of crop systems and how they affect subsequent crops will encourage and adopt a worthy crop rotation, which will promote soil health and increase crop productivity. These findings make it easier to create customized interventions that use microbial populations to maintain soil health, decrease the emissions of greenhouse gases (GHGs) and crop production, and the efficient use of resources in intercropping systems. According to references, we propose intercropping in agroecosystems, as depicted in the conceptual . Soil health typically refers to the capacity of soil to operate as a crucial living system that supports biological productivity, preserves both soil and plant health, enhances environmental quality, and maintains an agroecosystem . Soil is a highly intricate and complex, multifunctional system including gaseous, liquid, and solid elements that interact through various chemical, physical, and biological processes. Healthy soil sustains agroecosystems and facilitates the provision of essential services to ecosystems . In order to evaluate and sustain agroecosystems, it is essential to consider the biological, physical, and chemical functions and traits of the soil, particularly the biological ones that serve as sensitive indicators of soil health and quality . Microbiological and biochemical markers show that the variety and activity of soil microbes are important for the long-term health of agroecosystems because they keep soil health functions going, like cycling carbon and nutrients . Microbial indicators exhibit more sensitivity than physical traits to environmental alterations such as soil usage and management; hence, they enable early detection and forecasting of changes and disturbances in environmental sustainability . Related studies recognize soil microbial biomass, which includes fungi, bacteria, and algae, as a primary biological indicator of soil health and as a crucial source of nutrient cycling and delivery relative to plant demand . Additionally, through photosynthesis, plants fix and transfer carbon as carbohydrates into the food web, making it one of the most critical biological processes on Earth . Both agricultural and forestry soils widely use soil respiration and microbial biomass as bio-indicators of soil health. We expect biochemical indicators to effectively integrate the combined impact on soil’s chemical, physical, and biological processes and characteristics, making them suitable for diverse management and environmental conditions. This paragraph aims to elucidate and augment our focus on soil quality and related research, particularly with biochemical soil quality indicators such as enzymatic and microbial activity . Due to their role in nutrient cycling through microbial processes, soil enzyme activities frequently serve as sensitive indicators of soil biochemical quality and as measures of ecological quality. They can effectively respond to both anthropogenic and natural alterations in soil and are easily quantifiable . However, soil enzymes are important in the functioning of soil ecosystems, including nutrient and carbon cycling and maintenance . For example, soil enzymes have an important role in C ( β -glucosidase and galactosidase), N (urease), and P (acid phosphatases and alkaline) cycling, and these enzymes play crucial roles in the breakdown of OM and release nutrients, which are important for soil health and plant growth promotion. Dick and Tabatabai reported that soil enzymes serve as reliable indicators of soil quality because of their close association with soil biological processes, their ease of measurement, and their swift response to alterations in soil management. In summary, soil enzymes are valuable tools for assessing both long and short-term changes in soil quality and management practices. Furthermore, indicators of soil health for sustaining agroecosystems need to be associated with soil processes and responsive to alterations in management and environmental conditions. Soil biological traits, such as microbial biomass and its activities, serve as sensitive and rapid indicators that capture responses and data from diverse environmental conditions . We use soil biological metrics, such as microbial density, activity, and biomass, as indicators of soil health . Furthermore, a specific drawback of employing soil microbial characteristics as soil health indicators is the technological limitations encountered in the research of soil microbial populations. Molecular techniques that are more advanced, especially next-generation sequencing (NGS) methods (such as shotgun sequencing for structural and amplicon sequencing and functional microbial diversity investigations), have made it possible to study the part of soil microbial communities that cannot be grown in a lab. This is because most soil microorganisms do not grow on media or in a lab setting. However, while analyzing the data, we must acknowledge the limitations of these novel methodologies. In conclusion, we can use soil to focus on agroecosystem maintenance and enhance rhizosphere processes; soil health indicators are used to inform management techniques, optimize fertilizer application, and promote sustainable crop productivity. Integrated agricultural practices are essential for improving rhizospheric processes, which are important for sustainable crop yield and soil health. Integrated agricultural practices established soil properties (biological, chemical, and physical properties) that influence nutrient cycling and soil microbial community composition and structure and enhance the rhizospheric process . However, to reduce adverse effects, integrative biofertilizers, crop rotation, green manuring, intercropping, and mulching practices should be promoted and tested worldwide to improve the rhizosphere processes and agroecosystems resilience. The disturbance of rhizosphere processes and depletion of organic matter are restraining agricultural productivity around the globe via improper agricultural practices. In short, integrated agricultural practices (IAP) are based on the integration of different crops, livestock wastes, and fertilizers into production systems that, through valuable management practices, maintain a high level of soil fertility, productivity, and quality, reduce external inputs of agrochemical and fertilizers, and enhance the functions of soil inside biological cycles and its processes. Mixed-crop livestock and other available resources, such as seed primed—the basis of integrated agricultural practices—allow the most efficient and effective use of natural resources and biological cycles to improve below-ground activities . Sarkar et al. reported that microbial-assisted nutrient management is a sustainable, eco-friendly, and cost-effective option under integrated agricultural systems. The usage of less amount of agrochemicals provides support organically or naturally to protect the environment from nutrient denitrification or runoff, land degradation, and soil pollution. Duarah et al. stated that reducing the amount of synthetic NPK fertilizer application and seed priming (with bacteria) can significantly improve nitrogen-use efficiency. However, Entesari et al. witnessed that bio-priming seeds of soybean with Trichoderma sp., prior to planting, enhanced the nutrient status of the crop and improved the crop yield. Phosphorus (P), an essential nutrient, is often fixed in the soil and forms complexes with metals, limiting its availability to plants. However, certain fungal and bacterial communities have a trait called P solubilization, which allows them to convert unavailable or fixed phosphorus into forms that plants can absorb. Kim et al. reported that incorporating green manure can help re-mobilize phosphorus and make non-exchangeable potassium available, circulating nutrients more efficiently. This integrated approach significantly enhances phosphorus availability, a key element for plant growth and the activity of soil microorganisms. Integrated Agricultural Practices (IAP) provide numerous advantages by carefully selecting crops and utilizing available resources to develop creative, sustainable systems. These practices effectively achieve several critical objectives: minimizing disease and insect problems, reducing inputs and energy requirements, and lowering the reliance on agrochemicals and pesticides. Additionally, IAP mitigates risks associated with climate variability and economic fluctuations, making agricultural systems more resilient. Improved rhizosphere processes foster healthy soils, which are vital for sustaining ecosystem services that benefit plants, animals, and humans alike. By optimizing rhizospheric functions, agroecosystems can better support nutrient cycling, enhance soil biodiversity, and maintain productivity, ensuring the sustainable production of food and fiber. Ultimately, this integrated approach is essential for strengthening agroecosystem resilience, safeguarding long-term agricultural sustainability, and addressing the challenges of future global food security. Based on the evidence, recognizing the importance of rhizosphere processes and their interactions with microorganisms is essential for promoting sustainable agroecosystems. The rhizosphere fosters various microbial populations that augment nutrient accessibility, regulate pathogens, and promote soil structure. Some important things that must be performed to rebuild microbial communities, improve soil health, and make nutrient cycling easier are making green manure, intercropping, and using microbial inoculants. These solutions diminish dependence on synthetic fertilizers, augment carbon sequestration, and improve plant resilience to abiotic stress, fostering steady agricultural yields and economic advantages. Optimizing rhizosphere functions enables agroecosystems to enhance food and fiber production, preserve soil fertility, and ensure long-term sustainability, fostering resistance to environmental challenges and safeguarding global food security. However, the findings from this review provide valuable insights to agronomists, soil scientists, and crop managers on optimizing rhizosphere processes to improve agroecosystem resilience. In closing, the rhizosphere will be central to future ecological innovations in agriculture, driving both resilience and sustainability. However, further experimental research is required to fully unlock the prospect of the rhizosphere. Future studies should focus on integrating advanced technologies, such as microbial genomics and precision agriculture, to optimize rhizosphere management for more resilient and sustainable agroecosystems. The development of smart agricultural strategies, coupled with policy support, will be vital for addressing the growing challenges posed by climate change and soil degradation. |
Telenursing in the sexual function of women with breast cancer: A study protocol | eb069b72-3642-4abb-997c-b66b6d27bf23 | 9704939 | Internal Medicine[mh] | There is a consensus in the literature that most breast cancer patients have sexual problems. These problems are due to the diagnosis or cancer treatment, including surgery, chemotherapy, radiotherapy, and hormonal therapy, which interfere directly or indirectly with sexual functioning, whether through psychological, social, or biological aspects. Changes in the sexual functioning of breast cancer patients start from treatment and may last for a long period. The younger the patient, the more severe the symptoms or sexual problems will be. These problems directly impact the quality of life of these patients and their relationships. One of the major problems surrounding this issue is the fact that many healthcare practitioners do not address the sexuality complaints of patients with breast cancer during their appointments. Therefore, most patients seek information from other sources, thus highlighting communication barriers in the professional-patient relationship and the maintenance of a taboo around sexuality and its relevance to women, especially in the context of breast cancer treatment. This problem may have been further aggravated by the pandemic context experienced in the last 2 years, especially due to the need to reduce personal interactions to minimize the risk of COVID-19 transmission. Implementing healthcare and health assessment tools that do not expose professionals or cancer patients to COVID-19 contributed to the increase in telehealth activities delivered via phone or video. Few studies have examined how educational interventions by phone impact the sexual function of women with breast cancer undergoing treatment, especially at the beginning of cancer treatment and led by nurses. The International Council of Nurses considers telenursing a service that enables the maintenance of effective communication with clients with chronic noncommunicable diseases, in addition to providing an effective intervention in the promotion and education for a healthy life. There is, therefore, a need to evaluate the impact of educational interventions performed through telenursing on the sexual functioning of women with breast cancer more deeply, aiming to guide and inform these patients about the sexual problems arising from cancer diagnosis and treatment and help them deal with these problems. This study aims to test the effect of telenursing counseling on the sexual functioning of women undergoing breast cancer treatment.
2.1. Study design This is a randomized clinical trial (RCT) with 2 parallel groups, with an allocation ratio of 1:1, with implementation between February 2022 and February 2023. This protocol was developed according to the guidelines of the Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) (Supplemental data, Supplemental Digital Content, http://links.lww.com/MD/xxx ). 2.2. Study setting The study will be conducted in 2 reference hospitals in the state of Ceará, Brazil. One is a university hospital linked to the Federal University of Ceará. The other is a reference center for diagnosing, treating, and monitoring cancer in Ceará. 2.3. Subjects The study population will consist of women with breast cancer undergoing cancer treatment recruited using the following inclusion criteria: stage I, II, III or IV breast cancer; oncological treatment (surgery, chemotherapy, radiotherapy, or hormone therapy), as the literature indicates that all treatments interfere directly or indirectly in sexual functioning and that these effects may begin from treatment and remain for a long period; having at least one telephone number; and having a partner or spouse. The following subjects will be excluded: patients in treatment for sexual dysfunction or climacteric symptoms to reduce confounding variables in the assessment of the intervention’s outcome; treatment for another type of cancer; hearing impairment, as it makes communication by phone impossible; and having a medical diagnosis of any mental disorders since cognitive functioning is necessary to understand and fill in the data collection instruments. The discontinuity criteria are patients’ withdrawal from participating in the research after the commencement of data collection, death during the study, phone change after follow-up, and not answering phone calls after 5 attempts on different dates and times. There will be no change in the participant’s allocation to minimize the risk of bias. 2.4. Sample size G*power software, Version 3.1.9.6 (German University Heinrich-Heine Universitaet, Duesseldorf) , was used to calculate the sample size by selecting “Test F” as the family of the test, considering that the analysis of variance (ANOVA) of repeated measures with intra- and intergroup interaction will be used to verify the null hypothesis. The sample size estimation was made using an equation based on ANOVA, which took into account an effect of 0.085 (representing an effect size of at least 0.30 in the groups) and a power of 80% in the repeated measures design (2 groups, alpha = 0.05, nonsphericity correction = 0.5). Given the parameters above, the G*power software suggested 45 participants in each group (90 participants in total). Taking into account possible sample losses, a percentage of 20% was added, which totaled a sample of 108 participants. This effect size was stipulated by a previous study, which identified an intermediate effect size (approximately 30%) for sexual functioning among the groups (control and intervention). The sphericity corresponds to one of the assumptions the data must meet so that ANOVA can be used, ranging from 0.5 to 1, with 0.5 being the most rigorous. 2.5. Recruitment and data collection The patients will be approached in person at the study locations by a team of 2 researchers who will contribute to the data collection process. These researchers received a training of approximately 8 hours on the objectives and methodological procedures of the study, where simulations involving the data collection instrument were also performed. These researchers are also members of a research group that conducts research on mastectomized women at the Nursing Department of the Federal University of Ceará. After formal consent, sociodemographic and clinical data will be collected from the medical records and through the application of an adapted questionnaire. The questionnaire analyzes the following variables: age, education, color, marital status, length of the relationship, the average number of sexual intercourse/month, occupational status, presence of sexual problems before the diagnosis of breast cancer, city of origin, family income, religion, children, and the number of residents at home. Moreover, the following clinical data will be collected: weight, height, body mass index, menstruation status, presence of comorbidities, risk factors for breast cancer, time of diagnosis, primary tumor, lymph node and metastasis staging, previous surgery, therapeutic intervention currently implemented, and therapeutic interventions previously performed (Table ). Sexual functioning, which corresponds to the main outcome, will be analyzed using the Female Sexual Function Index (FSFI), an instrument of a specific nature and multidimensional scope that evaluates female sexual function. The scale has a possible maximum of 36 and a minimum of 2, which is obtained by summing the weighted scores of each of the 6 domains (desire, subjective arousal, orgasm, satisfaction, and pain). Higher scores reflect a better degree of sexual function (Fig. ). 2.6. Allocation Eligible participants will be randomly allocated to the intervention group or control group, with an allocation ratio of 1:1, using a simple randomization technique. After completion of the baseline evaluations, eligible patients will be randomized by a member of the research group that is not linked to the study. This member will assign a random numbering to each participant and then carry out a draw from the participants who will compose each group. Then, the principal researcher will carry out the intervention after being informed about the allocation of subjects in the groups. 2.7. Interventions The control group will not receive any intervention but may receive orientations about sexuality by healthcare professionals such as social workers, nurses, psychologists, or physicians (oncologists, mastologists, or gynecologists), eventually considering that none of the institutions where the study will be undertaken has specific protocols or routines aimed at sexual and reproductive health. Three telenursing counseling sessions will be carried out over 6 weeks in the intervention group. This period was chosen for being considerate enough to address the issues planned and to establish a bond and a relationship of trust between the researchers and the participants. Most educational interventions aimed at the sexual functioning of women with breast cancer are carried out in a period of 3 to 24 weeks. The sessions will take place at intervals of at least 15 days, considered enough for contacting all participants but not so long that would cause forgetfulness of the information provided in each contact. The calls will be made at times and days arranged with the participants and the researcher and are estimated to last 30 to 40 minutes. The telephone counseling will be conducted exclusively by the principal researcher to minimize the risks of divergence in the approach. The principal researcher has worked as a nurse practitioner in the public health system for approximately seven years, has a specialization degree in public health, has a master’s degree in nursing, and has worked with sexual dysfunction in women with breast cancer for 4 years. The content of the counseling sessions will be based on the guidelines provided by the American Cancer Society for managing female sexual problems related to cancer. The contents were organized into 3 main topics: Cancer, sex, and the female body; How can treatments interfere with sexual life? And: managing female sexual problems related to cancer. The counseling sessions will be based on the constructs of social-cognitive theory. Telenursing uses technologies to provide nursing care and direct nursing practice. Moreover, since 2008, the International Council of Nurses has considered telenursing a service that provides nurses with the administration of care to patients living in rural or remote areas and enables them to maintain effective communication with clients who have noncommunicable chronic diseases, providing an effective strategy to promote self-care and a healthy lifestyle. The telephone intervention eliminates visual contact, providing anonymity during the interactions. Furthermore, telephone calls enable greater inclusion of patients who live in rural areas or locations that are difficult to reach and who are often not prone to in-person interventions. Finally, telephone calls are considered an economical intervention method. 2.8. Relevant concomitant care permitted or prohibited during the trial Concomitant treatments for sexual dysfunction or menopausal symptoms will not be allowed during participation in the study since the comparison between the control and intervention groups does not allow cointerventions (medications, therapies, or behaviors) to avoid interferences regarding the outcome. 2.9. Outcome The study’s primary outcome will be sexual functioning, which will be assessed using the FSFI. The FSFI is an instrument built and validated in English and currently has at least 20 translations in different languages. The initial validation of the FSFI obtained a Cronbach’s alpha > 0.90 and reliability coefficient in the high general retest for all domains that make up the scale ( R = 0.79 to 0.86), demonstrating good construct validity with a statistically significant difference between the averages of groups with and without sexual desire disorder ( P < .001). The data above support the psychometric reliability and validity of the FSFI. In 2007, the FSFI was translated and validated into Portuguese (Brazil), obtaining a Cronbach’s alpha of 0.92 (95% CI: 0.90–0.93) and showing adequate internal consistency. The FSFI is a self-report questionnaire composed of 19 questions to evaluate female sexual functioning, which uses a Likert scale ranging from 0 to 5 and has 6 domains: desire, subjective arousal, lubrication, orgasm, satisfaction and pain (dyspareunia). Sexual function will be analyzed at baseline ( T 0), 6 weeks after the intervention ( T 1), and 12 weeks after the intervention ( T 2). In other studies that performed an intervention to improve the sexual functioning of women with breast cancer, the outcome was evaluated between 4 weeks and 12 months after the intervention. 2.10. Blinding Due to the nature of the intervention, it is impossible to blind the participants and the researcher responsible for providing and monitoring the intervention. Therefore, only research assistants who will evaluate the outcome after the intervention and the statistician who will perform the data analysis will be blinded concerning the participants’ group allocation. 2.11. Strategies to improve adherence to interventions Short Message Notifications will be sent to remember the subjects of the date and time of the scheduled sessions as a strategy to increase participants’ adherence to the study. Adherence will be assessed by participation in the telephone counseling sessions, which will be recorded in a specific form, including information such as the date and time of the intervention, duration, the topic addressed, and other relevant notes. In addition, a postintervention evaluation will include questions about the frequency, duration, and content offered through the counseling sessions. 2.12. Provisions for posttrial care Adverse events will be evaluated by the number of participants who report possible worsening symptoms in the primary endpoint. In addition, participants will indicate whether they have experienced worsening symptoms of any kind in the postintervention survey. 2.13. Statistical methods The data will be compiled and analyzed in the JASP program (Copyright 2013–2021 University of Amsterdam) Version JASP 0.15. Descriptive statistics will be obtained for the independent variables (means for continuous variables, absolute and relative frequencies for categorical variables, and their respective confidence intervals). An ANOVA statistical test for repeated measures will be completed to assess the extent to which the control and intervention groups differ concerning the outcome variable (sexual functioning) in the periods analyzed ( T 0, T 1, and T 2). All statistical analyses will be performed in collaboration with external statisticians, who will not be informed about the allocation of the groups. The dataset generated or analyzed during this study will not be publicly available due to the ethics review act. However, it will be available from the corresponding author upon reasonable request and under the Brazilian guidelines for research collaboration and data transfer. The study results will be submitted for publication in a peer-reviewed nursing journal. No professional writer will be involved. 2.14. Ethics approval and consent to participate Ethical approval was obtained by the Research Ethics Committee of the Federal University of Ceará, Brazil, under opinion no. 461609 and CAAE no. 43072721.9.0000.5054; and the Maternity School Assis Chateaubriand, Ceará, Brazil, with opinion number 4742687 and certificate of ethical appreciation number 43072721.9.3002.5050. Written informed consent will be obtained from all participants. All participants will receive a unique code number. The code will be stored separately from the survey data and will only be accessible to members of the research team. All data will be stored and processed under Brazilian law no. 13.709/2018 on data protection. The (Brazilian Clinical Trials Registry in Portuguese) will be notified of any changes in the trial registry if important changes regarding the conduction or protocol of the RCT are needed.
This is a randomized clinical trial (RCT) with 2 parallel groups, with an allocation ratio of 1:1, with implementation between February 2022 and February 2023. This protocol was developed according to the guidelines of the Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) (Supplemental data, Supplemental Digital Content, http://links.lww.com/MD/xxx ).
The study will be conducted in 2 reference hospitals in the state of Ceará, Brazil. One is a university hospital linked to the Federal University of Ceará. The other is a reference center for diagnosing, treating, and monitoring cancer in Ceará.
The study population will consist of women with breast cancer undergoing cancer treatment recruited using the following inclusion criteria: stage I, II, III or IV breast cancer; oncological treatment (surgery, chemotherapy, radiotherapy, or hormone therapy), as the literature indicates that all treatments interfere directly or indirectly in sexual functioning and that these effects may begin from treatment and remain for a long period; having at least one telephone number; and having a partner or spouse. The following subjects will be excluded: patients in treatment for sexual dysfunction or climacteric symptoms to reduce confounding variables in the assessment of the intervention’s outcome; treatment for another type of cancer; hearing impairment, as it makes communication by phone impossible; and having a medical diagnosis of any mental disorders since cognitive functioning is necessary to understand and fill in the data collection instruments. The discontinuity criteria are patients’ withdrawal from participating in the research after the commencement of data collection, death during the study, phone change after follow-up, and not answering phone calls after 5 attempts on different dates and times. There will be no change in the participant’s allocation to minimize the risk of bias.
G*power software, Version 3.1.9.6 (German University Heinrich-Heine Universitaet, Duesseldorf) , was used to calculate the sample size by selecting “Test F” as the family of the test, considering that the analysis of variance (ANOVA) of repeated measures with intra- and intergroup interaction will be used to verify the null hypothesis. The sample size estimation was made using an equation based on ANOVA, which took into account an effect of 0.085 (representing an effect size of at least 0.30 in the groups) and a power of 80% in the repeated measures design (2 groups, alpha = 0.05, nonsphericity correction = 0.5). Given the parameters above, the G*power software suggested 45 participants in each group (90 participants in total). Taking into account possible sample losses, a percentage of 20% was added, which totaled a sample of 108 participants. This effect size was stipulated by a previous study, which identified an intermediate effect size (approximately 30%) for sexual functioning among the groups (control and intervention). The sphericity corresponds to one of the assumptions the data must meet so that ANOVA can be used, ranging from 0.5 to 1, with 0.5 being the most rigorous.
The patients will be approached in person at the study locations by a team of 2 researchers who will contribute to the data collection process. These researchers received a training of approximately 8 hours on the objectives and methodological procedures of the study, where simulations involving the data collection instrument were also performed. These researchers are also members of a research group that conducts research on mastectomized women at the Nursing Department of the Federal University of Ceará. After formal consent, sociodemographic and clinical data will be collected from the medical records and through the application of an adapted questionnaire. The questionnaire analyzes the following variables: age, education, color, marital status, length of the relationship, the average number of sexual intercourse/month, occupational status, presence of sexual problems before the diagnosis of breast cancer, city of origin, family income, religion, children, and the number of residents at home. Moreover, the following clinical data will be collected: weight, height, body mass index, menstruation status, presence of comorbidities, risk factors for breast cancer, time of diagnosis, primary tumor, lymph node and metastasis staging, previous surgery, therapeutic intervention currently implemented, and therapeutic interventions previously performed (Table ). Sexual functioning, which corresponds to the main outcome, will be analyzed using the Female Sexual Function Index (FSFI), an instrument of a specific nature and multidimensional scope that evaluates female sexual function. The scale has a possible maximum of 36 and a minimum of 2, which is obtained by summing the weighted scores of each of the 6 domains (desire, subjective arousal, orgasm, satisfaction, and pain). Higher scores reflect a better degree of sexual function (Fig. ).
Eligible participants will be randomly allocated to the intervention group or control group, with an allocation ratio of 1:1, using a simple randomization technique. After completion of the baseline evaluations, eligible patients will be randomized by a member of the research group that is not linked to the study. This member will assign a random numbering to each participant and then carry out a draw from the participants who will compose each group. Then, the principal researcher will carry out the intervention after being informed about the allocation of subjects in the groups.
The control group will not receive any intervention but may receive orientations about sexuality by healthcare professionals such as social workers, nurses, psychologists, or physicians (oncologists, mastologists, or gynecologists), eventually considering that none of the institutions where the study will be undertaken has specific protocols or routines aimed at sexual and reproductive health. Three telenursing counseling sessions will be carried out over 6 weeks in the intervention group. This period was chosen for being considerate enough to address the issues planned and to establish a bond and a relationship of trust between the researchers and the participants. Most educational interventions aimed at the sexual functioning of women with breast cancer are carried out in a period of 3 to 24 weeks. The sessions will take place at intervals of at least 15 days, considered enough for contacting all participants but not so long that would cause forgetfulness of the information provided in each contact. The calls will be made at times and days arranged with the participants and the researcher and are estimated to last 30 to 40 minutes. The telephone counseling will be conducted exclusively by the principal researcher to minimize the risks of divergence in the approach. The principal researcher has worked as a nurse practitioner in the public health system for approximately seven years, has a specialization degree in public health, has a master’s degree in nursing, and has worked with sexual dysfunction in women with breast cancer for 4 years. The content of the counseling sessions will be based on the guidelines provided by the American Cancer Society for managing female sexual problems related to cancer. The contents were organized into 3 main topics: Cancer, sex, and the female body; How can treatments interfere with sexual life? And: managing female sexual problems related to cancer. The counseling sessions will be based on the constructs of social-cognitive theory. Telenursing uses technologies to provide nursing care and direct nursing practice. Moreover, since 2008, the International Council of Nurses has considered telenursing a service that provides nurses with the administration of care to patients living in rural or remote areas and enables them to maintain effective communication with clients who have noncommunicable chronic diseases, providing an effective strategy to promote self-care and a healthy lifestyle. The telephone intervention eliminates visual contact, providing anonymity during the interactions. Furthermore, telephone calls enable greater inclusion of patients who live in rural areas or locations that are difficult to reach and who are often not prone to in-person interventions. Finally, telephone calls are considered an economical intervention method.
Concomitant treatments for sexual dysfunction or menopausal symptoms will not be allowed during participation in the study since the comparison between the control and intervention groups does not allow cointerventions (medications, therapies, or behaviors) to avoid interferences regarding the outcome.
The study’s primary outcome will be sexual functioning, which will be assessed using the FSFI. The FSFI is an instrument built and validated in English and currently has at least 20 translations in different languages. The initial validation of the FSFI obtained a Cronbach’s alpha > 0.90 and reliability coefficient in the high general retest for all domains that make up the scale ( R = 0.79 to 0.86), demonstrating good construct validity with a statistically significant difference between the averages of groups with and without sexual desire disorder ( P < .001). The data above support the psychometric reliability and validity of the FSFI. In 2007, the FSFI was translated and validated into Portuguese (Brazil), obtaining a Cronbach’s alpha of 0.92 (95% CI: 0.90–0.93) and showing adequate internal consistency. The FSFI is a self-report questionnaire composed of 19 questions to evaluate female sexual functioning, which uses a Likert scale ranging from 0 to 5 and has 6 domains: desire, subjective arousal, lubrication, orgasm, satisfaction and pain (dyspareunia). Sexual function will be analyzed at baseline ( T 0), 6 weeks after the intervention ( T 1), and 12 weeks after the intervention ( T 2). In other studies that performed an intervention to improve the sexual functioning of women with breast cancer, the outcome was evaluated between 4 weeks and 12 months after the intervention.
Due to the nature of the intervention, it is impossible to blind the participants and the researcher responsible for providing and monitoring the intervention. Therefore, only research assistants who will evaluate the outcome after the intervention and the statistician who will perform the data analysis will be blinded concerning the participants’ group allocation.
Short Message Notifications will be sent to remember the subjects of the date and time of the scheduled sessions as a strategy to increase participants’ adherence to the study. Adherence will be assessed by participation in the telephone counseling sessions, which will be recorded in a specific form, including information such as the date and time of the intervention, duration, the topic addressed, and other relevant notes. In addition, a postintervention evaluation will include questions about the frequency, duration, and content offered through the counseling sessions.
Adverse events will be evaluated by the number of participants who report possible worsening symptoms in the primary endpoint. In addition, participants will indicate whether they have experienced worsening symptoms of any kind in the postintervention survey.
The data will be compiled and analyzed in the JASP program (Copyright 2013–2021 University of Amsterdam) Version JASP 0.15. Descriptive statistics will be obtained for the independent variables (means for continuous variables, absolute and relative frequencies for categorical variables, and their respective confidence intervals). An ANOVA statistical test for repeated measures will be completed to assess the extent to which the control and intervention groups differ concerning the outcome variable (sexual functioning) in the periods analyzed ( T 0, T 1, and T 2). All statistical analyses will be performed in collaboration with external statisticians, who will not be informed about the allocation of the groups. The dataset generated or analyzed during this study will not be publicly available due to the ethics review act. However, it will be available from the corresponding author upon reasonable request and under the Brazilian guidelines for research collaboration and data transfer. The study results will be submitted for publication in a peer-reviewed nursing journal. No professional writer will be involved.
Ethical approval was obtained by the Research Ethics Committee of the Federal University of Ceará, Brazil, under opinion no. 461609 and CAAE no. 43072721.9.0000.5054; and the Maternity School Assis Chateaubriand, Ceará, Brazil, with opinion number 4742687 and certificate of ethical appreciation number 43072721.9.3002.5050. Written informed consent will be obtained from all participants. All participants will receive a unique code number. The code will be stored separately from the survey data and will only be accessible to members of the research team. All data will be stored and processed under Brazilian law no. 13.709/2018 on data protection. The (Brazilian Clinical Trials Registry in Portuguese) will be notified of any changes in the trial registry if important changes regarding the conduction or protocol of the RCT are needed.
Differences in levels of sexual function among women allocated to the control and intervention groups in the analyzed periods ( T 0, T 1 and T 2). In addition, there were differences in the scores of the domains of sexual function (desire, excitement, orgasm, satisfaction and pain).
Some studies have already been developed using telehealth as a tool to apply educational interventions for breast cancer patients. However, different outcome variables have been studied, such as psychological suffering and coping, depression, fatigue, quality of life, and chemotherapy-related symptoms. A telephone counseling program assessed the improvement in psychosocial outcomes posttreatment of women with breast cancer to improve sexual health through variables such as anguish, depression, sexual dysfunction, and personal growth. Another study used a couple-based telephone intervention to address sexual issues. Furthermore, researchers have developed a multimodal nursing care program based on WeChat, consisting of chat software for the early rehabilitation of women after breast cancer surgery. However, none of these studies have specifically evaluated telenursing. The current study protocol describes the procedures for the clinical trial of telephone counseling. Information and communication technologies and follow-up and monitoring systems, with an emphasis on telenursing, improve the health and self-care outcomes of people with one or more Chronic Noncommunicable Diseases. In addition, telenursing is considered a resource to improve the quality of life of these patients and reduce hospitalizations and costs. This tool has been used to educate patients and health professionals. The clinical trial design has several significant strengths. The first is that the RCT is considered the gold standard of clinical research, allowing more reliable conclusions regarding the effectiveness of interventions. In addition, the study has 2 parallel arms for comparing the intervention to usual care and will include a follow-up assessment of the outcome 12 weeks after the intervention in addition to the measurement that will take place immediately after the end of the intervention (6 weeks). Finally, the quality of the research data will possibly be high due to the use of validated instruments (such as the FSFI) for measuring sexual function. Regarding the weaknesses of the study, we can mention the nonmeasurement of the prevalence rate of sexual problems in the population and the uncertainty regarding the inclusion and retention rate, which can impact the time needed to achieve the sample size required for adequate statistical significance. Another limitation is the use of a nonactive treatment in the control group, which can generate a risk of general and nonspecific effects.
This evidence-based nursing care strategy can be used to improve the sexual function of breast cancer patients and consequently their quality of life and marital relationship.
Data curation: Iarlla Silva Ferreira. Investigation: Romel Jonathan Velasco Yanez. Project administration: Iarlla Silva Ferreira. Resources: Ana Fátima Carvalho Fernandes. Software: Romel Jonathan Velasco Yanez. Writing – original draft: Iarlla Silva Ferreira and Ana Fátima Carvalho Fernandes. Writing – review & editing : Régia Christina Moura Barbosa Castro and Andrea Rodrigues Bezerra.
|
Pediatric genetics: rare is common | fe35a45a-3b9e-4565-813a-7cc85b8d5fdb | 7304359 | Pediatrics[mh] | C.G. contributed to this article in her personal capacity. The views expressed are her own and do not necessarily represent the views of the National Institutes of Health or the United States Government.
|
Smooth muscle contractility of laser-enucleated prostate tissues and impacts of preoperative α | 5867ba2a-21f3-499f-8a45-50ef74aafbcb | 11811036 | Surgical Procedures, Operative[mh] | Prostate smooth muscle contraction is considered essential in the pathophysiology and medical treatment of voiding symptoms suggestive of benign prostatic hyperplasia (BPH) , . Increased prostate smooth muscle tone may contribute to urethral obstruction, resulting in impaired bladder emptying, and finally in symptoms , . The α 1 -adrenoceptor antagonists (α 1 -blockers), applied for rapid symptom improvement, represent the first line option for medical treatment and are believed to act by inhibition of α 1 -adrenergic prostate smooth muscle contraction , . The phosphodiesterase-5 inhibitor tadalafil is available as an alternative and is believed to improve symptoms by smooth muscle relaxation as well . Treatment with 5α-reductase inhibitors (5ARI) is recommended for prevention of progression, complications and surgery in BPH . However, improvements by available drugs underlay obvious ceilings. The α 1 -blockers reduce symptom scores by maximally 30–50% and enhance the maximum urinary flow rate (Q max ) by not more than 40% , . These improvements are not far from placebos, reducing symptom scores by 10–34%, and increasing the Q max up to 28% , . Tadalafil decreases symptom scores to similar extent as α 1 -blockers, but does not enhance Q max in most studies . During prolonged application 5ARIs may improve symptoms as well, but benefits are hardly or not additive with α 1 -blockers. Low adherence, attributed to insufficient efficacy and unbalanced side effects accounts for the progression and complications, hospitalization, and finally contributes to high numbers of surgery due to BPH , . Surgery often becomes inevitable with progression of BPH and despite drug treatment, in patients with imminent complications but also as last resort treatment for adequate symptom improvement , . Transurethral resection of the prostate (TURP) was the standard surgery for BPH for decades, while holmium and thulium laser enucleation of the prostate (HoLEP, ThuLEP) are increasingly established alternatives , . Limited drug efficacy in voiding symptoms suggestive of BPH raised ongoing preclinical research addressing prostate smooth muscle contraction. Currently available drugs and drug candidates were developed based on experimental studies investigating their effects on contractions of prostate tissues in vitro, including human tissues. Tissue models included human tissues from radical prostatectomy (RP) for prostate cancer (PCa), or from TURP. However, TURP tissues have been supposed to be traumatized, by heat-induced denaturation during surgery, while RP tissues are not specifically representative for BPH, and do not cover medication-refractory voiding symptoms in BPH. To assess the potential of samples from laser enucleation for preclinical investigations, this study aims to examine the contractility of prostate tissues from HoLEP and ThuLEP. Study design, strategy and aims This study was carried out in accordance with the Declaration of Helsinki of the World Medical Association and has been approved by the ethics committee of the Ludwig-Maximilians University, Munich, Germany (approval number 22-0608, from 08-10-2022). Informed consent was obtained from all patients. Samples and data were collected and analyzed pseudonymized. This study included two parts. Firstly, tissues were collected from laser enucleation (HoLEP, ThuLEP) and TURP, and randomly assigned to examination of electric field stimulation- (EFS)- and noradrenaline-induced contractions. The primary aims of this part were to explore whether tissues from laser nucleation are still contractile in organ bath experiments, and a comparison to TURP and RP tissues. Experiments with TURP tissues were discontinued after experiments with tissues from 43 patients, as it became obvious that a high percentage did not react to KCl. Experiments with laser-enucleated tissues were continued until tissues from 49 patients were examined with EFS, 59 patients with noradrenaline, and 2 patients with KCl but without showing contractions, all obtained from a total of 85 laser-enucleated patients. Laser-enucleated and TURP samples without reaction to KCl were included in frequency and concentration response curves and for calculation of E max values, i. e. contractions in these samples were rated as 0 mN at each frequency and each noradrenaline concentration. KCl-induced contractions in samples from laser enucleation and TURP were compared to values from RP tissues in two of our previous study results , . Further, the E max values for EFS and noradrenaline in samples from laser enucleation were compared to E max values from our previous study results with RP tissues published 2018–2023 – . EFS- and noradrenaline-induced contractions in laser-enucleated tissues were analyzed for different subgroups, including separation for patients with and without catheterization for urinary retention, and preoperative treatment with α 1 -blockers, after completion of all experiments. Secondly, tissues from laser enucleation were collected from 21 further patients (not participating in the first part), and examined with EFS or noradrenaline, after the addition of α 1 -blockers or solvent (controls) in the organ bath. Thus, these experiments were planned and implemented after completion of the first part, as it turned out that laser-enucleated tissues are contractile in the organ bath. This second part was to assess, whether these tissues from laser enucleation are still suitable to examine effects of anticontractile drug candidates. Holmium and thulium laser enucleation of the prostate HoLEP enucleation was performed in a three-lobe technique, using the VersaPulse ® 100W Holmium Laser (Lumenis Ltd., Yokneam, Israel) with a frequency of 53 Hz and a power setting of 1.2 kJ, or the 150 W CyberHo Holmium laser paired with a 550 micron end-firing laser fiber, utilizing energy settings of 2.0 J per puls and a frequency of 50 Hz. For ThuLEP enucleation, a Dornier Thulio ® p-Tm:YAG with 570 micron end-firing disposable fiber was used with energy settings of 1.5 J per puls and a frequency of 50 Hz, resulting in 75 W. Tissue morcellation was performed using dual 5 mm reciprocating hollow metal blades. The same three-lobe technique was used for both the holmium and thulium laser enucleation . Enucleation of the median lobe started distally and progressed proximally towards the bladder neck. Once detached, the median lobe was pushed into the bladder, and prostatic attachments were released from the bladder neck, allowing the median lobe to fall into the bladder. Enucleation of the lateral lobes proceeded similarly, beginning with the right lateral lobe at the level of the verumontanum. At the 5 o’clock position of the prostate apex, the lateral lobe was detached from the surgical capsule, and pushed into the urinary bladder. Dissection of the left lateral lobe followed a similar protocol, starting at the 7 o’clock position. Identical to the other side, the side flap on the surgical capsule was released using the laser and pushed into the bladder. Tissue morcellation was initiated by the insertion of an offset nephroscope fitted with a soft-tissue morcellator into the bladder, under adequate bladder filling. Tissues were shredded on the circular knife, effectively cutting the adenoma into smaller pieces, which were subsequently suctioned out of the bladder and collected. Transurethral resection of the prostate TURP was performed by bipolar resection. Compared to monopolar systems this leads to a lower resection temperature, and consequently to lower thermal damage in the surrounding tissue. Following the insertion of the resectoscope through the urethra, a high-frequency electrical current was used to remove overgrown tissue of the prostate, layer by layer until reaching the surgical capsule. After excess tissue was removed using a glass syringe, blood vessels were sealed to stop bleeding, and the resectoscope was removed. Data from tissues from radical prostatectomy For comparison of tissues from TURP and laser enucleation with tissues from RP, E max values for EFS- and noradrenaline-induced contractions were compiled from control groups in previous studies with RP tissues from our lab (2018–2023) – . E max values were collected from each single experiment, with each value representing the mean of two samples from the same prostate, or the value of a single determination if only one sample was available in an independent experiment. Values were obtained from the control groups of these previous studies and were consequently obtained with solvents (mostly dimethylsulfoxid, DMSO) in varying amounts. Tissues were collected from periurethral zones, as previously described , . Conditions for interim storage and transport were similar to conditions for tissues from surgery for BPH in this study, with the exception that tissues from RP were macroscopically inspected and sampled by pathologists. Organ bath experiments Collected tissues varied qualitatively between surgeries (Fig. ). Macerations from laser-enucleation consisted of numerous small tissue shreds (dozens to hundreds per collected sample), with the largest measuring around 0.5–10 × 4 mm in size (Fig. ). Tissue shreds matching the required size for organ bath experiments (approximately 6 × 3 × 3 mm) were either used directly without further cutting or prepared by cutting largest available shreds. TURP chips were typically larger than shreds from laser-enucleation, with most chips reaching ≥ 1 cm in length, but often including flat or narrow protrusions (Fig. ). Strips for experiments were prepared from the largest available chips, specifically from interior regions of chunk-like parts and excluding margins (presumed to be most traumatized), extensions and flattened areas. Organ bath experiments were performed as recently described for RP tissues , . Tissue strips were mounted in organ baths, with four chambers per device (model 720 M, Danish Myotechnology, Aahus, Denmark) containing 10 ml Krebs–Henseleit solution (37 °C, pH 7.4) continuously gassed with carbogen (95% O 2 and 5% CO 2 ). After adjustment of a stable pretension of 4.9 mN within 45 min , , tissues were contracted by 80 mM KCl, by the addition of a 2 M KCl solution. As soon as a maximum plateau contraction was obtained, high molar KCl was washed out, resulting in a new baseline. Subsequently, tissues were used directly for frequency response curves by EFS or concentration response curves for noradrenaline. In the second study part, α 1 -blockers or solvent (for controls) were added after washout of KCl, and frequency or concentration response curves were constructed 30 min later. With each strip, only one frequency or concentration response curve was recorded. For construction of frequency and concentration response curves (without α 1 -blockers or solvent), strips were intuitively allocated to EFS and noradrenaline. Channels showing no reaction to KCl were not further examined by EFS or with noradrenaline, and included to analyses by rating as 0 mN at each frequency and noradrenaline concentration, except of a separate analysis as indicated. From a total of laser-enucleated tissues from 85 patients, 49 were examined by EFS and 59 with noradrenaline. In tissues from two patients, none of the strips contracted with KCl, which were included in data analyses by rating EFS- and noradrenaline-induced contractions as zero, resulting in tissues from 51 patients analyzed for EFS-induced, and from 61 patients analyzed for noradrenaline-induced contractions. Tissues from most patients were assessed by double or multiple determinations. In 46 out of 51 EFS experiments, 2–4 strips were examined per patient, and the remaining five as single determination with only one strip. In 53 out of 61 noradrenaline experiments, again multiple strips were examined per patient (2–4 strips for tissues from 49 patients, 5–8 strips for tissues from 4 patients). From a total of TURP tissues from 43 patients, 7 were examined by EFS and 26 with noradrenaline. In tissues from 16 patients, none of the examined strips contracted with KCl, which were rated as zero for analyses of EFS- and noradrenaline-induced contractions, resulting in TURP tissues from 23 patients analyzed for EFS, and from 42 patients analyzed for noradrenaline. In 21 out of 23 EFS experiments, multiple strips were examined per patient (2–4 strips for tissues from 18 patients, 6–8 strips for tissues from 3 patients). In 41 out of 42 noradrenaline experiments, multiple strips were examined per patient (2–4 strips for tissues from 28 patients, 5–8 strips for tissues from 13 patients). Ex vivo effects of α 1 -blockers were assessed in paired samples, i. e. α 1 -blockers or solvent were added to tissue strips from the same patient, being examined in the same experiment. Double determinations for the solvent and antagonist group were possible in 30 out of a total of 41 experiments. In the remaining experiments, the amount of available tissues did not allow the filling of two channels for both groups or single samples did not contract with KCl. However, these experiments included three samples per experiment, split to the control and antagonist group. Agonist- and EFS-induced contractions are expressed as percentage of 80 mM KCl-induced contractions, as this may correct individual variations and heterogeneities, and variables such as strip size or smooth muscle content. E max values, EC 50 values for agonists, and frequencies inducing 50% of the maximum EFS-induced contraction (EF 50 ) were calculated separately for each single experiment by curve fitting , using GraphPad Prism 6 (GraphPad Software Inc., San Diego, CA, USA). The software sends error messages, if curve fitting is not possible, or if results from curve fitting are suspected as “ambiguous”. In addition, values from curve fitting were checked manually for plausibility, as recommended in the “GraphPad Curve Fitting Guide” (GraphPad Software Inc.). Results in the first study part were marked as “ambiguous” in one EFS and one noradrenaline experiment. Following manual inspection, these values were considered plausible. As curve fitting was not possible with samples without contractions, E max for these tissues was set to 0 mN. Concentration response curves addressing ex vivo effects of α 1 -blockers included noradrenaline concentrations up to 3 mM, to allow detection of rightshifts and recovery at high agonist concentrations, and to allow curve fitting in antagonist groups. Consequently, control curves frequently included downhill parts at high agonist concentrations, which had to be excluded in 8 out of 21 experiments with noradrenaline, to allow plausible curve fitting. Values from 2 curves with tamsulosin, one with noradrenaline and one with EFS, were labelled as “ambiguous”, but provided plausible values after the exclusion of untypical values. Data for catheterization were available from 81 and for premedication from 82 out of 85 patients. Contraction data from patients with unavailable clinical data were excluded from group analyses. Statistical analyses Data in frequency and concentration response curves are means with standard deviation (SD). Single values in scatter plots are either means from all strips examined per tissue, or are values from each single strip. Group differences and effect sizes in the text are reported as mean differences (MD) with 95% confidence intervals (95% CI). Calculation of MDs and 95% CIs, and statistical analyses were performed using GraphPad Prism 6. Distribution of values within data sets including KCl-induced contractions, E max , EC 50 and EF 50 values were assessed by the D’Agostino & Pearson omnibus normality test (alpha = 0.05). Groups showing Gaussian distribution were compared using parametric tests, while non-parametric tests were applied to data sets containing at least one group not passing the normality test. Comparisons of KCl-induced contractions between three groups, and comparisons of previously reported E max values to E max values in the current study were performed by Dunn’s multiple comparison after one-way ANOVA with Kruskal Wallis test, allowing comparison of multiple groups without normal distribution. E max , EC 50 and EF 50 values in grouping analyses (i. e., between two groups) were compared by unpaired, two-tailed Mann Whitney test if data were not normally distributed in at least one of both groups, and by unpaired, two-tailed t test if data were normally distributed in both groups. E max , EC 50 and EF 50 values between paired groups (i. e., with/without ex vivo application of tamsulosin or silodosin) were compared by paired, two-tailed Wilcoxon matched-pairs signed rank test if data were not normally distributed in at least one of both groups, and by paired, two-tailed t test if data were normally distributed in both groups. Comparison of whole frequency and concentration response curves was performed by two-way analysis of variance (ANOVA), without multiple comparison and as data sets included three variables (concentration, contraction, treatment) . P values < 0.05 were considered significant. P values ≥ 0.05 are not indicated. The present study and analyses show an exploratory design, as typical features of a hypothesis-testing study are lacking, including a clear preset study plan, blinding, or biometric calculation of group sizes . Consequently, P values reported here are descriptive, but not hypothesis-testing . While the formation of group sizes was not driven by power calculations, 10 independent experiments per series were consistently found sufficient to detect biologically relevant differences or to detect drug effects in our previous organ bath experiments. Consequently, and as splitting into up to 4 subgroups was intended, while it was anticipated that clinical data would not be available from some patients, experiments with tissues from 50 to 60 patients were aimed with noradrenaline and EFS in part one, and from at least 10 patients per series in part two, without interim analyses. This study was carried out in accordance with the Declaration of Helsinki of the World Medical Association and has been approved by the ethics committee of the Ludwig-Maximilians University, Munich, Germany (approval number 22-0608, from 08-10-2022). Informed consent was obtained from all patients. Samples and data were collected and analyzed pseudonymized. This study included two parts. Firstly, tissues were collected from laser enucleation (HoLEP, ThuLEP) and TURP, and randomly assigned to examination of electric field stimulation- (EFS)- and noradrenaline-induced contractions. The primary aims of this part were to explore whether tissues from laser nucleation are still contractile in organ bath experiments, and a comparison to TURP and RP tissues. Experiments with TURP tissues were discontinued after experiments with tissues from 43 patients, as it became obvious that a high percentage did not react to KCl. Experiments with laser-enucleated tissues were continued until tissues from 49 patients were examined with EFS, 59 patients with noradrenaline, and 2 patients with KCl but without showing contractions, all obtained from a total of 85 laser-enucleated patients. Laser-enucleated and TURP samples without reaction to KCl were included in frequency and concentration response curves and for calculation of E max values, i. e. contractions in these samples were rated as 0 mN at each frequency and each noradrenaline concentration. KCl-induced contractions in samples from laser enucleation and TURP were compared to values from RP tissues in two of our previous study results , . Further, the E max values for EFS and noradrenaline in samples from laser enucleation were compared to E max values from our previous study results with RP tissues published 2018–2023 – . EFS- and noradrenaline-induced contractions in laser-enucleated tissues were analyzed for different subgroups, including separation for patients with and without catheterization for urinary retention, and preoperative treatment with α 1 -blockers, after completion of all experiments. Secondly, tissues from laser enucleation were collected from 21 further patients (not participating in the first part), and examined with EFS or noradrenaline, after the addition of α 1 -blockers or solvent (controls) in the organ bath. Thus, these experiments were planned and implemented after completion of the first part, as it turned out that laser-enucleated tissues are contractile in the organ bath. This second part was to assess, whether these tissues from laser enucleation are still suitable to examine effects of anticontractile drug candidates. HoLEP enucleation was performed in a three-lobe technique, using the VersaPulse ® 100W Holmium Laser (Lumenis Ltd., Yokneam, Israel) with a frequency of 53 Hz and a power setting of 1.2 kJ, or the 150 W CyberHo Holmium laser paired with a 550 micron end-firing laser fiber, utilizing energy settings of 2.0 J per puls and a frequency of 50 Hz. For ThuLEP enucleation, a Dornier Thulio ® p-Tm:YAG with 570 micron end-firing disposable fiber was used with energy settings of 1.5 J per puls and a frequency of 50 Hz, resulting in 75 W. Tissue morcellation was performed using dual 5 mm reciprocating hollow metal blades. The same three-lobe technique was used for both the holmium and thulium laser enucleation . Enucleation of the median lobe started distally and progressed proximally towards the bladder neck. Once detached, the median lobe was pushed into the bladder, and prostatic attachments were released from the bladder neck, allowing the median lobe to fall into the bladder. Enucleation of the lateral lobes proceeded similarly, beginning with the right lateral lobe at the level of the verumontanum. At the 5 o’clock position of the prostate apex, the lateral lobe was detached from the surgical capsule, and pushed into the urinary bladder. Dissection of the left lateral lobe followed a similar protocol, starting at the 7 o’clock position. Identical to the other side, the side flap on the surgical capsule was released using the laser and pushed into the bladder. Tissue morcellation was initiated by the insertion of an offset nephroscope fitted with a soft-tissue morcellator into the bladder, under adequate bladder filling. Tissues were shredded on the circular knife, effectively cutting the adenoma into smaller pieces, which were subsequently suctioned out of the bladder and collected. TURP was performed by bipolar resection. Compared to monopolar systems this leads to a lower resection temperature, and consequently to lower thermal damage in the surrounding tissue. Following the insertion of the resectoscope through the urethra, a high-frequency electrical current was used to remove overgrown tissue of the prostate, layer by layer until reaching the surgical capsule. After excess tissue was removed using a glass syringe, blood vessels were sealed to stop bleeding, and the resectoscope was removed. For comparison of tissues from TURP and laser enucleation with tissues from RP, E max values for EFS- and noradrenaline-induced contractions were compiled from control groups in previous studies with RP tissues from our lab (2018–2023) – . E max values were collected from each single experiment, with each value representing the mean of two samples from the same prostate, or the value of a single determination if only one sample was available in an independent experiment. Values were obtained from the control groups of these previous studies and were consequently obtained with solvents (mostly dimethylsulfoxid, DMSO) in varying amounts. Tissues were collected from periurethral zones, as previously described , . Conditions for interim storage and transport were similar to conditions for tissues from surgery for BPH in this study, with the exception that tissues from RP were macroscopically inspected and sampled by pathologists. Collected tissues varied qualitatively between surgeries (Fig. ). Macerations from laser-enucleation consisted of numerous small tissue shreds (dozens to hundreds per collected sample), with the largest measuring around 0.5–10 × 4 mm in size (Fig. ). Tissue shreds matching the required size for organ bath experiments (approximately 6 × 3 × 3 mm) were either used directly without further cutting or prepared by cutting largest available shreds. TURP chips were typically larger than shreds from laser-enucleation, with most chips reaching ≥ 1 cm in length, but often including flat or narrow protrusions (Fig. ). Strips for experiments were prepared from the largest available chips, specifically from interior regions of chunk-like parts and excluding margins (presumed to be most traumatized), extensions and flattened areas. Organ bath experiments were performed as recently described for RP tissues , . Tissue strips were mounted in organ baths, with four chambers per device (model 720 M, Danish Myotechnology, Aahus, Denmark) containing 10 ml Krebs–Henseleit solution (37 °C, pH 7.4) continuously gassed with carbogen (95% O 2 and 5% CO 2 ). After adjustment of a stable pretension of 4.9 mN within 45 min , , tissues were contracted by 80 mM KCl, by the addition of a 2 M KCl solution. As soon as a maximum plateau contraction was obtained, high molar KCl was washed out, resulting in a new baseline. Subsequently, tissues were used directly for frequency response curves by EFS or concentration response curves for noradrenaline. In the second study part, α 1 -blockers or solvent (for controls) were added after washout of KCl, and frequency or concentration response curves were constructed 30 min later. With each strip, only one frequency or concentration response curve was recorded. For construction of frequency and concentration response curves (without α 1 -blockers or solvent), strips were intuitively allocated to EFS and noradrenaline. Channels showing no reaction to KCl were not further examined by EFS or with noradrenaline, and included to analyses by rating as 0 mN at each frequency and noradrenaline concentration, except of a separate analysis as indicated. From a total of laser-enucleated tissues from 85 patients, 49 were examined by EFS and 59 with noradrenaline. In tissues from two patients, none of the strips contracted with KCl, which were included in data analyses by rating EFS- and noradrenaline-induced contractions as zero, resulting in tissues from 51 patients analyzed for EFS-induced, and from 61 patients analyzed for noradrenaline-induced contractions. Tissues from most patients were assessed by double or multiple determinations. In 46 out of 51 EFS experiments, 2–4 strips were examined per patient, and the remaining five as single determination with only one strip. In 53 out of 61 noradrenaline experiments, again multiple strips were examined per patient (2–4 strips for tissues from 49 patients, 5–8 strips for tissues from 4 patients). From a total of TURP tissues from 43 patients, 7 were examined by EFS and 26 with noradrenaline. In tissues from 16 patients, none of the examined strips contracted with KCl, which were rated as zero for analyses of EFS- and noradrenaline-induced contractions, resulting in TURP tissues from 23 patients analyzed for EFS, and from 42 patients analyzed for noradrenaline. In 21 out of 23 EFS experiments, multiple strips were examined per patient (2–4 strips for tissues from 18 patients, 6–8 strips for tissues from 3 patients). In 41 out of 42 noradrenaline experiments, multiple strips were examined per patient (2–4 strips for tissues from 28 patients, 5–8 strips for tissues from 13 patients). Ex vivo effects of α 1 -blockers were assessed in paired samples, i. e. α 1 -blockers or solvent were added to tissue strips from the same patient, being examined in the same experiment. Double determinations for the solvent and antagonist group were possible in 30 out of a total of 41 experiments. In the remaining experiments, the amount of available tissues did not allow the filling of two channels for both groups or single samples did not contract with KCl. However, these experiments included three samples per experiment, split to the control and antagonist group. Agonist- and EFS-induced contractions are expressed as percentage of 80 mM KCl-induced contractions, as this may correct individual variations and heterogeneities, and variables such as strip size or smooth muscle content. E max values, EC 50 values for agonists, and frequencies inducing 50% of the maximum EFS-induced contraction (EF 50 ) were calculated separately for each single experiment by curve fitting , using GraphPad Prism 6 (GraphPad Software Inc., San Diego, CA, USA). The software sends error messages, if curve fitting is not possible, or if results from curve fitting are suspected as “ambiguous”. In addition, values from curve fitting were checked manually for plausibility, as recommended in the “GraphPad Curve Fitting Guide” (GraphPad Software Inc.). Results in the first study part were marked as “ambiguous” in one EFS and one noradrenaline experiment. Following manual inspection, these values were considered plausible. As curve fitting was not possible with samples without contractions, E max for these tissues was set to 0 mN. Concentration response curves addressing ex vivo effects of α 1 -blockers included noradrenaline concentrations up to 3 mM, to allow detection of rightshifts and recovery at high agonist concentrations, and to allow curve fitting in antagonist groups. Consequently, control curves frequently included downhill parts at high agonist concentrations, which had to be excluded in 8 out of 21 experiments with noradrenaline, to allow plausible curve fitting. Values from 2 curves with tamsulosin, one with noradrenaline and one with EFS, were labelled as “ambiguous”, but provided plausible values after the exclusion of untypical values. Data for catheterization were available from 81 and for premedication from 82 out of 85 patients. Contraction data from patients with unavailable clinical data were excluded from group analyses. Data in frequency and concentration response curves are means with standard deviation (SD). Single values in scatter plots are either means from all strips examined per tissue, or are values from each single strip. Group differences and effect sizes in the text are reported as mean differences (MD) with 95% confidence intervals (95% CI). Calculation of MDs and 95% CIs, and statistical analyses were performed using GraphPad Prism 6. Distribution of values within data sets including KCl-induced contractions, E max , EC 50 and EF 50 values were assessed by the D’Agostino & Pearson omnibus normality test (alpha = 0.05). Groups showing Gaussian distribution were compared using parametric tests, while non-parametric tests were applied to data sets containing at least one group not passing the normality test. Comparisons of KCl-induced contractions between three groups, and comparisons of previously reported E max values to E max values in the current study were performed by Dunn’s multiple comparison after one-way ANOVA with Kruskal Wallis test, allowing comparison of multiple groups without normal distribution. E max , EC 50 and EF 50 values in grouping analyses (i. e., between two groups) were compared by unpaired, two-tailed Mann Whitney test if data were not normally distributed in at least one of both groups, and by unpaired, two-tailed t test if data were normally distributed in both groups. E max , EC 50 and EF 50 values between paired groups (i. e., with/without ex vivo application of tamsulosin or silodosin) were compared by paired, two-tailed Wilcoxon matched-pairs signed rank test if data were not normally distributed in at least one of both groups, and by paired, two-tailed t test if data were normally distributed in both groups. Comparison of whole frequency and concentration response curves was performed by two-way analysis of variance (ANOVA), without multiple comparison and as data sets included three variables (concentration, contraction, treatment) . P values < 0.05 were considered significant. P values ≥ 0.05 are not indicated. The present study and analyses show an exploratory design, as typical features of a hypothesis-testing study are lacking, including a clear preset study plan, blinding, or biometric calculation of group sizes . Consequently, P values reported here are descriptive, but not hypothesis-testing . While the formation of group sizes was not driven by power calculations, 10 independent experiments per series were consistently found sufficient to detect biologically relevant differences or to detect drug effects in our previous organ bath experiments. Consequently, and as splitting into up to 4 subgroups was intended, while it was anticipated that clinical data would not be available from some patients, experiments with tissues from 50 to 60 patients were aimed with noradrenaline and EFS in part one, and from at least 10 patients per series in part two, without interim analyses. Potassium-induced contractions Potassium-induced contractions of laser-enucleated tissues from HoLEP and ThuLEP (n = 85 patients) were higher compared to tissues from TURP (n = 43 patients) (HoLEP/ThuLEP 2.49 mN [2.07–2.92]; TURP 0.72 mN [0.38–1.06]; MD 1.77 mN [0.76–2.78]) (Fig. a). Potassium-induced contractions of RP tissues (RP 3.32 mN [3–3.64]; n = 189 patients) were higher compared to HoLEP/ThuLEP tissues (MD 0.83 mN [0.13–1.54]), and compared to TURP tissues (MD 2.6 mN [1.69–3.52]) (Fig. a). Included to these analyses are tissues and samples showing no reaction to KCl. In 2 out of 85 tissues from laser enucleation (2.4%), none of the examined single samples responded to KCl, while 16 out of 43 tissues from TURP (37.2%) were completely unresponsive to KCl with each examined sample. These differences persisted, if contractions of all single samples were analyzed instead of means of each prostate. Again, the difference between laser-enucleated and RP tissues was smaller (HoLEP/ThuLEP 2.57 mN [2.27–2.87]; RP 3.33 mN [3.1–3.55]; MD 0.76 mN [0.29–1.26]), than the difference between TURP tissues and both other groups (TURP 0.58 mN [0.38–0.77]; MD 2.75 mN [2.12–3.38] vs. RP; MD 1.99 mN [1.37–2.61] vs. HoLEP/ThuLEP) (Fig. b). If samples from HoLEP/ThuLEP tissues showing no reaction to KCl (i. e., 0 mN) were excluded (what was done with RP tissues) (Fig. c), contractions were again similar between tissues from HoLEP/ThuLEP and RP (HoLEP/ThuLEP 2.99 mN [2.66–3.32]; RP 3.33 mN [3.1–3.55]; MD 0.33 mN [-0.18 to 0.846]), but still substantially lower in TURP tissues. The percentage of single samples without reaction to highmolar KCl amounted to 60% in TURP tissues and 14% in tissues from laser enucleation. EFS-induced contractions EFS induced frequency-dependent contractions in laser-enucleated tissues (n = 51 patients) (Fig. a). E max values were calculated by curve fitting and compared to E max values from control groups (i. e., preincubated with solvents, but without drugs) in our previous studies with RP tissues (2018–2023). E max values from HoLEP/ThuLEP tissues were not statistically different to E max values for EFS-induced contractions of RP tissues in 6 from 21 of these previous studies (Fig. a). In 15 from 21 of studies with RP tissues, E max values for EFS-induced contractions were higher compared to E max values in tissues from HoLEP/ThuLEP (Fig. a). With an average E max of 47% [35–60] of KCl-induced contractions, contractions of HoLEP/ThuLEP tissues ranged lower than E max values in all studies with RP tissues, but with high overlap and with the closest E max values in two studies with RP tissues ranging at 68% [49–88] and 69% [55–84] of KCl-induced contractions (Fig. a). The variability and range of maximum EFS-induced contractions in RP tissues was high across studies, with the highest E max values mounting to 235% [150–320] and to 166% [107–224] of KCl-induced contractions (Fig. a). With an E max of 27% [7–48], EFS-induced contractions of TURP tissues were obviously lower compared to laser-enucleated and to RP tissues (Fig. a). Noradrenaline-induced contractions Noradrenaline induced concentration-dependent contractions in laser-enucleated tissues (n = 60 patients) (Fig. b). E max values were calculated by curve fitting and compared to E max values from control groups (i. e., preincubated with solvents, but without drugs) in our previous studies with RP tissues (2018–2023). E max values in tissues from HoLEP/ThuLEP were similar (i. e., not statistically different) to E max values for noradrenaline-induced contractions of RP tissues in 7 from 20 of these previous studies (Fig. b). In the 13 other from 20 of studies with RP tissues, E max values for noradrenaline-induced contractions were higher compared to E max values observed with tissues from HoLEP/ThuLEP (Fig. b). With an average E max of 99.7% [85–114] of KCl-induced contractions, contractions of HoLEP/ThuLEP tissues ranged lower than E max in 19 of the 20 studies with RP tissues, but with high overlap and with the closest E max values ranging at 92% [71–112] and 118% [103–133] of KCl-induced contractions in two studies with RP tissues (Fig. b). The variability and range of maximum noradrenaline-induced contractions in RP tissues was high across studies, with the highest E max values mounting to 260% [112–407] and to 217% [150–284] of KCl-induced contractions (Fig. b). With an E max of 56% [34–79], noradrenaline-induced contractions of TURP tissues were obviously lower compared to laser-enucleated and to RP tissues (Fig. b). Grouping of EFS-induced contractions of HoLEP/ThuLEP tissues EFS-induced contractions were similar in tissues from patients without (n = 25) and with catheter (n = 22) (Fig. a). In patients receiving preoperative treatment with α 1 -blockers (n = 21), contractions were by trend (though, not significantly) lower compared to patients without treatment (n = 26) (E max 68% [41–94] of KCl without α 1 -blocker, 41% [27–55] with α 1 -blocker, MD 27% [0–54]) (Fig. b). In patients without catheter, contractions were similar in tissues from patients with (n = 17) or without (n = 8) treatment with α 1 -blockers (Fig. c). However, in patients with catheter, contractions were substantially lower in patients with α 1 -blocker treatment (n = 8), compared to patients without α 1 -blocker treatment (n = 13) (Fig. d). Lower contractions in α 1 -blocker-treated patients were reflected by decreased E max values (E max 64% [36–92] of KCl without α 1 -blocker, 18% [11–25] with α 1 -blocker, MD 46% [11–81]) (Fig. d). Without α 1 -blocker treatment, contractions were similar in tissues from patients without (n = 8) and with (n = 13) catheter (Fig. e). In patients with α 1 -blocker treatment, contractions were lower in patients with catheter (n = 8), compared to tissues from patients without catheter (n = 17) (E max 47% [22–72] of KCl without catheter, 18% [11–25] with catheter, MD 29% [-7 to 65]) (Fig. f). Grouping of noradrenaline-induced contractions in HoLEP/ThuLEP tissues Noradrenaline-induced contractions were similar in tissues from patients without (n = 31) and with catheter (n = 26) (Fig. a). In patients receiving preoperative treatment with α 1 -blockers (n = 34), contractions were by trend (though, not significantly) lower compared to patients without treatment (n = 24) (Fig. b). While E max values for noradrenaline were similar between both groups, EC 50 values for noradrenaline were increased in tissues from patients with α 1 -blocker pretreatment, compared to tissues from patients without pretreatment (pEC 50 6.01 [0.08–0.41] without α 1 -blocker, 5.44 [0.07–0.34] with α 1 -blocker, MD 0.57 [0.11–0.54]) (Fig. b). In patients without catheter, contractions were similar in tissues from patients with (n = 22) or without (n = 9) treatment with α 1 -blockers (Fig. c). While E max values were similar between both groups, EC 50 values for noradrenaline were increased by trend in tissues with α 1 -blocker treatment, compared to tissues from patients without treatment (pEC 50 5.79 [5.11–6.46] without α 1 -blocker, 5.48 [5.21–5.745] with α 1 -blocker, MD 0.31 [-0.4 to 1.02]) (Fig. c). In patients with catheter, contractions were substantially lower in patients with α 1 -blocker treatment (n = 11), compared to patients without α 1 -blocker treatment (n = 15) (Fig. d). Lower contractions in α 1 -blocker-treated patients were reflected by decreased (though, not significantly) E max values (122% [94–151] of KCl without α 1 -blocker, 82% [48–116] with α 1 -blocker, MD 40% [2–82] of KCl), and by increased EC 50 values for noradrenaline (pEC 50 6 [5.79–6.2] without α 1 -blocker, 5.45 [5.23–5.68] with α 1 -blocker, MD 0.54 [0.28–0.81]) (Fig. d). Without α 1 -blocker treatment, contractions were higher in tissues from patients with catheter (n = 15) compared to tissues from patients without catheter (n = 9) (Fig. e). Changes were reflected by E max values, which were increased by trend (80% [43–116] of KCl without catheter, 122% [94–151] of KCl with catheter, MD 43% [-1 to 86] of KCl), while EC 50 values for noradrenaline were similar between both groups (Fig. e). In patients with α 1 -blocker treatment, contractions were similar in tissues from patients without (n = 22) and with (n = 11) catheter (Fig. f). Ex vivo effects of α 1 -blockers on contractions in laser-enucleated tissues Effects of tamsulosin and silodosin, applied ex vivo in the organ bath, were examined in separate series with laser-enucleated tissues from 21 further patients. Both antagonists caused right shifts of concentration response curves for noradrenaline, increased EC 50 values but unchanged E max values for noradrenaline, and decreased E max values for EFS (Fig. ), without that further grouping was required. Tamsulosin increased the EC 50 values for noradrenaline, from -6.09 [-6.32 to -5.86] in controls to -4.23 [-5.1 to -3.40] with tamsulosin (MD 1.87 [1.1–2.63] without reducing E max values, and decreased the E max for EFS-induced contractions from 91% [61–121] of KCl-induced contractions in controls, to 39% [19–59] with tamsulosin (MD -52% [-90 to -15]) without reducing EF 50 values for EFS (Fig. a). Silodosin increased the EC 50 values for noradrenaline, from -6.38 [-7.02 to -5.75] in controls to -3.89 [-4.64 to -3.13] with tamsulosin (MD 2.499 [1.87–3.13] without reducing E max values, and decreased the E max for EFS-induced contractions from 69% [43–95] of KCl-induced contractions in controls, to 28% with tamsulosin (MD -40% [-71 to -9]) without reducing EF 50 values for EFS (Fig. b). Potassium-induced contractions of laser-enucleated tissues from HoLEP and ThuLEP (n = 85 patients) were higher compared to tissues from TURP (n = 43 patients) (HoLEP/ThuLEP 2.49 mN [2.07–2.92]; TURP 0.72 mN [0.38–1.06]; MD 1.77 mN [0.76–2.78]) (Fig. a). Potassium-induced contractions of RP tissues (RP 3.32 mN [3–3.64]; n = 189 patients) were higher compared to HoLEP/ThuLEP tissues (MD 0.83 mN [0.13–1.54]), and compared to TURP tissues (MD 2.6 mN [1.69–3.52]) (Fig. a). Included to these analyses are tissues and samples showing no reaction to KCl. In 2 out of 85 tissues from laser enucleation (2.4%), none of the examined single samples responded to KCl, while 16 out of 43 tissues from TURP (37.2%) were completely unresponsive to KCl with each examined sample. These differences persisted, if contractions of all single samples were analyzed instead of means of each prostate. Again, the difference between laser-enucleated and RP tissues was smaller (HoLEP/ThuLEP 2.57 mN [2.27–2.87]; RP 3.33 mN [3.1–3.55]; MD 0.76 mN [0.29–1.26]), than the difference between TURP tissues and both other groups (TURP 0.58 mN [0.38–0.77]; MD 2.75 mN [2.12–3.38] vs. RP; MD 1.99 mN [1.37–2.61] vs. HoLEP/ThuLEP) (Fig. b). If samples from HoLEP/ThuLEP tissues showing no reaction to KCl (i. e., 0 mN) were excluded (what was done with RP tissues) (Fig. c), contractions were again similar between tissues from HoLEP/ThuLEP and RP (HoLEP/ThuLEP 2.99 mN [2.66–3.32]; RP 3.33 mN [3.1–3.55]; MD 0.33 mN [-0.18 to 0.846]), but still substantially lower in TURP tissues. The percentage of single samples without reaction to highmolar KCl amounted to 60% in TURP tissues and 14% in tissues from laser enucleation. EFS induced frequency-dependent contractions in laser-enucleated tissues (n = 51 patients) (Fig. a). E max values were calculated by curve fitting and compared to E max values from control groups (i. e., preincubated with solvents, but without drugs) in our previous studies with RP tissues (2018–2023). E max values from HoLEP/ThuLEP tissues were not statistically different to E max values for EFS-induced contractions of RP tissues in 6 from 21 of these previous studies (Fig. a). In 15 from 21 of studies with RP tissues, E max values for EFS-induced contractions were higher compared to E max values in tissues from HoLEP/ThuLEP (Fig. a). With an average E max of 47% [35–60] of KCl-induced contractions, contractions of HoLEP/ThuLEP tissues ranged lower than E max values in all studies with RP tissues, but with high overlap and with the closest E max values in two studies with RP tissues ranging at 68% [49–88] and 69% [55–84] of KCl-induced contractions (Fig. a). The variability and range of maximum EFS-induced contractions in RP tissues was high across studies, with the highest E max values mounting to 235% [150–320] and to 166% [107–224] of KCl-induced contractions (Fig. a). With an E max of 27% [7–48], EFS-induced contractions of TURP tissues were obviously lower compared to laser-enucleated and to RP tissues (Fig. a). Noradrenaline induced concentration-dependent contractions in laser-enucleated tissues (n = 60 patients) (Fig. b). E max values were calculated by curve fitting and compared to E max values from control groups (i. e., preincubated with solvents, but without drugs) in our previous studies with RP tissues (2018–2023). E max values in tissues from HoLEP/ThuLEP were similar (i. e., not statistically different) to E max values for noradrenaline-induced contractions of RP tissues in 7 from 20 of these previous studies (Fig. b). In the 13 other from 20 of studies with RP tissues, E max values for noradrenaline-induced contractions were higher compared to E max values observed with tissues from HoLEP/ThuLEP (Fig. b). With an average E max of 99.7% [85–114] of KCl-induced contractions, contractions of HoLEP/ThuLEP tissues ranged lower than E max in 19 of the 20 studies with RP tissues, but with high overlap and with the closest E max values ranging at 92% [71–112] and 118% [103–133] of KCl-induced contractions in two studies with RP tissues (Fig. b). The variability and range of maximum noradrenaline-induced contractions in RP tissues was high across studies, with the highest E max values mounting to 260% [112–407] and to 217% [150–284] of KCl-induced contractions (Fig. b). With an E max of 56% [34–79], noradrenaline-induced contractions of TURP tissues were obviously lower compared to laser-enucleated and to RP tissues (Fig. b). EFS-induced contractions were similar in tissues from patients without (n = 25) and with catheter (n = 22) (Fig. a). In patients receiving preoperative treatment with α 1 -blockers (n = 21), contractions were by trend (though, not significantly) lower compared to patients without treatment (n = 26) (E max 68% [41–94] of KCl without α 1 -blocker, 41% [27–55] with α 1 -blocker, MD 27% [0–54]) (Fig. b). In patients without catheter, contractions were similar in tissues from patients with (n = 17) or without (n = 8) treatment with α 1 -blockers (Fig. c). However, in patients with catheter, contractions were substantially lower in patients with α 1 -blocker treatment (n = 8), compared to patients without α 1 -blocker treatment (n = 13) (Fig. d). Lower contractions in α 1 -blocker-treated patients were reflected by decreased E max values (E max 64% [36–92] of KCl without α 1 -blocker, 18% [11–25] with α 1 -blocker, MD 46% [11–81]) (Fig. d). Without α 1 -blocker treatment, contractions were similar in tissues from patients without (n = 8) and with (n = 13) catheter (Fig. e). In patients with α 1 -blocker treatment, contractions were lower in patients with catheter (n = 8), compared to tissues from patients without catheter (n = 17) (E max 47% [22–72] of KCl without catheter, 18% [11–25] with catheter, MD 29% [-7 to 65]) (Fig. f). Noradrenaline-induced contractions were similar in tissues from patients without (n = 31) and with catheter (n = 26) (Fig. a). In patients receiving preoperative treatment with α 1 -blockers (n = 34), contractions were by trend (though, not significantly) lower compared to patients without treatment (n = 24) (Fig. b). While E max values for noradrenaline were similar between both groups, EC 50 values for noradrenaline were increased in tissues from patients with α 1 -blocker pretreatment, compared to tissues from patients without pretreatment (pEC 50 6.01 [0.08–0.41] without α 1 -blocker, 5.44 [0.07–0.34] with α 1 -blocker, MD 0.57 [0.11–0.54]) (Fig. b). In patients without catheter, contractions were similar in tissues from patients with (n = 22) or without (n = 9) treatment with α 1 -blockers (Fig. c). While E max values were similar between both groups, EC 50 values for noradrenaline were increased by trend in tissues with α 1 -blocker treatment, compared to tissues from patients without treatment (pEC 50 5.79 [5.11–6.46] without α 1 -blocker, 5.48 [5.21–5.745] with α 1 -blocker, MD 0.31 [-0.4 to 1.02]) (Fig. c). In patients with catheter, contractions were substantially lower in patients with α 1 -blocker treatment (n = 11), compared to patients without α 1 -blocker treatment (n = 15) (Fig. d). Lower contractions in α 1 -blocker-treated patients were reflected by decreased (though, not significantly) E max values (122% [94–151] of KCl without α 1 -blocker, 82% [48–116] with α 1 -blocker, MD 40% [2–82] of KCl), and by increased EC 50 values for noradrenaline (pEC 50 6 [5.79–6.2] without α 1 -blocker, 5.45 [5.23–5.68] with α 1 -blocker, MD 0.54 [0.28–0.81]) (Fig. d). Without α 1 -blocker treatment, contractions were higher in tissues from patients with catheter (n = 15) compared to tissues from patients without catheter (n = 9) (Fig. e). Changes were reflected by E max values, which were increased by trend (80% [43–116] of KCl without catheter, 122% [94–151] of KCl with catheter, MD 43% [-1 to 86] of KCl), while EC 50 values for noradrenaline were similar between both groups (Fig. e). In patients with α 1 -blocker treatment, contractions were similar in tissues from patients without (n = 22) and with (n = 11) catheter (Fig. f). 1 -blockers on contractions in laser-enucleated tissues Effects of tamsulosin and silodosin, applied ex vivo in the organ bath, were examined in separate series with laser-enucleated tissues from 21 further patients. Both antagonists caused right shifts of concentration response curves for noradrenaline, increased EC 50 values but unchanged E max values for noradrenaline, and decreased E max values for EFS (Fig. ), without that further grouping was required. Tamsulosin increased the EC 50 values for noradrenaline, from -6.09 [-6.32 to -5.86] in controls to -4.23 [-5.1 to -3.40] with tamsulosin (MD 1.87 [1.1–2.63] without reducing E max values, and decreased the E max for EFS-induced contractions from 91% [61–121] of KCl-induced contractions in controls, to 39% [19–59] with tamsulosin (MD -52% [-90 to -15]) without reducing EF 50 values for EFS (Fig. a). Silodosin increased the EC 50 values for noradrenaline, from -6.38 [-7.02 to -5.75] in controls to -3.89 [-4.64 to -3.13] with tamsulosin (MD 2.499 [1.87–3.13] without reducing E max values, and decreased the E max for EFS-induced contractions from 69% [43–95] of KCl-induced contractions in controls, to 28% with tamsulosin (MD -40% [-71 to -9]) without reducing EF 50 values for EFS (Fig. b). To the best of our knowledge, this study is the first to address the smooth muscle contractility of tissues obtained by laser enucleation. Preclinical investigation of prostate smooth muscle contraction was crucial in the development of currently available drugs for treatment of voiding symptoms, including α 1 -blockers and the phosphodiesterase-5 inhibitor tadalafil. Meanwhile, limitations of these medications became evident , initiating ongoing searches for novel targets and new candidate compounds, together with attempts to understand the reasons accounting for these limits. Consequently, contraction studies are of continuous interest. The findings of this study identify tissues from laser enucleation as a new model for investigation of prostate smooth muscle contraction, in patients with medication-refractory voiding symptoms in BPH. Typically, patients selected for surgery for BPH are characterized by severe symptoms, while treatment with α 1 -blockers is recommended for patients with moderate to severe symptoms . Ablative surgery is performed if complications of BPH are experienced, but also for adequate relief from lower urinary tract symptoms (LUTS) or postvoid urine in non-responders for medical treatment, or if medical treatment is refused but active treatment requested , . Thus, effective medical treatment is not available for these populations. Previous studies addressing prostate smooth contractions either used tissues from TURP or RP for PCa. TURP tissues cover the same or a similar patient group as laser-enucleated tissues. However, their use was often supposed to be limited by heat-induced traumatization, reducing their contractility in vitro. In our hands, TURP tissues were characterized by lower contractions compared to laser-enucleated tissues, and by an exceeding rate of complete non-responsiveness to contractile stimulation. While previous studies obtained results from TURP tissues, also with unquestioned relevance, information about exclusion and percentage of non-contractile tissues was rarely reported. In turn, tissues from RP for PCa, but without previous surgery for BPH were widely used in our previous studies. According to the prevalence of BPH in this age group , BPH and mild symptoms are likely in these patients, but these tissues from RP do not specifically cover the context of severe or medication-refractory voiding symptoms. As contractions of laser-enucleated tissues approached ranges of contractions seen with RP tissues, without high rates of non-responsiveness due to traumatization by surgery, these tissues may provide a suitable model to study prostate smooth muscle contraction in patients with severe and medication-refractory LUTS. Finally, it allows to correlate in vitro findings with clinical data in the future, while BPH-specific data including international prostate symptom score (IPSS), Q max or postvoid residual urine volume are not routinely assessed in patients undergoing RP for PCa. The reasons accounting for the divergent impacts of preoperative α 1 -blocker treatment, seen between patients with and without catheterization for urinary retention are still elusive at this stage. Pretreatment with α 1 -blockers resulted in reduced contractions by EFS and noradrenaline in tissues from catheterized patients, but not in tissues from patients without catheterization. Certainly, this difference reflects substantial, yet unknown heterogeneity between both populations, even though both groups receive surgery for BPH. Reasons may include 1) different tissue responsiveness to α 1 -blockers, 2) different tissue conditions affecting the drug availability in tissues ex vivo, or 3) differences affecting drug metabolism and bioavailability in vivo, and 4) further reasons and combinations. A different tissue responsiveness to α 1 -blockers could be imparted by divergent receptor expression, or by unknown differences in receptor regulation . Considering that ex vivo application of α 1 -blockers caused full effects, without grouping of patients, such differences in tissue responsiveness may be regarded as unlikely to impart the difference seen between catheterized and uncatheterized patients. Nevertheless, this cannot fully be excluded, as the experimental design differed for preoperative and ex vivo application of α 1 -blockers, including comparison of different patients in one series, but a comparison of paired samples from the same patients in the other series. Divergent tissue conditions may provide plausible explanations for the different impact of preoperative α 1 -blocker treatment in both groups. BPH may include stromal, glandular and mixed hyperplasia, but their specific contributions to symptoms and drug responsiveness are not understood and have been poorly documented in preclinical and clinical studies . Similar, prostatic fibrosis is a just recently emerging topic, and may include progressive deposition of extracellular matrix (ECM) compounds – . Drug penetration into tissues may reduce with increasing ECM content, depending on but also independently from vascularization in these tissue parts. This may reduce drug availability in vivo and thus, affect ex vivo contractility. Other factors related to tissue conditions may account as well. Different stromal-epithelial contents may determine the washout of α 1 -blockers during ex vivo procedures. Washout may be less effective in stromal compartments, and thus in tissues with high stromal content, compared to spongy tissues with high glandular content. With all due caution and if proving true, this may point to impaired washout of α 1 -blockers during ex vivo handling and thus, to predominant stromal hyperplasia in patients with catheterization (i.e. with urinary retention), and to predominant glandular hyperplasia in patients without. High individual variation in tissue consistency becomes obvious during routine work with TURP and laser ablation, ranging from soft to stiff tissues. In the run-up of the study, it was speculated that α 1 -blockers in the tissues, resulting from medication for voiding symptoms, will be washed out during surgery, intravesical maceration, transport and storage in custodiol solution, and finally by Krebs-Hensleit solution in the organ bath. Together, different tissue composition by cellular and non-cellular constituents, but also the phenotype of BPH may affect drug penetration and their removal in tissues, and may decide about clinical characteristics including urinary retention. Notably, the observable ex vivo effects from preoperative α 1 -blocker treatment seen in tissues from catheterized patients prove that the treatment affects prostate smooth muscle contractility, but without improving symptoms or complications, so these patients are evidentially medication-refractory. Results from application of α 1 -blockers ex vivo demonstrated that tissues from laser-enucleation are suitable for investigation of drug effects in the context of BPH. Application of α 1 -blockers in the organ bath still resulted in full potential effects, including increases in EC 50 values for α 1 -adrenergic agonists and recovery at high agonist concentrations. Considering that these patients needed surgery, despite treatment with α 1 -blockers and despite the effects of α 1 -blockers ex vivo, this raises the question, of whether inhibition of α 1 -adrenergic contractions, or smooth muscle contractions at all is a suitable strategy in these patients. For α 1 -adrenergic contractions, this is not the case. In addition to α 1 -adrenoceptors, prostate smooth muscle contraction can be induced by endothelins and thromboxane, to the maximum possible force. These non-adrenergic contractions were proposed to maintain full smooth muscle tone and thus, symptoms in medication-refractory LUTS , . Consequently, it appears possible that full inhibition of adrenergic contractions is insufficient for clinical effects, even if smooth muscle tone contributes to symptoms. On the other hand, it appears possible as well, that symptom severity is unrelated to smooth muscle contraction, even though this concept has been claimed for decades, but in view that a causative role of bladder outlet obstruction and symptoms has been challenged . While technical traumatization had lower effects on contractility in laser-enucleated tissues than in TURP tissues, influences of chopping, maceration and handling can not be fully excluded. TURP has been the gold standard in surgery for BPH for decades , . Laser-enucleation still has a niche existence in comprehensive healthcare , , but may gain in popularity, reaching a share of 20% in surgeries for BPH performed in France in 2018, and may increasingly replace TURP or emerged as the preferred option in centers with corresponding expertise – . HoLEP and ThuLEP equally relieve voiding symptoms, with high efficacy and safety . Heat generation in HoLEP and ThuLEP occurs by absorption of laser radiation by the prostate tissue, leading to heating and vaporization of water within the tissue. Comprehensive data allowing direct comparisons to TURP are limited for laser-enucleation, but heat development and propagation within the prostate and in the suprapubic region appear limited with HoLEP and ThuLEP – . Ex vivo, increases in temperatures across the instrument shaft and within the enucleation cavity remained below 5 K at any examined irrigation (apart from 0 ml/min), which is insufficient to cause tissue damage . Injurious temperatures are probably not attained during clinical application , . The depth of necrotic zones and tissue coagulation typically varies between studies and conditions for TURP , and probably for laser enucleation as well, where it may range from 0.1 to 4 mm , . Together, higher contractions of laser-enucleated tissues, compared to tissues from TURP may reflect a lower degree of heat-induced traumatization during surgery. While the observed differences between subgroups are of potential clinical relevance, the study is associated with limitations. The differences between patients with and without catheterization were surprising and obvious, but their investigation was not the primary aim. Thus, the initial, primary endpoint was the contractility in laser-enucleated tissues, in comparison to widely-used human tissue models. Unlike previous studies in which contractions were measured with solvent (as controls for test compounds), contractions were measured without further intervention in the current study. As another limitation, the study design does not take into account that male LUTS include storage symptoms caused by the urinary bladder, in addition to voiding symptoms attributed to BPH, or that conditions similar to voiding symptoms can be caused by an underactive bladder, instead of obstruction. A non-negligible number of patients with voiding symptoms (approximately 50%) also show detrusor overactivity and associated storage symptoms including urgency or frequency , . Clinical conditions and etiology of storage symptoms are highly variable, and may include or affect cholinergic voiding contractions of the detrusor, and non-cholinergic detrusor microcontractions initiating the micturition reflex . Storage symptoms in male mixed LUTS may develop secondary to obstruction or independently from it, and persist after desobstructive surgery. In no case, however, detrusor contractions are induced by α 1 -adrenoceptors, explaining why storage symptoms are resistant to α 1 -blockers . Consequently, symptom resistance to BPH-specific drugs may have been attributed to a certain, though unknown part of the study population, in particular in participants with high initial storage symptom scores, or with an underactive bladder. In fact, definite separation of voiding symptoms caused by obstruction from symptoms resulting from underactive or overactive bladder requires diagnosis by invasive urodynamics, or specific assessment of storage symptom scores , . In real world settings, decisions for desobstructive surgery are commonly based on routine care for male LUTS, not including diagnosis by invasive urodynamics . It has been estimated, that 18–28% of patients undergoing prostate surgery for LUTS have no obstruction . Symptoms in these patients may be predominantly attributed to bladder malfunction, and the surgery may be potentially unnecessary , . Thus, studies using tissues from desobstructive surgery addressing truly medication-refractory voiding symptoms should integrate data from diagnosis for storage symptoms. Smooth muscle contractions are intact in laser-enucleated prostate tissues, allowing investigation of prostate smooth muscle contraction in the context of medication-refractory voiding symptoms. Other than in TURP tissues, intrasurgical tissue traumatization does not limit investigation of contractility. Impacts of preoperative α 1 -blocker treatment divergently affected ex vivo contractility in tissues from patients with and without catheterization, reflecting fundamental differences in tissue conditions between patients with and without urinary retention. Heterogeneity in BPH differentially affects the risk for urinary retention in both subgroups, by unknown factors and despite shared surgery by laser-enucleation. Supplementary Information. |
Increasing bowel cancer screening using SMS in general practice: the SMARTscreen cluster randomised trial | 2b62fab2-9040-42c8-932e-ac0a93add355 | 10962502 | Family Medicine[mh] | Colorectal cancer (CRC) in Australia contributes to a notable burden of disease that remains consistently high, with approximately 15 000 new diagnoses of CRC and 5000 CRC-related deaths per year (11% of all cancer deaths). Only two-thirds of people diagnosed with CRC are expected to live beyond 5 years after diagnosis, and treatment costs the Australian Government A$1.1 billion (£571 million) per year. This high burden of disease is despite CRC being considered a cancer that can be both prevented (through the detection and removal of pre-cancerous polyps) and detected early on (stage-one CRC has a 99% relative survival rate). The Australian National Bowel Cancer Screening Program (NBCSP) is an Australian Government initiative aimed at reducing morbidity and mortality from CRC by offering a free, non-invasive screening test to detect early signs of CRC. Established in 2006, and expanded in subsequent years, it involves all Australians aged 50–74 years being mailed a self-collection faecal immunochemical test (FIT) kit every 2 years. Although this is a simple and cost-effective screening method for CRC, completion and returning of kits remains low at 40.9%. Participation is lowest in the 50–60-year-old age groups, with only 33.4% returning their kits. It has been estimated that increasing uptake of the NBCSP from 40% to 60% would prevent an additional 37 300 diagnoses of CRC and 24 800 deaths from CRC, as well as reducing the annual health budget expenditure by A$2.1 billion by 2040. Effective interventions to increase screening uptake have included endorsement of CRC screening by a GP, – TV public-service announcement campaigns featuring relatable people discussing their personal screening experience, and the provision of practical, easy-to-follow instructions about completing the NBCSP kit. , Short message service (SMS) messaging has been demonstrated to have an impact on screening uptake for first-time screeners in the UK, as well as in Alaska Natives and people of American Indian ethnicity. A systematic review by Uy et al demonstrated that SMS reminders increased breast and cervical screening. Each of these different interventions (that is, GP-endorsement, positive reinforcement of screening by relatable people, and clear instructions) have been tested individually, but not in combination. Evidence supports the use of combining effective health-promotion interventions to increase the uptake of screening, , which was the justification for the SMARTscreen trial. The trial examined whether general practice patients aged 50–60 years who were due for screening and received the SMARTscreen intervention were more likely to complete CRC screening than similar general practice patients in the control arm. The secondary aim, not reported here, was to evaluate whether sending the SMS was acceptable and feasible to general practice staff and their patients.
A detailed trial protocol was published before practices were recruited, and the trial was conducted with no substantive changes to the protocol. CONSORT guidelines were used for reporting the trial. Trial design This was a stratified cluster randomised controlled superiority trial. General practices were randomly allocated, in a 1:1 ratio, to two trial arms, stratified by practice location and practice size according to the number of patients. The unit of randomisation was at practice level for practical reasons, including the technical implementation of the intervention, the fact that the outcome could only be collected in aggregate form from the practice’s electronic health records (EHRs), and to minimise the risk of contamination; for example, if two family members from the same practice were randomised to different arms of the trial. Setting General practices were located in the Western Victorian Primary Health Network (WVPHN) geographical catchment area in regional Victoria, Australia. This region comprises a mixture of cities and towns, with a growing population of older, and therefore eligible, people, and >200 general practices. Participants Eligibility criteria General practices were eligible if they were: from the WVPHN region (which covers the geographical location classified as Modified Monash Model [MMM] 1 to 5); and had at least two full-time equivalent GPs. People from eligible practices were included if they were ‘active patients’; that is, if they: had had at least three medical consultations at the practice within the previous 2 years (as defined by the Royal Australian College of General Practitioners); were aged 50–60 years old (inclusive); and were due to receive the NBCSP kit in the 6-month intervention period. The due date for the NBCSP kit was determined either by birthday (NBCSP kits are sent every 2 years from a patient’s 50th birthday until they are 74 years old), or 2 years after a previous NBCSP kit had been received; if a person completes a kit, the date the next kit is due will be 2 years after the date of completion. Individuals were excluded if they: did not have a recorded mobile phone number in the EHR; had a diagnosis of CRC in their medical records; or had opted out of receiving SMS messages from their practice. Intervention SMARTscreen aimed to test the effect on screening participation of a combination of four evidence-based interventions delivered in an SMS from a general practice to their patients, by comparing it with usual care. Usual care included screening with the NBCSP. The trial was restricted to 50–60-year-olds, as they have a high level of daily smartphone use and that age group had the lowest screening rates. The SMS intervention included: an endorsement of the NBCSP by the person’s own general practice; a video of a person telling a positive story about how completing the kit affected them; an animated instructional video demonstrating how to complete the kit; and a link to information about CRC, the NBCSP and the importance of screening. The SMARTscreen intervention was co-designed by the investigator team, which included clinical, academic, consumer, and industry representatives. Using existing evidence-based resources, along with an iterative process, the investigator team developed an SMS that included a text message with an embedded weblink, which opened to a customised webpage with four screens designed to increase screening. Each weblink had a unique alphanumeric code to enable tracking of the number of times each webpage was opened. An SMS was sent from the person’s general practice to each eligible individual in the last week of the month before their kit was due. Details on how patients were identified are outlined later in this article. The trial intervention period started on 1 January 2021 and ended on 30 June 2021. The outcome assessment ran over a 12-month timeframe, from 1 January 2021 until 31 December 2021. Some practices had technical difficulties installing the software required for the trial, which delayed their start date by 1 month; as such, their intervention period started on 1 February 2021 and ended on 31 July 2021, and their follow-up period ran from 1 February 2021 until 31 January 2022. In the first 3 months of the intervention period, the videos in the SMS weblink needed to be actively clicked to start playing; after 3 months, the technology changed and the video content automatically played when the weblink was opened. As such, after the first 3 months, it was no longer possible for the authors to measure whether people intentionally opened the video to watch it. Outcome The primary outcome was the difference, between the intervention and control arms, in the proportion of eligible people who completed a FIT for up to 12 months after the SMS was sent, as recorded in the EHR. Data collection The proportion of eligible people who completed a FIT was calculated by dividing the total number of eligible people who had an NBCSP FIT kit result recorded in the EHR at participating practices (numerator) by the total number of individuals eligible to receive a FIT from the NBCSP at each practice during the study period (denominator). Individuals eligible to receive a FIT in the next 6 months (denominator) were extracted from general practice EHRs using the CAT4 clinical audit tool, along with specific filters that were developed to determine when the patients’ kits were due, based on either their birthday or 2 years after their previous kit was returned. Individual-level data were stored at the general practice and each SMS was sent by the practice directly to the eligible people. Using a data-collection template, the trial coordinator at each practice recorded the number of individuals who were eligible to receive a FIT kit by the month the SMS would be due, at ages 50, 52, 54, 56, 58, 60 years, or as undetermined age, and sex (male/female). No identifying information was extracted. To determine the number of individuals with an NBCSP FIT kit result recorded in the EHR, an audit of the EHRs in each general practice was conducted by the trial coordinator. Information was collected on the patient’s month and year of birth, sex (male/female), and date of the FIT result. The date for the recorded FIT result was defined as the date the general practice received a copy of it in the EHR from the NBCSP, and had to fall between 1 January 2021 and 31 December 2021, or between 1 February 2021 and 31 January 2022 for practices with a delayed start. Within this dataset, the authors also determined the month that individuals were due to receive the SMS, based on the month of their birthday. For statistical analysis of the primary outcome, the two data sources containing the number of individuals eligible for the trial (denominator) and number of eligible individuals that had a recorded FIT result in the EHR (numerator) were counted for each general practice and merged at the aggregate level by general practice. The geographical location of the general practice was classified using the MMM, and the number of active patients aged 50–60 years was collected for each general practice. Sample size For 80% power with a two-sided 5% alpha level, the authors required 1400 eligible general practice patients from 20 practices (70 patients per practice) to detect a 10% increase in proportion with recorded bowel cancer screening in the intervention arm (50%) compared with the control arm (40%), assuming an intra-cluster correlation coefficient of 0.008. Randomisation Sequence generation General practices were randomised at a ratio of 1:1 into intervention or control arms. The allocation sequence was computer generated by a statistician, stratified by geographical remoteness (MMM 1–3 and 4–5) and general practice size (<1000 active patients versus ≥1000 active patients), using random permuted block sizes of two and four in each stratum. Allocation concealment The block sizes were not disclosed until data collection and primary analysis had been completed. Before randomising practices, the trial coordinator randomly assigned the codes 1 and 2 to the intervention and control arms, and created a masked key, which was kept separate from the random-allocation schedule. After practices consented to the trial and baseline data were collected, the statistician randomly assigned the practice to arm 1 or 2 using the random-allocation schedule. The study coordinator was then notified, and it was then that they unmasked the allocation and informed the practice to which trial arm they had been allocated. Implementation The trial coordinator met with the general practice staff via Zoom or in person, after which a delegate for the practice provided written consent to participate on behalf of the practice. All staff were given written information about the trial and the methods before consenting. Consent was only given if all GPs in the practice agreed to be involved. Blinding Practices and the trial coordinator could not be blinded for pragmatic reasons. The masked key for the trial arm allocation was only revealed to the statistician and those investigators not involved in the trial implementation after the blinded review of the primary analysis. Statistical methods The characteristics of each general practice and their eligible patients were summarised using counts and percentages by trial arm. Primary analysis was intention to treat; this included all general practices and eligible individuals, irrespective of whether they received all, or part of, the intended intervention. For the primary analysis, individuals were excluded if they were outside of the age range, or for whom a FIT result was recorded in January 2021 if their general practice started the intervention in February 2021. Generalised linear models, using the logit and identity link function, with binomial family, adjusted for the randomisation stratification factors (geographical remoteness and general practice size) were used to estimate the odds ratio and difference in proportions for the primary outcome between the two trial arms, respectively. Generalised estimating equations with robust standard errors were used with the models to allow for clustering by general practice. The estimated intervention effect was reported as the odds ratio and the percentage difference between trial arms with respective 95% confidence intervals (CIs). P -values were estimated using the logistic regression model described above. The intra-practice correlation coefficient was estimated using one-way analysis of variance and reported with 95% CI. Statistical analysis was conducted using Stata (version 17). After a blinded review of the data, before any statistical analysis, three additional sensitivity analyses, using the same statistical methods as above, were included using different criteria for defining the numerator of the primary outcome: sensitivity analysis 1 — all individuals identified with a FIT result in the EHR, including people who did not meet the eligibility criteria (that is, those outside of the age range or with FIT results recorded in the EHR before the start of the trial period); sensitivity analysis 2 — only those individuals with an identified FIT result in the EHR, whose birthday was within the 6-month intervention period; sensitivity analysis 3 — individuals as defined in sensitivity analysis 2, who had returned their FIT kit within 190 days from when their SMS was due (defined as the end of the data-collection month). Trial registration The trial was registered with the Australian New Zealand Clinical Trials Registry (ID: ACTRN12620001020976). Patient and public involvement (PPI) statement Two members of the public who were self-described as cancer PPI representatives were directly involved in the design of the intervention and contributed to the trial methodology including supervision. They were active co-investigators and co-authors of this article.
This was a stratified cluster randomised controlled superiority trial. General practices were randomly allocated, in a 1:1 ratio, to two trial arms, stratified by practice location and practice size according to the number of patients. The unit of randomisation was at practice level for practical reasons, including the technical implementation of the intervention, the fact that the outcome could only be collected in aggregate form from the practice’s electronic health records (EHRs), and to minimise the risk of contamination; for example, if two family members from the same practice were randomised to different arms of the trial.
General practices were located in the Western Victorian Primary Health Network (WVPHN) geographical catchment area in regional Victoria, Australia. This region comprises a mixture of cities and towns, with a growing population of older, and therefore eligible, people, and >200 general practices.
Eligibility criteria General practices were eligible if they were: from the WVPHN region (which covers the geographical location classified as Modified Monash Model [MMM] 1 to 5); and had at least two full-time equivalent GPs. People from eligible practices were included if they were ‘active patients’; that is, if they: had had at least three medical consultations at the practice within the previous 2 years (as defined by the Royal Australian College of General Practitioners); were aged 50–60 years old (inclusive); and were due to receive the NBCSP kit in the 6-month intervention period. The due date for the NBCSP kit was determined either by birthday (NBCSP kits are sent every 2 years from a patient’s 50th birthday until they are 74 years old), or 2 years after a previous NBCSP kit had been received; if a person completes a kit, the date the next kit is due will be 2 years after the date of completion. Individuals were excluded if they: did not have a recorded mobile phone number in the EHR; had a diagnosis of CRC in their medical records; or had opted out of receiving SMS messages from their practice.
General practices were eligible if they were: from the WVPHN region (which covers the geographical location classified as Modified Monash Model [MMM] 1 to 5); and had at least two full-time equivalent GPs. People from eligible practices were included if they were ‘active patients’; that is, if they: had had at least three medical consultations at the practice within the previous 2 years (as defined by the Royal Australian College of General Practitioners); were aged 50–60 years old (inclusive); and were due to receive the NBCSP kit in the 6-month intervention period. The due date for the NBCSP kit was determined either by birthday (NBCSP kits are sent every 2 years from a patient’s 50th birthday until they are 74 years old), or 2 years after a previous NBCSP kit had been received; if a person completes a kit, the date the next kit is due will be 2 years after the date of completion. Individuals were excluded if they: did not have a recorded mobile phone number in the EHR; had a diagnosis of CRC in their medical records; or had opted out of receiving SMS messages from their practice.
SMARTscreen aimed to test the effect on screening participation of a combination of four evidence-based interventions delivered in an SMS from a general practice to their patients, by comparing it with usual care. Usual care included screening with the NBCSP. The trial was restricted to 50–60-year-olds, as they have a high level of daily smartphone use and that age group had the lowest screening rates. The SMS intervention included: an endorsement of the NBCSP by the person’s own general practice; a video of a person telling a positive story about how completing the kit affected them; an animated instructional video demonstrating how to complete the kit; and a link to information about CRC, the NBCSP and the importance of screening. The SMARTscreen intervention was co-designed by the investigator team, which included clinical, academic, consumer, and industry representatives. Using existing evidence-based resources, along with an iterative process, the investigator team developed an SMS that included a text message with an embedded weblink, which opened to a customised webpage with four screens designed to increase screening. Each weblink had a unique alphanumeric code to enable tracking of the number of times each webpage was opened. An SMS was sent from the person’s general practice to each eligible individual in the last week of the month before their kit was due. Details on how patients were identified are outlined later in this article. The trial intervention period started on 1 January 2021 and ended on 30 June 2021. The outcome assessment ran over a 12-month timeframe, from 1 January 2021 until 31 December 2021. Some practices had technical difficulties installing the software required for the trial, which delayed their start date by 1 month; as such, their intervention period started on 1 February 2021 and ended on 31 July 2021, and their follow-up period ran from 1 February 2021 until 31 January 2022. In the first 3 months of the intervention period, the videos in the SMS weblink needed to be actively clicked to start playing; after 3 months, the technology changed and the video content automatically played when the weblink was opened. As such, after the first 3 months, it was no longer possible for the authors to measure whether people intentionally opened the video to watch it.
The primary outcome was the difference, between the intervention and control arms, in the proportion of eligible people who completed a FIT for up to 12 months after the SMS was sent, as recorded in the EHR.
The proportion of eligible people who completed a FIT was calculated by dividing the total number of eligible people who had an NBCSP FIT kit result recorded in the EHR at participating practices (numerator) by the total number of individuals eligible to receive a FIT from the NBCSP at each practice during the study period (denominator). Individuals eligible to receive a FIT in the next 6 months (denominator) were extracted from general practice EHRs using the CAT4 clinical audit tool, along with specific filters that were developed to determine when the patients’ kits were due, based on either their birthday or 2 years after their previous kit was returned. Individual-level data were stored at the general practice and each SMS was sent by the practice directly to the eligible people. Using a data-collection template, the trial coordinator at each practice recorded the number of individuals who were eligible to receive a FIT kit by the month the SMS would be due, at ages 50, 52, 54, 56, 58, 60 years, or as undetermined age, and sex (male/female). No identifying information was extracted. To determine the number of individuals with an NBCSP FIT kit result recorded in the EHR, an audit of the EHRs in each general practice was conducted by the trial coordinator. Information was collected on the patient’s month and year of birth, sex (male/female), and date of the FIT result. The date for the recorded FIT result was defined as the date the general practice received a copy of it in the EHR from the NBCSP, and had to fall between 1 January 2021 and 31 December 2021, or between 1 February 2021 and 31 January 2022 for practices with a delayed start. Within this dataset, the authors also determined the month that individuals were due to receive the SMS, based on the month of their birthday. For statistical analysis of the primary outcome, the two data sources containing the number of individuals eligible for the trial (denominator) and number of eligible individuals that had a recorded FIT result in the EHR (numerator) were counted for each general practice and merged at the aggregate level by general practice. The geographical location of the general practice was classified using the MMM, and the number of active patients aged 50–60 years was collected for each general practice.
For 80% power with a two-sided 5% alpha level, the authors required 1400 eligible general practice patients from 20 practices (70 patients per practice) to detect a 10% increase in proportion with recorded bowel cancer screening in the intervention arm (50%) compared with the control arm (40%), assuming an intra-cluster correlation coefficient of 0.008.
Sequence generation General practices were randomised at a ratio of 1:1 into intervention or control arms. The allocation sequence was computer generated by a statistician, stratified by geographical remoteness (MMM 1–3 and 4–5) and general practice size (<1000 active patients versus ≥1000 active patients), using random permuted block sizes of two and four in each stratum. Allocation concealment The block sizes were not disclosed until data collection and primary analysis had been completed. Before randomising practices, the trial coordinator randomly assigned the codes 1 and 2 to the intervention and control arms, and created a masked key, which was kept separate from the random-allocation schedule. After practices consented to the trial and baseline data were collected, the statistician randomly assigned the practice to arm 1 or 2 using the random-allocation schedule. The study coordinator was then notified, and it was then that they unmasked the allocation and informed the practice to which trial arm they had been allocated. Implementation The trial coordinator met with the general practice staff via Zoom or in person, after which a delegate for the practice provided written consent to participate on behalf of the practice. All staff were given written information about the trial and the methods before consenting. Consent was only given if all GPs in the practice agreed to be involved. Blinding Practices and the trial coordinator could not be blinded for pragmatic reasons. The masked key for the trial arm allocation was only revealed to the statistician and those investigators not involved in the trial implementation after the blinded review of the primary analysis.
General practices were randomised at a ratio of 1:1 into intervention or control arms. The allocation sequence was computer generated by a statistician, stratified by geographical remoteness (MMM 1–3 and 4–5) and general practice size (<1000 active patients versus ≥1000 active patients), using random permuted block sizes of two and four in each stratum.
The block sizes were not disclosed until data collection and primary analysis had been completed. Before randomising practices, the trial coordinator randomly assigned the codes 1 and 2 to the intervention and control arms, and created a masked key, which was kept separate from the random-allocation schedule. After practices consented to the trial and baseline data were collected, the statistician randomly assigned the practice to arm 1 or 2 using the random-allocation schedule. The study coordinator was then notified, and it was then that they unmasked the allocation and informed the practice to which trial arm they had been allocated.
The trial coordinator met with the general practice staff via Zoom or in person, after which a delegate for the practice provided written consent to participate on behalf of the practice. All staff were given written information about the trial and the methods before consenting. Consent was only given if all GPs in the practice agreed to be involved.
Practices and the trial coordinator could not be blinded for pragmatic reasons. The masked key for the trial arm allocation was only revealed to the statistician and those investigators not involved in the trial implementation after the blinded review of the primary analysis.
The characteristics of each general practice and their eligible patients were summarised using counts and percentages by trial arm. Primary analysis was intention to treat; this included all general practices and eligible individuals, irrespective of whether they received all, or part of, the intended intervention. For the primary analysis, individuals were excluded if they were outside of the age range, or for whom a FIT result was recorded in January 2021 if their general practice started the intervention in February 2021. Generalised linear models, using the logit and identity link function, with binomial family, adjusted for the randomisation stratification factors (geographical remoteness and general practice size) were used to estimate the odds ratio and difference in proportions for the primary outcome between the two trial arms, respectively. Generalised estimating equations with robust standard errors were used with the models to allow for clustering by general practice. The estimated intervention effect was reported as the odds ratio and the percentage difference between trial arms with respective 95% confidence intervals (CIs). P -values were estimated using the logistic regression model described above. The intra-practice correlation coefficient was estimated using one-way analysis of variance and reported with 95% CI. Statistical analysis was conducted using Stata (version 17). After a blinded review of the data, before any statistical analysis, three additional sensitivity analyses, using the same statistical methods as above, were included using different criteria for defining the numerator of the primary outcome: sensitivity analysis 1 — all individuals identified with a FIT result in the EHR, including people who did not meet the eligibility criteria (that is, those outside of the age range or with FIT results recorded in the EHR before the start of the trial period); sensitivity analysis 2 — only those individuals with an identified FIT result in the EHR, whose birthday was within the 6-month intervention period; sensitivity analysis 3 — individuals as defined in sensitivity analysis 2, who had returned their FIT kit within 190 days from when their SMS was due (defined as the end of the data-collection month).
The trial was registered with the Australian New Zealand Clinical Trials Registry (ID: ACTRN12620001020976).
Two members of the public who were self-described as cancer PPI representatives were directly involved in the design of the intervention and contributed to the trial methodology including supervision. They were active co-investigators and co-authors of this article.
Participant flow and recruitment shows the trial profile. Between January 2021 and July 2021, 21 general practices of 22 that were approached were recruited into the trial; of these, 11 were allocated to the intervention arm and 10 were allocated to the control arm. Sixteen practices started the intervention period on 1 January 2021 as planned. Of the five practices that started their intervention period 1 month later, one was waiting for the CAT4 clinical audit tool to be installed, and the other four had technical issues (the clinical audit was not working and required repairs). In total, there were 5451 eligible individuals: 2914 (53.5%) in the intervention practices and 2537 (46.5%) in the control practices. Baseline data Characteristics of general practices and eligible individuals were similar for the intervention and control arms ( and ). Numbers analysed The primary analysis included all 21 practices and all 5451 eligible general practice patients who were due to receive a FIT kit during the 6-month intervention period. This included 269 (9.2%) patients in the intervention arm, who did not receive the SMS. In total, 1765 individuals (1166 in the intervention arm and 599 in the control arm) had a FIT result recorded in the EHR. Of these, 39 individuals were excluded from the numerator in the primary analysis as they did not meet eligibility criteria: five were born in 1972 and were just outside the eligible age range, and 34 had a FIT result recorded in their EHR in January 2021, but their practice started the intervention in February 2021. Outcomes and estimations In total, 39.2% of individuals in the intervention arm had a FIT result recorded in the EHR compared with 23.0% of controls — an absolute increase of 16.5% (95% CI = 2.02 to 30.9; P = 0.03) in the 12-month follow-up period . The estimated intervention effect was similar when all individuals identified with a FIT result were analysed (sensitivity analysis 1) . Sensitivity analyses 2 and 3, where stricter definitions were applied to define the numerator (the denominator remained the same), showed that there was still a greater percentage of individuals who were due to receive a FIT kit in the intervention arm compared with the control arm. Harms No harms or unintended consequences were reported for anyone involved in the trial.
shows the trial profile. Between January 2021 and July 2021, 21 general practices of 22 that were approached were recruited into the trial; of these, 11 were allocated to the intervention arm and 10 were allocated to the control arm. Sixteen practices started the intervention period on 1 January 2021 as planned. Of the five practices that started their intervention period 1 month later, one was waiting for the CAT4 clinical audit tool to be installed, and the other four had technical issues (the clinical audit was not working and required repairs). In total, there were 5451 eligible individuals: 2914 (53.5%) in the intervention practices and 2537 (46.5%) in the control practices.
Characteristics of general practices and eligible individuals were similar for the intervention and control arms ( and ).
The primary analysis included all 21 practices and all 5451 eligible general practice patients who were due to receive a FIT kit during the 6-month intervention period. This included 269 (9.2%) patients in the intervention arm, who did not receive the SMS. In total, 1765 individuals (1166 in the intervention arm and 599 in the control arm) had a FIT result recorded in the EHR. Of these, 39 individuals were excluded from the numerator in the primary analysis as they did not meet eligibility criteria: five were born in 1972 and were just outside the eligible age range, and 34 had a FIT result recorded in their EHR in January 2021, but their practice started the intervention in February 2021.
In total, 39.2% of individuals in the intervention arm had a FIT result recorded in the EHR compared with 23.0% of controls — an absolute increase of 16.5% (95% CI = 2.02 to 30.9; P = 0.03) in the 12-month follow-up period . The estimated intervention effect was similar when all individuals identified with a FIT result were analysed (sensitivity analysis 1) . Sensitivity analyses 2 and 3, where stricter definitions were applied to define the numerator (the denominator remained the same), showed that there was still a greater percentage of individuals who were due to receive a FIT kit in the intervention arm compared with the control arm.
No harms or unintended consequences were reported for anyone involved in the trial.
Summary The SMARTscreen combination SMS resulted in the proportion of patients who participated in the NBCSP being 16.5% higher (95% CI = 2.02 to 30.9) in the intervention arm than the control arm, over 12 months. Given that it has been estimated that increasing screening participation by 10% could prevent 27 000 incident CRC diagnoses and 16 800 cancer deaths, and that an additional A$200 million expenditure could be saved over the next 20 years in the Australian population, these results provide compelling evidence for conducting a larger trial in the broader Australian population. Strengths and limitations There were limitations to the data the authors were able to collect. Information about NBCSP results were extracted from the general practice EHRs, which may be incomplete. In addition, the percentage of eligible individuals who completed a FIT were likely to be underestimated, as only 80% of people aged 50–60 years old attend general practice every year, and not everyone nominates a GP to receive their NBCSP results. In the authors’ recent trial, 29.7% (131/441) of people did not have their NBSCP kit results in their patient record when compared with their NBCSP records; this was similar in both arms of the same trial (28.4% in the control arm, 30.8% in the intervention arm) (unpublished data). As such, the authors expect that the underestimation would be similar for both intervention and control practices, and the estimated absolute intervention effect would be unbiased. Nevertheless, the percentage of kits returned in the intervention arm was higher (39.2%) than the national average for 50–60-year-olds during the trial period (33.4% based on complete NBCSP data ). The de-identified outcome data related to the total number of individuals eligible for the trial (denominator) and whether their FIT results were recorded in the EHR (numerator) were collected separately, so the authors were unable to link the individual-level data between the two data sources, but could link the data at the aggregate level for each practice. This limited the ability to conduct sub-group analyses to explore whether there were differences in the intervention effect by patient characteristics (for example, age groups). Another limitation was not knowing whether people were receiving the SMS just before receiving their kit. The authors estimated when the kits would be received and timed the SMS for the month before this, according to the NBCSP rules. However, during the trial period, the NBCSP delivered kits up to 6 months after people’s birth dates, not always when they were due; this meant that it was possible that people did not receive the SMS just before they were due to receive the kit, but the authors had no way of assessing this. The authors are addressing these limitations by using data from the newly established National Cancer Screening Register in a follow-on trial: SMARTERscreen. This trial is developing methods for collecting de-identified individual-level data directly from the register to ensure that the SMS is sent at the right time and the data collected are more complete; in this way, more-granular data can be provided about individual responses to the SMS, and it should be possible to ascertain the number of individuals with a FIT result, as recorded in the register. Another potential limitation was the fact that the 12-month data collection period for each general practice was fixed from the time the practice entered the trial. However, the observation period for individuals varied, ranging from a minimum of 6 months to a maximum of 12 months, depending on the time between the SMS being sent and the end of the 12-month trial data-collection period. NBCSP monitoring data show that, if people are going to return their kit, most do so within 4 months; given that there was a minimum timeframe of 6 months in the study reported here, the authors were confident they would capture the bulk of returned kits. This was the same for both trial arms. Unexpected problems were also encountered because the trial was conducted during the COVID-19 pandemic (2020–2021). However, co-design of the SMS was limited to existing evidence-based resources, expert opinion from the multidisciplinary investigator team, and consultation with PPI representatives. The authors were unable to conduct in-person consultations or focus groups with PPI as planned. The recruitment of general practices was also challenging during the COVID-19 pandemic, with general practice facing unprecedented demands, including repeated lockdowns. Despite these challenges, the required number of practices were recruited with no attrition, and practice and patient characteristics between trial arms were similar. The study was conducted in partnership with the WVPHN, and the sample was drawn from within the Western District of Victoria region. People living in that district live in, mostly, rural areas — half the practices were in areas categorised as MMM 4–5 (that is, either medium-sized or small rural towns) — and, as such, the findings might not be generalisable to the entire Australian population, the majority of whom live in metropolitan areas. As interventions have been demonstrated to be more effective in under-served groups, this result might not translate to other sub-populations with higher baseline screening rates. Also, the study was limited to the practices’ ‘active patients’, who were defined as having attended the practice at least three times within the previous 2 years. The authors excluded patients who were not regular attenders, as they assumed that ‘non-active’ patients would be less likely to nominate a GP in the trial when returning their NBCSP kit. The results demonstrated that the SMARTscreen intervention led to the proportion of patients returning a kit being 16.5% higher; however, the screening participation data could be an underestimate for the reasons highlighted above, which would suggest the health and financial benefits might be even greater than suggested by the findings. Comparison with existing literature These results demonstrate that developing a complex intervention that combines effective tools — namely, GP endorsement, narrative motivational videos, clear instructional videos, and information about bowel cancer screening — delivered via an SMS can have an effect on screening uptake that is equally as strong, or potentially greater, when packaged into one easily accessible intervention than using the individual components alone. The individual components found in previous research included a GP endorsement in a letter (11.8% increase), a mass-media campaign using positive narrative videos (15% increase), and an SMS alone (0.6% to 15%). – The use of a multifaceted SMS intervention is supported by a systematic review of interventions to increase CRC screening uptake, which found that there was a compounded effect on screening when using a combination intervention instead of single interventions. Implications for research If expanded across the country, this intervention has the potential to result in large clinical and economic beneficial outcomes. SMARTscreen provided robust evidence to support a larger trial to test for the intervention’s effectiveness in the broader Australian community. The Australian National Health and Medical Research Council has funded a trial with a larger population as a direct result of the study reported here. The new trial, SMARTERscreen, will capture data using the new National Cancer Screening Register to provide complete datasets, along with patient characteristics such as age, sex, and location, to provide more individualised results. This will allow for a more-tailored or targeted approach to screening to be provided and, potentially, for there to be a focus on specific groups that currently under-screen. The results of SMARTERscreen will be available in 2025.
The SMARTscreen combination SMS resulted in the proportion of patients who participated in the NBCSP being 16.5% higher (95% CI = 2.02 to 30.9) in the intervention arm than the control arm, over 12 months. Given that it has been estimated that increasing screening participation by 10% could prevent 27 000 incident CRC diagnoses and 16 800 cancer deaths, and that an additional A$200 million expenditure could be saved over the next 20 years in the Australian population, these results provide compelling evidence for conducting a larger trial in the broader Australian population.
There were limitations to the data the authors were able to collect. Information about NBCSP results were extracted from the general practice EHRs, which may be incomplete. In addition, the percentage of eligible individuals who completed a FIT were likely to be underestimated, as only 80% of people aged 50–60 years old attend general practice every year, and not everyone nominates a GP to receive their NBCSP results. In the authors’ recent trial, 29.7% (131/441) of people did not have their NBSCP kit results in their patient record when compared with their NBCSP records; this was similar in both arms of the same trial (28.4% in the control arm, 30.8% in the intervention arm) (unpublished data). As such, the authors expect that the underestimation would be similar for both intervention and control practices, and the estimated absolute intervention effect would be unbiased. Nevertheless, the percentage of kits returned in the intervention arm was higher (39.2%) than the national average for 50–60-year-olds during the trial period (33.4% based on complete NBCSP data ). The de-identified outcome data related to the total number of individuals eligible for the trial (denominator) and whether their FIT results were recorded in the EHR (numerator) were collected separately, so the authors were unable to link the individual-level data between the two data sources, but could link the data at the aggregate level for each practice. This limited the ability to conduct sub-group analyses to explore whether there were differences in the intervention effect by patient characteristics (for example, age groups). Another limitation was not knowing whether people were receiving the SMS just before receiving their kit. The authors estimated when the kits would be received and timed the SMS for the month before this, according to the NBCSP rules. However, during the trial period, the NBCSP delivered kits up to 6 months after people’s birth dates, not always when they were due; this meant that it was possible that people did not receive the SMS just before they were due to receive the kit, but the authors had no way of assessing this. The authors are addressing these limitations by using data from the newly established National Cancer Screening Register in a follow-on trial: SMARTERscreen. This trial is developing methods for collecting de-identified individual-level data directly from the register to ensure that the SMS is sent at the right time and the data collected are more complete; in this way, more-granular data can be provided about individual responses to the SMS, and it should be possible to ascertain the number of individuals with a FIT result, as recorded in the register. Another potential limitation was the fact that the 12-month data collection period for each general practice was fixed from the time the practice entered the trial. However, the observation period for individuals varied, ranging from a minimum of 6 months to a maximum of 12 months, depending on the time between the SMS being sent and the end of the 12-month trial data-collection period. NBCSP monitoring data show that, if people are going to return their kit, most do so within 4 months; given that there was a minimum timeframe of 6 months in the study reported here, the authors were confident they would capture the bulk of returned kits. This was the same for both trial arms. Unexpected problems were also encountered because the trial was conducted during the COVID-19 pandemic (2020–2021). However, co-design of the SMS was limited to existing evidence-based resources, expert opinion from the multidisciplinary investigator team, and consultation with PPI representatives. The authors were unable to conduct in-person consultations or focus groups with PPI as planned. The recruitment of general practices was also challenging during the COVID-19 pandemic, with general practice facing unprecedented demands, including repeated lockdowns. Despite these challenges, the required number of practices were recruited with no attrition, and practice and patient characteristics between trial arms were similar. The study was conducted in partnership with the WVPHN, and the sample was drawn from within the Western District of Victoria region. People living in that district live in, mostly, rural areas — half the practices were in areas categorised as MMM 4–5 (that is, either medium-sized or small rural towns) — and, as such, the findings might not be generalisable to the entire Australian population, the majority of whom live in metropolitan areas. As interventions have been demonstrated to be more effective in under-served groups, this result might not translate to other sub-populations with higher baseline screening rates. Also, the study was limited to the practices’ ‘active patients’, who were defined as having attended the practice at least three times within the previous 2 years. The authors excluded patients who were not regular attenders, as they assumed that ‘non-active’ patients would be less likely to nominate a GP in the trial when returning their NBCSP kit. The results demonstrated that the SMARTscreen intervention led to the proportion of patients returning a kit being 16.5% higher; however, the screening participation data could be an underestimate for the reasons highlighted above, which would suggest the health and financial benefits might be even greater than suggested by the findings.
These results demonstrate that developing a complex intervention that combines effective tools — namely, GP endorsement, narrative motivational videos, clear instructional videos, and information about bowel cancer screening — delivered via an SMS can have an effect on screening uptake that is equally as strong, or potentially greater, when packaged into one easily accessible intervention than using the individual components alone. The individual components found in previous research included a GP endorsement in a letter (11.8% increase), a mass-media campaign using positive narrative videos (15% increase), and an SMS alone (0.6% to 15%). – The use of a multifaceted SMS intervention is supported by a systematic review of interventions to increase CRC screening uptake, which found that there was a compounded effect on screening when using a combination intervention instead of single interventions.
If expanded across the country, this intervention has the potential to result in large clinical and economic beneficial outcomes. SMARTscreen provided robust evidence to support a larger trial to test for the intervention’s effectiveness in the broader Australian community. The Australian National Health and Medical Research Council has funded a trial with a larger population as a direct result of the study reported here. The new trial, SMARTERscreen, will capture data using the new National Cancer Screening Register to provide complete datasets, along with patient characteristics such as age, sex, and location, to provide more individualised results. This will allow for a more-tailored or targeted approach to screening to be provided and, potentially, for there to be a focus on specific groups that currently under-screen. The results of SMARTERscreen will be available in 2025.
|
Lessons learned from regional training of paediatric nephrology fellows in Africa | 20bdeb01-99f6-48fe-8bb5-bd329f78a938 | 10243235 | Internal Medicine[mh] | Access to care for adult patients with kidney disease is a challenge in less well-resourced communities around the world and this situation is particularly bad for African children and adolescents with kidney disease, for whom even basic management resources for treatment of acute kidney injury (AKI) is often not available . Regions in Sub-Saharan Africa, with an estimated population of 430 million children in 2015, face the same increases in demand for health services to address the “triple burden of disease” (communicable and non-communicable diseases and injury) along with extremely challenging social determinants of health (e.g., undernutrition; lack of access to education; problems related to globalization) . A well-trained healthcare workforce is critical to optimize healthcare in less well-resourced countries and to achieve the sustainable development goals. In this context, training a paediatric specific nephrology workforce is critical to meet the increasing worldwide burden of kidney disease . The availability of nephrologists who are able to provide care for individuals with kidney disorders is much higher in high income countries (HIC) compared to low income countries (LIC) . Riaz et al. reported a global nephrologist density of 10.0 per million population (pmp) and a nephrology trainee density of 1.4 pmp; however, the distribution of these human resources is highly variable, with LIC reporting nephrologist and nephrology trainee densities of 0.2 pmp and 0.1p mp compared to HIC densities of 23.2 pmp and 3.8 pmp, respectively. African and South Asian regions do not have a workforce that is sufficient to meet current clinical needs or one that is sustainable with the potential for population growth. A recent publication has predicted an insufficient number of adult nephrologists globally by 2030, unless there is an increase in funded training posts and posts for qualified nephrologists in the public sector. The situation in paediatric nephrology is even more concerning . The shortages in the nephrology workforce can be attributed to many factors, including limited physician training capacity and migration of skilled workers across and between regions . Various organizations have attempted to address some aspects of this problem by training nephrologists for adult practice. For instance, as part of its building capacity and outreach initiative, the International Society of Nephrology (ISN) has a fellowship program in which trainees travel to an advanced nephrology center to obtain skills and training and then return to their home country to practice . In addition to enhanced patient care, positive impacts of the fellowships have included continuing medical education (CME), sister kidney centers (nephrology units in emerging economies teaming up with a supporting center), and clinical research educational ambassadors (receiving expert guidance and teaching) . Physicians from low- and middle-income countries (LMIC) who undergo paediatric and adult nephrology training in high resource areas such as North America and Europe, however, often face many challenges upon returning to their countries of origin. These can include an inability to apply what they learned or experienced as a result of differences in disease patterns, poor availability of equipment, as well as a lack of support from management, administration, and even colleagues . For example, there may be a lack of paediatric nephrology (PN) training in the management of endemic conditions such as AKI due to gastroenteritis and sepsis, and clinical skills are usually taught using highly specialized equipment and consumables that are not available in many African centers. In addition, due to regulations surrounding registration for clinical practice, many fellows training outside of Africa, cannot participate fully in “hands-on training” including performing kidney biopsies and bedside insertion of acute peritoneal dialysis (PD) catheters and hemodialysis (HD) lines. Following training, many frustrated doctors may, in turn, not return to their countries of origin, further compromising the nephrology workforce. Currently, South African regulations allow limited registration for fellowship trainees, which enables “hands-on” clinical practice under supervision in training institutions such as the Red Cross War Memorial Children’s Hospital (RCWMCH), situated in Cape Town. This center has developed a number of inexpensive and readily available improvisational interventions and devices that can be taught to African trainees to apply within their home settings . These include the use of multipurpose catheters, central lines, and chest drains, or nasogastric tubes for peritoneal dialysis (PD) access, in addition to locally made PD fluid . In the early 2000’s, the International Pediatric Nephrology Association (IPNA) approved Red Cross War Memorial Children’s Hospital (RCWMCH), University of Cape Town (UCT), as one of the first training centers for paediatric nephrology (PN) in a less well-resourced country and committed to funding a number of individual 6 to 24-month nephrology training fellowships. The ISN and the International Society for Peritoneal Dialysis (ISPD) have subsequently also assisted in providing pediatric nephrology fellowship funding, despite being predominantly adult programs. Very little has been published about paediatric nephrology fellowship training as a strategy to expand the nephrology workforce in LMIC. This paper aims to describe a paediatric nephrology training program within an African setting. Most of the fellows described in this paper originate from countries where AKI is common due to acute diarrhoeal diseases and infections including malaria. Chronic kidney disease (CKD) is common in their countries as a result of undiagnosed congenital kidney pathology due to a lack of antenatal screening and the poor availability of genetic services. In addition, chronic dialysis and transplantation has only been an option for children in a few countries in Africa. Finally, with the exception of South Africa, most of the other Sub-Saharan African countries require self-funding by families for the care of patients with paediatric kidney diseases, but prior to our training program, almost no pediatric nephrology services were available in Sub-Saharan Africa.
This paper presents a descriptive audit of the development, curriculum, funding support, outputs, and experience of a paediatric nephrology (PN) fellowship training program between 1999 and 2021. Study site The training program was initiated and largely conducted at the Red Cross War Memorial Children’s Hospital (RCWMCH), Cape Town, South Africa. RCWMCH is a dedicated tertiary level, 300-bed paediatric hospital with a 12-bed paediatric nephrology and transplant unit providing care for 400 in-patients per year and 2000 outpatients. The hands-on training includes instruction on kidney biopsies, bedside insertion of acute PD catheters and all forms of acute and chronic kidney replacement therapy (KRT) for patients ranging from infants to adolescents in the nephrology unit, as well as in the 30-bed paediatric intensive care unit and neonatal high care. Senior nephrology staff consisted of 2–3 consultants over the study period, who provided all training. Whereas most of the training occurred at RCWMCH, training also occurred at the PN Unit at Tygerberg Hospital/University of Stellenbosch and in the adult tertiary center, Groote Schuur Hospital (GSH). RCWMCH and GSH started an adolescent nephrology service in 2002, which was a relatively new concept and is still the only one of its kind in South Africa. Our fellows were also able to spend some time in this clinic learning how to manage adolescents with kidney problems and how to transition what they learned upon their return to their home institutions. We would hope that this would inspire paediatric and adult nephrologists to work more closely together (as occurs in ISN) and to set up similar programs with comparable content across Africa (context of training in paediatric nephrology training Supplemental Appendix ). In addition, 5 adult nephrology trainees from the GSH Adult Nephrology Department spent 1 month in the pediatric training program at RCWMCH following the completion of their adult nephrology sub-speciality examinations. This was intended to provide them with skills to manage children with kidney diseases in addition to their adult patients upon return to their home institutions in regions where paediatric nephrology care was not available. Participants All individuals who participated in PN training were included in this review. For acceptance into the training program, fellows had to meet well-defined criteria including having completed training in general pediatrics; no prior formal training in pediatric nephrology; and approval of the application by the IPNA fellowship committee. Data collection and analysis Data obtained include details of all individuals who participated in PN training at the study site, including gender, country and institution of origin, period of training, and whether any PN tertiary qualifications were achieved. All African PN fellows from outside South Africa also completed a survey on completion of their training (Supplement Appendix ). The themes addressed included pre- and post-training work opportunities, and areas of training that were deemed particularly useful, including specific skills and exposure to multidisciplinary teams. Experiences with hands-on training opportunities were also explored. Funding Funding for the training program originated from three main sources: the IPNA and the ISN, both of which provided 1–2 years of funding per trainee, and the ISPD which provided funding for shorter durations of training (3 months). Locally, the African Fellowship Program/Children’s Hospital Trust (APFP/CHT) provided substantial administrative and moral support for all fellowship programs, as well as additional funding in cases where there was a shortfall. Ethics This study complied with the ethical guidelines and principles of the Helsinki Declaration of 2008, South African Guidelines for Good Clinical Practice, and the MRC Ethical Guidelines for Research and was approved by the Human Research Ethics Committee of the University of Cape Town (Ethics: 646/2015). Fellows completing the questionnaire consented to the publication of the information provided.
The training program was initiated and largely conducted at the Red Cross War Memorial Children’s Hospital (RCWMCH), Cape Town, South Africa. RCWMCH is a dedicated tertiary level, 300-bed paediatric hospital with a 12-bed paediatric nephrology and transplant unit providing care for 400 in-patients per year and 2000 outpatients. The hands-on training includes instruction on kidney biopsies, bedside insertion of acute PD catheters and all forms of acute and chronic kidney replacement therapy (KRT) for patients ranging from infants to adolescents in the nephrology unit, as well as in the 30-bed paediatric intensive care unit and neonatal high care. Senior nephrology staff consisted of 2–3 consultants over the study period, who provided all training. Whereas most of the training occurred at RCWMCH, training also occurred at the PN Unit at Tygerberg Hospital/University of Stellenbosch and in the adult tertiary center, Groote Schuur Hospital (GSH). RCWMCH and GSH started an adolescent nephrology service in 2002, which was a relatively new concept and is still the only one of its kind in South Africa. Our fellows were also able to spend some time in this clinic learning how to manage adolescents with kidney problems and how to transition what they learned upon their return to their home institutions. We would hope that this would inspire paediatric and adult nephrologists to work more closely together (as occurs in ISN) and to set up similar programs with comparable content across Africa (context of training in paediatric nephrology training Supplemental Appendix ). In addition, 5 adult nephrology trainees from the GSH Adult Nephrology Department spent 1 month in the pediatric training program at RCWMCH following the completion of their adult nephrology sub-speciality examinations. This was intended to provide them with skills to manage children with kidney diseases in addition to their adult patients upon return to their home institutions in regions where paediatric nephrology care was not available.
All individuals who participated in PN training were included in this review. For acceptance into the training program, fellows had to meet well-defined criteria including having completed training in general pediatrics; no prior formal training in pediatric nephrology; and approval of the application by the IPNA fellowship committee.
Data obtained include details of all individuals who participated in PN training at the study site, including gender, country and institution of origin, period of training, and whether any PN tertiary qualifications were achieved. All African PN fellows from outside South Africa also completed a survey on completion of their training (Supplement Appendix ). The themes addressed included pre- and post-training work opportunities, and areas of training that were deemed particularly useful, including specific skills and exposure to multidisciplinary teams. Experiences with hands-on training opportunities were also explored.
Funding for the training program originated from three main sources: the IPNA and the ISN, both of which provided 1–2 years of funding per trainee, and the ISPD which provided funding for shorter durations of training (3 months). Locally, the African Fellowship Program/Children’s Hospital Trust (APFP/CHT) provided substantial administrative and moral support for all fellowship programs, as well as additional funding in cases where there was a shortfall.
This study complied with the ethical guidelines and principles of the Helsinki Declaration of 2008, South African Guidelines for Good Clinical Practice, and the MRC Ethical Guidelines for Research and was approved by the Human Research Ethics Committee of the University of Cape Town (Ethics: 646/2015). Fellows completing the questionnaire consented to the publication of the information provided.
Training content All fellows received comprehensive training in theoretical and clinical/practical aspects of paediatric nephrology, complemented by hands-on clinical training with prioritization of the management of AKI. A formal syllabus and weekly timetable were developed to address the educational content of the fellowship experience (Supplement Appendix ). Fellows could attain paediatric nephrology levels 1 and 2 training competence based upon the training content (Supplement Appendix ). Ultimately, there is potential for the development of a common paediatric kidney training syllabus across the African continent. Study population and demographics A total of 38 paediatric nephrology fellows were trained at RCWMCH from 1999 to 2021. Eleven (28.9%; 5 male) fellows were from South Africa (SA), with all but one funded by local provincial funding. Eight of the SA fellows were from the Western Cape province and three were from other SA provinces. All completed 2 years of training and achieved their PN sub-specialty accreditation. All but three SA fellows returned to their original centers, two of whom are working in private practice due to a shortage of public hospital posts, but remain as honorary lecturers for UCT/RCWMCH. Twenty-seven (71.1%; 14 male) fellows were from twelve anglophone African countries outside SA. The country of origin of the fellows is listed in Table . The program has trained an average of 2 PN fellows per year overall, with the number increasing to 3 in the last 5 years. A total of 28 fellows from Africa have been trained (Supplement Appendix ). The sources of funding for training are described in Table . The PN fellows from outside SA (African fellows) have had varied sources of funding, with IPNA and ISN supplying the majority of funds. Impact of COVID During the worldwide COVID-19 pandemic, the PN fellows continued their training at the RCWMCH uninterrupted. They were educated about the risks of COVID, provided with the required personal protective equipment (PPE) and afforded the same medical facilities as the local doctors, including access to COVID-19 testing and vaccines. The presence of COVID-19 did, however, preclude travel to their countries of origin for prolonged periods of time and compromise any short rotations to the adult hospitals, potentially impacting on the clinical experience gained in adult nephrology. Duration of training The duration of fellowship training for the African PN fellows varied, largely dependent on available funding. The training period was divided into two consecutive periods for analytical purposes: 2003–2012 (period 1) and 2013–2021 (period 2). Twelve African PN fellows were trained in the first period and 16 in the second period. During the first period, African PN fellows trained for a shorter duration with a mean of 11.9 (median 12) months; while training during the second period was longer with a mean and median training time of 18.4 and 23 months, respectively. The duration of training for the African Pediatric Nephrology fellows is documented in Table . Qualifications achieved Qualifications obtained by PN fellows during training included a post-graduate diploma in PN (University of Cape Town 1-year program) or a sub-specialty certificate in pediatric nephrology (College of Paediatrics of South Africa – USA board certification equivalent), which required at least 18–24 months of supervised training at RCWMCH to be eligible to sit for the written, oral, and OSCE exam. Many of the African PN fellows now require a recognized qualification to return to their home institution for positions of leadership either at their universities or their hospitals. Of those fellows who completed ≤ 12 months training, only 3/12 (25%) completed a diploma in PN, whereas 14/16 (88%) fellows training for > 15 months graduated with a sub-specialty certificate in PN. African PN fellows training at RCWMCH for more than 12 months can register with the University of Cape Town for a Master’s in Philosophy degree, which uses the College of Paediatric Certificate as the clinical examination and requires an additional research component. Nine (56%) African PN fellows, all of whom trained for > 18 months, achieved this MPhil degree, re-affirming that 18 months was the minimum time required for optimal clinical and research training. Research Research is encouraged as part of the training, but has not always been possible for short fellowship training periods, as acquisition of hands-on clinical skills takes precedence. Fellows who received more than 1 year of funded training were able to engage in research including topics such as patient audits of posterior urethral valves, pelvi-ureteric junction obstruction, crescentic nephritis, vitamin D status in CKD, drop out from chronic peritoneal dialysis, aminophylline affect on urine output, tuberculosis in paediatric kidney transplants, review of urodynamic studies, acute post-streptococcal glomerulonephritis, and severity of deranged electrolytes and kidney function. Social factors A majority of the fellows ( n = 25; 92.6%) were unable to see their families for the full duration of training. Only three of the fellows were able to bring their families to Cape Town with them. This is evidence of the commitment and dedication the fellows had with respect to learning and the provision of service to children with paediatric nephrology disorders in their home countries. Fellowship follow-up All 27 (100%) of the African PN fellows initially returned to their countries of origin for at least 2 years, supporting the concept of training “in Africa for Africans.” Subsequently, two PN fellows emigrated—one went to the Middle East in view of an unstable political situation in their home country, and the other moved to the UK as their partner was transferred there. Thus, 94.7% of trained African fellows remain in their home countries. Post-training survey results A voluntary survey was sent to all PN fellows coming from outside of South Africa following the completion of their training with a 100% (28/28) return rate (Supplement Appendix ). Work positions changed in all cases to a more senior position once fellowship training was completed with a number becoming heads of departments in their hospitals ( n = 4), lecturers in their universities ( n = 18), and vice dean ( n = 1). Time dedicated to paediatric nephrology Upon completion of fellowship training, the percentage of work effort dedicated to PN varied, with > 60% spending > 50% of their time in clinical PN (Table ) More detailed information regarding daily workload on return to home institutions including clinical, teaching, and administrative commitments is available in Supplement Appendix . Work facility The majority of PN fellows worked in state or university hospitals following the completion of their training (Supplement Appendix ) . Institutional support Upon completion of their PN fellowships and return to their countries of origin, 12 (50% fellows who completed this part of the questionnaire) received excellent institutional support, 10 (42%) received some support, and 2 (8%) received no support. Paediatric kidney facilities available on return from training was variable (Table ). AKI management PD remains a challenge in view of lack of dialysis solutions and standard PD catheters, resulting in the need to train fellows to have improvisation skills with homemade fluids and makeshift catheters (Fig. ). Community health care There was unanimous agreement that health care in the local community had been positively affected by PN training and the return of the trained fellow to their home institution. Qualitative comments pertaining to the impact of PN training were collected (Table ). Teaching All of the African PN fellows were required to provide teaching for both undergraduate and post-graduate students upon returning to their institutions. When asked if PN training prepared them for teaching, 12 (46%) felt they were well prepared to teach, 3 (12%) felt they were prepared to some extent, 2 (7%) felt they were not prepared, and 9 (35%) did not respond to this question. In general, fellows who stayed for a longer duration of training felt more confident with teaching. Benefits of the training program Overall, hands-on training was found to be very valuable and was deemed to be the most useful part of the training program on completion (Table ). Modifications to the training program The majority (16/28) of the fellows who were able to spend more than 1 year in training managed to obtain enough experience in conducting kidney biopsies, acute PD, HD (lines and technique), KRT, and research time. Although the program was deemed to be comprehensive, the fellows reported that they required a longer duration of training to conduct research or learn teaching skills. Overall, a program lasting 18–24 months was seen as the ideal time period to learn all that was required. Additional recommendations for improved training are provided in Supplement Appendix . Assessment of the training experience All (100%) of the fellows felt that they would strongly recommend the program in terms of its hands-on approach and acquisition of skills. Subjective feedback from fellows on their training experience with subsequent recommendations has been collected (Supplement Appendix ).
All fellows received comprehensive training in theoretical and clinical/practical aspects of paediatric nephrology, complemented by hands-on clinical training with prioritization of the management of AKI. A formal syllabus and weekly timetable were developed to address the educational content of the fellowship experience (Supplement Appendix ). Fellows could attain paediatric nephrology levels 1 and 2 training competence based upon the training content (Supplement Appendix ). Ultimately, there is potential for the development of a common paediatric kidney training syllabus across the African continent.
A total of 38 paediatric nephrology fellows were trained at RCWMCH from 1999 to 2021. Eleven (28.9%; 5 male) fellows were from South Africa (SA), with all but one funded by local provincial funding. Eight of the SA fellows were from the Western Cape province and three were from other SA provinces. All completed 2 years of training and achieved their PN sub-specialty accreditation. All but three SA fellows returned to their original centers, two of whom are working in private practice due to a shortage of public hospital posts, but remain as honorary lecturers for UCT/RCWMCH. Twenty-seven (71.1%; 14 male) fellows were from twelve anglophone African countries outside SA. The country of origin of the fellows is listed in Table . The program has trained an average of 2 PN fellows per year overall, with the number increasing to 3 in the last 5 years. A total of 28 fellows from Africa have been trained (Supplement Appendix ). The sources of funding for training are described in Table . The PN fellows from outside SA (African fellows) have had varied sources of funding, with IPNA and ISN supplying the majority of funds.
During the worldwide COVID-19 pandemic, the PN fellows continued their training at the RCWMCH uninterrupted. They were educated about the risks of COVID, provided with the required personal protective equipment (PPE) and afforded the same medical facilities as the local doctors, including access to COVID-19 testing and vaccines. The presence of COVID-19 did, however, preclude travel to their countries of origin for prolonged periods of time and compromise any short rotations to the adult hospitals, potentially impacting on the clinical experience gained in adult nephrology.
The duration of fellowship training for the African PN fellows varied, largely dependent on available funding. The training period was divided into two consecutive periods for analytical purposes: 2003–2012 (period 1) and 2013–2021 (period 2). Twelve African PN fellows were trained in the first period and 16 in the second period. During the first period, African PN fellows trained for a shorter duration with a mean of 11.9 (median 12) months; while training during the second period was longer with a mean and median training time of 18.4 and 23 months, respectively. The duration of training for the African Pediatric Nephrology fellows is documented in Table .
Qualifications obtained by PN fellows during training included a post-graduate diploma in PN (University of Cape Town 1-year program) or a sub-specialty certificate in pediatric nephrology (College of Paediatrics of South Africa – USA board certification equivalent), which required at least 18–24 months of supervised training at RCWMCH to be eligible to sit for the written, oral, and OSCE exam. Many of the African PN fellows now require a recognized qualification to return to their home institution for positions of leadership either at their universities or their hospitals. Of those fellows who completed ≤ 12 months training, only 3/12 (25%) completed a diploma in PN, whereas 14/16 (88%) fellows training for > 15 months graduated with a sub-specialty certificate in PN. African PN fellows training at RCWMCH for more than 12 months can register with the University of Cape Town for a Master’s in Philosophy degree, which uses the College of Paediatric Certificate as the clinical examination and requires an additional research component. Nine (56%) African PN fellows, all of whom trained for > 18 months, achieved this MPhil degree, re-affirming that 18 months was the minimum time required for optimal clinical and research training.
Research is encouraged as part of the training, but has not always been possible for short fellowship training periods, as acquisition of hands-on clinical skills takes precedence. Fellows who received more than 1 year of funded training were able to engage in research including topics such as patient audits of posterior urethral valves, pelvi-ureteric junction obstruction, crescentic nephritis, vitamin D status in CKD, drop out from chronic peritoneal dialysis, aminophylline affect on urine output, tuberculosis in paediatric kidney transplants, review of urodynamic studies, acute post-streptococcal glomerulonephritis, and severity of deranged electrolytes and kidney function.
A majority of the fellows ( n = 25; 92.6%) were unable to see their families for the full duration of training. Only three of the fellows were able to bring their families to Cape Town with them. This is evidence of the commitment and dedication the fellows had with respect to learning and the provision of service to children with paediatric nephrology disorders in their home countries.
All 27 (100%) of the African PN fellows initially returned to their countries of origin for at least 2 years, supporting the concept of training “in Africa for Africans.” Subsequently, two PN fellows emigrated—one went to the Middle East in view of an unstable political situation in their home country, and the other moved to the UK as their partner was transferred there. Thus, 94.7% of trained African fellows remain in their home countries.
A voluntary survey was sent to all PN fellows coming from outside of South Africa following the completion of their training with a 100% (28/28) return rate (Supplement Appendix ). Work positions changed in all cases to a more senior position once fellowship training was completed with a number becoming heads of departments in their hospitals ( n = 4), lecturers in their universities ( n = 18), and vice dean ( n = 1).
Upon completion of fellowship training, the percentage of work effort dedicated to PN varied, with > 60% spending > 50% of their time in clinical PN (Table ) More detailed information regarding daily workload on return to home institutions including clinical, teaching, and administrative commitments is available in Supplement Appendix .
The majority of PN fellows worked in state or university hospitals following the completion of their training (Supplement Appendix ) .
Upon completion of their PN fellowships and return to their countries of origin, 12 (50% fellows who completed this part of the questionnaire) received excellent institutional support, 10 (42%) received some support, and 2 (8%) received no support. Paediatric kidney facilities available on return from training was variable (Table ).
PD remains a challenge in view of lack of dialysis solutions and standard PD catheters, resulting in the need to train fellows to have improvisation skills with homemade fluids and makeshift catheters (Fig. ).
There was unanimous agreement that health care in the local community had been positively affected by PN training and the return of the trained fellow to their home institution. Qualitative comments pertaining to the impact of PN training were collected (Table ).
All of the African PN fellows were required to provide teaching for both undergraduate and post-graduate students upon returning to their institutions. When asked if PN training prepared them for teaching, 12 (46%) felt they were well prepared to teach, 3 (12%) felt they were prepared to some extent, 2 (7%) felt they were not prepared, and 9 (35%) did not respond to this question. In general, fellows who stayed for a longer duration of training felt more confident with teaching.
Overall, hands-on training was found to be very valuable and was deemed to be the most useful part of the training program on completion (Table ).
The majority (16/28) of the fellows who were able to spend more than 1 year in training managed to obtain enough experience in conducting kidney biopsies, acute PD, HD (lines and technique), KRT, and research time. Although the program was deemed to be comprehensive, the fellows reported that they required a longer duration of training to conduct research or learn teaching skills. Overall, a program lasting 18–24 months was seen as the ideal time period to learn all that was required. Additional recommendations for improved training are provided in Supplement Appendix .
All (100%) of the fellows felt that they would strongly recommend the program in terms of its hands-on approach and acquisition of skills. Subjective feedback from fellows on their training experience with subsequent recommendations has been collected (Supplement Appendix ).
The attractiveness of nephrology as a specialty has diminished over the past few decades leading to global concerns regarding the future of the specialty’s workforce, even more so in LMIC . There has been a call to boost recruitment of both adult and paediatric nephrologists by increasing exposure of medical students to nephrology, providing mentoring, improving the clinical experience, incorporating procedural skills, facilitating exchanges between trainees and senior nephrologists, adapting active approaches to identify dissatisfaction and burnout, increasing renumeration, and incentivizing advances in the field of nephrology . Some high-income countries rely on foreign-trained doctors to cover shortages in nephrology staffing. For example, a study from Oman showed that the majority of practicing nephrologists were expatriate physicians, with local doctors representing only 14% of the workforce . In the USA, a recent report showed that international medical graduates represented 47% of active nephrologists and 65% of nephrology trainees . By comparison, our fellows had a 100% return rate to their home countries reversing the “brain drain” and providing PN knowledge to these countries. The Paediatric Nephrology Department at RCWMCH/UCT and the APFP/CHT have, with funding from international nephrology organizations, been able to assist in regional training of “African paediatric nephrology fellows in Africa for Africa” in 1 center. Invaluable experience was gained with “hands-on” patient examinations, diagnosis, and management, as well as procedures. Peritoneal dialysis for management of pediatric AKI using adapted techniques in the absence of the availability of a paediatric surgeon to place a Tenckhoff in a theater facility, has now become accepted as a safe and effective alternative, as recently published in the updated ISPD guidelines for management of children with AKI . In the first decade of training, most of the fellows stayed for 1 year or less; however, the training duration of the fellows in the second decade increased to 2 years as home institutions requested that their fellows complete training with formal qualifications. As trainers, we also strongly supported this philosophy in keeping with international norms of 2–4 years of nephrology training which often includes some clinical research training, the next logical step for our fellows following clinical training. In some cases, PN fellows have returned to their home institutions without completing their research training and then found it very difficult for them to complete it, as they have lacked clinical support to assist them in meeting the large workload of patients. The measure of success of this program, in addition to the fellows acquiring sub-specialty exam/post-graduate qualifications or master’s degrees, is the fact that there has been an initial “100% return rate” at 1 year to home institutions in Africa. The training program tried to ensure that the referring institution was prepared to support trained fellows and offer them a position upon their return home. This is an area where human healthcare resource planning can become more involved in local countries. More than two-thirds of the fellows returned to government/university positions, on occasion being part-time for those employed in private hospitals. As medical directors of kidney/dialysis units, communication skills, staff empowerment, allocation of resources, mentoring, team building, and strategic planning are all important skills and they learned these skills while training at RCWMCH. On return to their home centers, many of the fellows had their skills recognized which enabled them to take up more influential positions as heads of departments within their hospitals and universities. In terms of daily workload, the majority of the fellows spent their time post-training in clinical work as opposed to teaching, with very little time dedicated to administration. Unfortunately, the survey did not specifically ask about dedicated research time or whether this was included in clinical or administration time. Despite the recent publication of ISPD guidelines pertaining to the use of PD for management of AKI, overall there has been a decline in the use of PD in high-resource countries and this has resulted in the loss or absence of knowledge on PD leading to an unwarranted pessimistic view of this form of KRT . On review of the survey data pertaining to dialysis training for AKI and especially PD training with hands-on PD insertion techniques, the fellows generally felt it to be “extensively covered” and resulted in fellows being able to perform PD for the management of AKI in their own centers and characterizing it as a center strength. The benefits of PD in this setting have also been experienced by the “Saving Young Lives” initiative. Challenges pertaining to the provision of PD in LMICs include lack of PD catheters, consumables, and PD solutions . PN working in these regions have highlighted their needs to focus on pragmatic pathways to provide kidney support therapy. This program has provided that particular focus and it has been recognized as a strength by the fellows. In general, “hands-on training” was found to be the most useful part of the training, providing opportunities for fellows to acquire the skills necessary for practical management of common kidney diseases in Africa, as well as meeting on-call demands with mentor support. Supervised training for procedures such as kidney biopsies, placing bedside PD catheters, and setting up a PD system form the basis of what nephrologists do and promotes the training of others who often practice without the backup and support of interventional radiologists and surgeons specifically trained in pediatrics. On return to their home institutions, 50% of fellows felt that they did receive some support from their institutions, but the remainder felt that they needed significantly more. The most useful equipment included dialysis (PD and HD) and biopsy consumables and ultrasound machines. Other departments thought to be essential, but with limited time availability and resources included radiology (only office hour availability and absence of nuclear medicine and urodynamics), histology (office hours only and lack of EM and immunohistochemistry) and surgeons (shortage of paediatric trained surgeons and urologists, in particular). The role for training opportunities for allied and nursing healthcare workers (e.g., dialysis nurses, dialysis technicians, dietitians, and others) in their native languages also needs to be addressed because of their essential role in pediatric kidney care. Despite the challenges in their institutions, the fellows felt unanimously that their training had enabled them to positively affect health care in their community with a summary statement illustrating this: Yes, there has been a change as people are becoming more aware of paediatric kidney conditions and we are getting calls from across the country regarding management of various kidney conditions and whether or not to refer to the unit. An important part of their training was in the field of advocacy for children and adolescents with kidney disease. Whereas data collection with identification of patient outcomes and a review of the impact of training and resources on outcomes in LMIC is difficult, registries of kidney disease in children need to be established in Africa, as occurs with other adult and pediatric kidney patients to determine the true extent of children’s kidney disease and to facilitate the generation of successful diagnostic and treatment strategies . Teaching of both under- and post-graduate students was an essential part of the fellows’ responsibilities and in general, fellows who stayed for longer training, overall felt more confident with teaching. Whereas CME was also an essential part of training with many feeling it was adequate, suggestions have been made specifically to develop patient treatment pathways relevant to local conditions. In addition, the role of virtual webinars as a widely available avenue for CME education throughout LMICs requires close attention and evaluation. Recommended modifications to the program include recommendations that the program duration be a minimum of 18–24 months to allow for sub-specialty exams in PN and to permit acquisition of skills in HD (iline access and chronic HD), KRT in PICU, teaching, and research methodologies, with time to complete any research projects once they have started. Additional recommendations included development of different levels of training to allow for some local training followed by concentrated training in advanced nephrology once enough PN units have been established in the region, as well as post-graduate networking among fellows to promote the development of sister center programs. Many fellows are already on a WhatsApp group as a form of promoting a “PN Fellowship Network” in Africa and this support should be extended further (Table ). Social factor consideration The training of PN fellows funded by IPNA/ISN/ISPD and APFP has resulted in fellows not only becoming teachers and leaders in their own institutions/universities, but also leaders in IPNA, even as councilors. All the PN Fellows who have trained at RCWMCH deserve significant accolades for the sacrifices of family time (many up to 24 months) they have made to gain knowledge and training to return to their home countries with skills to support paediatric nephrology. Funding to allow visits back home should be seen as being essential considering the sacrifice these fellows make. Without the commitment of these fellows, African PN would be a poorer specialty. Following the initiation of a PN training program at RCWMCH, PN units in the Gauteng region (Johannesburg and Pretoria) and more recently Durban have also expanded this training program for fellows from other parts of Africa with the same dedication of the participants that we have witnessed.
The training of PN fellows funded by IPNA/ISN/ISPD and APFP has resulted in fellows not only becoming teachers and leaders in their own institutions/universities, but also leaders in IPNA, even as councilors. All the PN Fellows who have trained at RCWMCH deserve significant accolades for the sacrifices of family time (many up to 24 months) they have made to gain knowledge and training to return to their home countries with skills to support paediatric nephrology. Funding to allow visits back home should be seen as being essential considering the sacrifice these fellows make. Without the commitment of these fellows, African PN would be a poorer specialty. Following the initiation of a PN training program at RCWMCH, PN units in the Gauteng region (Johannesburg and Pretoria) and more recently Durban have also expanded this training program for fellows from other parts of Africa with the same dedication of the participants that we have witnessed.
This paediatric nephrology fellowship “based in Africa for Africa” has made it possible for physicians to get comprehensive PN clinical training as a result of their commitment, the commitment of those at RCWMCH and the funding support from multiple organizations. The collaborative experience has contributed to substantial improvement in the availability of pediatric kidney care in LMICs. Current IPNA fellows training situation Since the IPNA Fellowship Program was initiated almost 20 years ago, more than 260 fellows have completed their training, coming from more than 56 countries, most of them low-income countries. Currently, 41 training centers participate in this initiative, aimed to disseminate pediatric nephrology expertise to under-served areas of the world. During the current year, 13 fellows are being trained in centers located in South Africa, China, Singapore, France, and Brazil, and 14 fellows are about to start their training, half of them coming from the African continent, highlighting the relevance of this program for the region (personal communication Francisco Cano March 2023).
Since the IPNA Fellowship Program was initiated almost 20 years ago, more than 260 fellows have completed their training, coming from more than 56 countries, most of them low-income countries. Currently, 41 training centers participate in this initiative, aimed to disseminate pediatric nephrology expertise to under-served areas of the world. During the current year, 13 fellows are being trained in centers located in South Africa, China, Singapore, France, and Brazil, and 14 fellows are about to start their training, half of them coming from the African continent, highlighting the relevance of this program for the region (personal communication Francisco Cano March 2023).
Below is the link to the electronic supplementary material. Graphical abstract (PPTX 49 KB) Appendices 1–9 (DOCX 49 KB)
|
Body composition analysis using CT at three aspects of the lumbar third vertebra and its impact on the diagnosis of sarcopenia | 459f2e42-f394-4986-bf12-caef557dba24 | 11863840 | Surgical Procedures, Operative[mh] | Sarcopenia is a syndrome characterized by a progressive decrease in skeletal muscle mass, strength, and physical performance with aging . Both the European Working Group on Sarcopenia in Older People (EWGSOP) and the Asian Working Group for Sarcopenia (AWGS) have clarified the diagnosis of skeletal muscle strength, including convenient measurement methods and corresponding cutoff values . However, the diagnostic method for skeletal muscle mass has certain limitations, as its measurement methods, such as dual-energy X-ray absorptiometry (DXA) or bioelectrical impedance analysis (BIA), are not easily available in clinical practice. Additionally, the accuracy of DXA and BIA measurements depends largely on the accuracy of the equipment equations and evaluation conditions, such as temperature, humidity, and skin conditions. Therefore, there is a need for more convenient, accessible, and accurate methods for diagnosing muscle mass. Since the gold standard for skeletal muscle mass is derived from whole-body computed tomography (CT) , in clinical practice, the whole body of a patient is rarely scanned. Thus, we generally adopt the aspect of muscle area on CT to reflect whole-body muscle mass . The skeletal muscle area (SMA) of the thoracic 12th vertebra CT imaging and lumbar 3rd vertebra (L3) CT imaging have been widely applied to evaluate muscle mass in sarcopenia patients, and their correlation with whole-body muscle mass has been confirmed by several studies . Gastric cancer is the fifth most common malignancy and the fourth leading cause of cancer-related death worldwide and is mainly treated by curative gastrectomy . Several studies have found that sarcopenia is an independent risk factor for postoperative complications and overall survival in patients with gastric cancer . Patients with gastric cancer routinely undergo preoperative CT of the abdomen to assess gastric cancer staging. Thus, the SMA can be measured via L3-CT imaging without additional examinations to facilitate the clinical assessment of sarcopenia. Several research teams have used different L3-CT imaging cross-sections to diagnose skeletal muscle mass. Carey et al. used the superior aspect , Martin et al. intercepted the transverse aspect , and Zhuang et al. used the inferior aspect as the foundation for their diagnoses . Moreover, certain teams lack precise representations of the chosen cross-sections , while others have intercepted aspects that utilize the cutoffs of other teams that are inconsistent with the aspect from which the cutoff originates . This study compared data, including muscle area, on three aspects of L3-CT imaging and compared prognostic differences in the diagnosis of sarcopenia for the first time using different L3-CT imaging aspects.
Patients This study was approved by the Ethics Committee of the First Affiliated Hospital of Wenzhou Medical University. All patients who underwent radical gastrectomy for gastric cancer in the Department of Gastroenterology, The First Affiliated Hospital of Wenzhou Medical University, between July 2014 and February 2019 were included in this study. The inclusion criteria were as follows: (1) planned to undergo gastrectomy with curative intent for gastric cancer; (2) had preoperative abdominal CT available for review (no more than 1 month prior to surgery); and (3) agreed to participate in this study and signed an informed consent form. The exclusion criteria included (1) having physical deformities that prevented them from performing muscle strength or physical fitness tests and (2) those who underwent palliative surgery. All patients were routinely managed according to the 2010 Japanese Gastric Cancer Treatment Guidelines (ver. 3) . All procedures were performed by experienced surgeons who independently performed > 200 standard gastric cancer surgeries. Follow-up All patients were followed up within 1 month of surgery. Thereafter, patients were followed-up every 3 months for 2 years, every 6 months thereafter for up to 5 years, and every 1 year thereafter. Patients were contacted by telephone and scheduled to return to the hospital at the aforementioned time points to complete the follow-up program. The follow-up schedule included laboratory tests, ultrasound, computed tomography (CT), or endoscopy. The final follow-up date was November 2021. Assessment of the skeletal muscle index All patients underwent routine preoperative abdominal computed tomography. As shown in Fig. , L3-CT images of the superior, transverse, and inferior aspects was intercepted from the Picture Archiving and Communication System. According to the EWGSOP, we used skeletal muscle area to represent skeletal muscle mass . ImageJ (National Institutes of Health, Bethesda, MD, USA, 1.52v) was used to assess the skeletal muscle area using particular Hounsfield Unit (HU) criteria of -29 to 150. As needed, tissue boundaries were manually drawn for each of the three aspects. Muscle area was normalized by height squared and reported as the skeletal muscle index (SMI, cm 2 /m 2 ). Muscle strength and physical performance Preoperative grip strength and a usual gait speed of 6 m were measured according to the EWGSOP and AWGS definitions of sarcopenia to separately determine the muscle strength and physical performance of each patient. Patients were tested for preoperative grip strength using an electronic hand dynamometer (EH101; Camry Electronics Co., Ltd., Zhongshan, Guangdong, China) with a dominant hand squeeze. The time between the first and last steps over 6 m was used to measure the usual gait speed. Both parameters were measured within 7 days preoperatively, and the maximum value of three repeated tests was recorded. The diagnosis of sarcopenia Patients with low skeletal muscle mass, low muscle strength, and/or low physical performance were considered to have sarcopenia, as defined by the EWGSOP and the AWGS. In this study, sarcopenia was diagnosed as follows: (1) low muscle mass (L3 skeletal muscle index (SMI) ≤ 40.8 cm 2 /m 2 in males and ≤ 34.9 cm 2 /m 2 in females) ; (2) low muscle strength (grip strength < 28 kg in males and < 18 kg in females) ; and (3) low muscle performance (6 m usual gait speed < 1 m/s) . The cutoff value was based on the inferior aspect of the L3-CT image. Sarcopenia diagnosed in the muscle area from the superior aspect of L3-CT images was referred to as superior sarcopenia, the transverse aspect of L3-CT imaging as transverse sarcopenia, and the inferior aspect of L3-CT images was considered inferior sarcopenia. Data collection For each patient enrolled in this study, the following data were collected at the time of patient admission: preoperative patient characteristics, including age, sex, body mass index (BMI), Charlson comorbidity index, skeletal muscle mass(SMA at the three aspects of L3-CT), muscle strength(grip strength) and physical performance (6 m usual gait speed); surgical details, including laparoscopic-assisted surgery, combined organ resection, type of resection, and operative time, were collected at the end of the patient's surgery; postoperative outcomes were collected during the postoperative hospitalization and at postoperative outpatient follow-up, including tumor pathological features and postoperative complications (within 30 days postsurgery). Analysis The agreement of SMA and SMI among the three aspects of L3-CT was calculated using the Wilcoxon rank sum test, and the agreement among the three diagnoses of sarcopenia was analyzed using kappa tests. Student’s t test was used to compare continuous normally distributed data, and the Mann‒Whitney U test was applied to continuous nonnormally distributed data. Spearman correlations were used for evaluation of the association between non-normally distributed data. Categorical data were compared using the chi-square test or Fisher's exact test. Univariate analysis was used to assess the relationships between categorical variables. Univariate Cox proportional hazards models with all potential baseline predictors were constructed to calculate risk ratios (HR) and 95% CI. Variables with a trend (P < 0.05) in the univariate analysis were selected, and multivariate logistic regression or Cox proportional risk models were constructed using forward stepwise variable selection. The Kaplan‒Meier method was used to estimate survival curves, and the log-rank test was used to compare the data. The data analysis was performed using the statistical package IBM SPSS Statistics software (SPSS) version 25.0.
This study was approved by the Ethics Committee of the First Affiliated Hospital of Wenzhou Medical University. All patients who underwent radical gastrectomy for gastric cancer in the Department of Gastroenterology, The First Affiliated Hospital of Wenzhou Medical University, between July 2014 and February 2019 were included in this study. The inclusion criteria were as follows: (1) planned to undergo gastrectomy with curative intent for gastric cancer; (2) had preoperative abdominal CT available for review (no more than 1 month prior to surgery); and (3) agreed to participate in this study and signed an informed consent form. The exclusion criteria included (1) having physical deformities that prevented them from performing muscle strength or physical fitness tests and (2) those who underwent palliative surgery. All patients were routinely managed according to the 2010 Japanese Gastric Cancer Treatment Guidelines (ver. 3) . All procedures were performed by experienced surgeons who independently performed > 200 standard gastric cancer surgeries.
All patients were followed up within 1 month of surgery. Thereafter, patients were followed-up every 3 months for 2 years, every 6 months thereafter for up to 5 years, and every 1 year thereafter. Patients were contacted by telephone and scheduled to return to the hospital at the aforementioned time points to complete the follow-up program. The follow-up schedule included laboratory tests, ultrasound, computed tomography (CT), or endoscopy. The final follow-up date was November 2021.
All patients underwent routine preoperative abdominal computed tomography. As shown in Fig. , L3-CT images of the superior, transverse, and inferior aspects was intercepted from the Picture Archiving and Communication System. According to the EWGSOP, we used skeletal muscle area to represent skeletal muscle mass . ImageJ (National Institutes of Health, Bethesda, MD, USA, 1.52v) was used to assess the skeletal muscle area using particular Hounsfield Unit (HU) criteria of -29 to 150. As needed, tissue boundaries were manually drawn for each of the three aspects. Muscle area was normalized by height squared and reported as the skeletal muscle index (SMI, cm 2 /m 2 ).
Preoperative grip strength and a usual gait speed of 6 m were measured according to the EWGSOP and AWGS definitions of sarcopenia to separately determine the muscle strength and physical performance of each patient. Patients were tested for preoperative grip strength using an electronic hand dynamometer (EH101; Camry Electronics Co., Ltd., Zhongshan, Guangdong, China) with a dominant hand squeeze. The time between the first and last steps over 6 m was used to measure the usual gait speed. Both parameters were measured within 7 days preoperatively, and the maximum value of three repeated tests was recorded.
Patients with low skeletal muscle mass, low muscle strength, and/or low physical performance were considered to have sarcopenia, as defined by the EWGSOP and the AWGS. In this study, sarcopenia was diagnosed as follows: (1) low muscle mass (L3 skeletal muscle index (SMI) ≤ 40.8 cm 2 /m 2 in males and ≤ 34.9 cm 2 /m 2 in females) ; (2) low muscle strength (grip strength < 28 kg in males and < 18 kg in females) ; and (3) low muscle performance (6 m usual gait speed < 1 m/s) . The cutoff value was based on the inferior aspect of the L3-CT image. Sarcopenia diagnosed in the muscle area from the superior aspect of L3-CT images was referred to as superior sarcopenia, the transverse aspect of L3-CT imaging as transverse sarcopenia, and the inferior aspect of L3-CT images was considered inferior sarcopenia.
For each patient enrolled in this study, the following data were collected at the time of patient admission: preoperative patient characteristics, including age, sex, body mass index (BMI), Charlson comorbidity index, skeletal muscle mass(SMA at the three aspects of L3-CT), muscle strength(grip strength) and physical performance (6 m usual gait speed); surgical details, including laparoscopic-assisted surgery, combined organ resection, type of resection, and operative time, were collected at the end of the patient's surgery; postoperative outcomes were collected during the postoperative hospitalization and at postoperative outpatient follow-up, including tumor pathological features and postoperative complications (within 30 days postsurgery).
The agreement of SMA and SMI among the three aspects of L3-CT was calculated using the Wilcoxon rank sum test, and the agreement among the three diagnoses of sarcopenia was analyzed using kappa tests. Student’s t test was used to compare continuous normally distributed data, and the Mann‒Whitney U test was applied to continuous nonnormally distributed data. Spearman correlations were used for evaluation of the association between non-normally distributed data. Categorical data were compared using the chi-square test or Fisher's exact test. Univariate analysis was used to assess the relationships between categorical variables. Univariate Cox proportional hazards models with all potential baseline predictors were constructed to calculate risk ratios (HR) and 95% CI. Variables with a trend (P < 0.05) in the univariate analysis were selected, and multivariate logistic regression or Cox proportional risk models were constructed using forward stepwise variable selection. The Kaplan‒Meier method was used to estimate survival curves, and the log-rank test was used to compare the data. The data analysis was performed using the statistical package IBM SPSS Statistics software (SPSS) version 25.0.
Comparison of body composition data at three L3 aspects As shown in the Supplementary Fig. 1, we excluded unsuitable patients and finally 1116 patients were included in the study. As illustrated in Supplementary Fig. 2, a comparative analysis of the three aspects reveals that grip strength and step speed exhibit the highest correlation with SMA and SMI of the Inferior aspect. We divided all patients into two groups according to sex, and the SMA and SMI of males and females according to the three aspects are summarized in Table . As shown, there were significant differences in the SMA and SMI between these three dimensions for both men and women, all with increases from top to bottom. All three aspects remained highly correlated between the two aspects for both sexes (superior aspect vs. transverse aspect: SMA-R 2 = 0.956; SMI-R 2 = 0.935; transverse aspect vs. inferior aspect: SMA-R 2 = 0.952; SMI-R 2 = 0.930; inferior aspect vs. superior aspect: SMA-R 2 = 0.924; and SMI-R 2 = 0.887). Diagnostic consistency of three aspects of L3 We considered inferior sarcopenia the gold standard. As shown in Table and Fig. , the sensitivity of the superior aspect was 0.955, the specificity was 0.924, and the AUC value was 0.939. The sensitivity of the transverse aspect was 0.941, the specificity was 0.949, and the AUC value was 0.945, which suggests that the diagnosis of the transverse aspect and the superior aspect had a high degree of agreement with that of the inferior aspect. The kappa value for the transverse aspect was 0.803, and the kappa value for the superior aspect was 0.745; both factors also showed a high degree of consistency compared to that of the inferior aspect. Population heterogeneity in three types of sarcopenia As shown in Table , the prevalence of superior sarcopenia, transverse sarcopenia, and inferior sarcopenia were 19.7%, 17.4%, and 13.8%, respectively, which revealed that the superior and transverse aspects were used to screen more patients with sarcopenia than was the inferior aspect. However, there were no significant differences in hospitalization costs (superior sarcopenia vs. inferior sarcopenia, p = 0.622; transverse sarcopenia vs. inferior sarcopenia, p = 0.511), postoperative hospitalization time (superior sarcopenia vs. inferior sarcopenia, p = 0.335; transverse sarcopenia vs. inferior sarcopenia, p = 0.255), or other indicators between superior sarcopenia and transverse sarcopenia with inferior sarcopenia. Thus, we divided the patients into superior sarcopenia and transverse sarcopenia groups and found that patients with superior sarcopenia and without inferior sarcopenia, both men and women, had slightly higher SMI compared with those with inferior sarcopenia; moreover, their hospitalization cost was reduced by approximately 10%, and their overall hospitalization time was shortened by one day. Furthermore, patients with transverse sarcopenia without inferior sarcopenia had a slightly greater SMI in both men and women, compared with those with inferior sarcopenia, and while women with transverse sarcopenia without inferior sarcopenia had a slightly greater BMI than did those with inferior sarcopenia. Short-term postoperative complications We evaluated postoperative complications that occurred within 30 days after gastrectomy and graded the complications according to the Clavien system , including those of Grade II or higher. The results of the univariate and multivariate analyses of the predictors of postoperative complications are presented in Table . The univariate analysis showed that advanced age, superior sarcopenia, transverse sarcopenia, inferior sarcopenia, Charlson Comorbidity Index, TNM staging, combined organ removal, and open surgery were risk factors for postoperative complications. The multivariate analysis showed a higher dominance ratio for inferior sarcopenia (OR = 2.030, p < 0.001) than for superior sarcopenia (OR = 1.608, p = 0.005) and transverse sarcopenia (OR = 1.679, p = 0.004). Long-term postoperative survival outcome The median postoperative follow-up period was 59 months. The 5-year survival rates were 66.2%, 65.9%, and 65.5% for patients with superior, transverse, and inferior sarcopenia, respectively. As shown in Fig. , Kaplan‒Meier analysis revealed that overall survival (OS) (log-rank, superior sarcopenia, p = 0.0015; transverse sarcopenia, p = 0.0024; inferior sarcopenia, p = 0.003) was significantly shorter in patients with sarcopenia than in those without sarcopenia, regardless of the aspect-based diagnosis of sarcopenia. As shown in Table , multivariate Cox models showed that inferior sarcopenia (HR = 1.491, p = 0.004), histologic type, TNM staging, and resection type were independently associated with poorer overall survival. When using superior sarcopenia (HR = 1.408, p = 0.005) or transverse sarcopenia (HR = 1.376, p = 0.012) instead of inferior sarcopenia, the inclusion of sarcopenia remained in the multifactorial model. Compared to those of inferior sarcopenia, the risks of superior sarcopenia and transverse sarcopenia appeared to be lower.
As shown in the Supplementary Fig. 1, we excluded unsuitable patients and finally 1116 patients were included in the study. As illustrated in Supplementary Fig. 2, a comparative analysis of the three aspects reveals that grip strength and step speed exhibit the highest correlation with SMA and SMI of the Inferior aspect. We divided all patients into two groups according to sex, and the SMA and SMI of males and females according to the three aspects are summarized in Table . As shown, there were significant differences in the SMA and SMI between these three dimensions for both men and women, all with increases from top to bottom. All three aspects remained highly correlated between the two aspects for both sexes (superior aspect vs. transverse aspect: SMA-R 2 = 0.956; SMI-R 2 = 0.935; transverse aspect vs. inferior aspect: SMA-R 2 = 0.952; SMI-R 2 = 0.930; inferior aspect vs. superior aspect: SMA-R 2 = 0.924; and SMI-R 2 = 0.887).
We considered inferior sarcopenia the gold standard. As shown in Table and Fig. , the sensitivity of the superior aspect was 0.955, the specificity was 0.924, and the AUC value was 0.939. The sensitivity of the transverse aspect was 0.941, the specificity was 0.949, and the AUC value was 0.945, which suggests that the diagnosis of the transverse aspect and the superior aspect had a high degree of agreement with that of the inferior aspect. The kappa value for the transverse aspect was 0.803, and the kappa value for the superior aspect was 0.745; both factors also showed a high degree of consistency compared to that of the inferior aspect.
As shown in Table , the prevalence of superior sarcopenia, transverse sarcopenia, and inferior sarcopenia were 19.7%, 17.4%, and 13.8%, respectively, which revealed that the superior and transverse aspects were used to screen more patients with sarcopenia than was the inferior aspect. However, there were no significant differences in hospitalization costs (superior sarcopenia vs. inferior sarcopenia, p = 0.622; transverse sarcopenia vs. inferior sarcopenia, p = 0.511), postoperative hospitalization time (superior sarcopenia vs. inferior sarcopenia, p = 0.335; transverse sarcopenia vs. inferior sarcopenia, p = 0.255), or other indicators between superior sarcopenia and transverse sarcopenia with inferior sarcopenia. Thus, we divided the patients into superior sarcopenia and transverse sarcopenia groups and found that patients with superior sarcopenia and without inferior sarcopenia, both men and women, had slightly higher SMI compared with those with inferior sarcopenia; moreover, their hospitalization cost was reduced by approximately 10%, and their overall hospitalization time was shortened by one day. Furthermore, patients with transverse sarcopenia without inferior sarcopenia had a slightly greater SMI in both men and women, compared with those with inferior sarcopenia, and while women with transverse sarcopenia without inferior sarcopenia had a slightly greater BMI than did those with inferior sarcopenia.
We evaluated postoperative complications that occurred within 30 days after gastrectomy and graded the complications according to the Clavien system , including those of Grade II or higher. The results of the univariate and multivariate analyses of the predictors of postoperative complications are presented in Table . The univariate analysis showed that advanced age, superior sarcopenia, transverse sarcopenia, inferior sarcopenia, Charlson Comorbidity Index, TNM staging, combined organ removal, and open surgery were risk factors for postoperative complications. The multivariate analysis showed a higher dominance ratio for inferior sarcopenia (OR = 2.030, p < 0.001) than for superior sarcopenia (OR = 1.608, p = 0.005) and transverse sarcopenia (OR = 1.679, p = 0.004).
The median postoperative follow-up period was 59 months. The 5-year survival rates were 66.2%, 65.9%, and 65.5% for patients with superior, transverse, and inferior sarcopenia, respectively. As shown in Fig. , Kaplan‒Meier analysis revealed that overall survival (OS) (log-rank, superior sarcopenia, p = 0.0015; transverse sarcopenia, p = 0.0024; inferior sarcopenia, p = 0.003) was significantly shorter in patients with sarcopenia than in those without sarcopenia, regardless of the aspect-based diagnosis of sarcopenia. As shown in Table , multivariate Cox models showed that inferior sarcopenia (HR = 1.491, p = 0.004), histologic type, TNM staging, and resection type were independently associated with poorer overall survival. When using superior sarcopenia (HR = 1.408, p = 0.005) or transverse sarcopenia (HR = 1.376, p = 0.012) instead of inferior sarcopenia, the inclusion of sarcopenia remained in the multifactorial model. Compared to those of inferior sarcopenia, the risks of superior sarcopenia and transverse sarcopenia appeared to be lower.
This study compared skeletal muscle area among three aspects of L3-CT imaging and investigated, for the first time, the differences in predicting patient prognosis caused by the diagnosis of sarcopenia using muscle data obtained from different aspects of L3-CT imaging. This study showed that the SMA and the SMI increased sequentially from the top to the bottom of the L3-CT image, SMA and SMI at the inferior aspect exhibit the most robust correlation with grip strength and walking speed. However, there was a high correlation among these three aspects, and the results were consistent with those of previous studies . For this reason, more patients were diagnosed with sarcopenia in a descending sequence, from top to bottom, with more patients diagnosed with sarcopenia in the superior and transverse aspects than in the inferior aspect. Overall, there was high agreement in the diagnosis of the three aspects (superior sarcopenia vs. inferior sarcopenia, kappa value = 0.745, p < 0.001; transverse sarcopenia vs. inferior sarcopenia, kappa value = 0.803, p < 0.001). There were also significant differences in the SMA and SMI among the three aspects, with a certain consistency in diagnosis. Therefore, further analysis is needed to clarify whether this difference impacts the prediction of clinical outcomes. Several studies have shown that patients with sarcopenia have higher hospitalization costs , longer hospital stays , and shorter postoperative survival than patients without sarcopenia, in agreement with the findings of our study. We found no significant differences in postoperative length of stay and hospitalization costs among the three sarcopenia populations. After further splitting the patients into superior sarcopenia and transverse sarcopenia cohorts, patients with superior sarcopenia alone (negative for inferior sarcopenia) had a greater SMI in both sexes than did those with inferior sarcopenia. Patients in these groups also spent less on hospitalization than patients with inferior sarcopenia and had a slightly shorter length of postoperative hospitalization, while patients with transverse sarcopenia alone (negative for inferior sarcopenia) had the same SMI and postoperative hospitalization cost as patients with superior sarcopenia alone. Previous studies have revealed that a low SMI can be used as an independent risk factor for predicting postoperative length of stay , cost , complications and long-term prognosis , whereas patients with a high SMI have been found to have a better postoperative prognosis. This may partially explain the relatively better length of stay and cost performance of patients with superior sarcopenia and transverse sarcopenia. Further analysis of postoperative complications revealed that inferior sarcopenia had the best predictive ability (superior sarcopenia, OR = 1.608; transverse sarcopenia, OR = 1.769; inferior sarcopenia, OR = 2.030). According to our analysis of long-term survival, inferior sarcopenia appeared to retain high predictive power for survival (superior sarcopenia, HR = 1.408; transverse sarcopenia, HR = 1.376; inferior sarcopenia, HR = 1.491). Taken together, these findings showed that patients with superior and transverse sarcopenia had relatively high SMI, as described previously. This was perhaps because the superior and transverse sarcopenia cohort included more patients with suspected sarcopenia, for whom the short- and long-term prognostic performance was slightly better than that of patients with inferior sarcopenia. This has led to superior and transverse sarcopenia to present a relatively low risk predictive ability in prognostic predictions. At the root of this, despite the high correlation of SMA between the three aspects, differences remain in SMA values. We uniformly used a cutoff that was obtained according to the L3 inferior aspect as the basis for the diagnosis of low SMI in this study, and applying this cutoff to the superior aspect and transverse aspect may be the underlying cause of this difference. This study has several limitations. First, this was a single-center study, and a larger multicenter study is needed to validate our findings. In addition, although we clarified that the cutoff value should be consistent with the intercept, whether a specific aspect of L3 selection would yield a better prognostic value remains unclear.
In this study, we explored in detail the diagnosis of low muscle mass in patients with sarcopenia and found that when using a uniform cutoff at the inferior aspect, it may be possible that a lower SMI at the superior and transverse aspects, compared to the inferior aspect screened out more critical patients with suspected sarcopenia, which is clearly detrimental to the predictive power of the model. We recommend that when diagnosing low SMI, the aspect of interception should be uniform and consistent with the aspect of the truncation values. Since it has the potential to be incorporated into risk-scoring systems for postoperative prognosis and to improve clinical decision-making in patients, this study of patients with gastric cancer emphasizes the need for a standardized assessment of sarcopenia.
Supplementary Material 1: Supplementary Figure 1. Flowchart for exclusion of patients not suitable for enrollment. Supplementary Material 2: Supplementary Figure 2. Comparison of the correlation between grip strength, gait speed, and muscle data(SMA and SMI) at three aspects.
|
Dilemmas for the pathologist in the oncologic assessment of pancreatoduodenectomy specimens | 2cd0fef5-812c-4446-bc68-2139d9a79c76 | 5924671 | Pathology[mh] | Pancreatoduodenectomy (PD) specimens is are among the most complex resection specimens encountered by pathologists. PD is performed most often for oncological reasons, such as (pre)cancerous lesions of the pancreas, ampulla, duodenum, and distal bile duct. In this review, we provide an overview of topics relevant for current clinical practice, in terms of both macro- and microscopic assessment of PD specimens with a suspicion of pancreatic cancer. Careful evaluation is imperative to properly assess tumor origin for staging and to select postoperative treatment strategies. We describe how different grossing techniques may influence the assessment, and we discuss (potentially) relevant pathological parameters in terms of postoperative treatment and prognosis in relation to the current literature, i.e., tumor origin, completeness of resection, and tumor spread. The relevance of the different histopathological characteristics has been studied extensively. Based on a retrospective analysis of 555 patients who underwent pancreatic resection for pancreatic ductal adenocarcinoma (PDAC), Brennan et al. developed a nomogram in which they present clinical and histopathological variables relevant for survival . Resection margin status, degree of differentiation, number of tumor-negative and tumor-positive lymph nodes, T-stage, and tumor size were identified as relevant histopathological parameters on Cox multivariable analysis.
To improve the quality of pathological assessment, different proposals for standardization of gross dissection protocols and multicolor inking have been published over the past decades. In our experience, two of these protocols are most commonly used. The first is the axial sectioning method: each specimen is serially sliced perpendicular to the long axis of the duodenum over its entire craniocaudal length . The second method involves bi-valving of the specimen along the pancreatic duct and the common bile duct (CBD). After bi-sectioning, the two halves can be serially sliced in three different planes: either by axial slicing, multi-valving (serial slicing along each half of the pancreas), or bread loafing (parallel to the neck of the pancreas) . To the best of our knowledge, there are no prospective studies comparing the different protocols head-to-head regarding establishing tumor origin, completeness of resection, and evaluation of tumor spread. A direct comparison of the different grossing techniques regarding diagnostic assessment of relevant pathological parameters is therefore not possible in this review. There is no gold standard, although The Royal College of Pathologists (RCP) favors the axial slicing method . The axial slicing method, propagated by Verbeke, has several major advantages . For every surgical specimen the same protocol is used, making it easy to perform. The circumferential margins are readily assessable (Fig. ). However, some aspects can be more difficult to evaluate using this protocol. The optimal plane of the section for capturing the ampullary region cannot be ascertained beforehand; as a result, the ampullary region frequently happens to fall between sections, hindering accurate assessment of tumor origin . Moreover, the value of axial slicing in case of ampullary and cystic tumors has not been explored. This is relevant, as the origin of periampullary tumors and cystic tumors is often inconclusive on preoperative imaging. Furthermore, in case of an intraductal papillary mucinous neoplasm (IPMN), axial slicing does not allow to distinguish between lesions originating from the main pancreatic duct or side branches. Determining the precise tumor origin may be especially relevant for ampullary tumors: in a large cohort study, four subtypes of ampullary carcinomas were identified based on their origin, with differences in prognosis in this otherwise heterogeneous group of cancer . This subclassification has recently been adopted by the College of American Pathologists for synoptic reporting .. The articles reporting on this subclassification use the bi-sectioning method for grossing. It remains to be seen whether axial slicing is suited for this subclassification of ampullary cancers . Bi-sectioning of the head of the pancreas, as described by Adsay, is technically more difficult to perform . Briefly, small probes are inserted into the pancreatic duct and CBD. The head is sliced in the plane defined by the two probes, exposing both ducts longitudinally. The pancreatic duct is either approached from the ampullary orifice or from the pancreatic neck margin (Fig. ). Probing the entire pancreatic duct is not always possible due to occlusion by tumor compression, tumor growth or reactive changes. The bi-sectioning method can be modified if needed; in case of a suspicion of distal cholangiocarcinoma, the CBD is most important to visualize, whereas the pancreatic duct takes preference in cystic lesions. However, it does have important advantages compared to axial slicing. The periampullary region is always visualized and the primary origin of periampullary tumors (pancreas, duodenum, CBD, and ampulla of Vater) can be more reliably appreciated. For ampullary tumors, the tumor origin can be more accurately assessed, which facilitates subtyping of this tumor . Additionally, bi-sectioning allows much more accurate documentation of cystic tumors and their relationship to the ducts; after successful bi-sectioning, the main pancreatic duct can be completely evaluated, inked if appropriate and a distinction can be made between the CBD, main pancreatic duct, and side branches, facilitating diagnosis of main and/or side branch IPMN. Regarding the search for lymph nodes, the RCP propagates a minimum yield of 12 lymph nodes, as this number was optimal for accurate staging of node-negative patients . In order to maximize the number of harvested lymph nodes, Verbeke and Adsay describe different methods. Verbeke uses extensive perpendicular sampling of the lamellated pancreas with its surrounding soft tissue, whereas Adsay uses the so-called “orange peeling method” to procure a maximum number of lymph nodes . The orange peeling method consists of shaving off the all peripancreatic soft tissue after multicolor inking. The shaved off tissue derived from the superior mesenteric vessel margin is bread-loafed and submitted entirely. Potential tumor-involved margins can also be submitted in their entirety, enabling margin assessment. Manual lymph node dissection is performed on the remaining tissue. Manual dissection may reduce double counting of lymph nodes, as the retrieved lymph nodes can be submitted in one cassette. The left-over fatty tissue is then submitted for the detection of microscopic lymph nodes. The authors conclude that the orange peeling method considerably increases the lymph node yield; however, it is unclear against which “conventional” method of searching for lymph nodes it was compared. The increased lymph node yield might be explained by the fact that more fatty tissue is included for microscopic examination. Double counting of lymph nodes remains an issue of concern in both methods. In many institutions, a small sample of fresh tumor tissue is harvested for research purposes. To do so, multicolor inking and slicing should be performed on the freshly received specimen. The difficulty in distinguishing tumor tissue from chronic and fibrosing inflammation in the surrounding pancreatic tissue makes correct sampling challenging. A good view of the tumor is imperative. There is no literature on which method is optimal for fresh tumor sampling.
Site of origin, tumor size and tumor type are important parameters that should be considered when evaluating a PD specimen. Each parameter has its own challenges, and we review the parameters that should be considered when evaluating a PD specimen with a suspicion of PDAC. Site of origin When PD is performed for a malignant tumor, one of the first steps is to ascertain the site of origin of the primary tumor, which cannot always be easily assessed in this anatomically complex area. The papilla of Vater is the protrusion into the duodenal lumen caused by the ampulla of Vater, which is formed by dilated junction of the distal pancreatic duct and the distal CBD. The ampulla of Vater is surrounded by the sphincter of Oddi. In the periampullary region, three distinct types of epithelial lining are joining: the duodenal surface of the ampulla, lined by intestinal epithelium; the ampulla of Vater, lined by foveolar-like mucosa with scattered goblet cells; and the distal ends of the CBD and pancreatic duct, lined by pancreatobiliary-type epithelium. Pancreatic duct glands and the peribiliary glands harbor pancreatic stem cells and biliary tree progenitor cells, and these may contribute to tumor heterogeneity . Due to the complex anatomy, the periampullary area gives rise to a heterogeneous group of tumors, each with their own histologic features and biological behavior . The minor duodenal papilla, which drains the accessory duct of Santorini, is situated 2 cm proximal to the major papilla. It is usually identifiable by close inspection, unless it is obliterated by tumor or severe inflammation. The minor duodenal ampulla is lined with pancreaticobiliary-type epithelium, identical to the epithelium lining the distal CBD and pancreatic duct and surrounded by a smooth muscle layer. The muscle layer is known as the sphincter of Helly, although there is some debate whether it should be considered a proper sphincter. All tumors that can occur in the pancreatic duct and the major papilla have also been reported as occurring in the minor papilla and the duct of Santorini, and awareness of the possibility of a tumor in the minor papilla might contribute to better tumor subtyping . Adenocarcinomas in the periampullary region can arise from the duodenum, ampulla of Vater, distal CBD, or pancreatic duct. Importantly, different TNM staging and adjuvant therapies apply to each of these distinct tumors . In addition, the primary tumor site may be an in- or exclusion criterion for clinical trials. In practice, the primary site of origin of the tumor is mainly determined macroscopically, based on the location of the tumor bulk. In particular in voluminous tumors, the site of origin can be difficult to assess. In general, when the tumor origin cannot be determined, most pathologists choose a default “most likely” diagnosis of PDAC, as it is most frequently encountered and has a defined treatment strategy. It is generally accepted that patients with PDAC have a worse prognosis than patients with cholangiocarcinoma or ampullary carcinoma. In fact, survival of patients with PDAC may even be overestimated due to misclassified PDACs . Interestingly, two retrospective analyses of 510 and 198 PDs found that the histopathological subtype of periampullary adenocarcinomas may be a better predictor of patient survival than the site of origin . In these studies, patients with a pancreaticobiliary subtype of ampullary or cholangiocarcinoma had a survival like that of patients with PDAC, whereas the intestinal subtype was associated with longer survival. Unfortunately, neither study described the method of gross dissection. Tumor definitions PDAC is defined by the World Health Organization (WHO) as an infiltrating epithelial neoplasm with glandular (ductal) differentiation, usually demonstrating mucin production without a predominant component of any other histological subtype. An abundant desmoplastic stromal response is a typical feature . The morphologic features of extrahepatic cholangiocarcinoma are very similar to those of PDAC. Extrahepatic (distal) cholangiocarcinoma is defined by the WHO as a malignant epithelial tumor with glandular differentiation arising in the extrahepatic biliary system . This includes tumors arising in the intrapancreatic part of the CBD. It is often difficult to distinguish a tumor arising in the pancreas and secondarily involving the CBD from a tumor arising in the CBD and secondarily growing into the pancreas, in part because there are few distinct morphologic features pointing to either origin. When a distal cholangiocarcinoma or PDAC involves the entire ampulla, the pathologist faces a similar dilemma. Microscopic features that may point to a bile duct origin are dysplasia within the CBD, circumferential involvement of the bile duct by invasive carcinoma, intraglandular neutrophil-rich debris, and a small tubular growth pattern . The difficulty in determining the primary origin of periampullary tumors together with the lack of a clear guidance by the WHO is a source of confusion leading to a lack of conformity in diagnosis. The incidence of distal cholangiocarcinoma is likely underestimated, as in different series the estimated incidence shows a wide range in . Reevaluation of patients registered with PDAC also shows frequent misclassification of distal cholangiocarcinoma . Ampullary carcinoma is defined by the WHO as a gland-forming malignant epithelial neoplasm, originating in the ampulla of Vater. Only carcinomas either centered on the ampulla, or circumferentially surrounding it, or completely replacing the ampulla are considered ampullary carcinomas . In large tumors, for which this criterion can no longer be assessed, the presence of precursor lesions at the level of the ampulla may be of help. There is no specific subclassification for tumors arising from the different compartments of the ampulla of Vater. Owing to the lack of a clear description of what encompasses the “ampulla of Vater,” the significance of tumors arising from different sites within this complex region has not yet been elucidated. To reduce ambiguity of the entity, ampullary carcinomas are sometimes subclassified based on location into four categories, namely intra-ampullary, ampullary-ductal, periampullary duodenal, and ampullary carcinoma not otherwise specified. The categories were proposed after a retrospective analysis of 249 ampullary carcinomas, each category with a difference in survival . However, this subclassification needs further validation. Tumor size Tumor size (defined as the largest dimension of the tumor as assessed at pathology) is a well-established predictor of survival in PDAC and determines T-category for tumors limited to the pancreas. Generally, patients with a tumor size < 3 cm have a better prognosis, but this is mostly only significant in univariate analyses . Multivariate analysis with correction for spread into peripancreatic soft tissue and surrounding structures is occasionally applied . Saka et al. described staging based on tumor size—rather than T-category—as a viable method . Multivariate analysis was used, with adjustment for age, sex, International Union Against Cancer (UICC) N-stage, margin status, and lymphovascular/perineural invasion. T-category was not considered, but more than 95% of cases had peripancreatic soft tissue involvement. Cutoff values of ≤ 2, > 2–4, and > 4 cm were found highly significant in terms of prognosis, both in their own cohort of 223 PD specimens and in the SEER database. In the eighth edition of the TNM, which came into effect in January 2018, peripancreatic soft tissue involvement is no longer a factor in the determination of T-category. T1-T3 is dependent only on tumor size, whilst T4 requires tumor involvement of the coeliac axis, superior mesenteric artery, and/or common hepatic artery . Intraductal papillary mucinous neoplasm IPMN is a precursor lesion to PDAC that is regularly seen in clinical practice and has received much attention lately. Histologically, gastric-type, intestinal-type, oncocytic type, and pancreatobiliary-type IPMNs are discriminated. These different histological subtypes have been associated with different clinicopathological features, such as risk of high-grade dysplasia and malignant transformation. However, there is a debate about the clinical relevance of these subtypes since multiple subtypes are often present within the same IPMN and histological subtyping is difficult to reproduce in a substantial number of cases . IPMNs are also subclassified as main duct-type or side branch-type, based on the location of involvement of the pancreatic duct, which is assessed by imaging. In addition, some IPMNs involve both the main pancreatic duct and the side branches and are called mixed-type IPMNs. IPMNs confined to a side branch rarely evolve into malignancy and have a better prognosis than main duct and mixed-type IPMNs . There are no studies comparing the correlation between imaging findings and pathological findings in the distinction between main and side branch IPMN. Macroscopically, bi-valving will visualize the entire main pancreatic duct, potentially facilitating the determination of IPMN location. PDACs arising in an IPMN have a better prognosis than PDACs not associated with IPMN. When assessing the size of a tumor arising in an IPMN, only the invasive portion should be taken into account to determine the T-category . However, it is often difficult to discriminate the invasive from the non-invasive part macroscopically. In addition, multifocality can make it difficult to measure the diameter of the invasive component.
When PD is performed for a malignant tumor, one of the first steps is to ascertain the site of origin of the primary tumor, which cannot always be easily assessed in this anatomically complex area. The papilla of Vater is the protrusion into the duodenal lumen caused by the ampulla of Vater, which is formed by dilated junction of the distal pancreatic duct and the distal CBD. The ampulla of Vater is surrounded by the sphincter of Oddi. In the periampullary region, three distinct types of epithelial lining are joining: the duodenal surface of the ampulla, lined by intestinal epithelium; the ampulla of Vater, lined by foveolar-like mucosa with scattered goblet cells; and the distal ends of the CBD and pancreatic duct, lined by pancreatobiliary-type epithelium. Pancreatic duct glands and the peribiliary glands harbor pancreatic stem cells and biliary tree progenitor cells, and these may contribute to tumor heterogeneity . Due to the complex anatomy, the periampullary area gives rise to a heterogeneous group of tumors, each with their own histologic features and biological behavior . The minor duodenal papilla, which drains the accessory duct of Santorini, is situated 2 cm proximal to the major papilla. It is usually identifiable by close inspection, unless it is obliterated by tumor or severe inflammation. The minor duodenal ampulla is lined with pancreaticobiliary-type epithelium, identical to the epithelium lining the distal CBD and pancreatic duct and surrounded by a smooth muscle layer. The muscle layer is known as the sphincter of Helly, although there is some debate whether it should be considered a proper sphincter. All tumors that can occur in the pancreatic duct and the major papilla have also been reported as occurring in the minor papilla and the duct of Santorini, and awareness of the possibility of a tumor in the minor papilla might contribute to better tumor subtyping . Adenocarcinomas in the periampullary region can arise from the duodenum, ampulla of Vater, distal CBD, or pancreatic duct. Importantly, different TNM staging and adjuvant therapies apply to each of these distinct tumors . In addition, the primary tumor site may be an in- or exclusion criterion for clinical trials. In practice, the primary site of origin of the tumor is mainly determined macroscopically, based on the location of the tumor bulk. In particular in voluminous tumors, the site of origin can be difficult to assess. In general, when the tumor origin cannot be determined, most pathologists choose a default “most likely” diagnosis of PDAC, as it is most frequently encountered and has a defined treatment strategy. It is generally accepted that patients with PDAC have a worse prognosis than patients with cholangiocarcinoma or ampullary carcinoma. In fact, survival of patients with PDAC may even be overestimated due to misclassified PDACs . Interestingly, two retrospective analyses of 510 and 198 PDs found that the histopathological subtype of periampullary adenocarcinomas may be a better predictor of patient survival than the site of origin . In these studies, patients with a pancreaticobiliary subtype of ampullary or cholangiocarcinoma had a survival like that of patients with PDAC, whereas the intestinal subtype was associated with longer survival. Unfortunately, neither study described the method of gross dissection.
PDAC is defined by the World Health Organization (WHO) as an infiltrating epithelial neoplasm with glandular (ductal) differentiation, usually demonstrating mucin production without a predominant component of any other histological subtype. An abundant desmoplastic stromal response is a typical feature . The morphologic features of extrahepatic cholangiocarcinoma are very similar to those of PDAC. Extrahepatic (distal) cholangiocarcinoma is defined by the WHO as a malignant epithelial tumor with glandular differentiation arising in the extrahepatic biliary system . This includes tumors arising in the intrapancreatic part of the CBD. It is often difficult to distinguish a tumor arising in the pancreas and secondarily involving the CBD from a tumor arising in the CBD and secondarily growing into the pancreas, in part because there are few distinct morphologic features pointing to either origin. When a distal cholangiocarcinoma or PDAC involves the entire ampulla, the pathologist faces a similar dilemma. Microscopic features that may point to a bile duct origin are dysplasia within the CBD, circumferential involvement of the bile duct by invasive carcinoma, intraglandular neutrophil-rich debris, and a small tubular growth pattern . The difficulty in determining the primary origin of periampullary tumors together with the lack of a clear guidance by the WHO is a source of confusion leading to a lack of conformity in diagnosis. The incidence of distal cholangiocarcinoma is likely underestimated, as in different series the estimated incidence shows a wide range in . Reevaluation of patients registered with PDAC also shows frequent misclassification of distal cholangiocarcinoma . Ampullary carcinoma is defined by the WHO as a gland-forming malignant epithelial neoplasm, originating in the ampulla of Vater. Only carcinomas either centered on the ampulla, or circumferentially surrounding it, or completely replacing the ampulla are considered ampullary carcinomas . In large tumors, for which this criterion can no longer be assessed, the presence of precursor lesions at the level of the ampulla may be of help. There is no specific subclassification for tumors arising from the different compartments of the ampulla of Vater. Owing to the lack of a clear description of what encompasses the “ampulla of Vater,” the significance of tumors arising from different sites within this complex region has not yet been elucidated. To reduce ambiguity of the entity, ampullary carcinomas are sometimes subclassified based on location into four categories, namely intra-ampullary, ampullary-ductal, periampullary duodenal, and ampullary carcinoma not otherwise specified. The categories were proposed after a retrospective analysis of 249 ampullary carcinomas, each category with a difference in survival . However, this subclassification needs further validation.
Tumor size (defined as the largest dimension of the tumor as assessed at pathology) is a well-established predictor of survival in PDAC and determines T-category for tumors limited to the pancreas. Generally, patients with a tumor size < 3 cm have a better prognosis, but this is mostly only significant in univariate analyses . Multivariate analysis with correction for spread into peripancreatic soft tissue and surrounding structures is occasionally applied . Saka et al. described staging based on tumor size—rather than T-category—as a viable method . Multivariate analysis was used, with adjustment for age, sex, International Union Against Cancer (UICC) N-stage, margin status, and lymphovascular/perineural invasion. T-category was not considered, but more than 95% of cases had peripancreatic soft tissue involvement. Cutoff values of ≤ 2, > 2–4, and > 4 cm were found highly significant in terms of prognosis, both in their own cohort of 223 PD specimens and in the SEER database. In the eighth edition of the TNM, which came into effect in January 2018, peripancreatic soft tissue involvement is no longer a factor in the determination of T-category. T1-T3 is dependent only on tumor size, whilst T4 requires tumor involvement of the coeliac axis, superior mesenteric artery, and/or common hepatic artery .
IPMN is a precursor lesion to PDAC that is regularly seen in clinical practice and has received much attention lately. Histologically, gastric-type, intestinal-type, oncocytic type, and pancreatobiliary-type IPMNs are discriminated. These different histological subtypes have been associated with different clinicopathological features, such as risk of high-grade dysplasia and malignant transformation. However, there is a debate about the clinical relevance of these subtypes since multiple subtypes are often present within the same IPMN and histological subtyping is difficult to reproduce in a substantial number of cases . IPMNs are also subclassified as main duct-type or side branch-type, based on the location of involvement of the pancreatic duct, which is assessed by imaging. In addition, some IPMNs involve both the main pancreatic duct and the side branches and are called mixed-type IPMNs. IPMNs confined to a side branch rarely evolve into malignancy and have a better prognosis than main duct and mixed-type IPMNs . There are no studies comparing the correlation between imaging findings and pathological findings in the distinction between main and side branch IPMN. Macroscopically, bi-valving will visualize the entire main pancreatic duct, potentially facilitating the determination of IPMN location. PDACs arising in an IPMN have a better prognosis than PDACs not associated with IPMN. When assessing the size of a tumor arising in an IPMN, only the invasive portion should be taken into account to determine the T-category . However, it is often difficult to discriminate the invasive from the non-invasive part macroscopically. In addition, multifocality can make it difficult to measure the diameter of the invasive component.
While both the presence of perineural and vasoinvasive growth have long been established as poor prognostic factors for many malignancies including PDAC, little is published about the value of these parameters in PDAC. Although some studies have shown that perineural and vasoinvasive growth are predictive of a worse outcome in univariate or multivariate analysis , other studies did not confirm this . The incidence of perineural invasion varies between 31 and 92% across studies. The reported incidence of vascular invasion is lower, varying between 9 and 55% . As vascular elastic stains are not commonly used in the assessment of PDAC, vascular invasion can be easily missed.
Multiple names are used to designate the different margins of the PD specimen. (See Table ). Here, we use the names used by the RCP. The transection margins are the pancreatic neck margin, the CBD, the proximal stomach or duodenum and the distal duodenum or jejunum margin. The superior mesenteric vessel margin (including the superior mesenteric vein and artery margin) is considered a dissection margin. The superior mesenteric vessel margin is most frequently involved by tumor cells, most likely due to the lack of peripancreatic soft tissue in this area . The anterior surface is covered by peritoneum and considered a “free surface” rather than a dissection margin. Even so, involvement of this surface likely increases the risk of recurrence . According to the RCP, the anterior surface should be considered in margin assessment. In contrast, the College of American Pathologists does not consider the free anterior surface for tumor involvement . The same discussion applies to the posterior surface. As argued by some, the posterior-right aspect of the pancreas—where the pancreatic head transitions into the duodenum—is covered by smooth connective tissue, rendering it a free margin . However, many consider the posterior margin a dissection margin because the pancreas is dissected from the surrounding retroperitoneal soft tissue . In colon cancer, tumor extension into the overlying peritoneum is relevant for the T-category and is associated with decreased survival . Whether this is also true for pancreatic cancer has not been investigated to our knowledge. R1 resection The definition of a microscopic incomplete (R1) resection differs across countries and centers. The UICC defines R1 as microscopic residual disease, without further specifying the type of margin (transection or dissection) or the mode of propagation (direct or indirect). In Europe and Japan, the presence of tumor cells < 1 mm from the resection margin is generally considered an incomplete resection, whereas in the USA, a resection is considered incomplete only when tumor cells are present in the margin. The rule of 1 mm clearance is adopted from the circumferential margin assessment in rectal carcinoma. In pancreas resection specimens, studies have shown that an R0 resection only carries any prognostic value when it is defined as ≥ 1 mm margin clearance . Two studies found that a margin clearance of ≥ 1.5 or ≥ 2 mm, respectively, is an even better predictor of survival than a margin clearance of ≥ 1 mm . However, after correction for tumor size and other clinicopathological parameters, margin clearance only remained an independent prognostic factor in the study that used a margin clearance of ≥ 1.5 mm. Recently, a prospective study evaluated the relevance of resection margin status for survival in 561 patients, of whom the majority had received adjuvant treatment . Of these patients 80% had an R1 resection (< 1 mm clearance), of which 58% had direct margin involvement (0 mm margin clearance). In multivariate analysis, R1-status was independently associated with survival; a tumor clearance of ≥ 1 mm best identified the subgroup with favorable survival. The RCP pointed out that the ≥ 1 mm margin clearance should only apply to true transection and dissection margins, excluding the anterior free surface . The definition which is used for microscopic incomplete resection affects tumor sampling and the number of blocks to be taken. When an incomplete resection is defined as direct involvement of the margin, the peripancreatic tissue may be sampled without special care for tissue orientation (e.g., orange peeling) . However, when defining R1 as < 1 mm clearance, the relation between the inked margin and underlying (fatty) tissue must be preserved, necessitating extensive perpendicular sampling of margins that are threatened on macroscopic assessment R1 percentage as a quality parameter The percentage of R1 resections is often considered the most important quality parameter of pathologic assessment of PD specimens. In general, it is thought that meticulous assessment should result in an R1-percentage of 70% or higher . Verbeke states that more R1 resections are detected in the axial slicing method when compared to “traditional” (i.e., not axial slicing) grossing methods, and that the axial slicing method is therefore more sensitive for R1 resections . For the nine studies that are considered “traditional,” the grossing method is often not clearly described. Moreover, the definition of an R1 resection is not uniform: the included margins, as well as the definition of a positive margin differ, being 0 mm in some studies and < 1 mm in others. In our opinion, further studies are needed to adequately compare the different grossing techniques in terms of R1-percentage as a quality parameter. Moreover, uniform and validated definitions for R1 need to be specified. Indirect tumor growth within the 1-mm margin When tumor cells are present within 1 mm of the margin other than by direct tumor spread (i.e., by lymphangio-invasion, perineural invasion, or lymph node metastasis), it is unclear if this should be considered an incomplete resection. The RCP considers these cases to be R1-resections but offers no further explanation. In contrast, for the UICC, these cases are considered to be R0-resections, except when vessel wall invasion is present within 1 mm of the resection margin . Similarly, Verbeke argues that tumor cells present by perineural spread, lymphangio-invasion or lymph node metastasis within 1 mm of the margin qualifies the resection as complete, based on the following arguments . Firstly, the mode of propagation and behavior of these tumor cells is different from that of tumor cells that spread by direct invasion. Secondly, tumor cells within a lymph node are encapsulated, hence the 0-mm clearance approach seems to be appropriate. However, when tumor cells breach the lymph node capsule and infiltrate the surrounding soft tissue, the 1 mm rule becomes applicable. Thirdly, lymphovascular and perineural tumor invasion are reflective of regional spread, whilst R0 resection is commonly understood to indicate successful local clearance of tumor. Locoregional tumor recurrence because of lymph node metastasis or spread along peripheral nerves cannot be prevented by an R0 resection.
The definition of a microscopic incomplete (R1) resection differs across countries and centers. The UICC defines R1 as microscopic residual disease, without further specifying the type of margin (transection or dissection) or the mode of propagation (direct or indirect). In Europe and Japan, the presence of tumor cells < 1 mm from the resection margin is generally considered an incomplete resection, whereas in the USA, a resection is considered incomplete only when tumor cells are present in the margin. The rule of 1 mm clearance is adopted from the circumferential margin assessment in rectal carcinoma. In pancreas resection specimens, studies have shown that an R0 resection only carries any prognostic value when it is defined as ≥ 1 mm margin clearance . Two studies found that a margin clearance of ≥ 1.5 or ≥ 2 mm, respectively, is an even better predictor of survival than a margin clearance of ≥ 1 mm . However, after correction for tumor size and other clinicopathological parameters, margin clearance only remained an independent prognostic factor in the study that used a margin clearance of ≥ 1.5 mm. Recently, a prospective study evaluated the relevance of resection margin status for survival in 561 patients, of whom the majority had received adjuvant treatment . Of these patients 80% had an R1 resection (< 1 mm clearance), of which 58% had direct margin involvement (0 mm margin clearance). In multivariate analysis, R1-status was independently associated with survival; a tumor clearance of ≥ 1 mm best identified the subgroup with favorable survival. The RCP pointed out that the ≥ 1 mm margin clearance should only apply to true transection and dissection margins, excluding the anterior free surface . The definition which is used for microscopic incomplete resection affects tumor sampling and the number of blocks to be taken. When an incomplete resection is defined as direct involvement of the margin, the peripancreatic tissue may be sampled without special care for tissue orientation (e.g., orange peeling) . However, when defining R1 as < 1 mm clearance, the relation between the inked margin and underlying (fatty) tissue must be preserved, necessitating extensive perpendicular sampling of margins that are threatened on macroscopic assessment
The percentage of R1 resections is often considered the most important quality parameter of pathologic assessment of PD specimens. In general, it is thought that meticulous assessment should result in an R1-percentage of 70% or higher . Verbeke states that more R1 resections are detected in the axial slicing method when compared to “traditional” (i.e., not axial slicing) grossing methods, and that the axial slicing method is therefore more sensitive for R1 resections . For the nine studies that are considered “traditional,” the grossing method is often not clearly described. Moreover, the definition of an R1 resection is not uniform: the included margins, as well as the definition of a positive margin differ, being 0 mm in some studies and < 1 mm in others. In our opinion, further studies are needed to adequately compare the different grossing techniques in terms of R1-percentage as a quality parameter. Moreover, uniform and validated definitions for R1 need to be specified.
When tumor cells are present within 1 mm of the margin other than by direct tumor spread (i.e., by lymphangio-invasion, perineural invasion, or lymph node metastasis), it is unclear if this should be considered an incomplete resection. The RCP considers these cases to be R1-resections but offers no further explanation. In contrast, for the UICC, these cases are considered to be R0-resections, except when vessel wall invasion is present within 1 mm of the resection margin . Similarly, Verbeke argues that tumor cells present by perineural spread, lymphangio-invasion or lymph node metastasis within 1 mm of the margin qualifies the resection as complete, based on the following arguments . Firstly, the mode of propagation and behavior of these tumor cells is different from that of tumor cells that spread by direct invasion. Secondly, tumor cells within a lymph node are encapsulated, hence the 0-mm clearance approach seems to be appropriate. However, when tumor cells breach the lymph node capsule and infiltrate the surrounding soft tissue, the 1 mm rule becomes applicable. Thirdly, lymphovascular and perineural tumor invasion are reflective of regional spread, whilst R0 resection is commonly understood to indicate successful local clearance of tumor. Locoregional tumor recurrence because of lymph node metastasis or spread along peripheral nerves cannot be prevented by an R0 resection.
Tumor-positive lymph nodes Metastasis to regional lymph nodes is independently associated with poor survival in PDAC , although this has not been found in all series . According to the UICC, regional lymph nodes are grouped into anterior pancreatoduodenal, posterior pancreatoduodenal, inferior (including the lymph nodes around the superior mesenteric vessels), CBD, coeliac, infrapyloric, and superior and proximal mesentery lymph nodes . Metastasis in non-regional LNs is defined as distant metastasis (M1). The 7th edition of the UICC staging system only considers the presence or absence of regional nodal disease. In 2015, Basturk et al. analyzed in 227 PDACs the prognostic value of the other two substaging protocols used for gastrointestinal malignancies, for which lymph node assessment had been performed according to a standard protocol. Whilst the N-category of upper gastrointestinal malignancies (N0 no metastasis, N1 metastasis to 1–2 lymph nodes, N2 metastasis to > 2 lymph nodes) performed best, they found that the N-category of the lower gastrointestinal organs (N0 no metastasis, N1 metastasis to 1–3 lymph nodes, N2 metastasis to > 3 lymph nodes) had significantly more prognostic value than that used in the 7th edition of the UICC . In the eighth edition of the TNM, the N-category matches that of the lower gastrointestinal organs . The Japan Pancreas Society distinguishes three N-categories and gives a weighting factor according the location of the lymph nodes . In published series, microscopically tumor-positive lymph nodes are found in up to 80% of surgical specimens during routine examination . Tumor involvement of distant lymph nodes is associated with decreased survival in pancreatic cancer patients . However, extended lymphadenectomy is discouraged and seldom performed, as it has been shown to be of limited value in long term survival, whilst increasing morbidity . Direct invasion of lymph nodes Direct invasion of lymph nodes by continuous tumor growth is present in about 5–10% of PD specimens. Two articles report no difference in survival for patients with direct nodal invasion versus those with lymph node metastasis, warranting an interpretation of direct invasion as “regular” lymph node positivity. Another article found that patients with isolated lymph node involvement did have improved survival compared to patients with metastasis to regional lymph nodes . The biological mechanisms responsible for eventual differences between different modalities of tumor spread remain unclear . Nodal micrometastases A few studies have evaluated the implication of the presence of isolated tumor cells or micrometastases in lymph nodes . The presence of nodal micrometastases identified on immunohistochemistry appears to be an independent prognostic factor for patients that were considered node-negative on routine histological examination. These patients have similar survival curves as N1 patients on routine examination. Patients with tumor negative lymph nodes by immunohistochemistry have a markedly better survival compared to patients with micrometastases. These studies did however not differentiate between nodal micrometastases and isolated tumor cells. Lymph node ratio The lymph node ratio considers both the total number of lymph nodes and the number of positive lymph nodes. It has proven to be an important predictor for survival in pancreatic cancer, although the predictive value of this parameter remains proportional to the adequacy of lymph node yield . In addition, it has proven more predictive than the dichotomous presence or absence of nodal disease . Extra-nodal lymph node spread Extra-nodal tumor spread from tumor positive lymph nodes is an adverse prognostic factor in many tumor types, including rectal, thyroid, bladder and gastric cancer. Luchini et al. recently published a review on the significance of extra-nodal spread in PDAC . It was associated with a poor prognosis in terms of overall and disease-specific survival. For this reason, they argue that its presence should be considered in oncologic staging and the choice of therapeutic approach. They also concluded that extra-nodal tumor spread may be present in more than 50% of N1 patients. Peripancreatic soft tissue spread Spread of a tumor outside of the pancreas into the surrounding soft tissue or adjacent organs was a feature of T3 tumors according to theTNM 7 classification. The assessment of tumor spread into the peripancreatic soft tissue has a large margin of error, as the pancreas lacks a true capsule, meaning there is no clear demarcation between pancreatic and peripancreatic tissue. Peripancreatic soft tissue involvement appears to be present in most cases. Saka et al. showed that, when using the orange peeling method for the peripancreatic soft tissue, tumor invasion is nearly always present (> 95%) and therefore not a good predictor of survival . Another complicating factor when assessing tumor spread into peripancreatic soft tissue is atrophy and fatty degeneration of the exocrine pancreas. Islet cells may be of guidance in these cases, as their presence indicates the previous level of the exocrine pancreas. In the eighth edition of the TNM, peripancreatic soft tissue spread is no longer relevant for the T-category; only tumor size is considered in the definition of T1-T3.
Metastasis to regional lymph nodes is independently associated with poor survival in PDAC , although this has not been found in all series . According to the UICC, regional lymph nodes are grouped into anterior pancreatoduodenal, posterior pancreatoduodenal, inferior (including the lymph nodes around the superior mesenteric vessels), CBD, coeliac, infrapyloric, and superior and proximal mesentery lymph nodes . Metastasis in non-regional LNs is defined as distant metastasis (M1). The 7th edition of the UICC staging system only considers the presence or absence of regional nodal disease. In 2015, Basturk et al. analyzed in 227 PDACs the prognostic value of the other two substaging protocols used for gastrointestinal malignancies, for which lymph node assessment had been performed according to a standard protocol. Whilst the N-category of upper gastrointestinal malignancies (N0 no metastasis, N1 metastasis to 1–2 lymph nodes, N2 metastasis to > 2 lymph nodes) performed best, they found that the N-category of the lower gastrointestinal organs (N0 no metastasis, N1 metastasis to 1–3 lymph nodes, N2 metastasis to > 3 lymph nodes) had significantly more prognostic value than that used in the 7th edition of the UICC . In the eighth edition of the TNM, the N-category matches that of the lower gastrointestinal organs . The Japan Pancreas Society distinguishes three N-categories and gives a weighting factor according the location of the lymph nodes . In published series, microscopically tumor-positive lymph nodes are found in up to 80% of surgical specimens during routine examination . Tumor involvement of distant lymph nodes is associated with decreased survival in pancreatic cancer patients . However, extended lymphadenectomy is discouraged and seldom performed, as it has been shown to be of limited value in long term survival, whilst increasing morbidity .
Direct invasion of lymph nodes by continuous tumor growth is present in about 5–10% of PD specimens. Two articles report no difference in survival for patients with direct nodal invasion versus those with lymph node metastasis, warranting an interpretation of direct invasion as “regular” lymph node positivity. Another article found that patients with isolated lymph node involvement did have improved survival compared to patients with metastasis to regional lymph nodes . The biological mechanisms responsible for eventual differences between different modalities of tumor spread remain unclear .
A few studies have evaluated the implication of the presence of isolated tumor cells or micrometastases in lymph nodes . The presence of nodal micrometastases identified on immunohistochemistry appears to be an independent prognostic factor for patients that were considered node-negative on routine histological examination. These patients have similar survival curves as N1 patients on routine examination. Patients with tumor negative lymph nodes by immunohistochemistry have a markedly better survival compared to patients with micrometastases. These studies did however not differentiate between nodal micrometastases and isolated tumor cells.
The lymph node ratio considers both the total number of lymph nodes and the number of positive lymph nodes. It has proven to be an important predictor for survival in pancreatic cancer, although the predictive value of this parameter remains proportional to the adequacy of lymph node yield . In addition, it has proven more predictive than the dichotomous presence or absence of nodal disease .
Extra-nodal tumor spread from tumor positive lymph nodes is an adverse prognostic factor in many tumor types, including rectal, thyroid, bladder and gastric cancer. Luchini et al. recently published a review on the significance of extra-nodal spread in PDAC . It was associated with a poor prognosis in terms of overall and disease-specific survival. For this reason, they argue that its presence should be considered in oncologic staging and the choice of therapeutic approach. They also concluded that extra-nodal tumor spread may be present in more than 50% of N1 patients.
Spread of a tumor outside of the pancreas into the surrounding soft tissue or adjacent organs was a feature of T3 tumors according to theTNM 7 classification. The assessment of tumor spread into the peripancreatic soft tissue has a large margin of error, as the pancreas lacks a true capsule, meaning there is no clear demarcation between pancreatic and peripancreatic tissue. Peripancreatic soft tissue involvement appears to be present in most cases. Saka et al. showed that, when using the orange peeling method for the peripancreatic soft tissue, tumor invasion is nearly always present (> 95%) and therefore not a good predictor of survival . Another complicating factor when assessing tumor spread into peripancreatic soft tissue is atrophy and fatty degeneration of the exocrine pancreas. Islet cells may be of guidance in these cases, as their presence indicates the previous level of the exocrine pancreas. In the eighth edition of the TNM, peripancreatic soft tissue spread is no longer relevant for the T-category; only tumor size is considered in the definition of T1-T3.
An increasing proportion of patients receives neoadjuvant therapy, including chemotherapy and radiation therapy. The influence of these treatments on the clinical relevance and definition of R-status is still unknown . Reactive changes such as fibrosis induced by neoadjuvant therapy, in post-treatment resection specimens might complicate the macroscopic identification of the tumor. Microscopic assessment of these cases includes evaluation of the response to preoperative treatment; however, there is no consensus on how tumor regression should be graded . In a large retrospective study, Chatterjee compared 233 patients with no and minimal residual disease to patients with moderate and poor response . They found that patients with no or minimal residual disease have significantly improved survival. This advantage remained after multivariate analysis including pathologic tumor stage, margin status, and lymph node status. They conclude that histologic grading is an important prognostic factor. Several grading schemes for the assessment of residual tumor in post-treatment PD specimens have been proposed, including the College of American Pathologists (CAP) and Evans grading system . Recently, Lee et al. validated their own 3-tiered histologic tumor regression grading (HTGR) scheme (HTRG 0, no viable tumor; HTRG 1, < 5% viable tumor cells; HTRG 2, ≥ 5% viable tumor cells). In multivariate analysis, HTRG grade 0 or 1 was an independent prognostic factor for better disease-free survival, but not for overall survival . In a recent retrospective cohort study, 398 PDAC patients who underwent neoadjuvant therapy and PD were analyzed to validate the new size-based T-category definitions of the eight edition of the TNM . The authors showed that the new T stage system outperformed the old T stage system in patients after neoadjuvant treatment. Additionally, they found that a tumor size cutoff of 1.0 cm worked better for T2 than the proposed tumor size cutoff of 2.0 cm in this group of patients. Interestingly, all nine patients with a complete pathologic response showed no recurrence at the end of follow-up.
We presented a comprehensive overview of the dilemmas the pathologist may face in the assessment of a PD specimen. Several different approached for gross dissection have been proposed in the literature, each having its advantages and disadvantages about assessment of important clinicopathological parameters. Compared with axial slicing, bi-valving seems better suited for the assessment of tumor origin. The value in terms of prognosis of some the clinicopathological parameters (e.g., tumor size) has been evaluated by many studies. However, most of these studies were retrospective and the protocols and frequently used definitions lacked a standardized histopathological approach, both at the macroscopic and microscopic level. For several pathological parameters, e.g., completeness of resection, this has hampered clinical validation and further application. As a result, different organizations have published their own guidelines, which show divergence on potentially important issues. Pathologists and surgeons should be aware of these differences and of the uncertainties in histopathological assessment of PDs. Neoadjuvant therapy, which is increasingly administered, will also influence the assessment of the specimen and the interpretation of certain parameters in the resection specimen after chemotherapy . Further prospective studies are needed to validate the clinical relevance of the various dissection protocols and the interpretation of certain macroscopic and microscopic parameters.
|
COVID-19 pandemic and lockdown: Changing trends in Ophthalmology for in-patient and emergency services | 09bfe65b-17b9-4971-ab89-b5ddefc288f3 | 7942100 | Ophthalmology[mh] | This retrospective study was carried out in a government medical college and multispecialty tertiary care hospital in North India (Chandigarh). This hospital in addition to serving the population of Chandigarh, also serves neighboring states of North India. All routine and outpatient services have been suspended since the national lockdown (March 24, 2020). The hospital in addition to providing emergency care services also has been treating systemically ill COVID patients, in COVID dedicated areas. A special hospital infection control committee was formed and nodal officers were appointed in each individual department to look into infection control practices. Data of patients who underwent COVID testing was reported to nodal officers each day. Since March 25, 2020, Ophthalmic care delivery was managed via the hospital telemedicine portal and 24-h emergency services. At the time of submission of the manuscript, the hospital had still not resumed its outpatient services due to heavy COVID load. The study adheres to tenets of the Declaration of Helsinki. No identifiable parameters of patient information were used for data analysis. Retrospective data containing records of clinical presentation to the ophthalmology emergency services, in-patient records, operating room records and utilization of items was analyzed for the 6-month period (March 25 to September 30, 2020) post national lockdown. Triaging of patients was done and practice guidelines as issued by the All India Ophthalmological Society (AIOS) were followed. A detailed history was taken including contact and travel history, and patients were also enquired about their awareness regarding COVID. Personal protective equipment (PPE) kit was donned before examining COVID suspects in the emergency. All patients who required hospital admission underwent COVID testing prior to admission. The data pertaining to the 6 month study period was compared to retrospective data of the same time period in 2019. All data was entered in Microsoft Excel 2019 (Microsoft Corporation, Redmond, USA). Descriptive data was described using mean (±Standard deviation) or median (Range). P value <0.05 was considered statistically significant.
Emergency services A total of 1152 new patients were seen in emergency ophthalmology service from March 25 to September 30, 2020, as opposed to 1515 during the same time period last year (March 25 to September 30, 2020). A decrease of 23.9% was observed, due to the ongoing pandemic and lockdown. On triaging patients according to AIOS guidelines, 58.6% ( n = 676) were classified as emergency cases, 18.3% ( n = 211) as urgent cases and 23% ( n = 265) as routine cases. Patients requiring routine care were referred to tele-ophthalmology service. One of the earliest Indian studies that evaluated patient presentation in ophthalmic emergency during lockdown had observed similar proportions, reporting 73% as emergency or urgent, whereas 26.85% as routine care patients. The average number of patients seen in emergency per day was 5.96 in the study period as opposed to 8.02 per day in same time period last year. Average number of patients per day nosedived to 3.64 in May 2020 but rising trends were again observed as lockdown was partially lifted in June, 2020 (6.03). The trends have been rising each month since, as further restrictions were relaxed. Trends of average number of patients seen in emergency have been illustrated in . During this 6-month period, 324 patients directly accessed tele-ophthalmology services, forming only 21.95% of total patients that accessed eye care services. Average age of patients presenting to emergency service was 34 ± 7.2 years in this year's study period as opposed to 40.2 ± 5.6 years in the same time period last year. Majority of patients in 2020 study period were males (61.8%, n = 712), whereas the male predominance was lower in 2019 study period (56.1%, n = 850). Most common presenting complaint in emergency setting was eye trauma due to road traffic accidents (36.19%, n = 417), followed by dryness and allergy related symptoms (13.2%, n = 153), assault (9.2%, n = 106) and fall from height (9.4%, n = 109). Last year's data showed that eye trauma due to road traffic accidents (52.1%, n = 792) was most common presentation followed by fall from height (10.6%, n = 162) and trauma due to minor household accidents (7.1%, n = 108). A downward trend was seen in ocular trauma secondary to road traffic accident, fall from height, minor household accidental trauma, conjunctivitis cases and firecracker injury as compared to last year. Increased cases of assault, chemical injury, foreign body in eye, and dryness and allergy related problems were noted this year. A disturbing trend was noted in household violence this year (32% of all assault cases, n = 34) as compared to last year (14%, n = 14). The presenting complaints in emergency service have been summarized in . In-patient census Only 45 patients were admitted in the 6-month study period, as compared to 1300 last year in same time period. A decrease of 96.53% was noted this year. The most common indication for hospital admission in this year's study period was open globe injury, constituting 54.83% ( n = 24) of all admitted patients. Other indications were traumatic optic neuropathy needing intravenous steroids (17.7%, n = 8), retinoblastoma patients for chemotherapy (11.1%, n = 5), orbital cellulitis (11.1%, n = 5), severe fungal corneal ulcer (4.4%, n = 2) and closed globe injury with total hyphema (2.2%, n = 1). Last year's data showed the most common indication for admission was elective surgery (78.9%, n = 1026). The monthly admission trends of study periods in both years have been illustrated in . Operating room data A total of 1341 surgeries were performed in the time span mentioned in 2019, with 73 being emergency procedures. Most common elective surgery was phacoemulsification with intraocular lens implantation (39.6%, n = 531), whereas the most common emergency procedure was open globe injury repair (71.42% of emergency surgeries, n = 52). This year's study period showed that only 25 surgeries were performed, a reduction of 98.13% from past year. The most common indication was open globe injury repair (81.25%, n = 21), others being intravitreal anti-Vascular Endothelial Growth Factor (anti-VEGF) injections (12%, n = 3) and intravitreal chemotherapy (4%, n = 1). The monthly trend of surgeries has been summarized in . Material and resource consumption Data of material indent (from the hospital central store) were collected to understand monthly resource consumption in ophthalmology in-patient and emergency ward. The use of disposable gloves saw an 11 fold rise in April 2020 from April 2019 (1100 versus 100), whereas the number of triple layered surgical masks issued in April 2020 was 2400 as compared to only 100 in April 2019, i.e., 24 fold. N95/KN95 masks, previously not being used were put into use. Personal protective gear such as head caps, disposable gloves and PPE kits consumption increased over time . Use of surface disinfectants and sterilizers has also increased with passage of time as can be seen in . Use of sodium hypochlorite solution for cleaning increased by 6 times and of Povidone iodine hand scrub by 2.5 times, this September compared to September last year. Manpower at work Due to the COVID crisis and manpower crunch, postgraduate resident doctors are posted in COVID dedicated areas, on rotation. This has resulted in a decrease in the pool of residents actively working in the Ophthalmology department for emergencies. In September 2020, 11 of 14 junior doctors (78.5%) were posted in COVID areas at some point. elucidates the number of residents each month posted in COVID dedicated areas.
A total of 1152 new patients were seen in emergency ophthalmology service from March 25 to September 30, 2020, as opposed to 1515 during the same time period last year (March 25 to September 30, 2020). A decrease of 23.9% was observed, due to the ongoing pandemic and lockdown. On triaging patients according to AIOS guidelines, 58.6% ( n = 676) were classified as emergency cases, 18.3% ( n = 211) as urgent cases and 23% ( n = 265) as routine cases. Patients requiring routine care were referred to tele-ophthalmology service. One of the earliest Indian studies that evaluated patient presentation in ophthalmic emergency during lockdown had observed similar proportions, reporting 73% as emergency or urgent, whereas 26.85% as routine care patients. The average number of patients seen in emergency per day was 5.96 in the study period as opposed to 8.02 per day in same time period last year. Average number of patients per day nosedived to 3.64 in May 2020 but rising trends were again observed as lockdown was partially lifted in June, 2020 (6.03). The trends have been rising each month since, as further restrictions were relaxed. Trends of average number of patients seen in emergency have been illustrated in . During this 6-month period, 324 patients directly accessed tele-ophthalmology services, forming only 21.95% of total patients that accessed eye care services. Average age of patients presenting to emergency service was 34 ± 7.2 years in this year's study period as opposed to 40.2 ± 5.6 years in the same time period last year. Majority of patients in 2020 study period were males (61.8%, n = 712), whereas the male predominance was lower in 2019 study period (56.1%, n = 850). Most common presenting complaint in emergency setting was eye trauma due to road traffic accidents (36.19%, n = 417), followed by dryness and allergy related symptoms (13.2%, n = 153), assault (9.2%, n = 106) and fall from height (9.4%, n = 109). Last year's data showed that eye trauma due to road traffic accidents (52.1%, n = 792) was most common presentation followed by fall from height (10.6%, n = 162) and trauma due to minor household accidents (7.1%, n = 108). A downward trend was seen in ocular trauma secondary to road traffic accident, fall from height, minor household accidental trauma, conjunctivitis cases and firecracker injury as compared to last year. Increased cases of assault, chemical injury, foreign body in eye, and dryness and allergy related problems were noted this year. A disturbing trend was noted in household violence this year (32% of all assault cases, n = 34) as compared to last year (14%, n = 14). The presenting complaints in emergency service have been summarized in .
Only 45 patients were admitted in the 6-month study period, as compared to 1300 last year in same time period. A decrease of 96.53% was noted this year. The most common indication for hospital admission in this year's study period was open globe injury, constituting 54.83% ( n = 24) of all admitted patients. Other indications were traumatic optic neuropathy needing intravenous steroids (17.7%, n = 8), retinoblastoma patients for chemotherapy (11.1%, n = 5), orbital cellulitis (11.1%, n = 5), severe fungal corneal ulcer (4.4%, n = 2) and closed globe injury with total hyphema (2.2%, n = 1). Last year's data showed the most common indication for admission was elective surgery (78.9%, n = 1026). The monthly admission trends of study periods in both years have been illustrated in .
A total of 1341 surgeries were performed in the time span mentioned in 2019, with 73 being emergency procedures. Most common elective surgery was phacoemulsification with intraocular lens implantation (39.6%, n = 531), whereas the most common emergency procedure was open globe injury repair (71.42% of emergency surgeries, n = 52). This year's study period showed that only 25 surgeries were performed, a reduction of 98.13% from past year. The most common indication was open globe injury repair (81.25%, n = 21), others being intravitreal anti-Vascular Endothelial Growth Factor (anti-VEGF) injections (12%, n = 3) and intravitreal chemotherapy (4%, n = 1). The monthly trend of surgeries has been summarized in .
Data of material indent (from the hospital central store) were collected to understand monthly resource consumption in ophthalmology in-patient and emergency ward. The use of disposable gloves saw an 11 fold rise in April 2020 from April 2019 (1100 versus 100), whereas the number of triple layered surgical masks issued in April 2020 was 2400 as compared to only 100 in April 2019, i.e., 24 fold. N95/KN95 masks, previously not being used were put into use. Personal protective gear such as head caps, disposable gloves and PPE kits consumption increased over time . Use of surface disinfectants and sterilizers has also increased with passage of time as can be seen in . Use of sodium hypochlorite solution for cleaning increased by 6 times and of Povidone iodine hand scrub by 2.5 times, this September compared to September last year.
Due to the COVID crisis and manpower crunch, postgraduate resident doctors are posted in COVID dedicated areas, on rotation. This has resulted in a decrease in the pool of residents actively working in the Ophthalmology department for emergencies. In September 2020, 11 of 14 junior doctors (78.5%) were posted in COVID areas at some point. elucidates the number of residents each month posted in COVID dedicated areas.
The pandemic and lockdown created some unique challenges both for patients and health care services. Movement restriction, social distancing norms and fear of contracting the virus from the hospital were the common problems cited by patients. The ophthalmologists also had concerns regarding the extra risk of exposure as the patient and doctor come in close proximity whereas slit-lamp examination. An online survey amongst Indian ophthalmologists during the lockdown had revealed that majority of ophthalmologists were not seeing patients and there was near-total cessation of elective surgeries. Managing the available manpower, reducing risk of exposure and triaging patients also posed significant challenges. Health care sector had to modify practices to keep up with changing trends, including introduction of telemedicine services for addressing non-emergent issues. There are previous studies evaluating the impact of lockdown on ophthalmic services in India, but mostly from private centers. As the hospital had several COVID dedicated areas, routine services could not be resumed in our institute even after lockdown, due to heavy load of COVID positive cases. Some interesting trends were noted in patients presenting to emergency services. It was noted that many did not follow the requisite precautions needed and poor compliance with face masks was commonly observed. It was also noted that some patients delayed seeking medical help due to risk of exposure to the virus and lack of transport facilities. A decrement of 23.9% in number of patients visiting emergency setting was noted in this year's study period as compared to last year. A recent study from an apex institute in North India has also reported a decrease of 35.25% in ophthalmic emergency services since lockdown when compared to same time period last year. Nearly a quarter (23%; n = 265) of total patients that visited emergency setting were triaged as requiring routine care and were advised to follow-up on tele-consultation. As many patients require non-urgent care, we must create more awareness and promote use of tele-ophthalmology services. Movement restriction from neighboring states played a major role in reduction of patient load along with behavioral changes during lockdown. As restrictions were uplifted partially in June 2020, the average number of cases rose to 6.03 per day and have further risen to 8.83 per day in September 2020, as more restrictions were removed . A reduced compliance with follow-ups and medication was seen, probably due to restricted movement to pharmacies and hospital during lockdown. A male dominance and younger age of presentation during lockdown was noted in our study. These demographics are similar to those noted in a previous study in a tertiary care institute during lockdown. The gender and age bias could possibly be due to females being busier in household with children as schools were closed and elderly (being at higher risk of contracting infection) preferred staying at home. One positive outcome of the lockdown was the substantial decrease in the number of road traffic accident patients. A decrement of 47.3% from last year was seen whereas the proportion of patients with drunk driving also came down to 30% this year from 51.4% in last year's study period. Authors from a study in an apex ophthalmic institute have also reported decrement of 41.75% in cases of mechanical trauma during lockdown. Sixty patients (5.2%) had chemical injury with 17 of them reporting accidental splash of sanitizer in their eyes. As use of sanitizers has become common and the screen time on cellphones/laptops has increased, 153 patients (13.6%) presented with complaints of dryness and itching. After the initial prescription, such patients were advised to follow-up on tele-ophthalmology service. 106 patients (9.2%) came to emergency with history of assault, and 32% ( n = 34) of these were victims of domestic violence. This disturbing increase in cases of domestic violence from 14% in last year's study period was secondary to misdirected anger and frustration due to loss of work and social interaction with family and friends. Reports have noted a 10 year high in the number of domestic violence cases during the lockdown. This increase in number of domestic violence cases emphasized the need for a mental health helpline to help people cope up with the changed environment in COVID times. We also came across two patients with pressure cooker blast injuries. Both cases presented with severe open globe injury and evisceration had to be performed. In both cases, the patients were not well acquainted with proper use of pressure cooker and were trying their hand at cooking during the lockdown. We suggest use of cookers/cookware only under supervision for the untrained to avoid such mishaps. Seventy-eight health care personnel sought ophthalmic care for mask induced dry eyes, due to prolonged wearing of masks while on duty. The continuous breathing and re-breathing could be a reason behind the same. Thirteen health workers also reported sanitizer associated chemical injury. Of 45 patients admitted to eye ward, 25 patients (55.5%) were admitted for surgical purposes; 24 requiring surgical globe repair for open globe injury and 1 needing intravitreal chemotherapy for retinoblastoma. This was in contrast to last year, when 72% ( n = 1091) were admitted for surgical purposes. Due to the close contact between surgeon and patient during surgical time, there was apprehension at both ends towards surgical management. All patients who were operated underwent COVID testing one day before surgery. Only COVID negative patients were operated in emergency OT. Two patients needing open globe repair tested positive for COVID and were operated in dedicated COVID OT, after taking all necessary precautions. Post-operatively, they were managed in COVID ward. Data for the aforementioned study period for 2019, showed that of 1237 patients that were operated, 67% ( n = 829) surgeries were operated on in-patient basis whereas rest were on outpatient basis. This year only 21 surgeries were performed in the same time period, all on in-patient basis. Of the 24 open globe injury patients, 3 refused surgery and only 21 were operated. Our hospital also has a dedicated eye bank for donor cornea collection and processing. All eye bank related processes have been suspended since the national lockdown, due to fear of COVID transmission. Although a total of 189 donor corneas were collected and 44.9% of these utilized ( n = 85) for transplantation during last year's study period, this year the collection and transplantation numbers stand at zero. Another study from an apex North Indian institute reported decrease in cornea collection by 99.61% since announcement of lockdown. Another crucial aspect that must be considered in the present scenario is the availability and management of resources. Resource availability is not only an issue for over populated, developing nations but may also become a limiting factor in even in developed countries due to surging demands. Therefore, both material and manpower resources need to be managed properly for proper functioning of health care delivery system. Our hospital is also academically oriented, apart from providing health care services. Clinical and surgical learning has been disrupted due to closure of routine services. Academic classes were also suspended for a while, before resumption on online platforms. If we look at the manpower availability, several postgraduate doctors from Ophthalmology department were posted in COVID areas each month on rotation leading to depletion from the actively functional pool available for Ophthalmology services. Residents were apprehensive due to the risk of contracting COVID, especially those living at their own homes. Four Ophthalmology residents also contracted COVID during the month of August. Rest of the residents were posted in emergency areas on rotation to ensure learning and proper manpower management. Manpower management is thus a crucial aspect in current times to minimize exposure to other residents as well as manage the workload of emergency services. Material resources are needed for personal protection as well as sterilization/disinfection purposes. In pre-COVID times, a casual approach was seen amongst many ophthalmologists but practices have become much stringent since the pandemic. In this year's study period, 4405 N95/KN95 masks were used. If we look at the ratio of triple layered surgical masks to N95 masks, in April 2020 it was 3:1, whereas it became 3:5 by September 2020. The overall use of N95/KN95 masks has increased, whereas that of surgical masks has come down. The use of other personal protective gear such as head caps and PPE kits has also increased. An interesting aspect that came up in this study was the decreased availability of Chlorhexidine hand rub in this year's study period. Overall use of Chlorhexidine hand rub in this year's study period was 169 liters versus 201 liters in same time period in 2019 (decrease of 15.9%). This is perhaps due to the huge surge in demand of hand sanitizer solutions leading to less availability. This study has highlighted how the trends are changing and the need to adapt to them. The pandemic is here to stay and our practices should be modified accordingly. Resource building and management is a crucial factor that must be looked into.
There is uncertainty regarding how long the pandemic might last and its future repercussions. The pandemic and subsequent lockdown have certainly affected patient presentation, health care delivery and the pool of resources. Understanding the changing trends will help in better preparedness, if a similar cycle were to repeat in the future. Awareness regarding tele-ophthalmology will ensure that no patient faces delay in primary treatment due to the pandemic. We must look into optimum utilization of our current resources, keeping an eye on future needs at the same time. Efficient planning is needed to ensure smooth health care delivery that is safe for the patient as well as health care personnel. Chief limitation was no data from last year pertaining to fundus examination calls and in-patient calls in other departments. Emergency patients that were seen in OPD last year and those who were directly referred to tele-ophthalmology service this year were also not considered for analysis. Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest.
Nil.
There are no conflicts of interest.
|
Spatial proteomic differences in chronic traumatic encephalopathy, Alzheimer's disease, and primary age‐related tauopathy hippocampi | bb81ed1d-c185-495f-8dad-ebd2415600cf | 11848160 | Biochemistry[mh] | BACKGROUND Dementia is a growing public health concern globally, as the number of individuals with cognitive impairment is projected to reach ≈ 150 million by 2050. , , Alzheimer disease (AD) remains the most common cause of dementia worldwide, but despite recent advances in biomarkers and therapeutics, the precise pathogenesis of this disorder remains unclear. It has also recently become evident that other neurodegenerative pathologies frequently coexist with AD neuropathologic change (ADNC), and a significant portion of the clinical symptoms that have been attributed solely to AD may be due to the additive or synergistic effects of these concomitant pathologies. , , , ADNC is characterized by the presence and extent of two interrelated protein deposits. Hyperphosphorylated tau (p‐tau) immunoreactive neurofibrillary tangles (NFTs), composed of paired helical filaments of 3R‐ and 4R‐tau isoforms, begin to accumulate in the brainstem during early adulthood , before progressing through the entorhinal region, hippocampus, and limbic system to the neocortex in well‐defined Braak NFT stages that correspond roughly with the severity of clinical symptoms. Amyloid beta (Aβ) diffuse and neuritic plaques progress from the neocortex to limbic structures, deep gray nuclei, brainstem, and cerebellum in Thal phases, initially coexisting in the same space as NFTs within the entorhinal cortex and hippocampus. Primary age‐related tauopathy (PART) is currently classified as a separate entity in which 3R/4R NFTs develop in the hippocampus and entorhinal cortex, corresponding approximately with Braak NFT stages I through IV, with minimal extension beyond the medial temporal lobe. , Unlike ADNC, these NFTs form in a Aβ‐independent manner and have early involvement of the CA2 hippocampal subfield, a region that is relatively spared in ADNC until later in the disease process. , , PART has been separated into definite PART, in cases which are completely devoid of Aβ (Thal phase 0, Consortium to Establish a Registry for Alzheimer's Disease [CERAD] neuritic plaque [NP] score “absent”) and possible PART, which have minimal Aβ deposition (Thal phase 1–2 and/or CERAD NP score “sparse”). While the density of hippocampal p‐tau deposition and presence of other comorbidities have been associated with cognitive decline, Braak staging does not appear to correlate with cognition in individuals with PART, and it is currently debated as to whether this process may represent a more “normal” aging pattern. , , , Similarly, chronic traumatic encephalopathy (CTE) is an Aβ‐independent 3R/4R‐tau immunoreactive tauopathy, although the p‐tau molecules adopt different secondary structural elements compared to ADNC and PART. CTE is a neurodegenerative consequence of exposure to repetitive head impacts, and the p‐tau distribution pattern is again different. The earliest stages involve p‐tau at the depths of sulci in the neocortex (often in the frontal cortex first) in a perivascular distribution with later spread to the hippocampus, where CTE preferentially involves the CA4 and CA2 hippocampal subregions. , , , , The apparent similarities in p‐tau composition and differences in Aβ and p‐tau distribution among ADNC, possible PART, definite PART, and CTE provide an opportunity to investigate the proteomic milieu of NFT‐ and non–NFT‐bearing neurons across hippocampal subregions. Previous studies using Nanostring GeoMx Digital Spatial Profiling (DSP) have demonstrated that the proteome of NFT‐ and non–NFT‐bearing neurons in possible PART (Thal phase 1–2) shares more in common with ADNC than with definite PART (Thal phase 0), highlighting the importance and influence of hippocampal Aβ on these processes. Spatial proteomic studies of NFTs in resilient and cognitively impaired ADNC subjects demonstrated differences in proteins associated with inflammatory response and Aβ processing, suggesting that Aβ may affect the protein composition of NFTs or may be regulated by proteins expressed in these neurons, with a subsequent impact on cognition. In this study, we investigate the proteome of NFT‐ and non–NFT‐bearing neurons in the entorhinal cortex, CA1, CA2, and CA4 hippocampal subregions in subjects with ADNC, possible PART, definite PART, and CTE to determine the relationships among proteins involved with synaptic/neuronal integrity, Aβ processing, proteostasis, autophagy, inflammation, gliosis, and oxidative stress in the context of disease state and Aβ. METHODS 2.1 Case selection A total of twenty subjects with brain tissue collected between 2016 and 2022 by the Mount Sinai Neuropathology Brain Bank and Research CoRE were evaluated. The cases were selected on the basis of neuropathologic diagnosis, pathologic disease stage, and the presence of sufficient numbers of neurons and p‐tau pathology in the entorhinal cortex and CA1, CA2, and CA4 hippocampal subregions. For all cases, hematoxylin and eosin (H&E), immunohistochemical (IHC), and Bielschowsky silver stains were re‐reviewed by two board‐certified neuropathologists (T.E.R. and J.M.W.) to ensure consistency in diagnosis and staging of ADNC, PART, and CTE (as well as other comorbidities) according to the most current diagnostic criteria. , , For each case, at least one section from frontal, temporal, parietal, and occipital neocortex, as well as bilateral hippocampi and cerebellar cortex were used to diagnose and stage ADNC, PART, and CTE. All subjects with CTE were male and had histories of American football participation, while none of the ADNC or PART cases had documented histories of head trauma. All included cases had p‐tau–positive neurofibrillary degeneration and a sufficient number of neurons in each of the evaluated hippocampal subregions (entorhinal cortex, CA1, CA2, and CA4). Other neuropathologic comorbidities, including cerebral amyloid angiopathy (CAA), cerebrovascular disease (CVD), Lewy body disease (LBD), limbic‐predominant age‐related TPD‐43 encephalopathy neuropathologic change (LATE‐NC), hippocampal sclerosis (HS), aging‐related tau astrogliopathy (ARTAG), and argyrophilic grain disease (AGD), were also reassessed for each case. The included cases were eight subjects with ADNC (five with high‐level ADNC and three with intermediate‐level ADNC), seven cases with PART (three cases with definite PART and four cases with possible PART), and five cases with CTE (all high level, stage III–IV; Table ). Of note, NFT‐ and non–NFT‐bearing neurons in the entorhinal, CA1, and CA2 subregions of all PART cases and seven of the ADNC cases were analyzed with digital spatial profiling proteomic panels in a previous publication ( https://creativecommons.org/licenses/by‐nc/4.0/ ). For the current report, CTE cases, an additional ADNC case, and an additional subregion relevant to CTE (CA4) were included. 2.2 IHC Standard histologic and IHC workups were performed for complete diagnostic evaluation on each case. Five µm thick formalin‐fixed paraffin‐embedded (FFPE) sections of the posterior hippocampus at the level of the lateral geniculate nucleus (LGN) were mounted on charged slides and baked at 70°C. H&E, Bielschowsky silver stains, p‐tau IHC (AT8; MN1020, Thermo Fisher Scientific), and Aβ IHC (6E10; SIG‐39320, Covance, Inc.) were performed on all sections using a Leica Bond III automated immunostainer (Leica Biosystems), according to the manufacturer's protocols. All slides were digitally scanned as previously described. , , RESEARCH IN CONTEXT Systematic review : The authors reviewed the literature using online journal sources. Proteomic studies involving Alzheimer's disease neuropathologic change and primary age‐related tauopathy have been conducted, but investigations into the specific hippocampal subregion differences and impact of amyloid beta (Aβ), particularly with regard to chronic traumatic encephalopathy, have not. Interpretation : These data provide a framework for understanding the interactions between phosphorylated tau (p‐tau) and Aβ at the level of the hippocampus, and shed light on mechanisms which may be involved in countering the effects of Aβ. Numerous proteins associate specifically with neurons in Aβ‐positive or Aβ‐negative pathologies, suggesting that many of the pathways investigated here influence Aβ deposition or are affected by the presence of Aβ. Future directions : Further investigation into additional brain regions affected by p‐tau and Aβ, additional proteinopathies (including 4R tauopathies), comorbid neuropathologies, and unbiased protein profiling methods will help unravel the underlying pathogenesis of these disorders and aid in development of diagnostic biomarkers and therapies. 2.3 Computer‐assisted assessment of p‐tau Hippocampal subfields (CA1, CA2, CA4, and entorhinal cortex) were segmented and annotated using Aperio ImageScope software, as previously described. , , , The CA2 hippocampal subfield was neuroanatomically defined as the region of the cornu ammonis with the most compact neuronal density. The CA1 subfield was defined as the region between the distal CA2 boundary and the hippocampal fissure. The CA4 subfield was defined as the region enclosed by, but not including, the dentate gyrus. Each subregion was selected using the Aperio ImageScope pen tool and any defects in the tissue were excluded using the Aperio ImageScope negative pen tool. P‐tau burden was determined using Aperio ImageScope positive pixel count version 9 default parameters (intensity threshold for weak positive pixels [upper limit] = 220, intensity threshold for weak positive pixels [lower limit] = 175, intensity threshold for medium positive pixels [lower limit] = 100, intensity threshold for strong positive pixels [lower limit] = 0). Positive and strong positive pixels were divided by total pixels to establish a ratio of positive pixels/total pixels (positive pixel proportion). 2.4 NanoString GeoMx DSP FFPE hippocampal sections of the posterior hippocampus at the level of the LGN were de‐paraffinized and incubated with four morphology markers (specific fluorescently labeled antibodies to aid in selecting regions of interest [ROIs]) according to the manufacturer's instructions using protocols developed for DSP assays ( https://nanostring.com/products/geomx‐digital‐spatial‐profiler/geomx‐protein‐assays/ ). The morphology markers consisted of SYTO 83 (Thermo Fisher Scientific) to label all nucleic acid, and fluorescently labeled antibodies against AT8 (p‐tau, Ser202 & Thr205; conjugated with an Alexa Fluor 594 [AF594] antibody labeling kit from Thermo Fisher Scientific), Aβ (6E10; Alexa Fluor 488 [AF488]; Thermo Fisher Scientific), and neuronal nuclei (HuD E‐1; Alexa Fluor 647; Thermo Fisher Scientific; Figure ). Each ROI consisted of a 50 µm diameter circle (1963.5 µm 2 ), which was centered on a neuronal cell body, encompassing an NFT‐bearing neuron (as well as its immediate microenvironment), a non–NFT‐bearing neuron (as well as its immediate microenvironment), or an area of neuron‐free background microenvironment, as previously described in detail. , For each case, five individual NFT‐bearing neurons, five non–NFT‐bearing neurons (AT8‐negative/“normal neurons”), and two areas of neuron‐free background microenvironment were manually selected by T.E.R. and J.M.W. for high‐resolution multiplex proteomic profiling from the CA1, CA2, and CA4 subregions and the entorhinal cortex (Figure in supporting information). For each slide, a total of 48 ROIs were analyzed. This resulted in a total of 40 NFT‐bearing neurons and 40 non–NFT‐bearing neurons from each subregion of ADNC subjects (320 neurons across all subregions and cases), 35 NFT‐bearing neurons and 35 non–NFT‐bearing neurons from each subregion of PART subjects (280 neurons across all subregions; 120 definite PART and 160 possible PART), and 25 NFT‐bearing neurons and 25 non–NFT‐bearing neurons from each subregion of CTE subjects (200 neurons across all subregions), for a total of 800 analyzed neurons. Each slide was incubated with a cocktail containing 76 total antibodies (70 protein targets and 6 included for normalization) conjugated to unique ultraviolet (UV)‐photocleavable oligonucleotide tags (Table in supporting information). Quantitative levels of each of these antibodies were analyzed in each ROI using NanoString's GeoMx Digital Spatial Profiler System (NanoString Technologies). , After incubation with the antibody cocktail, each ROI was illuminated with UV light, cleaving the oligonucleotides from the antibodies bound to their respective proteins. Oligonucleotides were collected in a 96‐well plate and hybridized to four‐color, six‐spot optical barcodes and analyzed on NanoString's nCounter platform, resulting in quantification of each antibody present in each ROI. These digital counts were normalized by a signal‐to‐noise ratio (SNR) using mouse immunoglobulin G (IgG)2a and rabbit IgG, as internal control housekeeper proteins (Histone H3, S6, GAPDH) have been shown to display altered expression in NFTs in ADNC. , All ROIs were re‐normalized from our previous analysis with the context of an additional hippocampal subregion (CA4) and an additional neuropathologic entity (CTE). The levels of proteins related to LBD (α‐synuclein and p‐α‐synuclein [S129]) and LATE‐NC (TDP‐43 and p‐TDP‐43 [S409/S410]) were specifically assessed in all neurons to minimize any potential proteomic impact of these comorbidities within the selected NFT‐ and non–NFT‐bearing neurons. 2.5 Statistics All statistical analysis and principal component analysis was performed using GraphPad Prism version 9 (GraphPad Software, Inc.). Correlations between protein quantification and pathology or Aβ were made with Pearson correlation coefficient r . Differences between protein levels were evaluated using multiple t tests. Statistical significance was set at α = 0.05. All comparisons were adjusted for multiple testing using false discovery rate (FDR) correction. Case selection A total of twenty subjects with brain tissue collected between 2016 and 2022 by the Mount Sinai Neuropathology Brain Bank and Research CoRE were evaluated. The cases were selected on the basis of neuropathologic diagnosis, pathologic disease stage, and the presence of sufficient numbers of neurons and p‐tau pathology in the entorhinal cortex and CA1, CA2, and CA4 hippocampal subregions. For all cases, hematoxylin and eosin (H&E), immunohistochemical (IHC), and Bielschowsky silver stains were re‐reviewed by two board‐certified neuropathologists (T.E.R. and J.M.W.) to ensure consistency in diagnosis and staging of ADNC, PART, and CTE (as well as other comorbidities) according to the most current diagnostic criteria. , , For each case, at least one section from frontal, temporal, parietal, and occipital neocortex, as well as bilateral hippocampi and cerebellar cortex were used to diagnose and stage ADNC, PART, and CTE. All subjects with CTE were male and had histories of American football participation, while none of the ADNC or PART cases had documented histories of head trauma. All included cases had p‐tau–positive neurofibrillary degeneration and a sufficient number of neurons in each of the evaluated hippocampal subregions (entorhinal cortex, CA1, CA2, and CA4). Other neuropathologic comorbidities, including cerebral amyloid angiopathy (CAA), cerebrovascular disease (CVD), Lewy body disease (LBD), limbic‐predominant age‐related TPD‐43 encephalopathy neuropathologic change (LATE‐NC), hippocampal sclerosis (HS), aging‐related tau astrogliopathy (ARTAG), and argyrophilic grain disease (AGD), were also reassessed for each case. The included cases were eight subjects with ADNC (five with high‐level ADNC and three with intermediate‐level ADNC), seven cases with PART (three cases with definite PART and four cases with possible PART), and five cases with CTE (all high level, stage III–IV; Table ). Of note, NFT‐ and non–NFT‐bearing neurons in the entorhinal, CA1, and CA2 subregions of all PART cases and seven of the ADNC cases were analyzed with digital spatial profiling proteomic panels in a previous publication ( https://creativecommons.org/licenses/by‐nc/4.0/ ). For the current report, CTE cases, an additional ADNC case, and an additional subregion relevant to CTE (CA4) were included. IHC Standard histologic and IHC workups were performed for complete diagnostic evaluation on each case. Five µm thick formalin‐fixed paraffin‐embedded (FFPE) sections of the posterior hippocampus at the level of the lateral geniculate nucleus (LGN) were mounted on charged slides and baked at 70°C. H&E, Bielschowsky silver stains, p‐tau IHC (AT8; MN1020, Thermo Fisher Scientific), and Aβ IHC (6E10; SIG‐39320, Covance, Inc.) were performed on all sections using a Leica Bond III automated immunostainer (Leica Biosystems), according to the manufacturer's protocols. All slides were digitally scanned as previously described. , , RESEARCH IN CONTEXT Systematic review : The authors reviewed the literature using online journal sources. Proteomic studies involving Alzheimer's disease neuropathologic change and primary age‐related tauopathy have been conducted, but investigations into the specific hippocampal subregion differences and impact of amyloid beta (Aβ), particularly with regard to chronic traumatic encephalopathy, have not. Interpretation : These data provide a framework for understanding the interactions between phosphorylated tau (p‐tau) and Aβ at the level of the hippocampus, and shed light on mechanisms which may be involved in countering the effects of Aβ. Numerous proteins associate specifically with neurons in Aβ‐positive or Aβ‐negative pathologies, suggesting that many of the pathways investigated here influence Aβ deposition or are affected by the presence of Aβ. Future directions : Further investigation into additional brain regions affected by p‐tau and Aβ, additional proteinopathies (including 4R tauopathies), comorbid neuropathologies, and unbiased protein profiling methods will help unravel the underlying pathogenesis of these disorders and aid in development of diagnostic biomarkers and therapies. Systematic review : The authors reviewed the literature using online journal sources. Proteomic studies involving Alzheimer's disease neuropathologic change and primary age‐related tauopathy have been conducted, but investigations into the specific hippocampal subregion differences and impact of amyloid beta (Aβ), particularly with regard to chronic traumatic encephalopathy, have not. Interpretation : These data provide a framework for understanding the interactions between phosphorylated tau (p‐tau) and Aβ at the level of the hippocampus, and shed light on mechanisms which may be involved in countering the effects of Aβ. Numerous proteins associate specifically with neurons in Aβ‐positive or Aβ‐negative pathologies, suggesting that many of the pathways investigated here influence Aβ deposition or are affected by the presence of Aβ. Future directions : Further investigation into additional brain regions affected by p‐tau and Aβ, additional proteinopathies (including 4R tauopathies), comorbid neuropathologies, and unbiased protein profiling methods will help unravel the underlying pathogenesis of these disorders and aid in development of diagnostic biomarkers and therapies. Computer‐assisted assessment of p‐tau Hippocampal subfields (CA1, CA2, CA4, and entorhinal cortex) were segmented and annotated using Aperio ImageScope software, as previously described. , , , The CA2 hippocampal subfield was neuroanatomically defined as the region of the cornu ammonis with the most compact neuronal density. The CA1 subfield was defined as the region between the distal CA2 boundary and the hippocampal fissure. The CA4 subfield was defined as the region enclosed by, but not including, the dentate gyrus. Each subregion was selected using the Aperio ImageScope pen tool and any defects in the tissue were excluded using the Aperio ImageScope negative pen tool. P‐tau burden was determined using Aperio ImageScope positive pixel count version 9 default parameters (intensity threshold for weak positive pixels [upper limit] = 220, intensity threshold for weak positive pixels [lower limit] = 175, intensity threshold for medium positive pixels [lower limit] = 100, intensity threshold for strong positive pixels [lower limit] = 0). Positive and strong positive pixels were divided by total pixels to establish a ratio of positive pixels/total pixels (positive pixel proportion). NanoString GeoMx DSP FFPE hippocampal sections of the posterior hippocampus at the level of the LGN were de‐paraffinized and incubated with four morphology markers (specific fluorescently labeled antibodies to aid in selecting regions of interest [ROIs]) according to the manufacturer's instructions using protocols developed for DSP assays ( https://nanostring.com/products/geomx‐digital‐spatial‐profiler/geomx‐protein‐assays/ ). The morphology markers consisted of SYTO 83 (Thermo Fisher Scientific) to label all nucleic acid, and fluorescently labeled antibodies against AT8 (p‐tau, Ser202 & Thr205; conjugated with an Alexa Fluor 594 [AF594] antibody labeling kit from Thermo Fisher Scientific), Aβ (6E10; Alexa Fluor 488 [AF488]; Thermo Fisher Scientific), and neuronal nuclei (HuD E‐1; Alexa Fluor 647; Thermo Fisher Scientific; Figure ). Each ROI consisted of a 50 µm diameter circle (1963.5 µm 2 ), which was centered on a neuronal cell body, encompassing an NFT‐bearing neuron (as well as its immediate microenvironment), a non–NFT‐bearing neuron (as well as its immediate microenvironment), or an area of neuron‐free background microenvironment, as previously described in detail. , For each case, five individual NFT‐bearing neurons, five non–NFT‐bearing neurons (AT8‐negative/“normal neurons”), and two areas of neuron‐free background microenvironment were manually selected by T.E.R. and J.M.W. for high‐resolution multiplex proteomic profiling from the CA1, CA2, and CA4 subregions and the entorhinal cortex (Figure in supporting information). For each slide, a total of 48 ROIs were analyzed. This resulted in a total of 40 NFT‐bearing neurons and 40 non–NFT‐bearing neurons from each subregion of ADNC subjects (320 neurons across all subregions and cases), 35 NFT‐bearing neurons and 35 non–NFT‐bearing neurons from each subregion of PART subjects (280 neurons across all subregions; 120 definite PART and 160 possible PART), and 25 NFT‐bearing neurons and 25 non–NFT‐bearing neurons from each subregion of CTE subjects (200 neurons across all subregions), for a total of 800 analyzed neurons. Each slide was incubated with a cocktail containing 76 total antibodies (70 protein targets and 6 included for normalization) conjugated to unique ultraviolet (UV)‐photocleavable oligonucleotide tags (Table in supporting information). Quantitative levels of each of these antibodies were analyzed in each ROI using NanoString's GeoMx Digital Spatial Profiler System (NanoString Technologies). , After incubation with the antibody cocktail, each ROI was illuminated with UV light, cleaving the oligonucleotides from the antibodies bound to their respective proteins. Oligonucleotides were collected in a 96‐well plate and hybridized to four‐color, six‐spot optical barcodes and analyzed on NanoString's nCounter platform, resulting in quantification of each antibody present in each ROI. These digital counts were normalized by a signal‐to‐noise ratio (SNR) using mouse immunoglobulin G (IgG)2a and rabbit IgG, as internal control housekeeper proteins (Histone H3, S6, GAPDH) have been shown to display altered expression in NFTs in ADNC. , All ROIs were re‐normalized from our previous analysis with the context of an additional hippocampal subregion (CA4) and an additional neuropathologic entity (CTE). The levels of proteins related to LBD (α‐synuclein and p‐α‐synuclein [S129]) and LATE‐NC (TDP‐43 and p‐TDP‐43 [S409/S410]) were specifically assessed in all neurons to minimize any potential proteomic impact of these comorbidities within the selected NFT‐ and non–NFT‐bearing neurons. Statistics All statistical analysis and principal component analysis was performed using GraphPad Prism version 9 (GraphPad Software, Inc.). Correlations between protein quantification and pathology or Aβ were made with Pearson correlation coefficient r . Differences between protein levels were evaluated using multiple t tests. Statistical significance was set at α = 0.05. All comparisons were adjusted for multiple testing using false discovery rate (FDR) correction. RESULTS 3.1 Patient demographics and general pathologic characteristics No significant differences were noted between the ages of the neuropathologically defined cohorts; the mean ages were 80.1 ± 3.6 years, 76.8 ± 3.1, 80.3 ± 2.9, and 72.4 ± 2.3 for the ADNC, possible PART, definite PART, and CTE subjects, respectively (Table ). There was a significant difference in the sex of the groups ( P = 0.0285), although this may have been driven in large part by the fact that all of the included CTE subjects were male. , No significant difference was noted between the sex of ADNC and PART subjects. The majority of cases had some degree of cognitive impairment (dementia or mild cognitive impairment [MCI]), and the majority of the cases had some form of neuropathologic comorbidity, including LATE‐NC, LBD, CVD, AGD, ARTAG, and HS, which is expected for a population in this age range. Notably, case 3 had stage 3 LATE‐NC and HS, while case 18 had diffuse neocortical LBD, stage 2 LATE‐NC, and HS (Table ). None of the PART or CTE cases displayed neuritic plaques, while all but one of the ADNC cases had moderate or frequent neuritic plaques in the neocortex by CERAD criteria. By definition, definite PART cases had an absence of Aβ plaques in all assessed brain regions (Thal 0), while the included possible PART cases all had Thal 2. CTE cases were split between those devoid of any Aβ and those with a degree of hippocampal Aβ. Notably, the ADNC cases were all Thal phase 3 or higher, so all had concurrent Aβ plaques and NFTs in the hippocampus. In terms of p‐tau distribution, all included cases had some degree of neurofibrillary degeneration in each of the assessed subregions (Figure ). Using AT8‐stained slides, we quantitatively assessed the relative burden of p‐tau across all hippocampal subregions. Similar to previous studies, , , subjects with definite PART had a significantly lower CA1 positive pixel proportion compared to ADNC, possible PART, and CTE subjects, but statistically equivalent levels of CA2 p‐tau pathology with these groups (Figure in supporting information). In addition, there was a non‐significant trend toward lower CA4 p‐tau pathology in definite and possible PART compared to both ADNC and CTE. Definite PART cases had a significantly higher CA2/CA1 ratio than ADNC, possible PART, and CTE, and CTE had a higher CA2/CA1 ratio than ADNC (Figure ). These findings are consistent with previous studies demonstrating that the CA2 subregion is affected by p‐tau–positive neurofibrillary degeneration earlier in the disease courses of PART and CTE compared to ADNC. , , , , 3.2 Proteomic differences between NFT‐bearing neurons and non–NFT‐bearing neurons Across neuropathologic diagnoses and hippocampal subregions, NFT‐bearing neurons and their immediate microenvironments contained significantly higher levels of p‐tau epitopes S214 ( P < 0.0001), S396 ( P < 0.0001), and S404 ( P = 0.0011), compared to non–NFT‐bearing neurons. NFT‐bearing neurons (and their immediate microenvironments) also had significantly elevated levels of proteins involved in autophagy pathways, including p62 ( P < 0.0001), cathepsin D (CTSD; P < 0.0001), glucocerebrosidase (GBA; P = 0.0045), and glycoprotein non‐metastatic melanoma protein B (GPNMB; P = 0.0101), as well as proteins involved in Aβ processing, including neprilysin ( P = 0.0001), and gliosis, including vimentin ( P < 0.0001). 3.3 Principal component analysis of NFT‐ and non–NFT‐bearing neurons We next assessed all proteins in the panel as a whole to determine global differences between these neuropathologically defined entities (Figure ). A principal component analysis (PCA) plot of PC1 (explaining 44.6% of the variance) and PC2 (13.4% of the variance) demonstrated separation between ADNC and CTE in NFT‐bearing neurons (Figure ), non–NFT‐bearing neurons (Figure ), and all combined neurons (Figure ). Neurons from subjects with definite PART segregated more closely with CTE while those from subjects with possible PART segregated more closely with ADNC (Figure 2B, 2D, and ), suggesting that the proteome of individual neurons and their respective microenvironments may be significantly influenced by the presence or absence of Aβ, or conversely, that the proteomic composition of these neurons and their surrounding microenvironments may be regulating the presence or absence of Aβ. Interestingly, the proteins which contributed the most to PC1 were proteins involved with (1) inflammation and microglial function, including epithelial membrane protein 1 (EMP1; specifically elevated in definite PART), Mer proto‐oncogene tyrosine kinase (MERTK; elevated in both definite PART and CTE), colony stimulating factor 1 receptor (CSF1R; specifically elevated in definite PART), and CD9 (elevated in both definite PART and CTE); (2) oxidative stress including Park7 (elevated in both definite PART and CTE); and (3) autophagy and protein degradation, including ubiquitin (elevated in both definite PART and CTE), B‐cell lymphoma 2‐associated athanogene 3 (BAG3; specifically elevated in CTE), microtubule‐associated protein 1A/1B light chain 3B (LC3B; elevated in both definite PART and CTE), and Park5 (elevated in both definite PART and CTE). In PC2, which appears to primarily separate ADNC and CTE, activated microglial markers and Aβ‐related proteins are positively correlated with ADNC, while homeostatic microglial proteins and certain astrocyte markers (including GFAP and S100B) are correlated with CTE, suggesting different glial responses between these disorders (Figure in supporting information). 3.4 Proteomic associations with Aβ Given the finding that the pathologies more closely associated with Aβ (ADNC and possible PART) appeared to cluster separately from those which are not associated with Aβ (CTE and definite PART), as well as our previous observations that the proteins within our panel of possible PART shared more in common with ADNC than definite PART, we evaluated the proteome in the context of Aβ. A large number of proteins were positively correlated with both definite PART and CTE (to the exclusion of ADNC and possible PART), including p‐tau (S404), amyloid precursor protein (APP), ApoA‐I, apolipoprotein E, insulin‐degrading enzyme (IDE), neprilysin, myelin basic protein (MBP), P2ry12, glial fibrillary acidic protein (GFAP), C4B, CD9, ionized calcium‐binding adaptor molecule 1 (IBA1), MERTK, P2RX7, S100B, TMEM119, ATG5, CTSD, heat shock cognate 71 kDa protein (HSC70), LC3B, Park5, ubiquitin, vacuolar protein sorting 35 (VPS35), and Park7, while A Disintegrin and Metalloproteinase 10 (ADAM10), CD45, and vimentin were positively correlated with both ADNC and possible PART (Figure ). Given this pattern, we grouped the ROIs as positive/negative for Aβ regardless of pathologic diagnosis, to identify which proteins were specifically positively or negatively correlated with the presence of Aβ. Notably, hippocampal Aβ was inversely correlated with numerous markers of synaptic health and neuronal integrity, including synaptophysin, neurofilament light (NfL), NeuN, and calbindin, as well as proteins involved with proteostasis and autophagy, including ATG5, HSC70, Park5, CTSD, and LC3B; proteins involved specifically with Aβ processing, including IDE, BACE1, neprilysin, ApoA‐I, and S100B; proteins involved with response to oxidative stress (Park7); and numerous markers of inflammation and microglia (Figure ). Conversely, a smaller number of proteins were positively correlated with hippocampal Aβ, including some involved with Aβ and Aβ processing (Aβ 1‐40 and ADAM10), oxidative stress (PTEN‐induced putative kinase 1 [PINK1]), inflammation (CD45 and CD68), and autophagy/proteostasis (BAG3 and transcription factor EB [TFEB]). 3.5 Difference in p‐tau epitopes among ADNC, PART, and CTE Direct comparison of NFT‐bearing neurons (and their respective immediate microenvironments) demonstrates numerous differences in p‐tau epitopes. There are elevated levels of p‐tau S214 and S396 in definite PART compared to the other neuropathologies, particularly in the CA1 and CA2 subregions, as well as elevated S404 in the CA1 subregion. CTE has higher levels of S214 and S404 in the entorhinal cortex compared to Aβ‐positive pathologies (Figure ). In contrast to NFT‐bearing neurons, there are minimal differences in p‐tau epitopes in non–NFT‐bearing neurons among neuropathologies, with CTE demonstrating elevated levels of p‐tau S214 and S404 in the entorhinal cortex compared to all other neuropathologies, and definite PART had lower p‐tau S199 compared to AD in the CA2 subregion (Figure ). 3.6 Proteomic differences among NFT‐bearing neurons in ADNC, PART, and CTE Direct comparison of NFT‐bearing neurons (and their respective immediate microenvironments; Figure ) between definite PART and CTE compared to ADNC and possible PART demonstrates similar expression patterns of proteins related to Aβ processing (IDE, neprilysin, and S100B), proteostasis and autophagy (ATG5, HSC70, Park5, and LC3B), inflammation and gliosis (IBA1, TMEM119, S100B, and GFAP), and oxidative stress (Park7), suggesting that many of these proteins may be involved in the control/regulation of Aβ deposition and distribution in these pathologies. There are differences in protein expression between definite PART and CTE, particularly notable in the CA4 subregion, however. These differences include (1) proteins involved in oxidative stress such as PINK1, which is significantly lower in CTE than other pathologies in the CA4 subregion and entorhinal cortex (Figure ); (2) proteins involved in autophagy and protein degradation, such as CTSD (Figure ), GBA, and ATG12 (Figure ); (3) proteins reflective of neuronal integrity, in particular NeuN in the CA4 subregion (Figure ); (4) proteins involved in Aβ processing, including ApoA‐I (Figure ); and (5) proteins involved with microglial function, including GPNMB (Figure ) as well as CSF1R (Figure ). 3.7 Proteomic differences among non–NFT‐bearing neurons (“normal neurons”) in ADNC, PART, and CTE Direct comparison of non–NFT‐bearing neurons (and their respective immediate microenvironments) demonstrates that the patterns of proteomic differences in normal neurons are generally similar to those identified in NFT‐bearing neurons, with the non–Aβ‐dependent pathologies (definite PART and CTE) having more similarities overall compared to the Aβ‐positive pathologies (ADNC and possible PART; Figure ). There are again notable differences between CTE and definite PART, particularly in the CA4 subregion. These differences include PINK1 (oxidative stress; Figure ), CD39 (inflammation and gliosis; Figure ), NeuN (neuronal integrity; Figure ), GPNMB (Figure ), and BACE1 (Aβ processing; Figure ). Patient demographics and general pathologic characteristics No significant differences were noted between the ages of the neuropathologically defined cohorts; the mean ages were 80.1 ± 3.6 years, 76.8 ± 3.1, 80.3 ± 2.9, and 72.4 ± 2.3 for the ADNC, possible PART, definite PART, and CTE subjects, respectively (Table ). There was a significant difference in the sex of the groups ( P = 0.0285), although this may have been driven in large part by the fact that all of the included CTE subjects were male. , No significant difference was noted between the sex of ADNC and PART subjects. The majority of cases had some degree of cognitive impairment (dementia or mild cognitive impairment [MCI]), and the majority of the cases had some form of neuropathologic comorbidity, including LATE‐NC, LBD, CVD, AGD, ARTAG, and HS, which is expected for a population in this age range. Notably, case 3 had stage 3 LATE‐NC and HS, while case 18 had diffuse neocortical LBD, stage 2 LATE‐NC, and HS (Table ). None of the PART or CTE cases displayed neuritic plaques, while all but one of the ADNC cases had moderate or frequent neuritic plaques in the neocortex by CERAD criteria. By definition, definite PART cases had an absence of Aβ plaques in all assessed brain regions (Thal 0), while the included possible PART cases all had Thal 2. CTE cases were split between those devoid of any Aβ and those with a degree of hippocampal Aβ. Notably, the ADNC cases were all Thal phase 3 or higher, so all had concurrent Aβ plaques and NFTs in the hippocampus. In terms of p‐tau distribution, all included cases had some degree of neurofibrillary degeneration in each of the assessed subregions (Figure ). Using AT8‐stained slides, we quantitatively assessed the relative burden of p‐tau across all hippocampal subregions. Similar to previous studies, , , subjects with definite PART had a significantly lower CA1 positive pixel proportion compared to ADNC, possible PART, and CTE subjects, but statistically equivalent levels of CA2 p‐tau pathology with these groups (Figure in supporting information). In addition, there was a non‐significant trend toward lower CA4 p‐tau pathology in definite and possible PART compared to both ADNC and CTE. Definite PART cases had a significantly higher CA2/CA1 ratio than ADNC, possible PART, and CTE, and CTE had a higher CA2/CA1 ratio than ADNC (Figure ). These findings are consistent with previous studies demonstrating that the CA2 subregion is affected by p‐tau–positive neurofibrillary degeneration earlier in the disease courses of PART and CTE compared to ADNC. , , , , Proteomic differences between NFT‐bearing neurons and non–NFT‐bearing neurons Across neuropathologic diagnoses and hippocampal subregions, NFT‐bearing neurons and their immediate microenvironments contained significantly higher levels of p‐tau epitopes S214 ( P < 0.0001), S396 ( P < 0.0001), and S404 ( P = 0.0011), compared to non–NFT‐bearing neurons. NFT‐bearing neurons (and their immediate microenvironments) also had significantly elevated levels of proteins involved in autophagy pathways, including p62 ( P < 0.0001), cathepsin D (CTSD; P < 0.0001), glucocerebrosidase (GBA; P = 0.0045), and glycoprotein non‐metastatic melanoma protein B (GPNMB; P = 0.0101), as well as proteins involved in Aβ processing, including neprilysin ( P = 0.0001), and gliosis, including vimentin ( P < 0.0001). Principal component analysis of NFT‐ and non–NFT‐bearing neurons We next assessed all proteins in the panel as a whole to determine global differences between these neuropathologically defined entities (Figure ). A principal component analysis (PCA) plot of PC1 (explaining 44.6% of the variance) and PC2 (13.4% of the variance) demonstrated separation between ADNC and CTE in NFT‐bearing neurons (Figure ), non–NFT‐bearing neurons (Figure ), and all combined neurons (Figure ). Neurons from subjects with definite PART segregated more closely with CTE while those from subjects with possible PART segregated more closely with ADNC (Figure 2B, 2D, and ), suggesting that the proteome of individual neurons and their respective microenvironments may be significantly influenced by the presence or absence of Aβ, or conversely, that the proteomic composition of these neurons and their surrounding microenvironments may be regulating the presence or absence of Aβ. Interestingly, the proteins which contributed the most to PC1 were proteins involved with (1) inflammation and microglial function, including epithelial membrane protein 1 (EMP1; specifically elevated in definite PART), Mer proto‐oncogene tyrosine kinase (MERTK; elevated in both definite PART and CTE), colony stimulating factor 1 receptor (CSF1R; specifically elevated in definite PART), and CD9 (elevated in both definite PART and CTE); (2) oxidative stress including Park7 (elevated in both definite PART and CTE); and (3) autophagy and protein degradation, including ubiquitin (elevated in both definite PART and CTE), B‐cell lymphoma 2‐associated athanogene 3 (BAG3; specifically elevated in CTE), microtubule‐associated protein 1A/1B light chain 3B (LC3B; elevated in both definite PART and CTE), and Park5 (elevated in both definite PART and CTE). In PC2, which appears to primarily separate ADNC and CTE, activated microglial markers and Aβ‐related proteins are positively correlated with ADNC, while homeostatic microglial proteins and certain astrocyte markers (including GFAP and S100B) are correlated with CTE, suggesting different glial responses between these disorders (Figure in supporting information). Proteomic associations with Aβ Given the finding that the pathologies more closely associated with Aβ (ADNC and possible PART) appeared to cluster separately from those which are not associated with Aβ (CTE and definite PART), as well as our previous observations that the proteins within our panel of possible PART shared more in common with ADNC than definite PART, we evaluated the proteome in the context of Aβ. A large number of proteins were positively correlated with both definite PART and CTE (to the exclusion of ADNC and possible PART), including p‐tau (S404), amyloid precursor protein (APP), ApoA‐I, apolipoprotein E, insulin‐degrading enzyme (IDE), neprilysin, myelin basic protein (MBP), P2ry12, glial fibrillary acidic protein (GFAP), C4B, CD9, ionized calcium‐binding adaptor molecule 1 (IBA1), MERTK, P2RX7, S100B, TMEM119, ATG5, CTSD, heat shock cognate 71 kDa protein (HSC70), LC3B, Park5, ubiquitin, vacuolar protein sorting 35 (VPS35), and Park7, while A Disintegrin and Metalloproteinase 10 (ADAM10), CD45, and vimentin were positively correlated with both ADNC and possible PART (Figure ). Given this pattern, we grouped the ROIs as positive/negative for Aβ regardless of pathologic diagnosis, to identify which proteins were specifically positively or negatively correlated with the presence of Aβ. Notably, hippocampal Aβ was inversely correlated with numerous markers of synaptic health and neuronal integrity, including synaptophysin, neurofilament light (NfL), NeuN, and calbindin, as well as proteins involved with proteostasis and autophagy, including ATG5, HSC70, Park5, CTSD, and LC3B; proteins involved specifically with Aβ processing, including IDE, BACE1, neprilysin, ApoA‐I, and S100B; proteins involved with response to oxidative stress (Park7); and numerous markers of inflammation and microglia (Figure ). Conversely, a smaller number of proteins were positively correlated with hippocampal Aβ, including some involved with Aβ and Aβ processing (Aβ 1‐40 and ADAM10), oxidative stress (PTEN‐induced putative kinase 1 [PINK1]), inflammation (CD45 and CD68), and autophagy/proteostasis (BAG3 and transcription factor EB [TFEB]). Difference in p‐tau epitopes among ADNC, PART, and CTE Direct comparison of NFT‐bearing neurons (and their respective immediate microenvironments) demonstrates numerous differences in p‐tau epitopes. There are elevated levels of p‐tau S214 and S396 in definite PART compared to the other neuropathologies, particularly in the CA1 and CA2 subregions, as well as elevated S404 in the CA1 subregion. CTE has higher levels of S214 and S404 in the entorhinal cortex compared to Aβ‐positive pathologies (Figure ). In contrast to NFT‐bearing neurons, there are minimal differences in p‐tau epitopes in non–NFT‐bearing neurons among neuropathologies, with CTE demonstrating elevated levels of p‐tau S214 and S404 in the entorhinal cortex compared to all other neuropathologies, and definite PART had lower p‐tau S199 compared to AD in the CA2 subregion (Figure ). Proteomic differences among NFT‐bearing neurons in ADNC, PART, and CTE Direct comparison of NFT‐bearing neurons (and their respective immediate microenvironments; Figure ) between definite PART and CTE compared to ADNC and possible PART demonstrates similar expression patterns of proteins related to Aβ processing (IDE, neprilysin, and S100B), proteostasis and autophagy (ATG5, HSC70, Park5, and LC3B), inflammation and gliosis (IBA1, TMEM119, S100B, and GFAP), and oxidative stress (Park7), suggesting that many of these proteins may be involved in the control/regulation of Aβ deposition and distribution in these pathologies. There are differences in protein expression between definite PART and CTE, particularly notable in the CA4 subregion, however. These differences include (1) proteins involved in oxidative stress such as PINK1, which is significantly lower in CTE than other pathologies in the CA4 subregion and entorhinal cortex (Figure ); (2) proteins involved in autophagy and protein degradation, such as CTSD (Figure ), GBA, and ATG12 (Figure ); (3) proteins reflective of neuronal integrity, in particular NeuN in the CA4 subregion (Figure ); (4) proteins involved in Aβ processing, including ApoA‐I (Figure ); and (5) proteins involved with microglial function, including GPNMB (Figure ) as well as CSF1R (Figure ). Proteomic differences among non–NFT‐bearing neurons (“normal neurons”) in ADNC, PART, and CTE Direct comparison of non–NFT‐bearing neurons (and their respective immediate microenvironments) demonstrates that the patterns of proteomic differences in normal neurons are generally similar to those identified in NFT‐bearing neurons, with the non–Aβ‐dependent pathologies (definite PART and CTE) having more similarities overall compared to the Aβ‐positive pathologies (ADNC and possible PART; Figure ). There are again notable differences between CTE and definite PART, particularly in the CA4 subregion. These differences include PINK1 (oxidative stress; Figure ), CD39 (inflammation and gliosis; Figure ), NeuN (neuronal integrity; Figure ), GPNMB (Figure ), and BACE1 (Aβ processing; Figure ). DISCUSSION The number of Americans over the age of 65 is expected to rise precipitously in the coming decades, resulting in a significant increase in the number of people with AD and other neurodegenerative disorders. Despite a wealth of studies investigating the clinical features, genomics, proteomics, metabolomics, inflammatory milieu, and other features to develop biomarkers and disease‐modifying therapeutics, the definitive underlying cause of many of these disorders remains unknown. Neuropathologically, the hallmarks of AD include 3R/4R‐tau immunoreactive NFTs and Aβ‐positive diffuse and neuritic plaques, as well as synapse and neuron loss. These NFTs are an interesting feature as they are shared by two other disorders, PART and CTE, which have different patterns of NFT distribution, a different sequence of affected brain regions, lack concurrent Aβ deposition, and may have significant differences in cognition and other clinical features. , , , , , , The particular similarities and differences among these three disorders provide an opportunity to study the similar feature (NFTs) in different hippocampal subregions to uncover underlying pathogenic interactions that may be associated with clinical features. Previously, we analyzed the proteome of ADNC and PART cases, which revealed striking similarities between ADNC and possible PART (NFTs, normal neurons, and their immediate microenvironments) compared to those of definite PART, suggesting the importance of Aβ in the proteomic composition of these neurons and/or the importance of these proteins in the regulation of Aβ. Herein, we investigate the expression of the same proteins in NFTs and normal neurons in five subjects with CTE compared to this previous cohort of ADNC, definite PART, and possible PART cases, to strategically use the similarities and differences in NFT and Aβ distribution in key hippocampal subregions to better understand the underlying pathogenesis and p‐tau propagation patterns of these disorders. Although there is considerable overlap among neuropathologies, PCA analysis of the proteome of both NFT‐bearing neurons and non–NFT‐bearing neurons reveals similarities between ADNC and possible PART (Aβ‐dependent diagnoses) and differences compared to CTE and definite PART (Aβ‐independent diagnoses; Figure ). Numerous proteins are simultaneously elevated in either ADNC and possible PART or CTE and definite PART (Figure ) and 35 proteins (50%) were significantly correlated (either positively or negatively) with the presence of Aβ (Figure ), suggesting the importance of Aβ to the proteomic composition of NFT‐bearing neurons, non–NFT‐bearing neurons, and their immediate microenvironments. Proteins thought to be reflective of neuronal and synaptic health (synaptophysin, NeuN, calbindin, and NfL) were all decreased in the presence of Aβ pathology. Proteins involved with amyloid processing and autophagy/protein degradation generally inversely correlated with the presence of Aβ (Figure ). ApoA‐I, a protein which has been shown to inversely correlate with cortical Aβ, , was significantly increased in definite PART in particular, while BACE1, which is involved with Aβ cleavage, , was significantly increased in CTE, and IDE, CTSD, and neprilysin, involved in Aβ degradation, , , were increased in both definite PART and CTE (Figures , ). Conversely, ADAM10, a protein involved in APP processing, was higher in the Aβ‐dependent pathologies. Similarly, the majority of proteins involved with autophagy and protein degradation processes, including ATG5, CTSD, , , , HSC70, , LC3B, and Park5 (also known as UCHL1) , are lower in the presence of Aβ, while only BAG3 and TFEB are higher. These findings suggest significantly altered protein processing (including Aβ processing) among these diseases, which may influence the Aβ deposition pattern and in turn the p‐tau distribution, or conversely, these findings may represent variably impaired or compensatory protein processing mechanisms. Proteins involved with response to oxidative stress were more variable; Park7 , was lower in subjects with Aβ, while PINK1 was elevated in subjects with ADNC and definite PART but not possible PART or CTE (and this difference was particularly pronounced in the entorhinal cortex and CA4 subregion [Figure , , and ]). Proteins involved with inflammation, gliosis, and microglial function also showed numerous differences that were largely correlated with Aβ. A few proteins, including CD45, CD68, and Ki‐67, are elevated in the presence of Aβ; however, the majority of inflammatory proteins, including P2ry12, GFAP, CD163, CD39, IBA1, S100B, and TMEM119, , , , , are elevated in Aβ‐independent pathologies, and could play a protective role. GPNMB, a protein involved in microglial function and the inflammatory response as well as autophagy and potentially Aβ clearance, , , was highest in definite PART across all subregions, consistent with the hypothesized protective effect. Other proteins involved with microglial function and autophagy, such as CSF1R, EMP1, GBA, and ATG12, were similarly selectively elevated in definite PART. Interestingly, GFAP and S100B have a very similar pattern across these diseases (Figure ), with very low relative expression in ADNC and possible PART, moderate expression in definite PART, and very high relative expression in CTE, and are two proteins heavily contributing to PC2 separating ADNC and CTE neurons (Figure and Figure ). These findings are consistent with previous observations that these proteins may be used as biomarkers of traumatic brain injury , and that S100B may specifically inhibit Aβ aggregation. There were several differences in protein expression within the CA2 and CA4 hippocampal subregions in CTE, regions which may be more susceptible to traumatic brain injury and repetitive head impacts. Altered protein expression reflective of neuronal integrity, microglial function, and autophagy within CA4 in CTE support the hypothesis that this region may be particularly vulnerable to trauma and to developing pathologies. For instance, tau pathology has been shown to preferentially occur within these regions , and TDP‐43 inclusions preferentially involve CA2, CA4, and the dentate gyrus in CTE. There are a number of limitations to the present study. Given the age range of the included subjects (Table ), we were unable to include a true age‐matched “control” cohort completely devoid of both p‐tau and Aβ pathology as almost all subjects above the age of 60 will display some degree of p‐tau pathology. , The number of cases and ROIs that were analyzed was limited by the DSP technology and budget, precluding a more thorough incorporation of brain regions, comorbid conditions, and spectrum of cognitive states to determine the effects of these proteinopathies in the context of clinical symptomology. The other main limitation was in the number of protein targets included in the DSP panel, restricting the number of potential discoveries, which could be mitigated in the future using a more unbiased protein analysis method. , , Nevertheless, while the DSP panel only included 70 protein targets, these targets were strategically selected from a wide variety of functional proteins involved in neurodegenerative disease (synaptic health, oxidative stress, autophagy, microglial markers, etc.), and so provides general coverage for a number of biologically relevant pathways of interest. Like our previous analyses of resilient/cognitively impaired ADNC and ADNC/PART, these data highlight subregion‐specific differences between diseases, as well as the interaction between the neuroimmune system, neurons, p‐tau, and Aβ. While there were significant differences in key proteins between definite PART and CTE (particularly within the CA4 hippocampal subregion), the majority of the differential protein expression could be separated based on the presence or absence of hippocampal Aβ, again suggesting that possible PART may be part of the ADNC spectrum. Definite PART, on the other hand, could be a separate entity from ADNC, , actually sharing more in common with CTE, although these proteomic similarities also raise the possibility that some of the hippocampal pathology observed in CTE (particularly in the CA2 subregion) may represent co‐morbid PART. , , , As this was an observational study in autopsy tissue, it is unclear if the observed differences in proteins related to Aβ processing, autophagy/protein degradation, inflammation, gliosis, and oxidative stress were the factors determining if (and how) Aβ gets deposited, or if these protein differences simply represent a physiologic response to Aβ deposition. In either case, these findings solidify the notion that while p‐tau and Aβ begin as apparently disparate processes in distinct brain regions, once they converge at the level of the hippocampus they are highly interrelated and influence one another's distribution in a manner that appears to cut across neuropathologic diagnostic categories. , Furthermore, these protein differences may help to define future biomarkers to differentiate different disease states and misfolded protein deposition patterns, as well as provide potential targets for future therapeutics aimed at disrupting or augmenting these pathologies. Overall, these data add to the current literature surrounding the complex biology of these neuropathologic processes and how they interrelate on a biochemical level. Conception of the work: Timothy E. Richardson, Jamie M. Walker; design of the work: Timothy E. Richardson, Miranda E. Orr, Jamie M. Walker; acquisition/analysis/interpretation of the data: Timothy E. Richardson, Miranda E. Orr, Timothy C. Orr, Susan K. Rohde, Alexander J. Ehrenberg, Victoria Flores‐Almazan, Robina Afzal, Carolina Maldonado‐Díaz, Satomi Hiya, Leyla Canbeldek, Lakshmi Shree Kulumani Mahadevan, Cheyanne Slocum, Jorge Samanamud, Kevin Clare, Nicholas Scibetta, Raquel T. Yokoda, Daniel Koenigsberg, Gabriel A. Marx, Justin Kauffman, Enna Selmanovic, Eleanor Drummond, Thomas Wisniewski, Alison M. Goate, Kevin F. Bieniek, Tiffany F. Kautz, Elena V. Daoud, Jamie M. Walker; obtained materials and cases: Timothy E. Richardson, Miranda E. Orr, Timothy C. Orr, Alexander J. Ehrenberg, Emma L. Thorn, Thomas D. Christie, Claudia De Sanctis, Adam Goldstein, Charles L. White, John F. Crary, Kurt Farre, Michael L. Alosco, Jesse Mez, Ann C. McKee, Thor D. Stein, Tiffany F. Kautz, Elena V. Daoud, Jamie M. Walker; obtained funding for the work: Timothy E. Richardson, Jamie M. Walker; drafted the work or substantially revised it: Timothy E. Richardson, Susan K. Rohde, Kurt Farre, Kevin F. Bieniek, Eleanor Drummond, Jamie M. Walker; all authors read and approved the final manuscript. Preliminary results of the data presented in this paper have been published in abstract form for the 2024 Alzheimer's Disease/Parkinson's Disease (AD/PD) conference. The authors declare that they have no competing interests, conflicts of interest, or other relevant disclosures. Author disclosures are available in the . Not applicable. Not applicable. Supporting Information Supporting Information |
Patients’ preferences for headache acute and preventive treatment | 1523d95a-aa9c-4fb8-8dcc-20b984191fae | 5630539 | Preventive Medicine[mh] | Several agents with ever-newer mechanisms of action and neurostimulation techniques are testing for acute or preventive treatment of migraine and cluster headache developing an explosive therapeutic environment . Apart from four injectable monoclonal anti-CGRP antibodies, new treatments include oral agents (CGRP antagonists, 5-HT1F agonists and mGlu5 receptor modulators) and several neurostimulation devices for both symptomatic and prophylactic treatment of migraine and cluster headache. Whether these treatments will be finally commercially available depends on the results of phase 3 clinical trials in progress . New treatments target to improve efficacy, safety, tolerance and adherence. Among other factors, adherence is related to treatment efficacy, safety, tolerance, duration and route of administration; it is very poor in migraine. Only one out of four migraineurs comply with the current available treatments for chronic migraine when a treatment is required for six months; and only one out of five migraineurs comply when the duration of the preventive treatment increases up-to one year . To improve adherence, the patients’ perspectives and preferences should be taken into account in the choice of treatment . There are three general models for decision-making regarding medical treatment: the paternalistic model, the informed model, and the shared model . The classic “paternalistic model” is one in which the physician makes medical decisions for the patient without substantial consideration of the patient’s preferences. The “informed model” involves the physician communicating information to the patient regarding treatment options, risks, and benefits. After being provided sufficient information, the patient ultimately makes an informed treatment decision based on their preferences. The “shared model” involves the physician discussing treatment options and preferences with the patient and then both parties actively participate in making a shared medical decision . Headache sufferers prefer the “shared model” approach to medical decision making in regards to the prescription of triptans , but the patients’ preferences for the preventive anti-migraine treatments have not been investigated so far. This study aimed to systematically and prospectively record patients’ preferences related to symptomatic and preventive treatments for migraine and other primary headache disorders in the context of patient-centered medicine. Since nocebo may affect patient choices, we also aimed to investigate this cofactor in our patient population by using the Q-No questionnaire . This is a project designed by the Hellenic Headache Society. Five outpatient headache centers in Athens participated in the survey. After explaining the scope of the survey, reaching an agreement and signing the associated consent form, the participants were invited to fulfill a self-administered questionnaire (maximum time 10 min). All questionnaires were collected and kept by the department nurse. The main questionnaire (questionnaire A) consisted of 11 questions . To assess the internal consistency of the questionnaire, a second questionnaire (questionnaire B) included the same 11 items but rephrased, was delivered to 10% of participants after fulfilling the first one. To test the consistency of answers (test-retest reliability) the main questionnaire A was applied once again in another proportion of participants (10%), a month later. The Q-No questionnaire was included in the questions that have been addressed to patients (four additional questions). It is a self-fulfilled questionnaire that predicts potential nocebo behavior in outpatients seeking neurological consultation. The Q-No predicts nocebo with 71.7% specificity, 67.5% sensitivity and 42.5% positive predictive value . The participants were consecutive outpatients seeking neurological consultation for their headaches. The inclusion criteria were: (i) both genders, age 18–65 years; (ii) diagnosis of any primary headache disorder according to IHC-IIIbeta ; (iii) current preventive pharmaceutical treatment for headache lasting for more than 3 months; (iv) other medical conditions and medication overuse were allowed; (v) patients should be able to understand the Greek language and signed a consent form. The participant professions were classified according to the International Standard Classification of Occupations (ISCO-08) into eleven categories: unemployed, managers, professionals, technicians and associate professionals, clerical support workers, service and sales workers, skilled agricultural, forestry and fishery workers, craft and related trades workers, plant and machine operators, and assemblers, elementary occupations and armed forces occupations. The education of participants was classified according to the International Standard Classification of (ISCDE 2011) into ten categories: less than primary, primary, lower secondary, upper secondary, post secondary non-tertiary, short-cycle tertiary, Bachelor’s or equivalent level, Master’s or equivalent level, Doctoral or equivalent level and not elsewhere classified. The ethical and the scientific committees of all five Headache Centers approved the study protocol and all patients signed a consent form. Statistics Categorical variables are expressed in frequencies and percentages. Chi-square test with continuity correction was used to assess the association between the categorical variables (nominal or ordinal). The odds ratio applied in order to measure the magnitude of association. Associations between dependent variables and independent variables were analyzed using logistic regression model. Subgroup analyses were performed for the primary headache disorder; age; sex; frequency of headaches (episodic versus chronic types); education; occupation; and nocebo. All statistical tests were two-sided and p values of 0.05 or less were considered as statistically significant. Statistical analyses were conducted using the Software IBM-SPSS (Statistical Package for the Social Sciences -Version 24). Categorical variables are expressed in frequencies and percentages. Chi-square test with continuity correction was used to assess the association between the categorical variables (nominal or ordinal). The odds ratio applied in order to measure the magnitude of association. Associations between dependent variables and independent variables were analyzed using logistic regression model. Subgroup analyses were performed for the primary headache disorder; age; sex; frequency of headaches (episodic versus chronic types); education; occupation; and nocebo. All statistical tests were two-sided and p values of 0.05 or less were considered as statistically significant. Statistical analyses were conducted using the Software IBM-SPSS (Statistical Package for the Social Sciences -Version 24). Questionnaires from 514 consecutive headache patients were collected during May and July 2016. Interestingly, no patient denied participating in the study. The descriptive demographics and the analysis of fulfilled questionnaires by primary headache disorder they were suffering from are summarized in Tables and , respectively. Forty-two participants (8.17%) re-fulfilled the questionnaire A after a month without any difference from the initial one. Forty-nine participants (9.5%) reported the same answers as well in a rephrased questionnaire. None of the participants had any previous experience with neurostimulation techniques. Most participants (80.9%) judged that the efficacy is more important than the safety or the route of administration of a symptomatic treatment for headache; the large majority (88.1%) preferred the oral than other routes of administration for the drugs; interestingly, they also preferred neurostimulation instead of any pharmaceutical treatment (67.3%). More participants (72.4%) rated that the efficacy is more important than the safety or the route of administration of a treatment for the prevention of headache disorders; they choose (53.8%) a pill once daily than other routes of drug administration (including monthly subcutaneous or intravenous injections); like for symptomatic treatment, they preferred an external device for neurostimulation (63.1%) instead of any pharmaceutical prophylactic treatment. Two hundred ninety one participants (56.6%) scored more than 15 on the Q-No questionnaire, indicating potential nocebo behaviors (Table ). Subgroup analyses Participants’ preferences for the preventive headache treatment varied by the frequency of headache attacks they were suffering from. Those they were suffered from chronic headache disorders reported more often that they preferred neurostimulation than daily pharmaceutical treatment versus those they were suffered from episodic headache disorders (OR = 1.5, 95% CI:[1.1–2.1]; p = 0.013). Among several types of primary headache disorders those participants they were suffered from chronic migraine reported more often as well that they preferred an external neurostimulation device than any pharmaceutical treatment for migraine prophylaxis versus those they were suffered from any other primary headache disorders (OR = 2.15, 95% CI:[1.4–3.4]; p < 0.01). Those participants they scored more than 15 in the Q-No questionnaire they preferred to use daily external neurostimulation than daily drug treatment (OR = 1.6, 95% CI:[1.1–2.3]; p < 0.05) for headache prevention. They also prefer to use acute neurostimulation for symptomatic headache treatment than drugs (OR = 1.7, 95%CI: [1.1–2.5], p = 0.008). Statistics did not reveal any other differences in patients’ preferences including analyses for gender, age, occupation and education (data not shown). Participants’ preferences for the preventive headache treatment varied by the frequency of headache attacks they were suffering from. Those they were suffered from chronic headache disorders reported more often that they preferred neurostimulation than daily pharmaceutical treatment versus those they were suffered from episodic headache disorders (OR = 1.5, 95% CI:[1.1–2.1]; p = 0.013). Among several types of primary headache disorders those participants they were suffered from chronic migraine reported more often as well that they preferred an external neurostimulation device than any pharmaceutical treatment for migraine prophylaxis versus those they were suffered from any other primary headache disorders (OR = 2.15, 95% CI:[1.4–3.4]; p < 0.01). Those participants they scored more than 15 in the Q-No questionnaire they preferred to use daily external neurostimulation than daily drug treatment (OR = 1.6, 95% CI:[1.1–2.3]; p < 0.05) for headache prevention. They also prefer to use acute neurostimulation for symptomatic headache treatment than drugs (OR = 1.7, 95%CI: [1.1–2.5], p = 0.008). Statistics did not reveal any other differences in patients’ preferences including analyses for gender, age, occupation and education (data not shown). In this survey the patients’ preferences for headache symptomatic and preventive treatment have been recorded. Almost four out of five headache sufferers reported that they cared for more efficacy than for the safety or route of administration of the symptomatic or preventive treatments they were taking. Although naïve to neurostimulation and to new injectable anti-migraine treatments, two out of three patients preferred to use an external neurostimulator rather a drug to treat their headaches, both acutely or prophylactically (including the injectable agents every month). More than one of two patients preferred to take a pill once a day than an injection once a month or every three months for pharmaceutical prevention, assuming that all treatments have a comparable efficacy and safety profile. The type of headache the patients were suffered from did not affect their choices with one exception: those they were suffered from chronic headaches and from chronic migraine reported more often that they preferred an external neurostimulation device for acute and prophylactic treatment. An external device was significantly more preferable among patients with potential nocebo behaviors compared to those with low risk for nocebo as well. Therefore it appears that headache patients insist to prefer and trust the traditional ways of treatments among drugs (a pill to be taken orally once a day). As someone may expect they do not want to take a pill many times per day for prophylactic treatment, nor get an injection for acute headache treatment. On the other hand, they indicated a clear preference favoring external neurostimulation, although naïve to this technique. What causes this preference remains unclear from the study data. Because it was declared that efficacy and safety are hypothetically equal between treatments it cannot be assumed that safety issues are hiding behind this choice. Yet safety stands a major issue for a chronic treatment. Nor life style reasons can explain this preference as well. A positive expectation favoring an entire novel treatment driven from outside the human body could serve as a potential explanation but further investigation towards this direction is needed. No other study has been conducted to record patient preferences related to neurostimulation in headache and pain versus traditional pharmaceutical treatments. There are only a few studies published the last decade investigating patients’ preferences for headache treatment, most of them focused on which drug category the patients may prefer . In one study that investigated patients’ preferences for migraine prevention, the patients rated efficacy as the most important aspect in preventive therapy and preferred treatment options with higher efficacy rates , like in the present study. In another prospective study, patients changed their preferences favoring a nasal formulation of zolmitriptan because of the speed of onset and the overall efficacy compared to conventional zolmitritpan tablets , again indicating that efficacy matters most for the symptomatic treatment. Therefore, in all studies performed including the present one, headache sufferers rate the efficacy as the most important aspect of the treatment. In addition, the participants of this study did not like to be treated intravenously either acutely for symptomatic treatment or repetitively for prophylaxis. The patients’ choices recorded here might predict a limited preference for the use of the novel injectable prophylactic treatments for migraine and cluster headache with monoclonal anti-bodies. The anecdotal enthusiastic participation in the clinical trials for these novel treatments (both for injectable and neurostimulation) across Europe including Greece may contradict with our results however. Two out of five participants scored more than 15 in the Q-No questionnaire indicating potential nocebo behaviors. In meta-analyses for nocebo in clinical trials, eight out of 20 patients treated with placebo experienced any adverse event. More importantly, one out of 20 patients treated with placebo withdrew treatment because of adverse events. The adverse events in placebo groups mirrored the adverse events expected of the active medication studied, confirming that pretrial suggestions induce the adverse events in placebo-treated patients. Nocebo was higher in preventive treatments than in symptomatic ones . This is the first report of real life data using the Q-No questionnaire , showing that one out of two headache sufferers are in high risk to express nocebo behaviors resulting in limited adherence. Primary headache disorders are usually treatable but due to safety and tolerability reasons, available preventive treatments have often limited success, even in the right hands . There is no doubt today that some of those headache sufferers, who will discontinue the treatment because of safety or tolerability, are powered by nocebo . Among other co-factors, patients’ negative expectation and previous unpleasant treatment experiences create negative believes for the treatment outcome and safety, generating nocebo. Physicians treating headache sufferers should acknowledge nocebo as a significant cofactor for treatment adherence and failure and plan techniques to border nocebo, such as patients’ education and close follow-up. Positive suggestions and continuous support increase patient’s compliance and decreases nocebo. Study strengths and limitations This is the first study reporting patient preferences related to headache treatment options that include external neurostimulation and monthly injectable agents that are under investigation for migraine and cluster headache treatment. A large proportion of headache sufferers display nocebo behaviors that may affect these preferences. There are several limitations however. Participants had not experienced these new treatments that are still under investigation and results of phase 3 trials are missing to better compare their efficacy and long term safety; the size population; and patient selection biases including cultural ones. Thus, the results may not be completely generalizable to other practices. Suggestions for further targeted research In future studies, subgroup analysis could be performed to determine whether prior experience with both neurostimulation and monoclonal antibodies, or the presence of depressive or anxiety symptoms impacts the congruence between patient expectations and actual practice regarding decision making at the time of a treatment prescription. This is the first study reporting patient preferences related to headache treatment options that include external neurostimulation and monthly injectable agents that are under investigation for migraine and cluster headache treatment. A large proportion of headache sufferers display nocebo behaviors that may affect these preferences. There are several limitations however. Participants had not experienced these new treatments that are still under investigation and results of phase 3 trials are missing to better compare their efficacy and long term safety; the size population; and patient selection biases including cultural ones. Thus, the results may not be completely generalizable to other practices. In future studies, subgroup analysis could be performed to determine whether prior experience with both neurostimulation and monoclonal antibodies, or the presence of depressive or anxiety symptoms impacts the congruence between patient expectations and actual practice regarding decision making at the time of a treatment prescription. Headache sufferers prefer the external neurostimulation rather than the pharmaceutical treatment for their headaches, those who suffer from chronic headaches and chronic migraine in particular. A large proportion of headache sufferers have noticed nocebo behaviors that may control their treatment choices. In the light of several novel up-coming treatments these patient preferences are important for clinicians, insurances and health policy makers. Headache sufferers prefer to use an external device to treat their headaches, both for symptomatic and preventive treatment. Regardless of the primary headache disorder they suffer from, patients prefer to use a pill once daily to prevent their headaches rather an injection once a month or every three months. Nocebo is very prevalent among headache sufferers and may affect their choices for the treatment. Headache health providers should explore personal patients’ preferences before treatment decision-making and manage potential nocebo behaviors. |
Insights into diagnostic errors in endocrinology: a prospective, case-based, international study | b5ef1a82-cb81-4f8b-9cb6-106b8fcda506 | 10709946 | Physiology[mh] | Diagnosing patients is a key competence of physicians. Establishing a correct diagnosis is the basis to select the best treatment for the patient. Nonetheless, diagnostic errors in medicine are frequent and can have serious consequences for patients and their health . An estimate from the National Academy of Medicine stated that most people will experience at least one diagnostic error in their lifetime, sometimes with severe consequences . Therefore, more efforts into understanding the nature of diagnostic errors are crucial in order to reduce their occurrence and develop effective interventions. There is consensus amongst researchers that diagnostic errors are caused by both system and cognitive factors . Cognitive factors are considered the most common factor . According to Graber et al. they account for 74% of errors and mainly occurred due to faulty synthesis, faulty data gathering, and faulty knowledge . Furthermore, errors often occur in the patient-physician encounter, including history taking and physical examination . There is not yet consensus regarding the type of cognitive errors causing misdiagnosis. While some studies have suggested that cognitive biases (short cuts in the reasoning process) are the most common , others suggest that a lack of knowledge is the more important underlying factor . Most of those studies involved a retrospective analysis of real clinical cases, which are sensitive to hindsight bias and may impact the physician´s critical assessment . Clinical Reasoning – the ability to solve clinical cases – is not a general problem-solving skill but it is case-specific. Therefore, it can be assumed that the kind of clinical encounters chosen for a study will influence the frequency and nature of diagnostic errors. One widely accepted theory explaining the cognitive processes in clinical reasoning is the dual processing theory . Cognitive processes are controlled by two systems: System I, which is intuitive, fast and automatic, and System II, which is analytical and logical. Depending on the clinical experience and the familiarity with a specific clinical case, a physician will primarily use system I (for routine cases) or system II (for more unusual cases). The endocrine field contains some common diseases, affecting millions of people each year, such as endocrine hypertension, diabetes mellitus or osteoporosis, but also very rare diseases (such as Cushing’s syndrome or pheochromocytoma), some of them potentially fatal, if misdiagnosed . While content specific endocrine knowledge is often important for a correct diagnosis, patients with endocrine diseases often first present in a general practice or in general internal medicine. It is unclear how precisely physicians in general internal medicine or general practice can diagnose endocrine more umcommon cases and whether they are able to correctly identify “red flags”. We, therefore, chose to focus on endocrine cases in this study. Specifically, we developed a mix of cases that included diseases that are known to be commonly underdiagnosed although they are quite frequent, such as primary aldosteronism and hyponatremia, and potentially life-threatening, rare diagnoses that require quick diagnosis (Cushing’s syndrome, pheochromocytoma and Addison’s disease). In order to study the reasoning process of general practitioners and general internists, we conducted a study in which participants prospectively solved endocrine cases in a virtual setting. The aim was to analyze the cognitive causes of diagnostic errors in for the participants unusual cases and to identify differences between correctly and incorrectly solved cases in the field of endocrinology. Design and participants From August 2019 until January 2020, 24 physicians practicing internal medicine or general medicine completed a total of 111 simulated online clinical cases. The cases were all endocrine, however the participants were unaware of this, they were only informed that they were internal medicine cases. The participants were chosen amidst one specific criterion, they had to be a physician practicing internal medicine, this included all subspecialties of internal medicine, as well as general medicine, excluding solely those practicing endocrinology, as it was expected that endocrinologists would make less errors, due to their better knowledge of endocrinological diseases and the goal of this study was to analyze as many diagnostic errors as possible. Apart from that all physicians practicing internal and or general medicine were included, regardless of their level of working experience, age or origin. The participants were recruited mainly through the listserv of the SIDM (society to improve diagnosing in medicine), as well as through flyers in the LMU (Ludwig MaximiIians University) hospital, as well as through directly contacting physicians from university clinics or general practitioner practices, using contact details provided on the respective websites. The participations did not receive a financial incentive. Case development The cases were all written by one author (JF) based on real patient cases. Three resident and attending physicians specialized in endocrinology reviewed the cases. In an initial pilot study there were ten cases, completed by four physicians practicing internal medicine. The responses of the pilot were excluded from the data analysis. The aim of the pilot was to test the cases and the feasibility of the study. The five cases where most errors were made in the pilot, hence the most difficult, were selected for the actual study. Two cases with frequent diseases and three cases with very rare diseases were chosen for the study (Table ). All cases are shown in the supplement. Study procedure Participants first completed a sociodemographic questionnaire. Subsequently they diagnosed the five simulated internal medicine (endocrine) clinical cases (Table ) on the online based platform CASUS . This platform (details are shown in the supplement Fig. ) enables the following of different steps of the diagnostic process. Each clinical case consisted of a patient history, a detailed physical examination and technical findings, i.e. results from laboratory and imaging, in the patient file (Table for contents). The information in the history taking and physical examination consisted of age, gender, body mass index (BMI), vital parameters (blood pressure, heart rate, respiratory rate, body temperature) pre-existing illnesses, history of alcohol and nicotine consumption, cardiovascular, abdominal, lung and lymph node examination, a neurological examination, and their general and nutritional state. They were instructed to only look at the technical findings they deemed useful or essential to finding the correct diagnosis, in order to simulate the limited resources such as technical examinations and financial means in medical practice. However, the number of technical examinations that could be seen was not restricted. The number of technical findings viewed was recorded on the platform. Participants then had to state a diagnosis for each case, including an explanation and also had to indicate their diagnostic confidence (on a scale of 1–10, where 1 was not confident at all and 10 was very confident). They were able to switch between cases as they wished and were instructed to spend – very roughly - about 30 min on all cases to simulate the scarcity of time in medical practice, hence the time expenditure per case was measured. However, the time on task was not restricted. Content and statistical analysis The content and the diagnostic steps were analyzed as described in detail in a previous study regarding diagnostic errors made by students . The CASUS platform allows for gathering data prospectively and then analyze the physician’s diagnostic skills and diagnostic process. In addition, the technical findings participants looked at and how much time they spent on each finding was monitored. This, along with the explanation as to why physicians chose a diagnosis, helped understand in which part of the diagnostic process errors occurred. The causes of diagnostic errors were ascribed to one error cause based on an already published classification, which was developed as an adaption of Graber’s diagnostic errors classification . The seven error categories are: inadequate knowledge base, inadequate diagnostic skills, faulty context generation, overestimating/underestimating, faulty triggering, misidentification and premature closure. For a detailed description on the development of these error categories see Braun et al. . More details on how the errors were assigned to a category can be found in the supplement (supplement Table ). Diagnostic explanations were qualitatively analyzed . Each diagnostic error was assigned to one category (Table ). We assigned each case only to one category by choosing the predominant error that finally caused the misdiagnosis. One investigator (JF) coded all errors. A second rater (LB) also independently coded all errors and explanations. The interrater coefficient analyzed with Cohens Kappa was 0.79. The causes of misdiagnoses were quantitatively assessed. Diagnoses were binary coded as correct or incorrect. Cases with correct and incorrect diagnosis were compared regarding the time spent on a case, number of technical findings viewed, and diagnostic confidence. Means and standard deviations were calculated to describe continuous variables. Absolute counts and percentage shares were applied for describing categorical variables. P -values of equal or less than 5% were considered significant. Statistical analysis was performed in SPSS 27. Differences between groups were tested by the Mann-Whitney-U-Test due to a lack of normal distribution. From August 2019 until January 2020, 24 physicians practicing internal medicine or general medicine completed a total of 111 simulated online clinical cases. The cases were all endocrine, however the participants were unaware of this, they were only informed that they were internal medicine cases. The participants were chosen amidst one specific criterion, they had to be a physician practicing internal medicine, this included all subspecialties of internal medicine, as well as general medicine, excluding solely those practicing endocrinology, as it was expected that endocrinologists would make less errors, due to their better knowledge of endocrinological diseases and the goal of this study was to analyze as many diagnostic errors as possible. Apart from that all physicians practicing internal and or general medicine were included, regardless of their level of working experience, age or origin. The participants were recruited mainly through the listserv of the SIDM (society to improve diagnosing in medicine), as well as through flyers in the LMU (Ludwig MaximiIians University) hospital, as well as through directly contacting physicians from university clinics or general practitioner practices, using contact details provided on the respective websites. The participations did not receive a financial incentive. The cases were all written by one author (JF) based on real patient cases. Three resident and attending physicians specialized in endocrinology reviewed the cases. In an initial pilot study there were ten cases, completed by four physicians practicing internal medicine. The responses of the pilot were excluded from the data analysis. The aim of the pilot was to test the cases and the feasibility of the study. The five cases where most errors were made in the pilot, hence the most difficult, were selected for the actual study. Two cases with frequent diseases and three cases with very rare diseases were chosen for the study (Table ). All cases are shown in the supplement. Participants first completed a sociodemographic questionnaire. Subsequently they diagnosed the five simulated internal medicine (endocrine) clinical cases (Table ) on the online based platform CASUS . This platform (details are shown in the supplement Fig. ) enables the following of different steps of the diagnostic process. Each clinical case consisted of a patient history, a detailed physical examination and technical findings, i.e. results from laboratory and imaging, in the patient file (Table for contents). The information in the history taking and physical examination consisted of age, gender, body mass index (BMI), vital parameters (blood pressure, heart rate, respiratory rate, body temperature) pre-existing illnesses, history of alcohol and nicotine consumption, cardiovascular, abdominal, lung and lymph node examination, a neurological examination, and their general and nutritional state. They were instructed to only look at the technical findings they deemed useful or essential to finding the correct diagnosis, in order to simulate the limited resources such as technical examinations and financial means in medical practice. However, the number of technical examinations that could be seen was not restricted. The number of technical findings viewed was recorded on the platform. Participants then had to state a diagnosis for each case, including an explanation and also had to indicate their diagnostic confidence (on a scale of 1–10, where 1 was not confident at all and 10 was very confident). They were able to switch between cases as they wished and were instructed to spend – very roughly - about 30 min on all cases to simulate the scarcity of time in medical practice, hence the time expenditure per case was measured. However, the time on task was not restricted. The content and the diagnostic steps were analyzed as described in detail in a previous study regarding diagnostic errors made by students . The CASUS platform allows for gathering data prospectively and then analyze the physician’s diagnostic skills and diagnostic process. In addition, the technical findings participants looked at and how much time they spent on each finding was monitored. This, along with the explanation as to why physicians chose a diagnosis, helped understand in which part of the diagnostic process errors occurred. The causes of diagnostic errors were ascribed to one error cause based on an already published classification, which was developed as an adaption of Graber’s diagnostic errors classification . The seven error categories are: inadequate knowledge base, inadequate diagnostic skills, faulty context generation, overestimating/underestimating, faulty triggering, misidentification and premature closure. For a detailed description on the development of these error categories see Braun et al. . More details on how the errors were assigned to a category can be found in the supplement (supplement Table ). Diagnostic explanations were qualitatively analyzed . Each diagnostic error was assigned to one category (Table ). We assigned each case only to one category by choosing the predominant error that finally caused the misdiagnosis. One investigator (JF) coded all errors. A second rater (LB) also independently coded all errors and explanations. The interrater coefficient analyzed with Cohens Kappa was 0.79. The causes of misdiagnoses were quantitatively assessed. Diagnoses were binary coded as correct or incorrect. Cases with correct and incorrect diagnosis were compared regarding the time spent on a case, number of technical findings viewed, and diagnostic confidence. Means and standard deviations were calculated to describe continuous variables. Absolute counts and percentage shares were applied for describing categorical variables. P -values of equal or less than 5% were considered significant. Statistical analysis was performed in SPSS 27. Differences between groups were tested by the Mann-Whitney-U-Test due to a lack of normal distribution. Participants 24 (18 male, 6 female) participants completed 111 cases in total, 9 cases were not completed. Their mean age was 45 years (SD ± 15.6). Most participants were general practitioners (36%) or working in general internal medicine (21%) whereas the remaining participants were specialized in other fields of internal medicine (Table ). Their working settings included both hospitals and practices. Results of the error analysis: frequency, nature and distribution of errors The physicians misdiagnosed 52 out of 111 times, with a total error frequency of 47%. The mean time expenditure per case was 9 min and 48% of the technical findings were viewed. The frequencies of different causes for errors is shown in Table . Overall, the most common error type in all completed cases were misidentification and faulty context generation. Amongst the five cases, case 3 (Addison’s disease) had the lowest error frequency (33%) and case 2 (ectopic Cushing’s) had the highest (73%). The leading cause of diagnostic error differed from case to case (Table ). In cases with rare diseases, a lack of knowledge was not more frequently a cause of errors compared to the cases with more frequent diseases. Misdiagnoses in different diseases The kind of misdiagnoses were evaluated to explore possible patterns (Table ). In all five cases, misdiagnoses were mostly very common diagnoses. For example, primary hyperaldosteronism was most often diagnosed as primary arterial hypertension. The misidentification with other rare diseases (pheochromocytoma) was less common. In the case of ectopic Cushing’s syndrome , physicians made the most misdiagnoses, mostly due to the fact, that they did not diagnose accurately enough: A lot of physicians diagnosed a Cushing’s syndrome, but did not classify it as an ectopic Cushing’s syndrome (but rather as Cushing’s syndrome, Cushing’s disease or pituitary tumor). When determining the error types for this case, various participants overlooked the lung mass in the thoracic X-ray, which was essential to finding the correct diagnosis. In the case SIADH , other common diagnoses were stated instead of the correct one (pregnancy, gastroenteritis). In the pheochromocytoma case, one of the clinical signs – tachycardia and atrial fibrillation – was stated as final diagnosis, although it was a symptom of the underlying disease. Correctly and incorrectly solved cases The time per case, the time spent on patient’s history or physical examination did not differ between correctly and incorrectly solved cases (Table ). Furthermore, the number of technical findings that were looked at by the physicians did not differ between the cases, but in correctly solved cases, physicians spent more time on these technical findings. The diagnostic confidence was very high both in the correctly and incorrectly solved cases (median diagnostic confidence: 8 out of 10). 24 (18 male, 6 female) participants completed 111 cases in total, 9 cases were not completed. Their mean age was 45 years (SD ± 15.6). Most participants were general practitioners (36%) or working in general internal medicine (21%) whereas the remaining participants were specialized in other fields of internal medicine (Table ). Their working settings included both hospitals and practices. The physicians misdiagnosed 52 out of 111 times, with a total error frequency of 47%. The mean time expenditure per case was 9 min and 48% of the technical findings were viewed. The frequencies of different causes for errors is shown in Table . Overall, the most common error type in all completed cases were misidentification and faulty context generation. Amongst the five cases, case 3 (Addison’s disease) had the lowest error frequency (33%) and case 2 (ectopic Cushing’s) had the highest (73%). The leading cause of diagnostic error differed from case to case (Table ). In cases with rare diseases, a lack of knowledge was not more frequently a cause of errors compared to the cases with more frequent diseases. The kind of misdiagnoses were evaluated to explore possible patterns (Table ). In all five cases, misdiagnoses were mostly very common diagnoses. For example, primary hyperaldosteronism was most often diagnosed as primary arterial hypertension. The misidentification with other rare diseases (pheochromocytoma) was less common. In the case of ectopic Cushing’s syndrome , physicians made the most misdiagnoses, mostly due to the fact, that they did not diagnose accurately enough: A lot of physicians diagnosed a Cushing’s syndrome, but did not classify it as an ectopic Cushing’s syndrome (but rather as Cushing’s syndrome, Cushing’s disease or pituitary tumor). When determining the error types for this case, various participants overlooked the lung mass in the thoracic X-ray, which was essential to finding the correct diagnosis. In the case SIADH , other common diagnoses were stated instead of the correct one (pregnancy, gastroenteritis). In the pheochromocytoma case, one of the clinical signs – tachycardia and atrial fibrillation – was stated as final diagnosis, although it was a symptom of the underlying disease. The time per case, the time spent on patient’s history or physical examination did not differ between correctly and incorrectly solved cases (Table ). Furthermore, the number of technical findings that were looked at by the physicians did not differ between the cases, but in correctly solved cases, physicians spent more time on these technical findings. The diagnostic confidence was very high both in the correctly and incorrectly solved cases (median diagnostic confidence: 8 out of 10). Causes of errors in endocrinology We were able to distinguish seven different cognitive error types. Overall, the most common error categories were misidentification, premature closure and faulty context generation. These findings are in line with previous studies : In this study, all misdiagnoses were assigned only to one category, so that a single root cause of the nature of the error was determined. It is a strength of the study that the reflections of the physicians are available. This is an insight that most studies, particularly retrospective ones, do not have. By analyzing the explanations, we could determine the cause of an error with more certainty. However, in other studies, it was also described that errors are often multifactorial . Hence it could be that certain error causes are interdependent. For instance, if physicians have a cognitive bias due to overconfidence, such as stereotyping based on certain information of the patient, this could lead to premature closure, where they do not look closely at further technical findings, as they have already come to a premature diagnosis, hence leading to over- and underestimating of certain information. Other studies suggest lack of diagnostic skills (e.g. interpretation of imaging) may even be an underlying factor of premature closure . Not all diagnostic errors in endocrinology are of the same severity. For example, in the case of a patient with ectopic Cushing’s syndrome, the most errors occurred. However, the most common misdiagnosis – Cushing’s syndrome - is not completely incorrect but just imprecise. In everyday clinical practice, it is important for general practitioners and physicians working in general internal medicine to identify the correct specialty from which the disease could originate, so that the patient can be transferred to a specialist. An endocrinologist will be able (in the majority of cases) to classify this patient correctly as a patient having ectopic Cushing’s syndrome. In clinical practice, this incorrectness will possibly not harm the patient. However, in other cases with rare diseases such as pheochromocytoma the misdiagnoses might have much severer consequences. This should be kept in mind when we analyze diagnostic errors: Not every error will harm a patient. Therefore, just the frequency of errors is not critical but the causes of the errors, the kind of misdiagnoses and the therapy decisions based on those errors. A wrong diagnosis can nevertheless result in a correct treatment as already shown . Correctly and incorrectly solved cases Interestingly, we observed very little differences regarding time on task or number of technical findings viewed between correctly and incorrectly solved cases, which suggests that there are no major differences in the reasoning process of correct and incorrect diagnosis, but it likely depends more on the knowledge of physicians. It is notable that in correctly solved cases, physicians spent more time on technical examinations such as laboratory results and imaging. Spending more time on these findings might prevent premature closure and faulty context generation. This finding is different from several previous studies where correct diagnoses were often based on a faster diagnostic process . One explanation could be that in this study the cases were endocrine and therefore outside the main clinical expertise of participants. They may have led to a closer review of the technical findings in cases where physicians were correct. An interesting finding is that the levels of confidence in the diagnosis were rather high in both correct and incorrect diagnosis. The fact that physicians have poor calibration between confidence and accuracy is in line with previous studies . However, the overall confidence levels seem very high, reflecting a strong overconfidence of physicians in this study. Also, a low learning motivation can be associated with overconfidence as recently shown , which should be kept in minding in previous studies. Strengths and limitations Strengths of this study are the prospective design and the comprehensive analysis of the causes of errors enabled by the study platform CASUS. Moreover, as many of the participants were general practitioners, who patients often initially consult, this study simulates the primary care situation. Also, we had a multicentric, international approach, which is another advantage of the study. The limitations of the study include the limited sample size, as a sample of 24 physicians may not accurately represent the broader population of healthcare professionals. It has to be considered that our sample size may not accurately represent the broader population of healthcare professionals. The selection of the participants might be influenced by availability or interest. However, as quite a large number of cases was analyzed and we were able to draw valid conclusions of that which might be addressed in upcoming larger studies. However, we qualitatively analyzed 111 cases, which is quite extensive. Furthermore, in this study, the CASUS platform, although realistic, does present an artificial setting. Therefore, we cannot be sure that the findings apply to clinical practice. It is a constraint of the study, whether the findings can truly be seen to represent cognitive processes in genuine clinical encounters. For example, participants could not profit from discussions with colleagues, which could help find the correct diagnosis . Additionally, in everyday clinical practice, errors can be multifactorial and more unpredictable as more variables influence the outcome. However, even in everyday practice there are errors only caused by cognitive factors (see Graber et al.) and at least for those instances, the results of the study seem applicable. Additionally, we focused on difficult endocrinological cases, therefore, our results regarding the causes of errors might not be applicable for more common and less difficult endocrinological encounters. However, the cases chosen for a study will always influence the results regarding diagnostic errors as clinical reasoning is indeed case-specific. Therefore, this is a general limitation not only affecting our study. A further limitation is that we may not capture the full spectrum of endocrinological cases encountered in clinical practice. Also, there is the possibility of certain participants being familiar with the casus interface, and hence allowing them to more easily navigate through the findings, however the instructions on how to navigate through the findings are easily comprehendible. Outlook and conclusion To the best of our knowledge, errors in endocrinology made by physicians were analyzed for the first time in a controlled setting in this prospective study. Predominant causes for errors in this specialty include misidentification, faulty context generation, premature closure and a lack of knowledge which is in line with previous findings on causes of diagnostic error. The process of correctly diagnosed cases did not differ much from the incorrectly diagnosed cases. This suggests that it is not the reasoning process that is different in cases with and without errors, but that it is the content specific knowledge that plays the most important role. Therefore, practice with a large variety of clinical cases (including endocrine cases) in continuing medical education seems advantageous for error reduction. We were able to distinguish seven different cognitive error types. Overall, the most common error categories were misidentification, premature closure and faulty context generation. These findings are in line with previous studies : In this study, all misdiagnoses were assigned only to one category, so that a single root cause of the nature of the error was determined. It is a strength of the study that the reflections of the physicians are available. This is an insight that most studies, particularly retrospective ones, do not have. By analyzing the explanations, we could determine the cause of an error with more certainty. However, in other studies, it was also described that errors are often multifactorial . Hence it could be that certain error causes are interdependent. For instance, if physicians have a cognitive bias due to overconfidence, such as stereotyping based on certain information of the patient, this could lead to premature closure, where they do not look closely at further technical findings, as they have already come to a premature diagnosis, hence leading to over- and underestimating of certain information. Other studies suggest lack of diagnostic skills (e.g. interpretation of imaging) may even be an underlying factor of premature closure . Not all diagnostic errors in endocrinology are of the same severity. For example, in the case of a patient with ectopic Cushing’s syndrome, the most errors occurred. However, the most common misdiagnosis – Cushing’s syndrome - is not completely incorrect but just imprecise. In everyday clinical practice, it is important for general practitioners and physicians working in general internal medicine to identify the correct specialty from which the disease could originate, so that the patient can be transferred to a specialist. An endocrinologist will be able (in the majority of cases) to classify this patient correctly as a patient having ectopic Cushing’s syndrome. In clinical practice, this incorrectness will possibly not harm the patient. However, in other cases with rare diseases such as pheochromocytoma the misdiagnoses might have much severer consequences. This should be kept in mind when we analyze diagnostic errors: Not every error will harm a patient. Therefore, just the frequency of errors is not critical but the causes of the errors, the kind of misdiagnoses and the therapy decisions based on those errors. A wrong diagnosis can nevertheless result in a correct treatment as already shown . Interestingly, we observed very little differences regarding time on task or number of technical findings viewed between correctly and incorrectly solved cases, which suggests that there are no major differences in the reasoning process of correct and incorrect diagnosis, but it likely depends more on the knowledge of physicians. It is notable that in correctly solved cases, physicians spent more time on technical examinations such as laboratory results and imaging. Spending more time on these findings might prevent premature closure and faulty context generation. This finding is different from several previous studies where correct diagnoses were often based on a faster diagnostic process . One explanation could be that in this study the cases were endocrine and therefore outside the main clinical expertise of participants. They may have led to a closer review of the technical findings in cases where physicians were correct. An interesting finding is that the levels of confidence in the diagnosis were rather high in both correct and incorrect diagnosis. The fact that physicians have poor calibration between confidence and accuracy is in line with previous studies . However, the overall confidence levels seem very high, reflecting a strong overconfidence of physicians in this study. Also, a low learning motivation can be associated with overconfidence as recently shown , which should be kept in minding in previous studies. Strengths of this study are the prospective design and the comprehensive analysis of the causes of errors enabled by the study platform CASUS. Moreover, as many of the participants were general practitioners, who patients often initially consult, this study simulates the primary care situation. Also, we had a multicentric, international approach, which is another advantage of the study. The limitations of the study include the limited sample size, as a sample of 24 physicians may not accurately represent the broader population of healthcare professionals. It has to be considered that our sample size may not accurately represent the broader population of healthcare professionals. The selection of the participants might be influenced by availability or interest. However, as quite a large number of cases was analyzed and we were able to draw valid conclusions of that which might be addressed in upcoming larger studies. However, we qualitatively analyzed 111 cases, which is quite extensive. Furthermore, in this study, the CASUS platform, although realistic, does present an artificial setting. Therefore, we cannot be sure that the findings apply to clinical practice. It is a constraint of the study, whether the findings can truly be seen to represent cognitive processes in genuine clinical encounters. For example, participants could not profit from discussions with colleagues, which could help find the correct diagnosis . Additionally, in everyday clinical practice, errors can be multifactorial and more unpredictable as more variables influence the outcome. However, even in everyday practice there are errors only caused by cognitive factors (see Graber et al.) and at least for those instances, the results of the study seem applicable. Additionally, we focused on difficult endocrinological cases, therefore, our results regarding the causes of errors might not be applicable for more common and less difficult endocrinological encounters. However, the cases chosen for a study will always influence the results regarding diagnostic errors as clinical reasoning is indeed case-specific. Therefore, this is a general limitation not only affecting our study. A further limitation is that we may not capture the full spectrum of endocrinological cases encountered in clinical practice. Also, there is the possibility of certain participants being familiar with the casus interface, and hence allowing them to more easily navigate through the findings, however the instructions on how to navigate through the findings are easily comprehendible. To the best of our knowledge, errors in endocrinology made by physicians were analyzed for the first time in a controlled setting in this prospective study. Predominant causes for errors in this specialty include misidentification, faulty context generation, premature closure and a lack of knowledge which is in line with previous findings on causes of diagnostic error. The process of correctly diagnosed cases did not differ much from the incorrectly diagnosed cases. This suggests that it is not the reasoning process that is different in cases with and without errors, but that it is the content specific knowledge that plays the most important role. Therefore, practice with a large variety of clinical cases (including endocrine cases) in continuing medical education seems advantageous for error reduction. Below is the link to the electronic supplementary material. Supplementary Material 1: Supplement Figure 1: The CASUS surface. Supplement Table 1: The different error categories with explanation. Supplement: Cases Case 1: Morbus Conn. Case 2: Ectopic Cushing’s syndrome(paraneoplastic due to small cell lung cancer). Case 3: Morbus Addison. Case 4: SIADH (caused by medication, due to citalopram). Case 5: Pheochromocytoma |
Obstetrician–gynecologists’ perspectives towards medication use during pregnancy: A cross-sectional study | 2f006442-a761-4195-b279-ef496fa8aedf | 9678598 | Gynaecology[mh] | Pregnant women undergo unique physiological changes that may affect the pharmacokinetic properties of various medications. Around 40% of pregnant women uses either over-the-counter (OTC) or prescribed medications during their pregnancy to treat chronic or acute conditions, such as nausea, vomiting, diabetes, asthma, and hypertension. Pharmacological agents contribute to significant, preventable congenital abnormalities, leading to a rise in public health concerns about using medications during pregnancy. To produce such an effect, the medication must possess certain properties that allow it to cross the placenta, including but not limited to being unbound, weak base, lipid-soluble, and having a low molecular weight. Also, the fetus’s stage of development is a crucial point to consider when using medication during pregnancy. Most pregnant women know that medication use during pregnancy is paramount, which leads them to seek medical advice before taking any medication. A vast majority of studies evaluated pregnant women’s knowledge and attitudes towards using medicines during their pregnancy. One of which was conducted in Saudi Arabia in 2014, which concluded that women claim to receive inadequate medication-related information from physicians and pharmacists; instead, they rely on medication leaflets to attain such information. Obstetrician–gynecologists are frequently faced with inadequate and imprecise information to make decisions for clinical management. Although some medications’ teratogenicity potential is well known, there is limited information on the safety of many other medications used during pregnancy due to ethical considerations. Pregnant and lactating women are typically excluded from clinical trials. A study published in 2010 in the United States examined Obstetrician–gynecologists’ knowledge and informational resources regarding the safety of medication use during pregnancy. Results showed that the number of years in practice was associated with their response choice to medication safety questions. Most responders indicated sufficient access to helpful information regarding medication teratogenicity potential. However, more than half of the participants selected the lack of a single comprehensive source of information as the most significant barrier. Another study evaluating community pharmacists’ knowledge about medication safety during pregnancy in Saudi Arabia found a significant difference between age groups and country of graduation in knowledge test scores. To the best of our knowledge, no studies were conducted to spot the knowledge of Obstetrician–gynecologists in Saudi Arabia and their access to information about the risks of medication use during pregnancy. Such a study is highly warranted due to the physicians’ knowledge and practice’s effect on the patients’ health. For that, this study aims to assess Obstetrician–gynecologists’ knowledge of the medication teratogenicity potential, their frequently used resources, and their residency training contribution to medication use during pregnancy. The present study is a cross-sectional, survey-based study targeting licensed obstetrician-gynecologists practising in Saudi Arabia. Saudi and non-Saudi practitioners were eligible to fill out the questionnaire. Over 6 months, data were collected using a validated self-administered web-based questionnaire developed by the American College of Obstetricians and Gynecologists. The questionnaire is organized into 5 domains. The first domain (7 items) includes the participants’ demographic data. The second domain focused on assessing the knowledge about prescription medications, OTC, dietary supplements, and herbal products in the first trimester (23 items). The third domain was about the references used to obtain appropriate and updated information on medication use during pregnancy (15 items). The fourth domain was to demonstrate the physician’s attitudes toward medication use during pregnancy (6 items). The last domain was regarding the rating of the participant’s training in medication use during pregnancy (6 items). The questions utilized in the questionnaire included multiple choice, check all that apply, Likert-like scale, and fill-in-the-blank questions. With almost 350 clinicians registered as Obstetrician–gynaecologist specialists or consultants in Saudi Arabia, the sample was calculated to be 184 with a 95% confidence interval and 5% confidence level) as follow: S S = [ Z 2 p ( 1 − p ) ] / C 2 = [ ( 1.96 ) 2 × 0.5 ( 1 − 0.5 ) ] / ( 0.05 ) 2 = 384.16 S S / [ 1 + { ( S S 1 ) / P o p } ] = 384.16 / [ 1 + { ( 384.161 ) / 350 } ] = 184 King Saud University Medical City’s Institutional Review Board approved this study (19/0929). Following ethical approval, an online survey was sent to the department of Obstetrics & Gynecology in 6 large hospitals around the Kingdom to be distributed among their employees. Reminders were sent to non-responders, and visits were conducted to some sites with low response rates. Data were analyzed using SPSS version 25. Categorical variables were presented as numbers and percentages, while continuous variables were presented as mean and SD if normally distributed. However, if not normally distributed, median and IQR were used. Shapiro–Wilk test was used to assess for normal distribution. Analyses were tested for significance using an α of 0.05. A total of 60 obstetrician–gynecologists, completed the survey, with a response rate of 33%. The flowchart for the inclusion and exclusion process is shown in Figure . Most participants were female (72%), with a median age of 42. The median years of practice among the participants were 13 years. Around 40% were full-time hospital practitioners, and most (85%) were working in the central region (i.e., Riyadh). Seventy per cent of the participants reported providing routine care/gynecologic exams. Characteristics of participants included in the study are presented in Figure and Supplemental Digital Content (Appendix 1, http://links.lww.com/MD/H763 ). 3.1. Assessment of medication use during the first trimester of pregnancy Participants’ assessment of 23 selected medications regarding fetus safety if taken during the first trimester is presented in Supplemental Digital Content (Appendix 2, http://links.lww.com/MD/H764 ). Regarding prescription medications (Fig. ), the majority (87%) agreed that Isotretinoin is contraindicated. However, 8.3% of them were not sure. For Alprazolam, 25% considered it unsafe, 35% indicated that it required a risk-benefit assessment, and 30% were unsure. Most participants (76.7%) consider acetaminophen safe to use. Regarding dietary supplements (Fig. ), 75% stated that vitamin A supplements are not safe during the first trimester. Around 2-thirds (60%) of respondents were unsure about the safety of herbal remedies during pregnancy. 3.2. Information resources utilized by obstetrician-gynecologists Regarding the information resources used to answer questions, online databases (e.g., Lexi and Micromedex) were chosen as the top resources utilized by obstetrician-gynecologists to obtain information about the teratogenicity of medications (45%), followed by pharmacist consultation, FDA label, and colleagues’ conversation (21.7%). Further information is provided in Table . 3.3. Obstetrician–gynecologists’ attitudes toward medication use during pregnancy A Likert-Like scale was used to assess the proportion of obstetrician-gynecologists agreeing or disagreeing with various statements related to the information on the use of medications during pregnancy. Forty-eight per cent strongly agreed that liability is a concern if there were to be an adverse pregnancy outcome following the use of medications. Additionally, 41% agreed on the lack of sufficient information about the safety of medication use during pregnancy, while 31% reported a lack of accessibility to the available information. Interestingly, 26.7% reported a lack of time to communicate the information available to patients as one of the drawbacks. Additional details are provided in Table . 3.4. Obstetrician–gynecologists’ rating of their training Participants were asked to rate their training on medication use during pregnancy, and the results are presented in Table . Those who had been in practice for more than 15 years were significantly more likely to rate themselves as well qualified ( P -value < 0.05). The majority adequately and significantly rated their training on prescribed medications (58.3%), OTC medications (45%) and dietary supplements or herbal remedies (32%) ( P value < .05). Participants’ assessment of 23 selected medications regarding fetus safety if taken during the first trimester is presented in Supplemental Digital Content (Appendix 2, http://links.lww.com/MD/H764 ). Regarding prescription medications (Fig. ), the majority (87%) agreed that Isotretinoin is contraindicated. However, 8.3% of them were not sure. For Alprazolam, 25% considered it unsafe, 35% indicated that it required a risk-benefit assessment, and 30% were unsure. Most participants (76.7%) consider acetaminophen safe to use. Regarding dietary supplements (Fig. ), 75% stated that vitamin A supplements are not safe during the first trimester. Around 2-thirds (60%) of respondents were unsure about the safety of herbal remedies during pregnancy. Regarding the information resources used to answer questions, online databases (e.g., Lexi and Micromedex) were chosen as the top resources utilized by obstetrician-gynecologists to obtain information about the teratogenicity of medications (45%), followed by pharmacist consultation, FDA label, and colleagues’ conversation (21.7%). Further information is provided in Table . A Likert-Like scale was used to assess the proportion of obstetrician-gynecologists agreeing or disagreeing with various statements related to the information on the use of medications during pregnancy. Forty-eight per cent strongly agreed that liability is a concern if there were to be an adverse pregnancy outcome following the use of medications. Additionally, 41% agreed on the lack of sufficient information about the safety of medication use during pregnancy, while 31% reported a lack of accessibility to the available information. Interestingly, 26.7% reported a lack of time to communicate the information available to patients as one of the drawbacks. Additional details are provided in Table . Participants were asked to rate their training on medication use during pregnancy, and the results are presented in Table . Those who had been in practice for more than 15 years were significantly more likely to rate themselves as well qualified ( P -value < 0.05). The majority adequately and significantly rated their training on prescribed medications (58.3%), OTC medications (45%) and dietary supplements or herbal remedies (32%) ( P value < .05). To our knowledge, this is the first study in the nation that assesses Obstetrician–gynecologists’ knowledge of medications’ teratogenicity potential as well as the impact of their residency training on their decisions. The resources routinely used were also assessed. For a medication to be desirable, it must fulfill the following criteria: safe, effective, and indicated. During pregnancy, women should refrain from taking medications as much as possible due to the teratogenicity risk. However, certain medical conditions require urgent or ongoing treatment, and deciding to use them is not without apprehension. Thus, obstetrician-gynecologists play a vital role in identifying when medications are warranted and which are safe to be given during each trimester, in addition to adequately counseling patients. To assist in decision-making, the Food and Drug Administration (FDA) formerly stratified the medications’ teratogenic effects into 5 categories (i.e., A, B, C, D, and X), possessing fewer safety profiles when moving downwards. However, it is challenging to assess the risk-benefit ratio using this classification. In 2015, the FDA updated their pregnancy and lactation rule to overcome this issue. Nevertheless, even with the new FDA stratification, it is extremely challenging for physicians to make treatment decisions in this population. That is due to the diversity in fetal damage manifested in the same medication when taken at different trimesters, and the exclusion of pregnant women from clinical trials due to ethical considerations, leaving great uncertainty. Therefore, safety information is commonly obtained from other sources such as animal experiments, nonclinical data, case reports, and epidemiological data, of which possess abundant limitations, adding to the ambiguity of treatment decisions in this population. In this study, the participant’s level of knowledge regarding medication teratogenicity potential was assessed and revealed a great variation. Most respondents reported inaccessibility to current information about medication teratogenicity risk and a lack of sufficient data, emphasizing the need for updated, accessible references to aid clinical decisions. A multidisciplinary team including clinical pharmacists in the services of Obstetrics and Gynecology as medication specialists would be of great benefit. Clinical pharmacists’ contributions to the field were reported in the literature, highlighting their role in preventing the incidence of toxicity and death. Their expertise allows them to help select appropriate medications and adequately counsel patients regarding the safety of different treatment modalities, dietary supplements, and herbals. That was supported by previous evidence, where they found clinical pharmacy services in Obstetrics and Gynaecology were associated with a high level of physician satisfaction and better patient care. When assessing participants’ knowledge about the safety of medications in the first trimester, the vast majority reported that Isotretinoin is contraindicated and acetaminophen is safe, which is consistent with the published literature. On the contrary, results varied with Alprazolam. That may be attributed to the weak level of evidence and lack of consensus on its effect on the fetus. Nevertheless, since Alprazolam falls into Category D and may be detrimental to the fetus, prospective studies with a large sample size to assess its effect may be difficult to conduct. Moreover, 75% of responders stated that Vitamin A dietary supplements are not safe in the first trimester, which is far higher than a study conducted amongst community pharmacists, in which 48.4% reported it unsafe. As for the safety of herbals, participants showed a lack of sufficient knowledge of their use in this patient population. This uncertainty is alarming as the use of herbal medicine prevalence in pregnant women in the Middle East ranges from 7% to 55%. These medications may harm the mother and child; thus, healthcare practitioners’ education is essential in this regard as it also contributes to proper patient education. Several limitations exist in our study. The response rate remained low despite many reminders and visits to our participants. That may be justified by the Obstetrician–gynecologists’ high-load nature of practice and busy service, hindering the data collection process. In addition, most responders were from the central region, affecting the results’ generalizability. Since the study used self-administrated questionnaires, desirability bias may arise. It is also important to note that there was no way of determining whether or not responders used their actual knowledge or used reference sources when filling out the questionnaire. A nationwide, paper-based study is recommended to overcome the limitations mentioned above and confirm the results of this study. Our study found that Obstetrician–gynecologists vary in their knowledge about medication and herbal remedies’ teratogencity risk. These findings highlighted the need to emphasize this during their training year and the importance of having this information readily available to health care providers in an updated form. This work was supported by the College of Prince Sultan Bin Abdulaziz for Emergency Medical Services Research Center, Deanship of Scientific Research, King Saud University, Riyadh, Saudi Arabia. Conceptualization: Mashael Alshebly and Sultan Alghadeer. Data curation: Bana Almadi. Formal analysis: Abdullah M. Mubarak. Funding acquisition: Sultan Alghadeer. Investigation: Haya Alturki and Jeelan Alghaith. Methodology: Sultan Alghadeer. Supervision: Mashael Alshebly and Sultan Alghadeer. Validation: Mashael Alshebly and Abdulrahman Alwhaibi. Visualization: Mashael Alshebly and Abdullah M. Mubarak. Writing – original draft: Haya Alturki and Jeelan Algaith. Writing – review and editing: Bana Almadi and Abdulrahman Alwhaibi. |
Expert consensus on apical microsurgery | 0fe77c35-2e25-4c36-9507-e95bbde0ab3b | 11693765 | Dentistry[mh] | Root canal therapy is currently the most common and effective method for treating periapical diseases, with a success rate of more than 80%, while the success rate of root canal retreatment can reach 50–80%. Developments in technology, materials, and equipment related to root canal therapy, especially the introduction of dental operative microscopes, have aided in the increase in treatment success rates. However, due to the complexity of the root canal system, the formation of extraradicular bacterial biofilms, and the occurrence of true cysts, some periapical diseases still cannot be cured. In such cases, combined surgical treatment, i.e., endodontic surgery, is needed. – Endodontic microsurgery was developed in the 1990s with the application of a dental operative microscope. The magnification and illumination provided by the microscope allow endodontic surgery to be performed using microscopic instruments, ultrasonic tips, and bioactive ceramic materials. , – Three main types of endodontic microsurgery are currently performed: apical microsurgery, periradicular microsurgery, and microscopic intentional replantation. The apical microsurgery is a surgical procedure on the root apex, including osteotomy, root-end resection, root-end preparation, and filling under the microscope. For the cases where apical microsurgery is not feasible, microscopic intentional replantation is indicated, i.e., insertion of a tooth into its alveolus after the tooth has been extracted for the purpose of performing treatment under a microscope, such as root-end filling(s) or perforation repair. Periradicular microsurgery, including root amputation and hemisection, is a surgical procedure for the removal of a root or root of a tooth. , , Compared with traditional apical surgery, apical microsurgery has clear advantages, such as precise identification of root apices, small osteotomy, shallow bell angle of root-end resection, clear exploration of the resected root surface, and accurate root-end preparation. Apical microsurgery is accurate and minimally invasive, produces few complications, and has a success rate of more than 90%. , – However, due to the lack of awareness and understanding of apical microsurgery by dental general practitioners and even endodontists, many clinical problems remain to be overcome, such as the blind expansion of indications, the nonstandardized nature of the operations, the presence of serious complications, and low efficacy. A search of the literature revealed no relevant studies in Chinese or English, including expert consensuses, guidelines, or specifications, related to apical microsurgery. Neither foreign nor domestic endodontic organizations, such as the American Association of Endodontists (AAE), the European Society of Endodontology (ESE), and the Society of Cariology and Endodontology of Chinese Stomatological Association, have issued expert consensuses, guidelines, or specifications related to apical microsurgery. To standardize the clinical application of apical microsurgery, the Society of Cariology and Endodontology, Chinese Stomatological Association, has gathered well-known domestic experts, who major in endodontics, periodontics, or oral surgery, to hold a series of special discussions, on the basis of extensive investigations of the research results and clinical experience at home and abroad, we proposed this paper after repeated discussion. The expert consensus aims to guide the orderly, reasonable, and correct clinical implementation of apical microsurgery to improve the level and efficacy of periapical disease treatment and to better preserve natural teeth. This document specifies the indications, contraindications, preoperative preparations, operational procedures, complication prevention measures, and efficacy evaluation of apical microsurgery and is applicable to dentists who perform apical microsurgery after systematic training.
The indications for apical microsurgery include the following: (1) Teeth that still have symptoms and/or positive signs after root canal treatment and retreatment; (2) Inability to gain the coronal access to implement root canal treatment and/or retreatment of the diseased teeth with the presence of symptoms and/or positive signs. , , –
Systemic conditions Patients with systemic diseases or risks should consult corresponding specialists to determine the feasibility of apical microsurgery and the corresponding precautions. , , , , , Uncontrolled hypertension, coronary heart disease, and other cardiovascular and cerebrovascular diseases. Elevated risks of secondary infection: infective endocarditis caused by organic heart disease or a state of immunosuppression due to malignant tumors, organ transplantation, or uncontrolled diabetes. Elevated bleeding risk: abnormal coagulation function caused by hemophilia, thrombocytopenic purpura, or other diseases. Existing risk of osteonecrosis of the jaw: previous radiotherapy or injection with intravenous or oral bisphosphonates. Other conditions making the patient unsuitable for surgery, including pregnancy and an inability to cooperate with surgery due to age or mental status. Local conditions If a patient has the following local conditions, the surgeon should carefully evaluate the feasibility of apical microsurgery. , , , – . Diseased tooth in the acute inflammatory stage. . Proximity of the root apex of the diseased tooth to important anatomical structures, such as blood vessels and nerves. . Difficult lip retraction and obstruction by soft tissues and hard tissues limit the surgical approach. . Poor oral hygiene and insufficient periodontal support. . A crown-to-root ratio greater than 1:1 after root end resection or further grinding due to vertical root fracture or external root resorption.
Patients with systemic diseases or risks should consult corresponding specialists to determine the feasibility of apical microsurgery and the corresponding precautions. , , , , , Uncontrolled hypertension, coronary heart disease, and other cardiovascular and cerebrovascular diseases. Elevated risks of secondary infection: infective endocarditis caused by organic heart disease or a state of immunosuppression due to malignant tumors, organ transplantation, or uncontrolled diabetes. Elevated bleeding risk: abnormal coagulation function caused by hemophilia, thrombocytopenic purpura, or other diseases. Existing risk of osteonecrosis of the jaw: previous radiotherapy or injection with intravenous or oral bisphosphonates. Other conditions making the patient unsuitable for surgery, including pregnancy and an inability to cooperate with surgery due to age or mental status.
If a patient has the following local conditions, the surgeon should carefully evaluate the feasibility of apical microsurgery. , , , – . Diseased tooth in the acute inflammatory stage. . Proximity of the root apex of the diseased tooth to important anatomical structures, such as blood vessels and nerves. . Difficult lip retraction and obstruction by soft tissues and hard tissues limit the surgical approach. . Poor oral hygiene and insufficient periodontal support. . A crown-to-root ratio greater than 1:1 after root end resection or further grinding due to vertical root fracture or external root resorption.
History and preoperative examination Systemic conditions. The patient’s past medical history, medication history, and allergy history, especially the history of anesthesia-related allergies, should be collected to evaluate systemic health status, to rule out systemic diseases that are not suitable for surgery, and to predict possible complications. Blood pressure should be measured, and a physician should be consulted if necessary. Blood tests. Routine blood test results, clotting time, infectious diseases (hepatitis B, hepatitis C, AIDS, and syphilis), and blood glucose levels should be recorded. Maxillofacial examination. Check whether there is swelling of the maxillofacial region. General oral examination. Examination of temporomandibular joint, width of mouth opening, oral hygiene status, occlusion, oral vestibular depth, muscle attachment, etc. should be performed. Examination of the diseased tooth. The condition of hard tissues, including the shape of the tooth crown, the presence of a restoration, the integrity and marginal adaptation of the restoration, should be assessed. The conditions of the periodontal tissues and mucosa, the color and morphological texture of the gingiva and mucosa, the presence of a sinus tract, and the location and source of the sinus tract should be examined. The periodontal probing depth, width of the attached gingiva, condition of the root furcation, and health status of the interdental papilla should be evaluated. Imaging examinations Periapical radiographs and cone beam computed tomography (CBCT) should be obtained. The parallelling projection technique is recommended for periapical radiographs. CBCT can be used to determine the extent of the lesion and to examine the diseased tooth and its anatomical relationship with the surrounding tissues. , Confirming clinical diagnosis and developing treatment plans A correct diagnosis of the diseased tooth should be made based on the patient’s chief complaints, medical history, and examination results. Systemic and oral health evaluations should be performed, and apical microsurgery should be selected according to the indications.
Systemic conditions. The patient’s past medical history, medication history, and allergy history, especially the history of anesthesia-related allergies, should be collected to evaluate systemic health status, to rule out systemic diseases that are not suitable for surgery, and to predict possible complications. Blood pressure should be measured, and a physician should be consulted if necessary. Blood tests. Routine blood test results, clotting time, infectious diseases (hepatitis B, hepatitis C, AIDS, and syphilis), and blood glucose levels should be recorded. Maxillofacial examination. Check whether there is swelling of the maxillofacial region. General oral examination. Examination of temporomandibular joint, width of mouth opening, oral hygiene status, occlusion, oral vestibular depth, muscle attachment, etc. should be performed. Examination of the diseased tooth. The condition of hard tissues, including the shape of the tooth crown, the presence of a restoration, the integrity and marginal adaptation of the restoration, should be assessed. The conditions of the periodontal tissues and mucosa, the color and morphological texture of the gingiva and mucosa, the presence of a sinus tract, and the location and source of the sinus tract should be examined. The periodontal probing depth, width of the attached gingiva, condition of the root furcation, and health status of the interdental papilla should be evaluated. Imaging examinations Periapical radiographs and cone beam computed tomography (CBCT) should be obtained. The parallelling projection technique is recommended for periapical radiographs. CBCT can be used to determine the extent of the lesion and to examine the diseased tooth and its anatomical relationship with the surrounding tissues. ,
A correct diagnosis of the diseased tooth should be made based on the patient’s chief complaints, medical history, and examination results. Systemic and oral health evaluations should be performed, and apical microsurgery should be selected according to the indications.
Medical preparations It is recommended that surgery be performed in a dental clinic with dedicated space and that the clinic room be disinfected. The equipment should include a dental operative microscope and an ultrasonic unit. The instruments should include 45-degree surgical handpiece and long surgical burs; incision, separation, exposure, and suturing instruments; minicurettes; micromirrors; a microexplorer; ultrasonic tips for root-end preparation; and micropluggers. Drugs and other materials include anesthetic drugs, disinfectants, bioactive materials, vasoconstrictors, and stains. Patient preparation Chlorhexidine compound mouthwash should be used, and anti-inflammatory and analgesic drugs should be administered taken orally if necessary. Antibiotics can be used prophylactically when there is a risk of infection. Local anesthesia Anesthesia should cover the diseased tooth plus two neighboring teeth. Infiltration anesthesia is recommended for maxillary teeth, and block and infiltration anesthesia are recommended for mandibular teeth. The local anesthesia is performed according to the standard of the Chinese Stomatological Association “Guideline for oral local anesthesia (T/CHSA 021—2023)”. Surgical area preparation After disinfecting the surgical area, a drape should be applied.
It is recommended that surgery be performed in a dental clinic with dedicated space and that the clinic room be disinfected. The equipment should include a dental operative microscope and an ultrasonic unit. The instruments should include 45-degree surgical handpiece and long surgical burs; incision, separation, exposure, and suturing instruments; minicurettes; micromirrors; a microexplorer; ultrasonic tips for root-end preparation; and micropluggers. Drugs and other materials include anesthetic drugs, disinfectants, bioactive materials, vasoconstrictors, and stains.
Chlorhexidine compound mouthwash should be used, and anti-inflammatory and analgesic drugs should be administered taken orally if necessary. Antibiotics can be used prophylactically when there is a risk of infection.
Anesthesia should cover the diseased tooth plus two neighboring teeth. Infiltration anesthesia is recommended for maxillary teeth, and block and infiltration anesthesia are recommended for mandibular teeth. The local anesthesia is performed according to the standard of the Chinese Stomatological Association “Guideline for oral local anesthesia (T/CHSA 021—2023)”.
After disinfecting the surgical area, a drape should be applied.
The clinical operating procedure of apical microsurgery includes seven main steps as shown in Fig. . Microscope positioning and use The relative positions of the microscope and the patient should be adjusted so that the operation can be performed under direct microscopic vision. When the resected root surface is inspected and the root end is prepared, the root canal can be observed from a reflected view by micromirror under microscope. Flap incision and suturing should be performed under low magnification, inspection should be performed under high magnification, and other operations should be performed under medium magnification. , , Flap design A full-thickness flap including the diseased tooth and two neighboring teeth should be created with horizontal and vertical incisions; the former should include incisions in the gingival sulcus and attached gingival incisions. A rectangular flap, consisting of a mesial and distal vertical incision and a horizontal incision in the gingival sulcus or the attached gingiva, is usually used for the anterior teeth (Figs. , ); a triangular flap consisting of a mesial vertical incision and a horizontal incision in the gingival sulcus is used for the posterior teeth (Fig. ). In aesthetically relevant areas, the use of the horizontal submarginal incision or the papilla base incision is recommended, to avoid possible gingival recession from horizontal sulcular incision. , – Flap incision and elevation After the surgical blade cuts through the gingiva, mucosa, and periosteum to the bone surface, the full-thickness flap is elevated with a periosteal elevator. Retractors of an appropriate shape should be used to rest on the bone surface, and the flap, lip, and cheeks can be pulled without tension to fully expose the surgical field. Root apex positioning Based on the preoperative CBCT images, the working length of the root canal treatment, the position of a sinus tract, and the alveolar bone eminence at the root, the location of the root apex should be determined precisely. Root apex exposure If the cortical bone at the apical area is destroyed, osteotomy is not necessary. If the cortical bone at the apical area is intact, a 45-degree surgical handpiece with long surgical bur, a trephine, or an ultrasonic osteotome, can be used for osteotomy at the apical area of the diseased tooth to expose the root apex. Root end resection, curettage, and inspection The pathological tissue or foreign bodies in the periapical lesion area should be scraped off. – Under sterile water cooling, approximately 3 mm of the root apex is resected, and the cross-section of the root should be positioned perpendicular to the long axis of the root or inclined ≤ 10° in the buccal direction. After resection, the residual pathological tissue is removed, and the resected root surface is smoothed. – Epinephrine cotton pellets, ferric sulfate, aluminum chloride, and calcium sulfate can be used for hemostasis via biological effect and/or mechanical compression. The resected root surface should be stained with methylene blue solution, rinsed with normal saline, dried, and inspected under a microscope at high magnification to clarify the presence of vertical root fracture, microleakage, the isthmus, the missing root canal, lateral canals, and perforation, etc. , , – Root end preparation and filling An ultrasonic tip of appropriate diameter and bending direction should be used to lightly “peck” the gutta percha to prepare a Class I cavity coaxial with the root into a minimum of 3 mm depth along the running direction of the root canal while irrigating and cooling. Overcutting of the dentin wall should be avoided, and the microplugger should be used to compact the filling at the bottom of the cavity. The root canal should be cleaned and dried, after which bioactive material is filled into the cavity using the microplugger. The filling material should be compressed layer by layer, and excess material outside of root canal should be removed. , – Bone crypt treatments The bone crypt should be rinsed with normal saline to determine whether any foreign bodies have been retained. Suturing The mucoperiosteal flap should be repositioned, aligned accurately, and sutured without tension. Vertical incisions are closed with interrupted sutures, and horizontal incisions are closed with sling or interrupted sutures. Suture removal The time needed for suture removal depends on the condition of the incision. It is generally recommended that sutures be removed 5 to 7 days after surgery.
The relative positions of the microscope and the patient should be adjusted so that the operation can be performed under direct microscopic vision. When the resected root surface is inspected and the root end is prepared, the root canal can be observed from a reflected view by micromirror under microscope. Flap incision and suturing should be performed under low magnification, inspection should be performed under high magnification, and other operations should be performed under medium magnification. , ,
A full-thickness flap including the diseased tooth and two neighboring teeth should be created with horizontal and vertical incisions; the former should include incisions in the gingival sulcus and attached gingival incisions. A rectangular flap, consisting of a mesial and distal vertical incision and a horizontal incision in the gingival sulcus or the attached gingiva, is usually used for the anterior teeth (Figs. , ); a triangular flap consisting of a mesial vertical incision and a horizontal incision in the gingival sulcus is used for the posterior teeth (Fig. ). In aesthetically relevant areas, the use of the horizontal submarginal incision or the papilla base incision is recommended, to avoid possible gingival recession from horizontal sulcular incision. , –
After the surgical blade cuts through the gingiva, mucosa, and periosteum to the bone surface, the full-thickness flap is elevated with a periosteal elevator. Retractors of an appropriate shape should be used to rest on the bone surface, and the flap, lip, and cheeks can be pulled without tension to fully expose the surgical field.
Based on the preoperative CBCT images, the working length of the root canal treatment, the position of a sinus tract, and the alveolar bone eminence at the root, the location of the root apex should be determined precisely.
If the cortical bone at the apical area is destroyed, osteotomy is not necessary. If the cortical bone at the apical area is intact, a 45-degree surgical handpiece with long surgical bur, a trephine, or an ultrasonic osteotome, can be used for osteotomy at the apical area of the diseased tooth to expose the root apex.
The pathological tissue or foreign bodies in the periapical lesion area should be scraped off. – Under sterile water cooling, approximately 3 mm of the root apex is resected, and the cross-section of the root should be positioned perpendicular to the long axis of the root or inclined ≤ 10° in the buccal direction. After resection, the residual pathological tissue is removed, and the resected root surface is smoothed. – Epinephrine cotton pellets, ferric sulfate, aluminum chloride, and calcium sulfate can be used for hemostasis via biological effect and/or mechanical compression. The resected root surface should be stained with methylene blue solution, rinsed with normal saline, dried, and inspected under a microscope at high magnification to clarify the presence of vertical root fracture, microleakage, the isthmus, the missing root canal, lateral canals, and perforation, etc. , , –
An ultrasonic tip of appropriate diameter and bending direction should be used to lightly “peck” the gutta percha to prepare a Class I cavity coaxial with the root into a minimum of 3 mm depth along the running direction of the root canal while irrigating and cooling. Overcutting of the dentin wall should be avoided, and the microplugger should be used to compact the filling at the bottom of the cavity. The root canal should be cleaned and dried, after which bioactive material is filled into the cavity using the microplugger. The filling material should be compressed layer by layer, and excess material outside of root canal should be removed. , –
The bone crypt should be rinsed with normal saline to determine whether any foreign bodies have been retained.
The mucoperiosteal flap should be repositioned, aligned accurately, and sutured without tension. Vertical incisions are closed with interrupted sutures, and horizontal incisions are closed with sling or interrupted sutures.
The time needed for suture removal depends on the condition of the incision. It is generally recommended that sutures be removed 5 to 7 days after surgery.
Pathological examination is recommended for removing granulation-like tissue or cystic wall-like tissue after periapical curettage. The pathological examination results should be recorded in the medical records.
Postoperative reactions After apical microsurgery, some patients may experience mild to moderate pain, swelling, and congestion; severe postoperative reactions are rare. Care and medication After surgery, antibacterial mouthwash should be used to maintain oral hygiene. A cold compress should be applied intermittently for 24 hours in the surgical area, and an intermittent hot compress can be used if swelling still occurs afterward. Analgesics should be taken orally when there is pain. Patients who experience maxillary sinus perforation during surgery should be instructed to sleep with the head facing down, to not blow their nose forcefully, to avoid swimming, and to take antibiotics to prevent infection for 5–7 days after surgery. ,
After apical microsurgery, some patients may experience mild to moderate pain, swelling, and congestion; severe postoperative reactions are rare.
After surgery, antibacterial mouthwash should be used to maintain oral hygiene. A cold compress should be applied intermittently for 24 hours in the surgical area, and an intermittent hot compress can be used if swelling still occurs afterward. Analgesics should be taken orally when there is pain. Patients who experience maxillary sinus perforation during surgery should be instructed to sleep with the head facing down, to not blow their nose forcefully, to avoid swimming, and to take antibiotics to prevent infection for 5–7 days after surgery. ,
Surgical area infections When there are signs of infection, treatment should be administered according to the principles for the treatment of surgical infections. Neighboring tooth injury Injuries of the roots of the neighboring teeth should be avoided during apical microsurgery to the greatest extent possible. In the event of root injury to a neighboring tooth, a sterile cotton pellet should be immediately used to protect the wound surface to avoid contamination. The cotton pellet must be removed before flap repositioning; no special treatment is needed, but the patient should be periodically followed-up. , Maxillary sinus perforation In the event of maxillary sinus perforation, a cotton pellet tied with a thread can be used to block the perforation to avoid entrance of foreign bodies into the sinus cavity, after which surgery can be continued; if the perforation is large, the use of an absorbable collagen membrane is recommended to repair the maxillary sinus perforation after root-end filling. – Nerve injury Nerve injury, a serious complication, mostly occurs in the mental nerve, followed by the inferior alveolar nerve. Accurate preoperative positioning and effective intraoperative protection of the neurovascular bundle are required to avoid irreversible damage. Other Other complications, including vascular injury, soft tissue laceration, incision dehiscence, and surgical site infection, should be treated according to standard surgical principles.
When there are signs of infection, treatment should be administered according to the principles for the treatment of surgical infections.
Injuries of the roots of the neighboring teeth should be avoided during apical microsurgery to the greatest extent possible. In the event of root injury to a neighboring tooth, a sterile cotton pellet should be immediately used to protect the wound surface to avoid contamination. The cotton pellet must be removed before flap repositioning; no special treatment is needed, but the patient should be periodically followed-up. ,
In the event of maxillary sinus perforation, a cotton pellet tied with a thread can be used to block the perforation to avoid entrance of foreign bodies into the sinus cavity, after which surgery can be continued; if the perforation is large, the use of an absorbable collagen membrane is recommended to repair the maxillary sinus perforation after root-end filling. –
Nerve injury, a serious complication, mostly occurs in the mental nerve, followed by the inferior alveolar nerve. Accurate preoperative positioning and effective intraoperative protection of the neurovascular bundle are required to avoid irreversible damage.
Other complications, including vascular injury, soft tissue laceration, incision dehiscence, and surgical site infection, should be treated according to standard surgical principles.
Follow-up Clinical and imaging examinations are regularly performed 3, 6, 12, and 24 months after surgery. For patients who still have periapical lesions 1 year after surgery, follow-up should be conducted annually; observation should continue until 4 years after surgery. Efficacy evaluation Surgical efficacy should be preliminarily evaluated 1 year after surgery, and finally determined 4 years after surgery. – , – , Apical radiographs should be routinely taken. For patients who still have symptoms and for whom preoperative CBCT was taken, the scan can be used to evaluate the healing status of the periapical lesions. Successful efficacy is indicated if the diseased tooth has no pain or swelling, there is good healing of soft tissues, there are no sinus openings, and there is no loss of function, and if imaging examinations show that the periapical lesions have disappeared or shrunk. Surgical failure is considered if the diseased teeth have clinical symptoms and signs and the imaging examination shows no change or expansion of the periapical lesions. For teeth without clinical symptoms and signs but whose imaging results reveal indeterminate healing, continued observation of the teeth is recommended. , –
Clinical and imaging examinations are regularly performed 3, 6, 12, and 24 months after surgery. For patients who still have periapical lesions 1 year after surgery, follow-up should be conducted annually; observation should continue until 4 years after surgery.
Surgical efficacy should be preliminarily evaluated 1 year after surgery, and finally determined 4 years after surgery. – , – , Apical radiographs should be routinely taken. For patients who still have symptoms and for whom preoperative CBCT was taken, the scan can be used to evaluate the healing status of the periapical lesions. Successful efficacy is indicated if the diseased tooth has no pain or swelling, there is good healing of soft tissues, there are no sinus openings, and there is no loss of function, and if imaging examinations show that the periapical lesions have disappeared or shrunk. Surgical failure is considered if the diseased teeth have clinical symptoms and signs and the imaging examination shows no change or expansion of the periapical lesions. For teeth without clinical symptoms and signs but whose imaging results reveal indeterminate healing, continued observation of the teeth is recommended. , –
The medical records, including clinical examination, radiologic images, consultant, informed consent, prescription, surgical procedure, pathological examination results, and follow-ups, should be standardized and saved.
Following the biological concepts, i.e. complete debridement, tight sealing of root canal system, and conservation of dental tissue, the apical microsurgery, combined with the magnification and illumination provided by the dental operate microscope with the proper use of micro instruments, ultrasonic retrotips and bioceramics as root-end filling materials, can treat the endodontic origin diseases precisely and less traumatically with high success rate. More and more natural and healthy teeth have been preserved successfully. There are many technical changes that added to the evolution of apical microsurgery, including piezoelectric surgery, static navigation, dynamic navigation, augmented reality-guided surgery, and robot-assisted surgery. , , , –
|
Oral health education strategies for patients living with cardiovascular disease within hospital settings: a scoping review | b45f5e74-f8ed-4831-bda3-f96eecb4d63d | 11220159 | Patient Education as Topic[mh] | Introduction Cardiovascular disease (CVD) is a global public health issue. It is the leading cause of death world-wide and whilst modifiable risk factors smoking status, healthy diet, active lifestyle, and alcohol intake are well known ; one that is rarely publicised is poor oral health. Up to 90% of any population with at least one tooth are living with periodontal disease, a preventable oral condition that plays an integral role in oral as well as systemic health. Preventing this and other oral diseases begins with oral health education and involves oral hygiene instructions. Optimal oral hygiene practices involve toothbrushing for at least 2 mins twice daily and cleaning interdentally once a day . Manually removing dental biofilm from oral tissues with these habits are the most efficient way to ensure good oral health and lower both local and systemic inflammation . Traditionally oral health education is delivered within dental settings. However, just under half of adults world-wide do not attend a dentist regularly and as such, do not receive this important messaging. The health outcomes of those living with CVD can be impacted by poor oral hygiene as never or rarely brushing teeth has been shown to significantly increase the risk of a CVD event . The mechanism behind this is a result of poor oral hygiene allowing dental biofilm to remain stagnant on oral tissues, initiating an immune response and involves vasodilation of gingival tissues to allow rapid movement of immune cells to the site . As such, even in healthy individuals, poor oral hygiene can lead to elevation of inflammatory markers high-sensitive C-reactive protein (hsCRP) and interleukin (IL)-6 in as little as 3 weeks . For those living with CVD, an elevation of these markers puts them at an increased risk of a future cardiac event . Vasodilation of gingival tissues also gives pathogenic oral bacteria within the dental biofilm access to the body and its systems via the blood stream . Once in the bloodstream these pathogens and their secretions can lodge in distant organs such as the lungs, kidneys, brain , and heart where they can initiate a localised inflammatory response . In the heart these pathogens have been shown to invade vessel walls and adhere to atherosclerotic lesions, leading to atherosclerosis ; the primary cause of CVD . Many barriers exist that prevent individuals from attending the dental clinic such as cost, ease of access , and anxiety to name a few, highlighting the need to expand oral health education to other areas of healthcare. Research has shown that increased oral hygiene in hospital wards decreases the incidence of non-ventilator hospital acquired pneumonia and shortens hospital stays . Within this setting, however, oral care is delegated to nurses who can face many challenges to providing this care . As such, oral health practitioners (OHP) would be an appropriate alternative to take on this responsibility . Equally, an emerging educational tool within healthcare has been the use of digital devices . Their ability to deliver health messages across all literacy levels has improved patient quality of life (QoL) ; and digitally delivered health information within patient waiting rooms has been shown to improve oral hygiene practices and promote healthy lifestyle behaviours in patients with CVD . Whether face-to-face or digitally delivered, oral health education provided in cardiology settings has the potential to improve oral health and reduce the risk profile of patients living with CVD. 1.1 Aims This review aims to identify and describe oral health education programmes provided to patients living with CVD within hospital wards and outpatient clinics; as well as discuss any effect they had on health outcomes.
Aims This review aims to identify and describe oral health education programmes provided to patients living with CVD within hospital wards and outpatient clinics; as well as discuss any effect they had on health outcomes.
Methods This review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis for Scoping Reviews (PRISMA-ScR) statement . See . The review protocol is registered in the Open Science Framework (OSF) registry. 2.1 Search strategy The initial search commenced 21st December 2022 and was repeated on 14th August 2023. The final search occurring 7th May 2024. Electronic databases used include Cochrane, Medline (via Ovid), and Scopus. No limitations were placed on language or publication period, and no human filter was applied. Grey literature included phrase searching via Google Scholar, as well as reviewing citation lists of relevant studies. The search strategy included a combination of the following terms “cardiovascular disease, heart disease, education relating to dental health, oral health, health promotion, digital education, video education, patient education, health knowledge, oral hygiene instruction, hospital, oral health, oral care, dental care, preventative dentistry, video recording, video-audio media.” Boolean operators (AND/OR), medical subject headings (MeSH), and truncations were also utilised. For the full search see . Every attempt was made to retrieve studies that were inaccessible including web searches, using The University of Sydney library resources, and lastly, attempting to contact corresponding authors. 2.2 Inclusion and exclusion criteria The population comprised of adults ≥18 years who had been or were recently hospitalised as a result of cardiovascular disease. The intervention included digital and/or traditional oral health education being delivered to these patients within hospital wards or out-patient clinics. Any studies where oral hygiene education took place outside of the hospital environment, or within a dental setting were excluded. Only peer-reviewed publications of randomised controlled trials, quasi-randomised controlled trials, observational studies, including cohort, case–control and cross-sectional studies were eligible for inclusion. Conference abstracts, case-studies/series, letters to the editors, and editorials were excluded. Any non-English or animal studies were manually excluded. 2.3 Screening and data extraction Using Covidence, a systematic review software , 7,133 identified citations were imported into the programme where 1,695 duplicates were automatically removed. A further 12 duplicates were manually removed leaving a total of 5,426 for screening. Titles and abstracts were screened independently by LC and one of four alternate reviewers (LR, AT, LQ, FX). Most results did not relate to oral health education, CVD, adults, and/or were not based within a hospital setting. As such they were deemed irrelevant. Twenty-three articles were identified for further assessment, however after full text screening, a total of 3 studies [1 quasi-randomised and 2 randomised controlled trials ] met the inclusion criteria (see ). Any conflicts arising during the screening process were resolved via group discussion. A data extraction tool developed within Covidence was completed independently first by LC, followed by either LR, FX, AT, or LQ. Key characteristics extracted were author(s), publication year, country, study design, aim(s), setting, mean age, sex, outcomes measures, a comparison of oral hygiene intervention, intervention duration, and outcomes. Due to the heterogeneity of the studies, data synthesis has been presented narratively with reference to supporting material.
Search strategy The initial search commenced 21st December 2022 and was repeated on 14th August 2023. The final search occurring 7th May 2024. Electronic databases used include Cochrane, Medline (via Ovid), and Scopus. No limitations were placed on language or publication period, and no human filter was applied. Grey literature included phrase searching via Google Scholar, as well as reviewing citation lists of relevant studies. The search strategy included a combination of the following terms “cardiovascular disease, heart disease, education relating to dental health, oral health, health promotion, digital education, video education, patient education, health knowledge, oral hygiene instruction, hospital, oral health, oral care, dental care, preventative dentistry, video recording, video-audio media.” Boolean operators (AND/OR), medical subject headings (MeSH), and truncations were also utilised. For the full search see . Every attempt was made to retrieve studies that were inaccessible including web searches, using The University of Sydney library resources, and lastly, attempting to contact corresponding authors.
Inclusion and exclusion criteria The population comprised of adults ≥18 years who had been or were recently hospitalised as a result of cardiovascular disease. The intervention included digital and/or traditional oral health education being delivered to these patients within hospital wards or out-patient clinics. Any studies where oral hygiene education took place outside of the hospital environment, or within a dental setting were excluded. Only peer-reviewed publications of randomised controlled trials, quasi-randomised controlled trials, observational studies, including cohort, case–control and cross-sectional studies were eligible for inclusion. Conference abstracts, case-studies/series, letters to the editors, and editorials were excluded. Any non-English or animal studies were manually excluded.
Screening and data extraction Using Covidence, a systematic review software , 7,133 identified citations were imported into the programme where 1,695 duplicates were automatically removed. A further 12 duplicates were manually removed leaving a total of 5,426 for screening. Titles and abstracts were screened independently by LC and one of four alternate reviewers (LR, AT, LQ, FX). Most results did not relate to oral health education, CVD, adults, and/or were not based within a hospital setting. As such they were deemed irrelevant. Twenty-three articles were identified for further assessment, however after full text screening, a total of 3 studies [1 quasi-randomised and 2 randomised controlled trials ] met the inclusion criteria (see ). Any conflicts arising during the screening process were resolved via group discussion. A data extraction tool developed within Covidence was completed independently first by LC, followed by either LR, FX, AT, or LQ. Key characteristics extracted were author(s), publication year, country, study design, aim(s), setting, mean age, sex, outcomes measures, a comparison of oral hygiene intervention, intervention duration, and outcomes. Due to the heterogeneity of the studies, data synthesis has been presented narratively with reference to supporting material.
Results 3.1 Characteristics Three studies were identified , and included a total of 245 participants. See . The studies took place in either Hong Kong or Japan and spanned between 2013 and 2019. Participants recruited into the studies were patients who were within either a cardiac surgical or stroke rehabilitation hospital ward or were attending a hospital out-patient rehabilitation clinic. Across all studies, the lowest mean age of participants was 66.6 ± 10.8, the highest 70.9 ± 11.1; 60.6–68% were male. Whilst Omori et al. did not discuss employment status, two thirds of Dai et al.’s (housewife: 12.8%, retired: 51.1%, and unemployed: 2.1%), and close to three quarters of Lam et al.’s subjects were not working (53.1% retired, 19.8% homemaker). All studies reported a lack of regular oral hygiene practices at baseline , and reported no significant difference in oral hygiene status between intervention groups at baseline. 3.2 Outcome measures The primary outcome measures for two studies included measures of oral hygiene status using the Silness and Löe plaque index and the gingival bleeding index . See , for indices criteria. Secondary outcomes included gingival bleeding at 6 months or oral functional status, assessed by patients’ ability to perform toothbrushing and insert/remove their dentures . The outcome measures for one study was the number of oral bacteria on the tongue, followed by oral hygiene status, periodontal parameters, tongue coating scores, self-efficacy scale for self-care (SESS) scores, and the incidence of postoperative atrial fibrillation (AF). This study also assessed plaque score by using O’Leary’s plaque control record method . 3.3 Interventions The education provided in each trial focused on oral hygiene instruction and was delivered by oral health practitioners or dental assistants. Each intervention differed in its methods and materials, with follow ups of 3-weeks , 3- and 6-months , and discharge (approximately 1 month) . See for intervention details. In the first study, stroke patients attending their outpatient rehabilitation were placed into one of two arms control: conventional oral hygiene care programme (COHCP) or intervention advanced oral hygiene care programme (AOHCP). The control arm receiving an oral hygiene care programme, manual toothbrush, regular toothpaste (Colgate), and one on one oral hygiene instruction with a dental assistant. Whilst the intervention arm received the care programme, toothpaste, and one on one instruction however, also received an electric toothbrush with manufacturer’s instructions and a 3-month supply of chlorhexidine (CHX) mouth rinse . Similarly, investigators from another study involving CVD patients within a surgical ward , placed them into one of two arms. Both arms received similar interventions including oral hygiene instruction, using disclosing solution, interdental brush use delivered by certified dental hygienists. The hygienists also provided post-operative oral care to a small number of participants in each group. However, the teaching method differed between the groups. The control group received skills-based teaching, the intervention arm received oral hygiene instruction via a modified behavioural six-step method . See for this method. The final study included three arms: one control and two interventions. Each arm received oral hygiene instruction, whilst the intervention arms also received chlorhexidine (CHX) mouth rinse alone or in combination with 2 x weekly assisted brushing. The hygiene instruction was performed by a registered dentist and CHX was administered by ward nurses. The intervention arm receiving assisted brushing, had this performed by trained ward nurses. Training involved a 30-min education session run by dental hygienists. The authors deemed it unethical to include a negative control group due to their high risk of developing aspirational pneumonia . 3.4 Findings All 3 studies reported a lack of regular oral hygiene practices at baseline . At their conclusion, all found an improvement in toothbrushing habits, whilst one reported a significant increase of interdental brush use. All study arms had significant reductions in oral hygiene measures including plaque scores ( p = <0.001) , improved periodontal parameters, and tongue coating scores . Gingival bleeding was also reduced in all arms of two studies (p 0.004) , ( p < 0.001) , however, one reported significance in the intervention groups only ( p = 0.032). One study assessed tongue bacterial numbers , and reported significantly less bacteria (x10 7 cfu/mL) on participant tongues ( p < 0.02), as well as fewer days with post-operative AF in the intervention group (1.5 ± 2.8 vs. 4.8 ± 7.6 p = 0.019). It also reported that 5 patients (4 control and 1 intervention) developed pneumonia , whereas no patients developed pneumonia in the other studies . A 6-month follow-up was conducted in one study and they reported the continuation of plaque reduction ( p = 0.05) and a further reduction in bleeding on probing ( p < 0.01) in their intervention arm . SESS was measured in one study, showing worsened scores in the control arm .
Characteristics Three studies were identified , and included a total of 245 participants. See . The studies took place in either Hong Kong or Japan and spanned between 2013 and 2019. Participants recruited into the studies were patients who were within either a cardiac surgical or stroke rehabilitation hospital ward or were attending a hospital out-patient rehabilitation clinic. Across all studies, the lowest mean age of participants was 66.6 ± 10.8, the highest 70.9 ± 11.1; 60.6–68% were male. Whilst Omori et al. did not discuss employment status, two thirds of Dai et al.’s (housewife: 12.8%, retired: 51.1%, and unemployed: 2.1%), and close to three quarters of Lam et al.’s subjects were not working (53.1% retired, 19.8% homemaker). All studies reported a lack of regular oral hygiene practices at baseline , and reported no significant difference in oral hygiene status between intervention groups at baseline.
Outcome measures The primary outcome measures for two studies included measures of oral hygiene status using the Silness and Löe plaque index and the gingival bleeding index . See , for indices criteria. Secondary outcomes included gingival bleeding at 6 months or oral functional status, assessed by patients’ ability to perform toothbrushing and insert/remove their dentures . The outcome measures for one study was the number of oral bacteria on the tongue, followed by oral hygiene status, periodontal parameters, tongue coating scores, self-efficacy scale for self-care (SESS) scores, and the incidence of postoperative atrial fibrillation (AF). This study also assessed plaque score by using O’Leary’s plaque control record method .
Interventions The education provided in each trial focused on oral hygiene instruction and was delivered by oral health practitioners or dental assistants. Each intervention differed in its methods and materials, with follow ups of 3-weeks , 3- and 6-months , and discharge (approximately 1 month) . See for intervention details. In the first study, stroke patients attending their outpatient rehabilitation were placed into one of two arms control: conventional oral hygiene care programme (COHCP) or intervention advanced oral hygiene care programme (AOHCP). The control arm receiving an oral hygiene care programme, manual toothbrush, regular toothpaste (Colgate), and one on one oral hygiene instruction with a dental assistant. Whilst the intervention arm received the care programme, toothpaste, and one on one instruction however, also received an electric toothbrush with manufacturer’s instructions and a 3-month supply of chlorhexidine (CHX) mouth rinse . Similarly, investigators from another study involving CVD patients within a surgical ward , placed them into one of two arms. Both arms received similar interventions including oral hygiene instruction, using disclosing solution, interdental brush use delivered by certified dental hygienists. The hygienists also provided post-operative oral care to a small number of participants in each group. However, the teaching method differed between the groups. The control group received skills-based teaching, the intervention arm received oral hygiene instruction via a modified behavioural six-step method . See for this method. The final study included three arms: one control and two interventions. Each arm received oral hygiene instruction, whilst the intervention arms also received chlorhexidine (CHX) mouth rinse alone or in combination with 2 x weekly assisted brushing. The hygiene instruction was performed by a registered dentist and CHX was administered by ward nurses. The intervention arm receiving assisted brushing, had this performed by trained ward nurses. Training involved a 30-min education session run by dental hygienists. The authors deemed it unethical to include a negative control group due to their high risk of developing aspirational pneumonia .
Findings All 3 studies reported a lack of regular oral hygiene practices at baseline . At their conclusion, all found an improvement in toothbrushing habits, whilst one reported a significant increase of interdental brush use. All study arms had significant reductions in oral hygiene measures including plaque scores ( p = <0.001) , improved periodontal parameters, and tongue coating scores . Gingival bleeding was also reduced in all arms of two studies (p 0.004) , ( p < 0.001) , however, one reported significance in the intervention groups only ( p = 0.032). One study assessed tongue bacterial numbers , and reported significantly less bacteria (x10 7 cfu/mL) on participant tongues ( p < 0.02), as well as fewer days with post-operative AF in the intervention group (1.5 ± 2.8 vs. 4.8 ± 7.6 p = 0.019). It also reported that 5 patients (4 control and 1 intervention) developed pneumonia , whereas no patients developed pneumonia in the other studies . A 6-month follow-up was conducted in one study and they reported the continuation of plaque reduction ( p = 0.05) and a further reduction in bleeding on probing ( p < 0.01) in their intervention arm . SESS was measured in one study, showing worsened scores in the control arm .
Discussion This review assessed the current evidence in relation to oral hygiene education programmes provided to patients within cardiology wards and/or outpatient clinics and found that oral health education is rarely provided in these settings. Poor oral health is a significant global public health issue . Essential to preventing poor oral health is oral health education. However, as patients can have major barriers to overcome when accessing dental care , oral health education urgently needs to expand beyond the dental clinic. Incorporating this education and improving oral health within other areas of health would greatly benefit everyone and would have a profound positive effect on patients living with CVD . At baseline, participants of all included studies had poor oral health as well as suboptimal hygiene habits . At the end of the study periods, the intervention groups saw significant improvement in both clinical oral health status and self-reported oral hygiene habits. These findings mirror other studies using oral health practitioners (OHP) to educate patients in non-dental settings including residential aged care and mental health facilities . Omori et al. specifically illustrated a significant increase of interdental brush use in their study [(94.3%) intervention vs. (54.3%) control]. The improvement of interdental brush use, as well as other outcome parameters within the intervention arm is likely due to the six-step teaching method. This method has been shown to improve health outcomes as clinicians collaborate with patients to set achievable goals . Skills-only based health education is more paternalistic in nature, excluding patients’ prior beliefs or understanding, and removing autonomy . As the control arm received this type skills-based education, it could be related to the worsening of SESS score in this group. The improvement of oral health reported in the included studies also had a positive impact on post-operative health outcomes such as a reduction or absence of common CVD post-operative complications: pneumonia and post-operative AF; which can lengthen hospital stays or cause premature death . Although good oral hygiene has been shown to reduce systemic inflammation none of the included studies reported on inflammatory markers such as hsCRP or IL-6. 4.1 Current education strategies The education strategies employed in all included studies involved traditional face-to-face education and many of the excluded studies provided this education to nursing staff only . Nurses can face many challenges when providing oral care. From a personal level, barriers can include staffing issues, lack of time or training, or aversion to this care . This is reflected by a study where nurses admitted to ceasing toothbrushing altogether after the study period, even after oral hygiene was proven to eliminate ventilator assisted pneumonia due to the lack of time and because oral care was of low priority . However, challenges can also arise at the organisational level where training, resources and/or appropriate staff numbers are not provided . Furthermore, patients themselves can prevent nurses from providing oral care with aggressive behaviour, care refusal, communication issues, or where oral health is not prioritised . As such OHPs including dentists, dental hygienists, and oral health therapists, could be an appropriate alternative to help ease this burden on nursing staff . The FDI World Dental Federation also recognises the benefit OHPs would have within primary healthcare settings, calling for them to be integrated into these settings globally by 2030 . Currently however, resulting from a lack of political leadership, low oral health prioritisation on political agendas, as well as staffing, infrastructure, and funding obstacles , few countries are taking steps to utilise them in this way . 4.2 Changing the status quo One cost-effective way to bring oral health education to patients in these settings is through the use of digital technology. Digital CVD education programmes have been used to improve heart health outcomes , including text-message health tips post cardiac event . Whilst currently oral health messages are not included, they could easily be incorporated into these existing education packages. Another solution could be providing oral health training to non-dental clinicians such as pharmacists, general practitioners, and other allied health professionals, as many individuals visit these clinicians when they have a dental issue . A long-term solution at the organisational level could be a collaboration between universities and hospitals forming placements for student OHPs. Placements such as this have shown to benefit both students, hospital staff, and patients alike . Students placed in hospital settings would provide oral health education, assist with oral care, and with mobile dental units becoming more readily available, urgent treatment could be completed bedside . Additionally, referral pathways within the hospital system could also be created, utilising hospital dental clinics where applicable. 4.3 Gaps in the literature The number of eligible studies that involved the direct provision of oral health education to patients with CVD within hospital settings were limited and predominately located in Hong Kong or Japan between 2013 and 2019. These findings highlight the need to conduct more studies in different global communities. All included studies reported generalised poor oral health in their participants at baseline, similar to recent research within a Romanian emergency hospital and an Australian cardiac rehabilitation clinic . Both concluding further oral health education in these spaces are needed . A lack of resources and funding means OHPs are absent from hospital multidisciplinary teams . Poor oral health is a modifiable risk factor for CVD however, Appropriately, priority is given to patients’ heart health in cardiology wards however, current literature acknowledges poor awareness of the links between oral and heart health in patients within cardiology wards and outpatient clinics . This could be related to the absence of an OHP within these settings and compounded by the limited oral health messages in CVD education . This review has discussed oral health education programmes provided to patients with CVD in hospital settings. It has highlighted gaps where an OHP and/or digital technologies would be ideally placed to bridge them. As such, there are opportunities for future research and implementation of oral health education programmes for patients with CVD within hospital settings. Preferably oral health education should form part of primary prevention strategies for good general health, however incorporating it as part of secondary prevention strategies should also be a priority. 4.4 Clinical significance This review has highlighted the significant role oral health education plays in improving the long-term oral health in patients within hospital settings, as well as lowering the risk of common post-operative and post-stroke complications. Despite the important role oral health can play in cardiovascular health, this review has highlighted a lack of oral health education available to patients with CVD and proposed simple strategies to deliver these messages. The implementation and standardisation of programmes such as these may help to empower at-risk patients at their most vulnerable to improve their oral health for better general health. 4.5 Strengths and limitations The small number of eligible studies was a major limitation for this review. The three included studies took place in Hong Kong or Japan and thus may not be generalisable to CVD patients globally. However, a strength of this study is it is the first known review to analyse oral health education programmes provided to patients with CVD in hospital settings, highlighting a lack of oral health education in these spaces.
Current education strategies The education strategies employed in all included studies involved traditional face-to-face education and many of the excluded studies provided this education to nursing staff only . Nurses can face many challenges when providing oral care. From a personal level, barriers can include staffing issues, lack of time or training, or aversion to this care . This is reflected by a study where nurses admitted to ceasing toothbrushing altogether after the study period, even after oral hygiene was proven to eliminate ventilator assisted pneumonia due to the lack of time and because oral care was of low priority . However, challenges can also arise at the organisational level where training, resources and/or appropriate staff numbers are not provided . Furthermore, patients themselves can prevent nurses from providing oral care with aggressive behaviour, care refusal, communication issues, or where oral health is not prioritised . As such OHPs including dentists, dental hygienists, and oral health therapists, could be an appropriate alternative to help ease this burden on nursing staff . The FDI World Dental Federation also recognises the benefit OHPs would have within primary healthcare settings, calling for them to be integrated into these settings globally by 2030 . Currently however, resulting from a lack of political leadership, low oral health prioritisation on political agendas, as well as staffing, infrastructure, and funding obstacles , few countries are taking steps to utilise them in this way .
Changing the status quo One cost-effective way to bring oral health education to patients in these settings is through the use of digital technology. Digital CVD education programmes have been used to improve heart health outcomes , including text-message health tips post cardiac event . Whilst currently oral health messages are not included, they could easily be incorporated into these existing education packages. Another solution could be providing oral health training to non-dental clinicians such as pharmacists, general practitioners, and other allied health professionals, as many individuals visit these clinicians when they have a dental issue . A long-term solution at the organisational level could be a collaboration between universities and hospitals forming placements for student OHPs. Placements such as this have shown to benefit both students, hospital staff, and patients alike . Students placed in hospital settings would provide oral health education, assist with oral care, and with mobile dental units becoming more readily available, urgent treatment could be completed bedside . Additionally, referral pathways within the hospital system could also be created, utilising hospital dental clinics where applicable.
Gaps in the literature The number of eligible studies that involved the direct provision of oral health education to patients with CVD within hospital settings were limited and predominately located in Hong Kong or Japan between 2013 and 2019. These findings highlight the need to conduct more studies in different global communities. All included studies reported generalised poor oral health in their participants at baseline, similar to recent research within a Romanian emergency hospital and an Australian cardiac rehabilitation clinic . Both concluding further oral health education in these spaces are needed . A lack of resources and funding means OHPs are absent from hospital multidisciplinary teams . Poor oral health is a modifiable risk factor for CVD however, Appropriately, priority is given to patients’ heart health in cardiology wards however, current literature acknowledges poor awareness of the links between oral and heart health in patients within cardiology wards and outpatient clinics . This could be related to the absence of an OHP within these settings and compounded by the limited oral health messages in CVD education . This review has discussed oral health education programmes provided to patients with CVD in hospital settings. It has highlighted gaps where an OHP and/or digital technologies would be ideally placed to bridge them. As such, there are opportunities for future research and implementation of oral health education programmes for patients with CVD within hospital settings. Preferably oral health education should form part of primary prevention strategies for good general health, however incorporating it as part of secondary prevention strategies should also be a priority.
Clinical significance This review has highlighted the significant role oral health education plays in improving the long-term oral health in patients within hospital settings, as well as lowering the risk of common post-operative and post-stroke complications. Despite the important role oral health can play in cardiovascular health, this review has highlighted a lack of oral health education available to patients with CVD and proposed simple strategies to deliver these messages. The implementation and standardisation of programmes such as these may help to empower at-risk patients at their most vulnerable to improve their oral health for better general health.
Strengths and limitations The small number of eligible studies was a major limitation for this review. The three included studies took place in Hong Kong or Japan and thus may not be generalisable to CVD patients globally. However, a strength of this study is it is the first known review to analyse oral health education programmes provided to patients with CVD in hospital settings, highlighting a lack of oral health education in these spaces.
Conclusion This review concludes there is a need for further development and evaluation of oral health education programmes within hospital settings in different countries. Many patients with CVD have poor oral hygiene which can increase their risk of a recurrent cardiac event. The provision of basic oral health education provided directly to patients significantly improved oral hygiene, minimised the risk of post-operative pneumonia, and lowered post-operative days with AF.
The original contributions presented in the study are included in the article/ , further inquiries can be directed to the corresponding author.
LC: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review & editing. LR: Formal analysis, Writing – review & editing. FX: Formal analysis, Writing – review & editing. LQ: Formal analysis, Writing – review & editing. AT: Formal analysis, Writing – review & editing. JW: Conceptualization, Writing – review & editing. SK: Supervision, Writing – review & editing.
|
Immunopeptidomics for autoimmunity: unlocking the chamber of immune secrets | b7654947-22cc-41c4-a2c7-72d3807506b8 | 11747513 | Biochemistry[mh] | T cells form a highly antigen-specific arm of the adaptive immune system . T cells achieve this specificity through their surface T Cell Receptors (TCRs), which are diversified through V(D)J recombination to generate a large number of unique clones , . T cells can either have TCRalpha and beta chains (αβT cells), or TCRgamma and delta chains (γδT cells). Classically, αβT cells recognize peptide epitopes presented on MHC/HLA, whereas γδT cells recognize non-peptide ligands on non-classical MHC such as CD1d and MR1 , . While presentation through and recognition of non-classical MHC is important, here, we will focus on presentation of peptide epitopes to T cells by classical MHC. For the purposes of this review, we will use MHC and HLA interchangeably, with HLA as the preferred terminology for human alleles. There are two distinct classes of T cells, CD8+ T cells recognize epitopes on class I MHC, whereas CD4+ T cells recognize epitopes on class II MHC. The epitopes presented on class I MHC are typically 8-12 amino acids in length, whereas those presented by class II MHC can range from 10-25 amino acids , . Class I MHC are found on all nucleated cells; whereas class II MHC are found typically on professional antigen presenting cells (APCs) such as B cells, Dendritic Cells, and Macrophages , . Central and peripheral tolerance mechanisms typically restrict self-reactive T cells. T cells that recognize self-antigens in MHC molecules can be unleashed either as a failure of tolerance and/or therapeutic blockade of immune checkpoints , . The recognition of self epitopes presented on MHC is critical for developing autoimmunity. The pathways and sources of epitopes presented on these molecules are distinct, and together extensively sample the extracellular and intracellular proteome – . The MHC Class I pathway presents peptides from intracellular sources, such as viral or endogenous proteins, to CD8+ T cells . These proteins are degraded by the proteasome, and the resulting peptides are transported into the ER by TAP, where they are loaded onto class I MHC molecules. In contrast, the class II MHC pathway captures extracellular proteins, which are internalized, processed in endosomes, and presented on class II MHC molecules to CD4+ T cells. Cross-presentation involves the uptake of extracellular antigens by dendritic cells, which then process these antigens via either a vacuolar pathway or a cytosolic pathway. In the vacuolar pathway, antigens are degraded within endosomal compartments and loaded onto class I MHC molecules. In the cytosolic pathway, Dendritic cells can translocate the extracellular antigens into the cytosol, where they are processed by the proteasome. This mechanism allows Dendritic cells to activate CD8+ T cells and initiate cytotoxic responses against pathogens or tumors that do not directly infect them. The ability of MHC molecules to process and present epitopes from self and foreign proteins is critical for adaptive immunity – . Another important property of MHC is its high degree of polymorphism, with thousands of class I (HLA-A, HLA-B, HLA-C) and class II (HLA-DP, HLA-DQ, HLA-DR) alleles. Even inbred mouse strains have a diverse set of MHC (class I alleles H2-K, H2-D, H2-L and class II alleles H2-A and H2-E). This serves to diversify the T cell response to pathogens at a population level but adds complexity to studying the repertoire of peptides presented by them – . The HLA locus is one of the highest disease-associated genetic loci in autoimmunity , . Genetic variation in HLA alleles can confer protection from disease, e.g., HLA-DR2 in T1D or can contribute to the risk of developing autoimmunity, e.g., HLA-DQ2, DQ8, DR3, DR4, and HLA-A*02:01 in Type 1 Diabetes , . Similarly, in RA, HLA-DRB1*04:01, 04:04, 04:05, 01:01, and 10:01 are risk alleles , . The same HLA allele can be both protective and pathogenic for different autoimmune diseases. One of the most well-studied examples of such alleles is HLA-DR15 (HLA-DRB1*15:01), which confers heightened risk of developing Multiple Sclerosis, but is protective in T1D – . Variations in HLA alleles can predispose to or confer protection from autoimmunity by influencing the repertoire of peptides presented on them, and thereby shaping the adaptive immune response. Different HLA alleles have distinct peptide-binding preferences, resulting in unique peptide repertoires presented on the cell surface – . A striking example of how HLA-specific peptide preferences can positively and negatively influence immunity is found in HLA-B*27:05, a class I HLA allele. HLA-B*27:05 is well known to be enriched in individuals who can naturally control HIV infection, possibly by limiting viral escape from T cells – . On the other hand, HLA-B*27:05 is also strongly associated with Ankolysing Spondylitis , by possibly presenting self-peptides that may mimic microbial peptides. Another key set of studies providing a mechanistic explanation of how minute variations in HLA can affect immune repertoires comes from T1D. The accepted animal model of T1D is Non-Obese Diabetic (NOD) mice, which bear the class II MHC allele, I-Ag7. I-Ag7 possesses similar binding properties to the human T1D risk alleles, HLA-DQ8 and DQ2 . Studies have shown that these alleles have a neutral residue at the 57th position of the MHC-II β chain , resulting in a unique peptide-binding pocket. This limits the diversity of islet associated self-peptide repertoire, thereby restricting the islet-reactive TCR repertoire , . Upon modification of I-Ag7 to change the β57 residue to D/E, the TCR repertoire to insulin was altered, suggesting that changes at a single position in HLA/MHC can have profound effects on autoimmunity. These studies underscore the importance of identifying the peptides presented by disease-associated MHC/HLA to gain a better understanding of autoimmunity. An extensive literature and public database search was done to highlight some key HLA associations, epitopes, and antigens in context to autoimmune diseases which are listed in Table . The importance of MHC and epitope presentation has been documented since the mid-20 th century. Seminal work on tissue rejection and transplantation immunology in the early to mid-20th century, led to the discovery of the MHC genetic locus, first in mice (H-2) and then in HLA locus in humans . In the 1960s and 1970s, the major focus was on describing the genetics, structure, and function of MHC, revealing their critical role in antigen presentation to T cells – . Subsequent studies throughout the latter half of the 20th century revealed the finer details of T cell recognition of class I and class II MHC, establishing the principle of MHC as peptide receptors that present degraded proteins . The molecular basis of MHC restriction was firmly established by elucidating the crystal structure of HLA-A2 , and nearly a decade later, in 1996, the TCR-peptide-MHC complex . Our understanding of the landscape of epitopes presented by MHC has evolved alongside these discoveries. In the past decade, interest in identifying the peptides bound to MHC, collectively referred to as the ‘immunopeptidome’, has exploded across the biomedical research community , . This has been enabled by the yet nascent field of ‘immunopeptidomics’, which lies at the intersection of immunology and proteomics, and uses high-resolution mass spectrometry (MS) to identify and quantify the peptide repertoire presented by MHC . Since its inception, immunopeptidomics has facilitated our understanding of T cell responses, and has greatly enhanced the identification and profiling of antigen-specific T cells . Here, we will review how immunopeptidomics has been deployed to understand key autoantigens in autoimmune disorders. First, we will lay out the fundamentals and the technical considerations of implementing immunopeptidomics. Second, we will discuss how immunopeptidomics has transformed our understanding of post-translationally modified (PTM) epitopes. Third, we will discuss how immunopeptidomics has facilitated the identification of autoantigens and autoreactive T cells in three exemplary autoimmune disorders. Finally, we will summarize the outstanding challenges and provide future perspectives on utilizing immunopeptidomics for autoantigen discovery. Immunopeptidomics relies on MS for identification of protein fragments, similar to proteomics. However, given that peptide-MHC is a binary complex that is non covalently bound, peptides must be decoupled from MHC prior to MS. The four major steps in immunopeptidomics are: 1) Pulldown of MHC complexes using MHC-specific antibodies immobilized on beads; 2) Elution of peptides off of MHC using non-enzymatic mild acid treatment; 3) Peptide purification with either C18 reverse-phase separation or size-based methods such as size exclusion chromatography or filtering through a specific molecular weight cut off filter. Purified peptides are then subjected to MS; and 4) Bioinformatic identification of peptides from MS spectra. Each of these steps needs to be carefully designed, with several experiment-specific considerations. These are highlighted in Fig. . Some of the major considerations are: the need for a large number of starting cell numbers or amount of tissue, levels of MHC/HLA expression on the target cells, and the availability of antibodies specific to MHC/HLA alleles under investigation. The pioneering study that developed immunopeptidomics reported only tens of HLA-bound peptides in a single analysis from billions of cells. Recent advances in MS methods have enabled detection of peptides from limited sample with sensitivity, thus allowing for detection of antigens from primary cells and rare cell polulations . At present day, samples less than a billion cells can yield tens of thousands of peptides, including those with PTMs . The technical advances are largely fueled by Liquid Chromatography and tandem MS (LC-MS/MS), as well as sophisticated computational tools to identify and quantify spectra. For instance, peptide purification techniques have evolved to minimize peptide loss. They can range from using a molecular weight cut off filters or size exclusion chromatography to more sophisticated C18 reverse-phase chromatography that can separate compounds based on hydrophobicity. Advances in sample preparation during MS have also led to increased yields of peptides. For instance, a recent report used acetonitrile fractionation followed by introduction of ion mobility during gas phase separation to increase the number of detected peptides from the same samples by 2-5 fold . Similarly, Gravel et al., developed ion mobility separation-based time-of-flight (TOFIMS) MS to increase the sensitivity of immunopeptidomics . Several computational tools have led to increased sensitivity of detection and wide accessibility to users. For instance, ‘Immunolyser’ is a web-based tool that allows a standardized and streamlined workflow for immunopeptidomics that is accessible to researchers without any prior experience in MS . Similarly, SysteMHC Atlas v2.0 is a resource that has collected over 1 million peptides across over 7000 MS studies, and developed a suite of computational tools for analyses of PTMs, which has led to identification of over 470,000 modified peptides . A key consideration for calling peptides is also the selection of the appropriate databases. For instance, using the annotated genome will miss a large number of peptides that may be derived from unannotated ORFs , . Moreover, implementing machine learning algorithms such as PROSIT, and MS2rescore, MSbooster have led to increased sensitivity and reduced false negative rates – . Finally, while all these advances have enhanced peptide detection individually, combining them in various ways has synergistically led to significantly better outputs from immunopeptidomics . PTMs refer to chemical modifications of amino acid side chains occurring after their translation . PTMs can profoundly impact the structure, function, and localization of proteins . Of the 400 different types of PTMs that have been described in humans, phosphorylation, acetylation, and ubiquitination occur most frequently and are the most-well studied – . PTMs also alter the immunogenicity of peptide epitopes, which can be important in autoimmune disorders such as T1D and RA. The first demonstration of a PTM epitope presented on MHC was in melanoma, where a Tyrosinase derived epitope was found to be deamidated . Since then, numerous studies have profiled PTM epitopes on MHC – . It is estimated that peptides containing PTMs make up ~10% of the human immunopeptidome – . Importantly, dysregulation of PTMs is increasingly being implicated in the pathogenesis of autoimmune disease , . The contribution of PTMs to autoimmunity is manifold, involving a combination of host genetic factors and environmental exposures. Mechanistically, PTM of self-proteins can generate new epitopes, known as neoepitopes, capable of eliciting robust immune responses and breaking immune tolerance . PTM epitopes can alter binding to MHC and/or TCRs, thereby creating immunogenic neoepitopes – . PTMs add a layer of complexity to immunopeptidomics studies because of their low abundance, altered spectral profiles, and computational hurdles in identification , . PTMs may also be detected as an artifact of ionization that occurs during MS. It is appreciated that the immunopeptidomes of class I and II MHC differ in the types, positions, and ratios of PTMs . These preferences likely reflect: a) distinct research questions and model systems used, b) inherent differences in antigen processing and presentation between class I and II MHC, and c) diversity of pathways leading to PTMs , . While PTMs are often a small fraction of the immunopeptidome, their importance as autoantigens is outsized. For instance, a major epitope known as 2.5HIP (a hybrid peptide formed by post-translational splicing of Insulin and Chromogranin), was shown to be an essential antigen in NOD mice. However, it is often not detected in immunopeptidomics datasets. The importance of PTMs may often be disease-specific, and therefore PTM identification might not be relevant in all cases. Recent advances in single cell TCR sequencing have led to a large number of disease-associated T cells being profiled, however, the knowledge of their cognate epitopes is lagging behind by orders of magnitude . Experimentally, there are two types of approaches used for T cell epitope discovery. Antigen-directed approaches start with a limited number of (typically <1000) peptides and aim to identify T cells responsive to them. For instance, Wang et al. characterized the immunopeptidome of HLA-DR15 and identified self-epitopes and their microbial derived mimics as autoantigens in Multiple Sclerosis . On the other hand, TCR-directed approaches start with key TCRs and screen them against large epitope libraries (up to 10^7). We have recently identified novel T1D autoantigens using cell-based epitope libraries that were derived from a mouse pancreatic islet immunopeptidomics study , . In both cases, the knowledge of the peptides that were actually presented on MHC/HLA augmented antigen discovery by narrowing down the possible universe of epitopes recognized by T cells under investigation , , . In case of autoimmunity, this is especially important, as the scale of potential self-epitopes is genome-wide. In addition to the ~20,000 annotated coding genes, there are >10,000 unannotated or non-canonical open reading frames that contribute to the immunopeptidome. Furthermore, epitopes with PTMs add to this landscape of potential autoantigens. Immunopeptidomics can be used to scale down the number of epitopes under investigation, allowing better throughput and more targeted antigen discovery. In the next section, we will describe how immunopeptidomics approaches have helped autoantigen discovery in specific cases of autoimmune diseases: T1D, SLE, and RA. Type 1 Diabetes T1D or Autoimmune Diabetes, is a chronic disease caused by insulin deficiency due to the destruction of the insulin-producing β cells in the pancreatic islets of Langerhans , . Autoantibodies against insulin, the 65-kDa form of glutamic acid decarboxylase (GAD65), insulinoma-associated protein 2 (IA-2), and zinc transporter 8 (ZnT8) are associated with T1D but their role in the pathophysiology of the disease is not clear . It has been shown that autoreactive CD4+ and CD8+ T cells infiltrate the pancreas and mediate destruction of β cells. CD4+ T cells can propagate a pro-inflammatory environment through cytokine secretion and enhancing the function of cytotoxic CD8 + T cells, which can directly kill β cells. T1D can be modeled in NOD mice, which share several key features of disease including the presence of islet autoantibodies and infiltration of autoreactive CD4+ and CD8 + T cells in islets – . Islet-infiltrating T cells in NOD mice and T1D are known to recognize β cell autoantigens, many of which are overlapping. The restricted MHC diversity as well as availability of samples and reproducibility of disease course have allowed robust immunopeptidomics studies in NOD mice , . In contrast, the high HLA heterogeneity and limited access to viable β cells have impeded similar studies in humans. Islets harvested from cadaveric donors with T1D have very few β cells remaining, and those from donors without T1D have naturally low levels of HLA expression in the absence of inflammation, making direct detection of presented peptides challenging. Moreover, even in inflamed islet, class II HLA expression on β cells is low – . Therefore, the characterization of the HLA bound peptides from human β cells has been limited , . To circumvent these issues, approaches such as stably transfected human non β cell lines expressing specific autoantigen(s) and cell surface HLA allotypes of interest , , or human β cell lines generated by targeted oncogenesis have been used. A recent finding in T1D was the presence of hybrid insulin peptides (HIPs), which consist of epitopes generated by post-translational splicing of Insulin with other proteins. Studies have shown that HIPs are autoantigens for pathogenic CD4+ T cells in the human T1D and in NOD mice , , confirming the notion that PTM epitopes are key autoantigens , . Importantly, while the initial experiments with HIPs used synthetic peptides, their presence in the proteomes and immunopeptidomes derived from β cells has reaffirmed that HIP formation and recognition is a natural process that occurs in T1D . Subsequent studies based off of these results have led to the identification of novel HIPs as diabetogenic epitopes. We wish to highlight two recent studies that have effectively combined immunopeptidomics and antigen discovery approaches to identify novel autoantigens in T1D. Gonzalez-Duque et al. performed class I HLA immunopeptidomics on a human β cell line, ECN90, and on islets, and identified ~3000 peptides including native peptides, PTM epitopes, splice variants, and transpeptidation products (which are similar to HIPs, but their existence is still debated). Using synthetic peptides and peptide-MHC multimers, the authors identified several novel autoantigens including insulin gene enhancer protein ISL-1 and UCN3. T cells recognizing these autoantigens were shown to be enriched in the pancreata of T1D donors as compared with non-diabetic donors . In the second study, Wan et al. performed class II MHC immunopeptidomics on pancreatic islets and draining lymph nodes in NOD mice, and identified >4000 peptides bound to I-Ag7 . They also identified many PTM epitopes, splice variants, and HIPs. Using this immunopeptidomics dataset, our group built epitope libraries presenting >4000 epitopes in I-Ag7, and identified targets of islet-infiltrating T cells de novo, and found a predominance of HIP-reactive T cells . These studies exemplify that combining immunopeptidomics with antigen discovery can be a powerful strategy for identifying autoantigens. Systemic Lupus Erythematosus SLE is a multisystem, chronic autoimmune disease involving a complex interaction of impaired apoptotic clearance, complement activation, and immune complex formation which leads to dysregulated innate and adaptive immunity – . SLE is characterized by the presence of autoantibodies to nuclear and cytoplasmic antigens . While the importance of B cells and anti-nuclear antibodies in SLE pathogenesis is appreciated, tissue infiltrating T cells also play a key role . The antigenic landscape of autoreactive T cells in SLE is poorly defined, with a handful of known autoantigens, such as histones, described to date , . Interestingly, histones and other nuclear proteins are broadly modified post-translationally , but whether these PTMs lead to immunogenic epitopes is not known. Proteomics profiling of tissues and plasma in SLE patients and mouse models have shown changes in the soluble proteome associated with inflammation and immune dysregulation , . Antibodies to canonical autoantigens in SLE such as like Smith, RO, La, and histones, have been detected in patient sera, and serve as disease biomarkers , . First identification of T cell specificities in SLE came from curating a small list of potential autoantigens from proteomics datasets. Critically, the link between SLE proteomes and SLE immunopeptidomes is missing, largely due to the lack of immunopeptidomics data from mice or humans. SLE, unlike T1D, has a high level of heterogeneity in disease course, target tissues, and environmental triggers, therefore making it challenging to hone in on the key antigen-presenting populations. Only a small number of studies have reported immunopeptidomes in mouse models of SLE. In the early 2000s, Freed et al. characterized the peptides eluted from class II MHC from spleens in a SLE-prone mouse model, I-Ak or I-Ek alleles in MRL/lpr mice. A very small number ( < 20) peptides were detected, including some from potential SLE autoantigens such as histones . The study only uncovered a small number of peptides due to technical limitations like lower limit of detection and low abundance peptides . We have recently performed immunopeptidomics on the kidneys of MRL/lpr mice, which is the primary pathologic site in SLE. We identified >3000 epitopes presented on I-Ek in kidneys of MRL/lpr mice, and has used interaction language models to predict potential immunogens. In concordance with the previous reports, we did indeed detect peptides derived from histones and ribosomal proteins . In addition, we have developed an algorithm that will be able to predict HLA restriction of peptides previously not studied, this will advance our ability to tackle the HLA diversity. We believe that with the recent technological advances in immunopeptidomics, time is ripe to deploy it for autoantigen discovery in SLE. However, several key considerations still need to be taken into account, including the HLA diversity in humans, availability of kidney tissue from patients, and possible PTMs. Moreover, profiling the immunopeptidome will need to be followed by experimental validation of their immunogenicity in mice and humans. The application of immunopeptidomics will be essential for identifying autoantigens in SLE that will serve as diagnostic and therapeutic targets. Rheumatoid Arthritis RA is a systemic, inflammatory autoimmune disease, characterized by immune infiltration into the synovial joints, leading to varying degrees of functional impairment among patients . A prognostic hallmark of RA is the presence of autoantibodies that recognize self-proteins harboring PTMs like citrullination, homocitrullination (carbamylation), and acetylation – . Genetic association with certain HLA-DR alleles and the presence of anti-citrullinated protein antibodies (ACPAs), suggests a pathophysiological role of CD4+ T cells in disease , . While CD4+ T cell infiltration in the synovial tissue is characteristic of RA, the precise autoantigens recognized by them are poorly defined , . Most studies in RA have focused predominantly on the HLA-DR immunopeptidome, given that multiple HLA-DR molecules are strongly associated with the disease. Among the most well-described risk alleles for RA is the shared epitope (SE), a set of HLA-DRB1 alleles containing a consensus amino acid sequence in residues 70-74 of the HLA-DRβ chain . SE positivity strongly correlates with ACPA positivity and is associated with earlier onset of RA, increased disease severity, and higher mortality , . SE-containing HLA-DR alleles are thought to have enhanced presentation of arthritogenic antigens, including citrullinated peptides, leading to selection autoreactive T cell repertoires – and promotion of ACPA formation . In a 2010 study, 1427 HLA-DR-presented peptides, derived from 166 source proteins were identified in the synovia of two RA patients . Another study examining clinical samples of synovial tissue, synovial fluid mononuclear cells, and peripheral blood mononuclear cells identified 1593 peptides originating from 870 source proteins . A key mechanistic insight into how certain HLA alleles influence the genetic risk was obtained through immunopeptidomics studies comparing 962 unique peptides bound to strongly RA-associated DRB1*01:01, DRB1*04:01, and DRB1*10:01 alleles and non-RA-associated DRB1*15:01 allele . It was found that the peptide repertoires differed largely in terms of size, protein origin, composition, and affinity, with only about 10% overlap among RA-associated allotypes. Such empirical data on allelic binding preferences will enhance bioinformatics-based prediction tools that infer peptide repertoires. For example, Darrah et al. used the NetMHCII-2.3 binding prediction algorithm combined with proteolytic mapping to predict binding affinities for peptides derived from native and citrullinated antigens to RA-associated SE alleles (i.e., DRB*01:01, *04:01, *04:05, and *10:01). They demonstrated that structural changes induced by citrullination alter susceptibility to proteolytic cleavage, thus modulating antigen processing and revealing cryptic epitopes . Additionally, Kaabinejadian et al. used MHCMotifDecon on existing immunopeptidomic datasets and found that the secondary DR alleles (HLA-DRB3, DRB4, and DRB5), often overlooked due to their strong linkage disequilibrium with the primary HLA-DRB1 allele, contributed significantly to the HLA-DR repertoire. They posit that secondary DR alleles, which display non-redundant and complementary peptide repertoires, warrant regard as functionally independent alleles in future studies . This mechanistic understanding of RA susceptibility and HLA-associated variability would not have been possible without immunopeptidomics. T1D or Autoimmune Diabetes, is a chronic disease caused by insulin deficiency due to the destruction of the insulin-producing β cells in the pancreatic islets of Langerhans , . Autoantibodies against insulin, the 65-kDa form of glutamic acid decarboxylase (GAD65), insulinoma-associated protein 2 (IA-2), and zinc transporter 8 (ZnT8) are associated with T1D but their role in the pathophysiology of the disease is not clear . It has been shown that autoreactive CD4+ and CD8+ T cells infiltrate the pancreas and mediate destruction of β cells. CD4+ T cells can propagate a pro-inflammatory environment through cytokine secretion and enhancing the function of cytotoxic CD8 + T cells, which can directly kill β cells. T1D can be modeled in NOD mice, which share several key features of disease including the presence of islet autoantibodies and infiltration of autoreactive CD4+ and CD8 + T cells in islets – . Islet-infiltrating T cells in NOD mice and T1D are known to recognize β cell autoantigens, many of which are overlapping. The restricted MHC diversity as well as availability of samples and reproducibility of disease course have allowed robust immunopeptidomics studies in NOD mice , . In contrast, the high HLA heterogeneity and limited access to viable β cells have impeded similar studies in humans. Islets harvested from cadaveric donors with T1D have very few β cells remaining, and those from donors without T1D have naturally low levels of HLA expression in the absence of inflammation, making direct detection of presented peptides challenging. Moreover, even in inflamed islet, class II HLA expression on β cells is low – . Therefore, the characterization of the HLA bound peptides from human β cells has been limited , . To circumvent these issues, approaches such as stably transfected human non β cell lines expressing specific autoantigen(s) and cell surface HLA allotypes of interest , , or human β cell lines generated by targeted oncogenesis have been used. A recent finding in T1D was the presence of hybrid insulin peptides (HIPs), which consist of epitopes generated by post-translational splicing of Insulin with other proteins. Studies have shown that HIPs are autoantigens for pathogenic CD4+ T cells in the human T1D and in NOD mice , , confirming the notion that PTM epitopes are key autoantigens , . Importantly, while the initial experiments with HIPs used synthetic peptides, their presence in the proteomes and immunopeptidomes derived from β cells has reaffirmed that HIP formation and recognition is a natural process that occurs in T1D . Subsequent studies based off of these results have led to the identification of novel HIPs as diabetogenic epitopes. We wish to highlight two recent studies that have effectively combined immunopeptidomics and antigen discovery approaches to identify novel autoantigens in T1D. Gonzalez-Duque et al. performed class I HLA immunopeptidomics on a human β cell line, ECN90, and on islets, and identified ~3000 peptides including native peptides, PTM epitopes, splice variants, and transpeptidation products (which are similar to HIPs, but their existence is still debated). Using synthetic peptides and peptide-MHC multimers, the authors identified several novel autoantigens including insulin gene enhancer protein ISL-1 and UCN3. T cells recognizing these autoantigens were shown to be enriched in the pancreata of T1D donors as compared with non-diabetic donors . In the second study, Wan et al. performed class II MHC immunopeptidomics on pancreatic islets and draining lymph nodes in NOD mice, and identified >4000 peptides bound to I-Ag7 . They also identified many PTM epitopes, splice variants, and HIPs. Using this immunopeptidomics dataset, our group built epitope libraries presenting >4000 epitopes in I-Ag7, and identified targets of islet-infiltrating T cells de novo, and found a predominance of HIP-reactive T cells . These studies exemplify that combining immunopeptidomics with antigen discovery can be a powerful strategy for identifying autoantigens. SLE is a multisystem, chronic autoimmune disease involving a complex interaction of impaired apoptotic clearance, complement activation, and immune complex formation which leads to dysregulated innate and adaptive immunity – . SLE is characterized by the presence of autoantibodies to nuclear and cytoplasmic antigens . While the importance of B cells and anti-nuclear antibodies in SLE pathogenesis is appreciated, tissue infiltrating T cells also play a key role . The antigenic landscape of autoreactive T cells in SLE is poorly defined, with a handful of known autoantigens, such as histones, described to date , . Interestingly, histones and other nuclear proteins are broadly modified post-translationally , but whether these PTMs lead to immunogenic epitopes is not known. Proteomics profiling of tissues and plasma in SLE patients and mouse models have shown changes in the soluble proteome associated with inflammation and immune dysregulation , . Antibodies to canonical autoantigens in SLE such as like Smith, RO, La, and histones, have been detected in patient sera, and serve as disease biomarkers , . First identification of T cell specificities in SLE came from curating a small list of potential autoantigens from proteomics datasets. Critically, the link between SLE proteomes and SLE immunopeptidomes is missing, largely due to the lack of immunopeptidomics data from mice or humans. SLE, unlike T1D, has a high level of heterogeneity in disease course, target tissues, and environmental triggers, therefore making it challenging to hone in on the key antigen-presenting populations. Only a small number of studies have reported immunopeptidomes in mouse models of SLE. In the early 2000s, Freed et al. characterized the peptides eluted from class II MHC from spleens in a SLE-prone mouse model, I-Ak or I-Ek alleles in MRL/lpr mice. A very small number ( < 20) peptides were detected, including some from potential SLE autoantigens such as histones . The study only uncovered a small number of peptides due to technical limitations like lower limit of detection and low abundance peptides . We have recently performed immunopeptidomics on the kidneys of MRL/lpr mice, which is the primary pathologic site in SLE. We identified >3000 epitopes presented on I-Ek in kidneys of MRL/lpr mice, and has used interaction language models to predict potential immunogens. In concordance with the previous reports, we did indeed detect peptides derived from histones and ribosomal proteins . In addition, we have developed an algorithm that will be able to predict HLA restriction of peptides previously not studied, this will advance our ability to tackle the HLA diversity. We believe that with the recent technological advances in immunopeptidomics, time is ripe to deploy it for autoantigen discovery in SLE. However, several key considerations still need to be taken into account, including the HLA diversity in humans, availability of kidney tissue from patients, and possible PTMs. Moreover, profiling the immunopeptidome will need to be followed by experimental validation of their immunogenicity in mice and humans. The application of immunopeptidomics will be essential for identifying autoantigens in SLE that will serve as diagnostic and therapeutic targets. RA is a systemic, inflammatory autoimmune disease, characterized by immune infiltration into the synovial joints, leading to varying degrees of functional impairment among patients . A prognostic hallmark of RA is the presence of autoantibodies that recognize self-proteins harboring PTMs like citrullination, homocitrullination (carbamylation), and acetylation – . Genetic association with certain HLA-DR alleles and the presence of anti-citrullinated protein antibodies (ACPAs), suggests a pathophysiological role of CD4+ T cells in disease , . While CD4+ T cell infiltration in the synovial tissue is characteristic of RA, the precise autoantigens recognized by them are poorly defined , . Most studies in RA have focused predominantly on the HLA-DR immunopeptidome, given that multiple HLA-DR molecules are strongly associated with the disease. Among the most well-described risk alleles for RA is the shared epitope (SE), a set of HLA-DRB1 alleles containing a consensus amino acid sequence in residues 70-74 of the HLA-DRβ chain . SE positivity strongly correlates with ACPA positivity and is associated with earlier onset of RA, increased disease severity, and higher mortality , . SE-containing HLA-DR alleles are thought to have enhanced presentation of arthritogenic antigens, including citrullinated peptides, leading to selection autoreactive T cell repertoires – and promotion of ACPA formation . In a 2010 study, 1427 HLA-DR-presented peptides, derived from 166 source proteins were identified in the synovia of two RA patients . Another study examining clinical samples of synovial tissue, synovial fluid mononuclear cells, and peripheral blood mononuclear cells identified 1593 peptides originating from 870 source proteins . A key mechanistic insight into how certain HLA alleles influence the genetic risk was obtained through immunopeptidomics studies comparing 962 unique peptides bound to strongly RA-associated DRB1*01:01, DRB1*04:01, and DRB1*10:01 alleles and non-RA-associated DRB1*15:01 allele . It was found that the peptide repertoires differed largely in terms of size, protein origin, composition, and affinity, with only about 10% overlap among RA-associated allotypes. Such empirical data on allelic binding preferences will enhance bioinformatics-based prediction tools that infer peptide repertoires. For example, Darrah et al. used the NetMHCII-2.3 binding prediction algorithm combined with proteolytic mapping to predict binding affinities for peptides derived from native and citrullinated antigens to RA-associated SE alleles (i.e., DRB*01:01, *04:01, *04:05, and *10:01). They demonstrated that structural changes induced by citrullination alter susceptibility to proteolytic cleavage, thus modulating antigen processing and revealing cryptic epitopes . Additionally, Kaabinejadian et al. used MHCMotifDecon on existing immunopeptidomic datasets and found that the secondary DR alleles (HLA-DRB3, DRB4, and DRB5), often overlooked due to their strong linkage disequilibrium with the primary HLA-DRB1 allele, contributed significantly to the HLA-DR repertoire. They posit that secondary DR alleles, which display non-redundant and complementary peptide repertoires, warrant regard as functionally independent alleles in future studies . This mechanistic understanding of RA susceptibility and HLA-associated variability would not have been possible without immunopeptidomics. Overall, immunopeptidomics has provided valuable insights into the pathogenesis of autoimmune diseases and has helped to define key autoantigens. While immunopeptidomics studies in autoimmune diseases are often not directly translatable, they have tremendous potential to fuel diagnostic and therapeutic approaches. Generation of large datasets of MHC/HLA-bound peptides have led to the advances in computational tools to predict epitope binding , , . These algorithms can be then used to predict potential autoantigens. Immunopeptidomics datasets have directly fueled systematic antigen discovery approaches which are leading to using antigen-specific TCR repertoires as diagnostic tools. Moreover, novel autoantigens can be directly used as immunogens to induce tolerance or as targets for immunomodulatory strategies. Furthermore, immunopeptidomics has revealed mechanistic aspects of development of autoimmunity, such as cross-reactivity with microbial epitopes, changes in immune landscapes associated with disease states, and PTM-driven alterations in HLA binding and T cell recognition. Identifying the peptides that are presented on MHC/HLA has led to several key advances in autoimmunity including autoantigen identification, confirmation of the biological relevance for HIPs, and to an increasing appreciation for PTM epitopes as pathogenic. Several key considerations remain in the design and utility of immunopeptidomics studies, as highlighted in Fig. . As the experimental techniques to elute and detect peptides and computational tools to analyze datasets advance, we envision that the numbers of identified peptides will continue growing exponentially. This will enhance our understanding of the pathogenic antigens across autoimmune disorders. In summary, as immunopeptidomics opens this chamber of secrets, a treasure trove of autoantigens will be discovered. |
Enzyme-labeled liquid-based cytology (ELLBC): a new noninvasive diagnostic method for bladder cancers | 9c5be67f-f84c-4965-9015-5029d54e7ece | 10978622 | Pathology[mh] | Uroepithelial carcinoma (UC) is the most common type of bladder cancer, accounting for 90% of cases. It is a prevalent malignancy affecting the urinary tract. According to the American Cancer Society's Annual Review of Cancer Statistics 2022, bladder cancer ranks fourth in terms of the number of new cases among men and eighth in terms of the number of deaths among all cancers (Siegel et al. ). Cystoscopy is considered the gold standard for diagnosing and monitoring bladder cancer. The diagnosis of bladder cancer relies on cystoscopy and histologic evaluation of tissue samples (Babjuk et al. ). However, cystoscopy is an invasive procedure that often leads to poor patient compliance. It can also cause discomfort, anxiety, and concerns about disease progression (Koo et al. ). Biomarkers have been integrated into clinical decision-making for various cancers, including breast cancer (Chan et al. ). Consequently, there has been significant research interest in noninvasive urine tests with high diagnostic accuracy (Miyake et al. ; Costantini et al. ). However, despite the use of biomarkers, achieving the desired diagnostic rate remains a challenge. Therefore, cytology continues to be the gold standard for bladder cancer surveillance in many medical practices (Ng et al. ). Routine urine cytology has a sensitivity of 48% (16% for low-grade and 84% for high-grade) and a specificity of 86% (Yafi et al. ). However, these results may be influenced by the differentiated morphology and number of cells in the sample, and the test has lower sensitivity for cancers with low-grade superficiality, indicating certain limitations. ELLBC utilizes enzyme histochemistry based on LBC to identify abnormally elevated acid phosphatase in bladder cancer cells. This method differs from traditional Papanicolaou staining and is currently employed for detecting exfoliated cells in pleural and abdominal fluid, as well as pericardial effusion (Li et al. ). ELLBC was developed to address the limitations of traditional smears, such as cloudy cells and blood contamination. In this study, we applied ELLBC for early noninvasive diagnosis of bladder cancer in a clinical setting and further validated its efficacy through a prospective cohort study. Additionally, we systematically evaluated the performance of this technique in diagnosing UC and compared it with routine urine exfoliative cytology. Patients and samples Fifty patients who were initially diagnosed with suspected bladder cancers, based on symptoms such as hematuria or bladder irritation and urinary ultrasound suggestive of bladder mass, were selected from the Second Affiliated Hospital of Anhui Medical University (Anhui, China) between January 2022 and December 2022 (Patients with a combination of other urologic tumors and a combination of severe infectious diseases were excluded). Non-morning urine samples were collected from the patients, and ELLBC and CC tests were performed. Before participating in the study, all patients signed a consent form that allowed clinicians to access their medical records for clinical information. The study protocol was approved by the Ethics Committee of the Second Affiliated Hospital of Anhui Medical University (Ethics Approval No. YX2022-064). The process is illustrated in Fig. . Urine collection and preparation staining After admission, 100 mL of fresh urine (non-morning urine) was collected from the patients using enzyme-free EP tubes for examination by ELLBC and CC, respectively. Only one urine sample is required for ELLBC and three samples are required on three consecutive days for CC. The CC slide preparation involved double centrifugation of the sediment obtained from the urine samples, followed by Papanicolaou staining. The slides were then sent to the pathologist for microscopic reading after completion of the smear and transparency operations. For ELLBC, 100 mL of collected fresh urine was centrifuged at 2000 rpm for 10 min, washed with enzyme-labeled liquid-based cytological preservation solution, and prepared until the cells were partially dry. The partially dry cell smear was then fixed in a fixing solution at room temperature for 4 min to remove red blood cells, rinsed with tap water for 2 min, placed in a staining solution of acid phosphatase at 37 °C for 5–10 min, rinsed with running water for 2 min, dyed with nuclear redyeing solution for 5 min, rinsed again with running water for 2 min, and finally sealed with neutral gum after thorough air drying. The cytology reports were completed using the Paris system and the results were categorized as positive or negative. Positive results indicated confidence or suspicion of malignancy, while negative results indicated findings other than malignancy, such as inflammation or benign conditions. The final diagnosis of the patient was based on the histopathological findings. The dyeing process of ELLBC is shown in Figure , and the composition of the acid phosphatase stain solution is described in Table . The cytology reports were completed using the Paris system, with results categorized as positive or negative. Positive results indicated confidence or suspicion of malignancy, while negative results indicated findings other than malignancy, such as inflammatory or benign conditions. In all cases, the diagnosis was confirmed by the same pathologist, and the final diagnosis of the patient was based on the histopathological findings, which served as the gold standard. Figure shows a representative image of the staining. Data analysis Pathologic findings, ELLBC findings and CC findings were collected for 50 patients. The diagnosis of ELLBC relies on morphologic observation and specific red staining of the cytoplasm, and the diagnosis of CC can only be determined by observation of the nucleoplasmic ratio of the cells. The results were considered positive if abnormal cells were observed in one sample in the ELLBC group and one of three samples in the CC group. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy of ELLBC and CC tests in diagnosing bladder cancer were calculated against the pathologic findings, which were considered the gold standard. The calculation formula can be found in Supplementary Table S2. Additionally, the sensitivities of the two cytologic techniques were determined for different grades of uroepithelial carcinoma in the bladder. The chi-square test was used to analyze the differences in underestimation and overestimation rates between CC and ELLBC in detecting lesions. Statistical analysis was performed using SPSS (version 26, SPSS Inc., Chicago, IL, USA). A significance level of p < 0.05 was used to determine statistical significance. Fifty patients who were initially diagnosed with suspected bladder cancers, based on symptoms such as hematuria or bladder irritation and urinary ultrasound suggestive of bladder mass, were selected from the Second Affiliated Hospital of Anhui Medical University (Anhui, China) between January 2022 and December 2022 (Patients with a combination of other urologic tumors and a combination of severe infectious diseases were excluded). Non-morning urine samples were collected from the patients, and ELLBC and CC tests were performed. Before participating in the study, all patients signed a consent form that allowed clinicians to access their medical records for clinical information. The study protocol was approved by the Ethics Committee of the Second Affiliated Hospital of Anhui Medical University (Ethics Approval No. YX2022-064). The process is illustrated in Fig. . After admission, 100 mL of fresh urine (non-morning urine) was collected from the patients using enzyme-free EP tubes for examination by ELLBC and CC, respectively. Only one urine sample is required for ELLBC and three samples are required on three consecutive days for CC. The CC slide preparation involved double centrifugation of the sediment obtained from the urine samples, followed by Papanicolaou staining. The slides were then sent to the pathologist for microscopic reading after completion of the smear and transparency operations. For ELLBC, 100 mL of collected fresh urine was centrifuged at 2000 rpm for 10 min, washed with enzyme-labeled liquid-based cytological preservation solution, and prepared until the cells were partially dry. The partially dry cell smear was then fixed in a fixing solution at room temperature for 4 min to remove red blood cells, rinsed with tap water for 2 min, placed in a staining solution of acid phosphatase at 37 °C for 5–10 min, rinsed with running water for 2 min, dyed with nuclear redyeing solution for 5 min, rinsed again with running water for 2 min, and finally sealed with neutral gum after thorough air drying. The cytology reports were completed using the Paris system and the results were categorized as positive or negative. Positive results indicated confidence or suspicion of malignancy, while negative results indicated findings other than malignancy, such as inflammation or benign conditions. The final diagnosis of the patient was based on the histopathological findings. The dyeing process of ELLBC is shown in Figure , and the composition of the acid phosphatase stain solution is described in Table . The cytology reports were completed using the Paris system, with results categorized as positive or negative. Positive results indicated confidence or suspicion of malignancy, while negative results indicated findings other than malignancy, such as inflammatory or benign conditions. In all cases, the diagnosis was confirmed by the same pathologist, and the final diagnosis of the patient was based on the histopathological findings, which served as the gold standard. Figure shows a representative image of the staining. Pathologic findings, ELLBC findings and CC findings were collected for 50 patients. The diagnosis of ELLBC relies on morphologic observation and specific red staining of the cytoplasm, and the diagnosis of CC can only be determined by observation of the nucleoplasmic ratio of the cells. The results were considered positive if abnormal cells were observed in one sample in the ELLBC group and one of three samples in the CC group. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy of ELLBC and CC tests in diagnosing bladder cancer were calculated against the pathologic findings, which were considered the gold standard. The calculation formula can be found in Supplementary Table S2. Additionally, the sensitivities of the two cytologic techniques were determined for different grades of uroepithelial carcinoma in the bladder. The chi-square test was used to analyze the differences in underestimation and overestimation rates between CC and ELLBC in detecting lesions. Statistical analysis was performed using SPSS (version 26, SPSS Inc., Chicago, IL, USA). A significance level of p < 0.05 was used to determine statistical significance. Pathological diagnosis showed that there was a significant difference in the detection rate of bladder tumors between ELLBC and CC groups (Fig. ). Pathological diagnosis showed that there was a significant difference in the detection rate of bladder tumors between ELLBC and CC groups; There was no significant difference in the detection of high-grade tumors between the two methods (Table A). The sensitivity of ELLBC was 82.86%, the specificity was 93.33%, the PPV was 96.67%, the NPV was 70.00%, and the accuracy was 86.00%; CC had a sensitivity of 37.14%, specificity of 80.00%, PPV of 81.25%, NPV of 35.29%, and accuracy of 50%; ELLBC combined with CC had a sensitivity of 88.57%, specificity of 73.33%, PPV of 88.57%, NPV of 73.33%, and accuracy of 84.00% (Table B). For clinical staging, ELLBC and CC showed a significant difference in diagnosis between stages I and II, but not in stages III and IV (Table C). We collected the diagnostic results of ELLBC and CC and plotted the Receiver Operating Characteristic (ROC) Curve of ELLBC, the resulting area under the curve was 0.881 for ELLBC (95% CI = 0.776–0.986), 0.586 for CC (95% CI = 0.417–0.754), and 0.810 for ELLBC combined with CC (95% CI = 0.663–0.956) (Fig. ). Smears prepared by ELLBC and CC The study's findings revealed that in the urine cytology of patients with chronic haematuria, the CC preparation's background was more muddy (Fig. ). The study's findings revealed that in the urine cytology of patients with chronic haematuria, the CC preparation's background was more muddy (Fig. ). Bladder cancer screening and post-treatment surveillance still rely on conventional urine cytology combined with cystoscopy, but cystoscopy shows low sensitivity in the face of low-grade uroepithelial cancers and small lesions (Miyake et al. ), whereas the diagnostic sensitivity and specificity of conventional urine cytology are low and show a large discrepancy between the detection of low-grade and high-grade uroepithelial cancers (Babjuk et al. ). The obvious disadvantage of bladder cancer surveillance is the expensive financial burden, which accounts for 5.78% of all out-of-pocket costs for cancer patients in the United States (Schafer et al. ),in the European Union this value is 3% (Leal et al. ). Therefore, an increase in the diagnostic rate of cytology is of significant significance for the reduction of healthcare costs for bladder cancer patients. The most criticized drawback of conventional urine cytology is its low detection rate (20–50%) (Yafi et al. ), and similar results were obtained in our study, which were more pronounced in low-grade uroepithelial carcinomas, which may be attributed to the fact that low-grade cancers have a similar cellular morphology to the normal uroepithelium (Barkan et al. ). Another possible cause of the CC detection rate is cellular or blood contamination; urine collected from patients with persistent hematuria has lower sensitivity for cytology and the presence of a large number of atypical cells that need to be interpreted (Tan et al. ; Konety et al. ), which adversely affects the interpretation of malignant cells,urine samples accompanied by a large amount of inflammatory infiltrate, cellular degeneration, or heavy bleeding may lead to overdiagnosis or an underdiagnosed basis for the diagnosis (Sweeney et al. ). In contrast, liquid-based cytology preparation can exclude blood cell interference (Remmerbach et al. ; Xu et al. ) and obtain a clearer background (Chou et al. ), and our use of acid phosphatase as an enzyme marker also yielded a more favorable result for the interpretation of malignant cells, as shown in Fig. , which shows that ELLBC has a clear advantage over CC in the diagnosis of patients with persistent hematuria. Six urinary biomarkers approved by the U.S. Food and Drug Administration for the diagnosis or surveillance of bladder cancer have been evaluated in recent studies: quantitative or qualitative nuclear matrix protein 22 (NMP22), qualitative or quantitative bladder cancer antigen (BTA), fluorescence in situ hybridization (FISH), and fluorescence immunohistochemistry (ImmunoCyt), with a sensitivity range of 0.57–0.82, and specificity ranging from 0.74 to 0.88 (Tabayoyong and Kamat ; Goutas et al. ). The sensitivity of ELLBC reached the upper limit of conventional assays, and the specificity was superior to conventional assays. Enzyme histochemical staining is a method to show the activity and location of endogenous enzymes in tissues or cells on slices or smears by using the characteristics of intracellular enzymes to catalyse substrates. Its characteristics for in situ detection enzyme expression and enzyme activity rather than itself. As early as 1952, the histochemistry of alkaline phosphatase was confirmed by azo-dye coupling method (Grogg and Pearse ). Related studies in the past few years have shown that enzyme histochemistry staining can assist in monitoring the therapeutic effect of radiofrequency ablation for primary breast cancer (Seki et al. ). The color reaction of azo dye can obtain clearer antigen cell localization than immunofluorescence technology. Therefore, immunostaining with enzyme-labeled probes can be used in complex situations that require dual antigen localization (Tsutsumi ). In urine cytology, it may be simpler and of higher economic value to use azo dye-coupled method to monitor abnormally elevated acid phosphatase in bladder cancer cells. Although ELLBC had better sensitivity when combined with CC, its area under the curve was still smaller than that of ELLBC, and ELLBC yielded better specificity when compared with CC and sensitivity, and better sensitivity was obtained when the two were combined for monitoring, but the overall efficiency was still not as good as with ELLBC alone. Our study ultimately collected urine samples from 50 patients, however, the cytological staining of the enzymatic labeling method has some limitations of its own, i.e., better results can be obtained by reading the film after the staining is completed, i.e., after the staining of the preparation is completed and stored at room temperature, the smears may show varying degrees of discoloration after 1–3 months, which may appear to have some impact on the determination of the results. The final number of cases we included was only 50, which led us to realize that the clinical stages were mostly Stage I and Stage II, and a small number of patients were postoperative recurrences. This demonstrates the excellence of ELLBC performance in early screening. We will expand the caseload and focus on patients with different clinical stages and postoperative recurrence in subsequent studies to improve the applicability of ELLBC in clinical diagnosis. One notable constraint of our study is its non-randomized trial design, which may have resulted in a lack of standardization regarding the number of passes for each method and a potential bias towards the smear method. Consequently, careful interpretation of the findings is warranted. Additionally, the study is limited by its small sample size and the absence of an on-site cytopathologist to evaluate sample adequacy during collection. Technical difficulties during sample preparation may have contributed to cellular issues, suggesting that ELLBC could offer improved applicability. In conclusion, the ELLBC technique has a high sensitivity to bladder malignant cancers, and the application of this technique contributes to the early screening of bladder cancer, which may compensate for the low detection rate of conventional cytology. Cytology is a simple and inexpensive technique that can be used as an effective diagnostic tool for early screening of bladder cancer. The findings of this study demonstrate the potential value of enzyme-labeled liquid-based cytology technology as a noninvasive diagnostic tool for bladder cancer. However, it is important to note that there are still significant opportunities for further advancements in the cytologic diagnosis of bladder cancer. It is anticipated that this research will contribute to the enhancement of clinical decision-making in this field. Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 181 kb) |
Primary Care Clinicians as COVID-19 Vaccine Ambassadors | 1e9cfe88-db63-4cae-9a09-c3a926292fe6 | 8772350 | Health Communication[mh] | Vaccine hesitancy was a significant problem in the US and throughout the world-even before the recent spread of COVID-19 misinformation. In fact, the World Health Organization (WHO) listed “vaccine hesitancy” as 1 of 10 threats to global health in 2019 because of a significant increase in measles cases, coupled with the threats to international campaigns to immunize against cervical cancer and also to finally end the spread of polio in Afghanistan and Pakistan. It is not surprising that many patients are wary about receiving the COVID-19 vaccine. The many common concerns include: (1) safety and efficacy concerns related to the “warp speed” of vaccine development for this novel virus, (2) the possibility of short and long term side effects, (3) unclear long term efficacy, (4) minority populations not being an integral part of the clinical trials, (5) historical medical experimentation on minority populations, and (6) the politicization of the COVID-19 virus. As part of preventive maintenance, pediatricians (and other primary care clinicians treating children) consider vaccinations to be a major component of their practice. The measles, mumps, rubella (MMR), and polio vaccine for instance, not only protect the individual patient and his/her family, but the public’s health at large. In one large study, most parents considered their children’s pediatricians to be the best source of vaccination information. Primary care clinicians who mainly care for adults on the other hand, may spend much of their medical practice focused on chronic disease management and the treatment of pain. They recommend annual Influenza vaccines and various boosters, as well as the Zoster, Pneumococcal and Haemophillus influenza type B vaccines. Although extremely important to an individual patient, the public health risk of a single patient unwilling to be vaccinated with one of these vaccines will not risk the public’s health. Primary care clinicians treating adults have little training in communication regarding vaccination aimed to develop herd immunity. After all, there was no vaccine for the Spanish Flu in 1918 and today’s clinicians were not practicing medicine in 1955 when the Salk polio mass vaccination program began in the US. The roll out of the H1N1 vaccine in 2009 was considerably smaller than today’s COVID-19 vaccination efforts but suffered from similar supply shortages and miscommunication. Clinicians in the US have been stretched to the limits in every way imaginable, and now they are met with another moral imperative: to ask their patients about their plan for obtaining the COVID-19 vaccine. This might include: (1) empathic listening and kindness to their patients (beneficence), (2) communicating the risk of not obtaining a COVID-19 vaccine (non-maleficence), (3) allowing their patients to make an informed decision without judgment (autonomy), and (4) ensuring that the COVID-19 vaccine is distributed equitably (justice). This is a tall order for clinicians, but the public will surely benefit as clinicians provide the necessary space for this conversation.
Historically, primary care clinicians have had difficulty communicating to their patients about controversial and potentially emotional topics. While some have participated in residency and post-licensure offerings in “Breaking Bad News,” these programs are usually targeted to oncologic and other sub-specialty clinicians. Suicidality, intimate partner violence, substance use, sexual health, and gender identity are some examples of topics that clinicians receive little training on relative to the importance of these conversations. , The risk of poor clinician-patient communication may result in less than optimal outcomes for the individual and/or family unit. Now, clinicians are asked to talk to their patients about issues that affect the patient, the public health and even the health of the planet. These include gun safety, climate change, and health and now the importance of the COVID-19 vaccine.
Because patients have their own personal narrative (including fears, disbelief, mistrust, anxiety, etc.) regarding the COVID-19 vaccination, it is important for clinicians to learn how to communicate with a variety of patients. In fact, there are strategies for primary care clinicians to consider in the delivery of communication regarding vaccination. Many patients are hyperaroused and anxious due to the impact of the pandemic, and a discussion of the vaccine is best delivered in a reassuring tone aimed at helping a patient to see that the vaccine is a mechanism toward calm and not something additional to worry about. Conversely, for those who are detached or pessimistic, it is important to speak to the potential dangers of worldwide under vaccination and the consequences if citizens do not do their part in reaching herd immunity through vaccination. For patients who are passive as a result of pandemic forces, it is important to actually introduce anxiety into the situation to trigger them toward taking action.
There are many possible solutions regarding how primary care clinicians might quickly learn how to effectively communicate with their patients regarding their COVID-19 vaccination and thereby alleviate any fears and distrust that they may have. Project ECHO, a virtual telementoring network, is an example of a synchronous educational training model that could be used to teach clinicians regarding COVID-19 vaccine communication. For instance, clinicians can attend a one-hour Vaccine Communication ECHO training session to receive guidance from expert clinician communicators to understand how to navigate difficult conversations with their patients. Project ECHO is currently piloting a similar communication ECHO on Climate Change and Human Health to educate clinicians regarding how to communicate with their patients on the interrelatedness of the warming climate and health. Large organizations might also consider placing a short series of videos on their asynchronous “Learning Central” sites so that clinicians can watch scenarios of patient communication encounters regarding vaccination hesitancy. Smaller ambulatory clinics might decide to hold in-person or virtual round-table discussions to communicate the importance of COVID-19 vaccination with their patients to ensure the health of the clinic as well as the surrounding community. Finally, written educational material can be prepared by large medical organizations, such as the American Academy of Family Physicians, the American College of Physicians and the Centers for Disease Control and Prevention regarding COVID-19 vaccination communication strategies for patients. It is important to understand the potential impact that vaccine communication through daily clinical practices may have. While a multi-pronged approach to encourage vaccination is important, and would likely include social media, the role of important societal influencers, and the prosocial communication of returning to pre-pandemic experiences, the literature points to the critical importance of the clinical encounter. Throughout this pandemic, clinicians, and patients have been navigating this novel virus together, however, bringing a successful end to this pandemic will take enormous collaboration across the global community. Primary care clinicians can be trusted ambassadors for encouraging their patients to obtain the COVID-19 vaccine, but they would benefit greatly from receiving a brief training in vaccination communication skills. Clinicians can successfully act locally in their communities while understanding their value in the global campaign to combat this pandemic. Perhaps soon, we can remove vaccine hesitancy from the WHO’s list of top ten global threats and develop herd immunity with the COVID-19 vaccines. Today is the perfect opportunity for primary care clinicians and patients to work together towards the goal of eradication of the COVID-19 virus and a return to some sense of normalcy.
|
Myco-fabricated silver nanoparticle by novel soil fungi from Saudi Arabian desert and antimicrobial mechanism | 23d4768e-f471-46f0-b680-8f2e54212aac | 11220002 | Microbiology[mh] | A great concern has been given to the potential uses of the metal nanoparticles (MN) that have size in the range 1–100 nm, due to their significant optical , physical , and chemical characteristics in relation to their bulk origin. Such unique properties facilitated their applications in various fields of science and technology. In the medical and food sectors, for both biosensing and microbial detection, MN could be ideal because of their biocompatibility, ability in electron transfer, and reaction catalysis , beside other applications in diagnosis and therapy. MN could rapidly detect the microbes and tumor cells disease-related biomarkers, however, they also demonstrate capability for single-cell evaluation . Biomedical and biotechnological applications are the cornerstone for MN usage. Some of their important biomedical applications, but not limited to, are anti-inflammatory, antioxidant, antibacterial, antifungal, cytotoxic, drug delivery and detecting agents , . However, bioremediation, biodeterioration and wastewater treatment are their most known biotechnological applications , . Among the MN, silver nanoparticles (AgNPs) attracted a great interest due to their chemical, physical and biological characteristics , . AgNPs have high ability against microbes, and cancer cells besides, they have been used also in wound dressings, diagnostics, catalysis, biosensors and drug delivery – . Alongside the aforementioned biomedical applications, AgNPs can be applied in different biotechnology lines, such as food preservation and water filtration, in addition to spray and sanitization , . With the emergence of antimicrobial resistance, the need for alternative antimicrobial agents to replace the conventional ones enhanced researchers to develop AgNPs as one of the good options for microbial mitigation. Previous studies have approved the ability of AgNPs as antimicrobial agents with significant effects against some multidrug- resistant bacteria , , . The unique and wide range of AgNPs applications are the factors that led to the development of various fabrication methods for efficient and safe usage – . Although chemical and physical methods are the best ways to develop nanoparticles with unique size and shape in a short time, however, their weakness points are the high cost and the production of hazardous by-products , . Therefore, the green route could be a good alternative because it is an easy, applicable, and eco-friendly approach. Such route includes using microbes and plants, beside other biomolecules as biogenic agents , . Stable nanoparticles with better distributed size and morphology were noted from the biogenic approach in relation to those produced by other methods . The molecules from biogenic agents that may cap the MN enhance their stability and dispersity . Generally, nanoparticles produced with the aid of biological routes are harmless, environmentally benign that can be used in agriculture, medicine, biochemical detection and related fields. On this regard, the use of microbes as one approach of the biogenic synthesis is gaining great interest. From which fungi can be used for AgNPs development by extra or intracellular approaches however, the extracellular processis easier and faster than the intracellular one . Previous studies demonstrated extracellular approach for AgNPs fabrication using fungi , Employment of fungi is an important fabrication approach due to their ability to produce greater amounts of proteins in relation bacteria and their high intracellular metal uptake and binding capacity . In this respect, it is expected that, fungi from extreme conditions might have unique compounds that enhance their ability to withstand harsh environments, but also could help efficiently in AgNPs production. Therefore, the current study was designed to provide new candidates in AgNPs fabrication as safe approach in providing high stable and active materials. Two fungal species ( Embellisia spp. and Gymnoascus spp.) isolated from the desert soils in Saudi Arabia, not addressed before, were selected. Hence, present study represents the first report on these two species as biogenic agents in extracellular AgNPs fabrication using fungal filtrate as source of their biomolecules. Thereafter, the fabricated AgNPs were characterized using UV, DLS, TEM, SEM, and FTIR and tested against some bacterial pathogenic strains. In a trial to detect the possible antibacterial mechanism of the myco-fabricated AgNPs, treated bacteria were investigated using TEM and SDS-PAGE analysis.
Materials The laboratory experiments have been done at the faculty of science laboratory, Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia. Where the silver nitrate, potato-dextrose agar (PDA), Czapek-Dox Agar with 1% chloramphenicol, nutrient agar and broth and Sabouraud broth (Oxoid, UK) were obtained. The soil fungi ( Embellisia spp. and Gymnoascus spp.) have been isolated from Saudi Arabian desert soils. Protein profile and TEM analysis were done at king Saud university laboratory. Isolation of fungi Two fungal strains ( Embellisia spp. and Gymnoascus spp.) were isolated from ALkharj and ALQasab respectively, Saudi Arabian desert soils. The soil was collected from 5 to 20 cm depth, then subjected to culture following the dilution plate method, as described elsewhere . Two culture media were used: PDA and Czapek-Dox Agar with 1% chloramphenicol. The plates were incubated for 7 days at 28 °C. Pure fungal colonies were kept at 4 °C for further analysis. Molecular identification using 18S rRNA gene. A PCR assay for partial amplification of the fungal 18S rRNA genes was performed as described previously with the primers NS1 F (5′ GTAGTCATATGCTTGTCTC 3′) and NS8 R (5′ TCCGCAGGTTCACCTACGGA 3′). The PCR reaction mixture contained 2 µL of 10 × PCR Buffer, 1.6 µL of 2.5 mM dNTPs, 1.0 µL of 10 pmol/µL from each primer, 0.2 µL of KOMA Taq (2.5 U/µL), 2 µL of 20 ng/µL DNA template and HPLC-grade water to adjust the reaction volume to 20 µL. The amplification reactions were done in a Biometra thermal cycler (Analytik Jena, Jena, Germany) with the reaction conditions mentioned elsewhere . Confirmation of successful amplification was assured by applying the PCR products to agarose gel electrophoresis. The PCR products were purified by PCR Cleanup Kit (MilliporeSigma, Burlington, MA, USA). Sequence analysis The purified PCR products were sent to a commercial company (Marogen Europe, Amsterdam, The Netherlands) for partial sequencing of the 18S rDNA using the same amplification primers. Sequencing at the company was performed by the BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, USA) and 3730xl DNA Analyzer automated DNA sequencing system (Applied Biosystems). The retrieved 18S rDNA sequences of the two fungal isolates were analyzed using the software Geneious Prime Version 2020.1.2 . Consent sequences were produced from forward and reverse sequences. The edited sequences were compared with other sequences of related reference strains in nucleotide database of the National Center of Biotechnology Information (NCBI) of the United States using the nucleotide platform of the Basic Local Alignment Search Tool (BLASTn). The sequences were also used to construct a phylogenetic tree using the Neighbor-Joining method with the software MEGA X . Fungal culture filtrate Separate culture flasks containing 500 mL of Sabouraud broth (Oxoid, UK) were inoculated by each fungal species isolate individually. The flasks were incubated in static incubator for 7 days at 28 °C, after which the fungal biomass was collected by filtration through Whatman filter paper No. 1 and washed with sterile distilled water to remove excess medium. The fungal biomasses were weighed, and 0.5 g were added to distilled water (10 mL) and incubated at 28 °C for 72 h. Finally, the biomass was separated from the filtrate by filtration. The aqueous filtrate containing the fungal metabolites was kept for further usage in the refrigerator at 4 °C . Myco-synthesis of AgNPs For the biosynthesis of AgNPs, 1 mM AgNO 3 was added to fungal aqueous filtrate at a ratio 1:1 and then boiled for 30 min. This mixture was incubated under natural sun light at 25 °C for 24 h, till a dark brown color was formed. The nanoparticles were separated by centrifugation of the mixture at 14,000 rpm for 15 min, then dried by dropping the precipitate on petri dishes at room temperature and kept for further analysis . Characterization of myco-synthesized AgNPs AgNPs was re-dispersed in 1 mL of distilled water for chracterization. A UV–visible spectrophotometer (BIOCHROM Libra S60PC, England) was used in the range of 200–600 nm for measuring the absorbance of the AgNPs with the fungal aqueous filtrate was used as blank, as described before . The morphology, average diameter and distribution were estimated by transmission electron microscopy, TEM (JEM-1011, JEOL, Tokyo, Japan) at 80 kV voltage. The pattern of the size distribution (size and poly dispersed index, PDI) and zeta potential were assessed by a dynamic light scattering (DLS) system by a Zetasizer (NANO ZSP, Malvern Instruments Ltd., SerialNumber: MAL1118778, version 7.11, Malvern, UK). For surface analysis of the synthesized NPs, a scanning electron microscope, SEM (JED-2200 series, JEOL) provided by energy-dispersive X-ray spectroscopy (EDX) was used to confirm the presence of elemental silver, as previously described . Functional groups of fungal aqueous filtrates responsible for AgNO 3 reduction were estimated with Fourier transform infrared (FT-IR) spectroscopy (SPECTRUM100, Perkin-Elmer, Wellesley, MA, USA) with a range 450–3500 cm −1 , as detailed earlier . Investigation of the antibacterial activity of myco-synthesized AgNPs To assess the antibacterial potential of the myco-synthesized AgNPs in vitro, four human pathogenic bacteria were tested ( Escherichia coli, Pseudomonas aeruginosa Staphylococcus aureus, and Klebsiella pneumoniae ). The assay was conducted by the agar well diffusion technique . Using sterile swabs, 100 µL of 24 h. old bacterial culture with a concentration equal to 0.5 McFarland (1.5 × 10 8 CFU mL −1 ) from each of the four tested strains were streaked on a plate of nutrient agar and left to dry. Under aseptic conditions, wells of 0.8 mm in diameter were made on the agar using a sterile cork-borer. Volumes of 50 µL of myco-synthesized AgNPs were added to the wells. The fungal filtrate and 5 µg ampicillin discs were used as negative and positive controls, respectively. The plates were incubated at 37 °C for 24 h. The experiment was performed in triplicate and the growth inhibition zones were estimated in millimeters (means ± standard variation), as described previously . The minimum inhibitory concentration (MIC) and the minimum bactericidal concentration (MBC) were estimated using the microdilution method in nutrient broth. The bacterial cultures of the four tested strains were individually diluted to 0.5 McFarland from which a volume of 10 µL was added to nutrient broth (350 µL). The isolates were tested against the myco-fabricated AgNPs solubilized in Dimethyl sulfoxide (DMSO) at different concentrations (0.031, 0.062, 0.125 and 0.250 mg mL −1 ). The negative control composed of nutrient broth inoculated with bacterial inoculum. Then, the plates were incubated at 37 °C for 24 h. The MIC was determined as the lowest concentration that inhibit observable bacterial growth where the MBC was verified when no growth appeared when treated bacteria were sub-cultured on nutrient agar plates after incubation period . Protein profile pattern by SDS-PAGE analysis The total cellular soluble proteins of K. pneumoniae were extracted before and after treatment with 1 mg mL −1 of AgNPs synthesized from each fungal strain for 24 h at 37 °C, then purified by TriFast (Peqlab, VWR company). The proteins were further fractionated by OmniPAGE Mini vertical electrophoresis unit provided with a power Pro 5 power supply (Cleaver Scientific, Warwickshire, UK) on a SERVAGel™ TG PRiME 10% (SERVA, Heidelberg, Germany) . Detection of ultrastructural changes in AgNPs treated bacteria One species ( K. pneumoniae ) was chosen for TEM analysis (JEM-1011, JEOL, Tokyo, Japan, at 80 kV voltage) to test the possible structural changes for cells treated with AgNPs fabricated by both fungal strains and incubated for 24 h. at 37 °C. Thereafter, bacterial suspension was centrifuged at 4000 rpm for 15 min. Then, 2.5% glutaraldehyde in 100 mM phosphate buffer, pH 7.0, was added to the pellet for sample fixation for 15–30 min. This was followed by applying 1% osmium tetroxide in 100 mM phosphate buffer for 1–2 h. at 4 °C, then washed five times with distilled water. Then, En bloc stain with 2% aqueous uranyl acetate was applied for ~ 2 h at 4 °C in the dark. Dehydration was done with series of acetone (30, 50, 70, 90 and 100%). After that, the samples were infiltrated in propylene oxide, followed by embedding in fresh resin (Epon mixture) for 12–24 h. at 60–70 °C. The ultrathin sections of the samples were put on 200-mesh copper grids after double staining with 2% uranyl acetate and lead citrate, then examined under a JEOL 100 CX electron microscope operating at 80 kV, as described previously . Statistical analysis All data were represented as mean and standard deviations. One-way analysis of variance (ANOVA) as well as graphs preparation were performed by Graph-bad Prism 9.1 software (Inc., La Jolla, CA, USA). ImagJ software was used for calculating the particles size and diameter from TEM images.
The laboratory experiments have been done at the faculty of science laboratory, Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia. Where the silver nitrate, potato-dextrose agar (PDA), Czapek-Dox Agar with 1% chloramphenicol, nutrient agar and broth and Sabouraud broth (Oxoid, UK) were obtained. The soil fungi ( Embellisia spp. and Gymnoascus spp.) have been isolated from Saudi Arabian desert soils. Protein profile and TEM analysis were done at king Saud university laboratory.
Two fungal strains ( Embellisia spp. and Gymnoascus spp.) were isolated from ALkharj and ALQasab respectively, Saudi Arabian desert soils. The soil was collected from 5 to 20 cm depth, then subjected to culture following the dilution plate method, as described elsewhere . Two culture media were used: PDA and Czapek-Dox Agar with 1% chloramphenicol. The plates were incubated for 7 days at 28 °C. Pure fungal colonies were kept at 4 °C for further analysis.
A PCR assay for partial amplification of the fungal 18S rRNA genes was performed as described previously with the primers NS1 F (5′ GTAGTCATATGCTTGTCTC 3′) and NS8 R (5′ TCCGCAGGTTCACCTACGGA 3′). The PCR reaction mixture contained 2 µL of 10 × PCR Buffer, 1.6 µL of 2.5 mM dNTPs, 1.0 µL of 10 pmol/µL from each primer, 0.2 µL of KOMA Taq (2.5 U/µL), 2 µL of 20 ng/µL DNA template and HPLC-grade water to adjust the reaction volume to 20 µL. The amplification reactions were done in a Biometra thermal cycler (Analytik Jena, Jena, Germany) with the reaction conditions mentioned elsewhere . Confirmation of successful amplification was assured by applying the PCR products to agarose gel electrophoresis. The PCR products were purified by PCR Cleanup Kit (MilliporeSigma, Burlington, MA, USA). Sequence analysis The purified PCR products were sent to a commercial company (Marogen Europe, Amsterdam, The Netherlands) for partial sequencing of the 18S rDNA using the same amplification primers. Sequencing at the company was performed by the BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, USA) and 3730xl DNA Analyzer automated DNA sequencing system (Applied Biosystems). The retrieved 18S rDNA sequences of the two fungal isolates were analyzed using the software Geneious Prime Version 2020.1.2 . Consent sequences were produced from forward and reverse sequences. The edited sequences were compared with other sequences of related reference strains in nucleotide database of the National Center of Biotechnology Information (NCBI) of the United States using the nucleotide platform of the Basic Local Alignment Search Tool (BLASTn). The sequences were also used to construct a phylogenetic tree using the Neighbor-Joining method with the software MEGA X .
The purified PCR products were sent to a commercial company (Marogen Europe, Amsterdam, The Netherlands) for partial sequencing of the 18S rDNA using the same amplification primers. Sequencing at the company was performed by the BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, USA) and 3730xl DNA Analyzer automated DNA sequencing system (Applied Biosystems). The retrieved 18S rDNA sequences of the two fungal isolates were analyzed using the software Geneious Prime Version 2020.1.2 . Consent sequences were produced from forward and reverse sequences. The edited sequences were compared with other sequences of related reference strains in nucleotide database of the National Center of Biotechnology Information (NCBI) of the United States using the nucleotide platform of the Basic Local Alignment Search Tool (BLASTn). The sequences were also used to construct a phylogenetic tree using the Neighbor-Joining method with the software MEGA X .
Separate culture flasks containing 500 mL of Sabouraud broth (Oxoid, UK) were inoculated by each fungal species isolate individually. The flasks were incubated in static incubator for 7 days at 28 °C, after which the fungal biomass was collected by filtration through Whatman filter paper No. 1 and washed with sterile distilled water to remove excess medium. The fungal biomasses were weighed, and 0.5 g were added to distilled water (10 mL) and incubated at 28 °C for 72 h. Finally, the biomass was separated from the filtrate by filtration. The aqueous filtrate containing the fungal metabolites was kept for further usage in the refrigerator at 4 °C .
For the biosynthesis of AgNPs, 1 mM AgNO 3 was added to fungal aqueous filtrate at a ratio 1:1 and then boiled for 30 min. This mixture was incubated under natural sun light at 25 °C for 24 h, till a dark brown color was formed. The nanoparticles were separated by centrifugation of the mixture at 14,000 rpm for 15 min, then dried by dropping the precipitate on petri dishes at room temperature and kept for further analysis .
AgNPs was re-dispersed in 1 mL of distilled water for chracterization. A UV–visible spectrophotometer (BIOCHROM Libra S60PC, England) was used in the range of 200–600 nm for measuring the absorbance of the AgNPs with the fungal aqueous filtrate was used as blank, as described before . The morphology, average diameter and distribution were estimated by transmission electron microscopy, TEM (JEM-1011, JEOL, Tokyo, Japan) at 80 kV voltage. The pattern of the size distribution (size and poly dispersed index, PDI) and zeta potential were assessed by a dynamic light scattering (DLS) system by a Zetasizer (NANO ZSP, Malvern Instruments Ltd., SerialNumber: MAL1118778, version 7.11, Malvern, UK). For surface analysis of the synthesized NPs, a scanning electron microscope, SEM (JED-2200 series, JEOL) provided by energy-dispersive X-ray spectroscopy (EDX) was used to confirm the presence of elemental silver, as previously described . Functional groups of fungal aqueous filtrates responsible for AgNO 3 reduction were estimated with Fourier transform infrared (FT-IR) spectroscopy (SPECTRUM100, Perkin-Elmer, Wellesley, MA, USA) with a range 450–3500 cm −1 , as detailed earlier .
To assess the antibacterial potential of the myco-synthesized AgNPs in vitro, four human pathogenic bacteria were tested ( Escherichia coli, Pseudomonas aeruginosa Staphylococcus aureus, and Klebsiella pneumoniae ). The assay was conducted by the agar well diffusion technique . Using sterile swabs, 100 µL of 24 h. old bacterial culture with a concentration equal to 0.5 McFarland (1.5 × 10 8 CFU mL −1 ) from each of the four tested strains were streaked on a plate of nutrient agar and left to dry. Under aseptic conditions, wells of 0.8 mm in diameter were made on the agar using a sterile cork-borer. Volumes of 50 µL of myco-synthesized AgNPs were added to the wells. The fungal filtrate and 5 µg ampicillin discs were used as negative and positive controls, respectively. The plates were incubated at 37 °C for 24 h. The experiment was performed in triplicate and the growth inhibition zones were estimated in millimeters (means ± standard variation), as described previously . The minimum inhibitory concentration (MIC) and the minimum bactericidal concentration (MBC) were estimated using the microdilution method in nutrient broth. The bacterial cultures of the four tested strains were individually diluted to 0.5 McFarland from which a volume of 10 µL was added to nutrient broth (350 µL). The isolates were tested against the myco-fabricated AgNPs solubilized in Dimethyl sulfoxide (DMSO) at different concentrations (0.031, 0.062, 0.125 and 0.250 mg mL −1 ). The negative control composed of nutrient broth inoculated with bacterial inoculum. Then, the plates were incubated at 37 °C for 24 h. The MIC was determined as the lowest concentration that inhibit observable bacterial growth where the MBC was verified when no growth appeared when treated bacteria were sub-cultured on nutrient agar plates after incubation period .
The total cellular soluble proteins of K. pneumoniae were extracted before and after treatment with 1 mg mL −1 of AgNPs synthesized from each fungal strain for 24 h at 37 °C, then purified by TriFast (Peqlab, VWR company). The proteins were further fractionated by OmniPAGE Mini vertical electrophoresis unit provided with a power Pro 5 power supply (Cleaver Scientific, Warwickshire, UK) on a SERVAGel™ TG PRiME 10% (SERVA, Heidelberg, Germany) .
One species ( K. pneumoniae ) was chosen for TEM analysis (JEM-1011, JEOL, Tokyo, Japan, at 80 kV voltage) to test the possible structural changes for cells treated with AgNPs fabricated by both fungal strains and incubated for 24 h. at 37 °C. Thereafter, bacterial suspension was centrifuged at 4000 rpm for 15 min. Then, 2.5% glutaraldehyde in 100 mM phosphate buffer, pH 7.0, was added to the pellet for sample fixation for 15–30 min. This was followed by applying 1% osmium tetroxide in 100 mM phosphate buffer for 1–2 h. at 4 °C, then washed five times with distilled water. Then, En bloc stain with 2% aqueous uranyl acetate was applied for ~ 2 h at 4 °C in the dark. Dehydration was done with series of acetone (30, 50, 70, 90 and 100%). After that, the samples were infiltrated in propylene oxide, followed by embedding in fresh resin (Epon mixture) for 12–24 h. at 60–70 °C. The ultrathin sections of the samples were put on 200-mesh copper grids after double staining with 2% uranyl acetate and lead citrate, then examined under a JEOL 100 CX electron microscope operating at 80 kV, as described previously .
All data were represented as mean and standard deviations. One-way analysis of variance (ANOVA) as well as graphs preparation were performed by Graph-bad Prism 9.1 software (Inc., La Jolla, CA, USA). ImagJ software was used for calculating the particles size and diameter from TEM images.
Fungal strains isolation and identification The 18S rRNA gene sequences analysis identified the two soil fungal isolates from Saudi Arabian desert as Embellisia spp. and Gymnoascus spp. The phylogenetic tree constructed from the two sequences together with corresponding to other strains worldwide is shown in Fig. . The sequences of these isolates have been deposited in the nucleotide sequence database of the GenBank of NCBI under accession numbers MN995544 for Embellisia spp. and MN995517 for Gymnoascus spp. Myco-synthesis of AgNPs The individual mixture of the aqueous filtrates from both fungal strains and the AgNO 3 has been changed from colorless to a stable brown color after incubation for 24 h. Such changes indicated the reduction of silver ions to AgNPs by fungal secondary metabolites. This conversion was the first sign that Embellisia spp. and Gymnoascus spp. were capable in AgNPs fabrication from AgNO 3 providing E-AgNPs and G-AgNPs, respectively. Myco-synthesized AgNPs characterization UV–visible spectrophotometry was used to further confirmation of the AgNPs synthesis from fungal filtrate. Strong beaks at 430 and 450 nm for E-AgNPs and G-AgNPs, respectively, were reported (Fig. A,B). The absorption peaks were detected in relation to the fungal filtrate (blank). The morphology and the size diameter of the myco-synthesized E-AgNPs and G-AgNPs were observed using TEM analysis which indicates well-dispersed spherical nanoparticles of 2–20 nm size diameter (Fig. A,B). Further, analysis was done by DLS, that determined an average diameter of 63.39 nm for E-AgNPs and 59.97 nm for G-AgNPs with a polydispersity index (PDI) of 0.28 and 0.40, respectively (Fig. ). EDX analysis confirmed the presence of the silver element in all myco-synthesized AgNPs. Strong signals at 3 keV were noted for Ag, along with carbon and oxygen peaks (Fig. ). The FTIR measurements were used to determine the presence of various functional groups in the fungal metabolites involved in AgNPs synthesis, capping, stabilization and consistency. The FTIR for the aqueous filtrate from Embellisia spp. showed eight peaks ranging from 3269.54 to 1633 cm −1 ; the peaks increased to ten for E-AgNPs ranging between 3282.16 and 1636.98 cm −1 (Fig. A). In Gymnoascus spp. filtrate, eleven peaks were detected ranging from 3258.51 to 1633.37 cm −1 , whereas eleven peaks were detected for G-AgNPs ranging between 3319.33 and 1637.15 cm −1 (Fig. B). Antibacterial activity of the myco-synthesized AgNPs When the E-AgNPs and G-AgNPs at concentrations of 1 mg mL −1 each were tested for antimicrobial activity against four human pathogenic bacteria. The maximum activity was recorded against K. pneumoniae , with growth inhibition zones of 5 and 3.8 mm for E-AgNPs and G-AgNPs, respectively. S. aureus was also sensitive to AgNPs with growth inhibition zones of 4.8 mm for E-AgNPs and 2.5 mm for G-AgNPs (Fig. A,B). The lowest activity of both AgNPs was noted against E. coil and P. aeruginosa . The activity of E-AgNPs was significantly higher than G-AgNPs (p < 0.001).Significant varitaion in the bacterial response to the myco-fabricated AgNPs was noted as well as their interaction. The MIC and MBC estimated for both myco-synthesized AgNPs by the micro-dilution method are shown in Table . The lowest MIC was recorded at 0.03 μg mL −1 against K. pneumoniae for both E-AgNPs and G-AgNPs, while the lowest MBC for both AgNPs types was 0.06 μg mL −1 against the same strain (Table ). SDS-PAGE and protein profile pattern Since the highest activities for E-AgNPs and G-AgNPs were noted against K. pneumoniae , therefore this strain was chosen for SDS-PAGE analysis to detect the protein profiling pattern before and after AgNPs treatment. The protein pattern from treated K. pneumoniae (Fig. ) displayed lower protein bands intensities as clear in lanes 3 and 4 compared with the untreated control (lane 2). The total number of protein bands of the control was 22 at sizes ranged from 15 to 160 kDa, whereas in E-AgNPs treated cells (lane 3) they were 16 bands with sizes ranged from 15 to 120 kDa. In lane 4, the protein from G-AgNPs treated bacteria showed only 8 bands ranging from 19 to 120 kDa. Ultrastructural changes detected in AgNPs treated bacteria Examination of K. pneumoniae cells by TEM analysis after treatment with the myco-synthesized AgNPs (Fig. ) showed damaged cell wall (Fig. B,C) and deformation and rupture of the bacterial cell (Fig. D).
The 18S rRNA gene sequences analysis identified the two soil fungal isolates from Saudi Arabian desert as Embellisia spp. and Gymnoascus spp. The phylogenetic tree constructed from the two sequences together with corresponding to other strains worldwide is shown in Fig. . The sequences of these isolates have been deposited in the nucleotide sequence database of the GenBank of NCBI under accession numbers MN995544 for Embellisia spp. and MN995517 for Gymnoascus spp.
The individual mixture of the aqueous filtrates from both fungal strains and the AgNO 3 has been changed from colorless to a stable brown color after incubation for 24 h. Such changes indicated the reduction of silver ions to AgNPs by fungal secondary metabolites. This conversion was the first sign that Embellisia spp. and Gymnoascus spp. were capable in AgNPs fabrication from AgNO 3 providing E-AgNPs and G-AgNPs, respectively.
UV–visible spectrophotometry was used to further confirmation of the AgNPs synthesis from fungal filtrate. Strong beaks at 430 and 450 nm for E-AgNPs and G-AgNPs, respectively, were reported (Fig. A,B). The absorption peaks were detected in relation to the fungal filtrate (blank). The morphology and the size diameter of the myco-synthesized E-AgNPs and G-AgNPs were observed using TEM analysis which indicates well-dispersed spherical nanoparticles of 2–20 nm size diameter (Fig. A,B). Further, analysis was done by DLS, that determined an average diameter of 63.39 nm for E-AgNPs and 59.97 nm for G-AgNPs with a polydispersity index (PDI) of 0.28 and 0.40, respectively (Fig. ). EDX analysis confirmed the presence of the silver element in all myco-synthesized AgNPs. Strong signals at 3 keV were noted for Ag, along with carbon and oxygen peaks (Fig. ). The FTIR measurements were used to determine the presence of various functional groups in the fungal metabolites involved in AgNPs synthesis, capping, stabilization and consistency. The FTIR for the aqueous filtrate from Embellisia spp. showed eight peaks ranging from 3269.54 to 1633 cm −1 ; the peaks increased to ten for E-AgNPs ranging between 3282.16 and 1636.98 cm −1 (Fig. A). In Gymnoascus spp. filtrate, eleven peaks were detected ranging from 3258.51 to 1633.37 cm −1 , whereas eleven peaks were detected for G-AgNPs ranging between 3319.33 and 1637.15 cm −1 (Fig. B).
When the E-AgNPs and G-AgNPs at concentrations of 1 mg mL −1 each were tested for antimicrobial activity against four human pathogenic bacteria. The maximum activity was recorded against K. pneumoniae , with growth inhibition zones of 5 and 3.8 mm for E-AgNPs and G-AgNPs, respectively. S. aureus was also sensitive to AgNPs with growth inhibition zones of 4.8 mm for E-AgNPs and 2.5 mm for G-AgNPs (Fig. A,B). The lowest activity of both AgNPs was noted against E. coil and P. aeruginosa . The activity of E-AgNPs was significantly higher than G-AgNPs (p < 0.001).Significant varitaion in the bacterial response to the myco-fabricated AgNPs was noted as well as their interaction. The MIC and MBC estimated for both myco-synthesized AgNPs by the micro-dilution method are shown in Table . The lowest MIC was recorded at 0.03 μg mL −1 against K. pneumoniae for both E-AgNPs and G-AgNPs, while the lowest MBC for both AgNPs types was 0.06 μg mL −1 against the same strain (Table ).
Since the highest activities for E-AgNPs and G-AgNPs were noted against K. pneumoniae , therefore this strain was chosen for SDS-PAGE analysis to detect the protein profiling pattern before and after AgNPs treatment. The protein pattern from treated K. pneumoniae (Fig. ) displayed lower protein bands intensities as clear in lanes 3 and 4 compared with the untreated control (lane 2). The total number of protein bands of the control was 22 at sizes ranged from 15 to 160 kDa, whereas in E-AgNPs treated cells (lane 3) they were 16 bands with sizes ranged from 15 to 120 kDa. In lane 4, the protein from G-AgNPs treated bacteria showed only 8 bands ranging from 19 to 120 kDa.
Examination of K. pneumoniae cells by TEM analysis after treatment with the myco-synthesized AgNPs (Fig. ) showed damaged cell wall (Fig. B,C) and deformation and rupture of the bacterial cell (Fig. D).
It was challenging for many scientists to find out new, simplified and ecofriendly methods for obtaining efficient AgNPs. The bio-fabrications are highly recommended due to their advantages over other synthetic methods , . The utilization of fungi for the synthesis of AgNPs is remarkably useful in relation to other fabrication methods due to simple handling and harmless effect. Several strains of Fusarium oxysporum mediated extracellular AgNPs formation depending on reductase/electron shuttle relationship and provided NPs ranging between 20 and 50 nm . In the present work the fungal filtrates from Embellisia spp. and Gymnoascus spp. indicated their ability in reducing Ag ions for self-assembly of AgNPs as detected by the formation of dark brown color when both compounds were mixed. Additionally, the recorded absorbance peaks at the UV–visible spectrum corresponded to the typical band of AgNPs that attributed to surface plasmon resonance, confirming the production of AgNPs . DLS analysis revealed that G-AgNPs particle size was smaller than that of E-AgNPs, which may reflect the diversity of fungal bioactive molecules involved in the reduction process, such as citric acid, peroxidases, homogeneous and heterogeneous proteins along with fungal enzymes, especially nitrate reductases . Although the size variation was not that high however, PDI indicated better mono-dispersibility and homogeneity for E-AgNPs than G-AgNPs. TEM indicated spherical shape of NPs and EDX analysis confirmed the presence of the silver element in G-AgNPs and E-AgNPs at 3 keV, which is typical for the absorption of metallic silver nano-crystallites. The signal band detected by FTIR analysis at 3269.54 and 3282.16 cm −1 corresponds to N–H, C–H, and O–H stretching vibrations, indicating the presence of primary amines in the fungal proteins alkynes, and alcohols, respectively . Intense absorption bands in FTIR at 1633.37 and 1637.15 cm −1 might be attributed to amide I due to carbonyl stretch in proteins C=O . Similar functional group detected at both fungal filtrate and AgNPs fabricated by their aid indicating the role of fungal biomolecules in the reduction and capping processes of AgNPs. The biological molecules from the fungal extract may carry out a double function by performing the reduction process and enhancing the AgNPs stability, as they could serve as good capping agents that covered the AgNPs surface and prevented their aggregation , . From these biomolecules, fungal enzymes are good reducing agents . Fungi could have electrostatic interactions between the metal ion and cell wall associated enzymes followed by subsequent enzymatic reduction of metal ions resulting in intracellular or extracellular production of NPs . The capability of fungal filtrates in NPs formation was previously approved for Phoma spp., Chaetomium globosum , and Chaetomium spp., which were able to reduce AgNO 3 into AgNPs with varying sizes and shapes . Also, Fusarium oxysporum and Verticillium spp. were capable of reducing Ag and gold into their nanoparticle forms depending on their extracellular proteins and secondary metabolites . Current findings showed potent antibacterial against the four tested human pathogenic bacteria however, the activity of E-AgNPs was higher than G-AgNPs. The reasons behind the variation in the NPs activity could be attributed to the variation in their PDI characteristics since no clear variations were noted in their size and morphology. PDI bellow 0.3 reflecting the mono dispersity and uniformity of the E-AgNPs that might enhance their activity compared to G-AgNPs where their PDI was 0.4. K. pneumoniae was the most sensitive tested strain, which is in accordance with previous findings by Sonbol et al. , who observed strong antibacterial activity of three myco-synthesized AgNPs. The antibacterial activity of morphologically modified AgNPs and chitosan loaded essential oils, were reported against multidrug resistant Gram negative bacteria and biofilm forming Acinetobacter baumannii, respectively , . The antibacterial ability of AgNPs could be related to the electrostatic interactions between positively charged Ag ions and the negatively charged bacterial membrane, which would lead to membrane permeability disruption and cell wall damage , . AgNPs could have also bound a specific group of enzymes (thiols) and affected the DNA capacity to replicate leading to cell death . In this study, the MIC and MBC indicated low fungal based AgNPs concentrations were needed to suppress the growth of both tested Gram-negative and Gram-positive bacteria. The efficiency of mucogenic AgNPs as antibacterial agents could be associated with their ability to enhance cellular oxidative stress causing destruction of the cell components leading to cell death. Polyacrylamide gel electrophoresis (SDS-PAGE) is a powerful technique for molecular analysis of cell proteins, especially in comparing protein patterns to detect qualitative, and quantitative changes, as well as clustering of bacterial groups . Proteomic differences of K. pneumoniae before and after treatment with fungal-based AgNPs might indicate significant polymorphism in the protein profiles that could be attributed to degradation or even block of the pathways for biosynthesis of certain proteins. Lower protein band intensity compared with untreated control might also be attributed to oxidative stress enhanced by NPs treatment. Besides, unfolding of the protein chain resulting from the reaction of AgNPs with thiol groups of the proteins is expected to cause degradation of the proteins. Therefore, it might explain the reduction in the number of protein bands from 22 in untreated to 16 and 8 in the treated bacteria with E-AgNPs and G-AgNPs, respectively. Different studies reported that AgNPs may be involved in preventing protein synthesis by preventing many translation factors , . Moreover, TEM imaging of K. pneumoniae treated with fungal-based AgNPs showed disruption within the cell membrane that led to damage in the outer membrane and deformation of the bacterial cell. As a result of interaction of free Ag + with several cellular sites, such as cytoplasmic membrane, cytoplasm and nuclear matrix. So, membrane permeability increased leading to loss of certain membrane ions as K + that impaired membrane integrity, and effects on the respiration process , . Other reports linked AgNPs toxicity to induce oxidative stress by motivating the development of ROS, which lead to cellular dysfunction and death . In general, the unique characters of NPs, such as their smaller size with larger surface area, their charges enable them to easily attach and penetrate bacterial cell walls and membranes make them a perfect tool from biological source to control adverse antibiotic resistant human pathogens.
The present investigation indicated the ability of two soil fungal isolates from Saudi Arabian desert ( Embellisia spp. and Gymnoascus spp.) for the biosynthesis of AgNPs. Both strains provided NPs with good physiochemical characteristics however, the PDI for E-AgNPs was low indicating higher homogeneity compared to G-AgNPs which might enhance their antibacterial activity. As an ecofriendly, safe, and cost-effective approach, both myco-synthesized AgNPs could be recommended as antibacterial agents specially against K. pneumoniae where their mode of action was reported by TEM and SDS PAGE analysis. However, further investigations are required to validate these findings by increasing the sample size of tested agents and study their antibacterial mechanisms. Overall, this study added new fungal isolates as bio mediators for nanoparticle synthesis with efficiency as antibacterial agents against some human pathogens.
|
Diagnosis and treatment of COVID-19 complicated with liver abscess: A case report and literature review | 469472a4-2e87-4b24-95f5-98abb731d8f6 | 11749645 | Surgical Procedures, Operative[mh] | At present, Novel Coronavirus Pneumonia (NCP) is spreading in many countries and regions. COVID-19 is characterized by rapid onset, high infectivity, rapid progression and variably effective treatment. Our hospital is a designated hospital for COVID-19 treatment in Shiyan City, receiving and treating suspected, mild, ordinary, severe and critical patients. In the course of a pandemic, a patient with severe fever and shock as the main clinical manifestation of COVID-19 complicated by liver abscess was admitted to our department. After literature review, no case of this type has been reported so far, which is now shared as follows:
2.1. Clinical data The patient was a retired female aged 63 years, from Maojian District, Shiyan City, Hubei Province. She came to the fever Clinic of our hospital on February 10, 2020 due to “fever for 12 days.” She denied travel to an affected area or close contact with a COVID-19 patient. Her past medical history was unremarkable. The patient had fever without obvious cause on January 29, 2020, with the highest temperature of 39 °C, accompanied by fatigue and loss of appetite, without cold or chills, without cough, sputum, palpitations, chest tightness, asthma dyspnea, nausea, vomiting, abdominal pain, diarrhea, urinary frequency, urgency, dizziness or sweating. After a self-administered antipyretic diclofenac sodium suppository 50 mg, body temperature decreased, but symptoms recurred with intermittent fever, accompanied by tachypnea. Chest computed tomography (CT) examination, at another hospital, suggested bilateral lung infection and raised the possibility of viral pneumonia, therefore novel coronavirus nucleic acid test was performed in the fever clinic of our hospital and was found to be positive. Physical examination: Body temperature:36.7 °C (after diclofenac sodium 50 mg rectally), pulse:132/min regular, respiration:22/min, blood pressure:80/64 mm Hg, clear consciousness, red throat with congestion, red tonsils, coarse breath sounds in both lungs, no dry or wet rales. Abdominal examination showed no obvious abnormality, there was no edema of lower limbs and negative pathological signs. 2.2. Auxiliary examinations Blood gas analysis (on oxygen 5 L/min) suggested hypoxemia and respiratory distress, including pH: 7.499; the partial pressure of carbon dioxide: 35.2 mm Hg; the partial pressures of oxygen: 92 mm Hg; base excess: 4 mmol/L; hydrogen carbonate: 27.7 mmol/L; sulfur dioxide: 98%; Lactate: 1.51 mmol/L; oxygenation index: 224 mm Hg. Novel coronavirus nucleic acid test was positive on February 10, 2020, and was negative on February 15, 17, and 20, 2020, respectively. There were erythrocyte sedimentation and mild abnormal liver function. Blood analysis indicated that white blood cell was 32.71 × 10 9 /L, neutrophils was 30.2 × 10 9 /L, NEUT% was 92.3%, lymphocyte (LYMPH) was 0.7 × 10 9 /L, LYMPH% was 3.6%. C-reaction protein was 195.37 mg/L, serum ferritin was 1766.80 ng/mL, interleukin-6 was 146.5 pg/mL, procalcitonin was 1.75 ng/mL. Immunoglobulin M was weakly positive for influenza A + B virus, immunoglobulin M was weakly positive for parainfluenza virus types 1, 2, and 3. Blood culture was negative. There were multiple infectious lesions in both lungs on chest CT, indicating viral pneumonia in combination with the previous medical history. There were also bilateral pleural effusions, right larger than left, and slightly reduced hepatic density. Enhanced CT or magnetic resonance was recommended for further examination. Hepatobiliary pancreas and spleen ultrasonography suggested a possible large abscess in the right lobe of liver. Ultrasound of the thoracoabdominal cavity showed bilateral pleural effusions, with a large amount on the right side but no ascites. Drainage fluid culture of liver abscess showed Klebsiella pneumoniae . Five tumor markers: carcino-embryonic antigen, 0.9 ng/mL; alpha-fetoprotein, 5.46 ng/mL; serum ferritin, 1938.80 ng/mL; Ca125, 1558.70 U/mL; squamous cell carcinoma antigen, 15.20 ng/mL. Pleural biochemistry: total protein, 53 g/L; alkaline phosphatase, 129 U/L; adenosine deaminase, 13 U/L; lactate dehydrogenase, 341 U/L; amylase, 30 U/L; rheumatoid factor, <1.00 KU/L; high-sensitivity C-reactive protein, 29.49 mg/L. Pleural fluid appearance: yellow, transparent, no clot, specific gravity of 1.028, qualitative protein , white blood cell count of 4.615 × 10 9 /L, mononuclear cells of 48%, multiple nuclear cell classification count of 52%. There was no bacterial growth in pleural effusion culture over 5 days. Electrocardiogram showed sinus tachycardia, with abnormal T waves, and myocardial ischemia was considered. Brain natriuretic peptide, myocardial markers, renal function, and blood glucose level were normal. Urine and fecal routine cultures, and 5 quantification of viral hepatitis B, combination of hepatitis, Novel Coronavirus antibody, and A/B antigen and A/B respiratory syncytial virus nucleic acid were all negative. Blood routine examination, infection combination, coagulation indices and chest and abdominal CT examination appearances are shown in Figures – . 2.3. Diagnosis and treatment process Medical treatment was according to novel Coronavirus Infection Pneumonia Treatment Scheme (Trial Edition 6), as the combination of clinical manifestations, Chest CT images and positive nucleic acid test results were diagnostic of COVID-19. The patient was admitted with typical manifestations of severe disease: high fever, oxygenation index of 224 mm Hg, rapid heart rate, and shock. The patient was seriously ill, requiring high flow oxygen (5 L/min by mask) and a multidisciplinary discussion (Medical, Infection, Respiratory, and Radiology Departments) was held to formulate rescue and a treatment plan. Management included rapid infusion of sodium chloride solution, and dextran 40 solution to combat shock, antibiotics: imipenem-cilastatin sodium injection, 1.0 gram (g) q8h, + vancomycin injection, 1.0 g q12h, later replaced by ceftazidime, 2.0 g q12h, Arpidol hydrochloride 0.2 g 3 times a day, and recombinant human interferon a-1b 5 million IU the antiviral by nebulization, vitamin C injection 4 g, Xuebijing injection 100 mL to reduce inflammation, Chinese medicine prescription granules, Thymus method 1.6 mg twice a week, spleen polypeptide by injection 6 mL to regulate immunity, 20% human serum albumin 100 mL for correction of low plasma proteins, reduced glutathione needle 2.4 g + isoglycyrrhizase needle 20 mL intravenously to protect liver enzymes, panxitora azole enteric capsules for prevention of stress ulceration, the bow for injection heparin calcium 1500 IU subcutaneously qd as anticoagulation, for prevention of thrombosis, correction of electrolyte disorders, lactulose oral liquid runchang purge, compound lactobacillus acidophilus + garlic soft capsule for adjustment of intestinal flora, inhaled acetylcysteine solution + suction cloth to Ned suspension liquid atomization inhalation as an expectorant, the antioxidant, ambroxol hydrochloride oral solution + Yumei capsules oral for cough and others supportive treatment. According to the results of chest and abdominal CT, bedside hepatobiliary and thoracic B-ultrasound, percutaneous liver puncture catheter drainage was performed for liver abscess on February 11, 2020, under the guidance of ultrasound. Brown abscess fluid, mixed with a small amount of hemorrhagic material was extracted, and drainage fluid culture + drug sensitivity was sent. About 600 mL of light red abscess fluid was drained the next day. A chest drain placed in the right thoracic cavity, under ultrasound guidance on February 18, 2020, drained 150 mL and pleural fluid was sent for routine examination, biochemical examination, swelling mark and culture. February 21, 2020, Chest and abdominal drainage bags and not drainage liquid, to root out the chest and abdominal cavity drainage tube, stable vital signs. Admitted to hospital after 23 days (March 4, 2020), patients with normal temperature is more than 3 days, will be coronavirus 3 consecutive negative nucleic acid detection, review the chest CT hint double lung disease stove did not see obvious change and liver abscess is absorbed before narrowing, chest, abdominal cavity no effusion on both sides. After discussion by Panel discussion, the patient was discharged from the hospital with isolation observation and follow-up. 2.4. Final diagnosis COVID-19 severe pneumonia; Bacterial liver abscess; Septic shock; Pleural effusions; Influenza; Abnormal liver function; Electrolyte disturbance; and Hypoproteinemia.
The patient was a retired female aged 63 years, from Maojian District, Shiyan City, Hubei Province. She came to the fever Clinic of our hospital on February 10, 2020 due to “fever for 12 days.” She denied travel to an affected area or close contact with a COVID-19 patient. Her past medical history was unremarkable. The patient had fever without obvious cause on January 29, 2020, with the highest temperature of 39 °C, accompanied by fatigue and loss of appetite, without cold or chills, without cough, sputum, palpitations, chest tightness, asthma dyspnea, nausea, vomiting, abdominal pain, diarrhea, urinary frequency, urgency, dizziness or sweating. After a self-administered antipyretic diclofenac sodium suppository 50 mg, body temperature decreased, but symptoms recurred with intermittent fever, accompanied by tachypnea. Chest computed tomography (CT) examination, at another hospital, suggested bilateral lung infection and raised the possibility of viral pneumonia, therefore novel coronavirus nucleic acid test was performed in the fever clinic of our hospital and was found to be positive. Physical examination: Body temperature:36.7 °C (after diclofenac sodium 50 mg rectally), pulse:132/min regular, respiration:22/min, blood pressure:80/64 mm Hg, clear consciousness, red throat with congestion, red tonsils, coarse breath sounds in both lungs, no dry or wet rales. Abdominal examination showed no obvious abnormality, there was no edema of lower limbs and negative pathological signs.
Blood gas analysis (on oxygen 5 L/min) suggested hypoxemia and respiratory distress, including pH: 7.499; the partial pressure of carbon dioxide: 35.2 mm Hg; the partial pressures of oxygen: 92 mm Hg; base excess: 4 mmol/L; hydrogen carbonate: 27.7 mmol/L; sulfur dioxide: 98%; Lactate: 1.51 mmol/L; oxygenation index: 224 mm Hg. Novel coronavirus nucleic acid test was positive on February 10, 2020, and was negative on February 15, 17, and 20, 2020, respectively. There were erythrocyte sedimentation and mild abnormal liver function. Blood analysis indicated that white blood cell was 32.71 × 10 9 /L, neutrophils was 30.2 × 10 9 /L, NEUT% was 92.3%, lymphocyte (LYMPH) was 0.7 × 10 9 /L, LYMPH% was 3.6%. C-reaction protein was 195.37 mg/L, serum ferritin was 1766.80 ng/mL, interleukin-6 was 146.5 pg/mL, procalcitonin was 1.75 ng/mL. Immunoglobulin M was weakly positive for influenza A + B virus, immunoglobulin M was weakly positive for parainfluenza virus types 1, 2, and 3. Blood culture was negative. There were multiple infectious lesions in both lungs on chest CT, indicating viral pneumonia in combination with the previous medical history. There were also bilateral pleural effusions, right larger than left, and slightly reduced hepatic density. Enhanced CT or magnetic resonance was recommended for further examination. Hepatobiliary pancreas and spleen ultrasonography suggested a possible large abscess in the right lobe of liver. Ultrasound of the thoracoabdominal cavity showed bilateral pleural effusions, with a large amount on the right side but no ascites. Drainage fluid culture of liver abscess showed Klebsiella pneumoniae . Five tumor markers: carcino-embryonic antigen, 0.9 ng/mL; alpha-fetoprotein, 5.46 ng/mL; serum ferritin, 1938.80 ng/mL; Ca125, 1558.70 U/mL; squamous cell carcinoma antigen, 15.20 ng/mL. Pleural biochemistry: total protein, 53 g/L; alkaline phosphatase, 129 U/L; adenosine deaminase, 13 U/L; lactate dehydrogenase, 341 U/L; amylase, 30 U/L; rheumatoid factor, <1.00 KU/L; high-sensitivity C-reactive protein, 29.49 mg/L. Pleural fluid appearance: yellow, transparent, no clot, specific gravity of 1.028, qualitative protein , white blood cell count of 4.615 × 10 9 /L, mononuclear cells of 48%, multiple nuclear cell classification count of 52%. There was no bacterial growth in pleural effusion culture over 5 days. Electrocardiogram showed sinus tachycardia, with abnormal T waves, and myocardial ischemia was considered. Brain natriuretic peptide, myocardial markers, renal function, and blood glucose level were normal. Urine and fecal routine cultures, and 5 quantification of viral hepatitis B, combination of hepatitis, Novel Coronavirus antibody, and A/B antigen and A/B respiratory syncytial virus nucleic acid were all negative. Blood routine examination, infection combination, coagulation indices and chest and abdominal CT examination appearances are shown in Figures – .
Medical treatment was according to novel Coronavirus Infection Pneumonia Treatment Scheme (Trial Edition 6), as the combination of clinical manifestations, Chest CT images and positive nucleic acid test results were diagnostic of COVID-19. The patient was admitted with typical manifestations of severe disease: high fever, oxygenation index of 224 mm Hg, rapid heart rate, and shock. The patient was seriously ill, requiring high flow oxygen (5 L/min by mask) and a multidisciplinary discussion (Medical, Infection, Respiratory, and Radiology Departments) was held to formulate rescue and a treatment plan. Management included rapid infusion of sodium chloride solution, and dextran 40 solution to combat shock, antibiotics: imipenem-cilastatin sodium injection, 1.0 gram (g) q8h, + vancomycin injection, 1.0 g q12h, later replaced by ceftazidime, 2.0 g q12h, Arpidol hydrochloride 0.2 g 3 times a day, and recombinant human interferon a-1b 5 million IU the antiviral by nebulization, vitamin C injection 4 g, Xuebijing injection 100 mL to reduce inflammation, Chinese medicine prescription granules, Thymus method 1.6 mg twice a week, spleen polypeptide by injection 6 mL to regulate immunity, 20% human serum albumin 100 mL for correction of low plasma proteins, reduced glutathione needle 2.4 g + isoglycyrrhizase needle 20 mL intravenously to protect liver enzymes, panxitora azole enteric capsules for prevention of stress ulceration, the bow for injection heparin calcium 1500 IU subcutaneously qd as anticoagulation, for prevention of thrombosis, correction of electrolyte disorders, lactulose oral liquid runchang purge, compound lactobacillus acidophilus + garlic soft capsule for adjustment of intestinal flora, inhaled acetylcysteine solution + suction cloth to Ned suspension liquid atomization inhalation as an expectorant, the antioxidant, ambroxol hydrochloride oral solution + Yumei capsules oral for cough and others supportive treatment. According to the results of chest and abdominal CT, bedside hepatobiliary and thoracic B-ultrasound, percutaneous liver puncture catheter drainage was performed for liver abscess on February 11, 2020, under the guidance of ultrasound. Brown abscess fluid, mixed with a small amount of hemorrhagic material was extracted, and drainage fluid culture + drug sensitivity was sent. About 600 mL of light red abscess fluid was drained the next day. A chest drain placed in the right thoracic cavity, under ultrasound guidance on February 18, 2020, drained 150 mL and pleural fluid was sent for routine examination, biochemical examination, swelling mark and culture. February 21, 2020, Chest and abdominal drainage bags and not drainage liquid, to root out the chest and abdominal cavity drainage tube, stable vital signs. Admitted to hospital after 23 days (March 4, 2020), patients with normal temperature is more than 3 days, will be coronavirus 3 consecutive negative nucleic acid detection, review the chest CT hint double lung disease stove did not see obvious change and liver abscess is absorbed before narrowing, chest, abdominal cavity no effusion on both sides. After discussion by Panel discussion, the patient was discharged from the hospital with isolation observation and follow-up.
COVID-19 severe pneumonia; Bacterial liver abscess; Septic shock; Pleural effusions; Influenza; Abnormal liver function; Electrolyte disturbance; and Hypoproteinemia.
This 63-year-old woman with COVID-19 admitted to our hospital, denied any history of underlying disease, and her blood glucose monitoring was fair. Initial symptoms were fever, loss of appetite and fatigue, suggestive of viral infection, with positive SARS serology, but by 12 days she had developed bibasal pulmonary changes including small pleural effusions. In hospital she had remittent high fever and developed sinus tachycardia, hypoxemia, respiratory distress, hemodynamic instability, myocardial ischemic changes on electrocardiogram and shock. Small bilateral pleural effusions were present on admission, but they increased significantly over 4 days, with a large right pleural effusion appearing 1 week later. After admission, liver function was abnormal chest CT suggested a large low-density shadow in the right lobe of the liver. The possibility of a tumor or abscess was considered, but abscess was confirmed by bedside color ultrasound. Laboratory investigations of the pat suggested multi-system damage; progressive increase in D-dimer, abnormal liver function, markedly increased white blood cell, neutrophil, and platelet counts, progressive elevation of ferritin and procalcitonin, prolonged prothrombin time and activated partial thromboplastin time. K pneumoniae was cultured from the liver abscess from sepsis, although no bacterial growth was observed in 2 blood cultures. Sepsis could still be diagnosed from the clinical manifestations. Because the patient had no obvious respiratory symptoms at the onset of fever, we consider it likely that she initially presented with COVID-19 and subsequently developed a liver abscess and sepsis. We consider it less likely that a bacterial pneumonia was the primary condition because of the imaging and clinical sequence. Although pleural effusion occurs relatively rarely in COVID-19 (in 5.6% of CT scans, it was reported in severe cases in some elderly patients with underlying diseases, although the mechanism was unclear ). In our patient, other causes including cancer, tuberculosis, autoimmune, cardiac and renal were ruled out; pleural effusion was an exudate with typically elevated protein, lactate dehydrogenase and inflammatory cells and it was likely related to the liver abscess. The presentation with COVID-19 and development of respiratory failure would normally indicate glucocorticoid treatment, however, this was withheld in the face of liver abscesses and bacterial infection. We have found only 3 case reports of bacterial liver abscess being diagnosed in patients with recent COVID-19, an amebic liver abscess and a 36-year-old man with a liver abscess who was SARS-COV 2 positive. The bacterial liver abscesses all occurred 4 to 6 weeks after initial COVID-19; 1 patient had received dexamethasone and 2 tocilizumab. Interestingly, K pneumoniae was isolated from blood and liver aspirate in a 78-year-old man 6 weeks after admission with severe COVID-19 pneumonia. Our patient appears unique in presenting with a liver abscess early into the illness and before receiving any immunosuppressive therapy. We present this case to highlight the importance of considering additional diagnoses in a patient with COVID-19 and to stress the importance of individualizing the treatment strategy (here withholding corticosteroid therapy) in accordance with the patients’ overall clinical picture.
The authors would like to express their gratitude to Edit Springs ( https://www.editsprings.cn ) for the expert linguistic services provided.
Conceptualization: Xueqiang Jiang. Formal analysis: Xueqiang Jiang, Yanwei Liu, Wan Wang, Congyu Zhang. Funding acquisition: Xueqiang Jiang, Wan Wang. Investigation: Fan Li. Methodology: Congyu Zhang. Project administration: Fan Li, Wan Wang. Resources: Yanwei Liu. Validation: Fan Li, Congyu Zhang. Visualization: Yanwei Liu. Writing – original draft: Fan Li, Xueqiang Jiang, Wan Wang. Writing – review & editing: Xueqiang Jiang, Wan Wang.
|
Creation of a rating scale to teach Less Invasive Surfactant Administration (LISA) in simulation | 2fcefad5-3d38-4fc3-8bb2-bb865be163c8 | 10868031 | Pediatrics[mh] | Less invasive surfactant administration (LISA) is now the standard for surfactant administration and must be learned by neonatologists. Simulation training is an appropriate way to teach this procedure. For this study, we created and validated the first scale for teaching and assessing the LISA procedure in simulation. This high-quality tool is reliable, reproducible, and easy to use Respiratory distress syndrome (RDS) particularly impacts preterm neonates and affects nearly all infants born at 28 weeks of gestation or less. The approach to respiratory support in preterm infants with RDS has evolved toward more frequent use of noninvasive methods, resulting in improved clinical outcomes and potentially lower hospitalization costs . In keeping with this trend, Less Invasive Surfactant Administration (LISA) techniques have become increasingly common. This technique involves injecting surfactant into the trachea through a thin catheter during laryngoscopy while maintaining noninvasive ventilation over the child's nose. Once the surfactant is injected, the catheter is removed . The main advantage of this method is that it avoids intubation and all the associated risks . The LISA procedure reduces the median days on mechanical ventilation, intubation-related lung injury, and oxygen requirement at 28 days of life compared to the INSURE method (Intubation SURfactant Extubation), which is a less invasive method . The LISA method also reduces the composite score for death and/or bronchopulmonary dysplasia. Since the latest update of the European Consensus Guidelines on the management of RDS in 2019, the LISA technique has been considered to be the recommended method for surfactant administration, provided that clinicians have sufficient experience . Recent reports indicate that only 8% of neonatologists in the United States and 11% of neonatal units in England claim to routinely use this method . This could be partly explained by a lack of consensus regarding clinical practice and training for pediatricians to acquire LISA competence . Training neonatologists in this new technique is essential, and simulation is a particularly appropriate approach to learning it. To validate simulation skills, a criterion-referenced assessment with a rating scale is required. The tools must be consensual and have good reliability and validity. The aim of this study was to develop and validate a rating scale for teaching the LISA method in simulation. Institutional review board This project was approved by the Research Ethics Committee of Sorbonne University, Paris, France (CER-2022-028). All participants received an information letter. Informed consent was obtained from all subjects and participants signed a waiver of image rights. This work used the Downing method . Five steps are necessary for the construction and validation of an assessment tool: Development of the rating scale -The content of the rating scale must be based on good clinical experience and a review of the literature. Each item of the rating scale must be related to the subject studied by the tool. The first version of the rating scale was drafted by HR and CC. The rating scale was then sent to several experts in the method under study. To constitute the LISA expert panel in charge of assessing the relevance of each item of the scale, 12 neonatologists were recruited in France and Belgium who had a high level of expertise (more than 5 years of practice) in regard to the LISA method and in simulation teaching. They were recruited by mail. They worked in accordance with the modified Delphi method. This method is used to reach a consensus between multiple experts regarding a single question . Following this modified Delphi method, the experts rated each item from 0 to 6. The responses of each expert were anonymized using a letter of the alphabet. In keeping with the modified Delphi method, each item with a median score of less than 4 was modified, and this step was repeated until all the items had been validated by the experts. After a first round of assessment by the experts, all items on the rating scale had medians above 4. Nevertheless, the wording of some of the items was modified according to the experts' comments, and the rating scale was submitted to them a second time. Figure shows the rating scale developed with the expert group. The final scale is composed of 25 items divided into eight categories. The different categories represent the main stages of surfactant administration using the LISA method. These stages specify the performance of the procedure itself and all the elements of preparation prior to the performance of the procedure. Each item is scored 0, 1, or 2 depending on whether the participant failed, passed part of the item, or passed the entire item perfectly. The point is awarded regardless of the order in which the gestures are performed for each step. The total score is out of 50 and can easily be reduced to 100. Below the rating scale, a reminder of the necessary equipment as well as the doses of medication and a link showing the insertion of the catheter between the vocal cords was added to facilitate the use of the scale in the departments after the training. The drugs required for sedation were not specified to allow each center to define their own sedation protocol. -The response process: The first version of the rating scale was used in the simulation so that the observers knew how to rate the simulator runs. We carried out a first simulation with a physician who had already been trained in the LISA method within P2ULSE (Plateforme Pluridisciplinaire Hospitalo-Universitaire de e-Learning et de Simulation de l'Est parisien). No items were changed following this test simulation session. Some items appeared to be difficult to evaluate, but they were kept intact in the scale for training reasons. This point will be addressed in the discussion section. The first observer used the scale to assess this run and then trained the second observer. Validation process for the obtained rating scale The last three points of the Downing method require simulation sessions. Population The aim was to recruit forty participants. The number of participants was based on previously published studies . The simulations were offered on a voluntary basis as training for LISA. All participants were pediatric residents or senior physicians from Paris and its surrounding areas. They were recruited by mail or phone. Those classified as “inexperienced” were those who had performed the entire LISA procedure less than five times. Those classified as “experienced” were participants who had performed the LISA procedure at least five times. Simulation session material All the simulation sessions took place at the Trousseau Hospital (Assistance Publique des Hôpitaux de Paris (APHP), Paris), in the P2ULSE (Plateforme Pluridisciplinaire Hospitalo-Universitaire de e-Learning et de Simulation de l'Est parisien) simulation laboratory in February 2022. A "Premature Ann" manikin (Laerdal®, Stavanger, Norway) was used for the simulation sessions. It is a 25-weeks-of-gestation low-fidelity manikin, including an exact replica of the airways and body size . The design provides a degree of realism to procedures such as LISA. The environment of the simulation platform was that of a Level II maternity ward , with a heated radiant table, a T-piece resuscitator (Neopuff®), surfactant with preparation materials (syringe, needles), a fine catheter (LISACath®, Chiesi), intubation materials, and a laryngoscope with a Magill 00 blade. The sessions were filmed using three video cameras and recorded on a computer and a secure hard drive. The simulation step As usual, the simulation sessions took place in four steps: prebriefing, briefing, simulation, and debriefing. The pre-briefing consisted of a PowerPoint® presentation explaining the relevance of LISA and all the steps required for the procedure. The participants were then given a presentation of the simulation room, in particular the manikin and the equipment available. The briefing consisted of a succinct presentation of the medical situation. Each participant was asked to endorse the role of a physician belonging to the French pediatric emergency medical service transport team, arriving at a Level II maternity hospital where a 600 g newborn infant had just been born. The patient was in respiratory distress on oxygen and continuous positive airway pressure of 5 cm H 2 O. A facilitator was present in the simulation room and played the role of an emergency transport nurse. The participant was then invited to enter the simulation room. The debriefing was not carried out as a group after each run, as the other participants had to remain unaware of the scenario used. Each participant was invited to debrief individually with the two instructors. Videos of the simulation sessions were then independently viewed and evaluated by the two blinded raters (HR and BG). The evaluators used the rating scale to rate each participant's performance. All this was to allow the three next steps of the Downing method to be performed: Internal structure study: The aim of this step was to study the reliability of the rating scale. To do this, we need to evaluate the scores that the two observers gave to the participants during the simulation sessions. In this section, we assessed whether the number of items in the scale was correct and to what extent items in the same group were similar to each other. Comparison: The aim of this step was also to evaluate the reliability of the scale under modification of its score according to different groups of participants. Students partake in simulation sessions, and two raters score their performance using the rating scale. Two different observers should be able to assign a similar score to the same simulator run. Consequence: An evaluation tool must be able to distinguish between failures and successes, in other words between the degree to which participants are familiar with this technique. This last step reinforces the validity of the rating scale. Statistical analysis Reliability analysis included internal consistency testing using a Cronbach’s alpha (CA) test and interrater reliability analysis using intraclass correlation coefficient (ICC), linear regression and its coefficient. F-test or t-test were used to compare scores, as appropriate. Validity analysis included a comparison of the mean scores obtained by technical novices and experts using a t-test.Statistical significance was assumed below a p -value threshold of 0.05. All the statistical tests were carried out on Excel® version 2205, published by Microsoft®. This project was approved by the Research Ethics Committee of Sorbonne University, Paris, France (CER-2022-028). All participants received an information letter. Informed consent was obtained from all subjects and participants signed a waiver of image rights. This work used the Downing method . Five steps are necessary for the construction and validation of an assessment tool: -The content of the rating scale must be based on good clinical experience and a review of the literature. Each item of the rating scale must be related to the subject studied by the tool. The first version of the rating scale was drafted by HR and CC. The rating scale was then sent to several experts in the method under study. To constitute the LISA expert panel in charge of assessing the relevance of each item of the scale, 12 neonatologists were recruited in France and Belgium who had a high level of expertise (more than 5 years of practice) in regard to the LISA method and in simulation teaching. They were recruited by mail. They worked in accordance with the modified Delphi method. This method is used to reach a consensus between multiple experts regarding a single question . Following this modified Delphi method, the experts rated each item from 0 to 6. The responses of each expert were anonymized using a letter of the alphabet. In keeping with the modified Delphi method, each item with a median score of less than 4 was modified, and this step was repeated until all the items had been validated by the experts. After a first round of assessment by the experts, all items on the rating scale had medians above 4. Nevertheless, the wording of some of the items was modified according to the experts' comments, and the rating scale was submitted to them a second time. Figure shows the rating scale developed with the expert group. The final scale is composed of 25 items divided into eight categories. The different categories represent the main stages of surfactant administration using the LISA method. These stages specify the performance of the procedure itself and all the elements of preparation prior to the performance of the procedure. Each item is scored 0, 1, or 2 depending on whether the participant failed, passed part of the item, or passed the entire item perfectly. The point is awarded regardless of the order in which the gestures are performed for each step. The total score is out of 50 and can easily be reduced to 100. Below the rating scale, a reminder of the necessary equipment as well as the doses of medication and a link showing the insertion of the catheter between the vocal cords was added to facilitate the use of the scale in the departments after the training. The drugs required for sedation were not specified to allow each center to define their own sedation protocol. -The response process: The first version of the rating scale was used in the simulation so that the observers knew how to rate the simulator runs. We carried out a first simulation with a physician who had already been trained in the LISA method within P2ULSE (Plateforme Pluridisciplinaire Hospitalo-Universitaire de e-Learning et de Simulation de l'Est parisien). No items were changed following this test simulation session. Some items appeared to be difficult to evaluate, but they were kept intact in the scale for training reasons. This point will be addressed in the discussion section. The first observer used the scale to assess this run and then trained the second observer. The last three points of the Downing method require simulation sessions. Population The aim was to recruit forty participants. The number of participants was based on previously published studies . The simulations were offered on a voluntary basis as training for LISA. All participants were pediatric residents or senior physicians from Paris and its surrounding areas. They were recruited by mail or phone. Those classified as “inexperienced” were those who had performed the entire LISA procedure less than five times. Those classified as “experienced” were participants who had performed the LISA procedure at least five times. Simulation session material All the simulation sessions took place at the Trousseau Hospital (Assistance Publique des Hôpitaux de Paris (APHP), Paris), in the P2ULSE (Plateforme Pluridisciplinaire Hospitalo-Universitaire de e-Learning et de Simulation de l'Est parisien) simulation laboratory in February 2022. A "Premature Ann" manikin (Laerdal®, Stavanger, Norway) was used for the simulation sessions. It is a 25-weeks-of-gestation low-fidelity manikin, including an exact replica of the airways and body size . The design provides a degree of realism to procedures such as LISA. The environment of the simulation platform was that of a Level II maternity ward , with a heated radiant table, a T-piece resuscitator (Neopuff®), surfactant with preparation materials (syringe, needles), a fine catheter (LISACath®, Chiesi), intubation materials, and a laryngoscope with a Magill 00 blade. The sessions were filmed using three video cameras and recorded on a computer and a secure hard drive. The simulation step As usual, the simulation sessions took place in four steps: prebriefing, briefing, simulation, and debriefing. The pre-briefing consisted of a PowerPoint® presentation explaining the relevance of LISA and all the steps required for the procedure. The participants were then given a presentation of the simulation room, in particular the manikin and the equipment available. The briefing consisted of a succinct presentation of the medical situation. Each participant was asked to endorse the role of a physician belonging to the French pediatric emergency medical service transport team, arriving at a Level II maternity hospital where a 600 g newborn infant had just been born. The patient was in respiratory distress on oxygen and continuous positive airway pressure of 5 cm H 2 O. A facilitator was present in the simulation room and played the role of an emergency transport nurse. The participant was then invited to enter the simulation room. The debriefing was not carried out as a group after each run, as the other participants had to remain unaware of the scenario used. Each participant was invited to debrief individually with the two instructors. Videos of the simulation sessions were then independently viewed and evaluated by the two blinded raters (HR and BG). The evaluators used the rating scale to rate each participant's performance. All this was to allow the three next steps of the Downing method to be performed: Internal structure study: The aim of this step was to study the reliability of the rating scale. To do this, we need to evaluate the scores that the two observers gave to the participants during the simulation sessions. In this section, we assessed whether the number of items in the scale was correct and to what extent items in the same group were similar to each other. Comparison: The aim of this step was also to evaluate the reliability of the scale under modification of its score according to different groups of participants. Students partake in simulation sessions, and two raters score their performance using the rating scale. Two different observers should be able to assign a similar score to the same simulator run. Consequence: An evaluation tool must be able to distinguish between failures and successes, in other words between the degree to which participants are familiar with this technique. This last step reinforces the validity of the rating scale. The aim was to recruit forty participants. The number of participants was based on previously published studies . The simulations were offered on a voluntary basis as training for LISA. All participants were pediatric residents or senior physicians from Paris and its surrounding areas. They were recruited by mail or phone. Those classified as “inexperienced” were those who had performed the entire LISA procedure less than five times. Those classified as “experienced” were participants who had performed the LISA procedure at least five times. All the simulation sessions took place at the Trousseau Hospital (Assistance Publique des Hôpitaux de Paris (APHP), Paris), in the P2ULSE (Plateforme Pluridisciplinaire Hospitalo-Universitaire de e-Learning et de Simulation de l'Est parisien) simulation laboratory in February 2022. A "Premature Ann" manikin (Laerdal®, Stavanger, Norway) was used for the simulation sessions. It is a 25-weeks-of-gestation low-fidelity manikin, including an exact replica of the airways and body size . The design provides a degree of realism to procedures such as LISA. The environment of the simulation platform was that of a Level II maternity ward , with a heated radiant table, a T-piece resuscitator (Neopuff®), surfactant with preparation materials (syringe, needles), a fine catheter (LISACath®, Chiesi), intubation materials, and a laryngoscope with a Magill 00 blade. The sessions were filmed using three video cameras and recorded on a computer and a secure hard drive. As usual, the simulation sessions took place in four steps: prebriefing, briefing, simulation, and debriefing. The pre-briefing consisted of a PowerPoint® presentation explaining the relevance of LISA and all the steps required for the procedure. The participants were then given a presentation of the simulation room, in particular the manikin and the equipment available. The briefing consisted of a succinct presentation of the medical situation. Each participant was asked to endorse the role of a physician belonging to the French pediatric emergency medical service transport team, arriving at a Level II maternity hospital where a 600 g newborn infant had just been born. The patient was in respiratory distress on oxygen and continuous positive airway pressure of 5 cm H 2 O. A facilitator was present in the simulation room and played the role of an emergency transport nurse. The participant was then invited to enter the simulation room. The debriefing was not carried out as a group after each run, as the other participants had to remain unaware of the scenario used. Each participant was invited to debrief individually with the two instructors. Videos of the simulation sessions were then independently viewed and evaluated by the two blinded raters (HR and BG). The evaluators used the rating scale to rate each participant's performance. All this was to allow the three next steps of the Downing method to be performed: Internal structure study: The aim of this step was to study the reliability of the rating scale. To do this, we need to evaluate the scores that the two observers gave to the participants during the simulation sessions. In this section, we assessed whether the number of items in the scale was correct and to what extent items in the same group were similar to each other. Comparison: The aim of this step was also to evaluate the reliability of the scale under modification of its score according to different groups of participants. Students partake in simulation sessions, and two raters score their performance using the rating scale. Two different observers should be able to assign a similar score to the same simulator run. Consequence: An evaluation tool must be able to distinguish between failures and successes, in other words between the degree to which participants are familiar with this technique. This last step reinforces the validity of the rating scale. Reliability analysis included internal consistency testing using a Cronbach’s alpha (CA) test and interrater reliability analysis using intraclass correlation coefficient (ICC), linear regression and its coefficient. F-test or t-test were used to compare scores, as appropriate. Validity analysis included a comparison of the mean scores obtained by technical novices and experts using a t-test.Statistical significance was assumed below a p -value threshold of 0.05. All the statistical tests were carried out on Excel® version 2205, published by Microsoft®. Description of the population Forty participants were recruited to attend the simulation sessions. They were all resident or senior pediatricians working in Paris or the surrounding areas. They all volunteered to participate in the simulation sessions, which took place over 7 sessions in February 2022. There were 5 to 6 participants per simulation session. The average age of the participants was 31.6 years, and there were 11 men and 29 women. Their characteristics are presented in Table . Six trainees were "experienced" in the LISA technique and 34 were "inexperienced" according to our definition. Statistical validation of the rating scale Reliability analysis The overall CA was 0.72 for the entire rating scale. A CA coefficient above 0.9 indicates that the rating scale has repetitive items. A score below 0.7 indicates poor internal consistency and could be explained by discrepancies between items or a lack of items in the analysis. The overall ICC for the scale was 0.91. There was no significant difference between the mean scores of the two raters (31.5, standard deviation = 6.2 versus 31.1, standard deviation = 5.4, p = 0.80). In linear logistic regression, the coefficient of determination (R2) of the scores between the two raters was 0.99 (Fig. ). Validity analysis The mean score on the rating scale for the LISA procedure was 36.8 for the experienced trainees versus 30.5 for the nonexperienced learners ( p < 0.001). The average expert score was 36.8/50, or 73.6%. Given the small number of learners defined as “experienced”, the results obtained by the participants according to their intubation experience were compared. Intubation experience was defined as having performed more than 10 intubations. LISA requires the use of a laryngoscope and visualization of the vocal cords. These two parts of the procedure are the most delicate to perform, yet they are essential for intubation. The hypothesis was that participants with experience with intubation would do better on the LISA than those without such experience. The scores were significantly higher for the learners with intubation experience than for those without (33.9 versus 29.7, respectively; p = 0.016). Forty participants were recruited to attend the simulation sessions. They were all resident or senior pediatricians working in Paris or the surrounding areas. They all volunteered to participate in the simulation sessions, which took place over 7 sessions in February 2022. There were 5 to 6 participants per simulation session. The average age of the participants was 31.6 years, and there were 11 men and 29 women. Their characteristics are presented in Table . Six trainees were "experienced" in the LISA technique and 34 were "inexperienced" according to our definition. Reliability analysis The overall CA was 0.72 for the entire rating scale. A CA coefficient above 0.9 indicates that the rating scale has repetitive items. A score below 0.7 indicates poor internal consistency and could be explained by discrepancies between items or a lack of items in the analysis. The overall ICC for the scale was 0.91. There was no significant difference between the mean scores of the two raters (31.5, standard deviation = 6.2 versus 31.1, standard deviation = 5.4, p = 0.80). In linear logistic regression, the coefficient of determination (R2) of the scores between the two raters was 0.99 (Fig. ). Validity analysis The mean score on the rating scale for the LISA procedure was 36.8 for the experienced trainees versus 30.5 for the nonexperienced learners ( p < 0.001). The average expert score was 36.8/50, or 73.6%. Given the small number of learners defined as “experienced”, the results obtained by the participants according to their intubation experience were compared. Intubation experience was defined as having performed more than 10 intubations. LISA requires the use of a laryngoscope and visualization of the vocal cords. These two parts of the procedure are the most delicate to perform, yet they are essential for intubation. The hypothesis was that participants with experience with intubation would do better on the LISA than those without such experience. The scores were significantly higher for the learners with intubation experience than for those without (33.9 versus 29.7, respectively; p = 0.016). The overall CA was 0.72 for the entire rating scale. A CA coefficient above 0.9 indicates that the rating scale has repetitive items. A score below 0.7 indicates poor internal consistency and could be explained by discrepancies between items or a lack of items in the analysis. The overall ICC for the scale was 0.91. There was no significant difference between the mean scores of the two raters (31.5, standard deviation = 6.2 versus 31.1, standard deviation = 5.4, p = 0.80). In linear logistic regression, the coefficient of determination (R2) of the scores between the two raters was 0.99 (Fig. ). The mean score on the rating scale for the LISA procedure was 36.8 for the experienced trainees versus 30.5 for the nonexperienced learners ( p < 0.001). The average expert score was 36.8/50, or 73.6%. Given the small number of learners defined as “experienced”, the results obtained by the participants according to their intubation experience were compared. Intubation experience was defined as having performed more than 10 intubations. LISA requires the use of a laryngoscope and visualization of the vocal cords. These two parts of the procedure are the most delicate to perform, yet they are essential for intubation. The hypothesis was that participants with experience with intubation would do better on the LISA than those without such experience. The scores were significantly higher for the learners with intubation experience than for those without (33.9 versus 29.7, respectively; p = 0.016). The present study developed a rating scale for the LISA procedure in preterm neonates in simulation by the modified Delphi method. The validation process of the scale (including 40 participants in seven simulation sessions) found good internal consistency (CA=0.72) and very good reliability interraters (R2 0.99 and ICC=0.91). In pediatrics, several rating scales have already been developed and published. They concern technical skills such as placing an intraosseous catheter , carrying out a lumbar puncture , or performing intubation in a newborn , as well as nontechnical skills such as early identification of sepsis or the announcement of bad news . To our knowledge, this scale is the first developed for teaching the LISA method in simulation. Not all the studies systematically assessed validation as we did. The results are consistent with those obtained in studies that have published recognized tools, such as the study by Oriot et al, with an ICC score of 0.947, or the study by Diaz et al, where the CA score was 0.87. Nevertheless, this tool has some limitations. First, some items of the rating scale could not be properly evaluated ("neutral head position" and "nontraumatic supports"). The first item, "neutral head position," could not be assessed because the low-fidelity manikin used had the head in a neutral position from the start. The participants were informed during the prebriefing that it was necessary to verbally state the procedures that could not be performed in the simulation. The observers were informed that part of the evaluation would be based on the oral instructions given by the participants. Many participants did not verbally report that they were checking the child's head position. The assessors could, therefore, not evaluate whether the neutral position of the manikin’s head was the participant's choice during the simulation session, which was related to the specific technical characteristics of the manikin. The second item that could not be assessed was the item "nontraumatic laryngoscope insertion". This difficulty with evaluation was due to the angle of the cameras with which the video of the sequence was recorded. In addition, the traumatic nature of laryngoscope insertion could be subjective, as there may be no visible lesion in the manikin's mouth. If this scale is to be used for training, it will be necessary to place a camera close to the mouth of the manikin to check the operator's support. These two items have been retained in the rating scale because they are essential elements in the performance of surfactant injection. Indeed, this rating scale was created with a double objective. It had to be a criterion-referenced assessment scale used for learners and, therefore, had to be valid and reliable to allow objective evaluation of the learners' skills. It was also intended to be a teaching aid for beginners. More frequent use of simulation for training in the healthcare field has been encouraged by the publication of the report "To Err is Human" . The aim of systematically using simulation in healthcare training is twofold: to reduce the risk of error and to avoid performing the procedure on a patient for the first time. However, performing the procedure on patients remains the final objective after simulation training. Some elements, such as "neutral head position", are difficult to assess in a criterion-referenced manner in simulation but must nonetheless be taught to the student so that they can perform the procedure correctly on patients after having learned it in simulation. The LISA method of surfactant delivery has become the standard of care in neonatology. This procedure is highly technical and requires appropriate training, which is best provided by manikin simulation. As part of this training, it is necessary to develop a rating scale for this procedure to assess learners in a criterion-referenced manner. In the present scale, there are eight steps, with a total of 25 items. Each item is scored either 0, 1, or 2. It was decided to use a 0, 1, or 2 score because some items contain several assessable elements, such as dressing or pain management. During the validation process of this scale, it was found that there was a significant difference between those already trained in LISA versus those naive to the technique. The mean result of the participants already trained in LISA was 36.8/50, i.e., a success rate of 73.6% for the items. This could suggest that a success rate at least of 73.6% appears to be predictive of satisfactory achievement of the LISA method. Moreover, this rating scale also contains a "reminder" part comprising the materials needed to perform the procedure as well as the medication. This latter part, added to the checklist of the scale per se, could be considered to be a cognitive aid that makes the LISA assessment tool (rating scale + reminder) usable in neonatal intensive care units. This possibility underlines the proximity and educational continuity between the simulation platform and real-life practice. The present work with the experts revealed that some elements such as sedation and the use of atropine before the procedure are still a matter of debate. Sedation before the procedure is essential; however, studies are still needed to define the most appropriate sedation for the procedure and the type of patient. Randomized controlled trials are underway, and the French Society of Neonatology has recently recommended the use of propofol for LISA sedation . In the literature, there are no randomized trials evaluating the administration of atropine before LISA. In observational studies, no severe adverse events have been reported . The French Society of Neonatology suggests that atropine should be administered preventively or in the event of bradycardia . It was decided to make these two items (sedation and atropine use) optional for use according to local practice. If these items are to be assessed, they should be presented during the prebriefing. In a follow-up study, it would be of interest to apply this technique in an intensive care unit and to use the evaluation rate to perform the training. Similarly, to assess the pedagogical contribution of this scale, it might be interesting to evaluate the success rate after the accelerated training of novices in the simulation technique. The novices could then perform the procedure once they obtained the average score obtained by the experts in simulation, i.e., at least 36.8/50 (or 73.6%). The LISA method is now the recommended first-line method of surfactant administration. It is not yet used in all neonatal units, probably in part because it is not routinely taught. This rating scale, presented here, is the first scale to evaluate and teach the LISA method in simulation. The psychometric testing of the scale yielded good reliability and validity. This rating scale could be used to train beginners in the LISA method. A score of more than 36/50, i.e., a 72% success rate, appears to indicate a good ability to perform the procedure. Training future physicians dealing with preterm infants to perform LISA in simulation is crucial to improve the quality of care. Following this study, research is currently underway to define a minimum passing score for this criterion referenced evaluation instrument, with a view to its use in summative evaluation. |
Role of Regulatory T Cells in Regulating Fetal-Maternal Immune Tolerance in Healthy Pregnancies and Reproductive Diseases | 21bf784d-e221-45a0-a49a-7f20c8af18b8 | 7333773 | Pediatrics[mh] | Regulatory T cells (Tregs), a key subset of T lymphocytes, play a critical role in regulating the immune response and maintaining immune tolerance both in physiological and pathological processes. Many studies have shown that Tregs are compromised in patients with autoimmune diseases as well as in patients with graft-versus-host disease after receiving transplanted organs , however, these cells are activated to promote tumor growth and progression, leading to the failure of immunotherapies in cancer . Defects in the number of Tregs and their suppressive activity are involved in the development of various systemic or organ-specific autoimmune diseases, including thyroiditis , gastritis , type I diabetes (T1D) , systemic lupus erythematosus (SLE) , multiple sclerosis (MS) , rheumatoid arthritis (RA) , and inflammatory bowel disease (IBD) . During the course of pregnancy, the mother's systemic immune system is altered to tolerate the fetus, who expresses paternal major histocompatibility complex antigens. Many studies have supplied multiple lines of evidence that Tregs possess specific characteristics for preventing the development of a maternal immune response against the fetus and maintaining fetal-maternal tolerance. First, the proportion of Tregs in peripheral blood is significantly increased during pregnancy in both women and mice, and there is a specific recruitment of Tregs from maternal peripheral blood to the fetal-maternal interface, leading to a higher proportion of Tregs in the placental decidua than in the peripheral blood . Furthermore, a decreased proportion of Tregs has been proposed to be associated with pregnancy-related complication such as recurrent spontaneous abortion and pre-eclampsia . Second, antibody-mediated depletion of CD25 + Tregs has been shown to cause implantation failure in allogeneic mated mice . Conversely, the adoptive transfer of Tregs attenuates the high abortion rates in the well-studied CBA/J × DBA/2J abortion-prone murine model . Pregnancy is a physiological process greatly dependent on immune tolerance, which is regulated by the number of Tregs and their suppressive activity. This review of the current literature describes the role played by Tregs in regulating fetal-maternal immune tolerance. Furthermore, we demonstrate the relationship between a deficiency of Tregs and pregnancy-related complications, with the aim of identifying the mechanisms through which Tregs maintain fetal-maternal immune homeostasis, thus providing a potential target for treating pregnancy-related complications. Differentiation and Immunosuppressive Function of Tregs Tregs are divided into two populations, namely natural regulatory T cells (nTregs) and inducible regulatory T cells (iTregs). NTregs originate from the thymus in response to self-antigens, whereas iTregs are peripherally induced from T cells responsible for restraining immune responses to foreign antigens, such as commensal bacteria, food antigens and allergens . The mechanism underlying how Tregs are generated remains controversial. Although some studies have suggested that Tregs are anergic to TCR (T cell receptor) stimulation in vitro , the process involving the formation and selection of Tregs in the thymus is highly dependent on the TCR rearrangement, as evidenced by the observation that the development of Tregs is abrogated in TCR transgenic mice with RAG-2 deficiency . An increasing number of studies have suggested that Tregs are positively selected from autoreactive T cells that express specific TCR with the appropriate affinity for self-peptides . Unlike other T helper cells, Tregs lack the capacity to secrete specific cytokines, and it is therefore difficult to distinguish them from other T helper cells. Foxp3 is the most specific Tregs marker and is constitutively expressed in Tregs generated in both the thymus and the periphery irrespective of the mode or state of activation . The Foxp3 gene contains 11 exons and maintains a high degree of conservation between human and mouse genes . Mice genetically deficient in Foxp3 lose the ability to properly regulate Tregs activity and succumb to a fatal and severe lymphoproliferative autoimmune syndrome at 3–4 weeks of age . Similar to mice, humans carrying a FOXP3 mutant gene develop an autoimmune syndrome named IPEX (immune dysregulation, polyendocrinopathy, enteropathy, X-linked syndrome) . Beyond its role as an indispensable factor required for the development of Tregs, continuous Foxp3 expression is required for the latter's suppressive function. Research has shown that Tregs isolated from Foxp3 deficient mice lack suppressive function. However, transduction of Foxp3 endows CD4 + CD25 − T cells with the capacity to suppress the proliferation of CD4 + T cells . The suppressive function of Tregs is achieved via two mechanisms, namely a cell-contact dependent mechanism involving the recognition of co-stimulated molecules that directly suppress the expansion of effector T cells and a cell-contact independent mechanism involving the secretion of soluble cytokines that negatively regulate the immune response .
Tregs are divided into two populations, namely natural regulatory T cells (nTregs) and inducible regulatory T cells (iTregs). NTregs originate from the thymus in response to self-antigens, whereas iTregs are peripherally induced from T cells responsible for restraining immune responses to foreign antigens, such as commensal bacteria, food antigens and allergens . The mechanism underlying how Tregs are generated remains controversial. Although some studies have suggested that Tregs are anergic to TCR (T cell receptor) stimulation in vitro , the process involving the formation and selection of Tregs in the thymus is highly dependent on the TCR rearrangement, as evidenced by the observation that the development of Tregs is abrogated in TCR transgenic mice with RAG-2 deficiency . An increasing number of studies have suggested that Tregs are positively selected from autoreactive T cells that express specific TCR with the appropriate affinity for self-peptides . Unlike other T helper cells, Tregs lack the capacity to secrete specific cytokines, and it is therefore difficult to distinguish them from other T helper cells. Foxp3 is the most specific Tregs marker and is constitutively expressed in Tregs generated in both the thymus and the periphery irrespective of the mode or state of activation . The Foxp3 gene contains 11 exons and maintains a high degree of conservation between human and mouse genes . Mice genetically deficient in Foxp3 lose the ability to properly regulate Tregs activity and succumb to a fatal and severe lymphoproliferative autoimmune syndrome at 3–4 weeks of age . Similar to mice, humans carrying a FOXP3 mutant gene develop an autoimmune syndrome named IPEX (immune dysregulation, polyendocrinopathy, enteropathy, X-linked syndrome) . Beyond its role as an indispensable factor required for the development of Tregs, continuous Foxp3 expression is required for the latter's suppressive function. Research has shown that Tregs isolated from Foxp3 deficient mice lack suppressive function. However, transduction of Foxp3 endows CD4 + CD25 − T cells with the capacity to suppress the proliferation of CD4 + T cells . The suppressive function of Tregs is achieved via two mechanisms, namely a cell-contact dependent mechanism involving the recognition of co-stimulated molecules that directly suppress the expansion of effector T cells and a cell-contact independent mechanism involving the secretion of soluble cytokines that negatively regulate the immune response .
Cell-contact dependent suppressive activity is mediated via the recognition of co-stimulated molecules. In this process, Tregs function is highly dependent on the normal expression of molecules located on Tregs, and a deficiency of key molecules triggers the defective expansion and suppressive activity of Tregs, leading to a disturbance of immune homeostasis. IL-2 receptor α (IL-2Rα) and CTLA4 are the most important molecules involved in cell-contact dependent mechanism. Most Tregs abundantly express high-affinity IL-2 receptor α (CD25) and IL-2/IL-2R signaling provides indispensable signaling during the development and maturation of Tregs both in the thymus and in the periphery. Furthermore, the lack of the IL-2R cannot be compensated by other cytokine receptors . IL-2, IL-2Rα, and IL-2Rβ deficient mice all die from severe lymphoproliferation and autoimmune disease in early life. In addition, neutralization of circulating IL-2 by anti–IL-2 monoclonal antibodies inhibits Tregs proliferation and triggers a wide range of organ specific autoimmune diseases . IL-2-IL-2R signaling is essential for the development and maturation of both Tregs and Teff cells, however, low dose IL-2 is remarkably efficacious in promoting the expansion of Tregs rather than Teff cells, which possibly results from the higher affinity of IL-2R in Tregs . Based on the comparative activity and different sensitivity for IL-2, the consumption of IL-2 by Tregs has been shown to be a predominant mechanism involved in suppressing the expansion and activity of Teff cells and triggering Teff cell apoptosis due to IL2 deprivation . CTLA4, a key molecule constitutively expressed in Tregs, is crucial for maintaining T cells homeostasis and tolerance induction, and its expression is in part controlled by Foxp3 . Mice deficient in CTLA4 become sick by 2 weeks of age and moribund at 3–4 weeks of age, with diffuse and focal lymphocytic infiltration into various organs . Furthermore, specific deficiency of CTLA4 in Tregs results in the spontaneous development of systemic lymphoproliferation, multi-organ lymphocyte infiltrations, fatal T cell-mediated autoimmune diseases, and hyperproduction of immunoglobulin E in mice . CTLA-4-mediated suppressive regulation of T cell response and upregulation of Tregs activation are predominantly achieved by competition with CD28, a positive costimulatory molecule that shares common ligands (CD80/CD86) with CTLA4 . CTLA4 possesses significantly higher affinity in binding CD80/CD86 and CTLA4 rather than CD28 removes costimulatory ligands CD80/CD86 from APCs by a process of trans-endocytosis . These properties equip CTLA4 with the capacity to outcompete the ability of CD28 to serve as a negative immune regulator .
In addition to the cell-contact dependent mechanism, Tregs also exert suppressive activity in a cell-contact independent manner, mainly through the secretion of inhibitory cytokines. Unlike other T cells, Tregs fail to produce exclusive cytokines. However, certain cytokines, such as TGF-β and IL-10, secreted by Tregs are essential for the expansion and suppressive activity of Tregs. Several lines of evidence suggest that the addition of TGF-β enhances the conversion rate of native T cells into Tregs, and that TGF-β secreted by Tregs plays a partial role in maintaining suppressive properties by binding to the TGF-β receptor . Administration of neutralizing antibodies specific for TGF-β or specific deficiency of TGF-β expression in Tregs leads to a limitation or even abrogation of Tregs' suppressive activity . Strong evidence that the role of TGF-β to maintain Foxp3 expression is supported by the observation that the expression of Foxp3 is dramatically diminished in peripheral Tregs from TGF-β −/− mice and addition of TGF-β results in increased Foxp3 expression . Unlike TGF-β, the function of IL-10 in Tregs seems to be organ-specific. Recent studies have found that IL-10 and IL-35 produced by intratumoral Tregs cooperatively share a common BLIMP1 axis to promote the exhausted intratumoral T cell state and anti-tumor immunity, implying IL-10 and IL-35 contribute to maintaining immune tolerance . IL-10 is recognized as a potent suppressor of macrophage and T cell functions. Furthermore, IL-10 deficient mice are growth retarded and suffer from chronic enterocolitis . An increasing number of current studies have found that IL-10 is expressed in Tregs and plays an auxiliary role in promoting their expansion and function. IL-10 + Tregs are mostly located in intestinal tissues and are essential for limiting immune response-induced inflammation to the diverse intestinal microbiota, which may provide a reasonable explanation as to why IL-10 deficient mice or mice treated with anti-IL-10 receptor blockers succumb to intestinal inflammation . Although the Tregs-specific deficiency in IL-10 does not result in severe systemic autoimmunity, it does lead to immunological hyperreactivity at environmental interfaces, resulting in conditions such as spontaneous colitis, lung hyperreactivity, and skin hypersensitivity . Thus, while IL-10 production by Treg cells is not necessary for the regulation of systemic autoimmunity, it is essential for hindering excessive immune responses at local environmental interfaces. The suppressive activity of IL-10 is partly mediated via binding to IL-10R to restrain the Th17-induced inflammatory response, which plays a critical role in regulating intestinal homeostasis. This is illustrated by the observation that mice with IL-10R deficient Tregs produce high levels of IL-17 and are prone to developing severe colitis . Regulation of Fetal-Maternal Tolerance During Healthy Pregnancy For decades, many studies have shown that successful pregnancy depends on the homeostasis of fetal-maternal tolerance. Furthermore, failure of the maternal immune system to establish fetal-maternal tolerance is the predominant trigger in the development of pregnancy-related complications. Consequently, numerous therapeutic treatments aimed at suppressing the maternal immune system are employed in clinics. However, the effect of these therapies is not always apparent and is often accompanied by various side effects. It is therefore important to identify the cellular and molecular mechanisms responsible for establishing fetal-maternal immune tolerance in healthy and abnormal pregnancies to promote the development of targeted therapeutic interventions. The immune suppressive property of Tregs confers this cell population with a fundamental role in establishing the fetal-maternal immune tolerance necessary for successful pregnancy. Some studies consider pregnancy to be a process of mutual conversion between pro-inflammatory and anti-inflammatory conditions , therefore dividing pregnancy status into three distinct immunological states that correspond to different stages of fetal development: first, a pro-inflammatory stage associated with embryo implantation and placentation ; second, an anti-inflammatory-oriented stage associated with fetal growth ; and third, a switch from an anti-inflammatory to a pro-inflammatory stage necessary for the initiation of labor . Concurrent with the above stages is a dramatic change in the number of Tregs during the course of pregnancy. Following exposure of paternal alloantigens, circulating Tregs increase rapidly during the early pregnancy stage and peak during the second stage at which time trophoblast invasion of the maternal decidua is maximal; then, Tregs gradually decrease when labor begins . The change in the number of Tregs and crosstalk with other immune cells play a critical role throughout the entire course of pregnancy.
For decades, many studies have shown that successful pregnancy depends on the homeostasis of fetal-maternal tolerance. Furthermore, failure of the maternal immune system to establish fetal-maternal tolerance is the predominant trigger in the development of pregnancy-related complications. Consequently, numerous therapeutic treatments aimed at suppressing the maternal immune system are employed in clinics. However, the effect of these therapies is not always apparent and is often accompanied by various side effects. It is therefore important to identify the cellular and molecular mechanisms responsible for establishing fetal-maternal immune tolerance in healthy and abnormal pregnancies to promote the development of targeted therapeutic interventions. The immune suppressive property of Tregs confers this cell population with a fundamental role in establishing the fetal-maternal immune tolerance necessary for successful pregnancy. Some studies consider pregnancy to be a process of mutual conversion between pro-inflammatory and anti-inflammatory conditions , therefore dividing pregnancy status into three distinct immunological states that correspond to different stages of fetal development: first, a pro-inflammatory stage associated with embryo implantation and placentation ; second, an anti-inflammatory-oriented stage associated with fetal growth ; and third, a switch from an anti-inflammatory to a pro-inflammatory stage necessary for the initiation of labor . Concurrent with the above stages is a dramatic change in the number of Tregs during the course of pregnancy. Following exposure of paternal alloantigens, circulating Tregs increase rapidly during the early pregnancy stage and peak during the second stage at which time trophoblast invasion of the maternal decidua is maximal; then, Tregs gradually decrease when labor begins . The change in the number of Tregs and crosstalk with other immune cells play a critical role throughout the entire course of pregnancy.
Embryo implantation is the initial stage of pregnancy and involves apposition of the blastocyst and the uterine endometrium followed by attachment and invasion of the blastocyst into the endometrium, and reconstruction of the decidua by the invasive trophectoderm . The wide application of assisted reproductive technology, such as in vitro fertilization-embryo transplantation (IVF-ET) and intrauterine insemination (IUI), has enabled an analysis of earlier gestational stages from oocyte fertilization to implantation in humans. Adequate endometrial receptivity is considered a pivotal precondition for successful embryo implantation. Endometrial scratching before embryo transfer has been proposed as a clinical treatment to increase uterine receptivity, and some studies have demonstrated that endometrial scratching improves the pregnancy outcome by triggering an inflammatory response and enhancing angiogenesis at the implantation site, providing indirect evidence for the role played by inflammation during implantation . Studies based on human and animal experiments have demonstrated that the peri-implantation period is accompanied with the activation and infiltration of various immune cells . Uterine-specific natural killer (uNK) cells, macrophages (Mos), and dendritic cells (DCs) are recruited at the implantation site and exert prominent immune-regulatory effects during early pregnancy. uNK cells are the most abundant immune cells located in human decidua during early pregnancy, while Mos and DCs serve as antigen-presenting cells that infiltrate into the decidua. Crosstalk among these cells plays an essential role in regulating trophoblast invasion and in promoting spiral artery remodeling . The role played by Tregs during implantation is unclear. However, some studies have reported that a reduced number of Tregs is associated with implantation failure. Mice with a depletion of Tregs exhibit a significant defect in implantation, which is reversed following an adoptive transfer of Tregs . A study showed that compared with fertile women, endometrial tissue from women with unexplained infertility displayed a significant reduction Foxp3 mRNA expression, the fate-determining transcription factor especially expressed in Treg cells . Other evidence has also revealed a correlation between the level of Tregs in peripheral blood and the implantation rate. Women with implantation failure after IVF or artificial insemination by donor sperm (AID) had a significantly decreased percentage of Tregs compared with women with a successful pregnancy . Therefore, the presence of peripheral or local Tregs may create a limited but necessary immunomodulatory function during the course of implantation.
Successful implantation is followed by a phase of fetal growth and development. The establishment of fetal-maternal immune tolerance lays the foundation for this stage, with a shift from a pro-inflammatory immune response to a Th2/Treg-predominant anti-inflammatory immune tolerance . The proportion of Tregs begins to rise and peaks at this stage, and a paucity of Tregs could lead to pregnancy-related complications such as spontaneous abortion. Tregs exert a strong immunosuppressive function to maintain an anti-inflammatory environment and protect the fetus from maternal immunological rejection. Tregs can effectively suppress the expansion and activation of effector T cells via a classic cell-contact mechanism or by secreting suppressive cytokines as described previously. One study described a class of functionally distinct Tregs with expression of a co-inhibitory molecule TIGIT, which induces selective suppression of Th1 and Th17 cells but not Th2 cells. However, whether this Tregs subset is expanded and activated during pregnancy is still unknown . The pivotal role played by Tregs in fetal-maternal tolerance raises several questions about the mechanisms responsible for their expansion during pregnancy and underscores the need for studies investigating these mechanisms. Previous studies suggest that the activation and regulation of Tregs is primarily impacted by antigen exposure and the dynamic changes of steroid hormones that occur during pregnancy.
Investigators have proposed that exposure to male seminal fluid delivered during mating elicits the expansion of maternal Tregs, as evidenced by the increase in the number of Tregs within the period of time subsequent to mating and before embryo implantation . Immune tolerance to the fetus is necessary for successful pregnancy, and transmission of seminal fluid seems to play a priming role prior to implantation by promoting expansion of Tregs, thereby activating specific tolerance to paternal alloantigens. Seminal fluid contains various components, including a cellular fraction that contains sperm, leukocytes and epithelial cells and a non-cellular fraction of compounds such as TGF-β and prostaglandins. The cellular and acellular fractions in semen both contain several antigens, including classical class Ia, non-classical class Ib and minor antigens such as H-Y antigen, which drive an antigen-dependent expansion of Treg cells . The non-cellular components are also required to confer tolerance. As mentioned above, TGF-β is a critical cytokine for Tregs proliferation. One study found that intravaginal pre-treatment with TGF-β at mating enhances successful pregnancy in vivo in a well-established murine model . An in vitro experiment also indicated a role for prostaglandins in upregulating Foxp3 expression and enhancing Tregs function . Collectively, both sperm and seminal plasma may contribute to driving an expansion of Tregs and providing an immune-privileged environment that is beneficial for subsequent embryo implantation. Embryo implantation and fetal growth are the most important stages during pregnancy. Some studies have proposed that the implanted blastocyst should be considered a semi-allograft and constant immunosuppression is required for a pregnancy to be successful. Although a seemingly opposite pro-inflammatory process is involved in both implantation and initiation of labor, immunosuppression is an indispensable response to maintain immune homeostasis during the fetal growth stage, and this is highly dependent on the expansion and activation of Tregs triggered by the fetal alloantigens . When Tregs are depleted, fetal outcome is normal in syngeneic pregnancies rather than allogeneic pregnancies, suggesting that Tregs suppress maternal immune responses directed against fetal alloantigens rather than male-specific minor histocompatibility antigens . When encountered with parental alloantigens presented by a fetus, peripheral Tregs, generated extrathymically and induced by non-self-antigens, serve as the predominant subset suppressing immune response. The development of peripheral Tregs is dependent on the expression of a Foxp3 enhancer CNS1, a deficiency of which leads to an increased resorption of embryos in mice .
Serum levels of the pregnancy-associated hormones such as estrogen, progesterone, and human chorionic gonadotropin (HCG) increase dramatically during pregnancy. These hormones play an essential role in maintaining immune tolerance and in supporting successful pregnancy. Currently, there is increasing evidence that the mechanisms through which hormones contribute to immune homeostasis during pregnancy are in part due to the expansion of Tregs and their suppressive activity . Estrogen-based therapy has been reported to alleviate symptoms associated with several autoimmune diseases, such as collagen-induced arthritis , type1diabetes , and autoimmune encephalomyelitis . Furthermore, the mechanisms underlying these protective effects seem to be associated with changes in immune cells and cytokines . The number of Tregs in human peripheral blood change continuously during the menstrual cycle and peak before ovulation, which is concurrent with the change of the concentration of estrogen, suggesting that estrogen may be a powerful factor in promoting Tregs expansion . Some studies have demonstrated that the proliferation and suppressive activity of human Tregs observed with estrogen treatment is mediated through estrogen receptor α . In both in vivo and in vitro experiments, estrogen treatment triggers the expansion of Tregs. Furthermore, the addition of estrogen in combination with TCR stimulation enhances Foxp3 mRNA expression in CD4 + CD25 − T cells in vitro , suggesting that estrogen may potentially induce the conversion of CD4 + CD25 − T cells to Tregs . Progesterone, which is mainly produced by the placenta and is markedly elevated during pregnancy, functions as a regulator that maintains homeostasis at the maternal-fetal interface. Similar to estrogen, progesterone is considered to be another important hormone that promotes the expansion of Tregs and their suppressive capacity . The proportion of Tregs and the conversion rate of CD4 + CD25 − T cells into Tregs has been shown to increase significantly in the peripheral blood, spleen, and inguinal lymph nodes of ovariectomized mice after progesterone injection . Progesterone-mediated immune tolerance is achieved by progesterone binding to the glucocorticoid receptor rather than to the progesterone receptor . Progesterone promiscuously binds the glucocorticoid receptor and promotes immune suppression by inducing enrichment of Treg cells and triggering apoptosis of effector T cells, which is based on the preferred sensibility in effector T cells for glucocorticoid receptor-mediated T cells death compared with that in Tregs . Progesterone is also present at high levels in human cord blood where it has been reported to have an immune-suppressive function. Progesterone drives a shift of native cord blood T cells into suppressive Tregs, while impeding the conversion from native T cells into Th17 cells, another potential pathway through which progesterone regulates immune tolerance . HCG is another hormone that is increased during pregnancy, and is produced in the blastocyst after fertilization, reaching its maximum level at the 11th week and then gradually decreasing until birth . Khil et al. reported that HCG prevents the development of autoimmune-mediated diabetes in NOD mice by downregulating immune effector cells and cytokines and simultaneously upregulating the proportion of Tregs and the levels of TGF-β and IL-10, suggesting that HCG is an effective regulator for immune tolerance . Increased HCG during pregnancy provokes many Tregs-related responses including (1) augmenting the number of Tregs, (2) increasing their local and systemic suppressive function, (3) enhancing attraction of circulating Tregs into decidua, and (4) increasing the secretion of suppressive cytokines . HCG-mediated expansion of Tregs is achieved in part by retaining DCs in an immature state, leading to the generation of Tregs and a loss of the capacity to activate a T cell-mediated immune response . In vitro migration assays further confirmed the chemoattractant properties of HCG that promote migration of Tregs from the periphery into the uterus, which is potentially mediated via binding to HCG/LH receptors located on Tregs .
There is a decline in the number of Tregs as pregnancy progresses into the third gestational period. The reduction in the number of Tregs in late gestation may be a contributing factor for the initiation of spontaneous labor. This is supported by the finding that the proportion of Tregs in the decidua following a spontaneous vaginal delivery is significantly lower than that following an elective cesarean section . Shah et al. conducted a longitudinal analysis from 20-weeks gestational age to labor and observed a reduction in the number of activated Tregs (defined as Tregs with HLA-DR + ) and a significant shift toward a Th1/Th17 response with the onset of labor . Compared with women undergoing spontaneous term labor, the proportion of activated Tregs is significantly decreased in women in preterm labor . The change is similar to the reduction in activated Tregs observed in patients who experience an acute rejection after kidney transplantation, supporting that the reduction in the proportion and activity of Tregs promotes the conversion from an anti-inflammatory to a pro-inflammatory stage and plays a critical role in initiating spontaneous labor. The mechanism underlying the reduction in Tregs during labor remains an enigma. The alteration in hormone levels and in the microbial environment may be stimuli for activating an inflammatory response, however, the specific molecular mechanisms needs to be further investigated. The level of Tregs progressively decreases after delivery. However, there is a retention of “memory” Tregs with fetal specificity, which retain the ability to generate a more effective and accelerated suppressive response when re-exposed to the same fetal antigens in subsequent pregnancies . The primary pregnancy confers Tregs with a protective regulatory memory, which may provide an immunological basis for protection against complications such as pre-eclampsia in a subsequent pregnancy . Dysfunction of Tregs in Reproductive Diseases Since it has been determined that Tregs maintain fetal-maternal tolerance during the normal course of embryo implantation and pregnancy, it is of interest to investigate whether systemic and local maldistribution and dysfunction of Tregs play a role in the etiology of infertility and pregnancy-related complications. Increasing evidence suggests that a deficiency in the expansion and function of Tregs as well as an abnormal expression of key molecules are linked to pregnancy-related complications.
Since it has been determined that Tregs maintain fetal-maternal tolerance during the normal course of embryo implantation and pregnancy, it is of interest to investigate whether systemic and local maldistribution and dysfunction of Tregs play a role in the etiology of infertility and pregnancy-related complications. Increasing evidence suggests that a deficiency in the expansion and function of Tregs as well as an abnormal expression of key molecules are linked to pregnancy-related complications.
Recurrent spontaneous abortion (RSA), defined as the loss of three or more consecutive pregnancies, affects ~1% of women attempting to conceive . RSA is a complex pregnancy-related complication that is due to multiple factors including chromosomal abnormalities, congenital or acquired anatomical defects in the uterine fundus and cervix, and other endocrine diseases such as PCOS, diabetes, thyroid disorders, and others related to aberrant immune responses . Increasing evidences suggests that the proportion of various immune cells and cytokines is altered in patients with RSA, supporting that immune dysfunction may be a contributing factor to its etiology . Although there have been detailed guidelines describing clinical interventions for managing women with RSA, treatment based on immune rejection as a potential etiology is controversial, because no definite cellular and molecular mechanism has been discovered to date . The mechanisms through which Tregs contribute to RSA primarily involve an imbalance of the Th1/Th2/Th17/Treg cells paradigm and the abnormal proportion and activity of Tregs. Dysregulation of T lymphocyte homeostasis is also involved in the etiology of RSA. In peripheral blood from patients with RSA, the balance between Th1 and Th2 cells is disrupted in favor of Th1 cells, and the ratio of Th17/Treg cells is skewed toward Th17 cells . It is widely accepted that there is a close interaction between the expansion of Tregs and the secretion of IL-17. When IL-17 combines with the IL-17 receptor, Tregs are upregulated. Conversely, Tregs suppress the proliferation of Th17 cells and the secretion of IL-17 via cell-cell contact and via Il-10/TGF-β-mediated effects . However, this suppressive function of Tregs is abrogated in patients with RSA . Transfusion of Tregs into mice pretreated with IL-17 has been shown to significantly increase the expression of IL-10 and TGF-β, two key cytokines that mediate the suppressive activity of Tregs in decidua and lower the fetal resorption rates in mice . Furthermore, insufficient generation of pregnancy-induced Tregs triggers the accumulation of paternal alloantigen specific Th1 cells and directly results in the failure to establish appropriate maternal-fetal immune tolerance . Numerous studies have also confirmed that the reduction in the number of Tregs are involved in the pathogenesis of RSA . Sasaki et al. first reported the presence of Tregs in the decidua and demonstrated the proportion of Tregs in decidua from spontaneous abortions was significantly lower than that in decidua from induced abortion . Other studies have also demonstrated that the proportion of Tregs and the expression of Foxp3 in both the decidua and peripheral blood from patients with unexplained RSA patients are significantly lower than those from women with normal pregnancies . In addition to the reduction in number, Lourdes et al. reported that the suppressive function of Tregs is significantly impaired in RSA as assessed by a co-culture technique with CD4 + CD25 − T cells . Inadequate number of Tregs and downregulation of Treg cell activity impair the anti-inflammatory environment, weaken the immune tolerance against fetal rejection and thereby increase the risk of RSA.
Endometriosis is a benign gynecological disease affecting ~6–10% women of childbearing age, and is characterized by the implantation of endometrial tissues outside the uterus . Chronic pelvic pain, dysmenorrhea and infertility are the common symptoms occurring in patients with endometriosis . As multiple factors, including genetic and environmental factors, contribute to the development of endometriosis, the pathogenesis of endometriosis remains uncertain. Many theories have been proposed to explain how endometriosis develops, and one of the most widely accepted is the retrograde menstruation theory. This theory hypothesizes that fragments of endometrial tissue reflux to the peritoneum through the fallopian tubes during menstruation . However, this theory fails to explain why only a few women develop endometriosis even though retrograde menstruation is a common phenomenon occurring in most women of childbearing age . Therefore, other studies have postulated that a disturbed local and systemic immune response may be responsible for the development and progression of endometriosis . An aberrant immune environment that includes alternative activation of peritoneal macrophages , production of various cytokines , and reduction in natural killer cell cytotoxicity , all contribute to the survival and invasion of ectopic endometrial tissue. Dysregulation in T lymphocyte homeostasis is associated with the pathogenesis of endometriosis. The Th1/Th2 balance is altered in local and systemic immune conditions, such that there is skewing toward Th2 cells in endometriotic lesions, but skewing toward Th1 cells in peripheral blood . A disturbance in Tregs activity may be a more prominent mechanism involved in the etiology of endometriosis due to their immune-suppressive function, derangement of which could potentially promote the survival of ectopic endometrial lesions. However, evidence regarding the change in the proportion of Tregs in peripheral blood, peritoneal fluid, eutopic endometrium, and ectopic endometrial tissues among patients with endometriosis is inconsistent . The discrepancy may result from differences in patient selection, namely the patients with early or advanced endometriosis. Most studies suggest the proportion of Tregs is significantly increased in peritoneal fluid from women with endometriosis compared with control women . One study reported that the number of Tregs was increased in the peritoneal fluid and decreased in the peripheral blood, and another study found the number of Tregs was higher in peritoneal fluid than in peripheral blood, both indicating that active translocation of Tregs occurs from circulation to the local peritoneal cavity . However, some studies failed to find any difference in the proportion of Tregs in patients with endometriosis when compared with women without endometriosis . To bypass the confounding influence of interpatient variability, research has been carried out in an established animal model with endometriosis to identify abnormalities in the proportion of Tregs. In a study of baboons with induced endometriosis, the proportion of Tregs was decreased in peripheral circulation and eutopic endometrium but increased in ectopic tissue, which is consistent with Tregs' local immunosuppressive activity Tregs played . Tanaka et al. focused on the variation in resting and activated Tregs and put forth a new concept that the proportion of activated Tregs in the endometrioma rather than in the peritoneal fluid or peripheral blood is decreased, which may be temporal and associated with the angiogenesis and progression of endometriosis . However, a study showed the proportion of Tregs in ectopic endometrium was increased in patients with endometriosis compared with eutopic endometrium . Further research is required with an expanded sample size and more detailed subgroup analysis to better determine the role Tregs play in the pathogenesis of endometriosis. Change in the proportion of Tregs appears to contribute to the suppressed immune response against ectopic endometrial tissue, permitting implantation of endometrial tissue in the peritoneal cavity. Therefore, understanding the origin of local Tregs production may be provide new insights that will aid in the development of targeted therapies for women with endometriosis. The accumulation of Tregs in the peritoneal cavity may not only be a result of active translocation from the peripheral blood but may also be due to their local induction . Higher levels of IL-10 and TGF-β, two key cytokines responsible for regulating the proliferation and activity of Tregs, were found in the peritoneal fluid and serum of patients with endometriosis than in normal controls . Compared with serum levels, the level of cytokines in peritoneal fluid was significantly higher . Furthermore, IL-10 and TGF-β mRNA expression were significantly higher in ectopic lesions than eutopic endometrium from women with or without endometriosis, particularly in cases of advanced endometriosis . These results suggest Tregs and related cytokines maintain the local anti-inflammatory environment and play a crucial role in the development of endometriosis.
Preeclampsia is a common pregnancy-related complication that occurrs in 3–5% of pregnant women and can lead to iatrogenic preterm birth and fetal growth restriction . The precise etiology of preeclampsia remains unknown, although insufficient formation of uterine spiral arteries, over-activated inflammation, injured endothelial cells, and genetic factors have all been implicated . Interestingly, preeclampsia seems to be more common in primiparous than multiparous women, whereas the protective effect is abrogated with the change of partner. A meta-analysis compared the difference in the risk of preeclampsia in women who were impregnated by donor or partner sperm and found the risk was significantly increased in conceptions resulting from donor sperm . Furthermore, another study reported that prior and prolonged partner sperm exposure before pregnancy is associated with a significant reduction of the risk of preeclampsia . Taken together, these observations suggest that paternal antigens and sperm exposure induce an immune tolerance during the first pregnancy and offer effective protection against the development of preeclampsia with subsequent pregnancies, implying the adaptive immune response with alloantigen specificity and immunological memory is involved in the pathogenesis of preeclampsia . An increasing body of evidence suggests that an inadequate immune tolerance induced by Tregs-associated abnormalities play a pivotal role in the etiology of preeclampsia. Several studies have reported that, compared with normal pregnancy, both the number of Tregs and the ratio of Tregs to Th17 cells in peripheral blood are significantly reduced in preeclampsia . The increased ratio of Th17/Treg cells has also been confirmed by an analysis of Th17/Treg expression of related transcription factors and the secretion of Th17/Treg-related cytokines. Compared with healthy pregnant women, a reduction in the expression of Treg-specific transcription factor Foxp3 and an elevation in Th17-specific transcription factor RORγt in patients with preeclampsia has been reported . Furthermore, analysis of cytokine profiles have revealed a significant decrease in IL-10, and a significant increase in IL-17 levels in patients with preeclampsia . Taken together, these studies suggest that a shift occurs from Tregs to Th17 cells in the development of preeclampsia, leading to an abnormal immune state that triggers inflammation and an impairment of immune tolerance. The mechanism underlying the imbalance of Th17/Treg cells remains unclear. Eghbal-Fard et al. suggested the upregulation of miRNA in patients with preeclampsia may affect the differentiation and expansion of Th17/Treg cells by regulating the expression of specific transcription factors . In addition to the alteration in the proportion of Tregs, the immunosuppressive activity of Tregs is also altered in patients with preeclampsia. Darmochwal-Kolarz et al. reported the proliferation of effector T lymphocytes in patients with preeclampsia was significantly inhibited by Tregs isolated from healthy pregnant women. However, the suppressive response was lost if replaced with Tregs from patients with preeclampsia . The recruitment of Tregs from peripheral blood into decidua and the local expansion of decidual Tregs are important for maintaining fetal-maternal immune tolerance at the fetal-maternal interface. It has been well-established that the proportion of Tregs in decidua is decreased in preeclampsia . Though the reduction of decidual Tregs may be associated with an imbalance in systemic Tregs, local expansion may also play an important role. TCR repertoire analysis of decidual Tregs showed an insufficient clonal expansion of decidual Tregs in preeclampsia compared with healthy pregnancy . In normal pregnancy, induced rather than native Tregs are the dominant Tregs subset located in the decidua and are clonally expanded, while the expansive and suppressive capacity of iTregs is significantly impaired in preeclampsia . The local induction of Tregs depends on specific APCs within the decidual microenvironment. A significant reduction in the expression of HLA-G and ILT4 on decidual APCs is observed in preeclampsia compared with normal pregnancy, providing a possible clue to the lack of iTregs in preeclampsia . An aberrant proportion and type of Tregs in the decidua disturb the immune homeostasis during pregnancy and promote the development of preeclampsia. Tregs and Immune Therapy During Pregnancy Taken together, the above studies suggest that Tregs play a prominent role in regulating fetal-maternal immune tolerance, and a defect in the proportion and activity of Tregs is involved in the development of RSA, endometriosis, and preeclampsia. Thus, approaches designed to boost the proportion of Tregs or strengthen their suppressive function may lead to promising strategies for treating pregnancy-related diseases. Several Tregs-based target therapies are entering into clinical trials, including adoptive Treg cell therapy, Tregs-enhancing drugs, and low dose IL-2 administration . Administration of purified Tregs was firstly applied as Tregs-based target therapy. With the development of immune cell therapy, antigen-specific Tregs therapy was also proposed for treating autoimmune and graft-versus-host diseases. Phase I/II clinical trials aimed to explore the curative effect, and some have reported that Tregs administration alleviates clinical symptoms induced by autoimmunity . Some research has attempted to determine whether Tregs administration improves pregnancy outcomes. Yin et al. and Wang et al. examined the effectiveness of adoptive transfer of Tregs in preventing spontaneous abortion in mice models . Yin et al. established an abortion-prone pregnancy mice model with DBA/2J-mated pregnant CBA/J mice and performed adoptive transfer of freshly isolated and in vitro expanded Tregs from non-pregnant CBA/J mice. Wang et al. induced spontaneous abortions by administration of IL-17 in a CBA/J × BALB/c mouse model of normal pregnancy and performed adoptive transfer of in vitro expanded Tregs purified from pregnant CBA/J mice. These two studies demonstrated transfusion with in vitro expanded Tregs promotes immune suppressive activity, increases the secretion of suppressive cytokines and significantly reduces the rate of spontaneous abortion. Although Treg cell therapy has not been widely used in clinical practice, clinical research has initiated several non-specific immunotherapies partially regulating the proportion and activity of Tregs for the treatment of pregnancy-related diseases. Intravenous immunoglobulin G (IVIG) and paternal or third-party lymphocyte immunization therapy have been proposed for the treatment of patients with RSA due to the potential immunomodulatory effects. Although the benefit for these immunotherapies is controversial, a growing body of evidence suggests that they may increase rates of live birth and decrease rates of miscarriage . A variety of studies and clinical trials have reported both IVIG and lymphocyte immunization therapy correct the Tregs defect and rebalances the Th17/Treg paradigm in peripheral blood. Compared with a control group, the treatment triggers a shift toward Tregs in the Th17/Treg balance by enhancing the expansion of Tregs, promoting the secretion of suppressive cytokines, and inhibiting Th17 cells proliferation . Tregs-enhancing drugs are another type of Tregs-based target therapy. Rapamycin (Sirolimus) is an mTOR inhibitor, which acts as an immunosuppressive drug by selectively promoting the expansion of Tregs and inducing differentiation of T helper cells into Tregs. Royster et al. established a murine model with conditional knockdown of Tregs induced by diphtheria toxin. They found the deletion of Tregs decreased litter sizes and triggered embryo implantation failure, effects that were reversed after the treatment with rapamycin . A multicenter, double-blind, phase II randomized clinical trial administrated 2 mg/day of sirolimus for 2 days before embryo transfer to patients receiving IVF-ET therapy and who had a history of recurrent implantation failure. The study collected blood samples and assessed the ratio of Th17/Treg cells by flow cytometry 5–10 days prior to the initiation of an IVF cycle. Only patients with a high ratio of Th17/Treg cells were included in this trial. The trial reported that the administration of sirolimus reversed the imbalance in the ratio of Th17/Treg cells and significantly increased the rate of clinical pregnancy and live birth compared with those in the control group . Taken together, some studies have demonstrated the effectiveness of Tregs-based therapy in treating several autoimmune diseases and cases of organ transplantation. However, the methods cannot be directly applied for pregnancy-related diseases because the dynamic change in the immune state during pregnancy and the possibility of fetal drug toxicity must be taken into account. Most of the current treatments for pregnancy-related diseases focus on a reduction in an overactive immune response with the use of non-specific immunosuppressive therapy. This triggers the simultaneous activation of numerous immune cells and makes it difficult to control the dose and to evaluate the curative effect because of individual heterogeneity. Therefore, more studies should be conducted to further explore the effectiveness and safety of Tregs-based target therapies for the treatment of pregnancy-related diseases.
Taken together, the above studies suggest that Tregs play a prominent role in regulating fetal-maternal immune tolerance, and a defect in the proportion and activity of Tregs is involved in the development of RSA, endometriosis, and preeclampsia. Thus, approaches designed to boost the proportion of Tregs or strengthen their suppressive function may lead to promising strategies for treating pregnancy-related diseases. Several Tregs-based target therapies are entering into clinical trials, including adoptive Treg cell therapy, Tregs-enhancing drugs, and low dose IL-2 administration . Administration of purified Tregs was firstly applied as Tregs-based target therapy. With the development of immune cell therapy, antigen-specific Tregs therapy was also proposed for treating autoimmune and graft-versus-host diseases. Phase I/II clinical trials aimed to explore the curative effect, and some have reported that Tregs administration alleviates clinical symptoms induced by autoimmunity . Some research has attempted to determine whether Tregs administration improves pregnancy outcomes. Yin et al. and Wang et al. examined the effectiveness of adoptive transfer of Tregs in preventing spontaneous abortion in mice models . Yin et al. established an abortion-prone pregnancy mice model with DBA/2J-mated pregnant CBA/J mice and performed adoptive transfer of freshly isolated and in vitro expanded Tregs from non-pregnant CBA/J mice. Wang et al. induced spontaneous abortions by administration of IL-17 in a CBA/J × BALB/c mouse model of normal pregnancy and performed adoptive transfer of in vitro expanded Tregs purified from pregnant CBA/J mice. These two studies demonstrated transfusion with in vitro expanded Tregs promotes immune suppressive activity, increases the secretion of suppressive cytokines and significantly reduces the rate of spontaneous abortion. Although Treg cell therapy has not been widely used in clinical practice, clinical research has initiated several non-specific immunotherapies partially regulating the proportion and activity of Tregs for the treatment of pregnancy-related diseases. Intravenous immunoglobulin G (IVIG) and paternal or third-party lymphocyte immunization therapy have been proposed for the treatment of patients with RSA due to the potential immunomodulatory effects. Although the benefit for these immunotherapies is controversial, a growing body of evidence suggests that they may increase rates of live birth and decrease rates of miscarriage . A variety of studies and clinical trials have reported both IVIG and lymphocyte immunization therapy correct the Tregs defect and rebalances the Th17/Treg paradigm in peripheral blood. Compared with a control group, the treatment triggers a shift toward Tregs in the Th17/Treg balance by enhancing the expansion of Tregs, promoting the secretion of suppressive cytokines, and inhibiting Th17 cells proliferation . Tregs-enhancing drugs are another type of Tregs-based target therapy. Rapamycin (Sirolimus) is an mTOR inhibitor, which acts as an immunosuppressive drug by selectively promoting the expansion of Tregs and inducing differentiation of T helper cells into Tregs. Royster et al. established a murine model with conditional knockdown of Tregs induced by diphtheria toxin. They found the deletion of Tregs decreased litter sizes and triggered embryo implantation failure, effects that were reversed after the treatment with rapamycin . A multicenter, double-blind, phase II randomized clinical trial administrated 2 mg/day of sirolimus for 2 days before embryo transfer to patients receiving IVF-ET therapy and who had a history of recurrent implantation failure. The study collected blood samples and assessed the ratio of Th17/Treg cells by flow cytometry 5–10 days prior to the initiation of an IVF cycle. Only patients with a high ratio of Th17/Treg cells were included in this trial. The trial reported that the administration of sirolimus reversed the imbalance in the ratio of Th17/Treg cells and significantly increased the rate of clinical pregnancy and live birth compared with those in the control group . Taken together, some studies have demonstrated the effectiveness of Tregs-based therapy in treating several autoimmune diseases and cases of organ transplantation. However, the methods cannot be directly applied for pregnancy-related diseases because the dynamic change in the immune state during pregnancy and the possibility of fetal drug toxicity must be taken into account. Most of the current treatments for pregnancy-related diseases focus on a reduction in an overactive immune response with the use of non-specific immunosuppressive therapy. This triggers the simultaneous activation of numerous immune cells and makes it difficult to control the dose and to evaluate the curative effect because of individual heterogeneity. Therefore, more studies should be conducted to further explore the effectiveness and safety of Tregs-based target therapies for the treatment of pregnancy-related diseases.
Tregs are generally viewed as arising from a specific T cell lineage generated in the thymus or induced in peripheral organs. Being the most predominant immune-suppressive cells, a tremendous amount of research has focused on determining the molecular mechanisms responsible for inducing the expansion of Tregs and their activity in the periphery and in specific organs. This effort will provide new insights that will guide the improvement of Tregs-based targeted immune therapy. In recent years, increasing data has shown that the expansion of Tregs is triggered after exposure to the fetal alloantigens and changes dynamically over the course of pregnancy. Hormones such as estradiol and progesterone as well as HCG are significantly increased during pregnancy, and regulate the number and function of Tregs to sustain a proper pregnancy-related immune tolerance. Furthermore, various reproductive diseases such as recurrent miscarriage, endometriosis and preeclampsia result in part from the deficiency in the number and activity of Tregs. Therefore, modulating the immune response by boosting the number of Tregs and enhancing their activity may be a potential therapeutic strategy for managing these pregnancy-related complications.
All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.
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.
|
Cardiovascular disease burden due to productivity losses in European Society of Cardiology countries | d616ca57-eb4e-4051-98c2-bbd49ea291ae | 10785594 | Internal Medicine[mh] | What is already known on this topic Cardiovascular disease is the leading cause of death in the world, with 17.9 million deaths. Lost earnings due to premature death were €32 billion in 2015 across the European Union. What this study adds We estimate the lost earnings due to CVD-related premature death across 54 countries for 2018 at €62 billion. The total number of deaths due to CVD decreased from 4.8 million in 2000 to 4.6 million in 2010 and 4.4 million in 2018. We found considerable variation across countries concerning the economic impact of CVD related deaths. This highlights the potential gains from policies targeting prevention and care of cardiovascular diseases.
Cardiovascular disease (CVD, ICD-10 I00–I99) is the leading cause of death in the world, with 17.9 million deaths in 2019. CVD is the main source of years of life lost worldwide, accounting for 354.8 million years of life lost in 2019. The most common CVDs are coronary heart disease (CHD, ICD-10 I20–I25) and cerebrovascular disease (ICD-10 I60–I69), which were responsible for ∼50 and 35% of all CVD deaths, respectively. As well as the human costs of such deaths, early death due to CVD will negatively impact the workforce, and so the wider economy and society. Estimating the lost earnings (productivity losses) due to premature mortality provides an important societal perspective on the economic impact of CVD. Productivity losses consist of earnings associated with paid work that are lost if someone dies before their retirement age. Expected future earnings are assumed to reflect the individual's potential contribution to the economy. Estimating productivity costs is essential for policymakers who need to be aware of the societal impact of their decisions, since costs falling outside the healthcare budget can be considered equally important as those falling on the healthcare budget. In our previous work, we estimated productivity losses due to CVD-related premature mortality for the European Union (EU) to be €24 billion in 2003 and €32 billion in 2015, , after adjustments for employment rates and discounting. However, this work was restricted to the relatively affluent countries of the EU. These countries have advanced healthcare systems and often relatively low incidence of CVD, when compared with the rest of the world. Our study aimed to update and enhance our estimates of the economic impact of CVD premature mortality to society. First, we estimated the number of working years lost and productivity losses in 2018 for countries with a European Society of Cardiology (ESC) affiliated national cardiac society. Second, we broke down productivity losses in EU and non-EU countries and examined variations across countries. Finally, we examined trends in CVD deaths between 2000 and 2018.
Analysis framework and data sources We conducted a population-based cost analysis to evaluate productivity losses from premature death due to CVD. Although the ESC includes 57 National Cardiac Societies, we excluded countries with a population of <50 000 (San Marino), countries that are not members of the United Nations (Kosovo), and those with considerable political instability and conflict (Libya and Syria) in order to ensure reliability and validity of data. We adopted the same methodological framework for each of the 54 countries under study (28 EU member states as of 2018, and 26 non-EU member countries; see for the full list). This is the same approach used in other cost-of-illness studies evaluating CVD, , cancer, blood disorders, and dementia. We used the human capital approach to measure and value lost productivity due to premature mortality. This approach assumes that economic capital can be measured by an individual's ability to participate in paid employment. Within this framework, losses can be quantified by measuring the number of years by which working life is reduced due to mortality. Working years are valued using achievable gross earnings as a representation of the value of work to society. These costs are then adjusted for the proportion of the population who are employed. Country-specific productivity losses from premature death were estimated by calculating the sum of the age- and sex-specific products of the following (see for full references): number of CVD-related deaths; number of remaining work years at the time of death; annual gross earnings; economic activity and unemployment rates in the population (aged 15–79). Working years left at the time of death were estimated as the number of years up to age 79. The most frequent retirement age across EU and OECD countries is 65 years ( https://www.etk.fi/en/work-and-pensions-abroad/international-comparisons/retirement-ages/ & https://www.oecd-ilibrary.org/sites/99acb105-en/index.html?itemId=/content/component/99acb105-en ), but there are a significant number of individuals working up to age 79. The number of working years left at the time of death was adjusted by the likelihood of individuals in each age group (5-year age bands from 15 to 79 years old) being economically active and employed. In , we report mean annual gross earnings by sex. (Note that we used age- and sex-specific earnings in our analysis.) Future earnings lost were discounted to present values using a 3.5% annual rate. All costs were expressed in 2018 prices and converted to euros where applicable. , Finally, we examined the trends in all-cause and CVD-related mortality in the 54 ESC countries between 2000 and 2018 across all ages and for 15–64-year-olds. We estimated the proportion of CVD deaths of all deaths across all ages and for the 15–64-year-olds. lists the data sources of mortality data by country and year. Patient and public involvement Patients were not involved in our study as it comprised the analysis of secondary data. Statistical analysis Costs in 2018 were estimated individually by country, by country groupings (e.g. EU member or not, by World Bank income group—low income, lower middle income, upper middle income and high income), and in total across all ESC countries. We also estimated costs per capita, dividing total costs by the population of each country. As a sensitivity analysis and to facilitate comparisons across countries, we employed the purchasing power parity (PPP) method to adjust annual gross earnings. The PPP method allowed us to account for price differentials across countries. This was done by first converting costs in local currency into 2018 international dollars (Int$, an international dollar buys in each country a comparable amount of goods and services that US dollar buys in the USA) and then to PPP-adjusted euros (applying the EU-28 to US PPP index in 2018, i.e. 1.44 749). To explore variations in CVD-related productivity losses between countries in 2018, we undertook a series of ordinary least-squares (OLS) regression analyses (with robust SE) using CVD-related productivity losses per capita adjusted for price differentials (PPP adjusted). We explored the association between productivity losses and national income [gross domestic product (GDP)—per capita], proportion of GDP expenditure on healthcare, proportion of healthcare expenditure that was out-of-pocket expenditure, hospital beds per 1000 persons, and CVD incidence (crude rate per capita) in univariable and multivariable analyses. Diagnostic tests were performed for functional form (RESET test and link test). An explanatory variable was considered significant if its P value was less than 0.05. All regression analyses were performed using Stata version 15.0.
We conducted a population-based cost analysis to evaluate productivity losses from premature death due to CVD. Although the ESC includes 57 National Cardiac Societies, we excluded countries with a population of <50 000 (San Marino), countries that are not members of the United Nations (Kosovo), and those with considerable political instability and conflict (Libya and Syria) in order to ensure reliability and validity of data. We adopted the same methodological framework for each of the 54 countries under study (28 EU member states as of 2018, and 26 non-EU member countries; see for the full list). This is the same approach used in other cost-of-illness studies evaluating CVD, , cancer, blood disorders, and dementia. We used the human capital approach to measure and value lost productivity due to premature mortality. This approach assumes that economic capital can be measured by an individual's ability to participate in paid employment. Within this framework, losses can be quantified by measuring the number of years by which working life is reduced due to mortality. Working years are valued using achievable gross earnings as a representation of the value of work to society. These costs are then adjusted for the proportion of the population who are employed. Country-specific productivity losses from premature death were estimated by calculating the sum of the age- and sex-specific products of the following (see for full references): number of CVD-related deaths; number of remaining work years at the time of death; annual gross earnings; economic activity and unemployment rates in the population (aged 15–79). Working years left at the time of death were estimated as the number of years up to age 79. The most frequent retirement age across EU and OECD countries is 65 years ( https://www.etk.fi/en/work-and-pensions-abroad/international-comparisons/retirement-ages/ & https://www.oecd-ilibrary.org/sites/99acb105-en/index.html?itemId=/content/component/99acb105-en ), but there are a significant number of individuals working up to age 79. The number of working years left at the time of death was adjusted by the likelihood of individuals in each age group (5-year age bands from 15 to 79 years old) being economically active and employed. In , we report mean annual gross earnings by sex. (Note that we used age- and sex-specific earnings in our analysis.) Future earnings lost were discounted to present values using a 3.5% annual rate. All costs were expressed in 2018 prices and converted to euros where applicable. , Finally, we examined the trends in all-cause and CVD-related mortality in the 54 ESC countries between 2000 and 2018 across all ages and for 15–64-year-olds. We estimated the proportion of CVD deaths of all deaths across all ages and for the 15–64-year-olds. lists the data sources of mortality data by country and year.
Patients were not involved in our study as it comprised the analysis of secondary data.
Costs in 2018 were estimated individually by country, by country groupings (e.g. EU member or not, by World Bank income group—low income, lower middle income, upper middle income and high income), and in total across all ESC countries. We also estimated costs per capita, dividing total costs by the population of each country. As a sensitivity analysis and to facilitate comparisons across countries, we employed the purchasing power parity (PPP) method to adjust annual gross earnings. The PPP method allowed us to account for price differentials across countries. This was done by first converting costs in local currency into 2018 international dollars (Int$, an international dollar buys in each country a comparable amount of goods and services that US dollar buys in the USA) and then to PPP-adjusted euros (applying the EU-28 to US PPP index in 2018, i.e. 1.44 749). To explore variations in CVD-related productivity losses between countries in 2018, we undertook a series of ordinary least-squares (OLS) regression analyses (with robust SE) using CVD-related productivity losses per capita adjusted for price differentials (PPP adjusted). We explored the association between productivity losses and national income [gross domestic product (GDP)—per capita], proportion of GDP expenditure on healthcare, proportion of healthcare expenditure that was out-of-pocket expenditure, hospital beds per 1000 persons, and CVD incidence (crude rate per capita) in univariable and multivariable analyses. Diagnostic tests were performed for functional form (RESET test and link test). An explanatory variable was considered significant if its P value was less than 0.05. All regression analyses were performed using Stata version 15.0.
Productivity losses due to CVD in 54 ESC member countries As a whole, for the 54 ESC member countries under analysis, CVD accounted for 4.4 million deaths, 7.1 million working years lost, and €61.6 billion in productivity losses in 2018 (see and ). The average lost earnings per capita across the 54 countries were €57 (median €56, interquartile range 38–85), varying from €6 in Algeria to €133 in Switzerland. Approximately 60% (€37 billion) of all productivity losses occurred in EU-28 countries, but these accounted for only 42% (1.8 million) of deaths and 21% (1.5 million) of working years lost across all 54 countries. Productivity losses due to CVD in EU-28 in 2018 CVD accounted for 1.8 million deaths in the EU in 2018, representing 1.5 million potential years of work lost, which were estimated at €37.1 billion (see ). In terms of deaths per 1000 citizens, the average was 3.6 and varied from 1.8 in Belgium and Ireland to 10.2 in Bulgaria. In terms of working years lost per 1000 citizens, the average was 2.9 and varied from 0.8 in Belgium to 11.0 in Latvia. Averaged across EU-28 population, this corresponded to €72 in lost earnings per capita, varying from €34 in Belgium to €115 in Latvia. Adjusting for price differentials, the range of costs increased across the EU-28 from €32 in Belgium to €167 in Latvia (see and ). The main cause of CVD death was CHD, accounting for 34% (0.6 million deaths) of all deaths, followed by cerebrovascular diseases with 22% (0.4 million deaths). In terms of productivity losses, CHD and cerebrovascular disease accounted for 43% (€16.1 billion, €31 per capita) and 17% (€6.3 billion, €12 per capita), respectively, of all CVD-related losses. However, there was considerable variation across these countries in the proportion of losses due to CHD and cerebrovascular diseases. The proportion of losses attributable to CHD varied from 23% in Bulgaria to 63% in Cyprus. For cerebrovascular diseases, the proportion of total losses varied from 11% in Luxembourg to 27% in Portugal. Productivity losses due to CVD in ESC member countries not part of the EU CVD accounted for 2.6 million deaths in the remaining 26 ESC countries, representing 5.6 million potential years of work lost (see ). CVD cost €24.5 billion in lost earnings, after adjusting for employment rates and discounted to present values. In terms of deaths per 1000 citizens, the average was 4.5 and varied from 1.3 in Israel to 10.1 in Ukraine. In terms of working years lost per 1000 citizens, the average was 9.8 and varied from 1.7 in Norway to 30.3 in Georgia. This corresponded to €43 in lost earnings per capita in these countries, varying from €6 in Algeria to €133 in Switzerland. Adjusting for price differentials, the range of productivity losses increased across countries from €14 per capita in Algeria to €247 per capita in Georgia (see ). Using income categories from the World Bank, there was still considerable variation across countries in the same income group (see ). The main cause of CVD death was CHD, accounting for 60% of all deaths, followed by cerebrovascular diseases with 26%. In terms of productivity losses, CHD and cerebrovascular disease accounted for 52% (€12.7 billion, €22 per capita) and 20% (€4.9 billion, €9 per capita), respectively, of all CVD-related losses. However, there was considerable variation across these countries in the proportion of losses due to CHD and cerebrovascular diseases. The proportion of losses due to CVD attributable to CHD varied from 28% in Serbia to 83% in Lebanon. For cerebrovascular diseases, the proportion of total losses varied from 7% in Lebanon to 34% in North Macedonia. and report the ranking of countries by income category in terms of productivity losses per capita (PPP-adjusted) due to CHD and stroke. Amongst high- income countries, Lithuania (€71 due to CHD and €23 due to stroke) and Latvia (€67 due to CHD and €31 due to stroke) had the highest losses per capita, whereas Liechtenstein (€5 due to CHD and €5 due to stroke) and Belgium (€13 due to CHD and €4 due to stroke) reported the lowest losses for both conditions. Amongst upper-middle-income countries, Kazakhstan reported the highest losses per capita (€93 due to CHD and €50 due to stroke), whereas Algeria reported the lowest losses per capita (€9 due to CHD and €3 due to stroke). Finally, amongst lower-middle-income countries, Georgia reported the highest losses per capita (€130 due to CHD and €79 due to stroke), whereas Tunisia reported the lowest losses per capita (€13 due to CHD and €3 due to stroke). Associations with CVD-related productivity losses per capita in ESC countries reports the OLS results examining associations with productivity losses (PPP adjusted) across the 54 countries. In terms of single analysis, we found a positive significant association between CVD-related productivity losses and incidence of CVD amongst 15–64-year-olds ( P = 0.008), with each additional case per 1000 persons increasing CVD-related costs by €9.6 (see ). We also found significant associations between CVD productivity losses and healthcare expenditure as a proportion of GDP ( P = 0.031), with each additional 1% of healthcare expenditure decreasing CVD-related productivity losses by €0.5 (see ). However, we found no association between CVD-related productivity losses and CVD incidence and healthcare expenditure in the multivariable analysis. CVD deaths between 2000 and 2018 The total number of deaths due to CVD decreased from 4.8 million in 2000 to 4.6 million in 2010 and 4.4 million in 2018 (see and and ). Over the same time period, there was a significant increase in the number of deaths due to all causes, from 10.0 million in 2000 to 10.3 million in 2018. Hence, the proportion of all deaths due to CVD fell from 48% in 2000 to 46% in 2010 and 43% in 2018. About 20% of all CVD deaths in 2000 occurred between 15 and 64 years of age, falling to 18% between 2015 and 2018. Among 15–64-year-olds, CVD accounted for 33% of all deaths in 2000, 33% in 2010 and 32% in 2018. However, there was considerable variation across countries over time (see and and ). For example, five countries (Azerbaijan, Bulgaria, Egypt, Morocco, and Uzbekistan) reported the proportion of all deaths due to CVD above 40% in 2018, whereas France and Israel reported the lowest proportion at 13% and 12%, respectively.
As a whole, for the 54 ESC member countries under analysis, CVD accounted for 4.4 million deaths, 7.1 million working years lost, and €61.6 billion in productivity losses in 2018 (see and ). The average lost earnings per capita across the 54 countries were €57 (median €56, interquartile range 38–85), varying from €6 in Algeria to €133 in Switzerland. Approximately 60% (€37 billion) of all productivity losses occurred in EU-28 countries, but these accounted for only 42% (1.8 million) of deaths and 21% (1.5 million) of working years lost across all 54 countries.
CVD accounted for 1.8 million deaths in the EU in 2018, representing 1.5 million potential years of work lost, which were estimated at €37.1 billion (see ). In terms of deaths per 1000 citizens, the average was 3.6 and varied from 1.8 in Belgium and Ireland to 10.2 in Bulgaria. In terms of working years lost per 1000 citizens, the average was 2.9 and varied from 0.8 in Belgium to 11.0 in Latvia. Averaged across EU-28 population, this corresponded to €72 in lost earnings per capita, varying from €34 in Belgium to €115 in Latvia. Adjusting for price differentials, the range of costs increased across the EU-28 from €32 in Belgium to €167 in Latvia (see and ). The main cause of CVD death was CHD, accounting for 34% (0.6 million deaths) of all deaths, followed by cerebrovascular diseases with 22% (0.4 million deaths). In terms of productivity losses, CHD and cerebrovascular disease accounted for 43% (€16.1 billion, €31 per capita) and 17% (€6.3 billion, €12 per capita), respectively, of all CVD-related losses. However, there was considerable variation across these countries in the proportion of losses due to CHD and cerebrovascular diseases. The proportion of losses attributable to CHD varied from 23% in Bulgaria to 63% in Cyprus. For cerebrovascular diseases, the proportion of total losses varied from 11% in Luxembourg to 27% in Portugal.
CVD accounted for 2.6 million deaths in the remaining 26 ESC countries, representing 5.6 million potential years of work lost (see ). CVD cost €24.5 billion in lost earnings, after adjusting for employment rates and discounted to present values. In terms of deaths per 1000 citizens, the average was 4.5 and varied from 1.3 in Israel to 10.1 in Ukraine. In terms of working years lost per 1000 citizens, the average was 9.8 and varied from 1.7 in Norway to 30.3 in Georgia. This corresponded to €43 in lost earnings per capita in these countries, varying from €6 in Algeria to €133 in Switzerland. Adjusting for price differentials, the range of productivity losses increased across countries from €14 per capita in Algeria to €247 per capita in Georgia (see ). Using income categories from the World Bank, there was still considerable variation across countries in the same income group (see ). The main cause of CVD death was CHD, accounting for 60% of all deaths, followed by cerebrovascular diseases with 26%. In terms of productivity losses, CHD and cerebrovascular disease accounted for 52% (€12.7 billion, €22 per capita) and 20% (€4.9 billion, €9 per capita), respectively, of all CVD-related losses. However, there was considerable variation across these countries in the proportion of losses due to CHD and cerebrovascular diseases. The proportion of losses due to CVD attributable to CHD varied from 28% in Serbia to 83% in Lebanon. For cerebrovascular diseases, the proportion of total losses varied from 7% in Lebanon to 34% in North Macedonia. and report the ranking of countries by income category in terms of productivity losses per capita (PPP-adjusted) due to CHD and stroke. Amongst high- income countries, Lithuania (€71 due to CHD and €23 due to stroke) and Latvia (€67 due to CHD and €31 due to stroke) had the highest losses per capita, whereas Liechtenstein (€5 due to CHD and €5 due to stroke) and Belgium (€13 due to CHD and €4 due to stroke) reported the lowest losses for both conditions. Amongst upper-middle-income countries, Kazakhstan reported the highest losses per capita (€93 due to CHD and €50 due to stroke), whereas Algeria reported the lowest losses per capita (€9 due to CHD and €3 due to stroke). Finally, amongst lower-middle-income countries, Georgia reported the highest losses per capita (€130 due to CHD and €79 due to stroke), whereas Tunisia reported the lowest losses per capita (€13 due to CHD and €3 due to stroke).
reports the OLS results examining associations with productivity losses (PPP adjusted) across the 54 countries. In terms of single analysis, we found a positive significant association between CVD-related productivity losses and incidence of CVD amongst 15–64-year-olds ( P = 0.008), with each additional case per 1000 persons increasing CVD-related costs by €9.6 (see ). We also found significant associations between CVD productivity losses and healthcare expenditure as a proportion of GDP ( P = 0.031), with each additional 1% of healthcare expenditure decreasing CVD-related productivity losses by €0.5 (see ). However, we found no association between CVD-related productivity losses and CVD incidence and healthcare expenditure in the multivariable analysis.
The total number of deaths due to CVD decreased from 4.8 million in 2000 to 4.6 million in 2010 and 4.4 million in 2018 (see and and ). Over the same time period, there was a significant increase in the number of deaths due to all causes, from 10.0 million in 2000 to 10.3 million in 2018. Hence, the proportion of all deaths due to CVD fell from 48% in 2000 to 46% in 2010 and 43% in 2018. About 20% of all CVD deaths in 2000 occurred between 15 and 64 years of age, falling to 18% between 2015 and 2018. Among 15–64-year-olds, CVD accounted for 33% of all deaths in 2000, 33% in 2010 and 32% in 2018. However, there was considerable variation across countries over time (see and and ). For example, five countries (Azerbaijan, Bulgaria, Egypt, Morocco, and Uzbekistan) reported the proportion of all deaths due to CVD above 40% in 2018, whereas France and Israel reported the lowest proportion at 13% and 12%, respectively.
Across 54 ESC member countries, we estimated the total lost earnings due to CVD-related premature mortality to be €62 billion a year in 2018, of which 60% were incurred in 28 EU member countries. The higher the proportion of healthcare expenditure, the lower the productivity losses per capita, holding all else constant. The higher the incidence of CVD, the higher the productivity losses per capita, holding all else constant. However, these associations do not imply causality and lost significance in multivariable analysis. For the first time, this study assesses the mortality losses associated with CVD, CHD, and cerebrovascular disease across the whole of Europe, including countries in the European periphery, including Northern Africa, the Middle East, and former Soviet Republics in Asia. Furthermore, we were able to compare our results with previous work in which we estimated the productivity losses due to CVD mortality for the EU in earlier years. In 2003, we estimated these losses to be €24 billion for the 25 countries forming the EU at the time. These costs rose to €27 billion in 2009 (27 EU countries) and to €32 billion in 2015 (28 EU countries). In our current analysis, we estimated total productivity losses due to CVD mortality to be €37 billion for 2018. The increase in costs is explained by changes in employment rates and increase in average earnings from 2009 to 2018. For example, in 2018 prices, the productivity losses in 2015 were estimated at €34 billion compared with €37 billion in 2018 despite a decrease of about 75 000 CVD-related deaths. However, the number of potential years of work lost was similar in both years (1.5 million) due to an increase in employment rates since 2015. By using the same methodology to estimate the cost of CVD over time, it is possible to reliably compare the impact of CVD mortality on lost earnings over time. This comparative evidence is useful to decision-makers and health policy planners by informing evaluations of the impact of public health interventions, for example, in evaluating the societal benefits of addressing CVD risk factors. The accuracy of our study depends on the quality and availability of comparable CVD-related data across countries. Our study included 54 countries, for each of which we required the identification of comparable data on mortality, earnings, and employment rates by sex and age group. This required consulting and using a wide range of sources to conduct comprehensive analyses. Importantly, data on deaths due to CVD, CHD, and cerebrovascular disease were derived from two main sources: EUROSTAT for EU member states or those with close links to the EU; and the Global Burden of Disease for the remainder. As a result, some of the differences identified between mainly relatively wealthy Western European countries and non-EU countries in Eastern Europe, the Middle East, and North Africa could be due to differing methodologies on how disease-specific deaths were calculated. Given the large numbers of countries included in this study, and the fact that for EU countries comparisons can be made to over 15 years of previous data, these results are important to policy-makers. Data on costs of CVD premature mortality can aid decisions about the allocation of national resources, including service provision, prevention strategies, and future research funding. They allow for benchmarks to be set up across the over 50 countries under study in order to identify the most efficient public policy initiatives and healthcare systems capable of achieving the best CVD outcomes. We used the human capital approach to estimate productivity losses, as it is fairly transparent and can be quantified using available data, and is consistent with and permits comparisons with previous work on CVD and other conditions such as cancer, blood disorders, and dementia. However, it does have limitations, for example, by failing to value many socially valuable activities that are not remunerated as formal employment, such as much housework, caring, and volunteer activities. Alternative methods of valuing lives and life years, such as willingness to pay, have been proposed and discussed, but there is little consensus at present on which approach is best or on how these might be readily operationalized using existing data. , Furthermore, our estimates comprise productivity losses due to CVD mortality alone. This does not comprise CVD-related morbidity losses, which are likely to be significant in terms of individuals being declared incapacitated or disabled because of CVD or taking sickness leave for a defined time period. Our study provides a snapshot of the economic consequences outside the healthcare system posed by CVD in terms of lost earnings to 54 ESC member countries in 2018. Together with the evidence we have gathered over the last 15 years, it indicates that these costs are increasing over time, despite the decrease in CVD mortality. Comparative results across time and countries are important for policymakers to track the effect of public health interventions and treatments with the objective of contributing to maximizing social welfare in their countries.
qcad031_Supplemental_File Click here for additional data file.
|
A metabolic modeling-based framework for predicting trophic dependencies in native rhizobiomes of crop plants | 3c58b4a3-e894-439f-a347-00e050791cf8 | 11486489 | Microbiology[mh] | The rhizosphere serves as a hotspot for a diversity of interactions spanning from the secretion of organic compounds by plant roots to their uptake by the adjacent soil microbial community . These interactions form a complex network of metabolic exchanges whose structure and function has a considerable impact on plant health . Targeted secretion of exudates from plant roots’ into the environment is fundamental to the recruitment of specific microbes and the assembly of a plant-selected community (i.e., the rhizobiome) . Each plant has a unique profile of exudates guiding the formation of a specialized rhizobiome that is adapted to support its mineral absorption , secrete plant growth supporting compounds , and provide protection against soil-borne microorganisms that are detrimental to its health . A comprehensive understanding of the dynamics within the rhizosphere, considering both plant–microbe (PM) and microbe–microbe interactions, can guide the targeted assembly and maintenance of plant-beneficial soil microbial systems . The development of such microbiome-based, plant-beneficial strategies presents ecologically sound alternatives to conventional, chemical-based solutions in supporting plant health and productivity . `Omics data in general and specifically metagenomics data analyses can potentially provide keys for unraveling the black box of PM and microbe–microbe interactions in complex ecosystems such as the soil . Sequence-based analyses are, however, typically limited in terms of functional interpretation of community dynamics . Constraint-based modeling (CBM) is an approach that allows for the simulation of bacterial-metabolic activity in a given environment based on the constraints imposed by the annotated microbial genomes . This approach has long been used for studying the physiology and growth of single cells, represented as genome scale metabolic models (GSMMs), under varying conditions . Applying CBM over a GSMM can be used to assess the uptake and secretion of metabolites in the environment under study . Accordingly, when CBM is applied over multiple GSMMs, metabolite exchange profiles (secretion into and consumption from the environment) become interconnected. This sheds light on conditions supporting the growth of different bacterial groups within the community as well as their functional potential in the trophic network formed . Advancement of sequencing technologies alongside the development of automatic pipelines for GSMM construction has promoted an increased use of CBM for the modeling of communities with growing complexity . The relevance of CBM to the study of microbiomes has been demonstrated in a variety of ecosystems and recent works have shown that CBM-based predictions can guide the development of strategies for microbiome management . An accurate representation of microbial metabolic networks depends on the origin of the genomes analyzed. To date, most studies attempting to model the microbiome of specific ecosystems by GSMM represent native species using corresponding genome sequences from public depositories, a process which is usually referred to as 16S-based genome imputation . Genome recovery, or genome-resolved metagenomes, often referred to as metagenome assembled genomes (MAGs) allows one to obtain full genomes directly from metagenomes . Constructing GSMMs based on MAGs derived from a specific biological sample or directly from a native community enables a genuine view of the metabolic activities carried out in situ , hence bypasses the need in 16S-based genome imputation. Such an in silico representation of a native community (with respect to its environment) can be used to decrypt the myriad interconnected uptake and secretion exchange fluxes transpiring within the root-associated microbiome, spanning from root exudates to altered organic forms. The current study describes the recovery of 395 unique MAGs from metagenomes constructed for the native rhizosphere community of apple rootstocks cultivated in orchard soil affected by apple replant disease (ARD) . Soils were amended with Brassicaceae seed meal, or were not amended, supporting the development of either disease-suppressive or disease-conducive root microbiomes, respectively . MAGs were recovered from metagenomics data collected from these apple rhizosphere microbiomes. GSMMs constructed for the MAGs provide an in silico representation of highly abundant species in the native rhizosphere community. CBM simulations were then conducted in a rhizosphere-like environment, where microbial uptake-secretion fluxes were connected to form a directional trophic network. The aims of this study were twofold: first, to provide a general framework for delineating inter-species interactions occurring in the rhizosphere environment and second, to characterize the metabolic roles specific groups of bacteria fulfill in seed meal-amended (disease-suppressive) vs non-amended (disease-conducive) apple rhizobiome communities.
Assembly of a collection of MAGs representing a native microbial community from a soil agroecosystem Metagenomic sequencing obtained from the rhizosphere of apple rootstocks grown in orchard soil with a documented history of replant disease resulted in a total of approximately 2 billion quality reads (after filtration) at a length of 150 bp, as described in . Independent metagenomics assemblies of six different treatments yielded 1.4–2 million contigs longer than 2 kbp . Here, assemblies were binned using MetaWRAP into 296–433 high-quality MAGs for each of the six treatments ; completion and contamination thresholds were set to 90/5, respectively . De-replication (the process of combining highly similar genomes into a single representative genome) requiring 99% average nucleotide identity of the overall 2233 MAGs yielded a collection of 395 unique MAGs . Across samples, 30–36% of the raw reads were mapped to the MAG collection , in comparison to 61–71% mapped to the non-binned contigs . Using GTDB-Tk , taxonomic annotations were assigned at the phylum, order, genus, and species level for 395, 394, 237, and three of the de-replicated MAGs, respectively, reflecting the genuine diversity of the rhizosphere community, which include many uncharacterized species . Estimates of completion and contamination, total bin length and taxonomic affiliation for the MAG collections derived from each of the six assemblies, as well as the de-replicated MAGs, are provided in . As in previous reports , Proteobacteria, Acidobacteria, Actinobacteria, and Bacteroidetes were identified as the dominant phyla in the apple rhizosphere . Overall, the taxonomic distribution of the MAG collection corresponded with the profile reported for the same samples using alternative taxonomic classification approaches such as 16S rRNA amplicon sequencing and gene-based taxonomic annotations of the non-binned shotgun contigs . At the genus level, MAGs were classified into 143 genera in comparison to approximately 3000 genera that were identified for the same data based on gene-centric approaches and approximately 1000 genera based on amplicon sequencing of the same data. The functional capabilities of the bacterial genomes in the apple rhizosphere were initially assessed based on KEGG functional annotations of their gene catalogue . The 10 most frequent functional categories across the MAG collection were involved in primary metabolism, for example, carbohydrate metabolism, and the biosynthesis of essential cellular building blocks such as amino acids and vitamins. Specialized functional categories included those associated with autotrophic nutrition such as carbon fixation and metabolism of nitrogen, sulfur, and methane. Functional diversity was found to exist also when considering ubiquitous functions. For instance, though all bacteria are in need of the full set of amino acids, most genomes lack the full set of relevant biosynthesis pathways. The prevalence of biosynthetic pathways across MAGs (requiring at least a single relevant enzyme) ranges from 99.5% (e.g., glycine; missing in only two genomes) to 21% (e.g., tyrosine; detected in only 83 MAGs). Notably, the diversity of metabolic pathway completeness regarding amino acids and other essential cellular components suggests that the majority of bacterial soil species rely on an external supply of at least some of their obligatory nutritional demands. Construction of a simulation system for exploring environment-dependent metabolic performances and growth of rhizosphere bacteria Categorical classifications of discrete gene entities, such as pathway completeness analyses, have several inherent limitations as an approach for the contextualization and functional interpretation of genomic information. First, categorical classifications may underestimate the completeness of robust pathways with multiple redundant routes resulting in low pathway completeness. Second, pathway completeness analyses do not take into consideration the directionality and continuity of a biosynthetic process whose full conductance requires the availability of a specific environmental resource together with the required successive series of genes/reactions. Although functional potential can be inferred based on the static set of genes and enzymes present, actual metabolic performances of bacterial species in soil are dynamic and reflect multiple factors, including the availability of different nutritional sources. Such sources can be environmental inputs like root exudates or downstream exchange metabolites secreted by cohabiting bacteria. To better understand environment-dependent metabolic activity occurring in a native rhizosphere, a set of 395 GSMMs was constructed for the entire collection of the rhizosphere-bacterial community MAGs. All models were systematically subjected to validity and quality tests using MEMOTE , leaving a total of 243 GSMMs whose stoichiometric consistency was confirmed . On average, GSMMs included 1924 reactions, from which 203 were exchange reactions (specific reactions carrying out extracellular import and secretion of metabolites) and 1312 metabolites . Altogether, the GSMM set held 5152 unique metabolic reactions, 597 exchange reactions, and 2671 different compounds. The distribution of key model attributes (reactions, metabolites, and exchanges) across phyla is shown in . Model features were scalable with those reported by and . Additionally, comparison of GSMM scales indicated that the metabolic coverage (i.e., the number of reactions, which denote the potential of executing a metabolic function) of our data is within the same order of magnitude as described in recent large-scale automatic reconstructions . Next, species-specific rich (optimal) and poor (suboptimal) media were defined for each model in order to broadly assess GSMM growth capacity. A rich environment was defined as a medium containing all metabolites for which the model encodes an exchange reaction. A poor environment was defined as the minimal set of compounds enabling growth (see Methods). Essentially, poor environments contained species-specific carbon, nitrogen and phosphorous sources together with other trace elements. In addition to the two automatic media, we aimed to design a realistic simulating environment to explore the impact of the root exudates on the native community. To this end, metabolomics data from a set of studies which characterized the root exudates of Geneva 935 (G935) or Malling 26 (M26) rootstock cultivars – related to the G210 and M26 rootstocks whose rhizobiome was characterized here, were used to specify an array of apple root-derived compounds . The list of secreted compounds was consistent with other reported profiles of plant root exudates . The growth of each of the 243 GSMMs was simulated in each of the three species-specific environments (rich, poor, poor + exudates; exudates were added to corresponding poor media to ensure the exudates have a feasible effect). As expected, growth performances were higher on the poor medium supplemented with exudates in comparison to growth on the poor medium alone but lower than the growth in rich medium . Notably, GSMM growth patterns were not phylogenetically conserved and were inconsistent between related taxa. Moreover, ranking of models’ growth rate was inconsistent in the three different media (i.e., some models’ growth rates were markedly affected by the simulated media whereas others did not; ). This inconsistency indicates that the effect of exudates on community members is selective and differs between species (i.e., exudates increase the growth rates of some species more than others), as was previously reported . The number of active exchange fluxes in each medium corresponds with the respective growth performances displaying noticably higher number of potentially active fluxes in the rich enviroenment (also when applying loopless Flux Variability Analysis [FVA]) . Overall, simulations confirmed the existence of a feasible solution space for all the 243 models as well as their capacity to predict growth in the respective environemnt . As a next step toward conducting simulations in a genuine natural-like environment, we aimed to define a single ‘rhizosphere environment’ in which growth simulations for all models would take place. Unlike the species-specific root media (poor medium + exudates) which support growth of all models by artificially including multiple carbon sources that are derived from the automatic specifications of the poor medium, including such that are not provided by the root, this simulation environment was based on the root exudates as the sole carbon sources. By avoiding the inclusion of non-exudate organic metabolites, the true-to-source rhizosphere environment was designed to reveal the hierarchical directionality of the trophic exchanges in soil, as rich media often mask various trophic interactions taking place in native communities . Additionally, the rhizosphere environment also included an array of inorganic compounds used by the 243 GSMMs, which includes trace metals, ferric, phosphoric, and sulfuric compounds. Overall, the rhizosphere environment was composed of 60 inorganic compounds together with the 33 root exudates . The rhizosphere environment supported the growth of only a subset of the GSMMs that were capable of using plant exudates . Simulating growth succession and hierarchical trophic exchanges in the rhizosphere community To reflect the indirect effect of the root on the native community (i.e., to capture the effect of root-supported bacteria on the growth of further community members), we constructed the microbial community succession module (MCSM), a CBM-based algorithm aimed at predicting community-level trophic successions. MCSM utilizes FVA to simulate and enumerate the exchange fluxes of individual models, extending their secretion profiles beyond the standard FBA-based solutions commonly used in other CBM tools designed for modeling microbial interactions . Unlike certain CBM tools designed for modeling microbial community interactions , MCSM bypasses the need to define a community objective function, as the growth of each species is simulated individually. Trophic interactions are inferred by the extent to which exchange compounds secreted by bacteria could support the growth of other community members. MCSM iteratively grows the GSMMs in a defined environment, sums up their individual secretion profiles, and updates the initial simulation environment with those secreted compounds . Applying this algorithm to a microbiome in its native environment allows delineating the potential metabolic dependencies and interactions between bacterial species in a native community. Application of MCSM over the ‘rhizosphere environment’ (i.e., first iteration, root exudates, and ‘inorganic compounds’, ) supported the growth of 27 GSMMs ( and ). Then, the initial environment was updated with 145 additional compounds predicted to be secreted by the growing community members. The second iteration supported the growth of 33 additional species whose growth was supported by compounds predicted to be secreted by bacteria that grew in the first iteration. Following the second iteration, 25 new secreted compounds were added to the rhizosphere environment. The third iteration supported the growth of 11 additional GSMMs, with one additional compound secreted. After the third iteration, the updated environment did not support the growth of any new species. Overall, iterative growth simulations resulted in the successive growth of 71 species ( ; iterations 1–3). To enlarge the array of growing species, we tested the effect of the addition of organic phosphorous sources. Organic phosphorous is typically a limiting factor in soil and its utilization varies greatly between microbial species [i.e., different P sources were shown to have a selective effect on different microbial groups ]. The initial rhizosphere environment contained only inorganic phosphorous. During first to third MCSM simulations, nine organic P compounds were secreted to the simulation medium, which was updated accordingly. At the beginning of the fourth iteration, 31 additional organic P compounds were identified by screening the species-specific poor medium and were added to the medium ( ; organic phosphorous compounds). The additional organic P compounds supported the growth of nine additional GSMMs and led to the secretion of 13 new compounds, which were added to the environment. The fifth iteration supported the growth of four additional GSMMs and two new compounds were secreted. Final simulations in the cumulative rhizosphere environment were composed of all secreted compounds and led to the same secretion and growth profile as the previous iteration. Therefore, no further growth iterations were conducted. Overall, the successive iterations connected 84 out of 243 native members of the apple rhizosphere GSMM community via trophic exchanges. The inability of the remaining bacteria to grow, despite being part of the native root microbiome, possibly reflects the selectiveness of the root environment, which fully supports the nutritional demands of only part of the soil species, whereas specific compounds that might be essential to other species are less abundant . It is important to note that the specific exudate profile used here represents a snapshot of the root metabolome as root secretion profiles are highly dynamic, reflecting both environmental and plant developmental conditions. A possible complementary explanation to the observed selective growth might be the partiality of our simulation platform, which examined only plant–bacteria and bacteria–bacteria interactions while ignoring other critical components of the rhizosphere system such as fungi, archaea, protists, and mesofauna, as well as less abundant bacterial species, components all known to metabolically interact . Finally, the MAG collection, while relatively substantial, represents only part of the microbial community. Accordingly, the iterative growth simulations represent a subset of the overall hierarchical trophic exchanges in the root environment, necessarily reflecting the partiality of the dataset. In terms of the phylogenetic distribution of the models, 27 bacterial species grew on the first iteration (in which root exudates served as the sole organic sources). These bacteria represented 14 of the 17 phyla included in the initial model collection (consisting of 243 GSMMs) and maintained a distribution frequency similar to the original community. As in the full GSMM dataset (Community bar, ), most of the species which grew in the first iteration belonged to the phyla Acidobacteriota , Proteobacteria , and Bacteroidota . This result concurred with findings from the work of Zhalnina et al., which reported that bacteria assigned to these phyla are the primary beneficiaries of root exudates . Species from 3 out of the 17 phyla that did not grow in the first iteration – Elusimicrobiota , Chlamydiota , and Fibrobacterota , did grow on the second iteration . Members of these phyla are known for their specialized metabolic dependencies. Such is the case for example with members of the Elusimicrobiota phylum, which include mostly uncultured species whose nutritional preferences are likely to be selective . At the order level, bacteria classified as Sphingomonadales (class Alphaproteobacteria ), a group known to include typical inhabitants of the root environment , grew in the initial root environment. In comparison, other root-inhabiting groups including the orders Rhizobiales and Burkholderiales , did not grow in the first iteration. Rhizobiales and Burkholderiales did, however, grow in the second and third iterations, respectively, indicating that in the simulations, the growth of these groups was dependent on exchange metabolites secreted by other community members . Overall, 158 organic compounds were secreted throughout the MCSM simulation (from which 12 compounds overlapped with the original exudate medium). These compounds varied in their distribution and were mapped into 12 biochemical categories . Whereas plant secretions are a source of various organic compounds, microbial secretions provide a source of multiple vitamins and co-factors not secreted by the plant. Microbial-secreted compounds included siderophores (staphyloferrin, salmochelin, pyoverdine, and enterochelin), vitamins (pyridoxine, pantothenate, and thiamin), and coenzymes (coenzyme A, flavin adenine dinucleotide, and flavin mononucleotide) – all known to be exchange compounds in microbial communities . In addition, microbial secretions included 11 amino acids (arginine, lysine, threonine, alanine, serine, phenylalanine, tyrosine, leucine, glutamate, isoleucine, and methionine), also known as a common exchange currency in microbial communities . Some microbial-secreted compounds, such as phenols and alkaloids, were reported to be produced by plants as secondary metabolites . Additional information regarding mean uptake and secretion degrees of compounds classified to biochemical groups is found in . Conceptually, the rhizosphere microbiota can be classified into two trophic groups: primary exudate consumers, comprising microbial species that are direct beneficiaries from the root exudates, and secondary consumers, comprising microbial species whose growth may be provided directly via the uptake of metabolites secreted by other members of the soil microbial community. In the iterative MCSM simulations, compounds secreted by some of the primary consumers largely sustained the growth of secondary consumers, which were not able to grow otherwise. The full information on the secretion profiles and models’ growths is provided in . To validate the ability of MCSM to capture trophic dependencies and succession, we further tested whether it can track the well-documented example of cellulose degradation – a multi-step process conducted by several bacterial strains that go through the conversion of cellulose and its oligosaccharide derivatives into ethanol, acetate, and glucose, which are all eventually oxidized to CO 2 . Here, the simulation followed the trophic interactions in an environment provided with cellulose oligosaccharides (4 and 6 glucose units) on the first iteration . The formed trophic successions detected along iterations captured the reported multi-step process . Associating trophic exchanges with soil health The MAG collection analyzed in this study was constructed from shotgun libraries associated with apple rootstocks cultivated in orchard soil with a documented history of ARD and healthy/recovered (seed meal-amended) soils, providing a model system for disease-conducive vs disease-suppressive rhizosphere communities . Briefly, rhizobiome communities obtained from apple rootstocks grown in replant orchard soil leading to symptomatic growth (non-amended samples) were termed ‘sick’, whereas samples in which disease symptoms were ameliorated following an established soil amendment treatment , were termed ‘healthy’. Both sick and healthy plants were characterized by distinct differences in the structure and function of their rhizosphere microbial communities in the respective soil samples . In order to correlate microbial metabolic interactions with soil performance, GSMMs were classified into one of three functional categories based on differential abundance (DA) patterns of their respective MAGs: predominantly associated with ‘healthy’ soil (H), predominantly associated with ‘sick’ soil (S), and none-associated (NA) . The functionally classified GSMMs (H, S, and NA) were consolidated into a community network of metabolic interactions by linking their potential uptake and secretion exchange profiles (as predicted along growth iterations in ). The network was built as a directed bipartite graph, in which the 84 feasible GSMM nodes and the 203 metabolite nodes (27 root exudates, 146 microbial-secreted compounds, and 30 additional organic-P compounds) were connected by 9773 directed edges, representing the metabolic exchanges of organic compounds in the native apple rhizosphere community . Further information regarding node degrees is found in . The directionality of the network enabled its untangling into sub-network motifs stemming from a root exudate to exchange interactions, and ending with an unconsumed end-metabolite. Two types of sub-networks were detected : 3-component (PM) plant exudate–microbe–microbial-secreted metabolite; and 5-component (PMM; plant–microbe–microbe) plant exudate–microbe–intermediate microbial-secreted metabolite–microbe–microbial-secreted metabolite. Overall, the network included 45,972 unique PM paths and 571,605 unique PMM paths. Participation of GSMMs in PM paths ranged from 272 to 896 occurrences . GSSM participation in PMM paths ranged from 398 to 50,628 in the first microbe position (primary exudate consumer) and 1388–19,738 occurrences in the second microbe position (secondary consumer) . Frequency of GSMMs in the first position in PMM sub-network motifs was negatively correlated with the frequency of presence in second positions, possibly indicating species-specific preferences for a specific position/trophic level in the defined environment (Pearson = −0.279; p-value = 0.009, ). In order to explore the trophic preferences of bacteria associated with the different rhizosphere soil systems, the frequency of healthy (H), sick (S), or non-associated (NA) GSMMs in the PM and PMM sub-networks was compared . GSMMs classified as S initiated a significantly higher number of PMM sub-networks (located in the first position) than GSMMs classified as NA and H . H-classified PMM paths (first position) initiated a significantly higher number of sub-networks with GSMMs classified as NA compared to S-classified GSMMs, but no more than H-classified GSMMs (second position). Other PMM types did not show a significant effect at the second position. The higher number of trophic interactions formed by the S-classified primary exudate consumers in the PMM sub-network motifs suggests that non-beneficial bacteria may have a broader spectrum in terms of their utilization potential of root-secreted carbon sources compared to plant-beneficial bacteria. This might shed light on the dynamics of ARD, in which S-classified bacteria become increasingly dominant following long-term utilization of apple-root exudates, resulting in diminished capacity of the rhizosphere microbiome to suppress soil-borne pathogens . In order to predict exchanges with potential to support/suppress dysbiosis, the frequency of DA GSMM types (i.e., H or S) associated with metabolites (either consumed or secreted) in the PM paths was assessed . Considering consumed metabolites (root exudates), three and six compounds were found to be significantly more prevalent in H- and S-classified PM paths, respectively . Notably, the S-classified root exudates included compounds reported to support dysbiosis and ARD progression. For example, the S-classified compounds gallic acid and caffeic acid (3,4-dihidroxy-trans-cinnamate) are phenylpropanoids – phenylalanine intermediate phenolic compounds secreted from plant roots following exposure to replant pathogens . Though secretion of these compounds is considered a defense response, it is hypothesized that high levels of phenolic compounds can have autotoxic effects, potentially exacerbating ARD. Additionally, it was shown that genes associated with the production of caffeic acid were upregulated in ARD-infected apple roots, relative to those grown in γ-irradiated ARD soil , and that root and soil extracts from replant-diseased trees inhibited apple seedling growth and resulted in increased seedling root production of caffeic acid . As to the microbial-secreted compounds, a total of 79 unique compounds were found to be significantly overrepresented in either S (42 compounds) or H (41 compounds) classified PM paths . Several secreted compounds classified as healthy exchanges (H) were reported to be potentially associated with beneficial functions. For instance, the compounds L -sorbose (EX_srb__L_e) and phenylacetaladehyde (EX_pacald_e), both over-represented in H paths , have been shown to inhibit the growth of fungal pathogens associated with replant disease . Phenylacetaladehyde has also been reported to have nematicidal qualities . Combining both exudate uptake data and metabolite secretion data, the full H-classified PM path 4-hydroxybenzoate; GSMM_091; catechol ( ; the consumed exudate, the GSMM, and the secreted compound, respectively) provides an exemplary model for how the proposed framework can be used to guide the design of strategies which support specific, advantageous exchanges within the rhizobiome. The root exudate 4-hydroxybenzoate is metabolized by GSMM_091 (class Verrucomicrobiae , order Pedosphaerales ) to catechol. Catechol is a precursor of a number of catecholamines, a group of compounds which was recently shown to increase apple tolerance to ARD symptoms when added to orchard . This analysis (PM; ) leads to formulating the testable prediction that 4-hydroxybenzoate can serve as a selective enhancer of catecholamine synthesizing bacteria associated with reduced ARD symptoms, and therefore serve as a potential source for indigenously produced beneficial compounds. Conclusions In this study, we present a framework combining metagenomics analyses with CBM, which can be used to gain a deeper understanding of the functionality, dynamics, and division of labor among rhizosphere bacteria, and link their environment-dependent metabolism to biological significance. This exploratory framework aims to illuminate the black box of interactions occurring in the rhizospheres of crop plants and is based on the work of , in which a gene-centric analysis of metagenomics data from apple rhizospheres was conducted . We recovered high-quality sets of environment-specific MAGs, constructed the corresponding GSMMs, and simulated community-level metabolic interactions. By including authentic apple-root exudates in the models, we were able to begin untangling the highly complex plant–bacterial and bacterial–bacterial interactions occurring in the rhizosphere environment. More specifically , we used the framework to investigate a microbial community via examining its hierarchical secretion-uptake exchanges along multiple iterations . These analyses, which linked community-derived secretion profiles with the growth of other community members, demonstrated the successive, trophic-dependent nature of microbial communities. These interactions were elucidated via construction of a community-exchange network . Possible connections between root exudates, differentially abundant (DA) bacteria, their secreted end-products, and soil health were explored using the data derived from this network. From these analyses, we were able to associate different metabolic functionalities with diseased or healthy systems, and formulate new hypotheses regarding the general function of DA bacteria in the community. The framework we present is currently conceptual. Dealing with a highly complex system such as the rhizobiome inevitably comes with limitations. These limitations include the usage of automatic GSMM reconstruction, inherent caveats of CBM and the use of single-species GSMMs, the lack of transcriptomic and spatial-chemo-physical data, and the exclusion of competition over all its forms. Furthermore, a portion of the metabolomics data used in this framework was taken from a different source (different rootstock genotype), possibly introducing further bias to the analyses. This potential factor is due to the inherent discrepancy between the conditions from which genomics and metabolomics data were collected . Also not considered in this framework is the role of eukaryotes in the microbial-metabolic interplay. Moreover, the use of an automatic GSMM reconstruction tool (CarveMe; ), though increasingly used for depicting phenotypic landscapes, is generally less accurate than manual curation of metabolic models . This approach typically neglects specialized functions involving secondary metabolism and introduces additional biases such as the overestimation of auxotrophies . Nevertheless, manual curation is practically non-realistic for hundreds of MAGs, an expected outcome considering the volume of sequencing projects nowadays. As the primary motivation of this framework is the development of a tool capable of transforming high-throughput, low-cost genomic information into testable predictions, the use of automatic metabolic network reconstruction tools was favored, despite their inherent limitations, in pursuit of addressing the necessity of pipelines systematically analyzing metagenomics data. For these reasons, among others, the framework presented here is not intended to be used as a stand-alone tool for determining microbial function. The framework presented is designed to be used as a platform to generate educated hypotheses regarding bacterial function in a specific environment in conjunction with actual carbon substrates available in the particular ecosystem under study. The hypotheses generated provide a starting point for experimental testing required to gain actual, targeted, and feasible applicable insights . While recognizing its limitations, this framework is in fact highly versatile and can be used for the characterization of a variety of microbial communities and environments. Given a set of MAGs derived from a specific environment and environmental metabolomics data, this computational framework provides a generic simulation platform for a wide and diverse range of future applications. In the current study, the root environment was represented by a single pool of resources (metabolites). As genuine root environments are highly dynamic and responsive to stimuli, a single environment can represent at best a temporary snapshot of the conditions. Conductance of simulations with several sets of resource pools (e.g., representing temporal variations in exudation profile) can add insights on their effect on trophic interactions and community dynamics. In parallel, confirming predictions made in various environments will support an iterative process that will strengthen the predictive power of the framework and improve its accuracy as a tool for generating testable hypotheses. Similarly, complementing the genomics-based approaches done here with additional layers of 'omics information (mainly transcriptomics and metabolomics) can further constrain the solution space, deflate the number of potential metabolic routes and yield more accurate predictions of GSMMs’ performances . To summarize, we have constructed a framework enabling the analysis of metabolic interactions among microbes, as well as between microbes and their hosts, in their natural environment. Where recent studies begin to apply GSMM reconstruction and analysis starting from MAGs , this work applies the MAGs to GSMMs approach to conduct large-scale CBM analysis over high-quality MAGs derived from a native rhizosphere and explore the complex network of interactions in light of the functioning of the respective agroecosystem. The application of this framework to the apple rhizobiome yielded a wealth of preliminary knowledge about the metabolic interactions occurring within it, including novel information on putative functions performed by bacteria in healthy vs replant-diseased soil systems, and potential metabolic routes to control these functions. Overall, this framework aims to advance efforts seeking to unravel the intricate world of microbial interactions in complex environments including the plant rhizosphere. The framework is provided as a three stage-detailed pipeline in GitHub , copy archived at .
Metagenomic sequencing obtained from the rhizosphere of apple rootstocks grown in orchard soil with a documented history of replant disease resulted in a total of approximately 2 billion quality reads (after filtration) at a length of 150 bp, as described in . Independent metagenomics assemblies of six different treatments yielded 1.4–2 million contigs longer than 2 kbp . Here, assemblies were binned using MetaWRAP into 296–433 high-quality MAGs for each of the six treatments ; completion and contamination thresholds were set to 90/5, respectively . De-replication (the process of combining highly similar genomes into a single representative genome) requiring 99% average nucleotide identity of the overall 2233 MAGs yielded a collection of 395 unique MAGs . Across samples, 30–36% of the raw reads were mapped to the MAG collection , in comparison to 61–71% mapped to the non-binned contigs . Using GTDB-Tk , taxonomic annotations were assigned at the phylum, order, genus, and species level for 395, 394, 237, and three of the de-replicated MAGs, respectively, reflecting the genuine diversity of the rhizosphere community, which include many uncharacterized species . Estimates of completion and contamination, total bin length and taxonomic affiliation for the MAG collections derived from each of the six assemblies, as well as the de-replicated MAGs, are provided in . As in previous reports , Proteobacteria, Acidobacteria, Actinobacteria, and Bacteroidetes were identified as the dominant phyla in the apple rhizosphere . Overall, the taxonomic distribution of the MAG collection corresponded with the profile reported for the same samples using alternative taxonomic classification approaches such as 16S rRNA amplicon sequencing and gene-based taxonomic annotations of the non-binned shotgun contigs . At the genus level, MAGs were classified into 143 genera in comparison to approximately 3000 genera that were identified for the same data based on gene-centric approaches and approximately 1000 genera based on amplicon sequencing of the same data. The functional capabilities of the bacterial genomes in the apple rhizosphere were initially assessed based on KEGG functional annotations of their gene catalogue . The 10 most frequent functional categories across the MAG collection were involved in primary metabolism, for example, carbohydrate metabolism, and the biosynthesis of essential cellular building blocks such as amino acids and vitamins. Specialized functional categories included those associated with autotrophic nutrition such as carbon fixation and metabolism of nitrogen, sulfur, and methane. Functional diversity was found to exist also when considering ubiquitous functions. For instance, though all bacteria are in need of the full set of amino acids, most genomes lack the full set of relevant biosynthesis pathways. The prevalence of biosynthetic pathways across MAGs (requiring at least a single relevant enzyme) ranges from 99.5% (e.g., glycine; missing in only two genomes) to 21% (e.g., tyrosine; detected in only 83 MAGs). Notably, the diversity of metabolic pathway completeness regarding amino acids and other essential cellular components suggests that the majority of bacterial soil species rely on an external supply of at least some of their obligatory nutritional demands.
Categorical classifications of discrete gene entities, such as pathway completeness analyses, have several inherent limitations as an approach for the contextualization and functional interpretation of genomic information. First, categorical classifications may underestimate the completeness of robust pathways with multiple redundant routes resulting in low pathway completeness. Second, pathway completeness analyses do not take into consideration the directionality and continuity of a biosynthetic process whose full conductance requires the availability of a specific environmental resource together with the required successive series of genes/reactions. Although functional potential can be inferred based on the static set of genes and enzymes present, actual metabolic performances of bacterial species in soil are dynamic and reflect multiple factors, including the availability of different nutritional sources. Such sources can be environmental inputs like root exudates or downstream exchange metabolites secreted by cohabiting bacteria. To better understand environment-dependent metabolic activity occurring in a native rhizosphere, a set of 395 GSMMs was constructed for the entire collection of the rhizosphere-bacterial community MAGs. All models were systematically subjected to validity and quality tests using MEMOTE , leaving a total of 243 GSMMs whose stoichiometric consistency was confirmed . On average, GSMMs included 1924 reactions, from which 203 were exchange reactions (specific reactions carrying out extracellular import and secretion of metabolites) and 1312 metabolites . Altogether, the GSMM set held 5152 unique metabolic reactions, 597 exchange reactions, and 2671 different compounds. The distribution of key model attributes (reactions, metabolites, and exchanges) across phyla is shown in . Model features were scalable with those reported by and . Additionally, comparison of GSMM scales indicated that the metabolic coverage (i.e., the number of reactions, which denote the potential of executing a metabolic function) of our data is within the same order of magnitude as described in recent large-scale automatic reconstructions . Next, species-specific rich (optimal) and poor (suboptimal) media were defined for each model in order to broadly assess GSMM growth capacity. A rich environment was defined as a medium containing all metabolites for which the model encodes an exchange reaction. A poor environment was defined as the minimal set of compounds enabling growth (see Methods). Essentially, poor environments contained species-specific carbon, nitrogen and phosphorous sources together with other trace elements. In addition to the two automatic media, we aimed to design a realistic simulating environment to explore the impact of the root exudates on the native community. To this end, metabolomics data from a set of studies which characterized the root exudates of Geneva 935 (G935) or Malling 26 (M26) rootstock cultivars – related to the G210 and M26 rootstocks whose rhizobiome was characterized here, were used to specify an array of apple root-derived compounds . The list of secreted compounds was consistent with other reported profiles of plant root exudates . The growth of each of the 243 GSMMs was simulated in each of the three species-specific environments (rich, poor, poor + exudates; exudates were added to corresponding poor media to ensure the exudates have a feasible effect). As expected, growth performances were higher on the poor medium supplemented with exudates in comparison to growth on the poor medium alone but lower than the growth in rich medium . Notably, GSMM growth patterns were not phylogenetically conserved and were inconsistent between related taxa. Moreover, ranking of models’ growth rate was inconsistent in the three different media (i.e., some models’ growth rates were markedly affected by the simulated media whereas others did not; ). This inconsistency indicates that the effect of exudates on community members is selective and differs between species (i.e., exudates increase the growth rates of some species more than others), as was previously reported . The number of active exchange fluxes in each medium corresponds with the respective growth performances displaying noticably higher number of potentially active fluxes in the rich enviroenment (also when applying loopless Flux Variability Analysis [FVA]) . Overall, simulations confirmed the existence of a feasible solution space for all the 243 models as well as their capacity to predict growth in the respective environemnt . As a next step toward conducting simulations in a genuine natural-like environment, we aimed to define a single ‘rhizosphere environment’ in which growth simulations for all models would take place. Unlike the species-specific root media (poor medium + exudates) which support growth of all models by artificially including multiple carbon sources that are derived from the automatic specifications of the poor medium, including such that are not provided by the root, this simulation environment was based on the root exudates as the sole carbon sources. By avoiding the inclusion of non-exudate organic metabolites, the true-to-source rhizosphere environment was designed to reveal the hierarchical directionality of the trophic exchanges in soil, as rich media often mask various trophic interactions taking place in native communities . Additionally, the rhizosphere environment also included an array of inorganic compounds used by the 243 GSMMs, which includes trace metals, ferric, phosphoric, and sulfuric compounds. Overall, the rhizosphere environment was composed of 60 inorganic compounds together with the 33 root exudates . The rhizosphere environment supported the growth of only a subset of the GSMMs that were capable of using plant exudates .
To reflect the indirect effect of the root on the native community (i.e., to capture the effect of root-supported bacteria on the growth of further community members), we constructed the microbial community succession module (MCSM), a CBM-based algorithm aimed at predicting community-level trophic successions. MCSM utilizes FVA to simulate and enumerate the exchange fluxes of individual models, extending their secretion profiles beyond the standard FBA-based solutions commonly used in other CBM tools designed for modeling microbial interactions . Unlike certain CBM tools designed for modeling microbial community interactions , MCSM bypasses the need to define a community objective function, as the growth of each species is simulated individually. Trophic interactions are inferred by the extent to which exchange compounds secreted by bacteria could support the growth of other community members. MCSM iteratively grows the GSMMs in a defined environment, sums up their individual secretion profiles, and updates the initial simulation environment with those secreted compounds . Applying this algorithm to a microbiome in its native environment allows delineating the potential metabolic dependencies and interactions between bacterial species in a native community. Application of MCSM over the ‘rhizosphere environment’ (i.e., first iteration, root exudates, and ‘inorganic compounds’, ) supported the growth of 27 GSMMs ( and ). Then, the initial environment was updated with 145 additional compounds predicted to be secreted by the growing community members. The second iteration supported the growth of 33 additional species whose growth was supported by compounds predicted to be secreted by bacteria that grew in the first iteration. Following the second iteration, 25 new secreted compounds were added to the rhizosphere environment. The third iteration supported the growth of 11 additional GSMMs, with one additional compound secreted. After the third iteration, the updated environment did not support the growth of any new species. Overall, iterative growth simulations resulted in the successive growth of 71 species ( ; iterations 1–3). To enlarge the array of growing species, we tested the effect of the addition of organic phosphorous sources. Organic phosphorous is typically a limiting factor in soil and its utilization varies greatly between microbial species [i.e., different P sources were shown to have a selective effect on different microbial groups ]. The initial rhizosphere environment contained only inorganic phosphorous. During first to third MCSM simulations, nine organic P compounds were secreted to the simulation medium, which was updated accordingly. At the beginning of the fourth iteration, 31 additional organic P compounds were identified by screening the species-specific poor medium and were added to the medium ( ; organic phosphorous compounds). The additional organic P compounds supported the growth of nine additional GSMMs and led to the secretion of 13 new compounds, which were added to the environment. The fifth iteration supported the growth of four additional GSMMs and two new compounds were secreted. Final simulations in the cumulative rhizosphere environment were composed of all secreted compounds and led to the same secretion and growth profile as the previous iteration. Therefore, no further growth iterations were conducted. Overall, the successive iterations connected 84 out of 243 native members of the apple rhizosphere GSMM community via trophic exchanges. The inability of the remaining bacteria to grow, despite being part of the native root microbiome, possibly reflects the selectiveness of the root environment, which fully supports the nutritional demands of only part of the soil species, whereas specific compounds that might be essential to other species are less abundant . It is important to note that the specific exudate profile used here represents a snapshot of the root metabolome as root secretion profiles are highly dynamic, reflecting both environmental and plant developmental conditions. A possible complementary explanation to the observed selective growth might be the partiality of our simulation platform, which examined only plant–bacteria and bacteria–bacteria interactions while ignoring other critical components of the rhizosphere system such as fungi, archaea, protists, and mesofauna, as well as less abundant bacterial species, components all known to metabolically interact . Finally, the MAG collection, while relatively substantial, represents only part of the microbial community. Accordingly, the iterative growth simulations represent a subset of the overall hierarchical trophic exchanges in the root environment, necessarily reflecting the partiality of the dataset. In terms of the phylogenetic distribution of the models, 27 bacterial species grew on the first iteration (in which root exudates served as the sole organic sources). These bacteria represented 14 of the 17 phyla included in the initial model collection (consisting of 243 GSMMs) and maintained a distribution frequency similar to the original community. As in the full GSMM dataset (Community bar, ), most of the species which grew in the first iteration belonged to the phyla Acidobacteriota , Proteobacteria , and Bacteroidota . This result concurred with findings from the work of Zhalnina et al., which reported that bacteria assigned to these phyla are the primary beneficiaries of root exudates . Species from 3 out of the 17 phyla that did not grow in the first iteration – Elusimicrobiota , Chlamydiota , and Fibrobacterota , did grow on the second iteration . Members of these phyla are known for their specialized metabolic dependencies. Such is the case for example with members of the Elusimicrobiota phylum, which include mostly uncultured species whose nutritional preferences are likely to be selective . At the order level, bacteria classified as Sphingomonadales (class Alphaproteobacteria ), a group known to include typical inhabitants of the root environment , grew in the initial root environment. In comparison, other root-inhabiting groups including the orders Rhizobiales and Burkholderiales , did not grow in the first iteration. Rhizobiales and Burkholderiales did, however, grow in the second and third iterations, respectively, indicating that in the simulations, the growth of these groups was dependent on exchange metabolites secreted by other community members . Overall, 158 organic compounds were secreted throughout the MCSM simulation (from which 12 compounds overlapped with the original exudate medium). These compounds varied in their distribution and were mapped into 12 biochemical categories . Whereas plant secretions are a source of various organic compounds, microbial secretions provide a source of multiple vitamins and co-factors not secreted by the plant. Microbial-secreted compounds included siderophores (staphyloferrin, salmochelin, pyoverdine, and enterochelin), vitamins (pyridoxine, pantothenate, and thiamin), and coenzymes (coenzyme A, flavin adenine dinucleotide, and flavin mononucleotide) – all known to be exchange compounds in microbial communities . In addition, microbial secretions included 11 amino acids (arginine, lysine, threonine, alanine, serine, phenylalanine, tyrosine, leucine, glutamate, isoleucine, and methionine), also known as a common exchange currency in microbial communities . Some microbial-secreted compounds, such as phenols and alkaloids, were reported to be produced by plants as secondary metabolites . Additional information regarding mean uptake and secretion degrees of compounds classified to biochemical groups is found in . Conceptually, the rhizosphere microbiota can be classified into two trophic groups: primary exudate consumers, comprising microbial species that are direct beneficiaries from the root exudates, and secondary consumers, comprising microbial species whose growth may be provided directly via the uptake of metabolites secreted by other members of the soil microbial community. In the iterative MCSM simulations, compounds secreted by some of the primary consumers largely sustained the growth of secondary consumers, which were not able to grow otherwise. The full information on the secretion profiles and models’ growths is provided in . To validate the ability of MCSM to capture trophic dependencies and succession, we further tested whether it can track the well-documented example of cellulose degradation – a multi-step process conducted by several bacterial strains that go through the conversion of cellulose and its oligosaccharide derivatives into ethanol, acetate, and glucose, which are all eventually oxidized to CO 2 . Here, the simulation followed the trophic interactions in an environment provided with cellulose oligosaccharides (4 and 6 glucose units) on the first iteration . The formed trophic successions detected along iterations captured the reported multi-step process .
The MAG collection analyzed in this study was constructed from shotgun libraries associated with apple rootstocks cultivated in orchard soil with a documented history of ARD and healthy/recovered (seed meal-amended) soils, providing a model system for disease-conducive vs disease-suppressive rhizosphere communities . Briefly, rhizobiome communities obtained from apple rootstocks grown in replant orchard soil leading to symptomatic growth (non-amended samples) were termed ‘sick’, whereas samples in which disease symptoms were ameliorated following an established soil amendment treatment , were termed ‘healthy’. Both sick and healthy plants were characterized by distinct differences in the structure and function of their rhizosphere microbial communities in the respective soil samples . In order to correlate microbial metabolic interactions with soil performance, GSMMs were classified into one of three functional categories based on differential abundance (DA) patterns of their respective MAGs: predominantly associated with ‘healthy’ soil (H), predominantly associated with ‘sick’ soil (S), and none-associated (NA) . The functionally classified GSMMs (H, S, and NA) were consolidated into a community network of metabolic interactions by linking their potential uptake and secretion exchange profiles (as predicted along growth iterations in ). The network was built as a directed bipartite graph, in which the 84 feasible GSMM nodes and the 203 metabolite nodes (27 root exudates, 146 microbial-secreted compounds, and 30 additional organic-P compounds) were connected by 9773 directed edges, representing the metabolic exchanges of organic compounds in the native apple rhizosphere community . Further information regarding node degrees is found in . The directionality of the network enabled its untangling into sub-network motifs stemming from a root exudate to exchange interactions, and ending with an unconsumed end-metabolite. Two types of sub-networks were detected : 3-component (PM) plant exudate–microbe–microbial-secreted metabolite; and 5-component (PMM; plant–microbe–microbe) plant exudate–microbe–intermediate microbial-secreted metabolite–microbe–microbial-secreted metabolite. Overall, the network included 45,972 unique PM paths and 571,605 unique PMM paths. Participation of GSMMs in PM paths ranged from 272 to 896 occurrences . GSSM participation in PMM paths ranged from 398 to 50,628 in the first microbe position (primary exudate consumer) and 1388–19,738 occurrences in the second microbe position (secondary consumer) . Frequency of GSMMs in the first position in PMM sub-network motifs was negatively correlated with the frequency of presence in second positions, possibly indicating species-specific preferences for a specific position/trophic level in the defined environment (Pearson = −0.279; p-value = 0.009, ). In order to explore the trophic preferences of bacteria associated with the different rhizosphere soil systems, the frequency of healthy (H), sick (S), or non-associated (NA) GSMMs in the PM and PMM sub-networks was compared . GSMMs classified as S initiated a significantly higher number of PMM sub-networks (located in the first position) than GSMMs classified as NA and H . H-classified PMM paths (first position) initiated a significantly higher number of sub-networks with GSMMs classified as NA compared to S-classified GSMMs, but no more than H-classified GSMMs (second position). Other PMM types did not show a significant effect at the second position. The higher number of trophic interactions formed by the S-classified primary exudate consumers in the PMM sub-network motifs suggests that non-beneficial bacteria may have a broader spectrum in terms of their utilization potential of root-secreted carbon sources compared to plant-beneficial bacteria. This might shed light on the dynamics of ARD, in which S-classified bacteria become increasingly dominant following long-term utilization of apple-root exudates, resulting in diminished capacity of the rhizosphere microbiome to suppress soil-borne pathogens . In order to predict exchanges with potential to support/suppress dysbiosis, the frequency of DA GSMM types (i.e., H or S) associated with metabolites (either consumed or secreted) in the PM paths was assessed . Considering consumed metabolites (root exudates), three and six compounds were found to be significantly more prevalent in H- and S-classified PM paths, respectively . Notably, the S-classified root exudates included compounds reported to support dysbiosis and ARD progression. For example, the S-classified compounds gallic acid and caffeic acid (3,4-dihidroxy-trans-cinnamate) are phenylpropanoids – phenylalanine intermediate phenolic compounds secreted from plant roots following exposure to replant pathogens . Though secretion of these compounds is considered a defense response, it is hypothesized that high levels of phenolic compounds can have autotoxic effects, potentially exacerbating ARD. Additionally, it was shown that genes associated with the production of caffeic acid were upregulated in ARD-infected apple roots, relative to those grown in γ-irradiated ARD soil , and that root and soil extracts from replant-diseased trees inhibited apple seedling growth and resulted in increased seedling root production of caffeic acid . As to the microbial-secreted compounds, a total of 79 unique compounds were found to be significantly overrepresented in either S (42 compounds) or H (41 compounds) classified PM paths . Several secreted compounds classified as healthy exchanges (H) were reported to be potentially associated with beneficial functions. For instance, the compounds L -sorbose (EX_srb__L_e) and phenylacetaladehyde (EX_pacald_e), both over-represented in H paths , have been shown to inhibit the growth of fungal pathogens associated with replant disease . Phenylacetaladehyde has also been reported to have nematicidal qualities . Combining both exudate uptake data and metabolite secretion data, the full H-classified PM path 4-hydroxybenzoate; GSMM_091; catechol ( ; the consumed exudate, the GSMM, and the secreted compound, respectively) provides an exemplary model for how the proposed framework can be used to guide the design of strategies which support specific, advantageous exchanges within the rhizobiome. The root exudate 4-hydroxybenzoate is metabolized by GSMM_091 (class Verrucomicrobiae , order Pedosphaerales ) to catechol. Catechol is a precursor of a number of catecholamines, a group of compounds which was recently shown to increase apple tolerance to ARD symptoms when added to orchard . This analysis (PM; ) leads to formulating the testable prediction that 4-hydroxybenzoate can serve as a selective enhancer of catecholamine synthesizing bacteria associated with reduced ARD symptoms, and therefore serve as a potential source for indigenously produced beneficial compounds.
In this study, we present a framework combining metagenomics analyses with CBM, which can be used to gain a deeper understanding of the functionality, dynamics, and division of labor among rhizosphere bacteria, and link their environment-dependent metabolism to biological significance. This exploratory framework aims to illuminate the black box of interactions occurring in the rhizospheres of crop plants and is based on the work of , in which a gene-centric analysis of metagenomics data from apple rhizospheres was conducted . We recovered high-quality sets of environment-specific MAGs, constructed the corresponding GSMMs, and simulated community-level metabolic interactions. By including authentic apple-root exudates in the models, we were able to begin untangling the highly complex plant–bacterial and bacterial–bacterial interactions occurring in the rhizosphere environment. More specifically , we used the framework to investigate a microbial community via examining its hierarchical secretion-uptake exchanges along multiple iterations . These analyses, which linked community-derived secretion profiles with the growth of other community members, demonstrated the successive, trophic-dependent nature of microbial communities. These interactions were elucidated via construction of a community-exchange network . Possible connections between root exudates, differentially abundant (DA) bacteria, their secreted end-products, and soil health were explored using the data derived from this network. From these analyses, we were able to associate different metabolic functionalities with diseased or healthy systems, and formulate new hypotheses regarding the general function of DA bacteria in the community. The framework we present is currently conceptual. Dealing with a highly complex system such as the rhizobiome inevitably comes with limitations. These limitations include the usage of automatic GSMM reconstruction, inherent caveats of CBM and the use of single-species GSMMs, the lack of transcriptomic and spatial-chemo-physical data, and the exclusion of competition over all its forms. Furthermore, a portion of the metabolomics data used in this framework was taken from a different source (different rootstock genotype), possibly introducing further bias to the analyses. This potential factor is due to the inherent discrepancy between the conditions from which genomics and metabolomics data were collected . Also not considered in this framework is the role of eukaryotes in the microbial-metabolic interplay. Moreover, the use of an automatic GSMM reconstruction tool (CarveMe; ), though increasingly used for depicting phenotypic landscapes, is generally less accurate than manual curation of metabolic models . This approach typically neglects specialized functions involving secondary metabolism and introduces additional biases such as the overestimation of auxotrophies . Nevertheless, manual curation is practically non-realistic for hundreds of MAGs, an expected outcome considering the volume of sequencing projects nowadays. As the primary motivation of this framework is the development of a tool capable of transforming high-throughput, low-cost genomic information into testable predictions, the use of automatic metabolic network reconstruction tools was favored, despite their inherent limitations, in pursuit of addressing the necessity of pipelines systematically analyzing metagenomics data. For these reasons, among others, the framework presented here is not intended to be used as a stand-alone tool for determining microbial function. The framework presented is designed to be used as a platform to generate educated hypotheses regarding bacterial function in a specific environment in conjunction with actual carbon substrates available in the particular ecosystem under study. The hypotheses generated provide a starting point for experimental testing required to gain actual, targeted, and feasible applicable insights . While recognizing its limitations, this framework is in fact highly versatile and can be used for the characterization of a variety of microbial communities and environments. Given a set of MAGs derived from a specific environment and environmental metabolomics data, this computational framework provides a generic simulation platform for a wide and diverse range of future applications. In the current study, the root environment was represented by a single pool of resources (metabolites). As genuine root environments are highly dynamic and responsive to stimuli, a single environment can represent at best a temporary snapshot of the conditions. Conductance of simulations with several sets of resource pools (e.g., representing temporal variations in exudation profile) can add insights on their effect on trophic interactions and community dynamics. In parallel, confirming predictions made in various environments will support an iterative process that will strengthen the predictive power of the framework and improve its accuracy as a tool for generating testable hypotheses. Similarly, complementing the genomics-based approaches done here with additional layers of 'omics information (mainly transcriptomics and metabolomics) can further constrain the solution space, deflate the number of potential metabolic routes and yield more accurate predictions of GSMMs’ performances . To summarize, we have constructed a framework enabling the analysis of metabolic interactions among microbes, as well as between microbes and their hosts, in their natural environment. Where recent studies begin to apply GSMM reconstruction and analysis starting from MAGs , this work applies the MAGs to GSMMs approach to conduct large-scale CBM analysis over high-quality MAGs derived from a native rhizosphere and explore the complex network of interactions in light of the functioning of the respective agroecosystem. The application of this framework to the apple rhizobiome yielded a wealth of preliminary knowledge about the metabolic interactions occurring within it, including novel information on putative functions performed by bacteria in healthy vs replant-diseased soil systems, and potential metabolic routes to control these functions. Overall, this framework aims to advance efforts seeking to unravel the intricate world of microbial interactions in complex environments including the plant rhizosphere. The framework is provided as a three stage-detailed pipeline in GitHub , copy archived at .
Recovery of MAGs from metagenomics data constructed for apple rhizosphere microbiomes High coverage shotgun metagenomics sequence data were obtained from microbial DNA extracted from the rhizosphere of apple rootstocks cultivated in soil from a replant-diseased orchard . The experimental design included sampling of six different soil/apple rootstock treatments with five biological replicates each, as described in . Two different apple rootstocks (M26, ARD susceptible; G210, ARD tolerant) were grown in three different treatments: (1) orchard soil amended with Brassica napus seed meal, (2) orchard soil amended with Brassica juncea / Sinapis alba seed meal (BjSa), and (3) no-treatment control soil (NTC) (see ). Microbial DNA was extracted from rhizosphere soil and metagenomics data were assembled as described in . In each assembly contigs were binned to recover MAGs using MetaWRAP pipeline (v1.3.1), which utilizes several independent binners . The MAGs recovered by the different binners were collectively processed with the Bin_refinement module of metaWRAP, producing an output of a refined bin collection. A count table was constructed by mapping raw reads data of each assembly (e.g., BjSa; G210) to the bins, using BWA-MEM (Burrows-Wheeler Aligner - Maximum Exact Match) mapping software (version 0.7.17) with default parameters. DA of the reads associated with the respective bins in each assembly across the respective replicates was determined using the edgeR function implemented in R, requiring FDR adapted p-value <0.05. Read mapping information is shown in . Based on DA, MAGs were classified either as associated with healthy soil (H; BjSa DA), sick soil (S; NTC DA), or not-associated with either soil type (NA; not DA at any treatment site). Gene calling and annotation were performed with the Annotate_bins module of MetaWRAP. Pathway completeness was determined with KEGG Decoder v 1.0.8.2 based on the KO annotations extracted from Annotate_bins assignments. The quality of the genomes was determined with CheckM . For phylogenomic analyses and taxonomic classification of each bacterial and archaeal genome, we searched for and aligned 120 bacterial marker genes of the MAGs using the identity and align commands of GTDB-Tk v1.5.0 . MAGs were de-replicated using dRep v2.3.2 using the default settings and MAGs from the six assemblies were clustered into a single non-redundant set. Phylogenomic trees were rooted by randomly selecting a genome from the sister lineage to the genus as determined from the topology of the bacterial and archaeal GTDB R06-RS202 reference trees. Closely related GTDB taxa identified with the ‘classify_wf’ workflow were filtered using the taxa-filter option during the alignment step. Overall, a set of 395 high-quality genomes (≥90% completeness,<5% contamination) was used for downstream analyses. GSMM reconstruction, analysis, and characterization of the MAG collection GSMMs were constructed for each of the 395 MAGs using CarveMe v 1.5.1 a python-based tool for GSMM reconstruction. Installation and usage of CarveMe were done as suggested in the original CarveMe webpage ( https://carveme.readthedocs.io/en/latest/ ). The solver used for GSMM reconstruction is Cplex (v. 12.8.0.0). All GSMMs were drafted without gap filling as it might mask metabolic co-dependencies . Stoichiometric consistency of all GSMMs was systematically assessed via the standardized MEMOTE test suite, a tool for GSMM quality and completion assessment . GSMMs not stoichiometrically balanced were filtered out, as they might produce infeasible simulation results. Analyses and simulations of GSMMs, as well as retrieval of model attributes (reactions, metabolites, exchanges, etc.), were conducted via the vast array of methods found in COBRApy , a python coding language package for analyzing constraint-based reconstructions. For each GSMM, initial growth simulations took place in three different model-specific environments: rich medium, poor medium, and poor medium + exudates. The rich medium was composed of all the exchange reactions (i.e., exchange reaction for specific compounds) a model holds, gathered by the ‘exchanges’ attribute found in each model. The poor medium was composed of the minimal set of compounds required for a specific GSMM/species to grow at a fixed rate. This set of compounds was identified using the minimal_medium module from COBRApy (minimize components = True, growth rate of 0.1 biomass increase hour −1 ). Poor medium + exudates was defined as the poor medium with the addition of an array of apple-root exudates. These compounds were retrieved from two metabolomics studies characterizing the exudates of apple rootstocks grown in Lane Mountain Sand (Valley, WA) . Exudate compounds were aligned with the BIGG database in order to format them for use in COBRApy. For each media type, GSMM growth rates were calculated by solving each model using the summary method in COBRApy, which utilizes Flux Balance Analysis (FBA) for maximizing biomass. Construction of a common root environment medium and application of the MCSM In order to simulate the dynamics of the rhizosphere community in the root environment, a fourth growth medium representing a natural-like environment was defined and termed the ‘rhizosphere environment’. The rhizosphere environment was composed of two arrays of compounds: (1) the exudates (as described above) and (2) inorganic compounds essential for sustaining bacterial growth. This array was determined according to the minimal set of compounds identified for each GSMM (also described above). Rhizosphere environment components are provided in . Both sets of compounds were then consolidated into one array in which further simulations were conducted. The MCSM (which is the first module out of three comprising this computational framework) simulates the growth of a microbial community by iteratively growing the GSMMs in the community and adding compounds ‘secreted’ by the growing species to the simulation environment (i.e., the medium), thus enriching the medium/environment and supporting further growth. Unlike FBA, which is used for gathering an arbitrary solution regarding non-optimized fluxes, the MCSM uses FVA to determine exchange fluxes . FVA gathers the full range of exchange fluxes (both secretion and uptake) that satisfy the objective function of a GSMM (i.e., biomass increase). The FVA fraction of optimum was set to 0.9 (sustaining the objective function at 90% optimality, allowing a less restricted secretion profile). Secretion compounds added to the updated medium in each iteration were set to be given in optimal fluxes in next iteration, to ensure a metabolic effect based on the presence of specific metabolites in the environment, rather than their quantity. Flux boundaries of updated medium components were set to 1000 mmol/gDW hour (for a specific exchange compound; gDW, gram Dry Weight). MCSM was initially simulated in the rhizosphere environment medium. After each growth iteration, GSMM growth values and the compounds secreted by the species growing were collected, where the latter were added to the medium for the next iteration as described above. Growth iterations continued until no new compounds were secreted and no additional GSMMs had grown. After the third iteration, a set of organic phosphorous compounds (containing both carbon and phosphorous) was added to the environment. These compounds were gathered from the pool of model-specific minimal compounds selected for use in the poor medium. Information regarding the chemical formula of these compounds was gathered with the formula attribute of each compound object in COBRApy. Along MCSM iterations, secreted metabolites were classified into biochemical categories based on BRITE annotations or, in the absence of classification, manually. MCSM was further applied to inspect the framework’s ability to tracing cellulose degradation using cellulose medium . The tutorial for the MCSM stage of the framework workflow is found in GitHub , along with the GSMMs and media files. Instructions for conducting the analysis are in the README.md file. Construction of the exchange network and its untangling for screening sub-network motifs For each GSMM, a directed bipartite network representing all potential metabolic exchanges occurring within the rhizosphere community was constructed based on uptake and secretion data derived from MCSM iterations. Edges in the network were connected between GSMM nodes and metabolite nodes, with edge directionality indicating either secretion or uptake of a metabolite by a specific GSMM. The network was constructed with the networkx package, a python language package for the exploration and analysis of networks and network algorithms. Network-specific topography was obtained using the Kamada-kawai layout . Information regarding the degree and connectivity of the different node types was acquired from the graph object (G). Code for the network construction module in the framework is found at the project’s GitHub page under the name NETWORK.py. Untangling the exchange network into individual paths (i.e., sub-network motifs) was done using the all_shortest_paths function in networkx , applied over the exchanges network. Briefly, the algorithm screens for all possible shortest paths within the network, specifically screening for paths starting with an exudate node and ending secreted metabolite nodes (secreted by bacterial species but not consumed). This algorithm yielded two types of paths: (1) PM paths in which node positions one, two, and three represented exudate, microbe, and secreted metabolite, respectively, and (2) PMM paths. PMM paths (length of five nodes) were constructed based on PM paths (length of three nodes), in which positions four and five displayed unique (i.e., not found in PM paths) microbe and metabolite nodes, respectively. Code for the network untangling module in the framework is found at the project’s GitHub page under the name PATHS.py. Associating PM and PMM sub-network motifs features with soil health Sub-network motifs (PM and PMM) were functionally classified as associated with healthy soil (H), sick soil (S), or not-associated with either soil type (NA) based on differential abundance of the corresponding MAGs (PM) or MAGs combination (PMM) in the sub-network. Next, the GSMMs in both PM and PMM sub-networks were characterized according functional classification. For PMMs, the distribution of counts of classified sub-networks, at the different positions, was compared using the ANOVA test, followed by a Tukey test to significantly distinguish the groups. GSMM classifications were further projected on uptake and secreted metabolites in the pathway motifs. For simplification, the analysis focused only on PM paths because PMM paths incorporate PM paths, and exchanges within a PMM path do not directly reflect the effect of an exudate on the secreted end-product (but over the intermediate compound). On that account, start/end metabolites in PM paths were associated with H/S/NA paths based on one-sided hypergeometric test, comparing the frequency of each compound in a functionally characterized path type (either H or S) vs its frequency in NA classified paths and the reciprocal dataset.
High coverage shotgun metagenomics sequence data were obtained from microbial DNA extracted from the rhizosphere of apple rootstocks cultivated in soil from a replant-diseased orchard . The experimental design included sampling of six different soil/apple rootstock treatments with five biological replicates each, as described in . Two different apple rootstocks (M26, ARD susceptible; G210, ARD tolerant) were grown in three different treatments: (1) orchard soil amended with Brassica napus seed meal, (2) orchard soil amended with Brassica juncea / Sinapis alba seed meal (BjSa), and (3) no-treatment control soil (NTC) (see ). Microbial DNA was extracted from rhizosphere soil and metagenomics data were assembled as described in . In each assembly contigs were binned to recover MAGs using MetaWRAP pipeline (v1.3.1), which utilizes several independent binners . The MAGs recovered by the different binners were collectively processed with the Bin_refinement module of metaWRAP, producing an output of a refined bin collection. A count table was constructed by mapping raw reads data of each assembly (e.g., BjSa; G210) to the bins, using BWA-MEM (Burrows-Wheeler Aligner - Maximum Exact Match) mapping software (version 0.7.17) with default parameters. DA of the reads associated with the respective bins in each assembly across the respective replicates was determined using the edgeR function implemented in R, requiring FDR adapted p-value <0.05. Read mapping information is shown in . Based on DA, MAGs were classified either as associated with healthy soil (H; BjSa DA), sick soil (S; NTC DA), or not-associated with either soil type (NA; not DA at any treatment site). Gene calling and annotation were performed with the Annotate_bins module of MetaWRAP. Pathway completeness was determined with KEGG Decoder v 1.0.8.2 based on the KO annotations extracted from Annotate_bins assignments. The quality of the genomes was determined with CheckM . For phylogenomic analyses and taxonomic classification of each bacterial and archaeal genome, we searched for and aligned 120 bacterial marker genes of the MAGs using the identity and align commands of GTDB-Tk v1.5.0 . MAGs were de-replicated using dRep v2.3.2 using the default settings and MAGs from the six assemblies were clustered into a single non-redundant set. Phylogenomic trees were rooted by randomly selecting a genome from the sister lineage to the genus as determined from the topology of the bacterial and archaeal GTDB R06-RS202 reference trees. Closely related GTDB taxa identified with the ‘classify_wf’ workflow were filtered using the taxa-filter option during the alignment step. Overall, a set of 395 high-quality genomes (≥90% completeness,<5% contamination) was used for downstream analyses.
GSMMs were constructed for each of the 395 MAGs using CarveMe v 1.5.1 a python-based tool for GSMM reconstruction. Installation and usage of CarveMe were done as suggested in the original CarveMe webpage ( https://carveme.readthedocs.io/en/latest/ ). The solver used for GSMM reconstruction is Cplex (v. 12.8.0.0). All GSMMs were drafted without gap filling as it might mask metabolic co-dependencies . Stoichiometric consistency of all GSMMs was systematically assessed via the standardized MEMOTE test suite, a tool for GSMM quality and completion assessment . GSMMs not stoichiometrically balanced were filtered out, as they might produce infeasible simulation results. Analyses and simulations of GSMMs, as well as retrieval of model attributes (reactions, metabolites, exchanges, etc.), were conducted via the vast array of methods found in COBRApy , a python coding language package for analyzing constraint-based reconstructions. For each GSMM, initial growth simulations took place in three different model-specific environments: rich medium, poor medium, and poor medium + exudates. The rich medium was composed of all the exchange reactions (i.e., exchange reaction for specific compounds) a model holds, gathered by the ‘exchanges’ attribute found in each model. The poor medium was composed of the minimal set of compounds required for a specific GSMM/species to grow at a fixed rate. This set of compounds was identified using the minimal_medium module from COBRApy (minimize components = True, growth rate of 0.1 biomass increase hour −1 ). Poor medium + exudates was defined as the poor medium with the addition of an array of apple-root exudates. These compounds were retrieved from two metabolomics studies characterizing the exudates of apple rootstocks grown in Lane Mountain Sand (Valley, WA) . Exudate compounds were aligned with the BIGG database in order to format them for use in COBRApy. For each media type, GSMM growth rates were calculated by solving each model using the summary method in COBRApy, which utilizes Flux Balance Analysis (FBA) for maximizing biomass.
In order to simulate the dynamics of the rhizosphere community in the root environment, a fourth growth medium representing a natural-like environment was defined and termed the ‘rhizosphere environment’. The rhizosphere environment was composed of two arrays of compounds: (1) the exudates (as described above) and (2) inorganic compounds essential for sustaining bacterial growth. This array was determined according to the minimal set of compounds identified for each GSMM (also described above). Rhizosphere environment components are provided in . Both sets of compounds were then consolidated into one array in which further simulations were conducted. The MCSM (which is the first module out of three comprising this computational framework) simulates the growth of a microbial community by iteratively growing the GSMMs in the community and adding compounds ‘secreted’ by the growing species to the simulation environment (i.e., the medium), thus enriching the medium/environment and supporting further growth. Unlike FBA, which is used for gathering an arbitrary solution regarding non-optimized fluxes, the MCSM uses FVA to determine exchange fluxes . FVA gathers the full range of exchange fluxes (both secretion and uptake) that satisfy the objective function of a GSMM (i.e., biomass increase). The FVA fraction of optimum was set to 0.9 (sustaining the objective function at 90% optimality, allowing a less restricted secretion profile). Secretion compounds added to the updated medium in each iteration were set to be given in optimal fluxes in next iteration, to ensure a metabolic effect based on the presence of specific metabolites in the environment, rather than their quantity. Flux boundaries of updated medium components were set to 1000 mmol/gDW hour (for a specific exchange compound; gDW, gram Dry Weight). MCSM was initially simulated in the rhizosphere environment medium. After each growth iteration, GSMM growth values and the compounds secreted by the species growing were collected, where the latter were added to the medium for the next iteration as described above. Growth iterations continued until no new compounds were secreted and no additional GSMMs had grown. After the third iteration, a set of organic phosphorous compounds (containing both carbon and phosphorous) was added to the environment. These compounds were gathered from the pool of model-specific minimal compounds selected for use in the poor medium. Information regarding the chemical formula of these compounds was gathered with the formula attribute of each compound object in COBRApy. Along MCSM iterations, secreted metabolites were classified into biochemical categories based on BRITE annotations or, in the absence of classification, manually. MCSM was further applied to inspect the framework’s ability to tracing cellulose degradation using cellulose medium . The tutorial for the MCSM stage of the framework workflow is found in GitHub , along with the GSMMs and media files. Instructions for conducting the analysis are in the README.md file.
For each GSMM, a directed bipartite network representing all potential metabolic exchanges occurring within the rhizosphere community was constructed based on uptake and secretion data derived from MCSM iterations. Edges in the network were connected between GSMM nodes and metabolite nodes, with edge directionality indicating either secretion or uptake of a metabolite by a specific GSMM. The network was constructed with the networkx package, a python language package for the exploration and analysis of networks and network algorithms. Network-specific topography was obtained using the Kamada-kawai layout . Information regarding the degree and connectivity of the different node types was acquired from the graph object (G). Code for the network construction module in the framework is found at the project’s GitHub page under the name NETWORK.py. Untangling the exchange network into individual paths (i.e., sub-network motifs) was done using the all_shortest_paths function in networkx , applied over the exchanges network. Briefly, the algorithm screens for all possible shortest paths within the network, specifically screening for paths starting with an exudate node and ending secreted metabolite nodes (secreted by bacterial species but not consumed). This algorithm yielded two types of paths: (1) PM paths in which node positions one, two, and three represented exudate, microbe, and secreted metabolite, respectively, and (2) PMM paths. PMM paths (length of five nodes) were constructed based on PM paths (length of three nodes), in which positions four and five displayed unique (i.e., not found in PM paths) microbe and metabolite nodes, respectively. Code for the network untangling module in the framework is found at the project’s GitHub page under the name PATHS.py.
Sub-network motifs (PM and PMM) were functionally classified as associated with healthy soil (H), sick soil (S), or not-associated with either soil type (NA) based on differential abundance of the corresponding MAGs (PM) or MAGs combination (PMM) in the sub-network. Next, the GSMMs in both PM and PMM sub-networks were characterized according functional classification. For PMMs, the distribution of counts of classified sub-networks, at the different positions, was compared using the ANOVA test, followed by a Tukey test to significantly distinguish the groups. GSMM classifications were further projected on uptake and secreted metabolites in the pathway motifs. For simplification, the analysis focused only on PM paths because PMM paths incorporate PM paths, and exchanges within a PMM path do not directly reflect the effect of an exudate on the secreted end-product (but over the intermediate compound). On that account, start/end metabolites in PM paths were associated with H/S/NA paths based on one-sided hypergeometric test, comparing the frequency of each compound in a functionally characterized path type (either H or S) vs its frequency in NA classified paths and the reciprocal dataset.
|
Efficacy of non -thermal pressure plasma versus other modalities for disinfection of primary root canals | 7f488bea-6f19-4d83-9f20-e27aa1ae5e31 | 11725210 | Dentistry[mh] | Early exfoliation of primary teeth causes esthetic, biological, and functional problems . Thus the goal of treatment of deciduous teeth with large carious lesions is dental arch preservation, restoring healthy condition of the affected pulpal tissues, and maintaining normal development of permanent successors. Pulpectomy eliminates microorganisms from root canals by chemical irrigation and mechanical debridement . Its success depends on optimum root canal irrigation and effective disinfection. Enterococcus faecalis is considered a survivor of the chemo-mechanical steps, with its virulence factors that promote its adhesion to host cells. It can resist nutrient deprivation in endodontically treated teeth and attach to the collagen present in the dentin, showing resistance to chemo-mechanical procedures . Root canal is an enclosed complex anatomical space with morphological and microbiological challenges. Usual instrumentation techniques pile debris in the isthmus areas so; the ideal irrigant should be a biocompatible bactericidal, pulp tissue solvent, lubricating agent, smear layer and debris remover, and with sustained effect but without influencing dentin physical properties . Several adopted modes for root canal sanitization in deciduous teeth are found in the literature. Sodium hypochlorite (NaOCl) is the gold standard irrigant by virtue of its apparent antibacterial characteristics, ease of manipulation, decreased surface tension, and cheapness. However, it can negatively affect the permanent successors’ dental follicles in the presence of root resorption . Chlorhexidine is a sustained broad-spectrum antimicrobial irrigating solution with low toxicity. CHX is a synthetic cationic biguanide that interacts with the negative charges of the phosphate groups on the microbial wall, resulting in osmotic un-equilibrium of the cell . A diode laser has various applications in dentistry with promising disinfection outcomes . The bactericidal explicit impact of a diode laser (810 nm, penetrates more than 1 mm through dentinal thickness) depends on thermal affection; bacteria cannot resist laser exposure . Plasma medicine has drawn increased attention in biomedical fields including sterilization, surface modification, and cancer treatments. Plasma is the fourth state of matter, where atmosphere pressure nonequilibrium plasma (APNP) jet generates active radicles and charged particles that have antimicrobial activity in open space at room temperature . NTPP jet has been applied in various dental studies to effectively deactivate major pathogenic microorganisms in oral cavity and their biofilms such as Pseudomonas aeruinosa, Streptcoccus mutans, and enterococcus faecalis . Moreover the antimicrobial characteristics, NTPPs have shown anti-inflammatory trails, antifungal, antiparasitic, antiviral properties, and tissue repair . Also, it was reported that NTPP as a prosperous tool used as an adjunctive therapy against malignancy . NTPPs have two modes of action and both are efficient in dentistry; direct action by applying NTPP plasma plume to a set substrate, or circuitous action, through liquid activation by NTPP before its application to the substrate . Propolis is a wax-like biomaterial gained naturally from beehives; it has a wide range of biomedical implementations such as tissue regeneration, ulcers, oral infections, skin lesions, herpes, and wound healing . Various endodontic usage of propolis through intracanal irrigation intracanal medication , post-endodontic pain relief and vital pulp therapy have been examined . Therefore, this study compares the efficacy of NTPP, diode laser, propolis, and CHX for disinfection of deciduous anterior root canals infected with E. faecalis after sterilization by gamma radiation.
Teeth selection Forty primary anterior teeth were collected from faculty of dental medicine for girls’ Al-Azhar University outpatient dental clinics after obtaining a written consent from the accompanying parent for using the teeth for research purposes; research guidelines of ethics committee with approval code (REC-PD-24-03) were followed. Regulatory statement No human or living tissue participation in this study, so it was exempt from clinical trial registration. All teeth were picked according to the next criteria: A-Inclusion criteria Primary anterior teeth. Free from internal or external resorption. Complete root formation. Mechanical preparation of teeth Teeth cleansing was done to remove any tissue debris using an ultrasonic scaler, rinsed and disinfected with 1% Naocl, and then stored in thymol preservative till use. Access cavity was prepared using a round bur and tapered diamond bur (Jiangsu-China) till reaching the orifices of canals, pulp tissue was extirpated using H-file (Mani Japan) then root canal preparation was completed by Kido-s rotary files (India). ( ، ) Organic pulp tissue remnants were rinsed with 1 ml of 1% Naocl while the inorganic smear layer of root dentin was removed using 17% EDTA with each file; the canals were then flushed with 0.9% saline to remove residual irrigants and stop their effects . Teeth decoronation Teeth coronectomy below the cementoenamel junction was done till having a standardized 8 mm root length using a table motor with diamond size under coolant water spray. These root canals with standard lengths act as reservoirs for bacterial suspension . Light-cured composite resin was used to seal the apical foramen to prevent the extrusion of bacterial suspension and irrigant solutions. Samples sterilization Root samples were inserted into sterilization pouches after mechanical preparation and sterilized by gamma irradiation (Cobalt 60 irradiator of 1.774 KGY dose rate, total dose of 25 KGY, Egyptian Atomic Energy Authority). Sterility test Before injection of bacterial suspension, a sterility test was done to ensure bacterial (gram positive or gram negative) and fungal decontamination of the sterilized root samples. These were done by injection of the root canals with sterile saline, and then insertion of three sterile absorbent paper points into the root canals for one minute until they became saturated with the injected fluid and then transferred into:. Nutrient Agar (to exclude gram + ve bacterial contamination). MacConkey’s agar (to exclude gram -ve bacterial contamination). Sabouraud dextrose agar (to exclude fungal contamination). Plates incubation at 37° for 24 h under observation was done, to check effective sterilization if they remain clear. Sample size calculation Sample size has been determined to be ten samples per group referenced to feigning α = 0.05, β = 0.2, mean standard deviation of log colony-forming units to be 2.9 and practical size of 0.76 utilizing one-way analysis of variance (ANOVA) by G*Power 3.1.9.4 software. Specimens grouping They were divided into four groups; each group consisted of ten samples: Group I: irrigated with chlorhexidine. Group II: treated with diode laser (810 mm). Group III: irrigated with ethanolic extract of propolis. Group IV: treated with NTPP. . Propolis extracts preparation Extraction of propolis was carried out by maceration with 96% ethanol. Three hundred grams of propolis were measured by digital electronic scale and were combined with 300 ml 96% ethanol at 37 o C to obtain 100% (w) extract, then stored in a bottle closed tightly for one week. Then, the supernatant was filtered using a chromafil filter to eliminate impurities . Selection and preparation of bacterial microorganisms I-Bacterial reference strain Enterococcus faecalis references strain American Type of Culture Collection (ATCC 49533) was kindly provided by the regional center of mycology and biotechnology Al Azhar University of Egypt, to be used in this study, it was supplied as an actively growing culture on slope agar. II- Bacterial substructure preparation Microorganisms were preserved at -70℃ in brain heart infusion broth with 15% glycerol, freshly set subculture on bile auscline medium and previously incubated for 24 h at 37⁰ C. III- Inoculation of bacteria Forcible injections of 1 ml of bacterial suspension were carried out to be sure of reaching the entire working length of each root canal by a sterile syringe; samples then were placed individually submerged with 2 ml of brain heart infusion broth inside tightly sealed Eppendorf tubes (Fisherscl. Co.Uk) then incubated at 37℃ for 24 h for allowing bacterial multiplication . IV- Bacterial count First microbial sample (S1) was driven from each root canal by inserting three successive sizes of sterile absorbent paper points (Diadent- Co, Korea.) into each root canal for one minute till being fully satiated with bacteria , then removed and added to1ml saline in a sterile falcon tube. A sequent 10-fold dilution of microbial suspension in sterile saline (1/10, 1/100, 1/1000, 1/10000, 1/100000) was prepared using a micropipette, 0.1 ml from each dilution was plated on the brain heart infusion agar plates using the bacteriologic loop then incubated at 37°Cfor 24 h. Colony count through multiplying the amount of colony-forming units/plate by the dilution and volume factor . Canal disinfection Root canal samples were divided into four main groups ( n = 10), and a description of microbiological samples is shown in Table : Group I After the incubation with the bacterial suspension, root canals were irrigated with 5 ml of 2% chlorohexidine solution and left for 1 min, followed by the insertion of three sterile paper points using a sterile tweezer to take the second sample . (S2a). Group II After the incubation with the bacterial suspension, root canals were dried and irradiated by a diode laser (Elexxion, Claros Plcco, Germany) with output power 2w for 5s and a wavelength of 810 nm in continuous mode. An optical fiber 200 μm in diameter was inserted into a canal 1 mm shorter than the working length. Four times irradiation repetition at 10-second time interval with an energy density 2.68 J/mm 2 , and then the bacterial counting was performed . (S2b). Group III After the incubation with the bacterial suspension, they were irrigated with 5 ml of Ethanolic extract of propolis solution and left inside the root canal for 1 min then three sterile paper points were inserted by sterile tweezer in the root canals to take the second bacterial sample . (S2c) Group IV fter the incubation with bacterial suspension, root canals were subjected to experimental NTPP (Center of Plasma Technology-Faculty of Science- Al Azhar University-Cairo-Egypt) with ionized helium gas utilizing a cold plasma hand-piece with 16.1 kHz frequency, 3.4 kV input power, 4 L/min flow rate, and rigour of 4 for 1 min (Fig. , A-B),. A nozzle tip distance was 4 mm from the specimen exterior surface (Fig. , C), and then bacterial counting was performed (S2d) .
Forty primary anterior teeth were collected from faculty of dental medicine for girls’ Al-Azhar University outpatient dental clinics after obtaining a written consent from the accompanying parent for using the teeth for research purposes; research guidelines of ethics committee with approval code (REC-PD-24-03) were followed.
No human or living tissue participation in this study, so it was exempt from clinical trial registration. All teeth were picked according to the next criteria:
Primary anterior teeth. Free from internal or external resorption. Complete root formation.
Teeth cleansing was done to remove any tissue debris using an ultrasonic scaler, rinsed and disinfected with 1% Naocl, and then stored in thymol preservative till use. Access cavity was prepared using a round bur and tapered diamond bur (Jiangsu-China) till reaching the orifices of canals, pulp tissue was extirpated using H-file (Mani Japan) then root canal preparation was completed by Kido-s rotary files (India). ( ، ) Organic pulp tissue remnants were rinsed with 1 ml of 1% Naocl while the inorganic smear layer of root dentin was removed using 17% EDTA with each file; the canals were then flushed with 0.9% saline to remove residual irrigants and stop their effects .
Teeth coronectomy below the cementoenamel junction was done till having a standardized 8 mm root length using a table motor with diamond size under coolant water spray. These root canals with standard lengths act as reservoirs for bacterial suspension . Light-cured composite resin was used to seal the apical foramen to prevent the extrusion of bacterial suspension and irrigant solutions.
Root samples were inserted into sterilization pouches after mechanical preparation and sterilized by gamma irradiation (Cobalt 60 irradiator of 1.774 KGY dose rate, total dose of 25 KGY, Egyptian Atomic Energy Authority).
Before injection of bacterial suspension, a sterility test was done to ensure bacterial (gram positive or gram negative) and fungal decontamination of the sterilized root samples. These were done by injection of the root canals with sterile saline, and then insertion of three sterile absorbent paper points into the root canals for one minute until they became saturated with the injected fluid and then transferred into:. Nutrient Agar (to exclude gram + ve bacterial contamination). MacConkey’s agar (to exclude gram -ve bacterial contamination). Sabouraud dextrose agar (to exclude fungal contamination). Plates incubation at 37° for 24 h under observation was done, to check effective sterilization if they remain clear.
Sample size has been determined to be ten samples per group referenced to feigning α = 0.05, β = 0.2, mean standard deviation of log colony-forming units to be 2.9 and practical size of 0.76 utilizing one-way analysis of variance (ANOVA) by G*Power 3.1.9.4 software.
They were divided into four groups; each group consisted of ten samples: Group I: irrigated with chlorhexidine. Group II: treated with diode laser (810 mm). Group III: irrigated with ethanolic extract of propolis. Group IV: treated with NTPP. .
Extraction of propolis was carried out by maceration with 96% ethanol. Three hundred grams of propolis were measured by digital electronic scale and were combined with 300 ml 96% ethanol at 37 o C to obtain 100% (w) extract, then stored in a bottle closed tightly for one week. Then, the supernatant was filtered using a chromafil filter to eliminate impurities .
I-Bacterial reference strain Enterococcus faecalis references strain American Type of Culture Collection (ATCC 49533) was kindly provided by the regional center of mycology and biotechnology Al Azhar University of Egypt, to be used in this study, it was supplied as an actively growing culture on slope agar. II- Bacterial substructure preparation Microorganisms were preserved at -70℃ in brain heart infusion broth with 15% glycerol, freshly set subculture on bile auscline medium and previously incubated for 24 h at 37⁰ C. III- Inoculation of bacteria Forcible injections of 1 ml of bacterial suspension were carried out to be sure of reaching the entire working length of each root canal by a sterile syringe; samples then were placed individually submerged with 2 ml of brain heart infusion broth inside tightly sealed Eppendorf tubes (Fisherscl. Co.Uk) then incubated at 37℃ for 24 h for allowing bacterial multiplication . IV- Bacterial count First microbial sample (S1) was driven from each root canal by inserting three successive sizes of sterile absorbent paper points (Diadent- Co, Korea.) into each root canal for one minute till being fully satiated with bacteria , then removed and added to1ml saline in a sterile falcon tube. A sequent 10-fold dilution of microbial suspension in sterile saline (1/10, 1/100, 1/1000, 1/10000, 1/100000) was prepared using a micropipette, 0.1 ml from each dilution was plated on the brain heart infusion agar plates using the bacteriologic loop then incubated at 37°Cfor 24 h. Colony count through multiplying the amount of colony-forming units/plate by the dilution and volume factor .
Enterococcus faecalis references strain American Type of Culture Collection (ATCC 49533) was kindly provided by the regional center of mycology and biotechnology Al Azhar University of Egypt, to be used in this study, it was supplied as an actively growing culture on slope agar.
Microorganisms were preserved at -70℃ in brain heart infusion broth with 15% glycerol, freshly set subculture on bile auscline medium and previously incubated for 24 h at 37⁰ C.
Forcible injections of 1 ml of bacterial suspension were carried out to be sure of reaching the entire working length of each root canal by a sterile syringe; samples then were placed individually submerged with 2 ml of brain heart infusion broth inside tightly sealed Eppendorf tubes (Fisherscl. Co.Uk) then incubated at 37℃ for 24 h for allowing bacterial multiplication .
First microbial sample (S1) was driven from each root canal by inserting three successive sizes of sterile absorbent paper points (Diadent- Co, Korea.) into each root canal for one minute till being fully satiated with bacteria , then removed and added to1ml saline in a sterile falcon tube. A sequent 10-fold dilution of microbial suspension in sterile saline (1/10, 1/100, 1/1000, 1/10000, 1/100000) was prepared using a micropipette, 0.1 ml from each dilution was plated on the brain heart infusion agar plates using the bacteriologic loop then incubated at 37°Cfor 24 h. Colony count through multiplying the amount of colony-forming units/plate by the dilution and volume factor .
Root canal samples were divided into four main groups ( n = 10), and a description of microbiological samples is shown in Table :
After the incubation with the bacterial suspension, root canals were irrigated with 5 ml of 2% chlorohexidine solution and left for 1 min, followed by the insertion of three sterile paper points using a sterile tweezer to take the second sample . (S2a).
After the incubation with the bacterial suspension, root canals were dried and irradiated by a diode laser (Elexxion, Claros Plcco, Germany) with output power 2w for 5s and a wavelength of 810 nm in continuous mode. An optical fiber 200 μm in diameter was inserted into a canal 1 mm shorter than the working length. Four times irradiation repetition at 10-second time interval with an energy density 2.68 J/mm 2 , and then the bacterial counting was performed . (S2b).
After the incubation with the bacterial suspension, they were irrigated with 5 ml of Ethanolic extract of propolis solution and left inside the root canal for 1 min then three sterile paper points were inserted by sterile tweezer in the root canals to take the second bacterial sample . (S2c)
fter the incubation with bacterial suspension, root canals were subjected to experimental NTPP (Center of Plasma Technology-Faculty of Science- Al Azhar University-Cairo-Egypt) with ionized helium gas utilizing a cold plasma hand-piece with 16.1 kHz frequency, 3.4 kV input power, 4 L/min flow rate, and rigour of 4 for 1 min (Fig. , A-B),. A nozzle tip distance was 4 mm from the specimen exterior surface (Fig. , C), and then bacterial counting was performed (S2d) .
Data were statistically analyzed utilizing SPSS version 26 by one-way ANOVA for (general comparison) followed by Bonferroni post hoc analysis for pairwise comparisons at 0.05 significance level. I-Comparison within groups regarding colony-forming units (log CFUs / mL) before and after irrigation Table and Fig. represent the colony-forming units before and after irrigation (log CFUs/Ml). There was a significant difference between measured values before and after four irrigation types ( p < 0.001) for CHX, Diode Laser, NTPP, and ( P = 0.035) for Propolis. The highest values of colony reduction measured before and after irrigation were for NTPP (4.06 ± 0.88). II- Comparison between groups regarding reduction in colony forming units count after the irrigation Table and Fig. represent the reduction percentage in colony-forming units in each group after the irrigation compared with the baseline (before the irrigation). The highest reduction in colony-forming units count was noticed in NTPP group (98.79%), while the least reduction in colony-forming units count was noted in Propolis group (81.99%). Comparison of the four groups regarding bacterial count reduction by one-way ANOVA displayed a significant variation between groups ( P < 0.001). It showed that CHX was significantly more efficacious than Propolis in reducing bacterial count ( P = 0.001). Diode Laser was significantly more efficacious than Propolis in bacterial count reduction ( P < 0.001). NTPP was significantly more efficacious than CHX ( P < 0.001), Diode Laser ( P < 0.001), and Propolis ( P < 0.001) for decreasing bacterial count. No other considerable distinctions were recorded ( p > 0.05).
Table and Fig. represent the colony-forming units before and after irrigation (log CFUs/Ml). There was a significant difference between measured values before and after four irrigation types ( p < 0.001) for CHX, Diode Laser, NTPP, and ( P = 0.035) for Propolis. The highest values of colony reduction measured before and after irrigation were for NTPP (4.06 ± 0.88).
Table and Fig. represent the reduction percentage in colony-forming units in each group after the irrigation compared with the baseline (before the irrigation). The highest reduction in colony-forming units count was noticed in NTPP group (98.79%), while the least reduction in colony-forming units count was noted in Propolis group (81.99%). Comparison of the four groups regarding bacterial count reduction by one-way ANOVA displayed a significant variation between groups ( P < 0.001). It showed that CHX was significantly more efficacious than Propolis in reducing bacterial count ( P = 0.001). Diode Laser was significantly more efficacious than Propolis in bacterial count reduction ( P < 0.001). NTPP was significantly more efficacious than CHX ( P < 0.001), Diode Laser ( P < 0.001), and Propolis ( P < 0.001) for decreasing bacterial count. No other considerable distinctions were recorded ( p > 0.05).
This study compared the efficacy of NTPP, diode laser, propolis, and CHX for sanitization of deciduous anterior root canals colonized with E. faecalis after sterilization by gamma radiation. Primary Root canal samples were used to simulate the children’s mouth. This agreed with a previous study that compared the antimicrobial effect of different irrigants against E. faecalis in deciduous anterior teeth. Standardized root canals with 8 mm length to act as reservoirs for bacterial suspension were prepared from primary anterior root canals . Gamma ray (25 kGy) was used for sterilization as it completely eradicates all bacterial forms that may be found in the root canals, without detectable changes in dentinal tissues while other methods (dry heat, ethylene oxide, and autoclaving) of sterilization may affect dentin . E. faecalis was selected for the study as it is the most dominant species in reinfection cases . A classic colony counting technique was used in this study; to determine the antibacterial efficacy of the tested irrigant by counting the vital E. faecalis colonies on the media plates. Results proofed that all the used methods; reduced colony count, the highest reduction was noted in NTPP treated group, while the most minor reduction was noticed in Propolis treated group، this agrees with a previous research which reported that NTPP lessened E. faecalis colony count on a glass slab after two minutes. It was concluded that NTPP was more efficient than photodynamic therapy and diode laser for removal of E. faecalis in primary root canals and can act as an alternate to 2.5% NaOCl irrigation and this is in agreement with the current study findings. Considering the significance of rapid proceedings in pedodontics, a 99% reduction in colony-forming units after 1 min of NTPP application was reported . Such an agreeable result in root canal disinfection in a brief period; recommends its clinical application . NTPP generates reactive species, including ions, electrons, and neutral particles, which interact with microbial cells and biofilms present within the root canal system. The oxidative stress induced by these reactive species leads to cellular damage, disrupting the integrity of bacterial membranes and causing cell death. Additionally, NTPP promotes the formation of reactive oxygen and nitrogen species that enhance the antibacterial effect while preserving the dental tissue . Antimicrobial efficacy of NTPP was conditioned with time and 3 min of exposure time recorded superior results, a significant reduction of biofilm microorganisms on helium or He/O 2 NTPP exposure for 4, 6, and 8 min was also reported . This study used 810 nm diode lasers in accordance with a study which concluded that 810 nm diode laser causes a considerable decrease in E.faecalis colony count, in a continuous mode than pulse mode . Disinfection with 2% CHX dramatically reduced the microbial colony count; this agrees with a previous study that compared the effect of 2% CHX and diode laser on root canal disinfection and concluded that both techniques decreased bacterial count significantly whereas the diode laser was more efficacious than 2% chlorhexidine. The results showed that NTPP was more effective in reduction of bacterial count than CHX 2%. Nevertheless the antibacterial efficiency of the plasma device in another study was analogous to that of 2% CHX. Propolis was used as a natural alternative irrigant in the current study as it affects the bacterial cell wall, causing structural and functional damage. It produced the most minor reduction in bacterial count. NTPP, CHX, and diode laser were significantly efficient than Propolis for bacterial count reduction, this agrees with a previous study that found that Chlorhexidine was more effective than Propolis and there was enhanced antibacterial capacity with increasing concentrations. However, the results of this study disagreed with another research that deduced that diode Laser and Brazilian Propolis have equal effect as CHX in cavity disinfection. The results of this study can be justified by another study , which stated that diode Laser antibacterial mechanisms is due to thermal and photo disruptive effects that causes cell wall lethal damage; also it can be attributed to the greater depth of Laser radiation penetration into the dentin, surpassing the effect range of chemical disinfectants.
Under the current study operating conditions, utilizing non-thermal plasma has a distinct advantage for achieving total inactivation compared with the conventional procedures used in endodontic clinical applications in a short time interval.
Below is the link to the electronic supplementary material. Supplementary Material 1
|
InDEL instability in two different tumoral tissues and its forensic significance | 7239bf5e-6bfc-441b-9c2c-c8653f5f1cf3 | 11790770 | Forensic Medicine[mh] | Analysis of DNA polymorphisms in humans is a valuable tool for forensic identification, if there is a reference sample to compare with the sample in question. Where there is no reliable reference sample, it is necessary to find an alternative source such as an histopathology specimen. Some difficulties may arise when the accuracy of the reference sample is questionable with possible genetic variability, such as malignant tumors . This is because tumor tissues generally have mutations, which may affect genotype analysis. However, there are some situations where the use of tumor tissues may be the most advantageous option . The reason why STR loci are the first preferred method in forensic science laboratories is that there are approximately 1 million STR loci in the genome, and they show high polymorphism due to their different repeat numbers and content of these repeats and thus have high discrimination power . STRs are spread in non-coding regions of the genome and contain a higher mutation rate than other genomic regions. This is because DNA polymerase shifts during DNA replication and repair, causing the deletion or insertion of repeats in STRs. This change is defined as microsatellite instability (MSI) . Studies have shown that genetic instability is a very common phenomenon observed in many different tumors . It is known that tumor DNA harbors genetic changes not only in defined genes but also in repetitive DNA sequences. Loss of heterozygosity (LOH) and MSI are seen as genetic instability in many tumors . In addition to the mutative effects of tumors, formalin-fixed paraffin-embedded (FFPE) processes of archived tumoral tissues used in the analysis itself may cause degradation of the genetic material. The most important reason for this effect is that the tissues undergo chemical preparation processes that will lead to degradation before they are embedded in paraffin . InDels are one of the most common polymorphisms resulting from the insertion or deletion of one or more nucleotides in the human genome . InDels contain much lower mutation rates compared to STRs, and also have a shorter amplicon length (60–200 bp). This makes it possible to use InDels in degraded samples . In addition, it has been stated that InDels can be a potential genetic marker in forensic sciences due to their abundance in the genome and ease of analysis . The aim of this study is to observe whether the instability is seen in tumoral tissues at 36plex InDel Panel . For this purpose, InDel genetic profiles obtained from tumoral and non-tumoral tissues were compared. The determination of whether specific loci exhibit mutations or remain in their normal state can be made by contrasting the identified genetic changes in tumoral tissues with their corresponding normal counterparts. According to our knowledge, this is the first study to question the existence of genomic instability in InDels for forensic purposes using FFPE tissues. Sample collection A total of 47 cases, 26 of which were diagnosed as breast cancer and 21 as thyroid cancer, were included in the study. A total of 94 samples containing both tumoral and non-tumoral tissues, two paraffin blocks for each case, were selected and studied. As we were unable to reach these individuals during their active illness, we couldn’t take fresh biological tissues as a control group. The samples were collected from the archive of Istanbul University-Cerrahpaşa, Cerrahpaşa Faculty of Medicine, Department of Medical Pathology. Informed consent was obtained by contacting the owners of the samples. The ethical permission was approved by Clinical Research Ethics Evaluation Committee at İstanbul University-Cerrahpaşa, Cerrahpaşa Faculty of Medicine on March 2nd ,2021 (No. 17,853). The range values for the storage time of FFPE tissues were determined as 2–29 months (Median = 8 months) in cases with breast carcinoma and 1–38 months (Median = 19 months) in cases with thyroid carcinoma. In one case with breast tumor, the age, grade and storage period of the blocks could not be determined because the final report was not available. No significant difference was found between the breast and thyroid tumor groups in terms of block storage times. (Mann Whitney U = 334, p = 0.114). The age range of the cases was between 28 and 77 for breast cancer cases (mean = 53.1 ± 10.7), 28–63 (mean = 46.9 ± 9.8) for thyroid cases. Accordingly, the average age of the thyroid carcinoma patient group is statistically significantly lower than that of breast cancer cases (Mann Whitney U = 171.500; p = 0.045). Pathological diagnosis of all thyroid tissues was papillary thyroid carcinoma. Although different types of cancer were seen in breast tissues, most of them were invasive ductal carcinomas. Types of breast carcinomas are given in Table . The grades (breast) and thyroid types (thyroid) of the cases were recorded by reviewing their reports. Accordingly, 11 (44%) out of 26 breast cancer cases were classified as grade 2, and 14 (56%) were classified as grade 3. In one case, age, grade, and block storage period could not be determined because a report could not be obtained. For thyroid cancer cases, 1 (5.5%) was determined as Stage 1, 7 (38.9%) as Stage 2, and 10 (55.6%) as Stage 3. The stage could not be determined in the reports of 3 cases. Age and grade/Stage information for the cases are provided in Supplementary Table . Sample preparation and DNA extraction Sections of 2 μm thickness were prepared from paraffin blocks using a microtome. Slides were stained with hematoxylin-eosin dye using the standard method . Hematoxylin-eosin-stained slides were examined by the pathologist under a light microscope to distinguish paraffin blocks containing tumor and non-tumor tissue. DNA was extracted from both tumor and non-tumor tissue sections by using E.Z.N.A. Tissue DNA Kit (Omega Bio-Tek, Norcross, GA, USA). Before DNA isolation, tissues were deparaffinized using xylene and sequential alcohol solutions of decreasing concentration according to the kit protocol. DNA samples were quantified using the Quant-iT dsDNA High-Sensitivity (HS) Assay Kit (Invitrogen, CA, USA) by Qubit fluorometer (Invitrogen, CA, USA). PCR amplification In this study, 36-InDelplex panel developed by Filoğlu et al. was used for genotyping, and PCR was carried out according to the procedure of this study . In tumoral and the non-tumoral tissues, PCR was performed using 34 InDel loci located on autosomal chromosomes (rs34660708, rs34495360, rs2308135, rs2307789, rs2307521, rs2308112, rs2308163, rs1610919, rs2308137, rs16646, rs144389514, rs56168866, rs16671, rs33972805, rs25549, rs16722, rs2067304, rs140861207, rs1160981, rs4646006, rs6480, rs28369942, rs2307838, rs3062629, rs10590424, rs16458, rs10623496, rs1610937, rs16363, rs2067147, rs1160965, rs2307656, rs2308101, rs2308072, rs2067191), as well as the InDel locus on the Y chromosome (rs2032678) and the amelogenin locus (AMG-XY).These InDel loci are part of the 36-InDelplex panel developed by Filoğlu et al., demonstrating high polymorphism in the Turkish population and their applicability in forensic identification . PCR was performed using the SimpliAmp™ Thermal Cycler (Thermo Fisher Scientific, MA, USA) under the following conditions: initial denaturation at 95 °C for 11 min, 30 cycles of denaturation at 94 °C for 20 s and amplification at 62 °C for 3 min, followed by elongation at 60 °C for 1 hour. The components used in PCR are 4.2 µl Master Mix, 0.5 µl Taq Polymerase, 3 µl Primer Mix and 3 µl DNA, with a total volume of 10.7 µl. Complete profiles were obtained with a single PCR in most of the samples. However, PCR was repeated for a small number of samples to eliminate missing alleles, especially in some non-tumoral breast tissues. Specifically in some samples, DNA was initially diluted to a concentration of 1 ng/µL and added to the PCR mixture such as in Filoğlu et al’s study , and the full profile was obtained. However, in some other cases, profiles containing many missing alleles were obtained despite using up to 3ng/µL DNA input. In these cases, the PCR protocol was repeated by adding the DNA undiluted to the PCR mixture to obtain a complete profile. InDels genotyping Electrophoresis of PCR products was performed by using an eight capillary Applied Biosystems™ 3500 Genetic Analyzer (Thermo Fisher Scientific, MA, USA). First, a mixture containing 9.5 µl of Hi-Di™ Formamide and 0.5 µl of GeneScan™ LIZ 500 Size Standard for each sample was prepared and vortexed. Afterward, 10 µl of the mixture and 1 µl of PCR product per sample were loaded into the plate well. Samples were in the G-5 module of GS STR POP7 (1 ml) with 36 cm capillaries, using POP-7 polymer; the injection was carried out at 1.2 kV and 24 s, 60 °C for 30 minutes. GeneScan™ LIZ 500 Size Standard was used to verify the running conditions and detect InDel loci. After verifying the position of this standard for each sample, the locations of InDels were determined by comparing them with the standard reference(K562). The genotypes in the electropherogram were evaluated using GeneMapper Software v.5.0 (Thermo Fisher Scientific, CA, USA). InDel instability is defined as the emergence of a new allele or the replacement of an existing allele. For InDel markers, InDel instability is observed when the homozygous genotype observed in healthy tissue transforms into either heterozygosity or a distinct homozygous state in the tumor tissue. Furthermore, allelic deletion detected in the tumor tissue, into its corresponding non-tumoral tissue, was categorized as loss of heterozygosity. A total of 47 cases, 26 of which were diagnosed as breast cancer and 21 as thyroid cancer, were included in the study. A total of 94 samples containing both tumoral and non-tumoral tissues, two paraffin blocks for each case, were selected and studied. As we were unable to reach these individuals during their active illness, we couldn’t take fresh biological tissues as a control group. The samples were collected from the archive of Istanbul University-Cerrahpaşa, Cerrahpaşa Faculty of Medicine, Department of Medical Pathology. Informed consent was obtained by contacting the owners of the samples. The ethical permission was approved by Clinical Research Ethics Evaluation Committee at İstanbul University-Cerrahpaşa, Cerrahpaşa Faculty of Medicine on March 2nd ,2021 (No. 17,853). The range values for the storage time of FFPE tissues were determined as 2–29 months (Median = 8 months) in cases with breast carcinoma and 1–38 months (Median = 19 months) in cases with thyroid carcinoma. In one case with breast tumor, the age, grade and storage period of the blocks could not be determined because the final report was not available. No significant difference was found between the breast and thyroid tumor groups in terms of block storage times. (Mann Whitney U = 334, p = 0.114). The age range of the cases was between 28 and 77 for breast cancer cases (mean = 53.1 ± 10.7), 28–63 (mean = 46.9 ± 9.8) for thyroid cases. Accordingly, the average age of the thyroid carcinoma patient group is statistically significantly lower than that of breast cancer cases (Mann Whitney U = 171.500; p = 0.045). Pathological diagnosis of all thyroid tissues was papillary thyroid carcinoma. Although different types of cancer were seen in breast tissues, most of them were invasive ductal carcinomas. Types of breast carcinomas are given in Table . The grades (breast) and thyroid types (thyroid) of the cases were recorded by reviewing their reports. Accordingly, 11 (44%) out of 26 breast cancer cases were classified as grade 2, and 14 (56%) were classified as grade 3. In one case, age, grade, and block storage period could not be determined because a report could not be obtained. For thyroid cancer cases, 1 (5.5%) was determined as Stage 1, 7 (38.9%) as Stage 2, and 10 (55.6%) as Stage 3. The stage could not be determined in the reports of 3 cases. Age and grade/Stage information for the cases are provided in Supplementary Table . Sections of 2 μm thickness were prepared from paraffin blocks using a microtome. Slides were stained with hematoxylin-eosin dye using the standard method . Hematoxylin-eosin-stained slides were examined by the pathologist under a light microscope to distinguish paraffin blocks containing tumor and non-tumor tissue. DNA was extracted from both tumor and non-tumor tissue sections by using E.Z.N.A. Tissue DNA Kit (Omega Bio-Tek, Norcross, GA, USA). Before DNA isolation, tissues were deparaffinized using xylene and sequential alcohol solutions of decreasing concentration according to the kit protocol. DNA samples were quantified using the Quant-iT dsDNA High-Sensitivity (HS) Assay Kit (Invitrogen, CA, USA) by Qubit fluorometer (Invitrogen, CA, USA). In this study, 36-InDelplex panel developed by Filoğlu et al. was used for genotyping, and PCR was carried out according to the procedure of this study . In tumoral and the non-tumoral tissues, PCR was performed using 34 InDel loci located on autosomal chromosomes (rs34660708, rs34495360, rs2308135, rs2307789, rs2307521, rs2308112, rs2308163, rs1610919, rs2308137, rs16646, rs144389514, rs56168866, rs16671, rs33972805, rs25549, rs16722, rs2067304, rs140861207, rs1160981, rs4646006, rs6480, rs28369942, rs2307838, rs3062629, rs10590424, rs16458, rs10623496, rs1610937, rs16363, rs2067147, rs1160965, rs2307656, rs2308101, rs2308072, rs2067191), as well as the InDel locus on the Y chromosome (rs2032678) and the amelogenin locus (AMG-XY).These InDel loci are part of the 36-InDelplex panel developed by Filoğlu et al., demonstrating high polymorphism in the Turkish population and their applicability in forensic identification . PCR was performed using the SimpliAmp™ Thermal Cycler (Thermo Fisher Scientific, MA, USA) under the following conditions: initial denaturation at 95 °C for 11 min, 30 cycles of denaturation at 94 °C for 20 s and amplification at 62 °C for 3 min, followed by elongation at 60 °C for 1 hour. The components used in PCR are 4.2 µl Master Mix, 0.5 µl Taq Polymerase, 3 µl Primer Mix and 3 µl DNA, with a total volume of 10.7 µl. Complete profiles were obtained with a single PCR in most of the samples. However, PCR was repeated for a small number of samples to eliminate missing alleles, especially in some non-tumoral breast tissues. Specifically in some samples, DNA was initially diluted to a concentration of 1 ng/µL and added to the PCR mixture such as in Filoğlu et al’s study , and the full profile was obtained. However, in some other cases, profiles containing many missing alleles were obtained despite using up to 3ng/µL DNA input. In these cases, the PCR protocol was repeated by adding the DNA undiluted to the PCR mixture to obtain a complete profile. Electrophoresis of PCR products was performed by using an eight capillary Applied Biosystems™ 3500 Genetic Analyzer (Thermo Fisher Scientific, MA, USA). First, a mixture containing 9.5 µl of Hi-Di™ Formamide and 0.5 µl of GeneScan™ LIZ 500 Size Standard for each sample was prepared and vortexed. Afterward, 10 µl of the mixture and 1 µl of PCR product per sample were loaded into the plate well. Samples were in the G-5 module of GS STR POP7 (1 ml) with 36 cm capillaries, using POP-7 polymer; the injection was carried out at 1.2 kV and 24 s, 60 °C for 30 minutes. GeneScan™ LIZ 500 Size Standard was used to verify the running conditions and detect InDel loci. After verifying the position of this standard for each sample, the locations of InDels were determined by comparing them with the standard reference(K562). The genotypes in the electropherogram were evaluated using GeneMapper Software v.5.0 (Thermo Fisher Scientific, CA, USA). InDel instability is defined as the emergence of a new allele or the replacement of an existing allele. For InDel markers, InDel instability is observed when the homozygous genotype observed in healthy tissue transforms into either heterozygosity or a distinct homozygous state in the tumor tissue. Furthermore, allelic deletion detected in the tumor tissue, into its corresponding non-tumoral tissue, was categorized as loss of heterozygosity. Since the analysis kit consists of 36 loci, 936 loci in 26 breast cancer cases and 756 loci in 21 thyroid cancer cases were analyzed. If no allele peaks were observed at a locus, or if there was only one allele with a height not exceeding analysis thresholds (200 Relative Fluorescence Unit), loci were not included in the analysis. Profile could be obtained in all 47 cases. The number of loci that could not be analyzed after this evaluation is given in Table . As a result, 815 loci of breast and 685 loci of thyroid cancer cases could be compared. There was no statistically significant difference between the tissue group variables (tumor vs. non-tumoral block; χ2 = 4,2128, p > 0,05) and the cancer type variables (breast vs. thyroid; χ2 = 2,209, p > 0,05) in terms of degradation rates. For all 47 cases, the median value of cases with undetectable alleles, due to degradation, was determined as 1.5 when examining the distribution of 192 loci for each locus (Range 0–28). Accordingly, degradation-related undetectable cases are observed in only 15 out of 36 loci, with 3 or more cases, and 8 of these loci exhibit degradation in 10 or more cases. The numerical distribution of loci where 3 or more cases have undetectable alleles due to degradation is provided in Table . While the most significant allele loss is observed in ID13(rs144389514), there has been no allele loss due to degradation in loci ID26(rs2308163), ID3(rs2307521), ID16(rs16458), ID32(rs2308101), ID9(rs1610937), ID17(rs3062629), ID31(rs16671), ID27(rs2067147), ID15(rs16646), ID24(rs1610919), ID30(rs34495360), ID25(rs34660708), and ID29(rs25549). Mutation rates of loci for both tumor types The changes were observed in 75 of 815 loci in breast tissues (9,2%), of which %65.3 complete loss of heterozygosity, %20 of InDels instability, and %14,7 partial loss of heterozygosity. Whereas changes in thyroid cases were observed in 10 out of a total of 685 loci (%1,5). Distribution of changes were 3 (30%) complete, 3 (30%) partial loss of heterozygosity, and 4 (40%) InDels instability among these changes. In the point of tumor types, mutations causing genetic instability are shown in Table . The most changed loci are in order of frequency ID24, ID16 and ID26 loci. Examples of complete loss of heterozygosity, partial loss of heterozygosity and InDels instability were given in Fig. a, b and c, respectively. According to given rates at Table , the incidence of InDels mutations in breast cancer tissues is significantly higher than in thyroid cancer tissues (Chi-square = 40,4694; p < 0.001). On the other hand, there was no statistically significant difference between the two cancer/tissue types in terms of distribution according to mutation types (Chi-square = 4.647; p > 0.05). According to the Spearman correlation test results, no correlation was identified between the number of mutations detected in the cases and their ages ( r = 0.038, p = 0.800) or block storage times ( r = -0.045, p = 0.764). Furthermore, it was determined that the number of mutations did not exhibit a statistically significant difference between different grades in breast cancer and different stages in thyroid cancer (Mann Whitney U = 74, p = 0.893; Kruskal-Wallis = 571, p = 0.752, respectively). Efficiency of comparison on a case-by-case basis Mutational changes were observed 21 of 26 (80.8%) breast cancer cases whereas only 6 of 21 thyroid carcinoma expressed these changes (28.6%). Number of breast cancer cases with mutational changes was significantly higher than thyroid cancers (Fisher exact test; p < 0.001). Although the number of mutant loci was ranged from 1 to 10, 12 of 21 breast cancer cases had two or less than two mutations. On the other hand, the number of mutant loci was 3 in one of the six mutated thyroid cancer cases whereas 1 or 2 mutations were observed in the other five cases. Distribution of overall and detailed numbers of mutations on a case-by-case basis was given at Table and Supplementary Table . The changes were observed in 75 of 815 loci in breast tissues (9,2%), of which %65.3 complete loss of heterozygosity, %20 of InDels instability, and %14,7 partial loss of heterozygosity. Whereas changes in thyroid cases were observed in 10 out of a total of 685 loci (%1,5). Distribution of changes were 3 (30%) complete, 3 (30%) partial loss of heterozygosity, and 4 (40%) InDels instability among these changes. In the point of tumor types, mutations causing genetic instability are shown in Table . The most changed loci are in order of frequency ID24, ID16 and ID26 loci. Examples of complete loss of heterozygosity, partial loss of heterozygosity and InDels instability were given in Fig. a, b and c, respectively. According to given rates at Table , the incidence of InDels mutations in breast cancer tissues is significantly higher than in thyroid cancer tissues (Chi-square = 40,4694; p < 0.001). On the other hand, there was no statistically significant difference between the two cancer/tissue types in terms of distribution according to mutation types (Chi-square = 4.647; p > 0.05). According to the Spearman correlation test results, no correlation was identified between the number of mutations detected in the cases and their ages ( r = 0.038, p = 0.800) or block storage times ( r = -0.045, p = 0.764). Furthermore, it was determined that the number of mutations did not exhibit a statistically significant difference between different grades in breast cancer and different stages in thyroid cancer (Mann Whitney U = 74, p = 0.893; Kruskal-Wallis = 571, p = 0.752, respectively). Mutational changes were observed 21 of 26 (80.8%) breast cancer cases whereas only 6 of 21 thyroid carcinoma expressed these changes (28.6%). Number of breast cancer cases with mutational changes was significantly higher than thyroid cancers (Fisher exact test; p < 0.001). Although the number of mutant loci was ranged from 1 to 10, 12 of 21 breast cancer cases had two or less than two mutations. On the other hand, the number of mutant loci was 3 in one of the six mutated thyroid cancer cases whereas 1 or 2 mutations were observed in the other five cases. Distribution of overall and detailed numbers of mutations on a case-by-case basis was given at Table and Supplementary Table . InDels can be used as a genetic marker in forensic sciences . They are very common in the genome, provide successful results in degraded samples due to their small amplicon size, and have a lower mutation rate than STRs. The methods used for the analysis of STRs can also be used for InDels analysis. Therefore, InDels have been observed to be more expedient than STRs . Loss of heterozygosity is characterized by mutation of one allele followed by deletion of the remaining allele . In addition, the loss of heterozygosity also differs according to the amount of peak intensity. To be considered as complete loss of heterozygosity (cLOH), the peak intensity should be less than 0.5 and higher than 2, while partial loss of heterozygosity (pLOH) occurs when the peak size of an allele decreases by more than %50 . Effect of the paraffinization process on DNA analysis has been investigated in various studies before. In one recent example, Vitosevic showed that 3 gene regions, which are often used as reference genes in genetic analysis studies, can be amplified in paraffin blocks for up to 30 years . Indeed, the fixation and embedding process causes degradation of DNA due to its fragmentation and chemical modifications . These modifications may occur due to mechanisms such as tissue aging over time in the fixative, hydrolysis of DNA with formic acid, and formation of intramolecular methylene bridges . Covalent cross-links formed during fixation can interfere with extraction and downstream applications, impacting PCR, sequencing, and molecular analyses. Additionally, the fixation process often reduces overall DNA yield, posing challenges in high-throughput applications or when working with limited tissue samples. These limitations highlight the complexity and constraints associated with using FFPE tissues for DNA studies . In our study this degradation effect manifested itself as a randomly distributed reduction in peak sizes or complete loss of alleles in some of the InDel loci both in tumoral and non-tumoral blocks, as well. In many studies conducted with paraffin embedded blocks, it is reported that there is a decrease in the amount of DNA as well as in the quality of DNA . In order to cope with the disadvantages of using tissues embedded in paraffin blocks, procedures such as adjusting the coverage numbers, the amount of DNA template analyzed, pretreatment methodologies, optimizing proteinase K digestion conditions or simplifying DNA extraction procedures can be implemented . Kerick et al. showed that there was no difference between FFPE and snap frozen tissues in terms of detection of InDel mutations because of increasing the coverage levels to 40 × . In one study, they reported that successful results were obtained by adding 30 ng of DNA template to the PCR reaction . In our study, this problem is solved by increasing the amount of DNA. In certain samples, diluting DNA to 1 ng/µL, following Filoğlu’s study , yielded a complete profile, while in other cases, profiles with numerous missing alleles persisted even with 2 ng/µL or 3 ng/µL DNA. In these cases, the PCR and electrophoresis protocol was repeated by adding the DNA undiluted to the PCR mixture to obtain a complete profile. This adjustment proved effective because, even during fixation, certain DNA chains endure in an intact state. The increase of template DNA quantity enhances the likelihood of amplifying intact DNA chains, thereby achieving a more comprehensive and accurate profiling of the genetic material. In these cases, a complete profile could be obtained by adding DNA without dilution. As a result, an overall profile of 88.7% could be obtained in tumor and non-tumor tissues. From another perspective, although they show a certain degree of degradation, FFPE tissues are attractive tools for clinical studies because they are much easier to obtain than frozen tissues. However, this degradation effect may not cause a limitation as expected . Although Oh et al. obtained lower yielding and mapping results in FFPE samples compared to frozen samples in their exome sequencing studies, they showed that this did not have a statistically significant effect. Moreover, they showed that frozen samples could lead to higher off-target rate determinations compared to FFPEs, thus pointing to an advantageous aspect for FFPE tissues. They attribute this to DNA fragments that shorten with fragmentation and say that this will increase on-target coverage . This study’s primary aim is to determine whether the identification success of the InDel multiplex panel exhibits any difference between tumor tissues and normal tissues. In this retrospectively conducted study, the use of paraffin blocks as controls could have quickly demonstrated differences between normal and tumoral tissues. Indeed, in a study similar to ours, Oliveira and colleagues utilized tumor-free tissue within the tumor tissue block as a control . There was no significant difference between the degradation rates in terms of two tumor types we studied or whether the tissue contains tumoral or non-tumoral tissue. These ratios, which were between 6.2% and 7.5% in both cases (See Table ). Similarly in a study by Soo et al., small-sized DNA or RNA fragments were amplified successfully in formalin-fixed paraffin-embedded samples after they were kept for several years . Conversely, in a study by Guyard et al., they show that after 4 to 6 years the strong decrease in the amount of amplifiable DNA is due to fragmentation of DNA . In our study, there were medium-long waiting times distributed between 1 and 38 months compared to these studies, and more than 90% matching was possible (Supplementary Table ). As reported at " " section, the mutations were detected in only six of 21 thyroid cancer cases. Moreover, in these six cases, mutations were detected at only 1 or 2 loci. This finding suggests that archival FFPE tissues of thyroid tumor can be used in cases of forensic genetic identification. However, the presence of mutations in 21 of the 26 breast cancer cases examined suggests that paraffin blocks, in which this cancer type is detected, should be used carefully in forensic genetic identification. The most mutations observed in the ID24(rs1610919), ID26(rs2308163), and ID16 loci(rs16458), along with the prevalence of cLOH, results from these loci should be treated with suspicion. Particularly in malpractice cases, it is essential to acknowledge that such variations in these loci can be dismissed. However, as reported by Nam SK et al., depending on the nature of the case, it is still possible to make a reliable comparison in cases with a small number of mutations . In a study where Pereira et al. used 38 non-coding bi-allelic indel markers on different populations, they found random match probabilities (RMP) to be 1 in 10-17 billion even for the 25 most informative markers. Accordingly, even if an incomplete profile is obtained, the possibility of identification is high . Turajlic et al. reported that mutations at renal cell carcinomas are twice as much as all other cancer types and breast cancers have higher averages than thyroid cancers in terms of both InDel numbers and InDel ratio . Similarly, we observed higher instability rates in breast cancer cases than in thyroid cancer cases. On the other hand, we found that the number of mutations did not show a relationship with the grade for breast cancers or the stage for thyroid cancers. Moreover, no difference was detected according to the storage time of the blocks and the age of the cases. In this instance, it does not seem possible to say with certainty, type of malignancy may play an important role in the rate of mutational changes. Indeed, Wu et al. suggested that the number of InDel mutations which is described as ‘Tumor Mutational Burden (TMB)’ may differ between cancers and associated with prognosis and treatment response . Unlike this study, which took into account the coding regions, our study showed the effect of cancers on the analysis in terms of forensic identification, as it was performed on non-coding loci. An additional significant result indicated by our findings is that, when identification is required using an InDel panel from FFPE tissues, it is preferable to extract material from a non-tumor block rather than a tumor block. No similar study has been reported in the literature. In the closest study to ours, Oliveira et al. detected InDel mutations in the CDH1 gene that were not present in normal tissue in cases of Hereditary Diffuse Type Gastric Carcinoma . In their study, they dissected and compared both tumor and non-tumor tissue from the same FFPE block. Although their study focused on a gene that codes differently from ours, our findings underscore the preference for selecting these blocks whenever normal tissue is available. Tumorous FFPE tissues may be utilized in samples such as needle biopsy where no normal tissue block is present. It is essential to determine the mutation dynamics in cancer types for this to be possible. Cancerous transformation in tissues may adversely affect the success of forensic identification studies. It has been shown by previous studies that cancer can accompany disruptions in STR loci . For example, Peloso and his colleagues worked with archival FFPE tissues taken from 24 people with lung cancer as well as corresponding normal tissues obtained from lymph node sections of the same patients. Although allele drop-out was observed in at least one STR in 20 of 24 samples, allelic imbalance was observed in the STR loci of all samples. Additionally, a small portion of the samples showed loss of heterozygosity . Also, Anaian et al. observed DNA profiles from 12 gastric,12 breast and 10 colorectal formalin-fixed paraffin-embedded tissue (FFPET) samples, revealing 55 cases of partial loss of heterozygosity (pLOH), 15 cases of complete loss of heterozygosity (cLOH), and 13 cases of microsatellite instability (MSI) . On the other hand, other identification markers (e.g. SNPs, InDels) have been also tested in tumor samples . In the study of Tozzo et al., 61 (92.4%) of 66 frozen tumor samples (hepatic, gastric, breast, and colorectal cancer) were found to have at least one mutation in InDel loci . Majority of these mutations were LOHs of which 41.7% were only partial (pLOH) and MSI events constituted only 20.6% of all the mutations. These authors argued that because of these ratios, the use of STR loci for identification purposes would be more appropriate than InDel loci. On the contrary, in a study by Zhao et al., only pLOH and cLOH mutational events of InDels loci were observed in fresh tumoral tissue samples (colorectal and gastric cancer). The total mutation rate of InDels was 0.25% in tumoral tissues . Authors suggested that InDels might be more powerful than STR in source identification; unlike the results reported by Tozzo et al. Our results support the view that MSI is the least common type of InDel mutations in tumoral tissues. While only 22,35% of the mutations are of this type, the vast majority are LOH mutations. Moreover, in 27 of 47 cases, a mutation was detected in at least one of the loci compared to non-tumoral tissues (See Table ). A problem we are dealing with was more missing alleles in the profiles of the non-tumoral breast tissue samples which can be due to the high density of adipose tissue that leads to losses in terms of DNA yield. Similar to our study, in a study conducted by McDonough et al., it is stated that there is less DNA in non-tumoral breast tissues than the corresponding tumoral part . In the study conducted by Mee et al., it was observed that since non-tumoral breast tissue consists of almost completely fat, it has less DNA, RNA and protein amount compared to tumoral breast tissue . A limitation of our study is that it was carried out retrospectively on archival FFPE tissues. The degradation effect of paraffinization will be able to be determined more clearly in future studies using fresh biological tissues as the control group. In prospectively planned studies, the effects of prognostic variables regarding the tumor on both degradation and detection of mutations can be determined. The effectiveness of the panel can be increased by determining tumor-specific mutation distributions (Table ). As a result, complete profiles of 36 InDel loci were obtained in the majority of paraffinized tissues, both tumoral and non-tumoral. Multiple mutations were seen in some of the cases, while no mutations were found in almost half of the cases. In the vast majority of cases with mutations, only one or two mutations such as loss of heterozygosity, partial loss of heterozygosity and InDels instability were observed. These low InDel mutation rates compared to STR instability, make InDel analysis from paraffin blocks suitable for forensic genetic identification. However, researchers should keep in mind that there may be differences between the profiles of the tumoral tissues taken as reference and the actual case. In addition, by incorporating additional markers such as Single Nucleotide Polymorphisms (SNPs) and microhaplotypes with low mutation rates into the study alongside Indels, researchers can significantly enhance the discrimination power in identification processes. Low mutation rates reduce the likelihood of errors or inconsistencies in the analysis, ensuring the accuracy of the results over time. The combination of Indels, SNPs, and microhaplotypes creates a multi-marker approach that significantly boosts the discrimination power in identification processes. This increased power allows for more precise differentiation between individuals, particularly in forensic applications or population genetics studies. The 36-InDelplex panel was genotyped in 47 cases, 26 with breast cancer and 21 with thyroid cancer. Mutational changes were observed in 75 (9.2%) of 815 loci in breast tumor tissue and 10 (1.5%) out of 685 loci in thyroid. While the incidence of InDels mutations in breast tumor tissues was significantly higher than in thyroid tumor tissues, there was no statistically significant difference between tumor/non-tumor tissue type. Mutational changes were observed 21 of 26 (80.8%) breast cancer cases whereas at only 6 of 21 (28.6%) thyroid carcinoma cases. Low InDel mutation rates compared to STR instability, make InDel analysis from paraffin blocks suitable for forensic genetic identification. However, researchers should keep in mind that there may be differences between the profiles of the tumoral tissues taken as reference and the actual case. In addition, by incorporating additional markers such as SNPs and microhaplotypes with low mutation rates into the study alongside Indels, researchers can significantly enhance the discrimination power in identification processes. The reference samples should not only consist of tumor tissues, but also DNA should be obtained from the non-tumoral part of the paraffin block. Below is the link to the electronic supplementary material. Supplementary Material 1 (DOCX 24.1 KB) |
Systems biology and machine learning approaches identify drug targets in diabetic nephropathy | cddd8b84-0b9a-442f-947b-f76d4f6d734a | 8648918 | Pharmacology[mh] | Diabetic nephropathy (DN) is a major complication in diabetes mellitus and the leading cause of end-stage renal disease (ESRD). Despite the beneficial effects of current drugs such as angiotensin-converting enzyme inhibitors and angiotensin II receptor blockers, DN patients are still reaching ESRD . Therefore, it is critical to understand the molecular mechanisms of this disorder to develop more efficient therapeutic approaches. Systems biology with its quantitative and predictive viewpoints has provided a unique opportunity to explore the complex biological processes involved in the pathogenesis of chronic disorders , . It allows the generation of holistic maps of interactions between a variety of biomolecules involved in these processes. Considering the pivotal role of microRNAs (miRNAs) in the regulation of a bundle of functionally related genes , , we were motivated to study the network of DN-associated miRNAs with their targets. Although the role of individual miRNAs in DN has been previously explored , holistic evaluations have just been started . Valuable insights into the molecular mechanisms of complex disorders have been introduced using systems biology strategies, however, little progress has been made in the translation of this knowledge to the bedside. Although omics technologies, in line with advanced computational techniques, allowed the identification of lots of biomolecules with complex interactions in disease pathogenesis, the identification of appropriate therapeutic targets has remained elusive. To address this challenge, some investigators have proposed that the success of current FDA-approved drugs compared to many ingredients that failed during preclinical and clinical screenings can, at least partly, be attributed to the characteristics of their target proteins . Hence, these potential discriminating properties can be identified and exploited to predict novel drug targets (DT). Based on this assumption, several classic machine learning algorithms have been utilized for feature selection and DT prediction – . However, these studies suffer from several limitations such as ignoring the unequal frequency of DT and non-DT proteins, inappropriate machine performance measures, or unspecified details of utilized methods. We have here developed a next-generation machine learning method that considers high-level feature interactions and unbalanced DT/non-DT classes. This study aimed to predict novel targets for DN based on the holistic map of molecular pathogenesis. Several in silico and wet lab steps were pursued to identify the miRNA profile of the disorder and detect novel differentially expressed (DE) miRNAs in the cortex and medulla of diabetic kidneys. Moreover, miRNA-target interaction networks were inferred to identify central nodes and critical interactions. Pathway enrichment analysis also allowed the prediction of affected signaling pathways in this disorder. Next, to translate the findings of this study to clinical application, a high-performance machine learning framework, named "modified Group Method of Data Handling with Automatic Feature Selection (mGMDH-AFS)", was developed and validated for the prediction of DT in human proteome based on a variety of biochemical and network topology features. This classifier was then applied to candidate novel therapeutic targets in the constructed holistic map of DN. The design of this study is schematically presented in Fig. .
Diabetic nephropathy mouse model Male DBA/2 J mice, aged 6–10 weeks, were supplied from the Pasteur Institute of Iran (Tehran, Iran). All animal studies were performed according to the NIH guide for the care and use of laboratory animals . In addition, the study was conducted in compliance with the ARRIVE guidelines. All protocols were approved by the Isfahan University of Medical Sciences Ethics Committee (IR.MUI.MED.REC.1399.933). For five consecutive days, streptozotocin (STZ) was dissolved in sodium citrate buffer and intraperitoneally administered to pre-starved mice at the dose of 40 mg/kg (total amount: 200 mg/kg). Control mice received citrate buffer. Mice were supplied with 10% sucrose water during STZ injection and a few days after to avoid sudden hypoglycemia. As the mice with a lesser extent of diabetes usually do not show renal injury, one week after the final STZ injection, their non-fasting blood glucose was evaluated and those with blood glucose levels below 280 mg/dL were excluded from the study. Three months after the last dose of STZ, mice with the blood glucose range of 300–600 mg/dL were transferred to metabolic cages to collect 24-h urine. Urine albumin concentration was measured with an ELISA test (Exocell, Philadelphia, PA), and urine volume was used to calculate total albumin excretion. Urine-specific gravity, kidney weight, serum glucose, HbA1C, serum creatinine, and urea were also measured. Moreover, kidney tissues were sampled. Histopathologic parameters such as glomerular basement membrane thickening, mesangial matrix, increased mesangial cell proliferation, and diffuse mesangial sclerosis were assessed after Hematoxylin and Eosin (H&E) and Periodic Acid–Schiff (PAS) staining. For each pathology field, 20 serial glomeruli were evaluated, and the percentages of affected glomeruli were calculated. The Mann–Whitney U test was applied for statistical analysis. miRNA microarray The left kidneys were harvested, and the cortex and medulla were separated. The tissues were kept in RNAlater (Qiagen, Valencia, CA, USA) and stored at −70 °C until RNA extraction. The tissues were lysed in 1 mL QIAzol (Qiagen, Valencia, CA, USA) and homogenized with TissueLyser LT (Qiagen). Next, 250 µL chloroform (Merck, Darmstadt, Germany) was added and centrifuged at 12,000 rpm for 20 min at 4 °C after 15 min of incubation. An equal volume of cold ethanol (Merck) was added to the upper aqueous phase in a new tube and incubated at −20 °C overnight after which, the samples were centrifuged at 14,000 g for 45 min at 4 °C. To dislodge the pellet, 75% ethanol was added and centrifuged at 12,000 g for 15 min at 4 °C. The dried pellet was dissolved in double-distilled water. Using a BioPhotometer (Eppendorf, Hamburg, Germany), RNA concentration was measured at 260 nm. miRNA profiling was performed for the cortex and medulla of five DN and three control mice (total 16 samples). The quality of the microarray experiment was evaluated by the principal component analysis (PCA) and the hierarchical clustering. For hierarchical clustering, the correlation coefficient, and the average linkage methods, as the distance metric, were applied and the heat maps were plotted by ClusterMaker application of Cytoscape software version 3.2.0 . PCA was carried out with ggplot package of R . The miRNAs with logarithm to base two of fold change (log 2 FC) ≥ 0.5 or ≤ −0.5 were selected. Quantitative PCR Primers were designed with Gene Runner version 3.05 (Hastings Software Inc., Hastings, NY, USA) and Oligo version 7 (Molecular Biology Insights, Inc, USA) software. cDNA was synthesized using RevertAid First-Strand cDNA Synthesis Kit (Thermo Scientific, Vilnius, Lithuania) and PCR thermal cycler machine (Takara Bio, Shiga, Japan). The RNA samples, stem-loop specific primer, and double-distilled water were mixed and incubated at 75 °C for 5 min. The vials were immediately placed on ice, and ten mM dNTP mix, 5X reaction buffer, M-MuLV RT enzyme, and double-distilled water were added and spun briefly. Next, they were incubated at 42 °C for 60 min. The reaction was terminated by heating at 70 °C for 5 min. QuantiFast SYBR Green PCR kit (Takara Bio) and Rotor-Gene 6000 real-time PCR machine (Corbett, Sydney, Australia) were employed to assess the expression levels of 37 miRNAs as well as Snord70 and Snord68, as internal controls in the same samples used for the microarray experiment. The temperature profile consisted of an initial step of 95 °C for 10 min, followed by 40 cycles of 95 °C for 15 s and 60 °C for 1 min. The data were analyzed using REST 2009 software . MicroRNA target identification Experimentally validated targets of DE miRNAs were collected using miRTarBase database . Targets that were validated by strong evidence were selected. The validated targets were obtained from studies on murine samples for all miRNAs, except for mmu-miR-802-5p and mmu-miR-187-3p that did not have any mouse validated targets; thus, human targets with strong evidence were obtained. A list of genes with potential roles in DN was manually collected from the literature. TargetScan 6.2 and miRWalk databases were used for predicted, and validated miRNAs targeting these genes, respectively. The miRNAs that were broadly conserved among vertebrates and mammals were selected in TargetScan. Network construction and analysis Using CluePedia plugin version 2.1.7 of Cytoscape software, interaction networks for validated targets of DE miRNAs in the cortex and medulla were constructed by STRING resource . Confidence cut-off was set at 0.8; then, DE miRNAs were merged to the networks. The edges including activation, post-translational modification, binding, and inhibition were allowed to be shown. The topology of the networks was analyzed by NetworkAnalyzer , CytoNCA , and CentiScape applications of Cytoscape. Mann–Whitney U test was employed to compare the topology features using the SPSS statistical package, version 18.0 (SPSS Inc., Chicago, IL, USA). The modules were determined in the networks by the MCODE version 1.4.1 plugin of Cytoscape. Human protein–protein interactions were retrieved from HPRD version 9 and imported to the Cytoscape. Gene ontology and pathway enrichment analysis Pathway enrichment analysis was carried out using Cytoscape ClueGO plugin version 2.1.7 . In this analysis, Reactome and KEGG databases were selected for retrieving the data. In addition, the DAVID database was applied to collect pathways and biological processes from BioCarta and Gene Ontology (GO), respectively. P-value correction was performed and pathways and GO terms with the adjusted P -value ≤ 0.05 were selected. Drug target collection FDA-approved drugs and their target proteins were downloaded from the DrugBank database . Non-human targets were deleted. Collection of biochemical features Protein accession and annotation The UniProt ID and the official protein name for all human proteins were extracted from the UniProtKB/Swiss-Prot of the UniProt database . Using SPSS statistical package, the Chi-square test was employed to compare the biochemical features. Protein family Human proteins belonging to receptors, G-Couple Protein Receptors (GPCRs), nuclear hormone receptors, enzyme-linked receptors, tyrosine kinase receptors, serine/threonine receptors, ion channels, ligand, and voltage-gated ion channels, transporters, GTPases, ATPases, Phosphatase, and proteases were extracted from protein family of UniProt. In addition, human kinases were collected from the pkinfam data of UniProt. Post-translational modification Experimentally validated post-translational modifications (PTM) were extracted from the dbPTM database . Enzyme Enzymes were downloaded from the ENZYME database and categorized into oxidoreductase, transferase, hydrolase, lyase, isomerase, and ligase classes. Metabolic enzymes were extracted from Metabolic Enzyme Database . Transcription factor and cofactor The experimentally validated human transcription factors (TFs) list was extracted from the TFCheckpoint database . Transcription cofactors were collected from AnimalTFDB . Epigenetics regulator The epigenetics regulators including chromatin remodelers, histone modifiers, RNA/DNA modifiers, and scaffold proteins were extracted from EpiFactors . Transcriptional response to small molecules The list of up or down-regulated genes in the treatment of small molecules identified in the CMAP project was harvested from Enrichr . miRNA target All human genes identified as strongly validated miRNA targets were retrieved from miRTarBase . Mitochondrial protein Human proteins located in mitochondria were collected from miToCarta2.0 . Mutation A list of human mutated genes was downloaded from Online Mendelian Inheritance in Man (OMIM) . SNPs-trait association All SNP-trait associations with P -value ≤ 5 × 10 –8 were obtained from GWAS catalog v2.2.1 . SNPs mapped in non-coding and intragenic regions were deleted. Coding genes that were nearest to the SNPs were collected. Machine learning State-of-the-art Current machine learning methods including logistic regression (LR) , , , radial basis function (RBF) kernel support vector machine (SVM) , , generalized linear model (GLM) , and radial basis function network (RBFN) were utilized to predict potential drug targets. Discriminative features were selected in LR and GLM based on their statistical structures. Sequential Forward Selection (SFS) method was used in SVM and RBFN . The classifiers are briefly described below: (a) Logistic regression LR uses the following regression model for the prediction: [12pt]{minimal}
$$ ( {} ) = b0 + _{j = 1}^{{N_{f} }} {b_{j} x_{j} } + e$$ log p 1 - p = b 0 + ∑ j = 1 N f b j x j + e where p is the detection probability, and e denotes the binomial error term. Having fitted the model by tuning the parameters b j using the input features x j (N f is the number of features), each case with the estimated p ≥ 0.5 was classified as the DT, or non-DT otherwise. Upon normalization of the input features, those with low weights could be excluded (a.k.a., feature selection). (b) Support vector machine Hyperplanes are mainly used in SVM to separate data points of different classes . The data is transferred to higher dimensions using non-linear mappings (i.e., kernels). We used radial basis function kernels in this study. The method proposed by Wu and Wang was used to tune the radius of the RBF and the soft-margin parameter . Moreover, the SVM classifier was trained using sequential minimal optimization . (c) Generalized linear model GLM is a flexible linear regression that involves using a link function to relate the output of the model to the response variable. LR is one of the categories of GLM which employs the logit function i.e., log (p/(1−p)) as the linking function. GLM is suitable for the binomial model. Here, the Poisson linking function was utilized , . (d) Radial basis function neural networks RBF Networks encompass the following layers, the input as the entire features, a hidden layer with a non-linear RBF activation function, and a linear output layer with one node per category or class of data. Each output node calculates the score for the associated class and the class with the highest score is selected for each input sample. The RBF prototypes were estimated using the K-means clustering while the other network parameters were estimated using the Backpropagation algorithm . The mGMDH-AFS algorithm An mGMDH-AFS algorithm was developed based on inductive neural networks or Group Method of Data Handling (GMDH) that created more accessible models and provide more transparency , . In biological systems characterized by high dimensionality, it is crucial to perform feature selection to improve classification accuracy , . The GMDH algorithm was embedded with Particle Swarm Optimization (PSO), a population-based stochastic optimization algorithm , and a Relief-feature-weighting algorithm to estimate optimal model fitting. Briefly, categorical data were transferred to interval features using logistic regression , . The binary encoding was used for each categorical variable, and the logistic regression function parameters were estimated using iterative reweighted least-squares on the estimation set. It is, indeed, a form of non-linear data normalization. The I-RELIEF algorithm was employed prior to the classification. The weight of the features was iteratively estimated based on their capability to discriminate between neighboring patterns . Moreover, the oversampled training set was divided into estimation and validation sets to avoid over-fitting . The GMDH, proposed by Ivakhnenko was utilized in the current study. In this network, the pairwise interactions of each input feature (a.k.a., neurons) are calculated at each layer. The output of each neuron in the current layer is used as the input to the next layer, and the network is built layer by layer until no improvement is observed in the validation set (i.e., early stopping criterion). The overall structure of the algorithm is provided in Fig. . In this study, the top 10 neurons were selected at each layer. Instead of the traditional polynomial function widely used in the GMDH network, a matrix of nonlinear non-convex functions including exponential, sinusoid, and logarithmic forms was used to model the interaction. Thus, the PSO algorithm was used to estimate the parameters of the model rather than the least-square algorithm. Since the data is highly imbalanced, the Matthews correlation coefficient was used as the fitness function instead of the traditional RMSE. The output of the GMDH algorithm is a continuous variable ranging from zero to one. The optimal cut-off was then estimated on the training set using Receiver Operating Characteristic (ROC) curve . The MATLAB code is available online to interested readers: https://github.com/marateb/Drug-Targets-Classification . Validation The hold-out validation was used to assess the performance of the developed methods. The dataset was randomly split into two independent 70% training and 30% test sets . Also, four-fold cross-validation was employed for further performance assessment to guard against testing hypotheses suggested by the data (Type III errors) . The classifiers were assessed in terms of the performance indices such as sensitivity, specificity, precision, accuracy, and diagnostic odds ratio (DOR) whose definitions are given in Table . The Q-Cochran’s test and McNemar’s post-hoc test were used to identify whether the proposed system significantly outperformed the state-of-the-art. The Bonferroni correction was also applied for multiple comparisons and the adjusted P -values were then used for interpretation. The random permutation test was used to compare the performance of the mGMDH-AFS machine with real DT/non-DT classes and ten random sets whose class labels were randomly permuted. MATLAB version 8.6 (The MathWorks Inc., Natick, MA, USA) was used for offline processing.
Male DBA/2 J mice, aged 6–10 weeks, were supplied from the Pasteur Institute of Iran (Tehran, Iran). All animal studies were performed according to the NIH guide for the care and use of laboratory animals . In addition, the study was conducted in compliance with the ARRIVE guidelines. All protocols were approved by the Isfahan University of Medical Sciences Ethics Committee (IR.MUI.MED.REC.1399.933). For five consecutive days, streptozotocin (STZ) was dissolved in sodium citrate buffer and intraperitoneally administered to pre-starved mice at the dose of 40 mg/kg (total amount: 200 mg/kg). Control mice received citrate buffer. Mice were supplied with 10% sucrose water during STZ injection and a few days after to avoid sudden hypoglycemia. As the mice with a lesser extent of diabetes usually do not show renal injury, one week after the final STZ injection, their non-fasting blood glucose was evaluated and those with blood glucose levels below 280 mg/dL were excluded from the study. Three months after the last dose of STZ, mice with the blood glucose range of 300–600 mg/dL were transferred to metabolic cages to collect 24-h urine. Urine albumin concentration was measured with an ELISA test (Exocell, Philadelphia, PA), and urine volume was used to calculate total albumin excretion. Urine-specific gravity, kidney weight, serum glucose, HbA1C, serum creatinine, and urea were also measured. Moreover, kidney tissues were sampled. Histopathologic parameters such as glomerular basement membrane thickening, mesangial matrix, increased mesangial cell proliferation, and diffuse mesangial sclerosis were assessed after Hematoxylin and Eosin (H&E) and Periodic Acid–Schiff (PAS) staining. For each pathology field, 20 serial glomeruli were evaluated, and the percentages of affected glomeruli were calculated. The Mann–Whitney U test was applied for statistical analysis.
The left kidneys were harvested, and the cortex and medulla were separated. The tissues were kept in RNAlater (Qiagen, Valencia, CA, USA) and stored at −70 °C until RNA extraction. The tissues were lysed in 1 mL QIAzol (Qiagen, Valencia, CA, USA) and homogenized with TissueLyser LT (Qiagen). Next, 250 µL chloroform (Merck, Darmstadt, Germany) was added and centrifuged at 12,000 rpm for 20 min at 4 °C after 15 min of incubation. An equal volume of cold ethanol (Merck) was added to the upper aqueous phase in a new tube and incubated at −20 °C overnight after which, the samples were centrifuged at 14,000 g for 45 min at 4 °C. To dislodge the pellet, 75% ethanol was added and centrifuged at 12,000 g for 15 min at 4 °C. The dried pellet was dissolved in double-distilled water. Using a BioPhotometer (Eppendorf, Hamburg, Germany), RNA concentration was measured at 260 nm. miRNA profiling was performed for the cortex and medulla of five DN and three control mice (total 16 samples). The quality of the microarray experiment was evaluated by the principal component analysis (PCA) and the hierarchical clustering. For hierarchical clustering, the correlation coefficient, and the average linkage methods, as the distance metric, were applied and the heat maps were plotted by ClusterMaker application of Cytoscape software version 3.2.0 . PCA was carried out with ggplot package of R . The miRNAs with logarithm to base two of fold change (log 2 FC) ≥ 0.5 or ≤ −0.5 were selected.
Primers were designed with Gene Runner version 3.05 (Hastings Software Inc., Hastings, NY, USA) and Oligo version 7 (Molecular Biology Insights, Inc, USA) software. cDNA was synthesized using RevertAid First-Strand cDNA Synthesis Kit (Thermo Scientific, Vilnius, Lithuania) and PCR thermal cycler machine (Takara Bio, Shiga, Japan). The RNA samples, stem-loop specific primer, and double-distilled water were mixed and incubated at 75 °C for 5 min. The vials were immediately placed on ice, and ten mM dNTP mix, 5X reaction buffer, M-MuLV RT enzyme, and double-distilled water were added and spun briefly. Next, they were incubated at 42 °C for 60 min. The reaction was terminated by heating at 70 °C for 5 min. QuantiFast SYBR Green PCR kit (Takara Bio) and Rotor-Gene 6000 real-time PCR machine (Corbett, Sydney, Australia) were employed to assess the expression levels of 37 miRNAs as well as Snord70 and Snord68, as internal controls in the same samples used for the microarray experiment. The temperature profile consisted of an initial step of 95 °C for 10 min, followed by 40 cycles of 95 °C for 15 s and 60 °C for 1 min. The data were analyzed using REST 2009 software .
Experimentally validated targets of DE miRNAs were collected using miRTarBase database . Targets that were validated by strong evidence were selected. The validated targets were obtained from studies on murine samples for all miRNAs, except for mmu-miR-802-5p and mmu-miR-187-3p that did not have any mouse validated targets; thus, human targets with strong evidence were obtained. A list of genes with potential roles in DN was manually collected from the literature. TargetScan 6.2 and miRWalk databases were used for predicted, and validated miRNAs targeting these genes, respectively. The miRNAs that were broadly conserved among vertebrates and mammals were selected in TargetScan.
Using CluePedia plugin version 2.1.7 of Cytoscape software, interaction networks for validated targets of DE miRNAs in the cortex and medulla were constructed by STRING resource . Confidence cut-off was set at 0.8; then, DE miRNAs were merged to the networks. The edges including activation, post-translational modification, binding, and inhibition were allowed to be shown. The topology of the networks was analyzed by NetworkAnalyzer , CytoNCA , and CentiScape applications of Cytoscape. Mann–Whitney U test was employed to compare the topology features using the SPSS statistical package, version 18.0 (SPSS Inc., Chicago, IL, USA). The modules were determined in the networks by the MCODE version 1.4.1 plugin of Cytoscape. Human protein–protein interactions were retrieved from HPRD version 9 and imported to the Cytoscape.
Pathway enrichment analysis was carried out using Cytoscape ClueGO plugin version 2.1.7 . In this analysis, Reactome and KEGG databases were selected for retrieving the data. In addition, the DAVID database was applied to collect pathways and biological processes from BioCarta and Gene Ontology (GO), respectively. P-value correction was performed and pathways and GO terms with the adjusted P -value ≤ 0.05 were selected.
FDA-approved drugs and their target proteins were downloaded from the DrugBank database . Non-human targets were deleted.
Protein accession and annotation The UniProt ID and the official protein name for all human proteins were extracted from the UniProtKB/Swiss-Prot of the UniProt database . Using SPSS statistical package, the Chi-square test was employed to compare the biochemical features. Protein family Human proteins belonging to receptors, G-Couple Protein Receptors (GPCRs), nuclear hormone receptors, enzyme-linked receptors, tyrosine kinase receptors, serine/threonine receptors, ion channels, ligand, and voltage-gated ion channels, transporters, GTPases, ATPases, Phosphatase, and proteases were extracted from protein family of UniProt. In addition, human kinases were collected from the pkinfam data of UniProt. Post-translational modification Experimentally validated post-translational modifications (PTM) were extracted from the dbPTM database . Enzyme Enzymes were downloaded from the ENZYME database and categorized into oxidoreductase, transferase, hydrolase, lyase, isomerase, and ligase classes. Metabolic enzymes were extracted from Metabolic Enzyme Database . Transcription factor and cofactor The experimentally validated human transcription factors (TFs) list was extracted from the TFCheckpoint database . Transcription cofactors were collected from AnimalTFDB . Epigenetics regulator The epigenetics regulators including chromatin remodelers, histone modifiers, RNA/DNA modifiers, and scaffold proteins were extracted from EpiFactors . Transcriptional response to small molecules The list of up or down-regulated genes in the treatment of small molecules identified in the CMAP project was harvested from Enrichr . miRNA target All human genes identified as strongly validated miRNA targets were retrieved from miRTarBase . Mitochondrial protein Human proteins located in mitochondria were collected from miToCarta2.0 . Mutation A list of human mutated genes was downloaded from Online Mendelian Inheritance in Man (OMIM) . SNPs-trait association All SNP-trait associations with P -value ≤ 5 × 10 –8 were obtained from GWAS catalog v2.2.1 . SNPs mapped in non-coding and intragenic regions were deleted. Coding genes that were nearest to the SNPs were collected.
The UniProt ID and the official protein name for all human proteins were extracted from the UniProtKB/Swiss-Prot of the UniProt database . Using SPSS statistical package, the Chi-square test was employed to compare the biochemical features.
Human proteins belonging to receptors, G-Couple Protein Receptors (GPCRs), nuclear hormone receptors, enzyme-linked receptors, tyrosine kinase receptors, serine/threonine receptors, ion channels, ligand, and voltage-gated ion channels, transporters, GTPases, ATPases, Phosphatase, and proteases were extracted from protein family of UniProt. In addition, human kinases were collected from the pkinfam data of UniProt.
Experimentally validated post-translational modifications (PTM) were extracted from the dbPTM database .
Enzymes were downloaded from the ENZYME database and categorized into oxidoreductase, transferase, hydrolase, lyase, isomerase, and ligase classes. Metabolic enzymes were extracted from Metabolic Enzyme Database .
The experimentally validated human transcription factors (TFs) list was extracted from the TFCheckpoint database . Transcription cofactors were collected from AnimalTFDB .
The epigenetics regulators including chromatin remodelers, histone modifiers, RNA/DNA modifiers, and scaffold proteins were extracted from EpiFactors .
The list of up or down-regulated genes in the treatment of small molecules identified in the CMAP project was harvested from Enrichr .
All human genes identified as strongly validated miRNA targets were retrieved from miRTarBase .
Human proteins located in mitochondria were collected from miToCarta2.0 .
A list of human mutated genes was downloaded from Online Mendelian Inheritance in Man (OMIM) .
All SNP-trait associations with P -value ≤ 5 × 10 –8 were obtained from GWAS catalog v2.2.1 . SNPs mapped in non-coding and intragenic regions were deleted. Coding genes that were nearest to the SNPs were collected.
State-of-the-art Current machine learning methods including logistic regression (LR) , , , radial basis function (RBF) kernel support vector machine (SVM) , , generalized linear model (GLM) , and radial basis function network (RBFN) were utilized to predict potential drug targets. Discriminative features were selected in LR and GLM based on their statistical structures. Sequential Forward Selection (SFS) method was used in SVM and RBFN . The classifiers are briefly described below: (a) Logistic regression LR uses the following regression model for the prediction: [12pt]{minimal}
$$ ( {} ) = b0 + _{j = 1}^{{N_{f} }} {b_{j} x_{j} } + e$$ log p 1 - p = b 0 + ∑ j = 1 N f b j x j + e where p is the detection probability, and e denotes the binomial error term. Having fitted the model by tuning the parameters b j using the input features x j (N f is the number of features), each case with the estimated p ≥ 0.5 was classified as the DT, or non-DT otherwise. Upon normalization of the input features, those with low weights could be excluded (a.k.a., feature selection). (b) Support vector machine Hyperplanes are mainly used in SVM to separate data points of different classes . The data is transferred to higher dimensions using non-linear mappings (i.e., kernels). We used radial basis function kernels in this study. The method proposed by Wu and Wang was used to tune the radius of the RBF and the soft-margin parameter . Moreover, the SVM classifier was trained using sequential minimal optimization . (c) Generalized linear model GLM is a flexible linear regression that involves using a link function to relate the output of the model to the response variable. LR is one of the categories of GLM which employs the logit function i.e., log (p/(1−p)) as the linking function. GLM is suitable for the binomial model. Here, the Poisson linking function was utilized , . (d) Radial basis function neural networks RBF Networks encompass the following layers, the input as the entire features, a hidden layer with a non-linear RBF activation function, and a linear output layer with one node per category or class of data. Each output node calculates the score for the associated class and the class with the highest score is selected for each input sample. The RBF prototypes were estimated using the K-means clustering while the other network parameters were estimated using the Backpropagation algorithm . The mGMDH-AFS algorithm An mGMDH-AFS algorithm was developed based on inductive neural networks or Group Method of Data Handling (GMDH) that created more accessible models and provide more transparency , . In biological systems characterized by high dimensionality, it is crucial to perform feature selection to improve classification accuracy , . The GMDH algorithm was embedded with Particle Swarm Optimization (PSO), a population-based stochastic optimization algorithm , and a Relief-feature-weighting algorithm to estimate optimal model fitting. Briefly, categorical data were transferred to interval features using logistic regression , . The binary encoding was used for each categorical variable, and the logistic regression function parameters were estimated using iterative reweighted least-squares on the estimation set. It is, indeed, a form of non-linear data normalization. The I-RELIEF algorithm was employed prior to the classification. The weight of the features was iteratively estimated based on their capability to discriminate between neighboring patterns . Moreover, the oversampled training set was divided into estimation and validation sets to avoid over-fitting . The GMDH, proposed by Ivakhnenko was utilized in the current study. In this network, the pairwise interactions of each input feature (a.k.a., neurons) are calculated at each layer. The output of each neuron in the current layer is used as the input to the next layer, and the network is built layer by layer until no improvement is observed in the validation set (i.e., early stopping criterion). The overall structure of the algorithm is provided in Fig. . In this study, the top 10 neurons were selected at each layer. Instead of the traditional polynomial function widely used in the GMDH network, a matrix of nonlinear non-convex functions including exponential, sinusoid, and logarithmic forms was used to model the interaction. Thus, the PSO algorithm was used to estimate the parameters of the model rather than the least-square algorithm. Since the data is highly imbalanced, the Matthews correlation coefficient was used as the fitness function instead of the traditional RMSE. The output of the GMDH algorithm is a continuous variable ranging from zero to one. The optimal cut-off was then estimated on the training set using Receiver Operating Characteristic (ROC) curve . The MATLAB code is available online to interested readers: https://github.com/marateb/Drug-Targets-Classification . Validation The hold-out validation was used to assess the performance of the developed methods. The dataset was randomly split into two independent 70% training and 30% test sets . Also, four-fold cross-validation was employed for further performance assessment to guard against testing hypotheses suggested by the data (Type III errors) . The classifiers were assessed in terms of the performance indices such as sensitivity, specificity, precision, accuracy, and diagnostic odds ratio (DOR) whose definitions are given in Table . The Q-Cochran’s test and McNemar’s post-hoc test were used to identify whether the proposed system significantly outperformed the state-of-the-art. The Bonferroni correction was also applied for multiple comparisons and the adjusted P -values were then used for interpretation. The random permutation test was used to compare the performance of the mGMDH-AFS machine with real DT/non-DT classes and ten random sets whose class labels were randomly permuted. MATLAB version 8.6 (The MathWorks Inc., Natick, MA, USA) was used for offline processing.
Current machine learning methods including logistic regression (LR) , , , radial basis function (RBF) kernel support vector machine (SVM) , , generalized linear model (GLM) , and radial basis function network (RBFN) were utilized to predict potential drug targets. Discriminative features were selected in LR and GLM based on their statistical structures. Sequential Forward Selection (SFS) method was used in SVM and RBFN . The classifiers are briefly described below: (a) Logistic regression LR uses the following regression model for the prediction: [12pt]{minimal}
$$ ( {} ) = b0 + _{j = 1}^{{N_{f} }} {b_{j} x_{j} } + e$$ log p 1 - p = b 0 + ∑ j = 1 N f b j x j + e where p is the detection probability, and e denotes the binomial error term. Having fitted the model by tuning the parameters b j using the input features x j (N f is the number of features), each case with the estimated p ≥ 0.5 was classified as the DT, or non-DT otherwise. Upon normalization of the input features, those with low weights could be excluded (a.k.a., feature selection). (b) Support vector machine Hyperplanes are mainly used in SVM to separate data points of different classes . The data is transferred to higher dimensions using non-linear mappings (i.e., kernels). We used radial basis function kernels in this study. The method proposed by Wu and Wang was used to tune the radius of the RBF and the soft-margin parameter . Moreover, the SVM classifier was trained using sequential minimal optimization . (c) Generalized linear model GLM is a flexible linear regression that involves using a link function to relate the output of the model to the response variable. LR is one of the categories of GLM which employs the logit function i.e., log (p/(1−p)) as the linking function. GLM is suitable for the binomial model. Here, the Poisson linking function was utilized , . (d) Radial basis function neural networks RBF Networks encompass the following layers, the input as the entire features, a hidden layer with a non-linear RBF activation function, and a linear output layer with one node per category or class of data. Each output node calculates the score for the associated class and the class with the highest score is selected for each input sample. The RBF prototypes were estimated using the K-means clustering while the other network parameters were estimated using the Backpropagation algorithm .
An mGMDH-AFS algorithm was developed based on inductive neural networks or Group Method of Data Handling (GMDH) that created more accessible models and provide more transparency , . In biological systems characterized by high dimensionality, it is crucial to perform feature selection to improve classification accuracy , . The GMDH algorithm was embedded with Particle Swarm Optimization (PSO), a population-based stochastic optimization algorithm , and a Relief-feature-weighting algorithm to estimate optimal model fitting. Briefly, categorical data were transferred to interval features using logistic regression , . The binary encoding was used for each categorical variable, and the logistic regression function parameters were estimated using iterative reweighted least-squares on the estimation set. It is, indeed, a form of non-linear data normalization. The I-RELIEF algorithm was employed prior to the classification. The weight of the features was iteratively estimated based on their capability to discriminate between neighboring patterns . Moreover, the oversampled training set was divided into estimation and validation sets to avoid over-fitting . The GMDH, proposed by Ivakhnenko was utilized in the current study. In this network, the pairwise interactions of each input feature (a.k.a., neurons) are calculated at each layer. The output of each neuron in the current layer is used as the input to the next layer, and the network is built layer by layer until no improvement is observed in the validation set (i.e., early stopping criterion). The overall structure of the algorithm is provided in Fig. . In this study, the top 10 neurons were selected at each layer. Instead of the traditional polynomial function widely used in the GMDH network, a matrix of nonlinear non-convex functions including exponential, sinusoid, and logarithmic forms was used to model the interaction. Thus, the PSO algorithm was used to estimate the parameters of the model rather than the least-square algorithm. Since the data is highly imbalanced, the Matthews correlation coefficient was used as the fitness function instead of the traditional RMSE. The output of the GMDH algorithm is a continuous variable ranging from zero to one. The optimal cut-off was then estimated on the training set using Receiver Operating Characteristic (ROC) curve . The MATLAB code is available online to interested readers: https://github.com/marateb/Drug-Targets-Classification .
The hold-out validation was used to assess the performance of the developed methods. The dataset was randomly split into two independent 70% training and 30% test sets . Also, four-fold cross-validation was employed for further performance assessment to guard against testing hypotheses suggested by the data (Type III errors) . The classifiers were assessed in terms of the performance indices such as sensitivity, specificity, precision, accuracy, and diagnostic odds ratio (DOR) whose definitions are given in Table . The Q-Cochran’s test and McNemar’s post-hoc test were used to identify whether the proposed system significantly outperformed the state-of-the-art. The Bonferroni correction was also applied for multiple comparisons and the adjusted P -values were then used for interpretation. The random permutation test was used to compare the performance of the mGMDH-AFS machine with real DT/non-DT classes and ten random sets whose class labels were randomly permuted. MATLAB version 8.6 (The MathWorks Inc., Natick, MA, USA) was used for offline processing.
A combination of computational and experimental methods was employed to identify the miRNA profile in DN To explore DN pathogenesis, a mouse model of STZ-induced DN was established and after three months was validated using different functional (Fig. a) and histopathological (Fig. b-e) assessments. For constructing a holistic map of DN, we started with the profile of miRNA related to this disease as these molecules target functionally related genes so each variably expressed miRNA can be a clue to identify a group of related altered genes and functions , . miRNA microarray was performed on the cortex and medulla samples, separately, and the quality of microarray data was confirmed using unsupervised hierarchical clustering and PCA (Fig. a). We identified 7 and 12 miRNAs with |logFC|≥ 0.5 in cortex and medulla, respectively (Fig. b). To propose further miRNAs, a list of genes with a documented role in DN was provided (Table ). Using TargetScan and miRWalk databases, the predicted and validated miRNAs targeting the DN-associated genes were chosen, respectively (Table ). Based on the microarray, miRNA prediction, and validation, a total of 37 miRNAs were considered to be potentially related to DN (Fig. c). The alternations in the expression of these candidate miRNAs were examined in cortex and medulla samples by quantitative PCR (qPCR). Despite several optimizations, a reliable quantification was not achieved for 8 miRNAs (mmu-miR-711, mmu-miR-592-3p, mmu-miR-186-5p, mmu-miR-495-3p, mmu-miR-1192, mmu-miR-377-3p, mmu-miR-27b-3p, and mmu-miR-146b-5p) due to low or undetectable expression in the kidney or unavoidable technical problems. Among the remaining 29 miRNAs, qPCR data demonstrated the differential expression of 13 and 6 miRNAs in the cortex and medulla, respectively (Fig. d). A holistic miRNA-target interaction map was constructed, and key functions were inferred To investigate the role of differentially expressed miRNAs, the validated targets were identified as encompassing 108 and 56 genes for cortex and medulla, respectively (Table ). Then, the interactions between these genes and their targeting miRNAs were mapped (Fig. a, b). As expected for biological networks , both cortex and medulla networks followed a power law distribution. Graph theory measures such as degree, betweenness centrality, and closeness centrality were determined to identify central nodes in the networks. The top genes in terms of these centrality parameters were assumed as central (Table ). Interestingly, the critical role of the majority of the central nodes such as Hif1a, Vegfa, Sirt1, and Foxo1 has been shown in previous studies – . Similarly, there are experimental supports for the association of DN and miR-29, miR-34, miR-21, and miR-451, which we identified as central miRNAs – . This finding is in agreement with the concept that central network nodes drive critical functions . To identify key interactions in the cortex and medulla networks, modules were determined as sub-graphs of dense interactions (Fig. c). Two modules were related to the extra-cellular matrix aligned with DN histopathology , . Notably, miR-29, a key player in DN , regulates all elements of these two modules. Another module is related to the epigenetic control of Sirt1 on Foxo3 and Stat3, which are shown to be associated with DN pathogenesis – . Pathway enrichment analysis was performed to identify the signaling pathways associated with the miRNA validated targets. Using KEGG and Reactome databases, forty and eleven inter-connected pathways were enriched for cortex (Fig. a) and medulla (Fig. b), respectively. Further pathways, as well as GO biological process, were also identified using BioCarta and GO consortium for cortex (Table ) and medulla (Table ). The role of most enriched pathways, including TGFB, FGFR, EGFR, Notch, and hypoxia signaling has been shown in DN – . This analysis was validated by pathway enrichment analysis for ten random gene lists of similar sizes which yielded no or a few unrelated signaling pathways (data not shown). A novel high-performance machine learning method was developed to predict potential drug targets One of the main objectives of this research was to translate the findings on DN molecular pathogenesis into clinical applications. Thus, a machine learning approach was developed to predict which role player molecules could be suitable therapeutic targets. We followed the hypothesis that FDA-approved drug targets have unique properties compared to other proteins and these characteristics can be used to propose novel drug targets. Sixty-five biochemical features (Table ) were determined which were harvested from several databases, for all human proteins (#20,132). Moreover, 23 network topology features (Table ) were determined for the proteins in the human interactome obtained from HPRD (#9226). Based on DrugBank, 1443 proteins were then determined to be the targets of at least one FDA-approved drug. These proteins were considered as the DT group while the remaining 18,689 proteins were assigned to the non-DT class. Although the frequency or median of most biochemical (Fig. ) and topology (Fig. ) features are statistically different between DT and non-DT proteins, extensive distribution overlaps can be observed between these two protein classes. To assess the predictive value of these features for discrimination between DT and non-DT, different standard machine learning algorithms, namely logistic regression (LG), radial basis function kernel support vector machine (RBF-SVM), generalized linear model (GLM), and radial basis function network (RBFN) were utilized. The hold-out method measurements showed that the performance of none of the exploited methods was satisfying, neither with biochemical nor topology features (Fig. ). The failure of classical machines could be due to the unique nature of the data such as the considerable imbalanced size of the groups, significant overlap in the distribution of features in the two protein classes, and high dimensionality. The mentioned limitations were addressed by developing mGMDH-AFS, a high-performance machine learning method that considers high-level feature interactions. The performance of this novel tool was acceptable and significantly superior to the other standard machines, as revealed by hold-out validation (Fig. ). The prediction power of mGMDH-AFS was further evaluated by fourfold cross-validation (Fig. ). While topology features alone were not predictive for discrimination between DT and non-DT classes, biochemical features and the combination of biochemical and topology features were informative for the classification. The core topology features selected by mGMDH-AFS were included network degree, betweenness, and closeness. Also, being an enzyme, a receptor, an ion channel, or having post-translational modifications (PTM) were amongst the main biochemical features. To further assess the functionality of mGMDH-AFS, all human proteins were randomly labeled as 1 and 0 with the ratio of 1/0 the same as DT/non-DT classes. This procedure was repeated ten times to generate ten sets of randomly allocated proteins. As expected, the machine performance was significantly better with the real set compared to these random sets, indicating a reliable classification (Fig. ). After validating the mGMDH-AFS performance for all human proteins, this classifier was applied to predict potential DT in the constructed cortex and medulla networks. The algorithm calculated the probability of being DT for each protein in the networks. Using DrugBank to identify currently approved targets in these networks, an ROC curve analysis was carried out to evaluate machine performance (Fig. a). The top predicted drug targets for both cortex and medulla are Agtr1a, Egfr, Kit, Celsr3, Clic5, Tgfbr3, Acvr1b (Fig. b). Interestingly, the machine predicted Angiotensin II receptor, a well-known current drug target in DN, as the best candidate in the medulla and the 3rd best one in the cortex, supporting the validity of the developed approach. Some experimental supports were also found for other proposed targets such as EGF receptor and protein kinase C isoforms in previous studies on DN , .
To explore DN pathogenesis, a mouse model of STZ-induced DN was established and after three months was validated using different functional (Fig. a) and histopathological (Fig. b-e) assessments. For constructing a holistic map of DN, we started with the profile of miRNA related to this disease as these molecules target functionally related genes so each variably expressed miRNA can be a clue to identify a group of related altered genes and functions , . miRNA microarray was performed on the cortex and medulla samples, separately, and the quality of microarray data was confirmed using unsupervised hierarchical clustering and PCA (Fig. a). We identified 7 and 12 miRNAs with |logFC|≥ 0.5 in cortex and medulla, respectively (Fig. b). To propose further miRNAs, a list of genes with a documented role in DN was provided (Table ). Using TargetScan and miRWalk databases, the predicted and validated miRNAs targeting the DN-associated genes were chosen, respectively (Table ). Based on the microarray, miRNA prediction, and validation, a total of 37 miRNAs were considered to be potentially related to DN (Fig. c). The alternations in the expression of these candidate miRNAs were examined in cortex and medulla samples by quantitative PCR (qPCR). Despite several optimizations, a reliable quantification was not achieved for 8 miRNAs (mmu-miR-711, mmu-miR-592-3p, mmu-miR-186-5p, mmu-miR-495-3p, mmu-miR-1192, mmu-miR-377-3p, mmu-miR-27b-3p, and mmu-miR-146b-5p) due to low or undetectable expression in the kidney or unavoidable technical problems. Among the remaining 29 miRNAs, qPCR data demonstrated the differential expression of 13 and 6 miRNAs in the cortex and medulla, respectively (Fig. d).
To investigate the role of differentially expressed miRNAs, the validated targets were identified as encompassing 108 and 56 genes for cortex and medulla, respectively (Table ). Then, the interactions between these genes and their targeting miRNAs were mapped (Fig. a, b). As expected for biological networks , both cortex and medulla networks followed a power law distribution. Graph theory measures such as degree, betweenness centrality, and closeness centrality were determined to identify central nodes in the networks. The top genes in terms of these centrality parameters were assumed as central (Table ). Interestingly, the critical role of the majority of the central nodes such as Hif1a, Vegfa, Sirt1, and Foxo1 has been shown in previous studies – . Similarly, there are experimental supports for the association of DN and miR-29, miR-34, miR-21, and miR-451, which we identified as central miRNAs – . This finding is in agreement with the concept that central network nodes drive critical functions . To identify key interactions in the cortex and medulla networks, modules were determined as sub-graphs of dense interactions (Fig. c). Two modules were related to the extra-cellular matrix aligned with DN histopathology , . Notably, miR-29, a key player in DN , regulates all elements of these two modules. Another module is related to the epigenetic control of Sirt1 on Foxo3 and Stat3, which are shown to be associated with DN pathogenesis – . Pathway enrichment analysis was performed to identify the signaling pathways associated with the miRNA validated targets. Using KEGG and Reactome databases, forty and eleven inter-connected pathways were enriched for cortex (Fig. a) and medulla (Fig. b), respectively. Further pathways, as well as GO biological process, were also identified using BioCarta and GO consortium for cortex (Table ) and medulla (Table ). The role of most enriched pathways, including TGFB, FGFR, EGFR, Notch, and hypoxia signaling has been shown in DN – . This analysis was validated by pathway enrichment analysis for ten random gene lists of similar sizes which yielded no or a few unrelated signaling pathways (data not shown).
One of the main objectives of this research was to translate the findings on DN molecular pathogenesis into clinical applications. Thus, a machine learning approach was developed to predict which role player molecules could be suitable therapeutic targets. We followed the hypothesis that FDA-approved drug targets have unique properties compared to other proteins and these characteristics can be used to propose novel drug targets. Sixty-five biochemical features (Table ) were determined which were harvested from several databases, for all human proteins (#20,132). Moreover, 23 network topology features (Table ) were determined for the proteins in the human interactome obtained from HPRD (#9226). Based on DrugBank, 1443 proteins were then determined to be the targets of at least one FDA-approved drug. These proteins were considered as the DT group while the remaining 18,689 proteins were assigned to the non-DT class. Although the frequency or median of most biochemical (Fig. ) and topology (Fig. ) features are statistically different between DT and non-DT proteins, extensive distribution overlaps can be observed between these two protein classes. To assess the predictive value of these features for discrimination between DT and non-DT, different standard machine learning algorithms, namely logistic regression (LG), radial basis function kernel support vector machine (RBF-SVM), generalized linear model (GLM), and radial basis function network (RBFN) were utilized. The hold-out method measurements showed that the performance of none of the exploited methods was satisfying, neither with biochemical nor topology features (Fig. ). The failure of classical machines could be due to the unique nature of the data such as the considerable imbalanced size of the groups, significant overlap in the distribution of features in the two protein classes, and high dimensionality. The mentioned limitations were addressed by developing mGMDH-AFS, a high-performance machine learning method that considers high-level feature interactions. The performance of this novel tool was acceptable and significantly superior to the other standard machines, as revealed by hold-out validation (Fig. ). The prediction power of mGMDH-AFS was further evaluated by fourfold cross-validation (Fig. ). While topology features alone were not predictive for discrimination between DT and non-DT classes, biochemical features and the combination of biochemical and topology features were informative for the classification. The core topology features selected by mGMDH-AFS were included network degree, betweenness, and closeness. Also, being an enzyme, a receptor, an ion channel, or having post-translational modifications (PTM) were amongst the main biochemical features. To further assess the functionality of mGMDH-AFS, all human proteins were randomly labeled as 1 and 0 with the ratio of 1/0 the same as DT/non-DT classes. This procedure was repeated ten times to generate ten sets of randomly allocated proteins. As expected, the machine performance was significantly better with the real set compared to these random sets, indicating a reliable classification (Fig. ). After validating the mGMDH-AFS performance for all human proteins, this classifier was applied to predict potential DT in the constructed cortex and medulla networks. The algorithm calculated the probability of being DT for each protein in the networks. Using DrugBank to identify currently approved targets in these networks, an ROC curve analysis was carried out to evaluate machine performance (Fig. a). The top predicted drug targets for both cortex and medulla are Agtr1a, Egfr, Kit, Celsr3, Clic5, Tgfbr3, Acvr1b (Fig. b). Interestingly, the machine predicted Angiotensin II receptor, a well-known current drug target in DN, as the best candidate in the medulla and the 3rd best one in the cortex, supporting the validity of the developed approach. Some experimental supports were also found for other proposed targets such as EGF receptor and protein kinase C isoforms in previous studies on DN , .
Despite considerable efforts, diabetic patients progress to end-stage renal disease at alarming proportions, necessitating further investigations to design more efficient therapeutic approaches . Numerous studies have shown the role of individual elements in the pathogenesis of DN but systematic evaluations have rarely been attempted. In this study, miRNA-targets interaction maps were established in DN using system biology approaches. The central nodes, modules, and critical signaling pathways were determined. To translate the research findings to clinical applications, an innovative high-performance classifier was also developed to predict novel therapeutic targets in DN. The selection of miRNAs as the starting point for investigating DN molecular pathogenesis was based on the concept that they target functionally related genes. Indeed, each variably expressed miRNA can be a clue to identify a group of related altered functions , . Microarray profiling and bioinformatics predictions were employed in parallel to identify some miRNAs for further validation with qPCR. Among the identified DE miRNAs, the roles of miR-34a, miR-29b, miR-21a, miR145a, and miR-451a have been extensively shown in DN , , , , , indicating the validity of our approach. However, to the best of our knowledge, miR-802-5p, 182-5p, 210-3p, 31-5p, and 106a-5p are for the first time attributed to DN in the current study. One of the advantages of the present study is separate miRNA profiling for the cortex and medulla. Remarkably, the PCA plot demonstrated that the effect of anatomical area on miRNA expression was higher than the impact of the disease state. Although miR-802-5p and miR-34a are overexpressed in both cortex and medulla of DN mice, the other identified miRNAs are differential in either cortex or medulla. This finding is in line with recent studies revealing diverse mRNA and miRNA expression profiles for different anatomical parts of the normal kidney , . To provide evidence about the role of the identified miRNAs, their targets were collected and miRNA-target interaction networks were constructed. The constructed networks allowed a global systematic view towards the collaborations between miRNAs and their interactions with corresponding targets. Considering the complexity of biological interaction networks, it is critical to identify the elements with the highest influence on the outcome. Therefore, we performed topology analysis to determine the central nodes in the networks, as it has been previously shown that they drive key signaling pathways . Interestingly, the in vivo knockdown of miR-29, one of the central miRNAs in the cortex network, has been shown to halt the progression of DN . Notably, in the cortex network, this miRNA regulates two modules composed of extracellular matrix (ECM) elements. The identification of two modules associated with ECM in the cortex network is aligned with the histopathologic finding of fibrosis in the cortex. The role of DE miRNAs was further explored by performing pathway enrichment analysis to determine the signaling pathways associated with miRNA targets. Most of the enriched pathways, including TGFB, EGFR, and Notch signaling pathways were experimentally shown to be associated with DN – . Using this analysis, we could also predict novel pathways whose roles in DN remain to be confirmed in future studies. NGF signaling pathway which is related to diabetic neuropathy was amongst the enriched pathways . Similarly, the platelet degranulation pathway was detected as a potential role player in DN. Previous studies have demonstrated the role of this pathway in some pro-fibrotic disorders such as idiopathic pulmonary fibrosis and myelofibrosis , . The findings of this research were translated to clinical applications. It is hypothesized that the current FDA-approved drugs affect proteins with distinctive properties and these features can potentially be used to introduce novel drugs. Although some studies have shown that current DTs have unique features , , the construction of classifiers has remained a major challenge due to the overlap of feature distributions. To distinguish these two protein classes, some previous investigators have employed common machine learning approaches including LR, SVM, GLM, and RBFN , , , . In this study, however, these methods did not lead to satisfying results. This discrepancy could be assigned to the fact that these studies used equal size DT and non-DT classes to increase machine performance, or assessed the outputs with limited indices. To address the limitations of current classifiers, a high-performance machine learning method was developed. Unlike other methods, we considered the original highly imbalanced datasets for classification. However, learning from the imbalanced data is challenging . In this algorithm, a cost function other than the traditional mean-square-error metrics was utilized to avoid learning bias toward the majority class. The fitness function of the Matthews correlation coefficient was used whose robustness in imbalanced datasets has been proven , . In addition to the initial feature selection using the Relief algorithm, different interactions of the selected input features were considered during GMDH deep learning procedure. Indeed, biological systems are very complex and the nonlinear interactions of the non-redundant features could improve the performance of their classification systems . Based on different assessments, the performance of mGMDH-AFS is superior to that of the state-of-the-art approaches, suggesting it as a promising approach for therapeutic target prediction in complex disorders. After several validation steps, the developed machine learning algorithm was employed to analyze the cortex and medulla networks. It provided well-balanced diagnostic accuracy rates and resulted in a list of novel promising therapeutic targets for DN, which can be assessed in upcoming investigations. Among the top-ranked candidates in the cortex or medulla are Agtr1a, Egfr, Clic5, and Prkce. Interestingly, Agtr1a is the target of angiotensin receptor blockers which are currently in the market for DN treatment . It has been also shown that the inhibition of Egfr or Prkc isoforms by small molecules can prevent the progression of DN , . In conclusion, a combination of experimental and computational methods was exploited to generate a holistic map of DN and introduce novel therapeutic targets. The limitation of this work is the restriction of experimental data to miRNA profiling. As a future perspective, we plan to integrate other omics layers into the constructed networks to achieve more accurate insights. Notably, the proposed approach for drug target prediction could also be employed for other complex disorders as well.
Supplementary Tables. Supplementary Figures.
|
Bioactivity of Eugenol: A Potential Antibiotic Adjuvant with Minimal Ecotoxicological Impact | e10c4966-538c-4a03-99b1-29390a4efdf9 | 11241589 | Microbiology[mh] | Historically, natural products and their analogues have played a significant role in pharmacotherapy, particularly in treating infectious diseases . These products possess diverse active components, enabling them to target various bacterial strains, thus facilitating combinational therapy through multiple modes of action . Terpenes, derived from the mevalonate metabolic pathway, constitute a significant natural product family with diverse activities that have been extensively researched . They can kill or hinder microorganism growth depending on their concentration, despite their low water solubility . Monoterpenes’ antimicrobial efficiency varies based on their solubility and hydrogen bond ability, but it is not significantly impacted by the presence of double bonds or cyclic moieties in their structure . Aromatic compounds with alcohol groups, like eugenol (EUG), have strong inhibitory activity against microorganisms . EUG, classified as a phenylpropanoid, is a derivative of allyl chain-substituted guaiacol and it is biosynthesized by plants from phenylalanine . EUG exhibits weak acidic properties and slight solubility in water, while being readily soluble in organic solvents . It is typically derived from the natural essential oils found in plants belonging to the Lamiaceae , Lauraceae , Myrtaceae , and Myristicaceae families. EUG is notably abundant in clove oil ( Syzygium aromaticum ), serving as its principal constituent , but it can be synthetically manufactured too . This terpenoid has exhibited a range of beneficial properties, including antimicrobial activity against numerous human pathogens, encompassing a broad spectrum of Gram-positive and Gram-negative bacteria, fungi, and various parasites . Its potential as a new antibiotic against resistant bacteria is attributed to its natural origin and multifaceted mechanisms of action . Harnessing secondary plant metabolites, like EUG, as antimicrobial and/or adjuvant presents a cost-effective and innovative approach to developing novel strategies against antimicrobial resistance (AMR). Notably, very few instances of plant-based chemical resistance have been documented to date . A key approach to combatting AMR involves combining commercial antibiotics (ABXs) with substances such as adjuvants, some which are natural products, like monoterpenoids, to reduce the minimum inhibitory concentration (MIC) of the former. This established strategy is effective against resilient pathogens . When two substances interact, they can exhibit synergistic, additive, or antagonistic effects . Synergy occurs when the combined effect of two compounds surpasses the sum of their individual effects, with a lower dosage and toxicity, while antagonism is when a substance hinders the effect of another. On the other hand, addition happens when the effect of the mixture equals the adding of the individual effects. Synergy is the expected behavior when combining a commercial ABX and an adjuvant, particularly against resistant microorganisms . EUG was identified as Generally Recognized As Safe (GRAS) by prominent regulatory bodies such as the Food and Drug Administration (FDA) , the World Health Organization (WHO), and the Council of Europe . This recognition would underscore its safety for both human health and the environment , but might jeopardize future economical investment as a plausible adjuvant. Moreover, its safety profile has facilitated its application in pharmaceutical preparations and as a flavoring agent, stabilizer, antioxidant, antiseptic, and anesthetic in dental procedures . Analyzing soil and water ecotoxicity of new antimicrobials and their adjuvants is vital due to their significant environmental and health risks. Research highlights the role of environmental pollutants, including antimicrobials, in spreading AMR . By-products resulting from the modification or the degradation of antimicrobials have been identified in various environmental sources, emphasizing the need for monitoring . Antimicrobials in the environment could be toxic for algae and plants and inhibit microbial growth, affecting microbial diversity and thus negatively impacting soil enzymatic and metabolic activities . Therefore, assessing the ecotoxicity of new antimicrobials in soil and water is essential to understand their effects on ecosystems, microbial communities, and human health. Such analyses enable the risk assessment, mitigation strategy development, and protection of environmental and human well-being. Assessments of EUG’s antimicrobial and/or adjuvant activity should be accompanied by ecotoxicity studies, which are very scarce, addressing potential adverse effects on aquatic and terrestrial ecosystems and human and animal health, in line with the One Health strategy promoted by the WHO . Given these considerations, the objective of this research was doubled: (a) To assess the antimicrobial activity of EUG alone and its ability to enhance the bacterial inhibitory effect of commercial ABXs against clinically relevant Gram-positive and Gram-negative bacteria to find the synergistic combinations (in which the EUG would be the adjuvant) lowering MIC and thus allowing the reduction of ABXs’ doses. (b) To quantify the environmental impact of EUG by employing terrestrial and aquatic individual indicators alongside genetically sequenced microbial communities sourced from soil and river environments to obtain a more realistic environmental perspective of its potential use as an ABX adjuvant. By examining both the antimicrobial efficacy and environmental impact of EUG, this research provides new insights into its dual potential, enhancing antibiotic effectiveness while ensuring ecological safety.
2.1. Assessment of Eugenol Antimicrobial Activity EUG was tested against 14 Gram-positive ( Bacillus subtilis , Listeria monocytogenes , Streptococcus agalactiae , and Staphylococcus aureus ) and Gram-negative ( Acinetobacter baumannii , Enterococcus faecalis , Escherichia coli , Klebsiella aerogenes , Klebsiella pneumoniae , Pasteurella aerogenes , Pseudomonas aeruginosa , Salmonella enterica, and Serratia marcescens ) bacterial strains. Their MIC results are given in . EUG could not be tested against P. mirabilis since the DMSO concentration required to dissolve it was toxic for these bacteria, whereas it was not for the other bacteria , as mentioned in . Our study, that reports microdilution MIC values for the first time for Gram-positive bacteria such as L. monocytogenes , S. agalactiae, and S. aureus , established MICs of 1000 μg/mL (see ) for these strains. We compared our MIC values with those observed for different strains of the same bacteria using different methods. For example, the MIC for non-resistant strains of S. aureus determined by Walsh et al. varied from 106 to 1590 μg/mL, while those observed by Gallucci et al. and Hammer and Heel were approximately 33,430 μg/mL and from 127.2 to 8480 μg/mL, respectively. Perugini Biasi-Garbin et al. analyzed the MIC of several strains of S. agalactiae and found values ranging from 1325 to 5300 μg/mL, which are close to ours. The reported MIC values for L. monocytogenes varied from 67 to 1024 μg/mL according to different authors , some of which are also close to our values. For B. subtilis , no reported values for the isolated compound were found. Previous research suggested that EUG disrupts bacterial membranes, increasing their permeability . Additionally, EUG might inhibit ATPase. This could alter the efflux pump activity, especially in L. monocytogenes and S. agalactiae , resulting in the higher intracellular concentration of the ABX. Both in vivo and in vitro studies demonstrate reduced L. monocytogenes virulence upon EUG treatment , highlighting its potential as an adjuvant to conventional antibiotics, against L. monocytogenes and possibly other pathogens. Interestingly, L. monocytogenes strains did not develop resistance to EUG after exposure to sub-inhibitory concentrations . Concerning Gram-negative bacteria, reported MIC values for the same strains as ours using the macro- or microdilution methods were found for E. coli , P. aeruginosa , S. enterica, and S. marcescens . In E. coli , the MIC values exhibit considerable variations not only with our results, but also in the studies cited in . However, for S. enterica and S. marcescens, they are closer to our MIC values. In the case of P. aeruginosa, no comparison can be made because the value mentioned by the authors is not precisely determined and above 1000 or 2000 μg/mL . For the remaining bacteria, no values for the same strains were found with the macro- or microdilution methods, so this comparison was performed for different strains. A. baumannii ’s MIC ranged from 90.5 to 304 μg/mL and 318 μg/mL , slightly lower than ours. E. faecalis ’ MIC was 1060 μg/mL in the study conducted by Hammer and Heel , half of our value. Lastly, no EUG MIC was found for P. aerogenes and K. aerogenes . Our study was thus the first one to provide information about the inhibitory effect of EUG on these species. The variation of MIC values of Gram-negative bacteria is probably due to differences in the complexity of their outer membrane. Within the group of Gram-negative bacteria, the lowest MIC values were obtained for A. baumannii , K. aerogenes , K. pneumoniae , and P. aerogenes (500 μg/mL, ). These low values suggest a potential specificity of EUG against these bacteria. In the case of A. baumannii , consistently with the studies of Karumathil et al. , this could be attributed mainly to an inhibition of the AdeABC efflux pump. In K. pneumoniae , EUG was shown to induce membranous lipid peroxidation resulting from the production of ROS , leading to membrane disruption. ROS production caused by EUG might be specific to this bacterium, leading to lower MIC values. Except for P. mirabilis , for which experiments were not feasible (see for more information), the MIC values for the other Gram-negative bacteria exceeded 1000 μg/mL . Regarding E. faecalis , Hammer and Heel observed significant membrane disruption, while Morgaan et al. attributed EUG’s mode of action on E. coli to its lipophilic nature, causing bacterial membrane lipid disintegration, leading to energy depletion and cellular component leakage. Additionally, ATPase inhibition was noted in E. coli exposed to EUG . For P. aeruginosa , membrane disruption was also evident , although some studies suggested that EUG’s antimicrobial activity might prolong the lag phase of bacterial growth at sub-inhibitory concentrations, possibly due to interactions with primary non-vital bacterial targets . Aljuwayd et al. and Zhao et al. investigated the effects of EUG on S. Typhimurium ( S. enterica ) in vivo, observing the destruction of fimbriae, and reductions in adhesion molecules, and virulence factors. Aljuwayd et al. observed decreased pathogenicity and cell counts in chickens with EUG pretreatment, while Zhao et al. reported similar findings and noted EUG’s ability to maintain its intestinal barrier function and Tight Junctions (TJs) protein stability in mice. No specific information regarding the mode of action was available for S. marcescens and P. mirabilis . 2.2. Eugenol Combined with Commercial Antibiotics The checkboard assay revealed promising interactions between EUG and the conventional ABXs ampicillin (AMP), amoxicillin (AMO), gentamycin sulfate (GTM), streptomycin sulfate (STM), erythromycin (ERY), tetracycline hydrochloride (TC), and chloramphenicol (CHL). In general terms, 52% (9 out 17) of the experiments conducted on Gram-positive bacteria resulted in an ABX MIC reduction ≥ 75%, while this behavior was observed for only 26% (8 out of 30) of the tests on Gram-negative bacteria, . Six of these interactions were found to be synergistic: GTM-EUG- L. monocytogenes , GTM-EUG- S. aureus , CHL-EUG- A. baumannii , STM-EUG- S. enterica , TC-EUG- S. marcescens, and CHL-EUG- S. marcescens . All exhibited fractional inhibitory concentration (ΣFIC) values equal or below 0.50, calculated as described in . On the other hand, only two combinations resulted as antagonistic, which were STM-EUG- A. baumannii and CHL-EUG- K. aerogenes , both Gram-negative species. Although most of the interactions were additive, it must be highlighted that seven out of these additive combinations, AMP, AMO and TC-EUG- L. monocytogenes or GTM and CHL-EUG- S. agalactiae , STM-EUG- S. aureus, and CHL-EUG- K. pneumoniae , resulted in a pronounced decrease in the ABX dose (more than 75%, ). Additionally, a comprehensive kinetic growth study was conducted on the six synergistic interactions, as shown in , to unravel the behavior of each microorganism over a 24 h period under controlled conditions. It was observed that the combined administration of commercial ABXs and EUG, both at their MIC comb (see for a definition of MIC comb ) led to a complete inhibition of bacterial growth across all instances. These graphs also depict the total growth inhibition in experiments when natural and commercial ABXs were individually administered at their respective MIC alone (see for a definition of MIC alone ) and the non-inhibition when tested alone but at their respective MIC comb . However, even at subinhibitory concentrations of both the ABX and the natural product, some interesting facts can be extracted. The most pronounced effect observed in the graphs is related to the prolongation of the lag-phase. The lag-phase when the ABX was tested alone at its MIC comb was elongated for all experiments except for GTM-EUG- L. monocytogenes ( a) and CHL-EUG- A. baumannii ( c). However, when EUG was tested in these same circumstances, the lag-phase was modified only for S. enterica ( d). This fact was previously observed by Silva et al. on P. aeruginosa when EUG was tested at subinhibitory concentrations. They suggested that this behavior could be attributed to interactions between it and primary non-vital target sites of bacteria. 2.2.1. Aminoglycosides The precise synergistic mechanism between EUG and aminoglycosides remains unclear. However, previous studies suggest that the damage to the bacterial cell envelope induced by EUG may facilitate the entry of aminoglycosides (which use active transport systems to enter bacterial cells and inhibit protein synthesis ), thereby inhibiting bacterial growth at lower doses, albeit with variations depending on the microorganism . This mechanism would be especially efficient on Gram-positive bacteria, as observed for STM and GTM when acting together with EUG on S. agalactiae , L. monocytogenes, and S. aureus , , as shown in this work and previous research . For them, five of the six tested combinations resulted in an ABX MIC reduction ≥ 75%, while for the Gram-negative bacteria, this trend was observed only in two out of six. Within this group of bacteria, the most significant reduction in the aminoglycoside dosage was seen for STM against S. enterica , also displaying synergy ( and d), as previously demonstrated by Liu et al. , who noted the antibiofilm action of EUG in combination with STM. Additionally, while not synergistic but additive, EUG demonstrated substantial reductions in the aminoglycoside dosage when combined with STM against S. agalactiae , S. aureus, and K. pneumonia , as well as with GTM against S. agalactiae . This positions aminoglycosides as the antibiotic class most affected by their combination with EUG. Regarding the antagonistic interaction observed when EUG was combined with STM on A. baumannii , EUG might somehow alter the membrane potential and thus lead to a reduction in the active transport of STM into the cell, resulting in the observed antagonistic effect, . 2.2.2. Beta-Lactams Aminopenicillins are relatively small and hydrophilic molecules that use porins as essential entry points in these bacteria . Beta-lactams inhibit the activity of penicillin-binding proteins (PBPs), crucial for peptidoglycan cross-linking during cell wall synthesis, weakening the cell wall and altering the osmotic balance . This combined with EUG’s membrane-disrupting effects might lead to cell lysis . As expected, this effect was stronger in Gram-positive bacteria since the latter has a less complex envelop. Indeed, half of the tested interactions, versus 1 out of 10 for the Gram-negative bacteria, resulted in an ABX MIC reduction ≥ 75%. Although AMO and AMP did not exhibit synergistic interactions with EUG, they approached a value of ΣFIC = 0.5 for L. monocytogenes , showing a significant reduction in the ABX dose, although in an additive manner. Previous studies also found that EUG, when combined with AMP and AMO, led to an ABX dose reduction in a synergistic or additive manner on different strains of E. coli , P. aeruginosa , and S. Typhimurium or with AMO on E. faecalis . Our results may be especially significant for bacteria prone to developing resistance to beta-lactams through efflux pumps because the co-administration of ABX and EUG, which is known to inhibit efflux pumps activity, might result in a high intracellular ABX concentration . 2.2.3. Amphenicols The mechanism of action of CHL involves the reversible inhibition of bacterial protein synthesis by binding to the 50S subunit of the bacterial ribosome . It is a broad-spectrum antibiotic with high liposolubility, allowing it to enter bacteria through passive diffusion . The synergistic behavior observed for EUG when combined with CHL on A. baumannii ( and c) could involve the inhibition of the AdeABC efflux pump by EUG without increasing the membrane permeability . This could lead to the increased retention of CHL inside the cell, enhancing growth inhibition. Similarly, synergy between EUG and CHL was observed in this study against S. marcescens ( and f), proposing a possible shared mode of action involving efflux pump inhibition. Although there is limited research available on the remaining interactions in the selected microorganisms, EUG was found to decrease CHL MIC also in E. coli , P. aeruginosa, and S. enterica , as observed in . Unfortunately, we have no explanation concerning the antagonistic effect of CHL + EUG on K. aerogenes , since nothing is known about the mode of action of EUG in this species. 2.2.4. Tetracyclines TC is a broad-spectrum antibiotic that inhibits bacterial protein synthesis by binding to the 30S subunit of the bacterial ribosome, specifically at the A site, blocking the attachment of aminoacyl-tRNA molecules . This inhibition prevents the elongation of the polypeptide chain during translation. TC can also interfere with other bacterial processes, including essential metabolic pathways and cell membrane integrity, contributing to its bacteriostatic effect . The mechanism underlying the combined action of TC and EUG remains uncertain, although membrane disruption appears to be a plausible explanation, specifically in S. marcescens and L. monocytogenes , for which an ABX MIC reduction ≥ 75% was achieved ( and e). While there is limited literature on the combination of TC with EUG against these bacteria, a decrease in the ABX MIC for E. coli , P. aeruginosa , and S. Typhimurium has been previously reported . However, we did not investigate these combinations due to their initially low MIC (see . for detailed information). 2.2.5. Macrolides Macrolides, such as ERY, enter the cells via passive diffusion, inhibit protein chain elongation, and are effective against a broad spectrum of bacteria . When combined with EUG, ERY demonstrated MIC reductions across most tested bacteria, as previously determined on S. enterica , E. coli , and S. Typhimurium . Given EUG’s known modifications to the cell envelope, the mechanism of action of the EUG-ERY combination may be similar to that proposed for TC and EUG. 2.3. Eugenol Ecotoxicity The impact of EUG on soil and water environments is not extensively documented. EUG is reported to have effective insecticidal properties and it was approved by the Environmental Protection Agency (EPA) in the USA due to its minimum risk on mammals . In this report, some toxicity data are given, mostly on mammals, but there are also data (very few) referring to invertebrates that will be commented on later. To draw more comprehensive conclusions about EUG ecotoxicity, we have selected different soil and water non-target organisms together with microbiota extracted from water and soil, on which the metabolic effect of EUG was assessed. 2.3.1. Water Ecotoxicity of Eugenol Water ecotoxicity was analyzed through three indicators: Vibrio fischeri , Daphnia magna, and a complete water bacterial community, in which a comprehensive analysis of growth and metabolites’ intake was carried out. Average Well Color Development (AWCD) of water microbial populations and Community-Level Physiological Profiling (CLPP) after treatment with eugenol. The taxonomic categorization, spanning from the kingdom to species, was derived through the analysis of 16S rRNA sequences obtained from our river water samples. Across the levels of species, genus, family, order, class, phylum, and kingdom, the proportion of total reads ranged from 94.13% to 99.35%, while for species, it constituted 79.00%, . This microbial profile of the water sample is representative of a fluvial environment in the Mediterranean area and similar to that reported by other studies . Microbial water communities were exposed to EUG for a 7 day-period on the Biolog Ecoplates ® ( a). According to the statistical analysis, only after 48 h of exposure were significant differences ( p < 0.01) found between the control and an EUG concentration of 1000 mg/L. For the lowest concentrations (0.1, 10, and 100 mg/L), no significant decrease in the ability to degrade carbon sources was observed, since all values were similar to those of the control group. The microbial CLPP is understood as the metabolic capacity for degrading different carbon sources after toxic exposure . This method assesses the whole impact of toxic substances on these communities’ metabolism overall or by the group of metabolites. The changes in the CLPP allow us to identify the impact of EUG on the capacity of water microorganisms to metabolize the 31 most frequent organic carbon sources . For a better understanding of the metabolic changes, these carbon sources were grouped into polymers, amino acids, amines/amides, carbohydrates, and carboxylic and acetic acids . These results are shown in and . The metabolic ability of the water microbial communities is barely affected by EUG with respect to the control. As observed when analyzing the AWCD, only the highest concentration of EUG substantially changed the metabolic profile of these microorganisms. In fact, statistical differences ( p < 0.01) were detected between the control and 1000 mg/L for four of the five groups of metabolites (carbohydrates, polymers, amino acids, and amines/amides) from 48, 96, 120, and also 120 h, respectively. For the amino acids group, there was also significant differences between the control and 100 mg/L of EUG from 96 h. For the acids, no statistical difference was detected at any time for any of the EUG doses tested. When comparing the effects of EUG and commercial ABXs (the ones used in the antimicrobial experiments) on water microbiota, our EUG data revealed a weaker impact on AWCD. EUG and commercial ABXs caused a significant reduction in AWCD only at 1000 mg/L, but GTM, CHL, and TC also induced significant decreases at 100 mg/L . Regarding CLPP, EUG showed substantial metabolic differences compared to the control at 1000 mg/L for all metabolic groups except amino acids, which exhibited differences at 100 mg/L. In contrast, commercial ABXs had more pronounced effects on the microbial metabolism at all tested concentrations (from 0.1 to 1000 mg/L), with few exceptions such as AMP and CHL at lower concentrations . Our findings suggest that bacterial communities with high taxonomic diversity, like those in our samples, are less affected by EUG exposure compared to commercial ABXs . This could be attributed to antibiotic-degrading resistance mechanisms present in some community members, such as Pseudomonas sp. which is present in our sample , which may degrade EUG more efficiently than commercial ABXs. Consequently, while some species may be impacted, they may be replaced by resistant ones, resulting in no overall impact on population growth, as measured by AWCD, except at very high concentrations. Effect of EUG on water non-target organisms: Vibrio fischeri and Daphnia magna. A bioluminescence inhibition assay on V. fischeri and its dose–response curve revealed toxicity for these Gram-negative bacteria, as can be appreciated in a. The experiment’s significance was high ( p < 0.0001) and both EC 10 and EC 50 values were measured for the 30 min exposure test, being 0.708 mg/L (CI: 0.533–0.906) and 8.778 mg/L (CI: 7.454–10.370), respectively. The D. magna immobilization test dose–response curve can be observed in b. The results of the 24 h exposure test, with a high level of significance ( p < 0.0001), revealed EC 10 and EC 50 values of 0.824 mg/L (CI: 0.496–1.098) and 1.963 (CI: 1.571–2.444) mg/L, respectively. V. fischeri and D. magna are both valuable indicators of ecotoxicity, with the former’s bioluminescence enhancing its sensitivity to environmental stressors . Lal et al. and Gueretz et al. found lower EC 50 values for EUG in these organisms compared to our study, but their analysis lacked statistical rigor and employed fewer concentrations. In our study, with five doses and statistical analysis, slightly higher EC 50 values were found for both organisms, aligning closely with the findings of Baker and Grant (2018) for V. fischeri . Regardless of the reference values chosen, EUG demonstrates significantly lower EC 50 values compared to the ABXs examined in our study. For instance, EC 50 values for AMO ranged from 150 mg/L to 4000 mg/L , for AMP from 163 mg/L to 1056 mg/L , for TC from 6.70 mg/L to 173.8 mg/L , and for CHL from 20.68 mg/L to 1086 mg/L in immobilization tests lasting 24–48 h. The highest EC 50 value recorded was for GTM (>10,000 mg/L) , while the lowest was 8.21 mg/L for STM . No acute ecotoxicity was observed for ERY on V. fischeri . In the case of D. magna , EC 50 values for AMO ranged from >1000 mg/L to 6950 mg/L , for AMP were also >1000 mg/L, and for GTM and STM ranged from 875.5 mg/L to 947 mg/L . Reported EC 50 values for ERY and CHL were lower than those for AMO and AMP, but higher than those for EUG. Only Havelkova et al. provided a TC EC 50 value of 8.16 mg/L in a test with a longer exposure duration compared to ours. EUG continues to exhibit higher acute toxicity on non-target organisms such as D. magna compared to commercial ABXs. D. magna exhibits greater sensitivity to EUG compared to V. fischeri , consistent with prior research . Microcrustaceans generally appear more vulnerable to pollutants than bacteria . While the mode of action of EUG on V. fischeri remains unclear, it may involve mechanisms similar to those affecting other Gram-negative bacteria. Studies suggest that EUG acts as an ion channel blocker in D. magna , disrupting the proper function of its myogenic heart and causing a dose-dependent increase in the duration of muscle relaxation . Further research is necessary to fully understand EUG’s mode of action. 2.3.2. Soil Ecotoxicity of Eugenol Soil ecotoxicity was analyzed through a complete soil bacterial community, in which a comprehensive analysis of growth and metabolites intake was performed. In addition, soil ecotoxicity was also assessed on two indicators: Eisenia foetida and Allium cepa . Average Well Color Development (AWCD) of soil microbial populations and Community-Level Physiological Profiling (CLPP) after treatment with eugenol. Within the hierarchy of family, order, class, phylum, and kingdom, the proportion of total reads ranged from 93.71% to 99.49%, while for genus and species, it was 90.31% and 63.09%, respectively . The microbial profile detected in our soil sample is coherent with the type of microorganisms expected in a sample of these characteristics, as previously described by other authors . A similar effect as the one found in water was observed for the soil sample ( b) when AWCD was plotted versus a period of 168 h. From 24 to 168 h, statistical differences between the control and 1000 mg/L were found ( p < 0.01). However, at 168 h, the AWCD values between the control group and the dose of 100 mg/L were also significantly different. provides the data of the metabolic changes of the soil microorganisms after exposure to EUG for 168 h. For polymers, acids, and amino acids, significant differences ( p < 0.01) between the control and EUG at 1000 mg/L were found from 24 h. However, significant differences for amines/amides and carbohydrates were observed from 48 h and for this last group, significant differences between the control and an EUG dose of 100 mg/L was also detected after 72 h. To our knowledge, previous studies have not investigated the impact of EUG on soil microorganism communities. However, Pino-Otín et al. documented the AWCD and CLPP over time of different concentrations of AMO, AMP, STM, GTM, TC, ERY, and CHL on the soil microbiota. All commercial ABXs led to decreased AWCD at both 100 and 1000 mg/L, except for STM, which only reduced it at 100 mg/L . In contrast, EUG affected the soil microbiota mainly at 1000 mg/L, indicating lower ecotoxicity compared to commercial ABXs. Pino-Otín et al. observed the decreased metabolism of all five carbon sources with commercial ABXs, occurring either after 48 or 72 h across all concentrations. However, the influence on different carbon source groups varied, with amines and polymers showing the most significant changes. Conversely, EUG reduced the metabolism across all five groups equally, but only at 1000 mg/L. Similar to the water samples, soil bacterial communities exhibited higher resistance to EUG when composed of different species compared to single-species communities . Soil non-target organisms: Eisenia foetida and Allium cepa . A lethal dose of EUG for E. foetida was assessed by the 14-day mortality test, with this compound being toxic for its survival at very high concentrations. Up to 100 mg/L, the product does not induce mortality effects on E. foetida . After the 72 h exposure of A. cepa roots to EUG, its elongation was shortened, meaning that the EUG resulted as being toxic for the bulbs. The experiment showed a high degree of significance ( p < 0.0001) and the results are shown in c. The EC 10 values were measured as 1484.190 (CI: 1112.860–2039.888) mg/L and the EC 50 values correspond to 23.116 (CI: 20.061–26.743) mg/L. E. foetida , a common earthworm species, plays a crucial role in soil quality through its digging, feeding, and nutrient cycling activities, contributing to the soil structure and aeration. Its presence and behavior serve as indicators of soil health and contamination due to its susceptibility to pollutants, such as heavy metals . However, limited information exists on the effects of natural compounds on this species. In our study, the LC 50 for EUG was > 100 mg/kg, consistent with the findings by Almadiy and Nenaah , who observed no toxicity within a similar concentration range. Few studies have investigated the toxicity of antibiotics on E. foetida ; notably, TC effects were studied over a longer duration (56 days), yielding an LC 50 value of 2735 mg/kg . E. foetida exhibits notable resistance compared to other indicators, likely due to its complex multicellular nature. A. cepa is recognized as an indicator of soil, air, and water contamination, owing to its sensitivity to pH levels and nutrient requirements . The development and yield of onions depend directly on soil quality, particularly the presence of key metabolites like nitrogen, sulfur, and potassium. The responses of onions to these factors can indicate the soil nutrient status and deficiencies . Despite numerous studies on A. cepa , the effects of EUG on its roots remain unexplored, except for one study by Gogoi et al. , who found no toxicity of the essential oil containing EUG. Our study similarly observed the low toxicity of EUG on A. cepa root elongation (EC 50 = 23.116 mg/L), possibly due to its poor water solubility. In contrast, experiments with clove oil solutions containing EUG showed strong phytotoxicity on weeds, but they used an adjuvant (nonionic surfactants and paraffinic oil blend) to help dissolve it . While no specific ecotoxicity tests for ABXs on onions have been conducted to our knowledge, studies have examined how STM and TC affect these plants at the chromosome level . Our ecotoxicity findings indicate that EUG poses less harm to soil and water microbiota compared to commercial ABXs, possibly due to the faster metabolism by certain bacteria present in the samples. EUG demonstrated relatively low acute ecotoxicity for A. cepa and E. foetida ; however, direct comparisons with commercial ABXs are unavailable. In contrast, EUG exhibited higher acute ecotoxicity than some commercial ABXs for V. fischeri and D. magna . Nonetheless, the environmental impact of EUG would be minimal compared to commercial ABXs in case of being used as an adjuvant, as the proportion excreted into the environment without a mammalian metabolism is less than 0.1%, whereas for various ABXs, it ranges from 5% to 100% in human urine .
EUG was tested against 14 Gram-positive ( Bacillus subtilis , Listeria monocytogenes , Streptococcus agalactiae , and Staphylococcus aureus ) and Gram-negative ( Acinetobacter baumannii , Enterococcus faecalis , Escherichia coli , Klebsiella aerogenes , Klebsiella pneumoniae , Pasteurella aerogenes , Pseudomonas aeruginosa , Salmonella enterica, and Serratia marcescens ) bacterial strains. Their MIC results are given in . EUG could not be tested against P. mirabilis since the DMSO concentration required to dissolve it was toxic for these bacteria, whereas it was not for the other bacteria , as mentioned in . Our study, that reports microdilution MIC values for the first time for Gram-positive bacteria such as L. monocytogenes , S. agalactiae, and S. aureus , established MICs of 1000 μg/mL (see ) for these strains. We compared our MIC values with those observed for different strains of the same bacteria using different methods. For example, the MIC for non-resistant strains of S. aureus determined by Walsh et al. varied from 106 to 1590 μg/mL, while those observed by Gallucci et al. and Hammer and Heel were approximately 33,430 μg/mL and from 127.2 to 8480 μg/mL, respectively. Perugini Biasi-Garbin et al. analyzed the MIC of several strains of S. agalactiae and found values ranging from 1325 to 5300 μg/mL, which are close to ours. The reported MIC values for L. monocytogenes varied from 67 to 1024 μg/mL according to different authors , some of which are also close to our values. For B. subtilis , no reported values for the isolated compound were found. Previous research suggested that EUG disrupts bacterial membranes, increasing their permeability . Additionally, EUG might inhibit ATPase. This could alter the efflux pump activity, especially in L. monocytogenes and S. agalactiae , resulting in the higher intracellular concentration of the ABX. Both in vivo and in vitro studies demonstrate reduced L. monocytogenes virulence upon EUG treatment , highlighting its potential as an adjuvant to conventional antibiotics, against L. monocytogenes and possibly other pathogens. Interestingly, L. monocytogenes strains did not develop resistance to EUG after exposure to sub-inhibitory concentrations . Concerning Gram-negative bacteria, reported MIC values for the same strains as ours using the macro- or microdilution methods were found for E. coli , P. aeruginosa , S. enterica, and S. marcescens . In E. coli , the MIC values exhibit considerable variations not only with our results, but also in the studies cited in . However, for S. enterica and S. marcescens, they are closer to our MIC values. In the case of P. aeruginosa, no comparison can be made because the value mentioned by the authors is not precisely determined and above 1000 or 2000 μg/mL . For the remaining bacteria, no values for the same strains were found with the macro- or microdilution methods, so this comparison was performed for different strains. A. baumannii ’s MIC ranged from 90.5 to 304 μg/mL and 318 μg/mL , slightly lower than ours. E. faecalis ’ MIC was 1060 μg/mL in the study conducted by Hammer and Heel , half of our value. Lastly, no EUG MIC was found for P. aerogenes and K. aerogenes . Our study was thus the first one to provide information about the inhibitory effect of EUG on these species. The variation of MIC values of Gram-negative bacteria is probably due to differences in the complexity of their outer membrane. Within the group of Gram-negative bacteria, the lowest MIC values were obtained for A. baumannii , K. aerogenes , K. pneumoniae , and P. aerogenes (500 μg/mL, ). These low values suggest a potential specificity of EUG against these bacteria. In the case of A. baumannii , consistently with the studies of Karumathil et al. , this could be attributed mainly to an inhibition of the AdeABC efflux pump. In K. pneumoniae , EUG was shown to induce membranous lipid peroxidation resulting from the production of ROS , leading to membrane disruption. ROS production caused by EUG might be specific to this bacterium, leading to lower MIC values. Except for P. mirabilis , for which experiments were not feasible (see for more information), the MIC values for the other Gram-negative bacteria exceeded 1000 μg/mL . Regarding E. faecalis , Hammer and Heel observed significant membrane disruption, while Morgaan et al. attributed EUG’s mode of action on E. coli to its lipophilic nature, causing bacterial membrane lipid disintegration, leading to energy depletion and cellular component leakage. Additionally, ATPase inhibition was noted in E. coli exposed to EUG . For P. aeruginosa , membrane disruption was also evident , although some studies suggested that EUG’s antimicrobial activity might prolong the lag phase of bacterial growth at sub-inhibitory concentrations, possibly due to interactions with primary non-vital bacterial targets . Aljuwayd et al. and Zhao et al. investigated the effects of EUG on S. Typhimurium ( S. enterica ) in vivo, observing the destruction of fimbriae, and reductions in adhesion molecules, and virulence factors. Aljuwayd et al. observed decreased pathogenicity and cell counts in chickens with EUG pretreatment, while Zhao et al. reported similar findings and noted EUG’s ability to maintain its intestinal barrier function and Tight Junctions (TJs) protein stability in mice. No specific information regarding the mode of action was available for S. marcescens and P. mirabilis .
The checkboard assay revealed promising interactions between EUG and the conventional ABXs ampicillin (AMP), amoxicillin (AMO), gentamycin sulfate (GTM), streptomycin sulfate (STM), erythromycin (ERY), tetracycline hydrochloride (TC), and chloramphenicol (CHL). In general terms, 52% (9 out 17) of the experiments conducted on Gram-positive bacteria resulted in an ABX MIC reduction ≥ 75%, while this behavior was observed for only 26% (8 out of 30) of the tests on Gram-negative bacteria, . Six of these interactions were found to be synergistic: GTM-EUG- L. monocytogenes , GTM-EUG- S. aureus , CHL-EUG- A. baumannii , STM-EUG- S. enterica , TC-EUG- S. marcescens, and CHL-EUG- S. marcescens . All exhibited fractional inhibitory concentration (ΣFIC) values equal or below 0.50, calculated as described in . On the other hand, only two combinations resulted as antagonistic, which were STM-EUG- A. baumannii and CHL-EUG- K. aerogenes , both Gram-negative species. Although most of the interactions were additive, it must be highlighted that seven out of these additive combinations, AMP, AMO and TC-EUG- L. monocytogenes or GTM and CHL-EUG- S. agalactiae , STM-EUG- S. aureus, and CHL-EUG- K. pneumoniae , resulted in a pronounced decrease in the ABX dose (more than 75%, ). Additionally, a comprehensive kinetic growth study was conducted on the six synergistic interactions, as shown in , to unravel the behavior of each microorganism over a 24 h period under controlled conditions. It was observed that the combined administration of commercial ABXs and EUG, both at their MIC comb (see for a definition of MIC comb ) led to a complete inhibition of bacterial growth across all instances. These graphs also depict the total growth inhibition in experiments when natural and commercial ABXs were individually administered at their respective MIC alone (see for a definition of MIC alone ) and the non-inhibition when tested alone but at their respective MIC comb . However, even at subinhibitory concentrations of both the ABX and the natural product, some interesting facts can be extracted. The most pronounced effect observed in the graphs is related to the prolongation of the lag-phase. The lag-phase when the ABX was tested alone at its MIC comb was elongated for all experiments except for GTM-EUG- L. monocytogenes ( a) and CHL-EUG- A. baumannii ( c). However, when EUG was tested in these same circumstances, the lag-phase was modified only for S. enterica ( d). This fact was previously observed by Silva et al. on P. aeruginosa when EUG was tested at subinhibitory concentrations. They suggested that this behavior could be attributed to interactions between it and primary non-vital target sites of bacteria. 2.2.1. Aminoglycosides The precise synergistic mechanism between EUG and aminoglycosides remains unclear. However, previous studies suggest that the damage to the bacterial cell envelope induced by EUG may facilitate the entry of aminoglycosides (which use active transport systems to enter bacterial cells and inhibit protein synthesis ), thereby inhibiting bacterial growth at lower doses, albeit with variations depending on the microorganism . This mechanism would be especially efficient on Gram-positive bacteria, as observed for STM and GTM when acting together with EUG on S. agalactiae , L. monocytogenes, and S. aureus , , as shown in this work and previous research . For them, five of the six tested combinations resulted in an ABX MIC reduction ≥ 75%, while for the Gram-negative bacteria, this trend was observed only in two out of six. Within this group of bacteria, the most significant reduction in the aminoglycoside dosage was seen for STM against S. enterica , also displaying synergy ( and d), as previously demonstrated by Liu et al. , who noted the antibiofilm action of EUG in combination with STM. Additionally, while not synergistic but additive, EUG demonstrated substantial reductions in the aminoglycoside dosage when combined with STM against S. agalactiae , S. aureus, and K. pneumonia , as well as with GTM against S. agalactiae . This positions aminoglycosides as the antibiotic class most affected by their combination with EUG. Regarding the antagonistic interaction observed when EUG was combined with STM on A. baumannii , EUG might somehow alter the membrane potential and thus lead to a reduction in the active transport of STM into the cell, resulting in the observed antagonistic effect, . 2.2.2. Beta-Lactams Aminopenicillins are relatively small and hydrophilic molecules that use porins as essential entry points in these bacteria . Beta-lactams inhibit the activity of penicillin-binding proteins (PBPs), crucial for peptidoglycan cross-linking during cell wall synthesis, weakening the cell wall and altering the osmotic balance . This combined with EUG’s membrane-disrupting effects might lead to cell lysis . As expected, this effect was stronger in Gram-positive bacteria since the latter has a less complex envelop. Indeed, half of the tested interactions, versus 1 out of 10 for the Gram-negative bacteria, resulted in an ABX MIC reduction ≥ 75%. Although AMO and AMP did not exhibit synergistic interactions with EUG, they approached a value of ΣFIC = 0.5 for L. monocytogenes , showing a significant reduction in the ABX dose, although in an additive manner. Previous studies also found that EUG, when combined with AMP and AMO, led to an ABX dose reduction in a synergistic or additive manner on different strains of E. coli , P. aeruginosa , and S. Typhimurium or with AMO on E. faecalis . Our results may be especially significant for bacteria prone to developing resistance to beta-lactams through efflux pumps because the co-administration of ABX and EUG, which is known to inhibit efflux pumps activity, might result in a high intracellular ABX concentration . 2.2.3. Amphenicols The mechanism of action of CHL involves the reversible inhibition of bacterial protein synthesis by binding to the 50S subunit of the bacterial ribosome . It is a broad-spectrum antibiotic with high liposolubility, allowing it to enter bacteria through passive diffusion . The synergistic behavior observed for EUG when combined with CHL on A. baumannii ( and c) could involve the inhibition of the AdeABC efflux pump by EUG without increasing the membrane permeability . This could lead to the increased retention of CHL inside the cell, enhancing growth inhibition. Similarly, synergy between EUG and CHL was observed in this study against S. marcescens ( and f), proposing a possible shared mode of action involving efflux pump inhibition. Although there is limited research available on the remaining interactions in the selected microorganisms, EUG was found to decrease CHL MIC also in E. coli , P. aeruginosa, and S. enterica , as observed in . Unfortunately, we have no explanation concerning the antagonistic effect of CHL + EUG on K. aerogenes , since nothing is known about the mode of action of EUG in this species. 2.2.4. Tetracyclines TC is a broad-spectrum antibiotic that inhibits bacterial protein synthesis by binding to the 30S subunit of the bacterial ribosome, specifically at the A site, blocking the attachment of aminoacyl-tRNA molecules . This inhibition prevents the elongation of the polypeptide chain during translation. TC can also interfere with other bacterial processes, including essential metabolic pathways and cell membrane integrity, contributing to its bacteriostatic effect . The mechanism underlying the combined action of TC and EUG remains uncertain, although membrane disruption appears to be a plausible explanation, specifically in S. marcescens and L. monocytogenes , for which an ABX MIC reduction ≥ 75% was achieved ( and e). While there is limited literature on the combination of TC with EUG against these bacteria, a decrease in the ABX MIC for E. coli , P. aeruginosa , and S. Typhimurium has been previously reported . However, we did not investigate these combinations due to their initially low MIC (see . for detailed information). 2.2.5. Macrolides Macrolides, such as ERY, enter the cells via passive diffusion, inhibit protein chain elongation, and are effective against a broad spectrum of bacteria . When combined with EUG, ERY demonstrated MIC reductions across most tested bacteria, as previously determined on S. enterica , E. coli , and S. Typhimurium . Given EUG’s known modifications to the cell envelope, the mechanism of action of the EUG-ERY combination may be similar to that proposed for TC and EUG.
The precise synergistic mechanism between EUG and aminoglycosides remains unclear. However, previous studies suggest that the damage to the bacterial cell envelope induced by EUG may facilitate the entry of aminoglycosides (which use active transport systems to enter bacterial cells and inhibit protein synthesis ), thereby inhibiting bacterial growth at lower doses, albeit with variations depending on the microorganism . This mechanism would be especially efficient on Gram-positive bacteria, as observed for STM and GTM when acting together with EUG on S. agalactiae , L. monocytogenes, and S. aureus , , as shown in this work and previous research . For them, five of the six tested combinations resulted in an ABX MIC reduction ≥ 75%, while for the Gram-negative bacteria, this trend was observed only in two out of six. Within this group of bacteria, the most significant reduction in the aminoglycoside dosage was seen for STM against S. enterica , also displaying synergy ( and d), as previously demonstrated by Liu et al. , who noted the antibiofilm action of EUG in combination with STM. Additionally, while not synergistic but additive, EUG demonstrated substantial reductions in the aminoglycoside dosage when combined with STM against S. agalactiae , S. aureus, and K. pneumonia , as well as with GTM against S. agalactiae . This positions aminoglycosides as the antibiotic class most affected by their combination with EUG. Regarding the antagonistic interaction observed when EUG was combined with STM on A. baumannii , EUG might somehow alter the membrane potential and thus lead to a reduction in the active transport of STM into the cell, resulting in the observed antagonistic effect, .
Aminopenicillins are relatively small and hydrophilic molecules that use porins as essential entry points in these bacteria . Beta-lactams inhibit the activity of penicillin-binding proteins (PBPs), crucial for peptidoglycan cross-linking during cell wall synthesis, weakening the cell wall and altering the osmotic balance . This combined with EUG’s membrane-disrupting effects might lead to cell lysis . As expected, this effect was stronger in Gram-positive bacteria since the latter has a less complex envelop. Indeed, half of the tested interactions, versus 1 out of 10 for the Gram-negative bacteria, resulted in an ABX MIC reduction ≥ 75%. Although AMO and AMP did not exhibit synergistic interactions with EUG, they approached a value of ΣFIC = 0.5 for L. monocytogenes , showing a significant reduction in the ABX dose, although in an additive manner. Previous studies also found that EUG, when combined with AMP and AMO, led to an ABX dose reduction in a synergistic or additive manner on different strains of E. coli , P. aeruginosa , and S. Typhimurium or with AMO on E. faecalis . Our results may be especially significant for bacteria prone to developing resistance to beta-lactams through efflux pumps because the co-administration of ABX and EUG, which is known to inhibit efflux pumps activity, might result in a high intracellular ABX concentration .
The mechanism of action of CHL involves the reversible inhibition of bacterial protein synthesis by binding to the 50S subunit of the bacterial ribosome . It is a broad-spectrum antibiotic with high liposolubility, allowing it to enter bacteria through passive diffusion . The synergistic behavior observed for EUG when combined with CHL on A. baumannii ( and c) could involve the inhibition of the AdeABC efflux pump by EUG without increasing the membrane permeability . This could lead to the increased retention of CHL inside the cell, enhancing growth inhibition. Similarly, synergy between EUG and CHL was observed in this study against S. marcescens ( and f), proposing a possible shared mode of action involving efflux pump inhibition. Although there is limited research available on the remaining interactions in the selected microorganisms, EUG was found to decrease CHL MIC also in E. coli , P. aeruginosa, and S. enterica , as observed in . Unfortunately, we have no explanation concerning the antagonistic effect of CHL + EUG on K. aerogenes , since nothing is known about the mode of action of EUG in this species.
TC is a broad-spectrum antibiotic that inhibits bacterial protein synthesis by binding to the 30S subunit of the bacterial ribosome, specifically at the A site, blocking the attachment of aminoacyl-tRNA molecules . This inhibition prevents the elongation of the polypeptide chain during translation. TC can also interfere with other bacterial processes, including essential metabolic pathways and cell membrane integrity, contributing to its bacteriostatic effect . The mechanism underlying the combined action of TC and EUG remains uncertain, although membrane disruption appears to be a plausible explanation, specifically in S. marcescens and L. monocytogenes , for which an ABX MIC reduction ≥ 75% was achieved ( and e). While there is limited literature on the combination of TC with EUG against these bacteria, a decrease in the ABX MIC for E. coli , P. aeruginosa , and S. Typhimurium has been previously reported . However, we did not investigate these combinations due to their initially low MIC (see . for detailed information).
Macrolides, such as ERY, enter the cells via passive diffusion, inhibit protein chain elongation, and are effective against a broad spectrum of bacteria . When combined with EUG, ERY demonstrated MIC reductions across most tested bacteria, as previously determined on S. enterica , E. coli , and S. Typhimurium . Given EUG’s known modifications to the cell envelope, the mechanism of action of the EUG-ERY combination may be similar to that proposed for TC and EUG.
The impact of EUG on soil and water environments is not extensively documented. EUG is reported to have effective insecticidal properties and it was approved by the Environmental Protection Agency (EPA) in the USA due to its minimum risk on mammals . In this report, some toxicity data are given, mostly on mammals, but there are also data (very few) referring to invertebrates that will be commented on later. To draw more comprehensive conclusions about EUG ecotoxicity, we have selected different soil and water non-target organisms together with microbiota extracted from water and soil, on which the metabolic effect of EUG was assessed. 2.3.1. Water Ecotoxicity of Eugenol Water ecotoxicity was analyzed through three indicators: Vibrio fischeri , Daphnia magna, and a complete water bacterial community, in which a comprehensive analysis of growth and metabolites’ intake was carried out. Average Well Color Development (AWCD) of water microbial populations and Community-Level Physiological Profiling (CLPP) after treatment with eugenol. The taxonomic categorization, spanning from the kingdom to species, was derived through the analysis of 16S rRNA sequences obtained from our river water samples. Across the levels of species, genus, family, order, class, phylum, and kingdom, the proportion of total reads ranged from 94.13% to 99.35%, while for species, it constituted 79.00%, . This microbial profile of the water sample is representative of a fluvial environment in the Mediterranean area and similar to that reported by other studies . Microbial water communities were exposed to EUG for a 7 day-period on the Biolog Ecoplates ® ( a). According to the statistical analysis, only after 48 h of exposure were significant differences ( p < 0.01) found between the control and an EUG concentration of 1000 mg/L. For the lowest concentrations (0.1, 10, and 100 mg/L), no significant decrease in the ability to degrade carbon sources was observed, since all values were similar to those of the control group. The microbial CLPP is understood as the metabolic capacity for degrading different carbon sources after toxic exposure . This method assesses the whole impact of toxic substances on these communities’ metabolism overall or by the group of metabolites. The changes in the CLPP allow us to identify the impact of EUG on the capacity of water microorganisms to metabolize the 31 most frequent organic carbon sources . For a better understanding of the metabolic changes, these carbon sources were grouped into polymers, amino acids, amines/amides, carbohydrates, and carboxylic and acetic acids . These results are shown in and . The metabolic ability of the water microbial communities is barely affected by EUG with respect to the control. As observed when analyzing the AWCD, only the highest concentration of EUG substantially changed the metabolic profile of these microorganisms. In fact, statistical differences ( p < 0.01) were detected between the control and 1000 mg/L for four of the five groups of metabolites (carbohydrates, polymers, amino acids, and amines/amides) from 48, 96, 120, and also 120 h, respectively. For the amino acids group, there was also significant differences between the control and 100 mg/L of EUG from 96 h. For the acids, no statistical difference was detected at any time for any of the EUG doses tested. When comparing the effects of EUG and commercial ABXs (the ones used in the antimicrobial experiments) on water microbiota, our EUG data revealed a weaker impact on AWCD. EUG and commercial ABXs caused a significant reduction in AWCD only at 1000 mg/L, but GTM, CHL, and TC also induced significant decreases at 100 mg/L . Regarding CLPP, EUG showed substantial metabolic differences compared to the control at 1000 mg/L for all metabolic groups except amino acids, which exhibited differences at 100 mg/L. In contrast, commercial ABXs had more pronounced effects on the microbial metabolism at all tested concentrations (from 0.1 to 1000 mg/L), with few exceptions such as AMP and CHL at lower concentrations . Our findings suggest that bacterial communities with high taxonomic diversity, like those in our samples, are less affected by EUG exposure compared to commercial ABXs . This could be attributed to antibiotic-degrading resistance mechanisms present in some community members, such as Pseudomonas sp. which is present in our sample , which may degrade EUG more efficiently than commercial ABXs. Consequently, while some species may be impacted, they may be replaced by resistant ones, resulting in no overall impact on population growth, as measured by AWCD, except at very high concentrations. Effect of EUG on water non-target organisms: Vibrio fischeri and Daphnia magna. A bioluminescence inhibition assay on V. fischeri and its dose–response curve revealed toxicity for these Gram-negative bacteria, as can be appreciated in a. The experiment’s significance was high ( p < 0.0001) and both EC 10 and EC 50 values were measured for the 30 min exposure test, being 0.708 mg/L (CI: 0.533–0.906) and 8.778 mg/L (CI: 7.454–10.370), respectively. The D. magna immobilization test dose–response curve can be observed in b. The results of the 24 h exposure test, with a high level of significance ( p < 0.0001), revealed EC 10 and EC 50 values of 0.824 mg/L (CI: 0.496–1.098) and 1.963 (CI: 1.571–2.444) mg/L, respectively. V. fischeri and D. magna are both valuable indicators of ecotoxicity, with the former’s bioluminescence enhancing its sensitivity to environmental stressors . Lal et al. and Gueretz et al. found lower EC 50 values for EUG in these organisms compared to our study, but their analysis lacked statistical rigor and employed fewer concentrations. In our study, with five doses and statistical analysis, slightly higher EC 50 values were found for both organisms, aligning closely with the findings of Baker and Grant (2018) for V. fischeri . Regardless of the reference values chosen, EUG demonstrates significantly lower EC 50 values compared to the ABXs examined in our study. For instance, EC 50 values for AMO ranged from 150 mg/L to 4000 mg/L , for AMP from 163 mg/L to 1056 mg/L , for TC from 6.70 mg/L to 173.8 mg/L , and for CHL from 20.68 mg/L to 1086 mg/L in immobilization tests lasting 24–48 h. The highest EC 50 value recorded was for GTM (>10,000 mg/L) , while the lowest was 8.21 mg/L for STM . No acute ecotoxicity was observed for ERY on V. fischeri . In the case of D. magna , EC 50 values for AMO ranged from >1000 mg/L to 6950 mg/L , for AMP were also >1000 mg/L, and for GTM and STM ranged from 875.5 mg/L to 947 mg/L . Reported EC 50 values for ERY and CHL were lower than those for AMO and AMP, but higher than those for EUG. Only Havelkova et al. provided a TC EC 50 value of 8.16 mg/L in a test with a longer exposure duration compared to ours. EUG continues to exhibit higher acute toxicity on non-target organisms such as D. magna compared to commercial ABXs. D. magna exhibits greater sensitivity to EUG compared to V. fischeri , consistent with prior research . Microcrustaceans generally appear more vulnerable to pollutants than bacteria . While the mode of action of EUG on V. fischeri remains unclear, it may involve mechanisms similar to those affecting other Gram-negative bacteria. Studies suggest that EUG acts as an ion channel blocker in D. magna , disrupting the proper function of its myogenic heart and causing a dose-dependent increase in the duration of muscle relaxation . Further research is necessary to fully understand EUG’s mode of action. 2.3.2. Soil Ecotoxicity of Eugenol Soil ecotoxicity was analyzed through a complete soil bacterial community, in which a comprehensive analysis of growth and metabolites intake was performed. In addition, soil ecotoxicity was also assessed on two indicators: Eisenia foetida and Allium cepa . Average Well Color Development (AWCD) of soil microbial populations and Community-Level Physiological Profiling (CLPP) after treatment with eugenol. Within the hierarchy of family, order, class, phylum, and kingdom, the proportion of total reads ranged from 93.71% to 99.49%, while for genus and species, it was 90.31% and 63.09%, respectively . The microbial profile detected in our soil sample is coherent with the type of microorganisms expected in a sample of these characteristics, as previously described by other authors . A similar effect as the one found in water was observed for the soil sample ( b) when AWCD was plotted versus a period of 168 h. From 24 to 168 h, statistical differences between the control and 1000 mg/L were found ( p < 0.01). However, at 168 h, the AWCD values between the control group and the dose of 100 mg/L were also significantly different. provides the data of the metabolic changes of the soil microorganisms after exposure to EUG for 168 h. For polymers, acids, and amino acids, significant differences ( p < 0.01) between the control and EUG at 1000 mg/L were found from 24 h. However, significant differences for amines/amides and carbohydrates were observed from 48 h and for this last group, significant differences between the control and an EUG dose of 100 mg/L was also detected after 72 h. To our knowledge, previous studies have not investigated the impact of EUG on soil microorganism communities. However, Pino-Otín et al. documented the AWCD and CLPP over time of different concentrations of AMO, AMP, STM, GTM, TC, ERY, and CHL on the soil microbiota. All commercial ABXs led to decreased AWCD at both 100 and 1000 mg/L, except for STM, which only reduced it at 100 mg/L . In contrast, EUG affected the soil microbiota mainly at 1000 mg/L, indicating lower ecotoxicity compared to commercial ABXs. Pino-Otín et al. observed the decreased metabolism of all five carbon sources with commercial ABXs, occurring either after 48 or 72 h across all concentrations. However, the influence on different carbon source groups varied, with amines and polymers showing the most significant changes. Conversely, EUG reduced the metabolism across all five groups equally, but only at 1000 mg/L. Similar to the water samples, soil bacterial communities exhibited higher resistance to EUG when composed of different species compared to single-species communities . Soil non-target organisms: Eisenia foetida and Allium cepa . A lethal dose of EUG for E. foetida was assessed by the 14-day mortality test, with this compound being toxic for its survival at very high concentrations. Up to 100 mg/L, the product does not induce mortality effects on E. foetida . After the 72 h exposure of A. cepa roots to EUG, its elongation was shortened, meaning that the EUG resulted as being toxic for the bulbs. The experiment showed a high degree of significance ( p < 0.0001) and the results are shown in c. The EC 10 values were measured as 1484.190 (CI: 1112.860–2039.888) mg/L and the EC 50 values correspond to 23.116 (CI: 20.061–26.743) mg/L. E. foetida , a common earthworm species, plays a crucial role in soil quality through its digging, feeding, and nutrient cycling activities, contributing to the soil structure and aeration. Its presence and behavior serve as indicators of soil health and contamination due to its susceptibility to pollutants, such as heavy metals . However, limited information exists on the effects of natural compounds on this species. In our study, the LC 50 for EUG was > 100 mg/kg, consistent with the findings by Almadiy and Nenaah , who observed no toxicity within a similar concentration range. Few studies have investigated the toxicity of antibiotics on E. foetida ; notably, TC effects were studied over a longer duration (56 days), yielding an LC 50 value of 2735 mg/kg . E. foetida exhibits notable resistance compared to other indicators, likely due to its complex multicellular nature. A. cepa is recognized as an indicator of soil, air, and water contamination, owing to its sensitivity to pH levels and nutrient requirements . The development and yield of onions depend directly on soil quality, particularly the presence of key metabolites like nitrogen, sulfur, and potassium. The responses of onions to these factors can indicate the soil nutrient status and deficiencies . Despite numerous studies on A. cepa , the effects of EUG on its roots remain unexplored, except for one study by Gogoi et al. , who found no toxicity of the essential oil containing EUG. Our study similarly observed the low toxicity of EUG on A. cepa root elongation (EC 50 = 23.116 mg/L), possibly due to its poor water solubility. In contrast, experiments with clove oil solutions containing EUG showed strong phytotoxicity on weeds, but they used an adjuvant (nonionic surfactants and paraffinic oil blend) to help dissolve it . While no specific ecotoxicity tests for ABXs on onions have been conducted to our knowledge, studies have examined how STM and TC affect these plants at the chromosome level . Our ecotoxicity findings indicate that EUG poses less harm to soil and water microbiota compared to commercial ABXs, possibly due to the faster metabolism by certain bacteria present in the samples. EUG demonstrated relatively low acute ecotoxicity for A. cepa and E. foetida ; however, direct comparisons with commercial ABXs are unavailable. In contrast, EUG exhibited higher acute ecotoxicity than some commercial ABXs for V. fischeri and D. magna . Nonetheless, the environmental impact of EUG would be minimal compared to commercial ABXs in case of being used as an adjuvant, as the proportion excreted into the environment without a mammalian metabolism is less than 0.1%, whereas for various ABXs, it ranges from 5% to 100% in human urine .
Water ecotoxicity was analyzed through three indicators: Vibrio fischeri , Daphnia magna, and a complete water bacterial community, in which a comprehensive analysis of growth and metabolites’ intake was carried out. Average Well Color Development (AWCD) of water microbial populations and Community-Level Physiological Profiling (CLPP) after treatment with eugenol. The taxonomic categorization, spanning from the kingdom to species, was derived through the analysis of 16S rRNA sequences obtained from our river water samples. Across the levels of species, genus, family, order, class, phylum, and kingdom, the proportion of total reads ranged from 94.13% to 99.35%, while for species, it constituted 79.00%, . This microbial profile of the water sample is representative of a fluvial environment in the Mediterranean area and similar to that reported by other studies . Microbial water communities were exposed to EUG for a 7 day-period on the Biolog Ecoplates ® ( a). According to the statistical analysis, only after 48 h of exposure were significant differences ( p < 0.01) found between the control and an EUG concentration of 1000 mg/L. For the lowest concentrations (0.1, 10, and 100 mg/L), no significant decrease in the ability to degrade carbon sources was observed, since all values were similar to those of the control group. The microbial CLPP is understood as the metabolic capacity for degrading different carbon sources after toxic exposure . This method assesses the whole impact of toxic substances on these communities’ metabolism overall or by the group of metabolites. The changes in the CLPP allow us to identify the impact of EUG on the capacity of water microorganisms to metabolize the 31 most frequent organic carbon sources . For a better understanding of the metabolic changes, these carbon sources were grouped into polymers, amino acids, amines/amides, carbohydrates, and carboxylic and acetic acids . These results are shown in and . The metabolic ability of the water microbial communities is barely affected by EUG with respect to the control. As observed when analyzing the AWCD, only the highest concentration of EUG substantially changed the metabolic profile of these microorganisms. In fact, statistical differences ( p < 0.01) were detected between the control and 1000 mg/L for four of the five groups of metabolites (carbohydrates, polymers, amino acids, and amines/amides) from 48, 96, 120, and also 120 h, respectively. For the amino acids group, there was also significant differences between the control and 100 mg/L of EUG from 96 h. For the acids, no statistical difference was detected at any time for any of the EUG doses tested. When comparing the effects of EUG and commercial ABXs (the ones used in the antimicrobial experiments) on water microbiota, our EUG data revealed a weaker impact on AWCD. EUG and commercial ABXs caused a significant reduction in AWCD only at 1000 mg/L, but GTM, CHL, and TC also induced significant decreases at 100 mg/L . Regarding CLPP, EUG showed substantial metabolic differences compared to the control at 1000 mg/L for all metabolic groups except amino acids, which exhibited differences at 100 mg/L. In contrast, commercial ABXs had more pronounced effects on the microbial metabolism at all tested concentrations (from 0.1 to 1000 mg/L), with few exceptions such as AMP and CHL at lower concentrations . Our findings suggest that bacterial communities with high taxonomic diversity, like those in our samples, are less affected by EUG exposure compared to commercial ABXs . This could be attributed to antibiotic-degrading resistance mechanisms present in some community members, such as Pseudomonas sp. which is present in our sample , which may degrade EUG more efficiently than commercial ABXs. Consequently, while some species may be impacted, they may be replaced by resistant ones, resulting in no overall impact on population growth, as measured by AWCD, except at very high concentrations. Effect of EUG on water non-target organisms: Vibrio fischeri and Daphnia magna. A bioluminescence inhibition assay on V. fischeri and its dose–response curve revealed toxicity for these Gram-negative bacteria, as can be appreciated in a. The experiment’s significance was high ( p < 0.0001) and both EC 10 and EC 50 values were measured for the 30 min exposure test, being 0.708 mg/L (CI: 0.533–0.906) and 8.778 mg/L (CI: 7.454–10.370), respectively. The D. magna immobilization test dose–response curve can be observed in b. The results of the 24 h exposure test, with a high level of significance ( p < 0.0001), revealed EC 10 and EC 50 values of 0.824 mg/L (CI: 0.496–1.098) and 1.963 (CI: 1.571–2.444) mg/L, respectively. V. fischeri and D. magna are both valuable indicators of ecotoxicity, with the former’s bioluminescence enhancing its sensitivity to environmental stressors . Lal et al. and Gueretz et al. found lower EC 50 values for EUG in these organisms compared to our study, but their analysis lacked statistical rigor and employed fewer concentrations. In our study, with five doses and statistical analysis, slightly higher EC 50 values were found for both organisms, aligning closely with the findings of Baker and Grant (2018) for V. fischeri . Regardless of the reference values chosen, EUG demonstrates significantly lower EC 50 values compared to the ABXs examined in our study. For instance, EC 50 values for AMO ranged from 150 mg/L to 4000 mg/L , for AMP from 163 mg/L to 1056 mg/L , for TC from 6.70 mg/L to 173.8 mg/L , and for CHL from 20.68 mg/L to 1086 mg/L in immobilization tests lasting 24–48 h. The highest EC 50 value recorded was for GTM (>10,000 mg/L) , while the lowest was 8.21 mg/L for STM . No acute ecotoxicity was observed for ERY on V. fischeri . In the case of D. magna , EC 50 values for AMO ranged from >1000 mg/L to 6950 mg/L , for AMP were also >1000 mg/L, and for GTM and STM ranged from 875.5 mg/L to 947 mg/L . Reported EC 50 values for ERY and CHL were lower than those for AMO and AMP, but higher than those for EUG. Only Havelkova et al. provided a TC EC 50 value of 8.16 mg/L in a test with a longer exposure duration compared to ours. EUG continues to exhibit higher acute toxicity on non-target organisms such as D. magna compared to commercial ABXs. D. magna exhibits greater sensitivity to EUG compared to V. fischeri , consistent with prior research . Microcrustaceans generally appear more vulnerable to pollutants than bacteria . While the mode of action of EUG on V. fischeri remains unclear, it may involve mechanisms similar to those affecting other Gram-negative bacteria. Studies suggest that EUG acts as an ion channel blocker in D. magna , disrupting the proper function of its myogenic heart and causing a dose-dependent increase in the duration of muscle relaxation . Further research is necessary to fully understand EUG’s mode of action.
Soil ecotoxicity was analyzed through a complete soil bacterial community, in which a comprehensive analysis of growth and metabolites intake was performed. In addition, soil ecotoxicity was also assessed on two indicators: Eisenia foetida and Allium cepa . Average Well Color Development (AWCD) of soil microbial populations and Community-Level Physiological Profiling (CLPP) after treatment with eugenol. Within the hierarchy of family, order, class, phylum, and kingdom, the proportion of total reads ranged from 93.71% to 99.49%, while for genus and species, it was 90.31% and 63.09%, respectively . The microbial profile detected in our soil sample is coherent with the type of microorganisms expected in a sample of these characteristics, as previously described by other authors . A similar effect as the one found in water was observed for the soil sample ( b) when AWCD was plotted versus a period of 168 h. From 24 to 168 h, statistical differences between the control and 1000 mg/L were found ( p < 0.01). However, at 168 h, the AWCD values between the control group and the dose of 100 mg/L were also significantly different. provides the data of the metabolic changes of the soil microorganisms after exposure to EUG for 168 h. For polymers, acids, and amino acids, significant differences ( p < 0.01) between the control and EUG at 1000 mg/L were found from 24 h. However, significant differences for amines/amides and carbohydrates were observed from 48 h and for this last group, significant differences between the control and an EUG dose of 100 mg/L was also detected after 72 h. To our knowledge, previous studies have not investigated the impact of EUG on soil microorganism communities. However, Pino-Otín et al. documented the AWCD and CLPP over time of different concentrations of AMO, AMP, STM, GTM, TC, ERY, and CHL on the soil microbiota. All commercial ABXs led to decreased AWCD at both 100 and 1000 mg/L, except for STM, which only reduced it at 100 mg/L . In contrast, EUG affected the soil microbiota mainly at 1000 mg/L, indicating lower ecotoxicity compared to commercial ABXs. Pino-Otín et al. observed the decreased metabolism of all five carbon sources with commercial ABXs, occurring either after 48 or 72 h across all concentrations. However, the influence on different carbon source groups varied, with amines and polymers showing the most significant changes. Conversely, EUG reduced the metabolism across all five groups equally, but only at 1000 mg/L. Similar to the water samples, soil bacterial communities exhibited higher resistance to EUG when composed of different species compared to single-species communities . Soil non-target organisms: Eisenia foetida and Allium cepa . A lethal dose of EUG for E. foetida was assessed by the 14-day mortality test, with this compound being toxic for its survival at very high concentrations. Up to 100 mg/L, the product does not induce mortality effects on E. foetida . After the 72 h exposure of A. cepa roots to EUG, its elongation was shortened, meaning that the EUG resulted as being toxic for the bulbs. The experiment showed a high degree of significance ( p < 0.0001) and the results are shown in c. The EC 10 values were measured as 1484.190 (CI: 1112.860–2039.888) mg/L and the EC 50 values correspond to 23.116 (CI: 20.061–26.743) mg/L. E. foetida , a common earthworm species, plays a crucial role in soil quality through its digging, feeding, and nutrient cycling activities, contributing to the soil structure and aeration. Its presence and behavior serve as indicators of soil health and contamination due to its susceptibility to pollutants, such as heavy metals . However, limited information exists on the effects of natural compounds on this species. In our study, the LC 50 for EUG was > 100 mg/kg, consistent with the findings by Almadiy and Nenaah , who observed no toxicity within a similar concentration range. Few studies have investigated the toxicity of antibiotics on E. foetida ; notably, TC effects were studied over a longer duration (56 days), yielding an LC 50 value of 2735 mg/kg . E. foetida exhibits notable resistance compared to other indicators, likely due to its complex multicellular nature. A. cepa is recognized as an indicator of soil, air, and water contamination, owing to its sensitivity to pH levels and nutrient requirements . The development and yield of onions depend directly on soil quality, particularly the presence of key metabolites like nitrogen, sulfur, and potassium. The responses of onions to these factors can indicate the soil nutrient status and deficiencies . Despite numerous studies on A. cepa , the effects of EUG on its roots remain unexplored, except for one study by Gogoi et al. , who found no toxicity of the essential oil containing EUG. Our study similarly observed the low toxicity of EUG on A. cepa root elongation (EC 50 = 23.116 mg/L), possibly due to its poor water solubility. In contrast, experiments with clove oil solutions containing EUG showed strong phytotoxicity on weeds, but they used an adjuvant (nonionic surfactants and paraffinic oil blend) to help dissolve it . While no specific ecotoxicity tests for ABXs on onions have been conducted to our knowledge, studies have examined how STM and TC affect these plants at the chromosome level . Our ecotoxicity findings indicate that EUG poses less harm to soil and water microbiota compared to commercial ABXs, possibly due to the faster metabolism by certain bacteria present in the samples. EUG demonstrated relatively low acute ecotoxicity for A. cepa and E. foetida ; however, direct comparisons with commercial ABXs are unavailable. In contrast, EUG exhibited higher acute ecotoxicity than some commercial ABXs for V. fischeri and D. magna . Nonetheless, the environmental impact of EUG would be minimal compared to commercial ABXs in case of being used as an adjuvant, as the proportion excreted into the environment without a mammalian metabolism is less than 0.1%, whereas for various ABXs, it ranges from 5% to 100% in human urine .
3.1. Reagents Commercial antibiotics used for this research were AMP (CAS 69-53-4, purity > 96%), AMO (26787-78-0, purity > 96%), GTM (1405-41-0, purity > 96%), STP (3810-74-0, purity > 96%), ERY (114-07-8, purity 97.5%), TC (64-75-5, purity 99.2%), and CHL (56-75-7, purity 99.6%). AMP and AMO were acquired from Sigma Aldrich (Burlington, VT, USA), while GTM, ATP, ERY, and CHL from Acofarma, (Barcelona, Spain). On the other hand, the natural compound EUG (CAS 97-53-0, ≥98.5%) was also purchased from Sigma Aldrich (Burlington, VT, USA) and DMSO (CAS 67-68-5, >99.7%) from Fischer Bioreagents (Pittsburgh, PA, USA). 3.2. Bacterial Strains Growth Information concerning the strains used in this research and their growth conditions is shown in . Each strain was acquired lyophilized from Thermo Scientific (Waltham, MA, USA) in culture loops. The original strains were frozen in Cryoinstant Mix cryotubes (Deltalab, Barcelona, Spain) in accordance with the manufacturer’s recommendations to prevent mutations. After that, they were kept at −80 °C (Froilabo, Trust −80 °C, Collégien, France) and rehydrated in accordance with each microorganism’s technical information sheet for use , for which a summary is offered in . As previously described by Ferrando et al. , the bacterial inoculum was re-cultured from the cryotubes and incubated (J. P. Selecta, Barcelona, Spain) for 24 h in the conditions necessary for each optimum bacterial growth before any microorganism was used. After that, the culture was left overnight to reach the necessary bacterial optical density of 2.5 × 10 8 CFU/mL or 0.5 McFarland scale . 3.3. Minimum Inhibitory Concentration (MIC) The MIC of EUG was determined on each bacterial strain according to the Clinical and Laboratory Standards Institute . A stock solution of EUG was prepared by dissolving it into sterile water with 5% of DMSO, giving a concentration of 4000 μg/mL. The MIC of the ABXs cited in . and the toxicity of DMSO on the strains used in this work were analyzed in a previous study by Ferrando et al. . In that study, bacterial growth experiments were conducted in aqueous DMSO solutions with concentrations ranging from 0.04% to 20% ( v/v ). The objective was to identify the DMSO concentration that would allow the proper dissolution of the natural compound (cinnamaldehyde in that work, EUG in the current research) without adversely affecting bacterial growth. This optimal concentration was found to be 2.5%. At this concentration, bacterial growth was not affected for all bacteria except for P. mirabilis , which exhibited greater sensitivity to DMSO. A DMSO concentration of 0.16% would have been required to avoid the growth inhibition of P. mirabilis , but at this concentration, (as with cinnamaldehyde) it was not soluble. As mentioned in Ferrando et al. (2024) , not all possible ABX–EUG–Bacteria combinations were tested. Only ABXs with an MIC for a given bacteria type above 10 μg/mL were included in the study, whereas bacteria sensitive to DMSO concentrations below 2.5% were excluded because of the impossibility to solubilize EUG. According to these criteria, 47 combinations of ABX–EUG–bacteria were selected for the checkboard assay. To summarize, 100 μL of broth and 100 μL of EUG stock solution were added to each well of a 96-well round-bottom plate. After the bacterial suspension was adjusted using the 0.5 McFarland scale, 10 μL of it was added to the samples in two-fold dilutions. Additionally, negative and positive controls were introduced. The standard growth of bacteria in the absence of an antimicrobial agent was considered the positive control. The sole purpose of the negative control was to make sure that the culture broth was free of contamination or microbial development. Using a BioTekTM Synergy H1 Hybrid multimode microplate (Agilent, Madrid, Spain), absorbance was measured at 625 nm, at the optimal temperature for each bacterium, following the 24 h incubation period. Each experiment was carried out three times. 3.4. Bacterial Checkboard Assay The nature of the interactions between the natural chemical and each commercial antibiotic were investigated following the checkerboard assay technique . For this analysis, commercial ABX stock solutions were prepared by diluting them in sterile water at a concentration of 4 times their respective MICs. The procedure was the same for EUG, but the sterile water contained 5% of DMSO to guarantee its dissolution, as described in the previous section. Briefly, from the first to the seventh columns of the 96-well plate (round-bottom), EUG serial two-fold dilutions were added, while ABX solutions were added from rows A to G. As a result, the concentration of both types of antimicrobial drugs tested varied by well. A1 was the most concentrated well, while G7 was the least. After applying repeated two-fold dilutions to the plate, 10 μL of bacterium inoculum (0.5 McFarland) was added. Positive (antimicrobial-free bacterium) and negative (culture media without bacteria) controls were also constructed. To examine the interactions, two types of Fractional Inhibitory Concentration indices (FIC) were obtained, FIC A and FIC B (Equations (1) and (2), respectively). FIC A = (MIC of A in combination with B)/(MIC of A alone) (1) FIC B = (MIC of B in combination with A)/(MIC of B alone) (2) where A and B were each commercial ABX and EUG. The ∑FIC was calculated as the addition of both FICs. This value let us classify the interactions between the commercial ABX and EUG as follows: synergistic if ∑FIC ≤ 0.5, additive if ∑FIC was between 0.5 and 4, and antagonistic if ∑FIC ≥ 4 . Experiments were carried out in triplicate and in sterile flow chambers (Model MSC Advantage 1.2). 3.5. Bacterial Kinetic Growth Assay The procedure used was somewhat modified from that given by the Clinical and Laboratory Standards Institute . Wells in a 96-well plate (round-bottom) were filled with: (i) a commercial ABX solution at its MIC when tested alone (MIC alone ), (ii) a commercial ABX solution at its MIC when tested in combination with the natural product (MIC comb ), (iii) a natural compound solution at its MIC when tested alone, (iv) a natural product solution at its MIC when tested in combination with the commercial ABX, and (v) a solution made of natural compound and commercial ABX both at their respective MICs when tested in combination . At each bacterial optimum temperature , absorbance was measured every hour for 24 h at 625 nm using a SPECTROstar Nano from BMG Labtech (Madrid, Spain). Experiments were carried out in triplicate, and all results were presented as mean ± standard deviation. 3.6. Average Well Color Development (AWCD) Tests of Soil and Water Microorganisms 3.6.1. Water Samples Water samples were collected from the Gállego River (41°41′57″ N, 0°49′1″ W, Zaragoza, Spain) in June 2022. Procedures were carried out in situ according to standard procedures . The water temperature was 17 °C and its pH was neutral. Its physico-chemical characteristics are given in . Five liters of collected river water were used to isolate microorganisms for the genetic analysis using a 0.22 μm cellulose nitrate filter (Sartorius, Aubagne, France). The water was then resuspended in a sterile Falcon tube with 50 mL of sterilized phosphate-buffered saline (PBS) and centrifuged for 10 min at 5000× g to extract microorganisms for genetic analysis. The supernatant was removed, and the pellet was stored at −80 °C prior to sequencing. A 70 μm nylon filter (Becton Dickinson, Madrid, Spain) was used to filter 1 L of river water to remove debris, and then stored at 4 °C until it used for further experiments. 3.6.2. Soil Samples The procedure used for soil preparation was carried out according to earlier publication . Sample was obtained in June 2022 from a contaminant-free crop field (CITA, Zaragoza, Spain). The sample was first sieved to eliminate any contaminants larger than 2 mm, and it was then kept in darkness in sterile plastic bags. In a laminar flow biological safety hood to guarantee sterility (Model MSC Advantage 1.2, Thermo Fischer Scientific, Waltham, MA, USA), 95 mL of distilled, filtered (Sterifix ® 0.2 μm, Fischer Scientific, Waltham, MA, USA) water were added to 10 g of soil to begin soil microbial extraction, while mild magnetic stirring was performed for 30 min. Six falcon tubes were filled with 10 mL of the mixture’s supernatant after it had been allowed to stand for 1 h. The mixture was then sonicated for 1 min and centrifuged (at 1000× g for 10 min at 7 °C) with a Thermo Scientific Heraeus biofuge Primo R centrifuge (Waltham, MA, USA). The supernatants were collected and combined in a sterile tube. The pellets in the falcon tubes were then filled with 10 additional mL of filtered water and centrifuged again. This operation was repeated five times. The 60 mL finished leachate was then filtered to eliminate contaminants using a 70 μm nylon filter (Becton Dickinson, Madrid, Spain). 3.6.3. Biolog Ecoplates Preparation Using Biolog EcoPlates TM (Newark, NJ, USA) which contained three replicates of the 31 most important organic substrates for microbe metabolism, along with a water control, the microbial level physiological profile—defined as the metabolic capacity for degrading various carbon sources after substance exposure—was determined . This approach evaluates the full effect that potentially harmful compounds may have on the metabolism of these communities . At each concentration of EUG (0.1, 10, 100, and 1000 mg/L), wells of the Biolog EcoPlates TM were filled with 75 mL of soil leachate or processed water and 75 μL of EUG solutions. Three duplicates of each concentration were tested. The dilutions’ ultimate pH ranged from 6 to 7. J.P. Selecta (Barcelona, Spain) plates were sterilely incubated for 7 days at 25 °C. With the use of Gen5TM data analysis software (version number 1.08.4) and a BioTek Synergy H1 microplate reader (Agilent, Madrid, Spain), each well’s optical density (OD) was measured at t = 0 and every 24 h at 590 nm. With the OD values, the Average Well Color Development (AWCD) was calculated as follows : AWCD = ∑(AbsW − AbsC)/31 (3) where AbsW is the absorbance (or optical density) of each well with the carbon source and AbsC is the absorbance of the control well without it. According to comparable patterns of use, the 31 substrates were grouped into five functional classes: amines/amides, amino acids, carboxylic and acetic acids, polymers, and carbohydrates . Once this was completed, Equation (3) was applied to calculate the AWCD of each metabolic group . Repeated measures ANOVA test was performed to analyze the evolution of absorbance over time, with post hoc comparisons by Sidak if statistical significance was reached. ANOVA was carried out separately for each concentration and in a comparative way with the control as a factor. SPSS 28.0 software was used with a threshold value of p = 0.01 to accept or reject null hypothesis. 3.7. Genetic Analysis of Water and Soil Sample The prefiltered solution underwent subsequent filtration employing Sartorius (Madrid, Spain) 0.2 µm cellulose nitrate filters, which were thoroughly rinsed with PBS solution adjusted to a pH of 7.5. This resultant solution was then collected into Falcon tubes and subjected to centrifugation at 5000× g for a duration of 10 min. Subsequently, supernatants were meticulously decanted, and pellets were preserved by freezing at −80 °C for subsequent genetic analysis utilizing Froilabo (Paris, France), Trust −80 °C equipment . A quantity of 50 ng of DNA underwent amplification according to the 16S Metagenomic Sequencing Library Illumina 15,044,223 B protocol (Illumina, Paris, France) within the facilities of ADM BIOPOLIS laboratories (Parc Científica, Universitat de Valencia). In summary, the initial amplification phase involved the utilization of primers that incorporated: (1) a universal linker sequence facilitating the incorporation of amplicons with indexing and sequencing primers through the Nextera XT Index kit (Illumina, Paris, France), and (2) universal primers targeting the 16S rRNA gene . In the final amplification step, indexing sequences were integrated. The quantification of the 16S-based libraries was conducted via fluorimetry employing the Quant-iT™ PicoGreen™ dsDNA Assay Kit (Thermofisher, Waltham, MA, USA). The pooling of libraries occurred prior to sequencing on the MiSeq platform (Illumina, Paris, France) utilizing a configuration of 300 cycles for paired reads. The determination of pool size and quantity was conducted utilizing the Bioanalyzer 2100 (Agilent, Madrid, Spain) and the Library Quantification Kit for Illumina (Kapa Biosciences, Solo, Norway), respectively. A PhiX Control library (v3) (Illumina, Paris, France) was blended with the amplicon library at an anticipated ratio of 20%. Subsequently, sequencing data became available within an approximate timeframe of 56 h. Image analysis, base calling, and data quality assessment were executed on the MiSeq instrument utilizing MiSeq Control Software (MCS v3.1). 3.8. Vibrio Fischeri Bioluminiscence Assay Purchased lyophilizate bacteria from Macharey-Nagel (ref. 945 006) (Fischer Scientific, Waltham, MA, USA) were kept frozen at −18 °C. The protocol followed was UNE-EN ISO: 11348-3:2007 . Then, lyophilized V. fischeri bacteria were rehydrated with the reactivation solution provided by the company and stored at 4 °C for 5 min. Diluting pure EUG in an aqueous solution of 20 g/L NaCl yielded a stock solution of 4000 mg/L. Using the same solvent, successive dilutions of this were created, that is, 0.4, 4, 40, and 400 mg/L. The sample was strongly stirred for proper oxygenation and its pH was measured to make sure it stayed within the prescribed range (6–8.5). In addition, the appropriate volume of culture media (about 10 mL, purchased from Macharey-Nagel, reference 945 006), was added to the freeze-dried vial to create the bacterial solution. Four replicate measurements were taken for each sample dilution, and all samples and solutions were maintained at a temperature of 15 °C ± 1 °C. The goal of the experiment was to measure the bacterial solution’s bioluminescence following a brief (10 min) period of rest. Then, aliquots of EUG serial dilutions were added to a volume of the bacterial solution equal to 1 mL and the change in bioluminescence was assessed after 30 min of exposure. As a result, the concentrations tested (0.2, 2, 20, 200, and 2000 mg/L) were half of the serial dilutions described above. Using the XLSTAT (2014.5.03) program, the EC 50 and EC 10 values—effective concentrations of EUG that suppressed bioluminescence by 50% and 10%, respectively—and associated Confidence Intervals (CI) were calculated with the XLSTAT (2014.5.03) software from the dose–response curves for V. fischeri . 3.9. Daphnia Magna Assay Tests for D. magna (ref. DM121219, water flea, from Vidra Foc, Zaragoza, Spain) were conducted in accordance with standardized procedures and the Daphtoxkit FTM magna (1996) operating recommendations. The planktonic crustaceans were kept briefly, until usage, at 5 °C. D. magna eggs were incubated for 72 h between 20 and 22 °C in a TOXKIT model CH-0120D-AC/DC incubator (provided by ECOTEST, Valencia, Spain) with 6000 lx illumination. Spirulin provided in the Daphtokit was added to the crustaceans two hours before being exposed to EUG. EUG was evaluated at final concentrations of 0.1, 1, 10, and 100 mg/L in sterile freshwater . Additionally, this water served as a negative control. A 0.1 M NaOH solution was used to raise the pH to a range of 7–7.5. Each concentration was evaluated with five replicates of five organisms each. Daphnids were cultured at the required concentrations for 24 h in complete darkness at 20–22 °C. When gently agitated for 15 s, organisms were deemed immobile if they were unable to swim. Effective concentration values of EUG resulting in 50% and 10% (EC 50 and EC 10 , respectively) immobilization (inactive neonates) and their Confidence Intervals (CI) were obtained from the dose–response curves for D. magna mobility tests using the XLSTAT (2014.5.03) software, as mentioned in . 3.10. Eisenia Foetida Assay Adult E. foetida earthworms were purchased from Todo Verde (Ourense, Spain) and kept in sphagnum peat substrate (Spanish Flowers Company, Zaragoza, Spain) following recommendations from the provider for their optimum development. A controlled temperature of 18 to 25 °C, pH = 7.5 to 8, and 80% to 85% humidity for 15 days prior to testing was ensured. The experiments were conducted using standardized procedures . Earthworms that were at least 2 months old, had clitella, and weighed between 300 and 600 mg each were chosen for the tests . Tests were conducted in 750 cm 3 plastic jars with lids to avoid animal scape and to ensure the proper humidity. To allow ventilation and oxygen delivery, the jar lids were pierced with holes. The jars were filled with 750 g (wet weight) of standardized soil substrate, which was made up of industrial fine sand, sphagnum peat, and kaolin clay in a 7:1:2 ratio, respectively. Kaolin clay and sand were acquired from Imerys Ceramics (Civita Castellana, Italy), while sphagnum peat was purchased from Verdecora Vivarium (Zaragoza, Spain). Weighing the sample and drying it to a constant mass at 105 °C for 24 h allowed us to calculate the mixture’s water content. Deionized water was added to the medium and carefully mixed to adjust the total moisture content to 35–45% of the dry weight of the soil. Ten earthworms and EUG at different concentrations (0.1, 1, 10, 100, and 1000 mg/kg) were added to the jars and left for 14 days in a regulated environment (20 ± 2 °C, 80–85% relative humidity, and 400–800 lx of light). Each concentration was tested three times and negative controls without EUG were also prepared. Lethal concentration values of EUG LC 50 and LC 10 and their Confidence Intervals (CI) were obtained, as in , from the dose–response curves for E. foetida tests using the XLSTAT (2014.5.03) software. 3.11. Allium Cepa Assay Bulbs of A. cepa (variety Stuttgarter Riesen de 14/21) were acquired from Fitoagrícola Company (Castellón de la Plana, Spain) and stored until use in a dry environment, between 10 and 20 °C in the dark. The young bulbs were peeled before the test, preventing damage to the root ring. A. cepa acute toxicity experiments were carried out according to specific standardized procedures , which measured root elongation after 72 h of exposure to the test chemical . Mineral water from Aguas de San Martín de Veri S.A. (San Martín de Veri, Spain) was purchased and used as the growth medium for the bulbs. This medium was placed in 15 mL tubes, as it contains an adequate amount of Ca 2+ and Mg 2+ ( https://www.veri.es/es/el-producto accessed on 12 May 2024). In ecotoxicological experiments, twelve duplicates of each EUG concentration (0.2, 2, 20, 100, and 500 mg/L) were employed. Only water was used for the negative controls. The bulbs were placed in the top of each 15 mL tube and grown for 72 h at 25 °C in a dimly lit space. The tested solutions were renewed every 24 h. As in , using the XLSTAT (2014.5.03) program, the EC 50 and EC 10 values—effective concentrations of EUG that cause 50% and 10%, respectively, of root growth—as well as their Confidence Intervals (CI) were calculated from the dose–response curves for A. cepa .
Commercial antibiotics used for this research were AMP (CAS 69-53-4, purity > 96%), AMO (26787-78-0, purity > 96%), GTM (1405-41-0, purity > 96%), STP (3810-74-0, purity > 96%), ERY (114-07-8, purity 97.5%), TC (64-75-5, purity 99.2%), and CHL (56-75-7, purity 99.6%). AMP and AMO were acquired from Sigma Aldrich (Burlington, VT, USA), while GTM, ATP, ERY, and CHL from Acofarma, (Barcelona, Spain). On the other hand, the natural compound EUG (CAS 97-53-0, ≥98.5%) was also purchased from Sigma Aldrich (Burlington, VT, USA) and DMSO (CAS 67-68-5, >99.7%) from Fischer Bioreagents (Pittsburgh, PA, USA).
Information concerning the strains used in this research and their growth conditions is shown in . Each strain was acquired lyophilized from Thermo Scientific (Waltham, MA, USA) in culture loops. The original strains were frozen in Cryoinstant Mix cryotubes (Deltalab, Barcelona, Spain) in accordance with the manufacturer’s recommendations to prevent mutations. After that, they were kept at −80 °C (Froilabo, Trust −80 °C, Collégien, France) and rehydrated in accordance with each microorganism’s technical information sheet for use , for which a summary is offered in . As previously described by Ferrando et al. , the bacterial inoculum was re-cultured from the cryotubes and incubated (J. P. Selecta, Barcelona, Spain) for 24 h in the conditions necessary for each optimum bacterial growth before any microorganism was used. After that, the culture was left overnight to reach the necessary bacterial optical density of 2.5 × 10 8 CFU/mL or 0.5 McFarland scale .
The MIC of EUG was determined on each bacterial strain according to the Clinical and Laboratory Standards Institute . A stock solution of EUG was prepared by dissolving it into sterile water with 5% of DMSO, giving a concentration of 4000 μg/mL. The MIC of the ABXs cited in . and the toxicity of DMSO on the strains used in this work were analyzed in a previous study by Ferrando et al. . In that study, bacterial growth experiments were conducted in aqueous DMSO solutions with concentrations ranging from 0.04% to 20% ( v/v ). The objective was to identify the DMSO concentration that would allow the proper dissolution of the natural compound (cinnamaldehyde in that work, EUG in the current research) without adversely affecting bacterial growth. This optimal concentration was found to be 2.5%. At this concentration, bacterial growth was not affected for all bacteria except for P. mirabilis , which exhibited greater sensitivity to DMSO. A DMSO concentration of 0.16% would have been required to avoid the growth inhibition of P. mirabilis , but at this concentration, (as with cinnamaldehyde) it was not soluble. As mentioned in Ferrando et al. (2024) , not all possible ABX–EUG–Bacteria combinations were tested. Only ABXs with an MIC for a given bacteria type above 10 μg/mL were included in the study, whereas bacteria sensitive to DMSO concentrations below 2.5% were excluded because of the impossibility to solubilize EUG. According to these criteria, 47 combinations of ABX–EUG–bacteria were selected for the checkboard assay. To summarize, 100 μL of broth and 100 μL of EUG stock solution were added to each well of a 96-well round-bottom plate. After the bacterial suspension was adjusted using the 0.5 McFarland scale, 10 μL of it was added to the samples in two-fold dilutions. Additionally, negative and positive controls were introduced. The standard growth of bacteria in the absence of an antimicrobial agent was considered the positive control. The sole purpose of the negative control was to make sure that the culture broth was free of contamination or microbial development. Using a BioTekTM Synergy H1 Hybrid multimode microplate (Agilent, Madrid, Spain), absorbance was measured at 625 nm, at the optimal temperature for each bacterium, following the 24 h incubation period. Each experiment was carried out three times.
The nature of the interactions between the natural chemical and each commercial antibiotic were investigated following the checkerboard assay technique . For this analysis, commercial ABX stock solutions were prepared by diluting them in sterile water at a concentration of 4 times their respective MICs. The procedure was the same for EUG, but the sterile water contained 5% of DMSO to guarantee its dissolution, as described in the previous section. Briefly, from the first to the seventh columns of the 96-well plate (round-bottom), EUG serial two-fold dilutions were added, while ABX solutions were added from rows A to G. As a result, the concentration of both types of antimicrobial drugs tested varied by well. A1 was the most concentrated well, while G7 was the least. After applying repeated two-fold dilutions to the plate, 10 μL of bacterium inoculum (0.5 McFarland) was added. Positive (antimicrobial-free bacterium) and negative (culture media without bacteria) controls were also constructed. To examine the interactions, two types of Fractional Inhibitory Concentration indices (FIC) were obtained, FIC A and FIC B (Equations (1) and (2), respectively). FIC A = (MIC of A in combination with B)/(MIC of A alone) (1) FIC B = (MIC of B in combination with A)/(MIC of B alone) (2) where A and B were each commercial ABX and EUG. The ∑FIC was calculated as the addition of both FICs. This value let us classify the interactions between the commercial ABX and EUG as follows: synergistic if ∑FIC ≤ 0.5, additive if ∑FIC was between 0.5 and 4, and antagonistic if ∑FIC ≥ 4 . Experiments were carried out in triplicate and in sterile flow chambers (Model MSC Advantage 1.2).
The procedure used was somewhat modified from that given by the Clinical and Laboratory Standards Institute . Wells in a 96-well plate (round-bottom) were filled with: (i) a commercial ABX solution at its MIC when tested alone (MIC alone ), (ii) a commercial ABX solution at its MIC when tested in combination with the natural product (MIC comb ), (iii) a natural compound solution at its MIC when tested alone, (iv) a natural product solution at its MIC when tested in combination with the commercial ABX, and (v) a solution made of natural compound and commercial ABX both at their respective MICs when tested in combination . At each bacterial optimum temperature , absorbance was measured every hour for 24 h at 625 nm using a SPECTROstar Nano from BMG Labtech (Madrid, Spain). Experiments were carried out in triplicate, and all results were presented as mean ± standard deviation.
3.6.1. Water Samples Water samples were collected from the Gállego River (41°41′57″ N, 0°49′1″ W, Zaragoza, Spain) in June 2022. Procedures were carried out in situ according to standard procedures . The water temperature was 17 °C and its pH was neutral. Its physico-chemical characteristics are given in . Five liters of collected river water were used to isolate microorganisms for the genetic analysis using a 0.22 μm cellulose nitrate filter (Sartorius, Aubagne, France). The water was then resuspended in a sterile Falcon tube with 50 mL of sterilized phosphate-buffered saline (PBS) and centrifuged for 10 min at 5000× g to extract microorganisms for genetic analysis. The supernatant was removed, and the pellet was stored at −80 °C prior to sequencing. A 70 μm nylon filter (Becton Dickinson, Madrid, Spain) was used to filter 1 L of river water to remove debris, and then stored at 4 °C until it used for further experiments. 3.6.2. Soil Samples The procedure used for soil preparation was carried out according to earlier publication . Sample was obtained in June 2022 from a contaminant-free crop field (CITA, Zaragoza, Spain). The sample was first sieved to eliminate any contaminants larger than 2 mm, and it was then kept in darkness in sterile plastic bags. In a laminar flow biological safety hood to guarantee sterility (Model MSC Advantage 1.2, Thermo Fischer Scientific, Waltham, MA, USA), 95 mL of distilled, filtered (Sterifix ® 0.2 μm, Fischer Scientific, Waltham, MA, USA) water were added to 10 g of soil to begin soil microbial extraction, while mild magnetic stirring was performed for 30 min. Six falcon tubes were filled with 10 mL of the mixture’s supernatant after it had been allowed to stand for 1 h. The mixture was then sonicated for 1 min and centrifuged (at 1000× g for 10 min at 7 °C) with a Thermo Scientific Heraeus biofuge Primo R centrifuge (Waltham, MA, USA). The supernatants were collected and combined in a sterile tube. The pellets in the falcon tubes were then filled with 10 additional mL of filtered water and centrifuged again. This operation was repeated five times. The 60 mL finished leachate was then filtered to eliminate contaminants using a 70 μm nylon filter (Becton Dickinson, Madrid, Spain). 3.6.3. Biolog Ecoplates Preparation Using Biolog EcoPlates TM (Newark, NJ, USA) which contained three replicates of the 31 most important organic substrates for microbe metabolism, along with a water control, the microbial level physiological profile—defined as the metabolic capacity for degrading various carbon sources after substance exposure—was determined . This approach evaluates the full effect that potentially harmful compounds may have on the metabolism of these communities . At each concentration of EUG (0.1, 10, 100, and 1000 mg/L), wells of the Biolog EcoPlates TM were filled with 75 mL of soil leachate or processed water and 75 μL of EUG solutions. Three duplicates of each concentration were tested. The dilutions’ ultimate pH ranged from 6 to 7. J.P. Selecta (Barcelona, Spain) plates were sterilely incubated for 7 days at 25 °C. With the use of Gen5TM data analysis software (version number 1.08.4) and a BioTek Synergy H1 microplate reader (Agilent, Madrid, Spain), each well’s optical density (OD) was measured at t = 0 and every 24 h at 590 nm. With the OD values, the Average Well Color Development (AWCD) was calculated as follows : AWCD = ∑(AbsW − AbsC)/31 (3) where AbsW is the absorbance (or optical density) of each well with the carbon source and AbsC is the absorbance of the control well without it. According to comparable patterns of use, the 31 substrates were grouped into five functional classes: amines/amides, amino acids, carboxylic and acetic acids, polymers, and carbohydrates . Once this was completed, Equation (3) was applied to calculate the AWCD of each metabolic group . Repeated measures ANOVA test was performed to analyze the evolution of absorbance over time, with post hoc comparisons by Sidak if statistical significance was reached. ANOVA was carried out separately for each concentration and in a comparative way with the control as a factor. SPSS 28.0 software was used with a threshold value of p = 0.01 to accept or reject null hypothesis.
Water samples were collected from the Gállego River (41°41′57″ N, 0°49′1″ W, Zaragoza, Spain) in June 2022. Procedures were carried out in situ according to standard procedures . The water temperature was 17 °C and its pH was neutral. Its physico-chemical characteristics are given in . Five liters of collected river water were used to isolate microorganisms for the genetic analysis using a 0.22 μm cellulose nitrate filter (Sartorius, Aubagne, France). The water was then resuspended in a sterile Falcon tube with 50 mL of sterilized phosphate-buffered saline (PBS) and centrifuged for 10 min at 5000× g to extract microorganisms for genetic analysis. The supernatant was removed, and the pellet was stored at −80 °C prior to sequencing. A 70 μm nylon filter (Becton Dickinson, Madrid, Spain) was used to filter 1 L of river water to remove debris, and then stored at 4 °C until it used for further experiments.
The procedure used for soil preparation was carried out according to earlier publication . Sample was obtained in June 2022 from a contaminant-free crop field (CITA, Zaragoza, Spain). The sample was first sieved to eliminate any contaminants larger than 2 mm, and it was then kept in darkness in sterile plastic bags. In a laminar flow biological safety hood to guarantee sterility (Model MSC Advantage 1.2, Thermo Fischer Scientific, Waltham, MA, USA), 95 mL of distilled, filtered (Sterifix ® 0.2 μm, Fischer Scientific, Waltham, MA, USA) water were added to 10 g of soil to begin soil microbial extraction, while mild magnetic stirring was performed for 30 min. Six falcon tubes were filled with 10 mL of the mixture’s supernatant after it had been allowed to stand for 1 h. The mixture was then sonicated for 1 min and centrifuged (at 1000× g for 10 min at 7 °C) with a Thermo Scientific Heraeus biofuge Primo R centrifuge (Waltham, MA, USA). The supernatants were collected and combined in a sterile tube. The pellets in the falcon tubes were then filled with 10 additional mL of filtered water and centrifuged again. This operation was repeated five times. The 60 mL finished leachate was then filtered to eliminate contaminants using a 70 μm nylon filter (Becton Dickinson, Madrid, Spain).
Using Biolog EcoPlates TM (Newark, NJ, USA) which contained three replicates of the 31 most important organic substrates for microbe metabolism, along with a water control, the microbial level physiological profile—defined as the metabolic capacity for degrading various carbon sources after substance exposure—was determined . This approach evaluates the full effect that potentially harmful compounds may have on the metabolism of these communities . At each concentration of EUG (0.1, 10, 100, and 1000 mg/L), wells of the Biolog EcoPlates TM were filled with 75 mL of soil leachate or processed water and 75 μL of EUG solutions. Three duplicates of each concentration were tested. The dilutions’ ultimate pH ranged from 6 to 7. J.P. Selecta (Barcelona, Spain) plates were sterilely incubated for 7 days at 25 °C. With the use of Gen5TM data analysis software (version number 1.08.4) and a BioTek Synergy H1 microplate reader (Agilent, Madrid, Spain), each well’s optical density (OD) was measured at t = 0 and every 24 h at 590 nm. With the OD values, the Average Well Color Development (AWCD) was calculated as follows : AWCD = ∑(AbsW − AbsC)/31 (3) where AbsW is the absorbance (or optical density) of each well with the carbon source and AbsC is the absorbance of the control well without it. According to comparable patterns of use, the 31 substrates were grouped into five functional classes: amines/amides, amino acids, carboxylic and acetic acids, polymers, and carbohydrates . Once this was completed, Equation (3) was applied to calculate the AWCD of each metabolic group . Repeated measures ANOVA test was performed to analyze the evolution of absorbance over time, with post hoc comparisons by Sidak if statistical significance was reached. ANOVA was carried out separately for each concentration and in a comparative way with the control as a factor. SPSS 28.0 software was used with a threshold value of p = 0.01 to accept or reject null hypothesis.
The prefiltered solution underwent subsequent filtration employing Sartorius (Madrid, Spain) 0.2 µm cellulose nitrate filters, which were thoroughly rinsed with PBS solution adjusted to a pH of 7.5. This resultant solution was then collected into Falcon tubes and subjected to centrifugation at 5000× g for a duration of 10 min. Subsequently, supernatants were meticulously decanted, and pellets were preserved by freezing at −80 °C for subsequent genetic analysis utilizing Froilabo (Paris, France), Trust −80 °C equipment . A quantity of 50 ng of DNA underwent amplification according to the 16S Metagenomic Sequencing Library Illumina 15,044,223 B protocol (Illumina, Paris, France) within the facilities of ADM BIOPOLIS laboratories (Parc Científica, Universitat de Valencia). In summary, the initial amplification phase involved the utilization of primers that incorporated: (1) a universal linker sequence facilitating the incorporation of amplicons with indexing and sequencing primers through the Nextera XT Index kit (Illumina, Paris, France), and (2) universal primers targeting the 16S rRNA gene . In the final amplification step, indexing sequences were integrated. The quantification of the 16S-based libraries was conducted via fluorimetry employing the Quant-iT™ PicoGreen™ dsDNA Assay Kit (Thermofisher, Waltham, MA, USA). The pooling of libraries occurred prior to sequencing on the MiSeq platform (Illumina, Paris, France) utilizing a configuration of 300 cycles for paired reads. The determination of pool size and quantity was conducted utilizing the Bioanalyzer 2100 (Agilent, Madrid, Spain) and the Library Quantification Kit for Illumina (Kapa Biosciences, Solo, Norway), respectively. A PhiX Control library (v3) (Illumina, Paris, France) was blended with the amplicon library at an anticipated ratio of 20%. Subsequently, sequencing data became available within an approximate timeframe of 56 h. Image analysis, base calling, and data quality assessment were executed on the MiSeq instrument utilizing MiSeq Control Software (MCS v3.1).
Purchased lyophilizate bacteria from Macharey-Nagel (ref. 945 006) (Fischer Scientific, Waltham, MA, USA) were kept frozen at −18 °C. The protocol followed was UNE-EN ISO: 11348-3:2007 . Then, lyophilized V. fischeri bacteria were rehydrated with the reactivation solution provided by the company and stored at 4 °C for 5 min. Diluting pure EUG in an aqueous solution of 20 g/L NaCl yielded a stock solution of 4000 mg/L. Using the same solvent, successive dilutions of this were created, that is, 0.4, 4, 40, and 400 mg/L. The sample was strongly stirred for proper oxygenation and its pH was measured to make sure it stayed within the prescribed range (6–8.5). In addition, the appropriate volume of culture media (about 10 mL, purchased from Macharey-Nagel, reference 945 006), was added to the freeze-dried vial to create the bacterial solution. Four replicate measurements were taken for each sample dilution, and all samples and solutions were maintained at a temperature of 15 °C ± 1 °C. The goal of the experiment was to measure the bacterial solution’s bioluminescence following a brief (10 min) period of rest. Then, aliquots of EUG serial dilutions were added to a volume of the bacterial solution equal to 1 mL and the change in bioluminescence was assessed after 30 min of exposure. As a result, the concentrations tested (0.2, 2, 20, 200, and 2000 mg/L) were half of the serial dilutions described above. Using the XLSTAT (2014.5.03) program, the EC 50 and EC 10 values—effective concentrations of EUG that suppressed bioluminescence by 50% and 10%, respectively—and associated Confidence Intervals (CI) were calculated with the XLSTAT (2014.5.03) software from the dose–response curves for V. fischeri .
Tests for D. magna (ref. DM121219, water flea, from Vidra Foc, Zaragoza, Spain) were conducted in accordance with standardized procedures and the Daphtoxkit FTM magna (1996) operating recommendations. The planktonic crustaceans were kept briefly, until usage, at 5 °C. D. magna eggs were incubated for 72 h between 20 and 22 °C in a TOXKIT model CH-0120D-AC/DC incubator (provided by ECOTEST, Valencia, Spain) with 6000 lx illumination. Spirulin provided in the Daphtokit was added to the crustaceans two hours before being exposed to EUG. EUG was evaluated at final concentrations of 0.1, 1, 10, and 100 mg/L in sterile freshwater . Additionally, this water served as a negative control. A 0.1 M NaOH solution was used to raise the pH to a range of 7–7.5. Each concentration was evaluated with five replicates of five organisms each. Daphnids were cultured at the required concentrations for 24 h in complete darkness at 20–22 °C. When gently agitated for 15 s, organisms were deemed immobile if they were unable to swim. Effective concentration values of EUG resulting in 50% and 10% (EC 50 and EC 10 , respectively) immobilization (inactive neonates) and their Confidence Intervals (CI) were obtained from the dose–response curves for D. magna mobility tests using the XLSTAT (2014.5.03) software, as mentioned in .
Adult E. foetida earthworms were purchased from Todo Verde (Ourense, Spain) and kept in sphagnum peat substrate (Spanish Flowers Company, Zaragoza, Spain) following recommendations from the provider for their optimum development. A controlled temperature of 18 to 25 °C, pH = 7.5 to 8, and 80% to 85% humidity for 15 days prior to testing was ensured. The experiments were conducted using standardized procedures . Earthworms that were at least 2 months old, had clitella, and weighed between 300 and 600 mg each were chosen for the tests . Tests were conducted in 750 cm 3 plastic jars with lids to avoid animal scape and to ensure the proper humidity. To allow ventilation and oxygen delivery, the jar lids were pierced with holes. The jars were filled with 750 g (wet weight) of standardized soil substrate, which was made up of industrial fine sand, sphagnum peat, and kaolin clay in a 7:1:2 ratio, respectively. Kaolin clay and sand were acquired from Imerys Ceramics (Civita Castellana, Italy), while sphagnum peat was purchased from Verdecora Vivarium (Zaragoza, Spain). Weighing the sample and drying it to a constant mass at 105 °C for 24 h allowed us to calculate the mixture’s water content. Deionized water was added to the medium and carefully mixed to adjust the total moisture content to 35–45% of the dry weight of the soil. Ten earthworms and EUG at different concentrations (0.1, 1, 10, 100, and 1000 mg/kg) were added to the jars and left for 14 days in a regulated environment (20 ± 2 °C, 80–85% relative humidity, and 400–800 lx of light). Each concentration was tested three times and negative controls without EUG were also prepared. Lethal concentration values of EUG LC 50 and LC 10 and their Confidence Intervals (CI) were obtained, as in , from the dose–response curves for E. foetida tests using the XLSTAT (2014.5.03) software.
Bulbs of A. cepa (variety Stuttgarter Riesen de 14/21) were acquired from Fitoagrícola Company (Castellón de la Plana, Spain) and stored until use in a dry environment, between 10 and 20 °C in the dark. The young bulbs were peeled before the test, preventing damage to the root ring. A. cepa acute toxicity experiments were carried out according to specific standardized procedures , which measured root elongation after 72 h of exposure to the test chemical . Mineral water from Aguas de San Martín de Veri S.A. (San Martín de Veri, Spain) was purchased and used as the growth medium for the bulbs. This medium was placed in 15 mL tubes, as it contains an adequate amount of Ca 2+ and Mg 2+ ( https://www.veri.es/es/el-producto accessed on 12 May 2024). In ecotoxicological experiments, twelve duplicates of each EUG concentration (0.2, 2, 20, 100, and 500 mg/L) were employed. Only water was used for the negative controls. The bulbs were placed in the top of each 15 mL tube and grown for 72 h at 25 °C in a dimly lit space. The tested solutions were renewed every 24 h. As in , using the XLSTAT (2014.5.03) program, the EC 50 and EC 10 values—effective concentrations of EUG that cause 50% and 10%, respectively, of root growth—as well as their Confidence Intervals (CI) were calculated from the dose–response curves for A. cepa .
EUG exhibits antimicrobial activity against various Gram-positive and Gram-negative bacteria, with observed MIC values ranging from 500 to 2000 μg/mL using the microdilution method. Our study is the first to report MIC data for any strain of B. subtilis , K. aerogenes, and P. aerogenes and it is also the first one to provide MIC values obtained through the microdilution method on specific strains never tested before of L. monocytogenes , S. agalactiae , S. aureus , A. baumannii , E. faecalis, and K. pneumoniae . No correlation between the bacterial type and the MIC value was observed. Combination studies with commercial ABXs (ERY, STM, GTM, CHL, TC, AMO, and AMP) using the checkboard assay and the kinetic bacterial growth inhibition study revealed significant findings. Six of the interactions resulted in synergy, particularly with GTM, CHL, STM, and TC, across both bacterial types, with MIC reductions ranging from 75% to 88%. Additive interactions predominated (39 out of 47), allowing for significant reductions in ABX consumption (50% to 98%), suggesting their importance for future research. Only two interactions were found to be antagonistic. Although an antimicrobial adjuvant is not expected to exhibit such activity, the moderate–low bacterial inhibition capacity of EUG should not be considered an impediment to its potential use as such an adjuvant. EUG–ABXs combination results show its remarkable ability to reduce the ABXs’ dose while maintaining their inhibitory efficacy on pathogenic bacteria. In terms of acute ecotoxicity, EUG demonstrated a lower impact on soil non-target organisms, namely E. foetida and A. cepa , compared to commercial ABXs. Additionally, its effect on water and soil microbiota, which are more indicative of potential ecosystem effects, was weaker than that of commercial ABXs. Notably, significant effects of EUG were observed only at 1000 mg/L, whereas commercial ABXs showed effects at lower concentrations. However, only aquatic organisms such as D. magna and V. fischeri displayed more harmful effects from EUG compared to commercial ABXs. Nevertheless, as less than 0.1% of EUG is excreted by humans without undergoing metabolization (in contrast to an average of 50% for commercial ABXs (as mentioned in —Soil non-target organisms), the use of EUG as being antimicrobial and/or an adjuvant poses minimal ecotoxicological risk when compared to commercial ABXs. In conclusion, our study demonstrates that EUG combined with commercial ABXs exhibits promising antimicrobial activity against clinically relevant Gram-positive and -negative bacteria, suggesting its potential as an antibiotic adjuvant. Moreover, our findings indicate that EUG presents minimal ecotoxicological risk to both terrestrial and aquatic environments, offering valuable insights for its potential use in antimicrobial therapies while considering environmental sustainability.
|
Genomic characterisation and traceability analysis of a | 6a4d7e8c-13ed-4d93-b816-7f805f41fc2f | 11887133 | Microbiology[mh] | Clostridium botulinum (C. botulinum) is an anaerobic bacterium commonly found in soil and the intestines of animals, typically existing in a spore form. Under suitable environmental conditions, it can proliferate and produce various toxins. Its pathogenicity primarily arises from the production of neurotoxins such as botulinum toxin and tetanus toxin, while the bacterium itself lacks invasive properties .Foodborne illness caused by botulinum toxins remains a significant public health concern, often linked to improperly processed or stored foods. Different types of C. botulinum produce various neurotoxins, which are the primary cause of foodborne botulism . C. botulinum is classified into seven serotypes (A-G), with types A, B, E, and F primarily associated with human botulism, particularly A and B in foodborne cases .Recent studies have reported food poisoning incidents linked to botulinum toxin in food products. PCR detection of type A neurotoxin in two commercial canned food samples from Turkey underscores the critical importance of stringent quality control in food production .Furthermore, genomic sequencing in Northern Italy linked type B botulism to industrial asparagus cream, demonstrating the effectiveness of genomic tools in tracing foodborne pathogens .In a case report from Shenzhen, China, type B1 C. botulinum was isolated from the patient’s intestinal wash , highlighting the necessity for rapid identification in clinical settings. Abroad, type E botulinum toxin is predominantly found in marine environments and coastal soils of high-latitude regions .Studies in Finland indicate a high prevalence of type E in sediments, particularly in Arctic communities consuming traditional fermented foods .Type E botulinum toxin remains stable and active at temperatures below 0 °C .In mainland China, type E botulism cases are primarily linked to undercooked pork, beef, and fermented soy products .Therefore, a deeper understanding of C. botulinum and its toxins is crucial for assessing their impact on food safety and related foodborne illness outbreaks. This study aimed to investigate the sources and characteristics of C. botulinum implicated in foodborne illness outbreaks. Our team collected samples of leftover pickled pig liver, household waste containing vomit, and hospital vomitus. We isolated and identified pathogenic strains and their toxins as type E using enrichment culture media for nucleic acid extraction. Type E botulism was rare in low-latitude inland areas, prompting comprehensive whole-genome sequencing and sequence analysis to elucidate the genetic features and evolutionary relationships of the strain. These methods determined the cause of the outbreak, revealing genomic characteristics, functional annotations, and virulence factors of the pathogenic strains, while comparative genomic analysis explored relationships among the strains. This research provided valuable insights into the characteristics of C. botulinum and its role in foodborne illness events.
Brief case description and sample collection The three samples analyzed in this study were collected from a single food poisoning incident involving three patients who consumed home-prepared pickled pig liver stored in a refrigerator for approximately one week. Within 24 h of consumption, the patients exhibited symptoms including dizziness, weakness, numbness of the tongue, slurred speech, blurred vision, double vision, and vomiting. Upon hospitalization, they developed clinical signs of pharyngeal and respiratory muscle paralysis. Despite the use of ventilators, one patient succumbed to the illness due to ineffective treatment. The samples collected included one portion of the remaining pickled pig liver, one sample of household waste from the deceased patient (mixed with vomit), and one sample of vomit from the patient who passed away in the hospital. Ethics approval and consent to participate This study was conducted in compliance with the Declaration of Helsinki ( http://www.wma.net/en/30publications/10policies/b3/index.html ) and relevant national ethical guidelines. Informed consent was obtained from the patients or their legal representatives prior to sample collection. The study protocol was approved by the Institutional Review Board (IRB) of the Ethics Committee of the Kunming Center for Disease Control and Prevention (Approval No. NO2020012). All personal identifiers were anonymized to protect patient privacy. Isolation and identification of pathogenic strains The three samples were enriched using cooked meat medium as follows, 25 g of each sample were placed in sterile homogenization bags, and 225 mL of physiological saline was added to each bag. The samples were homogenized for 2 min. A 2 mL aliquot of the homogenized supernatant was inoculated into 10 mL of cooked meat medium and incubated anaerobically at 35 °C for 5 days. The phenotypic characteristics of the isolated strains were observed microscopically. To further characterize the strains and their toxin types, nucleic acids were extracted from the enriched supernatant, followed by qualitative detection using a C. botulinum nucleic acid detection kit (fluorescent PCR method) (#SKY-8145, Shenzhen Bioscience Co., Ltd., Shenzhen, China). The kit used the following primers: upstream primer F: GAATAYACWATAATWAATTGYAT, and downstream primers R1: TATCCTGYAAAGTCCAGATTATY, R2: TCARTTAARGTCCATATTAT, and R3: TCTTGYAAWGTCCAAATTAT. To determine the toxin types of the strains, qPCR was performed using detection kits for botulinum toxin genes BonT/A, BonT/B, BonT/E, and BonT/F. The C. botulinum type A nucleic acid detection kit (fluorescent PCR method) (#SKY-8146, Shenzhen Bioscience Co., Ltd., Shenzhen, China) used primers F: GTGATACAACCAGATGGTAGTTAT and R: AAAAAACAAGTCCCAATTATTAACTTT. The type B kit (#SKY-8147) used primers F: GAGATGTTTGTGAATATTATGATCCAG and R: GTTCATGCATTAATATCAAGGCTGG. The type E kit (#SKY-8148) used primers F: TATCCAAAATGATGCTTATATACCAAA and R: GGCACTTTCTGTGCATCTAAATA. The type F kit (#SKY-8149) used primers F: GCTTCATTAAAGAACGGAAGCAGTGCT and R: GTGGCGCCTTTGTACCTTTTCTAGG. The qPCR procedure included 10 min of pre-denaturation at 95 °C, followed by 5 s of denaturation at 95 °C, 60 s of annealing at 60 °C, and 8 s of extension at 72 °C, repeated for 45 cycles. Whole-genome sequencing and sequence analysis The whole genome DNA of the C. botulinum strain was sequenced by using the MinlON and Illumina MiSeq platforms. The MinlON sequence reads were assembled and corrected by Illumina MiSeq with unicycler. The prediction of ORFs and their annotations were performed using Prokka v1.14.6. The BLASTn was served for further sequence comparisons. The Illumina short reads and MinlON long reads were mapped to the sequence of KM001 by Burrows-Wheeler Aligner(BWA v0.7.17-r1188), and the corresponding depths were calculated by Mosdepth v0.3.3. Functional annotation of genome and phylogenetic trees The gene sequences predicted were compared using BLAST analysis with various functional databases such as COG, KEGG, and GO to annotate gene functions. After summarizing the function of genes from three aspects: cell components, molecular functions and biological processes, the secondary classification of the top 20 GOslims with the most annotations under each classification is selected for drawing. We utilized the Virulence Searcher tool to predict virulence factors (VF) within the genome. The whole genome sequences of C. botulinum of different genotypes in 23 different countries or regions were selected from the GenBank module of NCBI website (Table ). The 23 bacterial genomes were compared by HomBlocks, and the sequence was pruned by Gblock method to obtain the collinear block for the construction of evolutionary tree. Finally, execute 1000 bootstrap replicates using RAxML (v8.2.9) to infer the evolutionary maximum likelihood (ML) tree of bacteria. Comparative genomic analysis The whole genomes of four strains(as shown in Table ) were compared with KM001 by using ORTHOFINDER v2.5.5 ( https://github.com/davidemms/orthofinder ). The genome comparative analysis circle diagram was by CGView Comparison Tool (CCT) . Covariance analysis of genomes of similar strains was using mauve ( http://darlingLab.org/mauve/ ). In the analysis of common and unique genetic traits, a gene family is defined as a group of genes derived from a common ancestor. Single-copy and multi-copy gene families are identified through the clustering of homologous genes. These families are generally conserved across species and serve to elucidate interspecies relationships. Additionally, species-specific genes and families can be identified, potentially linked to unique phenotypic traits, allowing for the examination of evolutionary relationships at the genomic level. Using ORTHOFINDER (parameters: -M msa -A mafft -T raxml), we performed genetic family cluster analysis of the generated genomes and related species. The results of the gene family analysis were classified, and the number of genes in each category was tallied.
The three samples analyzed in this study were collected from a single food poisoning incident involving three patients who consumed home-prepared pickled pig liver stored in a refrigerator for approximately one week. Within 24 h of consumption, the patients exhibited symptoms including dizziness, weakness, numbness of the tongue, slurred speech, blurred vision, double vision, and vomiting. Upon hospitalization, they developed clinical signs of pharyngeal and respiratory muscle paralysis. Despite the use of ventilators, one patient succumbed to the illness due to ineffective treatment. The samples collected included one portion of the remaining pickled pig liver, one sample of household waste from the deceased patient (mixed with vomit), and one sample of vomit from the patient who passed away in the hospital.
This study was conducted in compliance with the Declaration of Helsinki ( http://www.wma.net/en/30publications/10policies/b3/index.html ) and relevant national ethical guidelines. Informed consent was obtained from the patients or their legal representatives prior to sample collection. The study protocol was approved by the Institutional Review Board (IRB) of the Ethics Committee of the Kunming Center for Disease Control and Prevention (Approval No. NO2020012). All personal identifiers were anonymized to protect patient privacy.
The three samples were enriched using cooked meat medium as follows, 25 g of each sample were placed in sterile homogenization bags, and 225 mL of physiological saline was added to each bag. The samples were homogenized for 2 min. A 2 mL aliquot of the homogenized supernatant was inoculated into 10 mL of cooked meat medium and incubated anaerobically at 35 °C for 5 days. The phenotypic characteristics of the isolated strains were observed microscopically. To further characterize the strains and their toxin types, nucleic acids were extracted from the enriched supernatant, followed by qualitative detection using a C. botulinum nucleic acid detection kit (fluorescent PCR method) (#SKY-8145, Shenzhen Bioscience Co., Ltd., Shenzhen, China). The kit used the following primers: upstream primer F: GAATAYACWATAATWAATTGYAT, and downstream primers R1: TATCCTGYAAAGTCCAGATTATY, R2: TCARTTAARGTCCATATTAT, and R3: TCTTGYAAWGTCCAAATTAT. To determine the toxin types of the strains, qPCR was performed using detection kits for botulinum toxin genes BonT/A, BonT/B, BonT/E, and BonT/F. The C. botulinum type A nucleic acid detection kit (fluorescent PCR method) (#SKY-8146, Shenzhen Bioscience Co., Ltd., Shenzhen, China) used primers F: GTGATACAACCAGATGGTAGTTAT and R: AAAAAACAAGTCCCAATTATTAACTTT. The type B kit (#SKY-8147) used primers F: GAGATGTTTGTGAATATTATGATCCAG and R: GTTCATGCATTAATATCAAGGCTGG. The type E kit (#SKY-8148) used primers F: TATCCAAAATGATGCTTATATACCAAA and R: GGCACTTTCTGTGCATCTAAATA. The type F kit (#SKY-8149) used primers F: GCTTCATTAAAGAACGGAAGCAGTGCT and R: GTGGCGCCTTTGTACCTTTTCTAGG. The qPCR procedure included 10 min of pre-denaturation at 95 °C, followed by 5 s of denaturation at 95 °C, 60 s of annealing at 60 °C, and 8 s of extension at 72 °C, repeated for 45 cycles.
The whole genome DNA of the C. botulinum strain was sequenced by using the MinlON and Illumina MiSeq platforms. The MinlON sequence reads were assembled and corrected by Illumina MiSeq with unicycler. The prediction of ORFs and their annotations were performed using Prokka v1.14.6. The BLASTn was served for further sequence comparisons. The Illumina short reads and MinlON long reads were mapped to the sequence of KM001 by Burrows-Wheeler Aligner(BWA v0.7.17-r1188), and the corresponding depths were calculated by Mosdepth v0.3.3.
The gene sequences predicted were compared using BLAST analysis with various functional databases such as COG, KEGG, and GO to annotate gene functions. After summarizing the function of genes from three aspects: cell components, molecular functions and biological processes, the secondary classification of the top 20 GOslims with the most annotations under each classification is selected for drawing. We utilized the Virulence Searcher tool to predict virulence factors (VF) within the genome. The whole genome sequences of C. botulinum of different genotypes in 23 different countries or regions were selected from the GenBank module of NCBI website (Table ). The 23 bacterial genomes were compared by HomBlocks, and the sequence was pruned by Gblock method to obtain the collinear block for the construction of evolutionary tree. Finally, execute 1000 bootstrap replicates using RAxML (v8.2.9) to infer the evolutionary maximum likelihood (ML) tree of bacteria.
The whole genomes of four strains(as shown in Table ) were compared with KM001 by using ORTHOFINDER v2.5.5 ( https://github.com/davidemms/orthofinder ). The genome comparative analysis circle diagram was by CGView Comparison Tool (CCT) . Covariance analysis of genomes of similar strains was using mauve ( http://darlingLab.org/mauve/ ). In the analysis of common and unique genetic traits, a gene family is defined as a group of genes derived from a common ancestor. Single-copy and multi-copy gene families are identified through the clustering of homologous genes. These families are generally conserved across species and serve to elucidate interspecies relationships. Additionally, species-specific genes and families can be identified, potentially linked to unique phenotypic traits, allowing for the examination of evolutionary relationships at the genomic level. Using ORTHOFINDER (parameters: -M msa -A mafft -T raxml), we performed genetic family cluster analysis of the generated genomes and related species. The results of the gene family analysis were classified, and the number of genes in each category was tallied.
C. botulinum identified as the causative agent of food poisoning Laboratory analysis of the three samples was conducted using anaerobic culturing on meat broth at 35 °C for five days. All 3 samples were culture positive. All samples demonstrated black coloration, gas production, and a distinct foul odor. Microscopic examination revealed the presence of Gram-positive bacilli with oval spores, larger than the vegetative cells and located at the terminal end, displaying a characteristic “tennis racket” shape consistent with C. botulinum . Real-time PCR results confirmed the presence of C. botulinum and type E botulinum toxin genes in all three samples. The genes for type A, B, and F botulinum toxins were negative, with Ct values greater than 37.5. The amplification curves for the C. botulinum nucleic acid detection samples were shown in Fig. A, and the amplification curves for the BonT/E toxin gene detection were shown in Fig. B. According to guidelines for food poisoning diagnosis and management, combined with epidemiological investigations, clinical presentations, and laboratory findings, this incident was classified as bacterial food poisoning due to the consumption of home-prepared pickled pig liver contaminated with type E botulinum toxin. C. botulinum is a Gram-positive anaerobic bacterium that is relatively rare and difficult to survive in normal environments. Its pathogenicity is mainly due to botulinum toxin, a potent neurotoxin that inhibits the release of acetylcholine at the neuromuscular junction, causing paralysis of major muscle groups, including ocular and respiratory muscles . Botulinum toxin is estimated to be 10,000 times more toxic than potassium cyanide, with a lethal dose for humans around 0.1 micrograms . These results confirm that the incident was caused by the same strain of C. botulinum producing type E toxin, isolated from these samples. The strain from sample 3, named KM001, underwent next-generation sequencing for whole genome analysis, while no further sequencing was performed on the other two samples. Genomic characteristics Whole genome sequencing of the isolated strain revealed a genome size of 3,713,470 bp, with an average sequencing depth of 428.6x for the second-generation sequencing and 87x for the third-generation sequencing, achieving a genome coverage of 99.17%. The analysis predicted 7,239 genes, including 3,509 coding sequences (CDs), 79 tRNA genes, and 31 rRNA genes. The detailed genome characteristics are illustrated in Fig. , which presents the GC skew, GC content, and annotations for rRNA and tRNA. Whole genome functional annotation To explore the complex biological functions associated with these genes, KEGG annotation was conducted, resulting in the classification of 1,840 genes into 44 metabolic pathways as shown in Fig. A. The most frequently annotated pathways included signaling and cellular processes, metabolism, and genetic information processing. Furthermore, the GO annotation analysis, which categorized genes into cell composition, molecular function, and biological processes, indicated that the primary activities involve oxidation-reduction processes and metabolic functions, as depicted in Fig. B. The prediction results of genomic islands (GIs) reveal that there are 131 segments in the strain’s genome predicted as putative alien genes (PUTAL), with the maximum length reaching 3734 bp. Most of these segments are short fragments below 3000 bp. In the COG analysis, the majority of contigs were enriched in the S category (Signal Transduction) (Fig. .C). In the analysis of the virulence factors of C. botulinum , a total of several virulence-related elements were identified (Fig. .D). Among these, the bont was found to be the most prominent, accounting for 27% of all predicted virulence factors. Specifically, when comparing the bont within different strains, it constituted 8% of the virulence factors detected in C. botulinum E3 strain Alaska E43, with a coverage of 98.67% and a consistency of 92.83%. In contrast, in C. botulinum 202 F, the bont represented 13% of the virulence factors, showing a coverage of 98.64% and a consistency of 92.37%. Additionally, the bont aligned with other strains, including C. botulinum B strain Eklund 17B (NRP) and Clostridium butyricum. Furthermore, other virulence factors such as nheA , nheB , asbD , asbE , asbF and dhbB were also identified, all corresponding to various strains of Bacillus thuringiensis. Phylogenetic analysis of C. botulinum The amino acid difference between serotypes A to G ranges from 37 to 70% . We obtained 23 complete genome sequences of C. botulinum from the NCBI database, and together with KM001, constructed a phylogenetic tree based on whole-genome sequences. The KM001 strain detected in this study showed similarity to published E-type strains available on GenBank. As shown in Fig. , this strain is most closely related to two isolates from Canada, namely E1 str E1 Dolman (2022) and E1 str BoNT E Beluga (2009). The strains isolated from food samples in Canada (salmon roe, seal meat in oil, and pickled herring) all belong to E-type. However, C. botulinum samples isolated from food in Japan, the United States, and Canada exhibited varying serotypes: A, B, and E, respectively. These serotypes are the most common ones affecting humans, with types A and B frequently found in soil samples from North America and other continents, while type E is predominantly associated with marine or estuarine sediments and the intestines of fish and marine mammals. In Alaska, Canada’s Arctic regions, and northern Russia, Inuit populations traditionally store ‘fermented’ seal or whale blubber outdoors. Inadequate storage conditions can facilitate the germination and growth of spores in partially anaerobic environments . The distribution of botulinum types across different countries may be influenced by local food characteristics and processing practices. A total of five genomes were compared to analyze differences among related bacterial genomes (Fig. .A). In comparative genomic analyses (Table ), each strain contains a substantial number of genes, ranging from 3,160 to 3,523. Notably, the 202 F strain has the highest gene count (3,523), while the E3str strain exhibits the lowest (3,160). The number of single genes varies across the strains, with the highest being in the v2.5 strain (469), and unique genes are most prominent in the 202 F strain (344). Interestingly, Bstr935198 and E3str strains lack unique genes entirely. The percentage of unclustered genes also varies, with the highest number found in the v2.5 strain (510), indicating a diverse genomic landscape. The gene families are relatively consistent across strains(Fig. .B). The number of genes within families ranges from 2,997 in the E3str strain to 3,406 in the Bstr strain, highlighting differences in gene family composition. The average number of genes per family is relatively similar across the strains(Fig. .C), ranging from 1.04 to 1.1, suggesting a stable family structure despite variations in total gene count.
identified as the causative agent of food poisoning Laboratory analysis of the three samples was conducted using anaerobic culturing on meat broth at 35 °C for five days. All 3 samples were culture positive. All samples demonstrated black coloration, gas production, and a distinct foul odor. Microscopic examination revealed the presence of Gram-positive bacilli with oval spores, larger than the vegetative cells and located at the terminal end, displaying a characteristic “tennis racket” shape consistent with C. botulinum . Real-time PCR results confirmed the presence of C. botulinum and type E botulinum toxin genes in all three samples. The genes for type A, B, and F botulinum toxins were negative, with Ct values greater than 37.5. The amplification curves for the C. botulinum nucleic acid detection samples were shown in Fig. A, and the amplification curves for the BonT/E toxin gene detection were shown in Fig. B. According to guidelines for food poisoning diagnosis and management, combined with epidemiological investigations, clinical presentations, and laboratory findings, this incident was classified as bacterial food poisoning due to the consumption of home-prepared pickled pig liver contaminated with type E botulinum toxin. C. botulinum is a Gram-positive anaerobic bacterium that is relatively rare and difficult to survive in normal environments. Its pathogenicity is mainly due to botulinum toxin, a potent neurotoxin that inhibits the release of acetylcholine at the neuromuscular junction, causing paralysis of major muscle groups, including ocular and respiratory muscles . Botulinum toxin is estimated to be 10,000 times more toxic than potassium cyanide, with a lethal dose for humans around 0.1 micrograms . These results confirm that the incident was caused by the same strain of C. botulinum producing type E toxin, isolated from these samples. The strain from sample 3, named KM001, underwent next-generation sequencing for whole genome analysis, while no further sequencing was performed on the other two samples.
Whole genome sequencing of the isolated strain revealed a genome size of 3,713,470 bp, with an average sequencing depth of 428.6x for the second-generation sequencing and 87x for the third-generation sequencing, achieving a genome coverage of 99.17%. The analysis predicted 7,239 genes, including 3,509 coding sequences (CDs), 79 tRNA genes, and 31 rRNA genes. The detailed genome characteristics are illustrated in Fig. , which presents the GC skew, GC content, and annotations for rRNA and tRNA.
To explore the complex biological functions associated with these genes, KEGG annotation was conducted, resulting in the classification of 1,840 genes into 44 metabolic pathways as shown in Fig. A. The most frequently annotated pathways included signaling and cellular processes, metabolism, and genetic information processing. Furthermore, the GO annotation analysis, which categorized genes into cell composition, molecular function, and biological processes, indicated that the primary activities involve oxidation-reduction processes and metabolic functions, as depicted in Fig. B. The prediction results of genomic islands (GIs) reveal that there are 131 segments in the strain’s genome predicted as putative alien genes (PUTAL), with the maximum length reaching 3734 bp. Most of these segments are short fragments below 3000 bp. In the COG analysis, the majority of contigs were enriched in the S category (Signal Transduction) (Fig. .C). In the analysis of the virulence factors of C. botulinum , a total of several virulence-related elements were identified (Fig. .D). Among these, the bont was found to be the most prominent, accounting for 27% of all predicted virulence factors. Specifically, when comparing the bont within different strains, it constituted 8% of the virulence factors detected in C. botulinum E3 strain Alaska E43, with a coverage of 98.67% and a consistency of 92.83%. In contrast, in C. botulinum 202 F, the bont represented 13% of the virulence factors, showing a coverage of 98.64% and a consistency of 92.37%. Additionally, the bont aligned with other strains, including C. botulinum B strain Eklund 17B (NRP) and Clostridium butyricum. Furthermore, other virulence factors such as nheA , nheB , asbD , asbE , asbF and dhbB were also identified, all corresponding to various strains of Bacillus thuringiensis.
C. botulinum The amino acid difference between serotypes A to G ranges from 37 to 70% . We obtained 23 complete genome sequences of C. botulinum from the NCBI database, and together with KM001, constructed a phylogenetic tree based on whole-genome sequences. The KM001 strain detected in this study showed similarity to published E-type strains available on GenBank. As shown in Fig. , this strain is most closely related to two isolates from Canada, namely E1 str E1 Dolman (2022) and E1 str BoNT E Beluga (2009). The strains isolated from food samples in Canada (salmon roe, seal meat in oil, and pickled herring) all belong to E-type. However, C. botulinum samples isolated from food in Japan, the United States, and Canada exhibited varying serotypes: A, B, and E, respectively. These serotypes are the most common ones affecting humans, with types A and B frequently found in soil samples from North America and other continents, while type E is predominantly associated with marine or estuarine sediments and the intestines of fish and marine mammals. In Alaska, Canada’s Arctic regions, and northern Russia, Inuit populations traditionally store ‘fermented’ seal or whale blubber outdoors. Inadequate storage conditions can facilitate the germination and growth of spores in partially anaerobic environments . The distribution of botulinum types across different countries may be influenced by local food characteristics and processing practices. A total of five genomes were compared to analyze differences among related bacterial genomes (Fig. .A). In comparative genomic analyses (Table ), each strain contains a substantial number of genes, ranging from 3,160 to 3,523. Notably, the 202 F strain has the highest gene count (3,523), while the E3str strain exhibits the lowest (3,160). The number of single genes varies across the strains, with the highest being in the v2.5 strain (469), and unique genes are most prominent in the 202 F strain (344). Interestingly, Bstr935198 and E3str strains lack unique genes entirely. The percentage of unclustered genes also varies, with the highest number found in the v2.5 strain (510), indicating a diverse genomic landscape. The gene families are relatively consistent across strains(Fig. .B). The number of genes within families ranges from 2,997 in the E3str strain to 3,406 in the Bstr strain, highlighting differences in gene family composition. The average number of genes per family is relatively similar across the strains(Fig. .C), ranging from 1.04 to 1.1, suggesting a stable family structure despite variations in total gene count.
In this study, we enriched and isolated C. botulinum from a foodborne illness outbreak, followed by real-time quantitative PCR detection and toxin typing for strain identification and whole-genome sequencing. A phylogenetic tree was constructed using 23 C. botulinum strains from the NCBI database. The results indicated that three individuals developed neurological symptoms, including dizziness, weakness, tongue numbness, slurred speech, blurred vision, and vomiting, after consuming pickled pig liver stored in a refrigerator for a week. The presence of C. botulinum and type E botulinum toxin in the pickled liver was confirmed, with the potent neurotoxin inhibiting acetylcholine release at the neuromuscular junction and causing paralysis of major muscle groups, including ocular and respiratory muscles .Type E botulism cases in mainland China are primarily reported from the Tibetan Plateau, where raw meat from autumn slaughtering of cattle, sheep, pigs, and marmots is susceptible to spore contamination and toxin production during storage. The region’s typical food preparation methods and lower boiling points at high altitudes further hinder toxin inactivation, along with the local practice of consuming raw meat, which increases the risk of botulism .Therefore, understanding the contamination pathways of C. botulinum in food and preventive measures is crucial. Pickled pig liver stored in a refrigerator for a week can serve as an optimal growth medium for C. botulinum under suitable conditions, resulting in toxin production. Strict adherence to hygiene standards during meat slaughtering and handling is essential. In early environmental studies, samples from Qinhuangdao, Shanghai, and Hainan Island showed that 3 out of every 5 tested positive for type E botulinum neurotoxin. Additionally, among 302 soil samples from the Tibetan Plateau (Qinghai and Tibet), 28 were positive, while 5 out of 249 soil samples from the Yunnan-Guizhou Plateau (Guizhou, Guangxi, and Yunnan) also detected C. botulinum .This indicates that type E C. botulinum is widespread in soil environments without clear regional characteristics, highlighting the importance of timely diagnosis and effective treatment for patients exhibiting typical symptoms. However, clinical diagnostic clues are often insufficient, as conditions like Guillain-Barré syndrome and myasthenia gravis may present similarly to botulism .Special tests are necessary to rule these out. Definitive diagnosis can be made by detecting botulinum toxin in food, gastric or intestinal contents, vomit, or feces. Most hospitals lack comprehensive commercial diagnostic kits such as PCR or ELISA and do not routinely use specialized culture media. In cases of severe symptoms or outbreaks, metagenomic next-generation sequencing (mNGS) can be employed for pathogen detection without predefined infection targets , thereby enhancing the evidence for diagnosis. Subsequently, we analyzed the functional genes of C. botulinum . Through whole-genome sequencing and functional annotation, the primary genes of this strain were found to be involved in metabolism, signal transduction, and genetic information processing. The functions of these genes elucidate the bacterium’s cellular behavior and survival strategies. KEGG annotation revealed that 1,840 genes were categorized into 44 metabolic pathways, with the highest enrichment observed in signaling and cellular processes as well as metabolism. This indicates that C. botulinum possesses a complex metabolic capacity, which is linked to its adaptability to various environments and pathogenic potential, enabling it to utilize a broader range of nutritional sources and enhancing its survival and toxin production in diverse ecological niches . However, this may also be related to the growth phase of the laboratory-cultured strains. A proteomic study of C. botulinum indicated that the most significantly expressed proteins between the exponential and stationary phases are enriched in functions related to metabolic processes. Proteins predominant in the exponential phase include ribosomal proteins and those involved in acetate fermentation and other growth-related functions such as transcription and translation. In contrast, the stationary phase is characterized by butyrate fermentation, amino acid metabolism, and the expression of neurotoxins in pathogenic strains . It is likely that our cultured strains are primarily in the exponential phase. GO annotation analysis further suggests that the main cellular activities involve redox processes and metabolic functions. In the analysis of C. botulinum virulence factors, bont accounts for 27% of the predicted virulence factors, with 3% of the bont sequences aligning with Clostridium butyricum, which is also capable of producing BoNT/E .Plasmids containing bont can be transferred through conjugation among different types of C. botulinum, even to non-toxic Clostridium strains . Virulence genes reflect genomic similarities and potential evolutionary relationships among different serotypes and related species. Comparative genomics reveal variations in genome size and gene composition among closely related strains, which may relate to their adaptability to different environments or hosts. The number of genes within gene families remains relatively stable (between 1.04 and 1.1). Although genomic sizes vary among different strains, the fundamental structure of gene families is conserved , possibly reflecting the core biological requirements of C. botulinum . This study has certain limitations. Firstly, only samples from fatal cases associated with the foodborne illness outbreak were isolated, which does not reflect the overall strains responsible for the poisoning. Secondly, the whole-genome analysis lacked functional validation of virulence factors and did not address evolutionary differences among C. botulinum strains from different regions. Additionally, the mechanisms by which virulence factors contribute to the pathogenicity of C. botulinum remain unclear.
This study analyzed samples from a foodborne illness outbreak, identifying C. botulinum as the causative agent. The strain was isolated, identified, subjected to whole-genome sequencing, and functionally analyzed. The findings provide significant insights into the biological characteristics, pathogenicity, and evolutionary relationships of C. botulinum . These results hold important implications for future food safety regulations, disease diagnosis and prevention, and microbiological research. However, this study also has certain limitations that require further exploration in subsequent research.
|
Phage biocontrol success of bacterial wilt depends on synergistic interactions with resident rhizosphere microbiota | 5ab80f53-7606-4fc5-9a7b-79398df545f7 | 11561305 | Microbiology[mh] | It is expected that the human population will reach approximately 10 billion people by 2050, resulting in an 80%–110% increase in the demand for food. To achieve this, it is crucial to reduce global losses to plant pathogens (Strange & Scott, ), especially in the face of climate change, which could affect pathogen virulence and disease development (Singh et al., ). Ralstonia solanacearum is one of the most destructive bacterial phytopathogens worldwide (Mansfield et al., ; Wang, Luo, et al., ), being a major threat to agriculture and food security. It can infect more than 250 plant species and is especially deadly to solanaceous plants, causing brown rot in potatoes and bacterial wilt disease in tomatoes (Elphinstone, ). Unfortunately, we still lack effective control measures against R. solanacearum , as traditional agronomic practices such as crop rotation and grafting (Lemaga et al., ), or chemical control methods (Nicolopoulou‐Stamati et al., ), have proven ineffective. Novel, effective and environmentally friendly control methods are hence urgently required. During the last decade, microbial biocontrol methods have shown promise in controlling R. solanacearum (Álvarez et al., ; Hu et al., ; Narasimha Murthy et al., ; Wei et al., ). These methods are often based on pathogen suppression by antagonistic microorganisms, such as bacteria and fungi, that compete for shared resources with the pathogen or produce antimicrobials that suppress pathogen growth (Beneduzi et al., ; Hu et al., ; Narasimha Murthy et al., ). Recently, the use of pathogen‐specific viruses, bacteriophages (phages for short), to biocontrol phytopathogenic bacteria has received more interest (Cao et al., ; Wamani et al., ; Yamada et al., ; Nion et al., ). Phages are efficient at killing bacteria (Wang et al., , ), which could provide additional time for the plant host immune system to combat the pathogen infection. Moreover, phages are self‐replicating and self‐limiting, only increasing their abundance when their host bacterium is present. Despite these advantages, phage biocontrol outcomes are variable (Hong et al., ; Wang et al., , ; Wang, Wang, et al., ), and it is not clear how phages should be applied to achieve more consistent biocontrol efficacy. Previous research has shown that the use of phage combinations and repeated application of phage can attain more efficient phage biocontrol results (Kalpage & De Costa, ; Wang et al., , ) and improve R. solanacearum density control in the rhizosphere. Furthermore, it has been shown that phages can have synergistic effects with rhizospheric bacteria, resulting in improved R. solanacearum suppression when applied together (Wang et al., , , ). While such phage‐bacteria synergies have previously been linked with increased phage application frequency (Wang et al., ), it is unclear if interactions with the existing rhizosphere microbiota depend on the identity of phages or whether phages are applied individually or in combinations. In this work, we tested if phage combinations are more effective at controlling bacterial wilt disease than single phages and if the R. solanacearum suppression could be linked with shifts in the suppressiveness of resident rhizosphere bacterial microbiota. In total, we used four phages that originated from the United Kingdom and China and applied them in all single‐phage and two‐phage combinations to control the R. solanacearum strain UW551 in the lab and greenhouse experiments (only 2‐phage combinations were used to keep the experimental design manageable). In addition to tracking bacterial wilt disease progression, we analysed changes in the bacterial community diversity and composition at the end of the greenhouse experiment and directly tested if any of the enriched rhizobacterial taxa showed pathogen suppression in vitro and in planta. Together, our results suggest that interactions between biocontrol phages and resident rhizosphere bacteria can result in variation in phage biocontrol outcomes. Bacterial and phage strains and preparation of stock cultures We used R. solanacearum UW551 strain as the model pathogen in this study, which was ordered from the National Culture Collection of Plant Pathogenic Bacteria ( https://www.fera.co.uk/ncppb ). Four phages were selected from the University of York collection for this study (Table ). Three phages were isolated from the UK river water samples using R. solanacearum strain UW551 as the host bacterium; R. solanacearum and its phages are naturally found in the UK rivers (Elphinstone & Matthews‐Berry, ). In addition, one phage originating from a tomato field in China was used in this study (Table ). Bacterial stocks were prepared by growing the bacteria in liquid CPG media (casamino acid 1 g/L, peptone 10 g/L, glucose 5 g/L, Merck, Gillingham, UK) for 72 h at 28°C with shaking (100 rpm) and preparing glycerol stocks (30% glycerol) that were stored at −80°C. Phage stocks were prepared by adding 60 μL phage supernatant stored at 4°C and 3 mL of UW551 bacterium at OD = 0.2, to 600 mL CPG into 1 L Erlenmeyer flasks. The cocultures were grown for 96 h at 28°C with shaking (100 rpm), centrifuged to pellet the bacteria (10 min at 5000 G) and the supernatant filtered through 0.2 μm filters to collect bacteria‐free phage lysate for the experiments (stored at 4°C in glass vials). Phage DNA extraction, genome assembly and annotation The phage DNA isolation kit (Norgen, Biotek corporation, Thorold, ON, Canada) was used to extract phage DNA for sequencing, following the manufacturer's instructions with the following modifications: 15 μL of 20 units Norgen DNase I was added to the phage supernatants and incubated at room temperature for 30 min; DNase I was inactivated by incubating at 75°C for 10 min; 10 μL of 20 μg/mL proteinase K was added to the phage supernatants and incubated for 1 h at 55°C; phage supernatants were incubated with lysis buffer at 65°C for 30 min; DNA was eluted into 50 μL nuclease‐free water. A Qubit high sensitivity dsDNA kit was used to measure the amount of extracted DNA. The samples were sequenced by MicrobesNG services using Illumina Novaseq 6000 with a 250‐bp paired‐end strategy. Raw reads were trimmed by MicrobesNG using Trimmomatic (Bolger et al., ) and the quality was assessed using in‐house scripts combined with the following software: Samtools, BedTools and bwa‐mem (Li & Durbin, ). Phage genomes were assembled following the pipeline outlined in Shen and Millard and Turner et al. . Briefly, Cutadapt (v2.3) (Martin, ) was also used to trim reads with less than 100 bp, and fastqc (v0.11.4) was used to check the read quality. After that, Seqtk (v 1.3; Li, ) was used to subsample paired‐end reads, so that genome assemblies would have around 100× coverage. Phage genomes were then subsequently assembled with the subsampled reads using SPAdes (Prjibelski et al., ). Appendix , Tables and include more information regarding the phage genome assembly. Raw data and genome assemblies are available in NCBI Sequence Read Archive (SRA) BioProject PRJNA1076092. Analysing genetic differences between phages based on MASH distance and orthologous genes Genetic distance between the four phages was analysed by MASH using sketch and triangle to find the distance based on pairwise comparisons. Amino acid sequences were used in the Synteny Imaging tool (SYNIMA) (Farrer, ) pipeline using Orthofinder. After that, kinfin (version 3 [Laetsch & Blaxter, ]) was used to calculate the number of orthogroups shared between the different phages, which were visualised using the R package UpSetR. Based on mash distances, high similarity between PY04 and PY065 phages was found (0.00134, Table ), while the rest of the pairwise comparisons showed mash distances of 1, indicating that these phages were completely different from each other based on the genome sequence data. Further SYNIMA analysis suggested that a missing HNH protein gene might affect the packaging and biocontrol activity of PY059 (See Text for more detailed genome comparisons between the phages). Testing phage infectivity in vitro To quantify the infectivity of phages on the UW551 host, both spotting and liquid assays were performed using all four phages individually (Wang et al., ). Liquid cultures were set up in a 96‐well plate by inoculating 170 μL of fresh CPG with 20 μL R. solanacearum OD = 0.1 (1E+02 CFUs/mL) and 10 μL of phages at 1E+06 PFUs/mL. Plates were incubated for 72 h at 28°C without shaking, and bacterial growth was measured as optical density (OD 600 nm) at 2‐h intervals. Phage infectivity was calculated by measuring bacterial growth inhibition by the phage relative to bacteria‐only cultures (UW551 + 10 μL of sterile H 2 0), using area under the growth curves derived from the whole growth curve. Construction of phage combination treatments for bacterial wilt biocontrol experiments in planta To compare the biocontrol efficiency of the phages, we tested the efficacy of each phage individually and in pairwise combination with all others, resulting in a total of four individual and six pairwise phage treatments. We used the Solanum lycopersicum (tomato) variety Moneymaker (Moles Seeds, Colchester, UK) in all plant experiments, due to its high susceptibility to R. solanacearum (Ji et al., ). Tomato seeds were sterilised in 1% bleach and rinsed with abundant water, after which they were directly grown in pots (approximately 51 × 48 × 50 mm) with approximately 120 g of John Innes No. 2 compost (free of Ralstonia ) for 3 weeks at 24–26°C (16 h light/8 h dark—with lights on when ambient light intensity went below 90 W.m −2 and off if ambient light intensity went above 250 W.m −2 ) at the quarantine glasshouses at Fera Science Ltd. (York, UK). The plants were watered daily except for the day before UW551 inoculation. The plants were inoculated with R. solanacearum strain UW551 after 3 weeks of growth, and phages were added to the phage treatment plants 1 day after the bacterial inoculation. The R. solanacearum strain was prepared for inoculation following the protocols reported by Khokhani et al. . First, we used CPG to grow the bacterium at 28°C for 72 h (100 rpm). One millilitre of culture was inoculated into 400 mL fresh CPG and maintained at 28°C for a further 24 h (100 rpm). After this, another 150 mL of fresh CPG was added, and cultures were grown for another 24 h in the same conditions. The OD 600 of the final bacterial culture was adjusted to 0.7 (approximately 8E+09 cells/mL; the pellet was not washed before application), and a total of 5 mL of bacterial culture was inoculated onto each plant via root drench by applying the inoculum close to the plant. The phage inoculum was prepared as described previously and diluted to the concentration of 1E+06 PFU/mL using sterile water. In the case of 1‐phage treatments, 5 mL of each phage solution was used, and with 2‐phage treatments, 2.5 mL of each phage was used to keep the total phage numbers the same between the treatments (therefore, a total phage volume of 5 mL was applied to all phage treatments). The bacterium‐only treatment was inoculated with 5 mL of sterile water instead of phage inoculum (UW551‐only) and the negative control with no microbial inoculum received 5 mL of CPG (simulates bacterial inoculation effect) followed by 5 mL of sterile water (simulating phage inoculation). A total of 12 independent tomato plants were used for every treatment (all plants were grown individually on separate trays). Disease progression was measured at the onset of disease, from 10 days post‐inoculation (dpi) to 21 dpi, using the disease index score (DI), which varies on a scale from 0 to 4, where 0 denotes healthy plants without symptoms, while indexes between 1 and 4 were used for plants with bacterial wilt symptoms present on 25%, 50%, 75% and 100% of the aerial parts, respectively. Rhizosphere soil collection for microbial community analysis at the end of the greenhouse experiment To assess the indirect effects of phages on soil rhizosphere bacteria, we explored bacterial species diversity and community composition at the end of the greenhouse experiment. The rhizosphere samples were collected at 21 dpi and processed on the same day: Plants were cut as close to the soil as possible, and the soil of the whole pot was homogenised. For each sample, approximately 2 g of the soil was stored at −80°C and were later used for DNA extraction using method 3 of the protocols described previously (Porteous & Armstrong, ; Tien et al., ). DNA concentrations were assessed with Nanodrop and stored at −80°C until further processing. Additional 2‐g subsamples of extracted soils were stored as 30% glycerol stocks for bacterial isolation, while another 2‐g subsamples were homogenised in 20 mL of sterile water, vortexed for 3 min and allowed to settle for 30 min to obtain soil washes. Ten millilitre of the soil washes were passed through a 0.2 μm filter (Sarstedt, Nümbrecht, Germany) and stored at 4°C for quantification of phage densities. Quantifying phage densities in rhizosphere soil samples Phage densities were quantified as PFU/mL using the double agar overlay plaque assay (Kropinski et al., ). After diluting samples on 96‐well microplates using three technical replicates per sample, 300 μL of UW551 bacterial cultures and 180 μL of all diluted phage suspensions were mixed with 10 mL of warm CPG soft agar. The suspensions were mixed by gently inverting the tubes six times; then, they were immediately poured over a round CPG agar plate and left to solidify. Plates were incubated at 28°C for 24 h, and subsequently, phage densities were determined as PFU/mL based on the number of visible plaques observed on the agar plates (Table ). Rhizosphere bacterial community analysis To compare changes in bacterial community composition and diversity between treatments at the end of the greenhouse experiment, 16S rRNA sequencing was performed by Novogene using a PE250 strategy on an Illumina NovaSeq6000 using the primers Forward GTGCCAGCMGCCGCGGTAA and Reverse GGACTACHVGGGTWTCTAAT targeting V4 region of the 16S ribosomal RNA. After first assessing the product sizes via PCR and gel runs, the same amount of PCR product was pooled for each sample, end‐repaired, A‐tailed and ligated to Illumina adapters. The library preparation and initial bioinformatic analyses were prepared by Novogene, where the raw data were spliced, and filtered to obtain clean sequence data. DADA2 (Callahan et al., ) was used to reduce noise and to obtain the final amplicon sequence variants (ASVs). Each ASV was then annotated to obtain species information using the SILVA database (Quast et al., ) and QIIME2's classify‐sklearn algorithm (Bokulich et al., ), using a pretrained Naive Bayes classifier. Each ASV was annotated at the kingdom, phyla, class, order, family, genus and species levels. Raw data are available in NCBI Sequence Read Archive (SRA) BioProject PRJNA1010659. We obtained an average of 119,002 raw read pairs for the 154 samples (after removing PY04 1‐phage treatment and one from the PY04 + PY045 2‐phage treatment due to poor quality) with an average length of 253 nucleotides, a GC content of 53.51%, Q20 of 99.55% and Q30 of 98.22% (Table ). To obtain rhizosphere bacterial density estimates based on sequence data, the amount of extracted DNA was normalised with the weight of the soil used for the DNA extraction per sample. Absolute abundance was calculated (Table ) and the taxonomy assignment of the ASVs was represented in Table . Pathogen abundances between treatments were calculated after rarefying the data by grouping together the two ASVs (ASV1 and ASV240) identified as Ralstonia (Tables and ). A Phyloseq object was created using the package phyloseq in R using the normalised absolute abundance, the taxa file and the metadata (total of 55,575 taxa and 154 samples). Rarefaction was performed using the function rarefy_even_depth after filtering (2747 taxa). To separate phage effects from other variables, we also sequenced the bacterial communities from the negative control plant replicates ( N = 12) which were not treated with R. solanacearum or phages and positive control plants that were inoculated with R. solanacearum only ( N = 12). Random forest analysis to identify bacterial taxa in shifted rhizosphere microbiomes After observing a shift in the bacterial microbiome composition in some of the plant replicate samples, we identified the specific taxa associated with these shifted communities. To achieve this, we performed a random forest analysis using the data of the samples treated with phages (1‐ and 2‐phage treatments, for a total of 118 plants, not including negative or positive controls) and categorised them into ‘shifted’ and ‘centred’ samples according to their position in the NMDS1 axis (as explained in the next section). The analysis was performed using the rarefied absolute count data (2747 taxa). The training set was created randomly by sampling 70% of all samples (80), while the remaining samples were used as the test data set to assess how well the model predicted the data (38 samples). The function randomForest from the package randomForest in R was used with default parameters, proximity = TRUE and 1000 trees. Average Mean Gini values were calculated for each bacterial genus to evaluate their relative importance in explaining the shift in microbiome composition. Isolation of bacteria from the end‐point rhizosphere samples and assessment of their biocontrol activity To experimentally validate if bacteria associated with shifted microbiomes could explain the suppression of R. solanacearum , 100‐fold dilutions of soil washes derived from both diseased and healthy plants (Table ) were plated onto TSA (tryptone‐soy agar), CPG agar and SMSA media which is a semi‐selective medium for R. solanacearum (Elphinstone et al., ). Plates were incubated at 28°C for 48 h, and single colonies streaked onto new media to isolate individual bacteria. Colony‐PCR was performed to identify the genera of >100 colonies by transferring a loop of the colony to 100 μL of sterile water, which was then boiled for 10 min and centrifuged before performing PCRs. The 16S rRNA sequence was amplified using the same primers that were used for the metabarcoding, and PCR products were purified using the Wizard® SV Gel and PCR Clean‐Up System (Promega, Southampton, UK) following the manufacturer's instructions. The products were quantified using a Nanodrop 2000, and Sanger‐sequenced using the forward primer. Based on the high Gini values obtained in the RF analysis, we focused on the following bacterial genera: four isolates assigned to Burkholderia‐Caballeronia‐Paraburkholderia sp., one Bacillus sp. isolate, two Pseudomonas sp. and two Rhodanobacter sp. isolates (Table ). Despite several attempts, we failed to isolate Sphingomonas bacteria from the samples even though it was one of the most abundant taxa in the shifted microbiomes based on the sequencing data. To test if these isolated taxa had a positive or negative effect on R. solanacearum , in vitro and in vivo assays were performed. We first assessed the direct inhibition of R. solanacearum growth with a spot assay using a method similar to phage density measurements; instead of spotting phages, four technical replicates of 10 μL of selected bacterial isolates (Table ) were inoculated on top of R. solanacearum lawns and pathogen inhibition was quantified as the diameter of inhibition halos observed on the plates. Liquid assays were also performed using a GFP‐tagged R. solanacearum UW551 strain obtained by transformation with the pCOMP‐PhII to introduce the recombination regions of the pRC plasmid, and then, the p‐COMP transformed colonies were used for a second transformation with pRCG‐Pps‐GFPuv (Cruz et al., ; Monteiro et al., ). In these assays, R. solanacearum growth was measured on 96‐well plates alone or in the presence of each of the 9 isolated rhizosphere taxa. All the strains were initially grown in CPG media for 48 h at 28°C with shaking (100 rpm) and then diluted to an OD = 0.1. Monocultures were prepared by inoculating 20 μL of bacteria cultures with 180 μL of CPG while cocultures with 20 μL of each bacterium were inoculated in 160 μL CPG ( N = 3 replicates per treatment). Pathogen relative density was assessed based on GFP fluorescence signal intensity at regular intervals for 48‐h growth period (excitation at 405 nm and emission at 509 nm). The potential biocontrol effects of a subset of bacterial taxa were tested in planta ( Pseudomonas P19, Rhodanobacter R55, Burkholderia B12, Burkholderia PB18). Plants were grown for 3 weeks at 20°C (±2°C) with 14 h light/10 h dark cycle and moved to PHCbi growth cabinets (MLR‐352) 3 days before bacterial inoculations to let plants adapt to the new environment of 24°C/20°C temperature and 16/8 h light/dark cycles. All bacterial inoculants were prepared as previously described. Plant inoculations were performed using the same density of R. solanacearum as in the greenhouse experiments (OD 600 = 0.7; and 5 mL) while the four potential biocontrol candidate strains were inoculated at a density of OD 600 = 0.35 (5 mL). Three‐week‐old plants were inoculated with one bacteria (5 mL of one bacterium) or both bacteria using 5 mL of each. Negative controls included plants grown without inoculated microbes and all treatments were replicated eight times. Analysing the importance of phages and resident bacterial taxa for disease development using partial least squares path modelling To understand the interactions between phages, resident rhizosphere bacterial communities and R. solanacearum abundances on disease development, we used a partial least squares path modelling (PLS‐PM) using the R package plspm. The cross‐loading between latent variables as well as the unidimensionality in the latent variables was tested during model construction. The final model was created then by assessing ‘Phages’ (reflective indicator included as the log (PFU/mL)), ‘Microbiome’ (reflective indicator included as NMDS1 which divided the samples into shifted and centred microbiomes), ‘ Ralstonia ’ (reflective indicator included as the pathogen abundance) and ‘Disease’ (reflective indicator included as two different variables: Disease index at 21 dpi and area under the diseased curve). To compare if there were differences between shifted and centred microbiome samples, the function plspm.groups was used with the ‘bootstrap’ method with 500 repetitions. Statistical analysis To assess mean differences between treatment groups, ANOVA or Kruskal–Wallis tests (based on the normality of the data) were used followed by Tukey's HSD test or Dunn's post hoc tests, respectively (package rstatix and FSA in R), and Bonferroni tests were used to adjust for multiple comparisons. The t _test function in rstatix package was used to compare only two groups. Pearson correlation was used to assess correlation between continuous variables. To measure the area under the disease progression curve or GFP fluorescence, we used the AUC function from the package pROC in R. We analysed the diversity of rhizosphere bacterial communities estimated as the number of ASVs, ASVs richness, Chao1 and Shannon diversity index (at ASV level) using the vegan package (Community Ecology Package [R package vegan version 2.6–4], 2022). Beta diversity was assessed with a non‐metric multidimensional scaling (NMDS) using the metaMDS function in the R vegan package which uses Bray–Curtis distances. NMDS1 values were used to divide the samples into ‘shifted’ (NMDS1 > 0) and ‘centred’ (NMDS1 < 0) groups. To check pairwise differences, we used pairwise Permutational Multivariate Analysis of Variance Using Distance Matrices (adonis) using the pairwise.adonis function from the package pairwiseAdonis in R based on Bray–Curtis distances with 999 permutations and p ‐values adjusted using Holm's method. All R analyses were conducted with version 4.3.2. We used R. solanacearum UW551 strain as the model pathogen in this study, which was ordered from the National Culture Collection of Plant Pathogenic Bacteria ( https://www.fera.co.uk/ncppb ). Four phages were selected from the University of York collection for this study (Table ). Three phages were isolated from the UK river water samples using R. solanacearum strain UW551 as the host bacterium; R. solanacearum and its phages are naturally found in the UK rivers (Elphinstone & Matthews‐Berry, ). In addition, one phage originating from a tomato field in China was used in this study (Table ). Bacterial stocks were prepared by growing the bacteria in liquid CPG media (casamino acid 1 g/L, peptone 10 g/L, glucose 5 g/L, Merck, Gillingham, UK) for 72 h at 28°C with shaking (100 rpm) and preparing glycerol stocks (30% glycerol) that were stored at −80°C. Phage stocks were prepared by adding 60 μL phage supernatant stored at 4°C and 3 mL of UW551 bacterium at OD = 0.2, to 600 mL CPG into 1 L Erlenmeyer flasks. The cocultures were grown for 96 h at 28°C with shaking (100 rpm), centrifuged to pellet the bacteria (10 min at 5000 G) and the supernatant filtered through 0.2 μm filters to collect bacteria‐free phage lysate for the experiments (stored at 4°C in glass vials). The phage DNA isolation kit (Norgen, Biotek corporation, Thorold, ON, Canada) was used to extract phage DNA for sequencing, following the manufacturer's instructions with the following modifications: 15 μL of 20 units Norgen DNase I was added to the phage supernatants and incubated at room temperature for 30 min; DNase I was inactivated by incubating at 75°C for 10 min; 10 μL of 20 μg/mL proteinase K was added to the phage supernatants and incubated for 1 h at 55°C; phage supernatants were incubated with lysis buffer at 65°C for 30 min; DNA was eluted into 50 μL nuclease‐free water. A Qubit high sensitivity dsDNA kit was used to measure the amount of extracted DNA. The samples were sequenced by MicrobesNG services using Illumina Novaseq 6000 with a 250‐bp paired‐end strategy. Raw reads were trimmed by MicrobesNG using Trimmomatic (Bolger et al., ) and the quality was assessed using in‐house scripts combined with the following software: Samtools, BedTools and bwa‐mem (Li & Durbin, ). Phage genomes were assembled following the pipeline outlined in Shen and Millard and Turner et al. . Briefly, Cutadapt (v2.3) (Martin, ) was also used to trim reads with less than 100 bp, and fastqc (v0.11.4) was used to check the read quality. After that, Seqtk (v 1.3; Li, ) was used to subsample paired‐end reads, so that genome assemblies would have around 100× coverage. Phage genomes were then subsequently assembled with the subsampled reads using SPAdes (Prjibelski et al., ). Appendix , Tables and include more information regarding the phage genome assembly. Raw data and genome assemblies are available in NCBI Sequence Read Archive (SRA) BioProject PRJNA1076092. Genetic distance between the four phages was analysed by MASH using sketch and triangle to find the distance based on pairwise comparisons. Amino acid sequences were used in the Synteny Imaging tool (SYNIMA) (Farrer, ) pipeline using Orthofinder. After that, kinfin (version 3 [Laetsch & Blaxter, ]) was used to calculate the number of orthogroups shared between the different phages, which were visualised using the R package UpSetR. Based on mash distances, high similarity between PY04 and PY065 phages was found (0.00134, Table ), while the rest of the pairwise comparisons showed mash distances of 1, indicating that these phages were completely different from each other based on the genome sequence data. Further SYNIMA analysis suggested that a missing HNH protein gene might affect the packaging and biocontrol activity of PY059 (See Text for more detailed genome comparisons between the phages). To quantify the infectivity of phages on the UW551 host, both spotting and liquid assays were performed using all four phages individually (Wang et al., ). Liquid cultures were set up in a 96‐well plate by inoculating 170 μL of fresh CPG with 20 μL R. solanacearum OD = 0.1 (1E+02 CFUs/mL) and 10 μL of phages at 1E+06 PFUs/mL. Plates were incubated for 72 h at 28°C without shaking, and bacterial growth was measured as optical density (OD 600 nm) at 2‐h intervals. Phage infectivity was calculated by measuring bacterial growth inhibition by the phage relative to bacteria‐only cultures (UW551 + 10 μL of sterile H 2 0), using area under the growth curves derived from the whole growth curve. To compare the biocontrol efficiency of the phages, we tested the efficacy of each phage individually and in pairwise combination with all others, resulting in a total of four individual and six pairwise phage treatments. We used the Solanum lycopersicum (tomato) variety Moneymaker (Moles Seeds, Colchester, UK) in all plant experiments, due to its high susceptibility to R. solanacearum (Ji et al., ). Tomato seeds were sterilised in 1% bleach and rinsed with abundant water, after which they were directly grown in pots (approximately 51 × 48 × 50 mm) with approximately 120 g of John Innes No. 2 compost (free of Ralstonia ) for 3 weeks at 24–26°C (16 h light/8 h dark—with lights on when ambient light intensity went below 90 W.m −2 and off if ambient light intensity went above 250 W.m −2 ) at the quarantine glasshouses at Fera Science Ltd. (York, UK). The plants were watered daily except for the day before UW551 inoculation. The plants were inoculated with R. solanacearum strain UW551 after 3 weeks of growth, and phages were added to the phage treatment plants 1 day after the bacterial inoculation. The R. solanacearum strain was prepared for inoculation following the protocols reported by Khokhani et al. . First, we used CPG to grow the bacterium at 28°C for 72 h (100 rpm). One millilitre of culture was inoculated into 400 mL fresh CPG and maintained at 28°C for a further 24 h (100 rpm). After this, another 150 mL of fresh CPG was added, and cultures were grown for another 24 h in the same conditions. The OD 600 of the final bacterial culture was adjusted to 0.7 (approximately 8E+09 cells/mL; the pellet was not washed before application), and a total of 5 mL of bacterial culture was inoculated onto each plant via root drench by applying the inoculum close to the plant. The phage inoculum was prepared as described previously and diluted to the concentration of 1E+06 PFU/mL using sterile water. In the case of 1‐phage treatments, 5 mL of each phage solution was used, and with 2‐phage treatments, 2.5 mL of each phage was used to keep the total phage numbers the same between the treatments (therefore, a total phage volume of 5 mL was applied to all phage treatments). The bacterium‐only treatment was inoculated with 5 mL of sterile water instead of phage inoculum (UW551‐only) and the negative control with no microbial inoculum received 5 mL of CPG (simulates bacterial inoculation effect) followed by 5 mL of sterile water (simulating phage inoculation). A total of 12 independent tomato plants were used for every treatment (all plants were grown individually on separate trays). Disease progression was measured at the onset of disease, from 10 days post‐inoculation (dpi) to 21 dpi, using the disease index score (DI), which varies on a scale from 0 to 4, where 0 denotes healthy plants without symptoms, while indexes between 1 and 4 were used for plants with bacterial wilt symptoms present on 25%, 50%, 75% and 100% of the aerial parts, respectively. To assess the indirect effects of phages on soil rhizosphere bacteria, we explored bacterial species diversity and community composition at the end of the greenhouse experiment. The rhizosphere samples were collected at 21 dpi and processed on the same day: Plants were cut as close to the soil as possible, and the soil of the whole pot was homogenised. For each sample, approximately 2 g of the soil was stored at −80°C and were later used for DNA extraction using method 3 of the protocols described previously (Porteous & Armstrong, ; Tien et al., ). DNA concentrations were assessed with Nanodrop and stored at −80°C until further processing. Additional 2‐g subsamples of extracted soils were stored as 30% glycerol stocks for bacterial isolation, while another 2‐g subsamples were homogenised in 20 mL of sterile water, vortexed for 3 min and allowed to settle for 30 min to obtain soil washes. Ten millilitre of the soil washes were passed through a 0.2 μm filter (Sarstedt, Nümbrecht, Germany) and stored at 4°C for quantification of phage densities. Phage densities were quantified as PFU/mL using the double agar overlay plaque assay (Kropinski et al., ). After diluting samples on 96‐well microplates using three technical replicates per sample, 300 μL of UW551 bacterial cultures and 180 μL of all diluted phage suspensions were mixed with 10 mL of warm CPG soft agar. The suspensions were mixed by gently inverting the tubes six times; then, they were immediately poured over a round CPG agar plate and left to solidify. Plates were incubated at 28°C for 24 h, and subsequently, phage densities were determined as PFU/mL based on the number of visible plaques observed on the agar plates (Table ). To compare changes in bacterial community composition and diversity between treatments at the end of the greenhouse experiment, 16S rRNA sequencing was performed by Novogene using a PE250 strategy on an Illumina NovaSeq6000 using the primers Forward GTGCCAGCMGCCGCGGTAA and Reverse GGACTACHVGGGTWTCTAAT targeting V4 region of the 16S ribosomal RNA. After first assessing the product sizes via PCR and gel runs, the same amount of PCR product was pooled for each sample, end‐repaired, A‐tailed and ligated to Illumina adapters. The library preparation and initial bioinformatic analyses were prepared by Novogene, where the raw data were spliced, and filtered to obtain clean sequence data. DADA2 (Callahan et al., ) was used to reduce noise and to obtain the final amplicon sequence variants (ASVs). Each ASV was then annotated to obtain species information using the SILVA database (Quast et al., ) and QIIME2's classify‐sklearn algorithm (Bokulich et al., ), using a pretrained Naive Bayes classifier. Each ASV was annotated at the kingdom, phyla, class, order, family, genus and species levels. Raw data are available in NCBI Sequence Read Archive (SRA) BioProject PRJNA1010659. We obtained an average of 119,002 raw read pairs for the 154 samples (after removing PY04 1‐phage treatment and one from the PY04 + PY045 2‐phage treatment due to poor quality) with an average length of 253 nucleotides, a GC content of 53.51%, Q20 of 99.55% and Q30 of 98.22% (Table ). To obtain rhizosphere bacterial density estimates based on sequence data, the amount of extracted DNA was normalised with the weight of the soil used for the DNA extraction per sample. Absolute abundance was calculated (Table ) and the taxonomy assignment of the ASVs was represented in Table . Pathogen abundances between treatments were calculated after rarefying the data by grouping together the two ASVs (ASV1 and ASV240) identified as Ralstonia (Tables and ). A Phyloseq object was created using the package phyloseq in R using the normalised absolute abundance, the taxa file and the metadata (total of 55,575 taxa and 154 samples). Rarefaction was performed using the function rarefy_even_depth after filtering (2747 taxa). To separate phage effects from other variables, we also sequenced the bacterial communities from the negative control plant replicates ( N = 12) which were not treated with R. solanacearum or phages and positive control plants that were inoculated with R. solanacearum only ( N = 12). After observing a shift in the bacterial microbiome composition in some of the plant replicate samples, we identified the specific taxa associated with these shifted communities. To achieve this, we performed a random forest analysis using the data of the samples treated with phages (1‐ and 2‐phage treatments, for a total of 118 plants, not including negative or positive controls) and categorised them into ‘shifted’ and ‘centred’ samples according to their position in the NMDS1 axis (as explained in the next section). The analysis was performed using the rarefied absolute count data (2747 taxa). The training set was created randomly by sampling 70% of all samples (80), while the remaining samples were used as the test data set to assess how well the model predicted the data (38 samples). The function randomForest from the package randomForest in R was used with default parameters, proximity = TRUE and 1000 trees. Average Mean Gini values were calculated for each bacterial genus to evaluate their relative importance in explaining the shift in microbiome composition. To experimentally validate if bacteria associated with shifted microbiomes could explain the suppression of R. solanacearum , 100‐fold dilutions of soil washes derived from both diseased and healthy plants (Table ) were plated onto TSA (tryptone‐soy agar), CPG agar and SMSA media which is a semi‐selective medium for R. solanacearum (Elphinstone et al., ). Plates were incubated at 28°C for 48 h, and single colonies streaked onto new media to isolate individual bacteria. Colony‐PCR was performed to identify the genera of >100 colonies by transferring a loop of the colony to 100 μL of sterile water, which was then boiled for 10 min and centrifuged before performing PCRs. The 16S rRNA sequence was amplified using the same primers that were used for the metabarcoding, and PCR products were purified using the Wizard® SV Gel and PCR Clean‐Up System (Promega, Southampton, UK) following the manufacturer's instructions. The products were quantified using a Nanodrop 2000, and Sanger‐sequenced using the forward primer. Based on the high Gini values obtained in the RF analysis, we focused on the following bacterial genera: four isolates assigned to Burkholderia‐Caballeronia‐Paraburkholderia sp., one Bacillus sp. isolate, two Pseudomonas sp. and two Rhodanobacter sp. isolates (Table ). Despite several attempts, we failed to isolate Sphingomonas bacteria from the samples even though it was one of the most abundant taxa in the shifted microbiomes based on the sequencing data. To test if these isolated taxa had a positive or negative effect on R. solanacearum , in vitro and in vivo assays were performed. We first assessed the direct inhibition of R. solanacearum growth with a spot assay using a method similar to phage density measurements; instead of spotting phages, four technical replicates of 10 μL of selected bacterial isolates (Table ) were inoculated on top of R. solanacearum lawns and pathogen inhibition was quantified as the diameter of inhibition halos observed on the plates. Liquid assays were also performed using a GFP‐tagged R. solanacearum UW551 strain obtained by transformation with the pCOMP‐PhII to introduce the recombination regions of the pRC plasmid, and then, the p‐COMP transformed colonies were used for a second transformation with pRCG‐Pps‐GFPuv (Cruz et al., ; Monteiro et al., ). In these assays, R. solanacearum growth was measured on 96‐well plates alone or in the presence of each of the 9 isolated rhizosphere taxa. All the strains were initially grown in CPG media for 48 h at 28°C with shaking (100 rpm) and then diluted to an OD = 0.1. Monocultures were prepared by inoculating 20 μL of bacteria cultures with 180 μL of CPG while cocultures with 20 μL of each bacterium were inoculated in 160 μL CPG ( N = 3 replicates per treatment). Pathogen relative density was assessed based on GFP fluorescence signal intensity at regular intervals for 48‐h growth period (excitation at 405 nm and emission at 509 nm). The potential biocontrol effects of a subset of bacterial taxa were tested in planta ( Pseudomonas P19, Rhodanobacter R55, Burkholderia B12, Burkholderia PB18). Plants were grown for 3 weeks at 20°C (±2°C) with 14 h light/10 h dark cycle and moved to PHCbi growth cabinets (MLR‐352) 3 days before bacterial inoculations to let plants adapt to the new environment of 24°C/20°C temperature and 16/8 h light/dark cycles. All bacterial inoculants were prepared as previously described. Plant inoculations were performed using the same density of R. solanacearum as in the greenhouse experiments (OD 600 = 0.7; and 5 mL) while the four potential biocontrol candidate strains were inoculated at a density of OD 600 = 0.35 (5 mL). Three‐week‐old plants were inoculated with one bacteria (5 mL of one bacterium) or both bacteria using 5 mL of each. Negative controls included plants grown without inoculated microbes and all treatments were replicated eight times. To understand the interactions between phages, resident rhizosphere bacterial communities and R. solanacearum abundances on disease development, we used a partial least squares path modelling (PLS‐PM) using the R package plspm. The cross‐loading between latent variables as well as the unidimensionality in the latent variables was tested during model construction. The final model was created then by assessing ‘Phages’ (reflective indicator included as the log (PFU/mL)), ‘Microbiome’ (reflective indicator included as NMDS1 which divided the samples into shifted and centred microbiomes), ‘ Ralstonia ’ (reflective indicator included as the pathogen abundance) and ‘Disease’ (reflective indicator included as two different variables: Disease index at 21 dpi and area under the diseased curve). To compare if there were differences between shifted and centred microbiome samples, the function plspm.groups was used with the ‘bootstrap’ method with 500 repetitions. To assess mean differences between treatment groups, ANOVA or Kruskal–Wallis tests (based on the normality of the data) were used followed by Tukey's HSD test or Dunn's post hoc tests, respectively (package rstatix and FSA in R), and Bonferroni tests were used to adjust for multiple comparisons. The t _test function in rstatix package was used to compare only two groups. Pearson correlation was used to assess correlation between continuous variables. To measure the area under the disease progression curve or GFP fluorescence, we used the AUC function from the package pROC in R. We analysed the diversity of rhizosphere bacterial communities estimated as the number of ASVs, ASVs richness, Chao1 and Shannon diversity index (at ASV level) using the vegan package (Community Ecology Package [R package vegan version 2.6–4], 2022). Beta diversity was assessed with a non‐metric multidimensional scaling (NMDS) using the metaMDS function in the R vegan package which uses Bray–Curtis distances. NMDS1 values were used to divide the samples into ‘shifted’ (NMDS1 > 0) and ‘centred’ (NMDS1 < 0) groups. To check pairwise differences, we used pairwise Permutational Multivariate Analysis of Variance Using Distance Matrices (adonis) using the pairwise.adonis function from the package pairwiseAdonis in R based on Bray–Curtis distances with 999 permutations and p ‐values adjusted using Holm's method. All R analyses were conducted with version 4.3.2. Phages varied in their ability to control R. solanacearum in vitro We found that all four phages could produce clear inhibition halos on R. solanacearum lawns grown on agar plates (data not shown). Similarly, all four phages reduced R. solanacearum densities in liquid media compared to the control treatments without phages (Figure ). However, different phages had different inhibitory effects (KW 4, N = 20 = 16.0, p = 0.03, Figure ): Phage PY059 showed the weakest pathogen density reduction, phage PY065 was the most effective and phages PY04 and PY045 showed intermediate efficiencies (Figure ). These results suggest that all phages had biocontrol ability in terms of pathogen density reduction in laboratory cultures in vitro. One‐ and two‐phage treatments were equally effective, and phage efficiency varied between and within phage treatments To validate phage efficiency in planta, we compared the biocontrol efficacy of all phages individually and in pairwise combinations using tomato as a model plant host and UW551 R. solanacearum strain as the pathogen (Figure ). Bacterial wilt disease symptoms appeared from 6 dpi in the pathogen‐only treatment (Table , Figure ), and disease symptoms were clearly delayed in the presence of phages (KW 11, N = 2160 = 284, p = 1.76E−54; Figure ). We also observed that the area under the disease progression curve was lower in two‐phage compared to one‐phage treatments (KW 3, N = 144 = 14.3, p = 0.0025, Figure ). However, the proportion of diseased plants in two‐phage treatments was significantly higher than one‐phage treatments at the end of the experiment (X‐squared = 35.186, df = 1, p = 1.498e−09; 33 of 72 (45.8%) vs. 15 of 48 (31.2%), respectively). Of the one‐phage treatments, PY045 was the most efficient phage even though this phage was not the most efficient at controlling R. solanacearum densities in vitro (Figure ). Of the two‐phage treatments, the most effective combination was the PY04 + PY045 treatment (Figure ), and these phages were also highly effective when applied as one‐phage treatments. In contrast, some combinations that consisted of highly efficient single phages (e.g. PY045 + PY065), failed to effectively control the disease (AUC compared with PY045 ( t (351) = −2.32, p = 0.02 and with PY065: t (349) = 0.416, p = ns; Figure )). Also, some other phage combinations (PY04 + PY059, PY045 + PY059, PY059 + PY065) failed to control the disease, potentially due to the low efficiency of PY059, which was included in all these combinations (Figure ). Overall, these results suggest that similar phage biocontrol efficiencies were attained when using phages in 1‐ and 2‐phage combinations, while variability in phage efficiency was large between and within phage treatments, indicative of strong phage identity effects and potential stochastic variation. We next compared if R. solanacearum (read counts) and phage (log (PFU/mL)) densities could predict successful phage biocontrol outcomes based on the last time point of the greenhouse experiment (see methods). We found that R. solanacearum densities correlated positively with disease severity in pathogen‐only ( R = 0.79, p < 0.001) and phage treatments (both 1‐ and 2‐phage treatments combined ( R = 0.62, p < 0.001)) (Figure ). Moreover, pathogen densities were significantly reduced in all phage treatments (KW 3, N = 142 = 30.4, p = 1.14E−6; Figure ), and when analysing phage‐treated plants only, R. solanacearum had higher densities in diseased plants when compared to healthy plants (KW 3, N = 142 = 57.6, p = 3.15E−13; Figure ). Pathogen densities were also higher in 2‐phage compared to 1‐phage treatments ( t (112) = −2.44, p = 0.0162; KW 3, N = 142 = 30.4, p = 1.14E−6; Figure ), and specifically PY04 + PY059 and PY045 + PY059 (Figure ) treatments showed comparable pathogen densities to the pathogen‐only control treatment. Overall, phage densities increased with the severity of disease symptoms ( R = 0.33, p < 0.001; Figure ). Relatively higher phage densities were recovered from two‐phage compared to one‐phage treatments ( t (80.6) = −4.3, p < 0.001; or considering all groups F 3, 138 = 60.92, p < 2E−16; Figure ) and only one of the phage combinations (PY045 + PY059) showed equally low phage densities to one‐phage treatments (Figure ). In general, phage densities were higher in phage‐treated diseased plants when compared to healthy plants ( F 2, 130 = 61.84, p = <2.0E−16; Figure ). No correlation between pathogen and phage densities was observed in one‐phage treatments. However, a negative correlation was observed in two‐phage treatments, where the reduction in R. solanacearum densities (log of the counts) was associated with an increase in phage densities (log (PFU/mL)) in diseased plants ( R = −0.38, p = 0.03; Figure ). These results suggest that while R. solanacearum densities predicted disease progression well during the greenhouse experiment, phage densities were less clearly linked with phage biocontrol efficiency. Moreover, while both one‐ or two‐phage treatments worked equally well, high variability in phage efficiency was observed within all phage treatments. Phage application shifts the rhizosphere microbiome diversity and composition in a subset of phage treatment replicates To better understand the variation underlying phage efficiency within phage treatments, we focused on comparing changes in the bacterial microbiome composition at the end of the greenhouse experiment at the level of individual plants (total of 154 sequenced rhizosphere microbiome samples). As the same standardised and well‐homogenised compost mix was used for all treatment replicates, potential changes could only arise due to treatment effects or stochastic chance events. No significant differences in the bacterial alpha diversity (Shannon index) were found between one‐ and two‐phage treatments, while both phage treatments had significantly lower diversities compared to the negative control treatment (KW 3, N = 142 = 18.1, p = 4.17E−4; Figure , Table ). Moreover, bacterial diversity was lower in the healthy plants when compared to diseased plants ( F 2, 130 = 22.88, p = <2.6E−09, Figure ). Significant differences were also found in terms of bacterial community composition assessed by NMDS analysis (stress = 0.11; Figure ) and pairwise Adonis test confirmed that one‐ and two‐phage treatments differed significantly from each other ( R 2 = 0.14, p ‐adj = 0.006), from the negative control (1‐phage: R 2 = 0.18, p ‐adj = 0.006; 2‐phage: R 2 = 0.05, p ‐adj = 0.024) and pathogen‐only treatment (1‐phage: R 2 = 0.2, p ‐adj = 0.006; 2‐phage: R 2 = 0.05, p ‐adj = 0.024) (Table ). Crucially, while all control microbiome samples clustered tightly together, some one‐ and two‐phage plant replicate samples showed a shift in their community composition along both NMDS1 and NMDS2 axes (Figure ). Interestingly, this shift in microbiome composition could be associated with a reduction in R. solanacearum densities along with increasing values on NDMS1 for both one‐phage ( R = 0.49, p = 5.1E−03; Figure ) and two‐phage treatment replicate samples ( R = 0.48, p = 2.2E−05; Figure ). Moreover, movement along the NDMS1 was positively correlated with higher phage abundances, especially in one‐phage treatment replicate samples ( R = 0.32, p = 0.029; Figure ). Negative values on NMDS2 correlated with a reduction in R. solanacearum densities, but only in one‐phage treatments ( R = 0.46, p = 0.001, Figure , Figure ). For further analyses, we classified our microbiome samples using the NMDS1 axis where samples with values >0 were considered as ‘shifted’ microbiome replicate samples, and samples with values <0 were considered as ‘centred’ microbiome replicate samples. Interestingly, while the shift in the microbiomes of a subset of phage treatment replicates was associated with reduced pathogen and increased phage densities, these replicate samples were not enriched in the healthy plant category (Figure ). However, some phage treatments were underrepresented in the shifted microbiome replicate subset (Figure ). For example, a significant underrepresentation of plants treated only with PY059 ( χ 2 = 4.3363, df = 1, p ‐value = 0.01865) or with PY04 + PY045, PY04 + PY059, PY045 + PY059 phage combinations was observed in the shifted plant replicate subset ( χ 2 (1, N = 57) = 3.6, p = 0.03; χ 2 (1, N = 57) = 4.76, p = 0.01; χ 2 (1, N = 57) = 4.76, p = 0.01, respectively). Moreover, no plants treated with PY045 + PY065 or PY059 + PY065 phage combinations were included in the shifted microbiome samples. To assess if certain bacterial taxa were enriched or reduced in the shifted microbiome replicate samples, we performed a random forest analysis using all 118 microbiome samples covering all phage treatment plant replicates (no negative or positive control samples). We randomly divided the data set into training ( N = 80) and test ( N = 38) data sets. Based on 1000 trees per data set, the first random forest model that differentiated taxa between shifted and centred microbiomes showed an out‐of‐bag (OOB) error rate of 0% and 0% classifying error in each of the categories, confirming the high robustness of the model (Figure ). The importance of each classifying taxa was explored at genera level using Gini values >0.3. The genera with the highest values were Lactobacillus , Sphingomonas , Pseudomonas and Rhodanobacter (Figure , Figure , Table ). Of the five most important taxa, Lactobacillus , Sphingomonas and Pseudomonas were significantly enriched and Rhodanobacter and Gemmatimonas significantly reduced in shifted microbiome replicate samples in both one‐phage and two‐phage treatments. Similarly, other taxa with lower Average Mean Gini values, such as Snodgrassella , Xanthomonas , Gilliamella , Oceanibaculum , Bifidobacterium , Weissella and Neisseria , were enriched in the shifted microbiome samples, while 0319‐6G20, Pseudolabrys , Devosia , Luteimonas , Kapabacteriales , SWB02, Terrimonas , Legionella , Burkholderia ‐ Caballeronia ‐ Paraburkholderia and Arenimonas were reduced in the shifted microbiome samples (Figure ). Together, these results suggest that phages drove shifts in the tomato rhizosphere microbiome composition in a subset of phage treatment replicates, which were also associated with reduced R. solanacearum and increased phage densities. Enriched bacterial taxa in shifted microbiomes can biocontrol R. solanacearum in vitro and in planta To test if the reduction in R. solanacearum densities within shifted microbiome replicates could be explained by enrichment of potentially pathogen‐suppressing taxa, we isolated >100 bacterial colonies at the end of the greenhouse experiment. We were able to isolate the following bacteria with high Gini values: four isolates of Burkholderia‐Caballeronia‐Paraburkholderia sp., one Bacillus sp. isolate, two Pseudomonas sp. and two Rhodanobacter sp. isolates that were likely fast growers in the isolation media used. Using in vitro lab experiments, we found that both Pseudomonas (strains P19 and P91), one Rhodanobacter (R55) and the three Burkholderia sp. (B12, PB10 and PB18) (Figure , Table ) showed clear inhibition halos when grown over a lawn of R. solanacearum strain. Similar inhibitory effects were found in co‐cultures grown in liquid media over a 48‐h time course (KW 9, N = 40 = 37.0, p = 2.6E−5), and this inhibition was especially prominent with both Pseudomonas strains and the Burkholderia B12 strain (Figure ). For the in planta disease suppression greenhouse assays, we selected Pseudomonas P19 and Burkholderia B12 isolates as potential biocontrol strains and also included Rhodanobacter R55 and Burkholderia PB18 as potential ‘negative control’ strains with expected low biocontrol potential based on our in vitro assays. We found that the Burkholderia B12 strain was able to completely inhibit bacterial wilt symptoms (Figure ), while Burkholderia PB18 strain was also quite effective, with only two of eight plants showing disease symptoms. In contrast, Pseudomonas P19 and Rhodanobacter R55 were unable to delay or reduce disease development. Due to the high biocontrol activity of Burkholderia B12, we compared the 16S sequence obtained by Sanger of this strain, with the metabarcoding data, to identify the ASV1852 corresponding with the highest similarity (99.6% similarity, E ‐value = 5.53E−130) and observed that the abundance of this taxon was significantly increased in the centred 1‐phage replicate samples, corresponding with the abundance profile seen for the whole genera (Figure ). These results demonstrate that changes in the composition of bacterial taxa in phage treatment microbiomes were also associated with functional changes in the suppressiveness of these communities. Partial least squares path modelling (PLS‐PM) suggests synergy between phages and the resident rhizosphere microbiota To explore direct and indirect relationships between different variables associated with bacterial wilt disease severity, a partial least squares path modelling (PLS‐PM) was used. Specifically, we explored the effects of phages and the shift in the microbiome composition on R. solanacearum densities and disease severity separately within the centred (NMDS1 < 0) and the shifted (NMDS1 > 0) microbiome groups, resulting in two separate PLS‐PM models (Figure ). The goodness of fit (GoF) for the shifted microbiome model reached 43.7%. As expected, we observed a significant positive effect of ‘ Ralstonia ’ on ‘Disease’ (0.62), indicating that pathogen abundances correlated positively with the disease severity. Moreover, the microbiome composition, reflected by NMDS1 value, had a negative effect on ‘ Ralstonia ’ (−0.4), indicative of the potential suppressive effect of microbiota on the pathogen. The ‘Phages’ had no significant effects on ‘Microbiome’ or ‘ Ralstonia ’ in this model, suggesting that phage densities were not clearly linked with disease progression. The goodness of fit (GoF) for the centred microbiome model was similar to the shifted microbiome model (GoF = 45.1%). However, a contrasting effect of the ‘Microbiome’ on ‘R alstonia ’ was found; microbiomes showed a positive association with pathogen densities (0.32), suggesting that some rhizobacteria may have facilitated disease progression. A bootstrap test with 500 iterations was conducted to confirm the contrasting effects of ‘Microbiome’ on ‘ Ralstonia ’ between the shifted and centred microbiome models ( t (116) = 4.96, p < 0.001). Moreover, ‘Phages’ were positively associated with ‘Microbiome’ (0.33) in the centred model, suggesting that phages had indirect effects on the rhizosphere microbiome composition in this sample group. Together, these models suggest that resident microbiome composition was an important factor contributing to the abundance of the pathogen and severity of disease incidence, explaining the variation in phage biocontrol efficiency within phage treatment replicates. R. solanacearum in vitro We found that all four phages could produce clear inhibition halos on R. solanacearum lawns grown on agar plates (data not shown). Similarly, all four phages reduced R. solanacearum densities in liquid media compared to the control treatments without phages (Figure ). However, different phages had different inhibitory effects (KW 4, N = 20 = 16.0, p = 0.03, Figure ): Phage PY059 showed the weakest pathogen density reduction, phage PY065 was the most effective and phages PY04 and PY045 showed intermediate efficiencies (Figure ). These results suggest that all phages had biocontrol ability in terms of pathogen density reduction in laboratory cultures in vitro. To validate phage efficiency in planta, we compared the biocontrol efficacy of all phages individually and in pairwise combinations using tomato as a model plant host and UW551 R. solanacearum strain as the pathogen (Figure ). Bacterial wilt disease symptoms appeared from 6 dpi in the pathogen‐only treatment (Table , Figure ), and disease symptoms were clearly delayed in the presence of phages (KW 11, N = 2160 = 284, p = 1.76E−54; Figure ). We also observed that the area under the disease progression curve was lower in two‐phage compared to one‐phage treatments (KW 3, N = 144 = 14.3, p = 0.0025, Figure ). However, the proportion of diseased plants in two‐phage treatments was significantly higher than one‐phage treatments at the end of the experiment (X‐squared = 35.186, df = 1, p = 1.498e−09; 33 of 72 (45.8%) vs. 15 of 48 (31.2%), respectively). Of the one‐phage treatments, PY045 was the most efficient phage even though this phage was not the most efficient at controlling R. solanacearum densities in vitro (Figure ). Of the two‐phage treatments, the most effective combination was the PY04 + PY045 treatment (Figure ), and these phages were also highly effective when applied as one‐phage treatments. In contrast, some combinations that consisted of highly efficient single phages (e.g. PY045 + PY065), failed to effectively control the disease (AUC compared with PY045 ( t (351) = −2.32, p = 0.02 and with PY065: t (349) = 0.416, p = ns; Figure )). Also, some other phage combinations (PY04 + PY059, PY045 + PY059, PY059 + PY065) failed to control the disease, potentially due to the low efficiency of PY059, which was included in all these combinations (Figure ). Overall, these results suggest that similar phage biocontrol efficiencies were attained when using phages in 1‐ and 2‐phage combinations, while variability in phage efficiency was large between and within phage treatments, indicative of strong phage identity effects and potential stochastic variation. We next compared if R. solanacearum (read counts) and phage (log (PFU/mL)) densities could predict successful phage biocontrol outcomes based on the last time point of the greenhouse experiment (see methods). We found that R. solanacearum densities correlated positively with disease severity in pathogen‐only ( R = 0.79, p < 0.001) and phage treatments (both 1‐ and 2‐phage treatments combined ( R = 0.62, p < 0.001)) (Figure ). Moreover, pathogen densities were significantly reduced in all phage treatments (KW 3, N = 142 = 30.4, p = 1.14E−6; Figure ), and when analysing phage‐treated plants only, R. solanacearum had higher densities in diseased plants when compared to healthy plants (KW 3, N = 142 = 57.6, p = 3.15E−13; Figure ). Pathogen densities were also higher in 2‐phage compared to 1‐phage treatments ( t (112) = −2.44, p = 0.0162; KW 3, N = 142 = 30.4, p = 1.14E−6; Figure ), and specifically PY04 + PY059 and PY045 + PY059 (Figure ) treatments showed comparable pathogen densities to the pathogen‐only control treatment. Overall, phage densities increased with the severity of disease symptoms ( R = 0.33, p < 0.001; Figure ). Relatively higher phage densities were recovered from two‐phage compared to one‐phage treatments ( t (80.6) = −4.3, p < 0.001; or considering all groups F 3, 138 = 60.92, p < 2E−16; Figure ) and only one of the phage combinations (PY045 + PY059) showed equally low phage densities to one‐phage treatments (Figure ). In general, phage densities were higher in phage‐treated diseased plants when compared to healthy plants ( F 2, 130 = 61.84, p = <2.0E−16; Figure ). No correlation between pathogen and phage densities was observed in one‐phage treatments. However, a negative correlation was observed in two‐phage treatments, where the reduction in R. solanacearum densities (log of the counts) was associated with an increase in phage densities (log (PFU/mL)) in diseased plants ( R = −0.38, p = 0.03; Figure ). These results suggest that while R. solanacearum densities predicted disease progression well during the greenhouse experiment, phage densities were less clearly linked with phage biocontrol efficiency. Moreover, while both one‐ or two‐phage treatments worked equally well, high variability in phage efficiency was observed within all phage treatments. To better understand the variation underlying phage efficiency within phage treatments, we focused on comparing changes in the bacterial microbiome composition at the end of the greenhouse experiment at the level of individual plants (total of 154 sequenced rhizosphere microbiome samples). As the same standardised and well‐homogenised compost mix was used for all treatment replicates, potential changes could only arise due to treatment effects or stochastic chance events. No significant differences in the bacterial alpha diversity (Shannon index) were found between one‐ and two‐phage treatments, while both phage treatments had significantly lower diversities compared to the negative control treatment (KW 3, N = 142 = 18.1, p = 4.17E−4; Figure , Table ). Moreover, bacterial diversity was lower in the healthy plants when compared to diseased plants ( F 2, 130 = 22.88, p = <2.6E−09, Figure ). Significant differences were also found in terms of bacterial community composition assessed by NMDS analysis (stress = 0.11; Figure ) and pairwise Adonis test confirmed that one‐ and two‐phage treatments differed significantly from each other ( R 2 = 0.14, p ‐adj = 0.006), from the negative control (1‐phage: R 2 = 0.18, p ‐adj = 0.006; 2‐phage: R 2 = 0.05, p ‐adj = 0.024) and pathogen‐only treatment (1‐phage: R 2 = 0.2, p ‐adj = 0.006; 2‐phage: R 2 = 0.05, p ‐adj = 0.024) (Table ). Crucially, while all control microbiome samples clustered tightly together, some one‐ and two‐phage plant replicate samples showed a shift in their community composition along both NMDS1 and NMDS2 axes (Figure ). Interestingly, this shift in microbiome composition could be associated with a reduction in R. solanacearum densities along with increasing values on NDMS1 for both one‐phage ( R = 0.49, p = 5.1E−03; Figure ) and two‐phage treatment replicate samples ( R = 0.48, p = 2.2E−05; Figure ). Moreover, movement along the NDMS1 was positively correlated with higher phage abundances, especially in one‐phage treatment replicate samples ( R = 0.32, p = 0.029; Figure ). Negative values on NMDS2 correlated with a reduction in R. solanacearum densities, but only in one‐phage treatments ( R = 0.46, p = 0.001, Figure , Figure ). For further analyses, we classified our microbiome samples using the NMDS1 axis where samples with values >0 were considered as ‘shifted’ microbiome replicate samples, and samples with values <0 were considered as ‘centred’ microbiome replicate samples. Interestingly, while the shift in the microbiomes of a subset of phage treatment replicates was associated with reduced pathogen and increased phage densities, these replicate samples were not enriched in the healthy plant category (Figure ). However, some phage treatments were underrepresented in the shifted microbiome replicate subset (Figure ). For example, a significant underrepresentation of plants treated only with PY059 ( χ 2 = 4.3363, df = 1, p ‐value = 0.01865) or with PY04 + PY045, PY04 + PY059, PY045 + PY059 phage combinations was observed in the shifted plant replicate subset ( χ 2 (1, N = 57) = 3.6, p = 0.03; χ 2 (1, N = 57) = 4.76, p = 0.01; χ 2 (1, N = 57) = 4.76, p = 0.01, respectively). Moreover, no plants treated with PY045 + PY065 or PY059 + PY065 phage combinations were included in the shifted microbiome samples. To assess if certain bacterial taxa were enriched or reduced in the shifted microbiome replicate samples, we performed a random forest analysis using all 118 microbiome samples covering all phage treatment plant replicates (no negative or positive control samples). We randomly divided the data set into training ( N = 80) and test ( N = 38) data sets. Based on 1000 trees per data set, the first random forest model that differentiated taxa between shifted and centred microbiomes showed an out‐of‐bag (OOB) error rate of 0% and 0% classifying error in each of the categories, confirming the high robustness of the model (Figure ). The importance of each classifying taxa was explored at genera level using Gini values >0.3. The genera with the highest values were Lactobacillus , Sphingomonas , Pseudomonas and Rhodanobacter (Figure , Figure , Table ). Of the five most important taxa, Lactobacillus , Sphingomonas and Pseudomonas were significantly enriched and Rhodanobacter and Gemmatimonas significantly reduced in shifted microbiome replicate samples in both one‐phage and two‐phage treatments. Similarly, other taxa with lower Average Mean Gini values, such as Snodgrassella , Xanthomonas , Gilliamella , Oceanibaculum , Bifidobacterium , Weissella and Neisseria , were enriched in the shifted microbiome samples, while 0319‐6G20, Pseudolabrys , Devosia , Luteimonas , Kapabacteriales , SWB02, Terrimonas , Legionella , Burkholderia ‐ Caballeronia ‐ Paraburkholderia and Arenimonas were reduced in the shifted microbiome samples (Figure ). Together, these results suggest that phages drove shifts in the tomato rhizosphere microbiome composition in a subset of phage treatment replicates, which were also associated with reduced R. solanacearum and increased phage densities. R. solanacearum in vitro and in planta To test if the reduction in R. solanacearum densities within shifted microbiome replicates could be explained by enrichment of potentially pathogen‐suppressing taxa, we isolated >100 bacterial colonies at the end of the greenhouse experiment. We were able to isolate the following bacteria with high Gini values: four isolates of Burkholderia‐Caballeronia‐Paraburkholderia sp., one Bacillus sp. isolate, two Pseudomonas sp. and two Rhodanobacter sp. isolates that were likely fast growers in the isolation media used. Using in vitro lab experiments, we found that both Pseudomonas (strains P19 and P91), one Rhodanobacter (R55) and the three Burkholderia sp. (B12, PB10 and PB18) (Figure , Table ) showed clear inhibition halos when grown over a lawn of R. solanacearum strain. Similar inhibitory effects were found in co‐cultures grown in liquid media over a 48‐h time course (KW 9, N = 40 = 37.0, p = 2.6E−5), and this inhibition was especially prominent with both Pseudomonas strains and the Burkholderia B12 strain (Figure ). For the in planta disease suppression greenhouse assays, we selected Pseudomonas P19 and Burkholderia B12 isolates as potential biocontrol strains and also included Rhodanobacter R55 and Burkholderia PB18 as potential ‘negative control’ strains with expected low biocontrol potential based on our in vitro assays. We found that the Burkholderia B12 strain was able to completely inhibit bacterial wilt symptoms (Figure ), while Burkholderia PB18 strain was also quite effective, with only two of eight plants showing disease symptoms. In contrast, Pseudomonas P19 and Rhodanobacter R55 were unable to delay or reduce disease development. Due to the high biocontrol activity of Burkholderia B12, we compared the 16S sequence obtained by Sanger of this strain, with the metabarcoding data, to identify the ASV1852 corresponding with the highest similarity (99.6% similarity, E ‐value = 5.53E−130) and observed that the abundance of this taxon was significantly increased in the centred 1‐phage replicate samples, corresponding with the abundance profile seen for the whole genera (Figure ). These results demonstrate that changes in the composition of bacterial taxa in phage treatment microbiomes were also associated with functional changes in the suppressiveness of these communities. To explore direct and indirect relationships between different variables associated with bacterial wilt disease severity, a partial least squares path modelling (PLS‐PM) was used. Specifically, we explored the effects of phages and the shift in the microbiome composition on R. solanacearum densities and disease severity separately within the centred (NMDS1 < 0) and the shifted (NMDS1 > 0) microbiome groups, resulting in two separate PLS‐PM models (Figure ). The goodness of fit (GoF) for the shifted microbiome model reached 43.7%. As expected, we observed a significant positive effect of ‘ Ralstonia ’ on ‘Disease’ (0.62), indicating that pathogen abundances correlated positively with the disease severity. Moreover, the microbiome composition, reflected by NMDS1 value, had a negative effect on ‘ Ralstonia ’ (−0.4), indicative of the potential suppressive effect of microbiota on the pathogen. The ‘Phages’ had no significant effects on ‘Microbiome’ or ‘ Ralstonia ’ in this model, suggesting that phage densities were not clearly linked with disease progression. The goodness of fit (GoF) for the centred microbiome model was similar to the shifted microbiome model (GoF = 45.1%). However, a contrasting effect of the ‘Microbiome’ on ‘R alstonia ’ was found; microbiomes showed a positive association with pathogen densities (0.32), suggesting that some rhizobacteria may have facilitated disease progression. A bootstrap test with 500 iterations was conducted to confirm the contrasting effects of ‘Microbiome’ on ‘ Ralstonia ’ between the shifted and centred microbiome models ( t (116) = 4.96, p < 0.001). Moreover, ‘Phages’ were positively associated with ‘Microbiome’ (0.33) in the centred model, suggesting that phages had indirect effects on the rhizosphere microbiome composition in this sample group. Together, these models suggest that resident microbiome composition was an important factor contributing to the abundance of the pathogen and severity of disease incidence, explaining the variation in phage biocontrol efficiency within phage treatment replicates. In this study, we compared how phage identities and their application either alone or in two‐phage combinations affected the efficiency of phage biocontrol against R. solanacearum phytopathogenic bacterium. We found that the species identity of applied phages and their combinations was more important in explaining phage efficiency compared to using phages either alone or as two‐phage combinations. Moreover, bacterial wilt disease biocontrol was better explained by a reduction in pathogen densities instead of corresponding increases in phage densities. Despite improved disease biocontrol, large between‐replicate variation was observed in both one‐ and two‐phage treatments, which was stochastic within treatments but was associated with non‐random changes in rhizosphere microbiome composition across all phage treatments, with certain taxa showing consistent enrichment or reduction in their relative abundances. While these microbiome shifts took place equally often in diseased and healthy plants, they were associated with a reduction in pathogen and an increase in phage densities, indicative of improved phage biocontrol. While our results are in line with previous research, which demonstrated that phage biocontrol applications can lead to taxonomic and functional changes in the resident rhizosphere microbiota (Wang et al., , ), we further show that such effects can be stochastic and arise quickly between initially identical plant replicates during one plant growth cycle. Interestingly, we found that both 1‐ and 2‐phage treatments reduced bacterial wilt disease incidence and pathogen densities to a similar degree. Overall, a reduction in R. solanacearum densities predicted bacterial wilt disease incidence well in both 1‐ and 2‐phage treatments. However, in contrast to previous studies (Wang et al., , ; Yang et al., ), phage densities predicted bacterial wilt disease incidence less well in our data, even though phages increased in their density in the presence of R. solanacearum during the experiment. A previous study showed that increasing the number of phages during phage biocontrol can lead to improved bacterial wilt disease control (Wang et al., ), where the best results were obtained when four phages were applied simultaneously. However, no significant benefits of using two‐phage combinations were observed in this study (Wang et al., ), which also aligns well with other studies reporting that phage cocktails do not always improve phage biocontrol efficiency relative to the application of single phages (Álvarez et al., ; Rabiey et al., ). Instead, previous results suggested that at least three phages might be needed to efficiently control R. solanacearum in the rhizosphere (Álvarez et al., ; Wang et al., ). Our results further suggest that the identity of phages and phage combinations had a big effect on their efficiency. While two of the phages used in this study (PY04 and PY065) were genetically highly similar, the other two phages (PY045 and PY059) were completely dissimilar. Interestingly, we observed that using dissimilar phages together in combination, such as PY04 + PY059, PY045 + PY059 and PY059 + PY065, led to poorer biocontrol outcomes in terms of higher R. solanacearum abundances and increased disease incidence. However, on some occasions, combining dissimilar phages, such as PY04 + PY045, led to very good biocontrol outcomes, which suggests that phage genetic dissimilarity alone is a poor predictor of combination efficiency. In line with these results, it was previously found that combining even highly genetically similar phages (>99%) can lead to significant improvement in phage biocontrol efficiency against R. solanacearum (Wang et al., ). Importantly, our results suggest that combining phages can reduce phage biocontrol efficiency relative to applying phages alone. This could be explained by the transition to persistent cells which help bacteria survive the phage stress by reducing their metabolic activity (Fernández‐García et al., ), phage‐resistant phenotypes arising by genetic mutation (Wang, Wang, et al., ) or more likely phage–phage competition from the same host cells (Refardt, ). Moreover, the reduced efficiency of some two‐phage treatments might have been due to the poor efficiency of phage PY059, which was observed to lack the HNH gene (Text ), which is required for efficient phage packaging during production of virions (Kala et al., ). Together, these results suggest that using more phages during phytopathogen biocontrol may not always result in improved disease suppression, and that combining phages based on their genetic dissimilarity might be a bad proxy for improving their efficiency. In addition to significant between‐phage treatment variation, we observed a large intra‐treatment variation in phage efficiency within all one‐ and two‐phage treatments. Interestingly, while a subset of phage treatment replicates showed a clear shift in their microbiome composition, these shifts were not consistently associated with healthy or diseased plants. However, they were only observed in phage treatments and were correlated with a reduction in pathogen and an increase in phage densities, indicative of improved phage biocontrol. Crucially, the shifted microbiomes were consistently associated with the enrichment or reduction of certain bacterial genera, including Sphingomonadaceae and Pseudomonadaceae, which have previously been associated with the suppressiveness of rhizosphere microbiome of the resistant ‘Hawaii 7996’ tomato cultivar against R. solanacearum (Kwak et al., ). We were able to isolate candidate bacterial taxa from our end‐point samples of the greenhouse experiment to test their disease suppressiveness against R. solanacearum in vitro and in planta experiments. While some of these strains suppressed R. solanacearum consistently both in vitro and in planta, some strains that worked well in vitro failed to control disease in planta. Despite several attempts, we failed to isolate Sphingomonas from our end‐point samples even though it was highly enriched based on sequence data. Overall, isolates classified as Burkholderia were the most effective at controlling R. solanacearum in planta validation experiments. Previous studies have shown that Burkholderia can be enriched in the rhizosphere in response to plant stimuli (Luo et al., ) and when controlling R. solanacearum with integrated biocontrol approaches (Hu et al., ). Members of Burkholderia can suppress various soil‐borne pathogens (Eberl & Vandamme, ) and could hence be an important taxon for soil suppressiveness to bacterial wilt. Interestingly, the most suppressive Burkholderia species we tested was enriched in centred microbiome samples in one‐phage treatments, suggesting that phages likely affected the suppressiveness of rhizosphere microbiota also in non‐shifted rhizosphere bacterial communities. A few previous studies have also found that the application of pathogen‐specific phages can lead to indirect changes in the microbiome composition (Wang et al., ) and enrichment of disease‐suppressive taxa belonging to Streptomyces and Nocardioides (Wang et al., ). As the applied phages were specific to R. solanacearum , the likely mechanism for the enrichment was the release of niche space by the phages that efficiently reduced the relative pathogen densities in the rhizosphere. In support of this, a recent study showed that changes in the resident microbiome taxa composition were clearer when phages could reduce the relative abundances of R. solanacearum more efficiently, which was also positively correlated with increased phage densities (Wang et al., ). Our results align well with this previous study as the shifts in the microbiome composition were more often observed when the reduction in the R. solanacearum relative abundances was clearer. However, the enriched taxa were different, which is not surprising as the compost substrate and associated resident bacterial communities were different between these experiments. Our findings also suggest that even though certain taxa responded consistently to phage application, in the shifted bacterial communities, composition in both 1‐ and 2‐phage treatments diverged, while centred communities remained very similar to each other. As several bacterial taxa, including Bacillus , Pseudomonas and Streptomyces have been linked with bacterial wilt disease‐suppressive soils in previous studies (Wang et al., , ; Wei et al., ; Yang et al., ), it is possible that any competitive taxa present in the rhizosphere microbiome might be able to respond and increase in relative abundance in response to phage infection‐mediated release of the niche space. The shift in the microbiome composition within phage treatments was likely stochastic as all the plants in the greenhouse experiment were started in identical conditions using the same batch of well‐mixed and homogenised compost substrate (John Innes No2). While the initial microbiomes were highly similar between all the plants (based on the tight clustering of negative control samples), a subset of plant replicates diverged and experienced a shift in their microbiome composition during one plant growth cycle. Crucially, the shifts in microbiome composition were not taxonomically random and the same bacterial taxa were observed to shift between different phage treatments. Even though we tried to make the starting conditions as similar as possible between the plant replicates, it is possible that some small differences in initial microbiomes were still present. Such small initial differences in microbiome composition have previously been shown to determine whether plants get infected by R. solanacearum or if they remain healthy during the subsequent tomato growth cycle (Wei et al., ). In another study, Gu et al. showed that initially homogeneous soils can quickly diverge into compositionally different microbiomes that later become associated with healthy or diseased plants. As we lack initial and temporal sampling of all plant rhizosphere microbiome replicates, we cannot rule out either of these hypotheses. However, as we used highly controlled experimental conditions as evidenced by tight clustering of negative control replicates, our findings are more likely explained by the scenario proposed by Gu et al. , where initially similar microbiomes rapidly diverged between plant replicates. To confirm this in the future, time‐series data are needed. Our findings have several important implications for phage biocontrol of phytopathogens. First, our results demonstrate that when applied through root drenching, phage ability to control R. solanacearum densities depends on the responses of the resident rhizosphere microbiota, and crucially, we found that this effect was stochastic among phage treatment replicates, introducing variability in phage biocontrol efficiency. This could have important repercussions for applying phages successfully in the field, where differences in the soil microbiome composition are likely to be greater (Hannula et al., ). Even though variability in rhizosphere microbiome composition could introduce more variability into the success of phage biocontrol, current data suggest that different resident bacterial taxa could provide disease suppression in synergy with applied phages (Wang et al., , ). It is hence possible that different rhizosphere microbiomes contain functionally redundant bacteria that can suppress soil‐borne pathogens and be enriched in response to phage application despite belonging to different taxonomic groups. Moreover, some of the disease‐suppressive taxa we validated showed reduction in their relative abundance in response to phage application. Hence, future work should explore phage effects at the bacterial community level to better understand the net effects of microbiome changes on disease suppression. Finally, what triggered the microbiome shifts in a subset of phage treatment replicates remains inconclusive. One potential and underexplored factor could be soil viromes, which have been linked with changes in the soil suppressiveness to R. solanacearum via effects on the disease‐suppressive bacterial taxa (Yang et al., ). Further work should also assess if these results depend on the phage application method, for example, by evaluating if rhizosphere changes also take place when the phages are applied as seed coating (Erdrich et al., ) instead of root drenches. In conclusion, our data suggest that the presence of resident microbiome can complicate phage biocontrol outcomes in plant rhizosphere by introducing variation in treatment efficiency. Sara Franco Ortega: Investigation; writing – original draft; methodology; visualization; software; formal analysis; data curation. Bryden Fields: Investigation; methodology; software; formal analysis; data curation. Daniel Narino Rojas: Investigation; methodology. Lauri Mikonranta: Investigation; methodology. Matthew Holmes: Investigation; methodology. Andrea L. Harper: Funding acquisition; writing – review and editing; project administration; supervision; resources. Ville‐Petri Friman: Project administration; supervision; resources; writing – review and editing; funding acquisition; conceptualization; visualization. The authors declare that they have no competing interests. Appendix S1 Table S2 Table S3 Table S4 Table S5 Table S6 Table S7 Table S8 Table S9 Table S10 Table S11 |
Robotics-assisted surgery in gynecology: A single-center experience with the Hugo™ RAS system in India | 5076808e-e8a0-4f9f-98ab-5aec96c08dc1 | 11877460 | Surgical Procedures, Operative[mh] | Minimally invasive surgical methods in gynecology are alternatives to traditional open surgery and offer numerous benefits compared to traditional laparotomy, including decreased estimated blood loss, a lower likelihood of complications, and shorter hospitalization and recovery periods. These methods mainly include laparoscopy, employed for managing a variety of benign gynecological conditions. Conventional laparoscopic surgery presents some limitations. The surgeon’s limited field of view requires him/her to rely on a camera assistant, who maneuvers the laparoscope based on verbal instructions from the surgeon. Moreover, owing to the 2D perspective provided, depth perception and spatial awareness are restricted. Although laparoscopes have been modified to provide high-definition 3D vision to address this issue, the need for a camera assistant remains, and the “camera shake” due to fatigue of the person holding the camera could disorient the surgeon. Laparoscope holders were among the first robots in laparoscopic surgery, with the AESOP ® 3000 robot introduced in 1994. The fixed positions of trocars used in conventional laparoscopy hinder the surgeon’s ability to maneuver with flexibility. The master–slave robots enhance the surgeon’s flexibility by providing six degrees of freedom of hand motion at the console. The da Vinci surgical system, introduced in the year 2000, was the first Food and Drug Administration-approved robotics-assisted surgery (RAS) system for general laparoscopic surgery. Since then, RAS systems have advanced, offering improved precision, dexterity, and ergonomics over conventional laparoscopic surgery, along with stabilized movements, tremor filtration, and motion scaling. , In RAS systems, the 3D high-definition magnified images enhance depth perception, the robotic arm carts offer improved dexterity with increased degrees of freedom, tremors are mitigated through vibration filtration, wristed instruments counter fulcrum effects, robot-held endoscopes with direct control reduce dependence on assistants, and adjustable seats and user-friendly interfaces afford surgeons greater comfort during procedures compared to conventional laparoscopic systems. Robotic-assisted techniques have found applications in a wide range of surgical fields, including hepatobiliary, pancreatic, endocrine, bariatric, anti-reflux, colorectal, pediatric, and gastric oncology surgeries, as well as hernia, prostatectomy, or abdominal wall reconstruction procedures. , In the field of gynecology, robotic surgery is not only used for benign conditions such as endometriosis, pelvic organ prolapse, pelvic pain, sacropexy, tubal recanalization, uterine fibroids, and abnormal uterine bleeding but also for treating malignant conditions such as cancers of the ovary, endometrium, and cervix and performing lymphadenectomy. , Patients undergoing gynecological procedures with RAS systems have been reported to exhibit early recoveries, shorter hospital stays, and fewer postoperative complications, including primary or secondary hemorrhage. The Hugo™ RAS system (Medtronic Inc, USA) was first used at Clínica Santa María (Chile) for robotic prostatectomy in June 2021, and since then, several hospitals worldwide have adopted this RAS system. The system comprises arm carts, a surgeon console, and a system tower and is compatible with wristed articulating instruments of the same company, a 3D endoscope, the VITOM ® 3D visualization system (Karl Storz SE & Co. KG, Germany), the Covidien ValleyLab™ FT10 high-frequency electrosurgical generator (Medtronic Inc, USA), and the VersaOne™ reusable positioning trocar system (User guide). The system provides high-resolution 3D images. The surgeon’s console has an interactive high-definition display with a highly adjustable ergonomic set-up. The surgeon’s wrist motion is translated into the robotic instrument’s actions with two hand controllers, facilitating fluid and precise movements. The sensors on the console track the surgeon’s 3D glasses, pausing instrument movement when the surgeon looks away from the display screen (User guide). The system’s tilt angle capability allows customization of the surgical approach according to the patient’s characteristics, especially in complex clinical scenarios. Previous reports on the Hugo™ RAS system have demonstrated favorable results with no technical failures or intraoperative complications in urological procedures , ; however, data on its use in gynecological surgeries is limited. This underscores the importance of surgeons sharing their experiences and outcomes to further solidify the evidence base for the application of robotic-assisted surgery in the field of gynecology. In India, the first surgical robot was installed in 2006. By 2019, there were 66 robotic centers with 71 systems installed and more than 500 surgeons skilled in robotic surgery. This article describes the initial experience of using the Hugo™ RAS system in gynecological surgeries at a single tertiary care center in India.
Study design and patient population This report presents a real-world single-center experience of using the Hugo™ RAS system on 20 patients who underwent surgeries for a range of gynecological conditions at the CARE Hospital in Banjara Hills, Hyderabad, from September 2022 to 2023. Since the study was aimed at reporting on the observations and experiences from a single center and was not intended to assess a specific hypothesis, no sample size calculation was performed. The STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) reporting guideline was employed for this study. Preprocedural evaluation The patients underwent preoperative assessment for eligibility and were included in the study if they were fit for general anesthesia, required surgery for gynecological conditions, were undergoing surgery with Hugo™ RAS system, and provided written informed consent per institution and geographic requirement. Patients were excluded from the study if they were, or were expected to be, unavailable for follow-up if local regulations prohibited their participation, or if they were already enrolled in, or had intentions to enroll in, concurrent investigations involving drugs or devices that could potentially introduce bias to the results of this study. Written informed consent was obtained from all patients, and the study was conducted in accordance with the principles of the Declaration of Helsinki. Patient position and port placements Patients were laid in a supine steep Trendelenburg position of >25° with their legs in the low-mid lithotomy position . There was no roll (0°), and the height of the bed was above 70 on the scale of the cart column. The endoscopic ports were placed near the patients’ umbilicus with a caudal offset of 5 cm between the right-hand surgeon port and the endoscopic port. The left-hand surgeon port was kept lateral to the endoscopic port at a distance of 14 cm. The reverse port was also placed lateral to the endoscopic port and the right-hand surgeon port at a distance of 8 cm from both. The assist port was placed 5 cm away from the endoscopic and left-hand surgeon ports. A distance of 2 cm was also maintained between any bony prominences and all the ports . For the procedure, the tilt was adjusted for the arms, the carts were placed in the range of 18–24 inches, and the brakes were set. The laser was set parallel to the bed, and the position button was used to move the arms to the desired positions while docking to the port at the angles prescribed in the Medtronic user guide . Once the placements of the arms were finalized, they were confirmed on the tower screen. Since the cases reported in this paper were the initial cases operated with the system, all surgeries were performed using three arms. However, as per the advanced practice, the technique is now modified and the surgeries currently performed at the center are by using both, three arms and two arms. In case of the two-arm technique, one arm is for unipolar and other for bipolar. Surgeries were performed by a single operator who had clinical experience of minimal access surgeries for about 28 years. The surgeries were assisted by a single assistant having clinical experience of minimal access surgeries for about 7 years. Outcome measures The outcome parameters evaluated were docking time (time from the first order given to position the robot to start at the console) and console time (time from the start of the procedure to suturing of the surgical incision), blood loss (intraoperative and postoperative drainage), and length of hospital stay. Postoperative complications were graded according to the Clavien-Dindo classification. Postoperative pain perception was evaluated using a visual analog scale (VAS) at 1, 6, and 12 h as well as 1 week after surgery. Data analysis Descriptive statistics were performed using R software version 4.3.2. Continuous variables were presented as mean ± standard deviation (SD).
This report presents a real-world single-center experience of using the Hugo™ RAS system on 20 patients who underwent surgeries for a range of gynecological conditions at the CARE Hospital in Banjara Hills, Hyderabad, from September 2022 to 2023. Since the study was aimed at reporting on the observations and experiences from a single center and was not intended to assess a specific hypothesis, no sample size calculation was performed. The STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) reporting guideline was employed for this study.
The patients underwent preoperative assessment for eligibility and were included in the study if they were fit for general anesthesia, required surgery for gynecological conditions, were undergoing surgery with Hugo™ RAS system, and provided written informed consent per institution and geographic requirement. Patients were excluded from the study if they were, or were expected to be, unavailable for follow-up if local regulations prohibited their participation, or if they were already enrolled in, or had intentions to enroll in, concurrent investigations involving drugs or devices that could potentially introduce bias to the results of this study. Written informed consent was obtained from all patients, and the study was conducted in accordance with the principles of the Declaration of Helsinki.
Patients were laid in a supine steep Trendelenburg position of >25° with their legs in the low-mid lithotomy position . There was no roll (0°), and the height of the bed was above 70 on the scale of the cart column. The endoscopic ports were placed near the patients’ umbilicus with a caudal offset of 5 cm between the right-hand surgeon port and the endoscopic port. The left-hand surgeon port was kept lateral to the endoscopic port at a distance of 14 cm. The reverse port was also placed lateral to the endoscopic port and the right-hand surgeon port at a distance of 8 cm from both. The assist port was placed 5 cm away from the endoscopic and left-hand surgeon ports. A distance of 2 cm was also maintained between any bony prominences and all the ports . For the procedure, the tilt was adjusted for the arms, the carts were placed in the range of 18–24 inches, and the brakes were set. The laser was set parallel to the bed, and the position button was used to move the arms to the desired positions while docking to the port at the angles prescribed in the Medtronic user guide . Once the placements of the arms were finalized, they were confirmed on the tower screen. Since the cases reported in this paper were the initial cases operated with the system, all surgeries were performed using three arms. However, as per the advanced practice, the technique is now modified and the surgeries currently performed at the center are by using both, three arms and two arms. In case of the two-arm technique, one arm is for unipolar and other for bipolar. Surgeries were performed by a single operator who had clinical experience of minimal access surgeries for about 28 years. The surgeries were assisted by a single assistant having clinical experience of minimal access surgeries for about 7 years.
The outcome parameters evaluated were docking time (time from the first order given to position the robot to start at the console) and console time (time from the start of the procedure to suturing of the surgical incision), blood loss (intraoperative and postoperative drainage), and length of hospital stay. Postoperative complications were graded according to the Clavien-Dindo classification. Postoperative pain perception was evaluated using a visual analog scale (VAS) at 1, 6, and 12 h as well as 1 week after surgery.
Descriptive statistics were performed using R software version 4.3.2. Continuous variables were presented as mean ± standard deviation (SD).
Patient characteristics All 20 patients (mean age 45.3 ± 6.5 (range 25–58) years) successfully underwent robotic surgery using the Hugo™ RAS system. Hysterectomy was performed on 18 patients, accessory and cavitated uterine mass excision on one patient, and myomectomy on one patient. Patient characteristics are shown in . All patients were followed up to 1-week post-surgery. Procedural outcomes The mean docking time was 6.3 ± 2.0 min, and the mean duration of the surgery was 86.9 ± 20.3 (range: 55.5–119.0) min. The mean blood loss (intraoperative and postoperative drainage) was 103.5 ± 62.4 (range: 50–250) mL; it was minimal in all cases with no requirement for blood transfusions. The duration of hospital stay was 2 days as per hospital norms. Postoperative complications All surgical procedures were uneventful, and no immediate complications were observed. Grade 1 complications according to the Clavien-Dindo classification were observed in two patients. These patients experienced minor late complications during the third week following surgery, characterized by vaginal spotting secondary to granulation tissue formation at the vaginal vault. Complications were addressed through a conservative approach on an outpatient department basis. Pain perception Postoperative pain perception declined over time; the mean VAS scores were 3.2 ± 0.4, 2.2 ± 0.4, and 1.0 ± 0.0 at 1, 6, and 12 h, respectively. No pain was perceived at 1 week after surgery.
All 20 patients (mean age 45.3 ± 6.5 (range 25–58) years) successfully underwent robotic surgery using the Hugo™ RAS system. Hysterectomy was performed on 18 patients, accessory and cavitated uterine mass excision on one patient, and myomectomy on one patient. Patient characteristics are shown in . All patients were followed up to 1-week post-surgery.
The mean docking time was 6.3 ± 2.0 min, and the mean duration of the surgery was 86.9 ± 20.3 (range: 55.5–119.0) min. The mean blood loss (intraoperative and postoperative drainage) was 103.5 ± 62.4 (range: 50–250) mL; it was minimal in all cases with no requirement for blood transfusions. The duration of hospital stay was 2 days as per hospital norms.
All surgical procedures were uneventful, and no immediate complications were observed. Grade 1 complications according to the Clavien-Dindo classification were observed in two patients. These patients experienced minor late complications during the third week following surgery, characterized by vaginal spotting secondary to granulation tissue formation at the vaginal vault. Complications were addressed through a conservative approach on an outpatient department basis.
Postoperative pain perception declined over time; the mean VAS scores were 3.2 ± 0.4, 2.2 ± 0.4, and 1.0 ± 0.0 at 1, 6, and 12 h, respectively. No pain was perceived at 1 week after surgery.
Robotic surgery is often seen as an extension or alternative to laparoscopic surgery rather than a definitively superior technology, despite having the benefits of being minimally invasive compared to open surgery, and there exists conflicting research on the comparative advantages and drawbacks of both robotic-assisted and laparoscopic approaches. This report documents the experience of using the Hugo™ RAS system for performing gynecological surgical procedures on 20 patients at a single center in India. Docking and operative times Compared with conventional laparoscopic methods, RAS is generally associated with a longer operative time, which includes the time required for placement of ports, docking time, console time, and closing time. The docking phase encompasses activities such as moving the robotic arms into the surgical field, setting the arms into their respective port sites, and inserting the robotic instruments in the abdomen, which account for approximately 3.5% of the operative time. A systematic review of reports on hysterectomies performed using RAS systems (2016–2021) presented a wide variation in the operative time (75.4–306.0 min). The docking time and console time for the Hugo™ RAS system in the present study were 6.3 ± 2.0 min and 86.9 ± 20.3 min, respectively. In a prospective observational study that used the Hugo™ RAS system for 192 gynecological surgical procedures performed at a single center in Italy, the docking time was reported to be 5.1 ± 2.0 min and the console time was 111.5 ± 40.2 min. The data indicate that robotic set-up and docking with the new Hugo™ RAS system can be performed efficiently and that the specific robotic docking learning curve is comparable to preexisting data for other platforms. The Hugo™ RAS system has four independent arm carts, which makes it easier to park the device away from the patient while enabling precise docking of the surgical arm at the desired location on the surgical field. This convenience stands in contrast to the da Vinci ® Surgical System, which features four arms attached to a single base. In a prospective observational study in the Netherlands, 65 robot-assisted operations, including 14 gynecological procedures, were performed using the da Vinci Xi system (Intuitive Surgical Inc., USA), and the docking time was reported to be 7.8 ± 2.7 min. The time required for docking and the surgery is largely influenced by the experience of the surgeon, the complexity of the surgery, and the techniques used for the surgery. , Nevertheless, the presence of a learning curve underscores the importance of surgeons receiving comprehensive training to proficiently handle the docking process and carry out successful procedures with the Hugo™ RAS system. Preclinical laboratory training is crucial for ensuring optimal performance by surgeons. A structured and certified training program beforehand, focusing on trocar positioning and familiarizing with operating room protocols through dry simulations with the new system, is important for achieving significant outcomes. This training also prepares surgeons to effectively manage any technical challenges that may arise during surgery. , Blood loss The patients enrolled in this study had minimal blood loss (103.5 ± 62.43 mL), and none of them required blood transfusion. There is a general trend of reduced blood loss with hysterectomy performed using a RAS system compared with conventional laparoscopic hysterectomy, , but there are also reports of no significant advantage of using a RAS system over conventional laparoscopy, with the estimated blood loss reported to be in the range of 50–237 mL with robotic hysterectomy and 50–230.5 mL in patients who underwent conventional laparoscopic hysterectomy. Length of hospital stay In the present study, the hospital stay after surgical procedures performed using the Hugo™ RAS system was 2 days per hospital norms, and no patient reported a longer duration. A systematic chart review of 152 women who underwent robotic hysterectomy for benign indications suggested that robotic assistance contributes to a short hospital stay even in patients with complex pathology. However, there is a lack of evidence suggesting a statistically significant difference in the length of hospital stay when comparing gynecological procedures conducted with and without the use of a RAS system. For example, in a retrospective analysis (2018–2019) of surgical outcomes for patients who underwent hysterectomy for early endometrial cancer or benign indications using the da Vinci Xi system, the length of hospital stay was 3.74 ± 1.92 days, while it was 3.71 ± 2.29 days for patients who underwent conventional laparoscopic hysterectomy with no significant difference ( p = 0.150). It is important to note that the length of hospital stay in the present study was as per the hospital norm for this particular study. The patient was hospitalized 1 day prior to the surgery for preprocedural tests and to decide on their inclusion/exclusion in the study. Post-surgery, the pain scores were evaluated, the bowel movement was assessed the following day, and the patient was discharged, resulting in a 36-h stay. However, there is a possibility that the patient could have been discharged on the same day of the surgery in the absence of such a norm. Nevertheless, consideration was also given to the insurance practices followed in India. As per insurance protocols, medical expenses can only be claimed if the patient is hospitalized for a minimum of 24 h. Postoperative complications Evidence from a retrospective cohort study on patients who underwent robotic-assisted laparoscopic hysterectomy ( n = 745) and conventional laparoscopic hysterectomy ( n = 688) suggested that while rates of major postoperative complications were similar between robotic and laparoscopic surgery, the robotic approach exhibited notable advantages with lower rates of minor postoperative complications and conversion to laparotomy. In this study, only two patients experienced postoperative complications when the Hugo™ RAS system was employed, and both cases involved late complications successfully addressed through a conservative approach. The variations in the incidences of postoperative complications could be due to confounding factors such as differences in patient characteristics, complexity of the surgery, presence of comorbidities, and the experience of the surgeon. – Obesity is a major risk factor for postoperative complications in patients undergoing hysterectomy, but robot-assisted laparoscopic hysterectomy is associated with lower morbidity and reduced length of hospital stay compared with conventional laparoscopic hysterectomy in obese patients. , Robot-assisted laparoscopic surgery is reported to be more feasible than conventional laparoscopic surgery for obese patients and those with poor respiratory compliance. A retrospective chart review of 1032 obese patients (Body Mass Index > 30 kg/m 2 ) at two academic institutions in the United States showed that most obese patients could well tolerate gynecological surgeries performed using a RAS system, and only 3% of the study population developed pulmonary complications. Pain Compared with laparoscopic surgery, robotic surgery is considered to be associated with less postoperative pain and reduced need for analgesia, which could be possibly due to the instrument’s wrist-like movement occurring within the abdominal cavity and the minimized trauma to the abdominal wall. In the present study, the VAS score for postoperative pain at 6 h was 2.2 with the Hugo™ RAS system. In a retrospective analysis of patients who underwent robotic-assisted radical prostatectomy during 2019–2021 using the da Vinci Xi system ( n = 100), the postoperative pain score was reported to be 2.5 at 6 h. Navigating robotic surgery challenges with the Hugo™ RAS system despite the advantages associated with robotic surgery, the challenges cannot be overlooked. The absence of tactile feedback, the surgeon’s distant positioning from the patient, and increased expenses relative to conventional laparoscopic surgery are some of the prominent disadvantages of robot-assisted laparoscopic surgery. Lack of haptic feedback has been a concern with the da Vinci systems. In contrast, the Hugo™ RAS system offers haptic feedback to surgeons, allowing them to identify and prevent tissue damage during the surgical procedure. Various safety mechanisms incorporated in the Hugo™ RAS system enhance surgical precision and minimize risks, thereby ensuring patient safety. The six hinges on the arm carts greatly increase the range of motion, creating a configured arrangement that minimizes the risk of collisions. The higher expenses associated with the da Vinci systems primarily arise from the acquisition and maintenance of the system, increased instrument costs, the use of semi-disposable tools, and prolonged operative times. The Hugo™ RAS system is more sustainable because the tower could be adapted for use with pure laparoscopy and one of the arms could function independently as a stationary assistant in traditional laparoscopic procedures, thereby reducing the need for trained personnel. Analysis of the cost of robotic-assisted hysterectomies at a community teaching hospital in the United States demonstrated a 15.5% reduction in the total cost and 14.3% in the operative cost of hysterectomy procedures after 5 years of experience with the RAS system. Thus, costs associated with robotic surgery are expected to diminish over time with improved surgical expertise, reduced operative duration and length of hospital stay, increased popularity of high-volume robotic centers, and adoption of minimally invasive surgery. Comparing the Hugo™ RAS system with other robotic surgical platforms is essential as it will help in identifying the advantages of Hugo™ RAS over other systems in terms of efficiency, precision, outcomes, and potential areas of improvement. Limitations of this study This report is limited in its scope as it only records the early experiences of the usage of the Hugo™ RAS system from a single center and has a small sample size of 20 patients. Given the limited case volume, surgeons utilizing the Hugo™ RAS system may still be in the initial stages of their learning process.
Compared with conventional laparoscopic methods, RAS is generally associated with a longer operative time, which includes the time required for placement of ports, docking time, console time, and closing time. The docking phase encompasses activities such as moving the robotic arms into the surgical field, setting the arms into their respective port sites, and inserting the robotic instruments in the abdomen, which account for approximately 3.5% of the operative time. A systematic review of reports on hysterectomies performed using RAS systems (2016–2021) presented a wide variation in the operative time (75.4–306.0 min). The docking time and console time for the Hugo™ RAS system in the present study were 6.3 ± 2.0 min and 86.9 ± 20.3 min, respectively. In a prospective observational study that used the Hugo™ RAS system for 192 gynecological surgical procedures performed at a single center in Italy, the docking time was reported to be 5.1 ± 2.0 min and the console time was 111.5 ± 40.2 min. The data indicate that robotic set-up and docking with the new Hugo™ RAS system can be performed efficiently and that the specific robotic docking learning curve is comparable to preexisting data for other platforms. The Hugo™ RAS system has four independent arm carts, which makes it easier to park the device away from the patient while enabling precise docking of the surgical arm at the desired location on the surgical field. This convenience stands in contrast to the da Vinci ® Surgical System, which features four arms attached to a single base. In a prospective observational study in the Netherlands, 65 robot-assisted operations, including 14 gynecological procedures, were performed using the da Vinci Xi system (Intuitive Surgical Inc., USA), and the docking time was reported to be 7.8 ± 2.7 min. The time required for docking and the surgery is largely influenced by the experience of the surgeon, the complexity of the surgery, and the techniques used for the surgery. , Nevertheless, the presence of a learning curve underscores the importance of surgeons receiving comprehensive training to proficiently handle the docking process and carry out successful procedures with the Hugo™ RAS system. Preclinical laboratory training is crucial for ensuring optimal performance by surgeons. A structured and certified training program beforehand, focusing on trocar positioning and familiarizing with operating room protocols through dry simulations with the new system, is important for achieving significant outcomes. This training also prepares surgeons to effectively manage any technical challenges that may arise during surgery. ,
The patients enrolled in this study had minimal blood loss (103.5 ± 62.43 mL), and none of them required blood transfusion. There is a general trend of reduced blood loss with hysterectomy performed using a RAS system compared with conventional laparoscopic hysterectomy, , but there are also reports of no significant advantage of using a RAS system over conventional laparoscopy, with the estimated blood loss reported to be in the range of 50–237 mL with robotic hysterectomy and 50–230.5 mL in patients who underwent conventional laparoscopic hysterectomy.
In the present study, the hospital stay after surgical procedures performed using the Hugo™ RAS system was 2 days per hospital norms, and no patient reported a longer duration. A systematic chart review of 152 women who underwent robotic hysterectomy for benign indications suggested that robotic assistance contributes to a short hospital stay even in patients with complex pathology. However, there is a lack of evidence suggesting a statistically significant difference in the length of hospital stay when comparing gynecological procedures conducted with and without the use of a RAS system. For example, in a retrospective analysis (2018–2019) of surgical outcomes for patients who underwent hysterectomy for early endometrial cancer or benign indications using the da Vinci Xi system, the length of hospital stay was 3.74 ± 1.92 days, while it was 3.71 ± 2.29 days for patients who underwent conventional laparoscopic hysterectomy with no significant difference ( p = 0.150). It is important to note that the length of hospital stay in the present study was as per the hospital norm for this particular study. The patient was hospitalized 1 day prior to the surgery for preprocedural tests and to decide on their inclusion/exclusion in the study. Post-surgery, the pain scores were evaluated, the bowel movement was assessed the following day, and the patient was discharged, resulting in a 36-h stay. However, there is a possibility that the patient could have been discharged on the same day of the surgery in the absence of such a norm. Nevertheless, consideration was also given to the insurance practices followed in India. As per insurance protocols, medical expenses can only be claimed if the patient is hospitalized for a minimum of 24 h. Postoperative complications Evidence from a retrospective cohort study on patients who underwent robotic-assisted laparoscopic hysterectomy ( n = 745) and conventional laparoscopic hysterectomy ( n = 688) suggested that while rates of major postoperative complications were similar between robotic and laparoscopic surgery, the robotic approach exhibited notable advantages with lower rates of minor postoperative complications and conversion to laparotomy. In this study, only two patients experienced postoperative complications when the Hugo™ RAS system was employed, and both cases involved late complications successfully addressed through a conservative approach. The variations in the incidences of postoperative complications could be due to confounding factors such as differences in patient characteristics, complexity of the surgery, presence of comorbidities, and the experience of the surgeon. – Obesity is a major risk factor for postoperative complications in patients undergoing hysterectomy, but robot-assisted laparoscopic hysterectomy is associated with lower morbidity and reduced length of hospital stay compared with conventional laparoscopic hysterectomy in obese patients. , Robot-assisted laparoscopic surgery is reported to be more feasible than conventional laparoscopic surgery for obese patients and those with poor respiratory compliance. A retrospective chart review of 1032 obese patients (Body Mass Index > 30 kg/m 2 ) at two academic institutions in the United States showed that most obese patients could well tolerate gynecological surgeries performed using a RAS system, and only 3% of the study population developed pulmonary complications. Pain Compared with laparoscopic surgery, robotic surgery is considered to be associated with less postoperative pain and reduced need for analgesia, which could be possibly due to the instrument’s wrist-like movement occurring within the abdominal cavity and the minimized trauma to the abdominal wall. In the present study, the VAS score for postoperative pain at 6 h was 2.2 with the Hugo™ RAS system. In a retrospective analysis of patients who underwent robotic-assisted radical prostatectomy during 2019–2021 using the da Vinci Xi system ( n = 100), the postoperative pain score was reported to be 2.5 at 6 h. Navigating robotic surgery challenges with the Hugo™ RAS system despite the advantages associated with robotic surgery, the challenges cannot be overlooked. The absence of tactile feedback, the surgeon’s distant positioning from the patient, and increased expenses relative to conventional laparoscopic surgery are some of the prominent disadvantages of robot-assisted laparoscopic surgery. Lack of haptic feedback has been a concern with the da Vinci systems. In contrast, the Hugo™ RAS system offers haptic feedback to surgeons, allowing them to identify and prevent tissue damage during the surgical procedure. Various safety mechanisms incorporated in the Hugo™ RAS system enhance surgical precision and minimize risks, thereby ensuring patient safety. The six hinges on the arm carts greatly increase the range of motion, creating a configured arrangement that minimizes the risk of collisions. The higher expenses associated with the da Vinci systems primarily arise from the acquisition and maintenance of the system, increased instrument costs, the use of semi-disposable tools, and prolonged operative times. The Hugo™ RAS system is more sustainable because the tower could be adapted for use with pure laparoscopy and one of the arms could function independently as a stationary assistant in traditional laparoscopic procedures, thereby reducing the need for trained personnel. Analysis of the cost of robotic-assisted hysterectomies at a community teaching hospital in the United States demonstrated a 15.5% reduction in the total cost and 14.3% in the operative cost of hysterectomy procedures after 5 years of experience with the RAS system. Thus, costs associated with robotic surgery are expected to diminish over time with improved surgical expertise, reduced operative duration and length of hospital stay, increased popularity of high-volume robotic centers, and adoption of minimally invasive surgery. Comparing the Hugo™ RAS system with other robotic surgical platforms is essential as it will help in identifying the advantages of Hugo™ RAS over other systems in terms of efficiency, precision, outcomes, and potential areas of improvement.
Evidence from a retrospective cohort study on patients who underwent robotic-assisted laparoscopic hysterectomy ( n = 745) and conventional laparoscopic hysterectomy ( n = 688) suggested that while rates of major postoperative complications were similar between robotic and laparoscopic surgery, the robotic approach exhibited notable advantages with lower rates of minor postoperative complications and conversion to laparotomy. In this study, only two patients experienced postoperative complications when the Hugo™ RAS system was employed, and both cases involved late complications successfully addressed through a conservative approach. The variations in the incidences of postoperative complications could be due to confounding factors such as differences in patient characteristics, complexity of the surgery, presence of comorbidities, and the experience of the surgeon. – Obesity is a major risk factor for postoperative complications in patients undergoing hysterectomy, but robot-assisted laparoscopic hysterectomy is associated with lower morbidity and reduced length of hospital stay compared with conventional laparoscopic hysterectomy in obese patients. , Robot-assisted laparoscopic surgery is reported to be more feasible than conventional laparoscopic surgery for obese patients and those with poor respiratory compliance. A retrospective chart review of 1032 obese patients (Body Mass Index > 30 kg/m 2 ) at two academic institutions in the United States showed that most obese patients could well tolerate gynecological surgeries performed using a RAS system, and only 3% of the study population developed pulmonary complications.
Compared with laparoscopic surgery, robotic surgery is considered to be associated with less postoperative pain and reduced need for analgesia, which could be possibly due to the instrument’s wrist-like movement occurring within the abdominal cavity and the minimized trauma to the abdominal wall. In the present study, the VAS score for postoperative pain at 6 h was 2.2 with the Hugo™ RAS system. In a retrospective analysis of patients who underwent robotic-assisted radical prostatectomy during 2019–2021 using the da Vinci Xi system ( n = 100), the postoperative pain score was reported to be 2.5 at 6 h. Navigating robotic surgery challenges with the Hugo™ RAS system despite the advantages associated with robotic surgery, the challenges cannot be overlooked. The absence of tactile feedback, the surgeon’s distant positioning from the patient, and increased expenses relative to conventional laparoscopic surgery are some of the prominent disadvantages of robot-assisted laparoscopic surgery. Lack of haptic feedback has been a concern with the da Vinci systems. In contrast, the Hugo™ RAS system offers haptic feedback to surgeons, allowing them to identify and prevent tissue damage during the surgical procedure. Various safety mechanisms incorporated in the Hugo™ RAS system enhance surgical precision and minimize risks, thereby ensuring patient safety. The six hinges on the arm carts greatly increase the range of motion, creating a configured arrangement that minimizes the risk of collisions. The higher expenses associated with the da Vinci systems primarily arise from the acquisition and maintenance of the system, increased instrument costs, the use of semi-disposable tools, and prolonged operative times. The Hugo™ RAS system is more sustainable because the tower could be adapted for use with pure laparoscopy and one of the arms could function independently as a stationary assistant in traditional laparoscopic procedures, thereby reducing the need for trained personnel. Analysis of the cost of robotic-assisted hysterectomies at a community teaching hospital in the United States demonstrated a 15.5% reduction in the total cost and 14.3% in the operative cost of hysterectomy procedures after 5 years of experience with the RAS system. Thus, costs associated with robotic surgery are expected to diminish over time with improved surgical expertise, reduced operative duration and length of hospital stay, increased popularity of high-volume robotic centers, and adoption of minimally invasive surgery. Comparing the Hugo™ RAS system with other robotic surgical platforms is essential as it will help in identifying the advantages of Hugo™ RAS over other systems in terms of efficiency, precision, outcomes, and potential areas of improvement.
This report is limited in its scope as it only records the early experiences of the usage of the Hugo™ RAS system from a single center and has a small sample size of 20 patients. Given the limited case volume, surgeons utilizing the Hugo™ RAS system may still be in the initial stages of their learning process.
Drawing from the initial experience, the use of the Hugo™ RAS system provides favorable outcomes for patients with gynecological conditions, demonstrating advantages in terms of short docking time and surgery duration, minimal blood loss, short length of hospital stay, few postoperative complications, and low pain perception. Further studies with more extensive sample sizes comparing the Hugo™ RAS system with other RAS systems and traditional laparoscopic procedures in the gynecological domain are needed.
sj-doc-1-whe-10.1177_17455057241302581 – Supplemental material for Robotics-assisted surgery in gynecology: A single-center experience with the Hugo™ RAS system in India Supplemental material, sj-doc-1-whe-10.1177_17455057241302581 for Robotics-assisted surgery in gynecology: A single-center experience with the Hugo™ RAS system in India by Manjula Anagani, Ravula Sindura Ganga and Snehalatha Paritala in Women’s Health
|
Pharmacogenetic and clinical risk factors for bevacizumab-related gastrointestinal hemorrhage in prostate cancer patients treated on CALGB 90401 (Alliance) | 2792b780-f9f1-45dd-bb1a-0015b9843e23 | 10912014 | Pharmacology[mh] | Bevacizumab is a recombinant humanized monoclonal antibody targeting circulating vascular endothelial growth factor (VEGF) and is approved to treat many solid tumor malignancies . VEGF is recognized as an essential regulator of normal and abnormal blood vessel growth . It induces factor III, von Willebrand factor and tissue plasminogen activator, and is involved in the synthesis of nitric oxide and prostacyclin . Inhibition of these processes may result in vascular instability contributing to thrombotic or hemorrhagic events, depending on the delicate balance of factors in a given patient . Furthermore, increased clot formation and vasoconstriction of the splanchnic vasculature can lead to bowel perforation . The significance of bevacizumab-related hemorrhage is highlighted by a Food and Drug Administration (FDA) issued black box warning . The estimated incidence of all-grade and high-grade (grade 3–5) hemorrhage is approximately 30 and 4%, respectively . Although bevacizumab-related fatal adverse events are rare (2%), the most common cause of treatment-related fatality are hemorrhagic events (25%), primarily involving the lungs, central nervous system (CNS) and gastrointestinal system . Studies of bevacizumab-treated patients also suggest an increased risk of hemorrhage at the site of the primary tumor . Furthermore, a recent pharmacovigilance study showed significantly higher risks of artery dissections or aneurysms in those receiving antiangiogenic drugs, including bevacizumab . Clinical factors proposed to increase overall bleeding risk (independent of treatment) include recent major bleed, anemia, thrombocytopenia, cancer and antiplatelet/anticoagulant use . Patients with cancer have increased bleeding risk when receiving anticoagulants compared to patients without cancer . Genetic mutations can also result in hereditary bleeding disorders including hemophilia A (factor VIII deficiency), hemophilia B (factor IX deficiency), von Willebrand disease and other rare conditions . One meta-analysis suggests an association between MTHFR gene polymorphisms and risk of intracranial hemorrhage , whereas other studies suggest a possible polygenic contribution to certain types of hemorrhagic stroke, such as subarachnoid hemorrhage , but the evidence base is lacking. Given the lack of data on risk factors for bevacizumab-related hemorrhage, the primary objective of this study was to identify potential clinical and pharmacogenetic risk factors for bevacizumab-related grade 2 or higher (2+) gastrointestinal hemorrhage using a genome-wide approach in metastatic castration-resistant prostate cancer patients treated on Cancer and Leukemia Group B (CALGB) 90401 . CALGB is now part of the Alliance for Clinical Trials in Oncology. CALGB 90401 was a placebo-controlled double-blinded phase III trial that equally randomized men with metastatic castration-resistant prostate cancer to receive docetaxel 75 mg/m 2 on day 1 of every 21-day cycle and prednisone 5 mg twice daily with or without bevacizumab 15 mg/kg on day 1 of every 21-day cycle, for up to two years . Randomization was stratified by age, prior history of arterial events, and Halabi prognostic risk groups . Patients were randomized to the trial from May 2005 to December 2007. Details on treatment and eligibility can be found in the original clinical trial publication . Patients from the CALGB 90401 parent study who provided IRB-approved informed consent for the pharmacogenetic substudy (CALGB 60404) were eligible for the genome-wide association (GWAS) analysis. Toxicity data were collected prospectively by the Alliance Statistics and Data Management Center at each treatment cycle on standardized forms that mandated reporting of all grade solicited toxicities, including gastrointestinal hemorrhage, and grade 3 or higher unsolicited toxicities, as defined by National Cancer Institute Common Toxicity Criteria for Adverse Events Version 3.0 (NCI-CTCAEv3.0) . A gastrointestinal hemorrhage event for the analysis was defined as any grade 2+ gastrointestinal hemorrhage defined by the NCI-CTCAEv3.0 as “symptomatic and medical intervention or minor cauterization indicated” and considered possibly, probably, or definitely related to therapy per provider report. Medical history was collected from the patient on standardized prestudy forms that documented prior history of peptic ulcer disease (PUD), hemorrhage and/or gastrointestinal perforation, and smoking. Prior history of hemorrhage/gastrointestinal perforation was limited to events that occurred within five years prior to enrollment. Baseline hemoglobin and antiplatelet/anticoagulant use was recorded along with dose and frequency. Genotyping A 10-mL sample of venous blood was collected prior to receiving treatment from all patients providing consent for the pharmacogenetic companion study . Genotyping was conducted using the HumanHap610-Quad Genotyping BeadChip (Illumina Inc., CA, USA) at the RIKEN Center for Genomic Medicine (Yokohama City, Japan). Genotype data are available at dbGaP ( https://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id=phs001002.v1.p1 ). Patients with genotype call rates <95% and SNPs with call rates <99%, poor genotype clustering, indeterminate or unreliable loci, deviation from Hardy-Weinberg equilibrium (HWE) and non-autosomal loci were excluded prior to analysis. After removing SNPs with a HWE p < 10 −8 and those with a minor allele frequency <0.05, 498,081 SNPs remained for the analysis. Eigensoft methods were used to visualize genetic ancestry and identify the subset of patients who were genetically estimated European , in whom the final pharmacogenetic analysis was conducted to avoid population stratification effects (i.e., when cases and controls are sampled from genetically different underlying populations where allele frequencies vary causing any associations observed to be due to sampling differences). Data quality was ensured by review of data by the Alliance Statistics and Data Management Center and by the study chairperson following Alliance policies. Statistical analysis Statistical analyses were conducted by the Alliance Statistics and Data Management Center on a data set locked on January 28, 2014. All analyses were performed using R software (version 3.1.0) and (version 2.37-7) , SAS software version 9.2 and GenABEL (version 1.8-0) . A competing risks model was utilized where the event of interest (grade 2+ gastrointestinal hemorrhage) was subject to three dependent informative censoring mechanisms: progression/death, other treatment-terminating adverse events, or ‘other’, which included lost to follow-up, withdrawal for reasons other than toxicity, or incomplete information . Patients who did not complete two years of therapy due to any of these competing risks, prior to experiencing the event of interest, were informatively censored at the cumulative bevacizumab cycles received. All analyses were conducted by considering the influence of bevacizumab therapy and cause-specific hazard. A multivariable Cox regression model was used to identify baseline covariates associated with gastrointestinal hemorrhage, including age (continuous), history of PUD (yes vs. no), history of hemorrhage (yes vs. no), history of smoking (yes vs. no), hemoglobin (continuous) and baseline antiplatelet/anticoagulant use (yes vs. no). The Log-rank statistic was used to test the association of all 498,081 SNPs that passed quality control with cumulative bevacizumab/placebo cycle at occurrence of grade 2+ gastrointestinal hemorrhage. A dominant genetic model was assumed as data suggest it is more robust against sampling deviations from HWE frequencies than the additive model and the higher likelihood within a GWAS to identify variants with low minor allele frequency (MAF) . The Bonferroni-corrected p-value threshold for significance was set at 1 × 10 −7 . The 1000 SNPs that had the strongest association with gastrointestinal hemorrhage were adjusted for baseline covariates with p < 0.1 in multivariable analysis. The 100 SNPs with the strongest association after adjustment for clinical covariates were further interrogated for biological function and cross referenced with previously described SNPs associated with hereditary bleeding disorders. HaploReg v2 was used to investigate linkage disequilibrium (LD) between SNPs and altered transcription factors (high LD defined as r 2 ≥ 0.90). PrediXcan is a gene-based association method that correlates imputed gene expression with the phenotype of interest to identify genes involved in the etiology of the phenotype . This methodology was used to test the genetically predicted gene expression using Depression Genes and Networks (DGN) whole blood elastic net model for association with grade 2+ gastrointestinal hemorrhage. The whole blood model was used because of its relevance to hemorrhage and large number of genotyped samples. The Bonferroni-corrected p-value threshold for significance was set at 9 × 10 −5 , based on 556 genes tested. To search for cumulative genetic effects and provide insight into mechanisms, functions, and pathways involved with bevacizumab-related hemorrhage, genes with an unadjusted p < 0.05 from the PrediXcan results were included in a pathway enrichment analysis using Ingenuity Pathway Analysis version 17199142 (Ingenuity ® Systems, Inc, www.ingenuity.com ). A phenome-wide association study (PheWAS) was performed to assess the association between genetic polymorphisms with phenotypes by linking genotyping from BioVu, an independent de-identified DNA data bank at Vanderbilt University, to a broad range of electronic medical record–derived clinical phenotypes . Briefly, phecodes are defined by hierarchical groupings of ICD-9 and ICD-10 codes. Cases are defined as individuals with 2 or more phecodes on unique dates and controls are individuals who do not have a phecode of interest. We used phecodes version 1.2 (available at https://phewascatalog.org/phecodes ). Genotyping was performed using the Illumina MEGA array and quality control protocol developed by the Vanderbilt Epidemiology Center in cooperation with Vanderbilt Technologies for Advanced Genomics Analysis and Research Design. Further details on the PheWAS approach can be found in prior publications . The PheWAS was conducted on 72,083 individuals of European ancestry and performed using a logistic regression, controlling for age and sex. Phecode associations with SNPs yielding uncorrected P- values lower than 0.05 were summarized. The Bonferroni-corrected p -value threshold for significance was set at 3.0 × 10 −5 , based on 1508 Phecodes tested. A 10-mL sample of venous blood was collected prior to receiving treatment from all patients providing consent for the pharmacogenetic companion study . Genotyping was conducted using the HumanHap610-Quad Genotyping BeadChip (Illumina Inc., CA, USA) at the RIKEN Center for Genomic Medicine (Yokohama City, Japan). Genotype data are available at dbGaP ( https://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id=phs001002.v1.p1 ). Patients with genotype call rates <95% and SNPs with call rates <99%, poor genotype clustering, indeterminate or unreliable loci, deviation from Hardy-Weinberg equilibrium (HWE) and non-autosomal loci were excluded prior to analysis. After removing SNPs with a HWE p < 10 −8 and those with a minor allele frequency <0.05, 498,081 SNPs remained for the analysis. Eigensoft methods were used to visualize genetic ancestry and identify the subset of patients who were genetically estimated European , in whom the final pharmacogenetic analysis was conducted to avoid population stratification effects (i.e., when cases and controls are sampled from genetically different underlying populations where allele frequencies vary causing any associations observed to be due to sampling differences). Data quality was ensured by review of data by the Alliance Statistics and Data Management Center and by the study chairperson following Alliance policies. Statistical analyses were conducted by the Alliance Statistics and Data Management Center on a data set locked on January 28, 2014. All analyses were performed using R software (version 3.1.0) and (version 2.37-7) , SAS software version 9.2 and GenABEL (version 1.8-0) . A competing risks model was utilized where the event of interest (grade 2+ gastrointestinal hemorrhage) was subject to three dependent informative censoring mechanisms: progression/death, other treatment-terminating adverse events, or ‘other’, which included lost to follow-up, withdrawal for reasons other than toxicity, or incomplete information . Patients who did not complete two years of therapy due to any of these competing risks, prior to experiencing the event of interest, were informatively censored at the cumulative bevacizumab cycles received. All analyses were conducted by considering the influence of bevacizumab therapy and cause-specific hazard. A multivariable Cox regression model was used to identify baseline covariates associated with gastrointestinal hemorrhage, including age (continuous), history of PUD (yes vs. no), history of hemorrhage (yes vs. no), history of smoking (yes vs. no), hemoglobin (continuous) and baseline antiplatelet/anticoagulant use (yes vs. no). The Log-rank statistic was used to test the association of all 498,081 SNPs that passed quality control with cumulative bevacizumab/placebo cycle at occurrence of grade 2+ gastrointestinal hemorrhage. A dominant genetic model was assumed as data suggest it is more robust against sampling deviations from HWE frequencies than the additive model and the higher likelihood within a GWAS to identify variants with low minor allele frequency (MAF) . The Bonferroni-corrected p-value threshold for significance was set at 1 × 10 −7 . The 1000 SNPs that had the strongest association with gastrointestinal hemorrhage were adjusted for baseline covariates with p < 0.1 in multivariable analysis. The 100 SNPs with the strongest association after adjustment for clinical covariates were further interrogated for biological function and cross referenced with previously described SNPs associated with hereditary bleeding disorders. HaploReg v2 was used to investigate linkage disequilibrium (LD) between SNPs and altered transcription factors (high LD defined as r 2 ≥ 0.90). PrediXcan is a gene-based association method that correlates imputed gene expression with the phenotype of interest to identify genes involved in the etiology of the phenotype . This methodology was used to test the genetically predicted gene expression using Depression Genes and Networks (DGN) whole blood elastic net model for association with grade 2+ gastrointestinal hemorrhage. The whole blood model was used because of its relevance to hemorrhage and large number of genotyped samples. The Bonferroni-corrected p-value threshold for significance was set at 9 × 10 −5 , based on 556 genes tested. To search for cumulative genetic effects and provide insight into mechanisms, functions, and pathways involved with bevacizumab-related hemorrhage, genes with an unadjusted p < 0.05 from the PrediXcan results were included in a pathway enrichment analysis using Ingenuity Pathway Analysis version 17199142 (Ingenuity ® Systems, Inc, www.ingenuity.com ). A phenome-wide association study (PheWAS) was performed to assess the association between genetic polymorphisms with phenotypes by linking genotyping from BioVu, an independent de-identified DNA data bank at Vanderbilt University, to a broad range of electronic medical record–derived clinical phenotypes . Briefly, phecodes are defined by hierarchical groupings of ICD-9 and ICD-10 codes. Cases are defined as individuals with 2 or more phecodes on unique dates and controls are individuals who do not have a phecode of interest. We used phecodes version 1.2 (available at https://phewascatalog.org/phecodes ). Genotyping was performed using the Illumina MEGA array and quality control protocol developed by the Vanderbilt Epidemiology Center in cooperation with Vanderbilt Technologies for Advanced Genomics Analysis and Research Design. Further details on the PheWAS approach can be found in prior publications . The PheWAS was conducted on 72,083 individuals of European ancestry and performed using a logistic regression, controlling for age and sex. Phecode associations with SNPs yielding uncorrected P- values lower than 0.05 were summarized. The Bonferroni-corrected p -value threshold for significance was set at 3.0 × 10 −5 , based on 1508 Phecodes tested. CALGB 90401 enrolled 1050 patients, of whom 1008 received treatment and were included in the clinical analysis (503 received bevacizumab and 505 received placebo). Patients in both arms received a median of eight cycles The GWAS included 616 genetically estimated European patients (314 received bevacizumab and 302 received placebo) (Fig. ). Baseline demographics were similar between the clinical and GWAS cohorts, and bevacizumab and placebo treated patients (Table ). Using a time-to-event competing risk model, the rate of grade 2+ gastrointestinal hemorrhage by treatment arm (bevacizumab and placebo) in the clinical cohort was 9.5% (48/503) and 3.8% (19/505) (HR = 2.7; 95% CI 1.53–4.93; P < 0.001) and in the GWAS cohort was 9.6% (30/314) and 2.0% (6/302) (HR = 5.2; 95% CI 2.09–15.52; P < 0.001), respectively. Clinical risk factors In multivariable analysis (Table ), bevacizumab treatment was confirmed to increase grade 2+ gastrointestinal hemorrhage risk compared to placebo in the clinical (HR 2.88; 95% CI: 1.64–5.06; P < 0.001) and GWAS cohorts (HR = 5.70; 95% CI: 2.17–14.97; P < 0.001). Of the remaining covariates, increasing age as a continuous variable also increased grade 2+ gastrointestinal hemorrhage risk in the clinical (HR 1.05; 95% CI 1.02–1.09; P = 0.002) and GWAS cohorts (HR = 1.06; 95% CI: 1.01–1.12; P = 0.01). History of PUD was significantly associated with grade 2+ gastrointestinal hemorrhage in the univariate analysis but did not maintain significance in the multivariable analysis (clinical cohort: HR 2.20, 95% CI: 0.98–4.90, P = 0.055; GWAS cohort: HR 2.61, 95% CI 0.87–7.83, P = 0.09). There was lack of association between history of hemorrhage, antiplatelet/anticoagulant use, smoking, and hemoglobin with increased hemorrhage risk (all p > 0.05). SNP analysis The top 10 SNPs ranked by p -value of their association with grade 2+ gastrointestinal hemorrhage, adjusted for bevacizumab treatment, age, and history of PUD, are reported in Table , along with the rsID, gene annotation, MAF, and adjusted cause-specific HR. Of 498,081 SNPs tested, one intergenic SNP (rs1478947, MAF = 0.06; HR 6.26; 95% CI 3.19–12.28; P = 9.40 × 10 −8 ) surpassed Bonferroni-corrected significance ( P = 1.0 × 10 −7 ) for association with grade 2+ gastrointestinal hemorrhage (see Fig. for Manhattan plot). The rate of grade 2+ gastrointestinal hemorrhage was 33.3% (13/39) and 6.2% (17/275) in bevacizumab-treated patients with the AA/AG and GG genotypes, respectively, while the incidence in the placebo arm was 2.9% (1/35) and 1.9% (5/267) (Fig. ). The one patient who carried the AA genotype was randomized to receive bevacizumab and experienced a grade 3 gastrointestinal hemorrhage during cycle 3, which resulted in treatment withdrawal. There was no significant interaction between SNP and treatment arm ( P = 0.23). Predicted gene expression analysis The top 10 genes ranked according to the statistical association between genotype-predicted gene expression (i.e., PrediXcan) and grade 2+ gastrointestinal hemorrhage by adjusted p -value, along with the corresponding adjusted cause-specific HR, are reported in Supplementary Table . No genes were significantly associated with gastrointestinal hemorrhage risk after adjusting for multiple comparisons. The top ranked gene, ITPA , encodes the protein inosine triphosphate pyrophosphohydrolase ( ITPA , HR = 5.77; 95% CI 2.29–14.56; P = 2.03 × 10 −4 ). Supplementary Table summarizes the top canonical pathways enriched from the gene-based pathway analysis, their respective p -values, and genes mapped to each pathway. Of these, Heme Biosynthesis from Uroporphyrinogen-III I, the third most highly enriched pathway ( P = 0.0038), was identified as having possible biological relevance to hemorrhage. PheWAS Of 1508 diagnosis phecodes tested, 66 are summarized with a p < 0.05, none of which were directly related to gastrointestinal hemorrhage, but several possibly indirectly related and associated with anemia and thrombocytopenia (Supplementary Table ). No associations surpassed Bonferroni correction. In multivariable analysis (Table ), bevacizumab treatment was confirmed to increase grade 2+ gastrointestinal hemorrhage risk compared to placebo in the clinical (HR 2.88; 95% CI: 1.64–5.06; P < 0.001) and GWAS cohorts (HR = 5.70; 95% CI: 2.17–14.97; P < 0.001). Of the remaining covariates, increasing age as a continuous variable also increased grade 2+ gastrointestinal hemorrhage risk in the clinical (HR 1.05; 95% CI 1.02–1.09; P = 0.002) and GWAS cohorts (HR = 1.06; 95% CI: 1.01–1.12; P = 0.01). History of PUD was significantly associated with grade 2+ gastrointestinal hemorrhage in the univariate analysis but did not maintain significance in the multivariable analysis (clinical cohort: HR 2.20, 95% CI: 0.98–4.90, P = 0.055; GWAS cohort: HR 2.61, 95% CI 0.87–7.83, P = 0.09). There was lack of association between history of hemorrhage, antiplatelet/anticoagulant use, smoking, and hemoglobin with increased hemorrhage risk (all p > 0.05). The top 10 SNPs ranked by p -value of their association with grade 2+ gastrointestinal hemorrhage, adjusted for bevacizumab treatment, age, and history of PUD, are reported in Table , along with the rsID, gene annotation, MAF, and adjusted cause-specific HR. Of 498,081 SNPs tested, one intergenic SNP (rs1478947, MAF = 0.06; HR 6.26; 95% CI 3.19–12.28; P = 9.40 × 10 −8 ) surpassed Bonferroni-corrected significance ( P = 1.0 × 10 −7 ) for association with grade 2+ gastrointestinal hemorrhage (see Fig. for Manhattan plot). The rate of grade 2+ gastrointestinal hemorrhage was 33.3% (13/39) and 6.2% (17/275) in bevacizumab-treated patients with the AA/AG and GG genotypes, respectively, while the incidence in the placebo arm was 2.9% (1/35) and 1.9% (5/267) (Fig. ). The one patient who carried the AA genotype was randomized to receive bevacizumab and experienced a grade 3 gastrointestinal hemorrhage during cycle 3, which resulted in treatment withdrawal. There was no significant interaction between SNP and treatment arm ( P = 0.23). The top 10 genes ranked according to the statistical association between genotype-predicted gene expression (i.e., PrediXcan) and grade 2+ gastrointestinal hemorrhage by adjusted p -value, along with the corresponding adjusted cause-specific HR, are reported in Supplementary Table . No genes were significantly associated with gastrointestinal hemorrhage risk after adjusting for multiple comparisons. The top ranked gene, ITPA , encodes the protein inosine triphosphate pyrophosphohydrolase ( ITPA , HR = 5.77; 95% CI 2.29–14.56; P = 2.03 × 10 −4 ). Supplementary Table summarizes the top canonical pathways enriched from the gene-based pathway analysis, their respective p -values, and genes mapped to each pathway. Of these, Heme Biosynthesis from Uroporphyrinogen-III I, the third most highly enriched pathway ( P = 0.0038), was identified as having possible biological relevance to hemorrhage. Of 1508 diagnosis phecodes tested, 66 are summarized with a p < 0.05, none of which were directly related to gastrointestinal hemorrhage, but several possibly indirectly related and associated with anemia and thrombocytopenia (Supplementary Table ). No associations surpassed Bonferroni correction. To our knowledge, this report is the first to investigate both clinical and pharmacogenetic risk factors for bevacizumab-related gastrointestinal hemorrhage, which was collected prospectively from a large, randomized phase III trial. Consistent with previous reports, bevacizumab treatment significantly increased gastrointestinal hemorrhage risk compared to placebo. Increasing age, prior PUD, and an intergenic SNP (rs1478947) were associated with cause-specific gastrointestinal hemorrhage risk in CALGB (Alliance) 90401. Incidence rates of grade 2 (and thus grade 2 or higher) gastrointestinal hemorrhage are infrequently reported in prior bevacizumab studies given these trials typically reported grade 3 or higher events and combine grade 1 and 2; however, grade 2 events were included in this analysis given the high risk for severe complications, hospitalization, and the need for medical intervention. While bevacizumab is not approved or used in patients with prostate cancer, a previous study reported that 1.7% and 3.4% of advanced ovarian cancer patients receiving chemotherapy plus placebo or chemotherapy plus bevacizumab experienced a grade 2+ gastrointestinal adverse event, including perforation, fistula, hemorrhage or necrosis, respectively; however, these patients were relatively younger and fewer had a history of PUD . Two large meta-analyses of bevacizumab-treated patients have reported incidence rates of overall grade 3+ hemorrhage of 2.8% and 3.5% , one of which reported the highest incidence for gastrointestinal hemorrhage (2.3%) . The incidence of grade 3+ gastrointestinal hemorrhage in the CALGB 90401 parent study (~6.0%) was relatively high compared to previous meta-analyses – this may be partially due to administration of higher bevacizumab doses and an older population. Compared to other solid tumors, patients with prostate cancer have similar rates of severe bleeding (<5% and mostly hematuria) , like the ~4% observed in our clinical cohort receiving placebo. However, bleeding rates are generally increased in cancer patients receiving anticoagulant treatment, those with metastatic disease, and patients with gastrointestinal primary malignancies and/or liver metastases . Older patients and those with a history of PUD were at increased risk for gastrointestinal hemorrhage during treatment. Consistent with these findings, previous studies have identified increasing age and history of PUD as risk factors for gastrointestinal hemorrhage in patients receiving other drugs known to increase bleeding risk, such as aspirin and nonsteroidal anti-inflammatory drugs . Previous studies have also demonstrated that bevacizumab treatment does not increase the risk of severe bleeding in patients receiving therapeutic anticoagulation compared to those not receiving anticoagulation . We did not observe an association between anticoagulant/antiplatelet use and gastrointestinal hemorrhage risk. Thrombocytopenia was not investigated as a risk factor since patients were required to have a minimum platelet count of 100,000/µl for enrollment and only 5% of patients had a platelet count of ≤150,000/µl. A genome-wide analysis was undertaken to identify SNPs associated with gastrointestinal hemorrhage risk. Previously reported SNPs associated with hereditary bleeding disorders, such as hemophilia A and B and von Willebrand’s disease , were not observed among the top 1000 SNPs in this analysis. No genes related to the top 100 SNPs were found to have a direct biological relationship to hemorrhage risk; however, one intergenic SNP (rs1478947) surpassed genome-wide significance. Rs1478947 is located on chromosome 8 at position 97471007 (Genome Reference Consortium Human Build 38)with a MAF 0.05 in HapMap CEU samples and 0.06 in the study population. Although rs1478947 genotype significantly increased gastrointestinal hemorrhage risk regardless of treatment arm, there was more than a 5-fold increase in toxicity rate in patients with the AA/AG versus GG genotype receiving bevacizumab compared to an approximately 1.5-fold increase in patients receiving placebo (Fig. ). This suggests the effect of rs1478947 may be potentiated in patients receiving bevacizumab; however, this could not be confirmed directly via statistical interaction analysis. Although the exact mechanism by which rs1478947 might increase gastrointestinal hemorrhage risk remains unclear, rs1478947 is in complete LD (r 2 = 1.0) with rs1478948 (located 30 base pairs upstream from rs1478947), variations of which may alter the binding motif for transcription factor hepatocyte nuclear factor-4 (HNF4). 26 HNF4A (located on chromosome 20) is localized to intestine, kidneys, and liver, whereas HNF4G is located on chromosome 8 (same as rs1478947) and primarily expressed in the small bowel . Prior reports demonstrated that HNF4 exerts a positive regulatory effect on clotting factor VII (fVII) expression and binding of this transcription factor is critical for normal fVII function . Mutations in the HNF4 binding site within the fVII promoter result in altered binding, severe fVII deficiency and an increased risk of bleeding . Further, mutations in the HNF4 binding site of factor IX also cause severe bleeding disorders . The Leyden phenotype of the severe bleeding disorder hemophilia B is caused by several point mutations within the promoter region, of which a number map in the HNF4 binding site . It could be hypothesized that increased gastrointestinal damage by presence of rs1478947 combined with bevacizumab may increase bleeding risk in the gastrointestinal tract. PrediXcan was used to test genetically predicted gene expression for association with gastrointestinal hemorrhage . The top ranked gene, ITPA , did not surpass Bonferroni corrected significance. Polymorphisms in ITPA have been shown to increase the risk of drug-induced anemia and thrombocytopenia , which are general risk factors for increased bleeding. Ingenuity Pathway Analysis of genes from the PrediXcan results identified several enriched canonical pathways, including Heme Biosynthesis from Uroporphyrinogen III I (Supplementary Table ). Uroporphyrinogen III is the precursor for synthesis of vitamin B12, chlorophyll, and heme. Defects in this pathway can result in porphyrias, which primarily result in neurological symptoms ; however, abnormalities in heme synthesis may also result in increased risk of hemorrhage as well as oxidative stress post-hemorrhage . PheWAS uses a reverse genetics approach to associate genetic variants with phenotypes by linking a database of deidentified genotypes to a broad range of electronic medical record-derived clinical phenotypes. The most common categories were endocrine/metabolic, hematopoietic, mental disorders, and circulatory system, but none surpassed statistical significance (Supplementary Table ). This analysis has limitations, including limited sample size, lack of available independent cohorts of bevacizumab-treated patients with prospectively collected toxicity data and banked DNA for replication, and absence of functional studies to determine biological plausibility. The success of any pharmacogenomics GWAS depends on effect size, allele frequency of genetic variants that impact the event of interest, the population (including race/ethnicity), and study design. GWAS by nature have little power due to multiple testing, thus requiring large sample sizes (e.g., >1000) to detect significant associations with low MAF. Importantly, large disease susceptibility GWAS (with sample sizes much larger than this analysis) have discovered and validated genetic predictors of complex phenotypes, including obesity , diabetes , and others, in intergenic regions of the DNA that have not yet been linked to a biological mechanism. It remains unknown how these variations mechanistically influence outcome, although one may speculate transcriptional regulation, long-range promoter regulation, variations in micro- or non-coding RNA, and other factors. Databases such as Encyclopedia of DNA Elements (Encode; https://genome.ucsc.edu/ENCODE ), The Genotype-Tissue Expression (GTEx) project ( http://www.gtexportal.org/home ), MiTranscriptome ( http://mitranscriptome.org ), and others, are built to increase the understanding of the biological consequences of intergenic variants; however, currently available data and understanding are still limited. Functional studies are critical to determine the biological plausibility of GWAS findings. Induced pluripotent stem cell-derived endothelial cell (iPSC-EC) responses may be a novel approach to evaluate sensitivity to anti-VEGF agents . Future studies can use iPSC-ECs to assess gene expression profiles associated with drug sensitivity or specific phenotypes. Lastly, this study was limited to genetically estimated Europeans based on the population enrolled on the parent prospective clinical trial and there is a critical need to conduct GWAS in diverse populations. Notably, the frequency of rs1478947 in Africans is 0.09 based on 1000 Genomes, thus replication in this population is critical. In conclusion, this analysis confirms the increased risk of gastrointestinal hemorrhage associated with bevacizumab treatment and increasing age. A potentially novel association was noted between an intergenic SNP, rs1478947, with gastrointestinal hemorrhage; however, the relatively low MAF, lack of a direct biological relationship to hemorrhage risk based on current literature, lack of functional studies, and unavailable replication cohort limit the interpretation of rs1478947 as a potential predictor of hemorrhage risk. This intronic variant, while removed during transcription, may serve a regulatory function, or involved in alternative splicing. It is in complete LD with rs1478948, which we hypothesize may have putative functionality in altering the binding motify of HNF4 (previously reported to be associated with hemorrhage risk). This finding may be particularly important in other solid tumors that are more prone to gastrointestinal hemorrhage and in which bevacizumab is more commonly used, like certain gastrointestinal malignancies (e.g., colorectal cancer). Future studies should focus on identifying and replicating factors that influence treatment-related hemorrhage risk, with the goal of building an algorithm to prospectively predict toxicity risk prior to initiating treatment. Understanding both clinical and pharmacogenetic risk factors for treatment-related hemorrhage is essential to mitigate risks and reduce the burden of this prevalent, and potentially fatal, complication. Supplementary Material |
Dietary Knowledge, Attitude, Practice Survey and Nutritional Knowledge-Based Intervention: A Cross-Sectional and Randomized Controlled Trial Study among College Undergraduates in China | 4778240c-9c02-47a0-ad5f-9e2258ddcfd4 | 11279395 | Health Literacy[mh] | A healthy diet habit is one of the essentials to a person’s physical and mental health, along with aging . The World Health Organization (WHO) believes that a healthy diet for adults should be rich in fruits, vegetables, beans, nuts and whole grains . Many previous studies have found that a healthy diet, such as the Mediterranean diet, can help to prevent obesity , regulate gut microbiota and reduce the incidence of cancer , chronic kidney diseases and many other diseases. Conversely, unhealthy diets and eating habits may increase the risk of obesity, cardiovascular disease, lung cancer and metabolic syndrome . In addition to physical health, many studies confirmed the correlation between healthy diet and mental health in adolescents and children . In the past decades, China’s economy has developed substantially. Along with these changes, the Chinese diet is also gradually changing from a plant-based to a meat-based diet . At the same time, the mortality and morbidity of many chronic diseases, such as cardiovascular diseases and diabetes, have dramatically increased and have become more prevalent in the young- and middle-aged groups . As an essential component in young people’s daily life, a healthy diet represents, and even likely leads, the formation of a healthy lifestyle in young people . However, contemporary young people, especially undergraduates, are facing challenges brought about by new lifestyles, which makes them more likely to develop bad dietary habits . Previous studies have shown that the dietary situation of global adolescents is problematic . Many studies have also suggested that the dietary status of college undergraduates is problematic . Therefore, more attention should be paid to undergraduates’ dietary status. The KAP model, which focuses on health literacy integrating knowledge, attitude and practice, points out that knowledge is the foundation for the establishment of positive and correct attitudes, and attitudes are the driving forces for behavior change . In a former study, health literacy was found to have a significant correlation with people’s health behaviors and the incidences of many diseases . Dietary literacy, as a special health literacy, is defined as the capacity to obtain, process and understand nutrition information and the materials needed to make appropriate decisions regarding one’s health , which is strongly related to one’s eating behaviors . College students tend to behave better in the ability of appraising and applying health information compared with those with less education . However, the dietary literacy status of undergraduates has been shown to remain unsatisfactory by many previous studies, which makes current college students a population of interest . And in order to solve this problem, a lot of studies have studied and found out several measures, among which nutrition education, in particular, was discovered as an effective way to improve college students’ dietary literacy . Located in the southeast of China, Zhejiang province has a high level of modernization, with numerous colleges and a large undergraduate population. In the previous study, Zhejiang was shown to have a high incidence rate of overall nutritional deficiency , indicating the unsatisfactory dietary status of college undergraduates. However, few researches were carried out to investigate the Chinese undergraduates’ dietary literacy status, especially in Zhejiang, from the perspective of KAP. Therefore, this study aimed to investigate the current status and the influencing factors of Chinese undergraduates’ dietary literacy, especially in Zhejiang, through a cross-sectional study and explore whether a nutritional lecture could improve their dietary literacy via a randomized controlled trial (RCT). We hypothesized that undergraduates’ dietary literacy was poor and influenced by multiple factors, which might be changed by the intervention of a targeted nutrition knowledge lecture. We hoped this study could have some implication for a positive and long-term significance for future health promotion policies.
2.1. Study Design This study was designed in two paths (as seen in ) as a cross-sectional study and a randomized controlled trial (RCT), which were reviewed and approved by the Medical Ethics Committee of Zhejiang Chinese Medical University, 20221011-3, 11 October 2022 and are registered with the clinical trials registry ( http://clinicaltrials.gov , ID: NCT05791500, accessed on 30 March 2023). 2.2. A Cross-Sectional Study Conducted by Questionnaire on Dietary Literacy in College Students Before the formal investigation, twenty-two college students were recruited from a few universities in China, including Zhejiang Chinese Medical University (ZCMU), Zhejiang University, Wenzhou Medical University, Zhejiang University of Technology and Ningbo University, to fill out a preliminary “Undergraduate dietary literacy KAP questionnaire” as an initial survey. According to the results, only 17.4% of college students had a relatively healthy diet (divided by 80% of the dietary literacy score). The required sample size was calculated to be 959, based on a two-sided z -test with a significance level of 5% and an allowable error of 0.024. Based on the global school-based student health survey (GSHS) questionnaire , an “Undergraduate dietary literacy KAP questionnaire” for data collection was designed, which contained a total of 35 questions and was divided into 4 parts: basic demographic characteristic questions of the respondents like sex, college year and major; dietary knowledge questions; dietary attitude questions and dietary practice questions. The detailed questionnaire can be seen in . Then, college undergraduates from multiple universities in Zhejiang were enrolled and asked to fill out the “Undergraduate dietary literacy KAP questionnaire” via the platform “Questionnaire Star” of WeChat https://www.wjx.cn/vm/YDgDLEJ.aspx# (accessed on 10 May 2022). After invalid questionnaire respondents (all options were filled with the same answer) were excluded, remaining valid questionnaires were scored to obtain dietary knowledge, attitude and practice scores according to the criteria in . To have a clear understanding of the current dietary KAP status among undergraduates, all the scores were standardized by the hundred-mark system and participants were ranked as “excellent”, “good”, “average” and “poor” based on 90 points, 80 points and 60 points. 2.3. An RCT Study Performed by a Nutritional Lecture In the RCT study, the sample size was calculated to be 90. The calculation was conducted with a probability of 50% and 17.4%, respectively, for college students in the nutritional lecture group and the control group owning healthy dietary KAP, based on type I error type ( α ) of 0.05 and type II error ( β ) of 0.1. Participants were recruited through a poster on the campus at ZCMU. Those undergraduates who were unable to complete the whole study (students in the 5th year in college or those who were going to graduate; students who did not live in dormitories arranged by the school; students who were suffering from illness) were excluded. All the participants were adult undergraduates and they signed the written consent. After that, participants were randomized to the nutritional lecture group and the control group via a random number generated by Excel 2019. Odd-numbered enrollees were assigned to the nutritional lecture group and even-numbered ones were assigned to the control group. Participants in the nutritional lecture group were requested to attend one single nutritional knowledge-based lecture, which was provided by a faculty from the Division of the Nutrition and Food Hygiene in the School of Public Health at ZCMU. The specific topics of the lecture mainly included knowledge on components of common dietary foods in daily life, how to choose healthy foods and the relationship between unhealthy diet and multiple chronic metabolic disease. Baseline data (day 0) and data after intervention (day 3 and day 100) were collected in two forms: “Undergraduate dietary literacy KAP questionnaire” and 72 h food records through food images. The scores of dietary knowledge and attitude were calculated from questionnaires in the same way as indicated above. The dietary behaviors of participants were evaluated more scientifically and accurately using two methods. One was assessed by dietary practice scores via the “Undergraduate dietary literacy KAP questionnaire”. The other was to collect food images of the 72 h dietary records from the participants, which were estimated by the Dietary Quality Index-International (DQI-I) method to obtain a dietary quality score . The detailed scoring rules for each item are shown in . 2.4. Statisitcal Analysis Categorical variables are presented as a sample percentage , and continuous variables are presented as means with standard deviation (SD) or medians with interquartile range (IQR) for the variables with normal or non-normal distribution, respectively. The normality of variable distribution was verified with the Shapiro–Wilk test before the statistical analysis. Then, t -tests or Wilcoxon tests were used to analyze the differences between the two groups. ANOVA analysis and Kruskal–Wallis tests were used to analyze the differences among multiple groups. Spearman rank correlation analysis and multivariate linear regression were applied to identify the relationship among dietary knowledge, attitude, practice and factors that affect dietary literacy. All the statistical methods in this study were two-tailed tests with a significance of 0.05 using SPSS 25.0. and GraphPad Prism 9.
This study was designed in two paths (as seen in ) as a cross-sectional study and a randomized controlled trial (RCT), which were reviewed and approved by the Medical Ethics Committee of Zhejiang Chinese Medical University, 20221011-3, 11 October 2022 and are registered with the clinical trials registry ( http://clinicaltrials.gov , ID: NCT05791500, accessed on 30 March 2023).
Before the formal investigation, twenty-two college students were recruited from a few universities in China, including Zhejiang Chinese Medical University (ZCMU), Zhejiang University, Wenzhou Medical University, Zhejiang University of Technology and Ningbo University, to fill out a preliminary “Undergraduate dietary literacy KAP questionnaire” as an initial survey. According to the results, only 17.4% of college students had a relatively healthy diet (divided by 80% of the dietary literacy score). The required sample size was calculated to be 959, based on a two-sided z -test with a significance level of 5% and an allowable error of 0.024. Based on the global school-based student health survey (GSHS) questionnaire , an “Undergraduate dietary literacy KAP questionnaire” for data collection was designed, which contained a total of 35 questions and was divided into 4 parts: basic demographic characteristic questions of the respondents like sex, college year and major; dietary knowledge questions; dietary attitude questions and dietary practice questions. The detailed questionnaire can be seen in . Then, college undergraduates from multiple universities in Zhejiang were enrolled and asked to fill out the “Undergraduate dietary literacy KAP questionnaire” via the platform “Questionnaire Star” of WeChat https://www.wjx.cn/vm/YDgDLEJ.aspx# (accessed on 10 May 2022). After invalid questionnaire respondents (all options were filled with the same answer) were excluded, remaining valid questionnaires were scored to obtain dietary knowledge, attitude and practice scores according to the criteria in . To have a clear understanding of the current dietary KAP status among undergraduates, all the scores were standardized by the hundred-mark system and participants were ranked as “excellent”, “good”, “average” and “poor” based on 90 points, 80 points and 60 points.
In the RCT study, the sample size was calculated to be 90. The calculation was conducted with a probability of 50% and 17.4%, respectively, for college students in the nutritional lecture group and the control group owning healthy dietary KAP, based on type I error type ( α ) of 0.05 and type II error ( β ) of 0.1. Participants were recruited through a poster on the campus at ZCMU. Those undergraduates who were unable to complete the whole study (students in the 5th year in college or those who were going to graduate; students who did not live in dormitories arranged by the school; students who were suffering from illness) were excluded. All the participants were adult undergraduates and they signed the written consent. After that, participants were randomized to the nutritional lecture group and the control group via a random number generated by Excel 2019. Odd-numbered enrollees were assigned to the nutritional lecture group and even-numbered ones were assigned to the control group. Participants in the nutritional lecture group were requested to attend one single nutritional knowledge-based lecture, which was provided by a faculty from the Division of the Nutrition and Food Hygiene in the School of Public Health at ZCMU. The specific topics of the lecture mainly included knowledge on components of common dietary foods in daily life, how to choose healthy foods and the relationship between unhealthy diet and multiple chronic metabolic disease. Baseline data (day 0) and data after intervention (day 3 and day 100) were collected in two forms: “Undergraduate dietary literacy KAP questionnaire” and 72 h food records through food images. The scores of dietary knowledge and attitude were calculated from questionnaires in the same way as indicated above. The dietary behaviors of participants were evaluated more scientifically and accurately using two methods. One was assessed by dietary practice scores via the “Undergraduate dietary literacy KAP questionnaire”. The other was to collect food images of the 72 h dietary records from the participants, which were estimated by the Dietary Quality Index-International (DQI-I) method to obtain a dietary quality score . The detailed scoring rules for each item are shown in .
Categorical variables are presented as a sample percentage , and continuous variables are presented as means with standard deviation (SD) or medians with interquartile range (IQR) for the variables with normal or non-normal distribution, respectively. The normality of variable distribution was verified with the Shapiro–Wilk test before the statistical analysis. Then, t -tests or Wilcoxon tests were used to analyze the differences between the two groups. ANOVA analysis and Kruskal–Wallis tests were used to analyze the differences among multiple groups. Spearman rank correlation analysis and multivariate linear regression were applied to identify the relationship among dietary knowledge, attitude, practice and factors that affect dietary literacy. All the statistical methods in this study were two-tailed tests with a significance of 0.05 using SPSS 25.0. and GraphPad Prism 9.
3.1. Basic Characteristics of Questionnaire Respondents After excluding three invalid questionnaires, a total of 1026 questionnaires were deemed valid. Among the responses, 944 were from ZCMU and 82 were from other universities nationwide. The numbers of male and female respondents were 271 and 755. The numbers of college undergraduates in their first, second, third, fourth and fifth year were 227, 511, 100, 144 and 44, respectively. The numbers of undergraduates majoring in the literature and arts, science and engineering, and medicine-related majors were 44, 78 and 904, respectively. The detailed results are shown in . 3.2. Current Dietary KAP Status of Questionnaire Respondents The dietary knowledge, attitude and practice of questionnaire participants were respectively 100.0 (33.3), 59.1 (13.6) and 71.7 (11.7). The rates of undergraduates with “excellent” dietary knowledge, attitude and practice were, respectively, 36.6%, 1.9% and 3.4%. The detailed results are shown in . 3.3. Dietary Literacy Discrepancy among Undergraduates of Different Sex, College Years and Majors We grouped and analyzed the differences in dietary knowledge, attitude and practice scores based on sex, college years and majors. First, the dietary knowledge scores of female and male participants were, respectively, 83.3 (33.3) and 83.3 (41.6). Female undergraduates showed better dietary knowledge scores than male counterparts ( p < 0.001). Second, dietary knowledge scores among the college undergraduates in the first, second, third, fourth and fifth year were 83.3 (41.7), 83.3 (33.3), 83.3 (33.3), 83.3 (25.0) and 95.8 (16.7), respectively. It was found that senior students behaved better than junior students in dietary knowledge ( p < 0.001). Additionally, students majoring in medicine had a higher level of dietary knowledge than students of non-medicine-related majors ( p < 0.001). Dietary knowledge scores among the majors of literature and art, science and engineering, and medicine-related majors were 66.7 (25.0), 83.3 (35.4) and 83.3 (33.3). As for different medicine-related majors, traditional Chinese medicine (TCM) students had a better grasp of dietary knowledge than preventive medicine students ( p < 0.001) and medical technology and information engineering (MTIE) students ( p < 0.05). Similarly, students majoring in nursing achieved higher dietary knowledge scores than preventive medicine and MTIE students ( p < 0.05). The detailed results are seen in . However, there were no significant differences in dietary attitude and practice scores among the students of different sex, college years and majors. The specific results are shown in . 3.4. The Relationship between Dietary Knowledge, Attitude and Practice As shown in the scatter plot in , there was a positive correlation between dietary attitude and dietary practice. Spearman rank correlation analysis also indicated a significant correlation between dietary practice and dietary attitude (r = 0.447, p < 0.001), while there was no significant correlation between dietary practice and dietary knowledge or between dietary knowledge and dietary attitude. Multiple linear regression analysis, including factors of college year, sex, major, dietary knowledge and attitude, showed that dietary attitudes had a significant impact on dietary practice ( β = 0.999, t = 15.092, p < 0.001), while dietary knowledge, different college years, majors and sex did not. The detailed results are shown in . 3.5. Basic Characteristics of the Participants in the RCT Study Ninety-nine undergraduate students from ZCMU were finally recruited in this RCT study and were randomized into the nutritional lecture group of fifty students or the group control of forty-nine students. Characteristics, including participants’ sex, years in colleges and major, were collected, and there was no significant difference between the two groups in terms of their basic characteristics, which are shown in . 3.6. The Impact of the Nutritional Lecture Shown in the Questionnaire As shown in , before intervention, there was no significant difference between the two groups in the baseline data of dietary knowledge, attitude and practice scores. After intervention, participants in the nutritional lecture group illustrated better dietary knowledge scores than those in the control group on day 3 ( p = 0.002) and day 100 ( p = 0.023); additionally, a significantly higher level of dietary attitude was shown in the intervention group on day 100 ( p = 0.044), but on day 3 the difference was not significant ( p = 0.058). And there was also no significant difference in dietary practice scores on day 3 ( p = 0.367) and day 100 ( p = 0.052). The detailed results are shown in . 3.7. Improvement Shown in the Food Images ANOVA analysis conducted on the dietary quality scores in the three visits revealed that nutritional lecture group participants showed a significant improvement on day 3 compared with the baseline (difference of 3.0, 95% CI of 0.3 to 5.7, p = 0.029). However, there was a significant decrease on day 100 compared with day 3 (difference of −4.0, 95% CI of −2.0 to −6.0, p < 0.001) and the improvement on day 3 seemed to have disappeared. The results are shown in . However, there was no statistical difference in the dietary quality scores between the two groups on day 0, day 3 and day 100. The detailed results are shown in .
After excluding three invalid questionnaires, a total of 1026 questionnaires were deemed valid. Among the responses, 944 were from ZCMU and 82 were from other universities nationwide. The numbers of male and female respondents were 271 and 755. The numbers of college undergraduates in their first, second, third, fourth and fifth year were 227, 511, 100, 144 and 44, respectively. The numbers of undergraduates majoring in the literature and arts, science and engineering, and medicine-related majors were 44, 78 and 904, respectively. The detailed results are shown in .
The dietary knowledge, attitude and practice of questionnaire participants were respectively 100.0 (33.3), 59.1 (13.6) and 71.7 (11.7). The rates of undergraduates with “excellent” dietary knowledge, attitude and practice were, respectively, 36.6%, 1.9% and 3.4%. The detailed results are shown in .
We grouped and analyzed the differences in dietary knowledge, attitude and practice scores based on sex, college years and majors. First, the dietary knowledge scores of female and male participants were, respectively, 83.3 (33.3) and 83.3 (41.6). Female undergraduates showed better dietary knowledge scores than male counterparts ( p < 0.001). Second, dietary knowledge scores among the college undergraduates in the first, second, third, fourth and fifth year were 83.3 (41.7), 83.3 (33.3), 83.3 (33.3), 83.3 (25.0) and 95.8 (16.7), respectively. It was found that senior students behaved better than junior students in dietary knowledge ( p < 0.001). Additionally, students majoring in medicine had a higher level of dietary knowledge than students of non-medicine-related majors ( p < 0.001). Dietary knowledge scores among the majors of literature and art, science and engineering, and medicine-related majors were 66.7 (25.0), 83.3 (35.4) and 83.3 (33.3). As for different medicine-related majors, traditional Chinese medicine (TCM) students had a better grasp of dietary knowledge than preventive medicine students ( p < 0.001) and medical technology and information engineering (MTIE) students ( p < 0.05). Similarly, students majoring in nursing achieved higher dietary knowledge scores than preventive medicine and MTIE students ( p < 0.05). The detailed results are seen in . However, there were no significant differences in dietary attitude and practice scores among the students of different sex, college years and majors. The specific results are shown in .
As shown in the scatter plot in , there was a positive correlation between dietary attitude and dietary practice. Spearman rank correlation analysis also indicated a significant correlation between dietary practice and dietary attitude (r = 0.447, p < 0.001), while there was no significant correlation between dietary practice and dietary knowledge or between dietary knowledge and dietary attitude. Multiple linear regression analysis, including factors of college year, sex, major, dietary knowledge and attitude, showed that dietary attitudes had a significant impact on dietary practice ( β = 0.999, t = 15.092, p < 0.001), while dietary knowledge, different college years, majors and sex did not. The detailed results are shown in .
Ninety-nine undergraduate students from ZCMU were finally recruited in this RCT study and were randomized into the nutritional lecture group of fifty students or the group control of forty-nine students. Characteristics, including participants’ sex, years in colleges and major, were collected, and there was no significant difference between the two groups in terms of their basic characteristics, which are shown in .
As shown in , before intervention, there was no significant difference between the two groups in the baseline data of dietary knowledge, attitude and practice scores. After intervention, participants in the nutritional lecture group illustrated better dietary knowledge scores than those in the control group on day 3 ( p = 0.002) and day 100 ( p = 0.023); additionally, a significantly higher level of dietary attitude was shown in the intervention group on day 100 ( p = 0.044), but on day 3 the difference was not significant ( p = 0.058). And there was also no significant difference in dietary practice scores on day 3 ( p = 0.367) and day 100 ( p = 0.052). The detailed results are shown in .
ANOVA analysis conducted on the dietary quality scores in the three visits revealed that nutritional lecture group participants showed a significant improvement on day 3 compared with the baseline (difference of 3.0, 95% CI of 0.3 to 5.7, p = 0.029). However, there was a significant decrease on day 100 compared with day 3 (difference of −4.0, 95% CI of −2.0 to −6.0, p < 0.001) and the improvement on day 3 seemed to have disappeared. The results are shown in . However, there was no statistical difference in the dietary quality scores between the two groups on day 0, day 3 and day 100. The detailed results are shown in .
There were several findings in our present study. First, in the cross-sectional study, we could see the problematic situation of dietary literacy in college students from the low percents of undergraduates with “excellent” and “good” dietary attitude and practice scores. Additionally, significant differences were discovered in dietary knowledge among the undergraduates in terms of different sex, college years and majors. Moreover, we revealed the determining role of dietary attitude on undergraduates’ dietary practice. Furthermore, in the RCT trial, dietary knowledge, attitude and dietary behaviors of the participants were significantly improved in a short period of time after a nutritional lecture. However, their dietary behaviors returned to the original state over several months after intervention. The college undergraduates’ dietary literacy situation seemed to be unsatisfactory. Many college students were found to hold a neutral attitude towards their eating health and attached little importance to the nutritional value of foods. Similar results have been reported in a previous study among adolescents in Chongqing, China . Additionally, many unhealthy eating behaviors such as skipping breakfasts were found among undergraduates, which is consistent with previous studies . In this study, undergraduates seemed to behave well in terms of dietary knowledge. But it did not mean Chinese college students had a good command of nutritional knowledge, because the basic and limited number of dietary knowledge questions might have been too easy for questionnaire respondents, especially students of medicine-related majors who were exposed to courses related to nutrition. The discrepancies in dietary knowledge among the undergraduates of different sex, college years and majors are of great interest to be explored. As to our finding that the female students showed better dietary knowledge than their male counterparts, it might be related to the fact that Chinese female undergraduates performed better than male undergraduates in learning psychology . Additionally, in order to keep fit, women seemed to pay more attention to contents related to healthy diets . Moreover, regarding the difference in dietary knowledge between the senior and junior students, one reason could be that nutrition education was deemed an important part for college students’ studying, especially for medical students , and that those junior undergraduates had not yet enrolled in the courses related to health promotion and nutrition while the senior undergraduates had already finished those courses. In former studies , these curricula were discovered to effectively improve students’ knowledge. Another reason could be that the senior students might be better educated and more knowledgeable compared with the junior ones after consecutive years of studying, thus leading to their higher dietary knowledge scores in the survey. Similar to the difference between upper-grade and lower-grade students, a couple of studies conducted in China reported the same difference in dietary knowledge between adults with a higher education level such as the level of doctor and those with a lower education level like the level of senior high school . They pointed out that education level might contribute to adults’ stronger interests in dietary knowledge and better understanding of nutritional requirements. The students of medicine-related majors illustrated a higher dietary knowledge level, consistent with a former study that explained that medical students might be more conscious of healthy eating, and they might also find it simpler to search for, find and comprehend information related to wellbeing . It might also be related to the fact that medical students were required to take courses related to nutrition in their programs , and they were more likely to be exposed to an environment that was filled with knowledge on healthy diets even in their out-of-classroom activities than those of non-medicine-related majors, which might eventually lead to the difference of dietary knowledge between them. One of the interesting results we saw was that the students majoring in TCM showed better dietary knowledge scores than the others, including preventive medicine, which was a surprise to us. To explore the causes, we found that TCM majors compared with other majors were very competitive, and had higher college admission scores in the college entrance examination than the others. They might be better educated and have a better learning ability, making them more capable in appraising and grasping health information compared with those with less education . In addition, the curricula of TCM are very intense, which may promote the students to study harder and more diligently. Despite the difference in dietary knowledge, in our study, there was no significant difference in dietary attitude scores among the undergraduate students regarding their sex, college year or major. This might suggest that the students who are female, senior or major in medicine-related programs, although having a deeper understanding and acquisition of dietary knowledge and being able to recognize healthy diets in their daily lives, might not realize the importance and necessity of healthy diets. They might think that, at their current and young age, they do not need to pay much attention to healthy eating, which ultimately leads to an insignificant difference in dietary attitude scores . Consequently, there was also no significant difference in dietary practice scores among the college students of different sex, college years and majors. Due to the lesser importance attached to healthy eating, they seem to prefer some delicious “junk food” over healthy foods that might have poor taste and flavor, which ultimately leads to insignificant differences in their dietary practice scores . The results of multiple linear regression shown in also explain the phenomenon of no differences in dietary attitude and practice between undergraduates with better and poorer dietary knowledge. It was shown that the students’ attitudes towards healthy diets are the determining factor that affects dietary behavior, while dietary knowledge is just the basic or secondary factor for the occurrence of dietary practice, which is consistent with the KAP health education model . We deeply recognize the importance of knowledge. However, we also realize that there may be a “gap” between knowledge or belief and behavior or action. Therefore, an intervention via a short-term targeted nutritional knowledge-based lecture was designed and aimed to illustrate how dietary knowledge improvement interacts with undergraduates’ dietary attitude and practice. From the results, we could see that one nutritional lecture might be an effective way to promote the occurrence of healthy dietary behaviors among undergraduate students in a short period of time by improving the participants’ knowledge, which seemed to be consistent with many previous studies . However, the effect was not long-lasting, which implies that, in order to maintain improved dietary literacy, more targeted and reinforced education is needed. Therefore, we explored several factors that might be required by effective nutrition education. Firstly, studies with longer durations seemed to achieve their started objectives better . One short nutritional lecture was probably not enough to improve undergraduates’ dietary attitudes and help them to establish a healthy dietary behavior pattern. Their dietary behaviors were improved in a short time, but not long after, they would return to their original status. Additionally, there was a lack of interaction between the investigators and participants, which might have led to the decrease in the effectiveness of a nutritional lecture . It was also possible that, due to the large time span between day 3 and day 100, there were many confounding factors; for example, it was the time period for students in ZCMU to return to school for exams on day 100. They had just passed a winter vacation and had not yet adapted to the differences in diet between home and school, with a decrease in attention to healthy eating under the pressure of final exams. And stress might result in irregular and unhealthy eating behaviors . For example, they might be more inclined to choose “junk food” with better taste and flavor or might overeat or even not eat. In the RCT study, the significant differences of dietary knowledge scores indicated the fact that the nutritional lecture truly improved the dietary knowledge of the participants in the nutritional lecture group. However, there was only a marginally significant difference in dietary attitude scores on day 3 between the two groups and, on day 100, there was a significant difference. This strange phenomenon could be explained by the Hawthorne effect . After just joining this study, participants might have thought that they were being observed by others and thus developed a tendency to change their behavior even without any interventions like a nutritional lecture in this study, ultimately resulting in insignificant differences between the nutritional lecture group and the control group. But, after a few months, participants in the control group returned to their original state, while those in the nutritional lecture group still maintained a better attitude towards healthy diets due to the influence of nutritional lectures, thereby leading to a difference between the two groups. Additionally, the p -value of the dietary attitudes between the two groups on day 3 was 0.056. Although it did not reach a significant level, it was very close. If the sample size of the study population is increased or the frequency of the nutritional lecture is increased in subsequent studies, it will be possible to obtain a positive result. Similarly, regarding the dietary practice scores shown in the questionnaires, there was also a marginally significant difference between the two groups on day 100. This might also have been due to the Hawthorne effect and the limited sample size of participants in this study. In this study, we combined the KAP health education model to investigate the dietary status of Chinese undergraduates, especially in Zhejiang province, from multiple perspectives of dietary knowledge, dietary attitude and dietary practice, and studied the relationship among the three factors along with many other underlying factors that might influence dietary literacy, which seemed not very clear in previous studies. Moreover, in addition to a cross-sectional survey, preliminary targeted nutritional knowledge-based interventions were carried out on participants recruited from ZCMU in this study to explore whether a nutritional knowledge-based lecture could improve dietary literacy. Additionally, in the RCT study, various methods, including questionnaires and dietary records based on food images, were used to estimate participants’ dietary behaviors more scientifically and accurately. After that, the intervention impact was evaluated, providing reference significance for more long-term and effective intervention measures in the future. However, there were still some limitations in our study. First, in the cross-sectional study, we adopted the method of convenience sampling when collecting the questionnaires. The “Undergraduate dietary literacy KAP questionnaire” was mainly distributed in universities in Zhejiang, with ZCMU as the main focus, and the proportion of women in both the questionnaire and the intervention participants was higher than men. These factors might have caused selection bias. Furthermore, the data after intervention were collected only twice, which could not illustrate clearly or well the trend of changes in undergraduates’ dietary literacy in both the intervention and the control groups. Lastly, the intervention measure adopted in this study was only one nutritional lecture. The longer term and far-reaching impact of nutrition education might not have been revealed.
In summary, this study revealed the unsatisfactory current dietary literacy status among Chinese undergraduate students, especially in Zhejiang, and demonstrated some discrepancies in dietary knowledge in college students in terms of sex, college year and major, discovering the determining role of dietary attitude on dietary practice. One targeted nutritional knowledge-based lecture was only effective short term regarding the improvement of their dietary literacy. In order to achieve substantial and long-lasting improvements in the healthy dietary behavior in college students in China, additional efforts are still needed, such as offering lasting rather than short-term nutrition courses among college students, strengthening the interaction between students and investigators during the process of education and setting nutrition labels in college canteens to help students establish a clearer understanding of what they eat every day.
|
Psychophysiological Response According to the Greenness Index of Subway Station Space | 2b0c0ea2-7dd4-4a65-8a24-05f0cbd89897 | 8272074 | Physiology[mh] | 1.1. Research Background and Purpose Subways are being used for public transportation to address space shortage attributed to rapid economic growth and urbanization. In South Korea, subway line 1 was first opened in Seoul in 1974, and Busan subway line 1 was opened in 1985. Subsequently, subway line 4 (Busan-Gimhae Light Rail Transit) and Donghae line were opened in Busan. The construction of a metropolitan rapid transit network is currently in progress, with the aim of building seven additional lines in the future. In addition, the number of Busan metropolitan railway daily transit passengers has increased from 98,000 in 1985 to 753,000 in 2010 and 938,000 in 2019. As subway spaces are expanded and the number of subway users increase, the environmental designs of subway station spaces have garnered significant research attention. Subway station users travelling via new sections and existing lines have negatively responded to the existing subway station, which is relatively old and unclean, and have advocated the creation of a cleaner and healthier space . Furthermore, the Ministry of Environment has prepared guidelines for developing ecological landscape of underground spaces according to Article 43 (2) of the Natural Environmental Conservation Act in 2009 and prescribed requirements to effectively create a green belt in underground spaces such as subways, which are regarded as important living spaces in the city . This shows that there is a need to improve existing closed and complex subway stations and propose strategies to modify subway stations for creating pleasant environments. Accordingly, the biophilic approach—the concept of bringing back nature in the city—has emerged as a natural and holistic approach in urban environments through design and governance in recent years . It is the most well-known greening and design approach among the public . The term “biophilic approach” was invented by E.O. Wilson, who was working as a professor at Harvard University, in 1984. Existing studies on the biophilic approach reported that exposing people to natural environments had positive effects on reducing their stress level, recovering their mental fatigue, and increasing their physical functions and abilities . In this regard, green spaces ensured in indoor environments can be advantageous for enhancing the physical and psychological health of people. The introduction of plants to underground spaces has been passively conducted so far, despite its advantages in enhancing aesthetics while serving decorative, constructive, and psychological and emotional functions. However, the Ministry of Environment has recently encouraged the use of green spaces in underground areas by laying out the general guidelines for designing urban railway platforms of subway stations with regard to four categories: functionality, convenience, environmental friendliness, and aesthetics. In the United States, the WELL Building Standard was developed in 2012, which emphasizes accessibility to nature for the health of people in buildings. Modern people who spend a considerable amount of time inside everyday are interested in creating a pleasant and healthy environment, and their desire to introduce natural elements into interior spaces, as primitive elements for psychological stability and relaxation, is increasing . Furthermore, subway users exhibit various use patterns apart from transportation, including meeting and resting; therefore, they are recognizing the necessity of environmental improvement through greening. Because subway station spaces applying interior landscape design resembled interior landscape spaces, a decrease in the concentration of dust and formaldehyde and an increase in humidity by approximately 3 to 4% were observed in these spaces . It was also found that an increase in the greening area on indoor walls had significant effects on reducing indoor temperature, PM1, and PM10 . Moreover, psychological restoration effects in subway station spaces applying interior landscape design were greater—by twice or more—than those in subway stations spaces not applying interior landscape design, on average . Thus, by developing a pleasant interior landscape in subway stations that play a key role in city transport operations, various positive effects are expected including the creation of spaces and the improvement in interior environment quality, energy saving, and emotional and psychological stability of users. The Gare de Lyon in Paris and the Lowline Project in New York City are case studies of interior landscaping in underground areas in which various interior landscape designs have been implemented, such as the “remote skylights” system delivering natural sunlight to trees and grass underneath city streets and a water supply system using fog as a water source. In South Korea, transportation authorities are attempting to improve the interior spaces of subway stations by applying new environmental designs, such as the installation of an LED vegetable garden, wall displays, and the construction of ceramic wall paintings . Recently, various efforts are being made to design an interior landscape in some subway stations at Seoul using new technologies, including a remote skylight system to control natural sunlight that has been employed by the Lowline Project in New York, smart farms, and plant-based biofilters. As such, new designs and technologies are being introduced to create a pleasant subway station spaces in Seoul as well as the world-centered cities. Physical environments in subway station spaces vary according to the characteristics of these spaces, such as the presence of natural light inflow, the area of a subway station, floor height, and degree of congestion . For this reason, subway station spaces should be classified according to their physical environments. Accordingly, an appropriate interior landscape plan considering these types of subway station spaces is required. Accordingly, the following research questions are presented: How can interior landscapes be designed to enhance the environmental landscape of subway stations as well as user satisfaction? How much interior space is required to achieve effective spatial, economic, and aesthetic satisfaction? How will subway users react to the interior landscape of subway stations? People can perceive stimuli coming from interior landscapes, such as a visual stimulus (e.g., a green view), an olfactory stimulus (e.g., the scent of flowers), and an auditory stimulus (e.g., the sound of leaves shaking), based on their five senses. Among these senses, the visual sense is regarded as the most crucial sense; visual perception, related to 2D or 3D shapes, colors, and texture, accounts for 80% of environmental perception . The greenness index (GI) reflects the amount of green space observed by an individual as compared to the entire landscape; this index reflects the amount, level of recognition, and satisfaction associated with green space . The GI of interior spaces has a significant effect on the visual preference and satisfaction of space users. According to previous studies on the GI, it is closely related to psychological stability and pleasantness . However, existing studies have established a range of green belts as a part of the city landscape, while focusing on the psychological impact, exterior GI preferences, and promotions strategies associated with the GI. Conversely, the preferred GI and interior landscape design were proposed, based on the response to the GI; subsequently, commercial spaces and workspaces were targeted . Therefore, to improve the environment of subway stations, users’ responses to the interior GI and preferences are required as reference data. Furthermore, by establishing interior landscape models of subway stations, creating 3D videos, and objectively measuring the response to the GI, determining the GI for 3D environments in a 2D aspect was attempted. Therefore, a 3D landscape model of the subway interior was established in this study. Further, the users’ psychological and physiological responses to the interior landscape environment and GI of a subway station were analyzed with the objective of proposing pleasant and healthy interior landscape strategies for subway stations. The findings of this study are expected to assist in planning and improving the interior landscape of subway stations globally. Furthermore, the findings of this study can be useful in practically analyzing the amount and recognition level of green belts. This is accomplished by providing human perspectives of interior spaces and utilizing the evaluation method of the GI via a 3D model, unlike the existing 2D measurements. 1.2. Research Scope and Methods This study proposed a plan for implementing a desirable interior landscape in subway station spaces to enhance environmental conditions and increase user satisfaction. To this end, we developed 3D landscape models of the subway interior and investigated the psychophysiological responses of research subjects based on the models . Specifically, we reviewed previous studies and classified subway stations as ground subway stations where natural light can be introduced and underground subway stations where natural light cannot be introduced based on the review result. Moreover, existing methods for applying interior landscape in subway station spaces were analyzed based on garden style, planter style, and wall style. Thus, arrangement was performed based on a combination of garden and wall styles, which did not disturb the traffic of subway users. The GI index was classified as 0%, 10%, 15%, and 20%. Through this process, eight 3D landscape models of the subway interior were implemented. Furthermore, we analyzed psychophysiological responses of 60 male and female adults in their 20 s and 30 s, who were the main subway users, based on the 3D landscape models implemented. To analyze physiological responses of subjects, it conducted an experiment on measuring their brainwaves according to the level of the GI and 3D landscape models. To analyze psychological responses of subjects, it carried out a survey on their preference for a 3D landscape model, their perception of interior landscape area, their evaluation of subway station space images, and attention restoration effect.
Subways are being used for public transportation to address space shortage attributed to rapid economic growth and urbanization. In South Korea, subway line 1 was first opened in Seoul in 1974, and Busan subway line 1 was opened in 1985. Subsequently, subway line 4 (Busan-Gimhae Light Rail Transit) and Donghae line were opened in Busan. The construction of a metropolitan rapid transit network is currently in progress, with the aim of building seven additional lines in the future. In addition, the number of Busan metropolitan railway daily transit passengers has increased from 98,000 in 1985 to 753,000 in 2010 and 938,000 in 2019. As subway spaces are expanded and the number of subway users increase, the environmental designs of subway station spaces have garnered significant research attention. Subway station users travelling via new sections and existing lines have negatively responded to the existing subway station, which is relatively old and unclean, and have advocated the creation of a cleaner and healthier space . Furthermore, the Ministry of Environment has prepared guidelines for developing ecological landscape of underground spaces according to Article 43 (2) of the Natural Environmental Conservation Act in 2009 and prescribed requirements to effectively create a green belt in underground spaces such as subways, which are regarded as important living spaces in the city . This shows that there is a need to improve existing closed and complex subway stations and propose strategies to modify subway stations for creating pleasant environments. Accordingly, the biophilic approach—the concept of bringing back nature in the city—has emerged as a natural and holistic approach in urban environments through design and governance in recent years . It is the most well-known greening and design approach among the public . The term “biophilic approach” was invented by E.O. Wilson, who was working as a professor at Harvard University, in 1984. Existing studies on the biophilic approach reported that exposing people to natural environments had positive effects on reducing their stress level, recovering their mental fatigue, and increasing their physical functions and abilities . In this regard, green spaces ensured in indoor environments can be advantageous for enhancing the physical and psychological health of people. The introduction of plants to underground spaces has been passively conducted so far, despite its advantages in enhancing aesthetics while serving decorative, constructive, and psychological and emotional functions. However, the Ministry of Environment has recently encouraged the use of green spaces in underground areas by laying out the general guidelines for designing urban railway platforms of subway stations with regard to four categories: functionality, convenience, environmental friendliness, and aesthetics. In the United States, the WELL Building Standard was developed in 2012, which emphasizes accessibility to nature for the health of people in buildings. Modern people who spend a considerable amount of time inside everyday are interested in creating a pleasant and healthy environment, and their desire to introduce natural elements into interior spaces, as primitive elements for psychological stability and relaxation, is increasing . Furthermore, subway users exhibit various use patterns apart from transportation, including meeting and resting; therefore, they are recognizing the necessity of environmental improvement through greening. Because subway station spaces applying interior landscape design resembled interior landscape spaces, a decrease in the concentration of dust and formaldehyde and an increase in humidity by approximately 3 to 4% were observed in these spaces . It was also found that an increase in the greening area on indoor walls had significant effects on reducing indoor temperature, PM1, and PM10 . Moreover, psychological restoration effects in subway station spaces applying interior landscape design were greater—by twice or more—than those in subway stations spaces not applying interior landscape design, on average . Thus, by developing a pleasant interior landscape in subway stations that play a key role in city transport operations, various positive effects are expected including the creation of spaces and the improvement in interior environment quality, energy saving, and emotional and psychological stability of users. The Gare de Lyon in Paris and the Lowline Project in New York City are case studies of interior landscaping in underground areas in which various interior landscape designs have been implemented, such as the “remote skylights” system delivering natural sunlight to trees and grass underneath city streets and a water supply system using fog as a water source. In South Korea, transportation authorities are attempting to improve the interior spaces of subway stations by applying new environmental designs, such as the installation of an LED vegetable garden, wall displays, and the construction of ceramic wall paintings . Recently, various efforts are being made to design an interior landscape in some subway stations at Seoul using new technologies, including a remote skylight system to control natural sunlight that has been employed by the Lowline Project in New York, smart farms, and plant-based biofilters. As such, new designs and technologies are being introduced to create a pleasant subway station spaces in Seoul as well as the world-centered cities. Physical environments in subway station spaces vary according to the characteristics of these spaces, such as the presence of natural light inflow, the area of a subway station, floor height, and degree of congestion . For this reason, subway station spaces should be classified according to their physical environments. Accordingly, an appropriate interior landscape plan considering these types of subway station spaces is required. Accordingly, the following research questions are presented: How can interior landscapes be designed to enhance the environmental landscape of subway stations as well as user satisfaction? How much interior space is required to achieve effective spatial, economic, and aesthetic satisfaction? How will subway users react to the interior landscape of subway stations? People can perceive stimuli coming from interior landscapes, such as a visual stimulus (e.g., a green view), an olfactory stimulus (e.g., the scent of flowers), and an auditory stimulus (e.g., the sound of leaves shaking), based on their five senses. Among these senses, the visual sense is regarded as the most crucial sense; visual perception, related to 2D or 3D shapes, colors, and texture, accounts for 80% of environmental perception . The greenness index (GI) reflects the amount of green space observed by an individual as compared to the entire landscape; this index reflects the amount, level of recognition, and satisfaction associated with green space . The GI of interior spaces has a significant effect on the visual preference and satisfaction of space users. According to previous studies on the GI, it is closely related to psychological stability and pleasantness . However, existing studies have established a range of green belts as a part of the city landscape, while focusing on the psychological impact, exterior GI preferences, and promotions strategies associated with the GI. Conversely, the preferred GI and interior landscape design were proposed, based on the response to the GI; subsequently, commercial spaces and workspaces were targeted . Therefore, to improve the environment of subway stations, users’ responses to the interior GI and preferences are required as reference data. Furthermore, by establishing interior landscape models of subway stations, creating 3D videos, and objectively measuring the response to the GI, determining the GI for 3D environments in a 2D aspect was attempted. Therefore, a 3D landscape model of the subway interior was established in this study. Further, the users’ psychological and physiological responses to the interior landscape environment and GI of a subway station were analyzed with the objective of proposing pleasant and healthy interior landscape strategies for subway stations. The findings of this study are expected to assist in planning and improving the interior landscape of subway stations globally. Furthermore, the findings of this study can be useful in practically analyzing the amount and recognition level of green belts. This is accomplished by providing human perspectives of interior spaces and utilizing the evaluation method of the GI via a 3D model, unlike the existing 2D measurements.
This study proposed a plan for implementing a desirable interior landscape in subway station spaces to enhance environmental conditions and increase user satisfaction. To this end, we developed 3D landscape models of the subway interior and investigated the psychophysiological responses of research subjects based on the models . Specifically, we reviewed previous studies and classified subway stations as ground subway stations where natural light can be introduced and underground subway stations where natural light cannot be introduced based on the review result. Moreover, existing methods for applying interior landscape in subway station spaces were analyzed based on garden style, planter style, and wall style. Thus, arrangement was performed based on a combination of garden and wall styles, which did not disturb the traffic of subway users. The GI index was classified as 0%, 10%, 15%, and 20%. Through this process, eight 3D landscape models of the subway interior were implemented. Furthermore, we analyzed psychophysiological responses of 60 male and female adults in their 20 s and 30 s, who were the main subway users, based on the 3D landscape models implemented. To analyze physiological responses of subjects, it conducted an experiment on measuring their brainwaves according to the level of the GI and 3D landscape models. To analyze psychological responses of subjects, it carried out a survey on their preference for a 3D landscape model, their perception of interior landscape area, their evaluation of subway station space images, and attention restoration effect.
2.1. Modularization and Establishment of Interior Landscape Model of Subway Station Space Based on the results of a previous study that investigated the interior landscape status of subway stations , it was identified that spaces where the interior landscape is installed are all waiting rooms regardless of space type. Furthermore, based on the physical environment of interior landscape space, it was found that there is a significant difference in the area and floor height of the waiting rooms. In terms of the floor height of the waiting rooms, there is a difference between the ground and underground subway station, and the floor height for the ground subway station is designed to be high to stimulate the inflow of natural light. Therefore, the representative space of the subway interior landscape model was selected to be the waiting room. The interior landscape models were established after separating the ground subway station, where natural lighting is enabled, and underground subway station, where natural lighting is not introduced, to evaluate the GI of different physical environments . To test and analyze the psychophysiological responses of research subjects with the presence of natural light inflow and a different GI, the area and floor height of the waiting rooms were maintained at 4958.94 m 2 and 3 m, respectively. The sunlight was set at 6500 K, which is the peak daylight hour; illumination intensity of subway station space was 500 lx ; and the walls, floor, and ceiling were given an achromatic color to easily recognize the presence of an interior landscape in the space. To simulate a subway waiting room for evaluating the GI, Auto Cad 2020, SketchUp 2019, and Twin motion 2020 programs were used, and a waiting room area of 4958.94 m 2 (room height of 3 m and volume of 14,876.81 m 3 ) was used, considering the area mean and median values of the Busan subway waiting room. To simulate the user walking from the subway station entrance to the square, passage, and ticket gate, a 30 s (30 s) video (considering the brainwave testing time) was used. Initially, the entire waiting room was the interior landscape space; however, because the video used in the experiment was limited to 30 s, the interior landscape of the waiting room space used was 1982.11 m 2 (5946.33 m 3 ) (excluding staircases and workspace). This granted users a realistic experience. During the experiments, subjects were informed that the interior landscape was applied to the entire subway waiting room, including four models that altered the interior landscape. The video was within 20° from the horizontal line at a point 1.5 m above ground based on the GI measurement to ensure subjects have a clear view of the interior landscape. The resolution of the video was 1920 × 1080 (2K Full HD). To view the GI, four representative spaces, namely, the entrance, square, passage, and ticket gate on the way from the waiting room to the platform, were selected as sample positions. The angle was adjusted to observe the interior landscape arrangement (to avoid obstruction by pillars), based on a previous study that investigated the interior GI. For the interior landscape design, the garden style, identified as the representative interior landscaping technique for the subway station in a previous study , was selected as the standard interior landscape model. The experiments were conducted in a mixed style, placing the pillars and walls at the center (planter style) and modularizing the interior landscape design to minimize pedestrian interference when using the subway station space. A previous study claimed that a two-column plantation is more effective than a one-column plantation when trees are planted in the garden-style; therefore, the two-column plantation was utilized. According to the results of a previous study , the landscape formed of a flat plane is preferred in the planter style (pillar/wall); the plantation was either used or excluded depending on the GI using a 2 m wall width and one pillar as standard. Sculpture and topography libraries in the Twinmotion program were used for the plants used in the interior landscape models. Five different types of plants with soft leaf shapes were selected according to findings from previous studies among green foliage plants widely utilized in case studies, which are identified to improve the psychological and physiological health of humans compared with plants of other colors . The volume of plants applied to the interior landscape models of underground subway station space was 236.66 m 3 (3.98% of the subway station area (volume)) for a GI of 10%, 648 m 3 (10.90%) for a GI of 15%, and 721.45 m 3 (12.13%) for a GI of 20%: for ground subway stations, 235.13 m 3 (3.95%) for a GI of 10%, 646.22 m 3 (10.87%) for a GI of 15%, and 721.40 m 3 (12.13%) for a GI of 20% . When organizing the interior landscape models for underground/ground subway station space, physical environment differences were observed for the windows and passages towards the platform (stair wall), and a permissible difference in the standard GI by each interior landscape design was accommodated. 2.2. Composition of Research Tools 2.2.1. Brainwave Measurement Tool An electroencephalogram (EEG) is a test for clinically detecting brain diseases, which has been used to diagnose epilepsy, sleep disturbance, comas, and encephalopathy in medical fields and perform research on neuroscience, cognition, physiology, and psychology in academic fields . Accordingly, various studies have been conducted to quantify inner responses to external stimulus as biosignals, including those on verifying effects of forest and gardens on reducing stress , identifying a difference in color emotions , and examining physiological effects of the fragrance of plants . In terms of the frequency range of brainwaves, α wave is in the range of 8–13 Hz and denotes a comfortable state such as an emotional break or rest . A high β wave (H.β) is in the range of 20–30 Hz and denotes awareness due to stress, such as tension and anxiety, and it occurs continuously . The prefrontal cortex reflects asymmetric characteristics because of the functional differentiation of the left brain and right brain . In this study, an analysis was conducted, emphasizing the absolute power values of α wave and high β wave to identify the asymmetric characteristics of the prefrontal cortex for the interior landscape model with different GI. The left and right asymmetric analyses of brainwaves were conducted by calculating the difference obtained after subtracting the left prefrontal (Fp1) frequency value from the right prefrontal (Fp2) frequency value of the absolute power value. As the α wave is inversely proportional to the level of brain activation , if the prefrontal α wave asymmetric value is positive , it implies that the left prefrontal cortex has been activated, which is related to a positive emotional state. However, if the prefrontal α wave asymmetric value is negative (−), it means the right prefrontal cortex has been activated, which is related to a negative emotional state . The effect of interior landscape installation and GI on prefrontal brainwave activities while watching the interior landscape model video for 8 min was investigated. The Neuro Harmony system, manufactured by Panaxto sand NeuroSpec 2.4 programs, was used to measure the EEG of the prefrontal cortex. The Neuro Harmony System showed a correlation of 0.916 ( p < 0.001) for left and right α, β, and θ waves with Grass System (USA), a well-known brainwave measuring instrument; thus, its reliability has been verified . The corresponding equipment was a two-channel brainwave measuring instrument and three dry electrodes plated with pure gold to measure the real-time brainwave activity of the left and right brains. This method has simple measurement due to its hairband form, and it is agreeable to the research participants. Fp1 (left) and Fp2 (right) of the prefrontal cortex were set as active electrodes, Fpz (center) was set as the ground electrode, and the earlobes were set as reference electrodes. In terms of the frequency range of brainwaves, α wave is in the range of 8–13 Hz and measures a comfortable state such as an emotional break or rest , and high β wave (H.β) is in the range of 20–30 Hz and measures awareness due to stress, such as tension and anxiety, and it continuously occurs . The occipital lobe is the area that primarily acquires visual information from the cerebrum, and the prefrontal lobe is the cortex, which analyzes color information obtained and is involved in a decision-making process. For this reason, this study used EEG values of the prefrontal lobe for analysis . 2.2.2. Questionnaire Development The preference for interior landscape models according to the GI of subway stations was investigated by preparing a questionnaire on the interior landscape model preference, attention restoration effect, and space image. In that process, the questions were reconstructed based on previous studies on the GI and interior landscaping, as well as guidelines for designing urban railway stations, amenities, and transfer facilities . The questionnaire comprised 118 questions (GI of 0%: 14 questions each and GI of 10%, 15%, and 20%: 15 questions each) on the preference for interior landscape models. For the interior landscape model preference eight questions (one question each), level of recognition of the interior landscape area six questions (GI of 10%, 15% and 20%: one question each), attention restoration effect 56 questions (seven questions each), and space image questions 48 questions (six questions each), and a 5-point Likert scale was employed. Subjects were asked to form questionnaires after watching videos on 3D landscape models of the subway interior according to the GI level. To analyze their preference for 3D landscape models of the subway interior, they watched videos on each 3D landscape model and expressed their preference for the corresponding interior environment. In terms of their perception of interior landscape area, they provided responses on the area of interior landscape installation that they perceived for each 3D landscape model. With regard to space images, the SD measurement method was utilized to analyze their perception of space images of each 3D landscape model. The answers for interior landscape model preferences were: “not preferred at all (1 point)”, “not preferred (2 points)”, “moderate (3 points)”, “preferred (4 points)”, and “strongly preferred (5 points).” The answers for the level of recognition of the interior landscape area were: “too low (1 point)”, “low (2 points)”, “suitable (3 points)”, “high (4 points)”, and “very high (5 points).” The answers for interior landscape model recognition level of the interior landscape area were: “strongly disagree (1 point)”, “disagree (2 points)”, “normal (3 points)”, “agree (4 points)”, “strongly agree (5 points)”. The attention restoration effect is subjectively measured based on the level of attention capacity, which is estimated to be restored when a person belongs to or looks at a certain environment . Previous studies measured the attention restoration effect by showing slides on the natural or urban landscape to research subjects and asking them about the possibility of attention restoration . The attention restoration effect can be measured based on recovery scales. The recovery scales consisted of seven items (coming to rest, renew energy, become myself again, lose all tension, order my thoughts again, put everything behind me, regain the ability to concentrate) . The original recovery scales were based on a 7-point scale ranging from “not at all (1 point)” to “the very much, for the affective states; very well, for the behavioral items (7 points).” However, in this study, a 5-point Likert scale was utilized to facilitate a more convenient comparison of preferences of research subjects for 3D landscape models of the subway interior, their perception of the area, and their space image evaluation. The answers for the attention restoration effect were, “not at all (1 point)”, “no (2 points)”, “moderate (3 points)”, “yes (4 points)”, and “absolutely (5 points).” The attention restoration points were calculated by summing up the points of seven attention restoration questions and calculating the average. Regarding questions on the space images, the SD (semantic differential) scale was used for the analysis. Contents of design guidelines for urban railway stations transfer facilities convenience facilities and previous studies were referred to. Six pairs of words describing the image, namely “dark–bright”, “uncomfortable–comfortable”, “not beautiful–beautiful”, “unpleasant–pleasant”, “unharmonious–harmonious”, “un-environmentally friendly–environmentally friendly” were employed. According to the SD measurement method, opposite adjectives were marked at both ends: “strongly”, scored “1 point and 5 points, “disagree” scored 2 points and 4 points, and “normal” scored 3 points. Before the correspondence analysis, preference, level of interior landscape area recognition, attention restoration effect, and space image measured as continuous variables (interval ratio scale) were recoded to discrete variables (nominal scale). As such, “not preferred at all (1 point)” and “not preferred (2 points)” were recoded to “not preferred GI (1)”, “moderate (3 points)” was recoded to “moderate GI (2)” and “preferred (4 points)” and “strongly preferred (5 points)” were recoded to “preferred GI (3).” For the level of interior landscape area recognition, “too low (1 point)” and “low (2 points)” were recoded to “small landscape area (1)”, “suitable (3 points)” was recoded to “suitable landscape area (2)”, and “high (4 points)” and “very high (5 points)” were recoded to “large landscape area (3).” For attention restoration effect, based on the total score (35 points), “7–20 points” was recoded to “low attention restoration effect (1)”, “21–27 points” was recoded to “moderate attention restoration effect (2)”, and “28–35 points” was recoded to “high attention restoration effect (3).” For space image, “1–2 points” were recoded to negative space image (e.g., “dark” for “bright–dark”), “moderate (3 points)” was recoded to “moderate (2)”, and “4–5 points” was recoded to positive space image (e.g., “bright” for “bright–dark”). To identify the relevance of interior landscape models providing a positive effect to research subjects among these, only the items with positive content among psychological response (“preferred GI (3)”, “suitable landscape area (2)”, “high attention restoration effect (3)”, “bright (3)”, “comfortable (3)”, “beautiful (3)”, “pleasant (3)”, “harmonious (3)”, and “environmentally friendly (3)”) were selected for analysis. 2.3. Experimental Method The experiments were conducted after approval by the bioethics committee of Pusan National University in South Korea. According to the 2015 South Korea census, people in their 20 s (30%) and 30 s (25%) mainly used the subway and transit systems for work and school commutes. Thus, these age groups were selected as the study subjects, with 60 people participating in the study (30 men and 30 women) . The experiments were conducted for approximately six days—from 6 November to 13 November 2020. In terms of statistical calculation, the sample mean distribution approximates normal distribution as the sample size of population increases, following central limit theorem; thus, if the sample size exceeds 30, it approaches a normal distribution pattern regardless of the population distribution. Therefore, in this study, the number of samples of one group was set to 30 to ensure reliability. Prior to the EEG experiment, research subjects were informed of this research, research participation methods, and the formation of the agreement. Approximately 10 min were required to complete these procedures. The brainwave measurement was conducted on the same group as the psychological response investigation. Participants’ brainwaves were measured eight times while observing the interior landscape models: once with closed eyes and seven times during the break time. Furthermore, each participant’s brainwaves were measured for 480 s, with 30 s allocated for a single measurement. Then, a survey on the eight 3D landscape models implemented was conducted for approximately 10–15 min. In total, 18–23 min was required, including the amount of time required for conducting the brainwave measurement experiment and survey. The study subjects were selected based on those who consented to the experiments and did not have brain diseases or mental illness, ophthalmologic diseases or a visual impairment, heart-related diseases, did not take medicine for underlying conditions such as diabetes, and had normal blood pressure. Prior to experiments, the research objective and methodology, experimental contents, and required time were explained to the selected research subjects. Alcohol consumption was prohibited for two days before the experiment, and sufficient sleep was required the day before the experiment. Caffeinated drinks and smoking, which could affect the autonomic nervous system and sensitivity, were prohibited for 2.5 h before the experiment to control the effect of these actions on experimental results . On the day of the experiment, color blindness and color weakness were identified using the “Korean color test table”, and the experimental procedure was explained along with instructions on opening and closing eyes. Moreover, mobile phones and jewelry were restricted to prevent the interruption of electromagnetic waves. Brainwaves were measured after excluding conditions that could cause artifact by awakening the subjects’ attention, such as yawning or blinking, to minimize the movement in a comfortable position . An experimental booth with a width of 900 mm, length of 900 mm, and height of 1335 mm was installed . The monitor was fixed at eye level, 1 m from the participant. A 27-inch monitor was used, and the resolution was set at 1920 × 1080. With regards to the psychological effects on test subjects, the external light was blocked using a blackout curtain, and fluorescent lights installed on the ceiling were turned on . The effect of noise from the chair where the test subject sat was minimized by setting the criterion as below 40 dB, and an interior temperature of 22 ± 3 °C and humidity of 50 ± 5% RH was fixed. The same laboratory facilities and environmental conditions were provided to all the participants. The EEG for interior landscape models with different GI was measured by establishing a stable state baseline with participants closing their eyes for 30 s. Then, the EEG was measured when stimulating interior landscape models with different GI for 30 s each. A break (30 s for each) was given after watching each of the eight interior landscape models to prevent a lack of concentration and tedium. The video with a GI of 0% was first presented to compare the effect of the presence of an interior landscape, and then the models with a GI of 10%, 15%, and 20% were randomly presented . Approximately 10–15 min was required to complete the survey. The brainwave signal was first converted to the power value through fast Fourier transform (FFT), which interprets complex waves into simple waves before converting it to a text file (txt). One-way analysis of variance (ANOVA) was performed using SPSS Statistics 25, and the significant values of each GI were deduced based on significance probability. Frequency analysis, χ 2 test and T -test, one-way ANOVA, and MDS (Multidimensional Scaling) correspondence analysis utilizing SPSS Statistics 25 were employed on the data. When the one-way ANOVA is performed, a post-hoc test analysis is required according to the test result for homogeneity of variances. In this study, Duncan was used to perform a post-hoc test analysis when the assumption of homogeneity of variance was satisfied based on the p -value of the F value being 0.05 or higher in the test result for homogeneity of variances. On the other hand, the Games–Howell test was used to perform a post-hoc test analysis when the assumption of homogeneity of variance was not satisfied based on the p -value of the F value being 0.05 or below in the test result for homogeneity of variances. When the homogeneity of variance is assumed, Tukey’s HSD test, the Schaffe method, and the Duncan multiple range test are more frequently preferred for the multiple comparison procedures . Among these methods, the Duncan multiple range test compares the ends of the mean of k group(s) adjacent to each other in stages. It shows a stronger ability for separating groups than Tukey’s HSD test and the Schaffe method. In addition, it controls the error rate related to test sets, although it performs comparison based on phased comparison order such as the Student–Newman–Keul method . The purpose of the ANOVA in-depth analysis is to see if there is a statistically significant difference between groups and which group on a given issue is significant. Therefore, a Duncan post-hoc test is used to evaluate the actual cluster level .
Based on the results of a previous study that investigated the interior landscape status of subway stations , it was identified that spaces where the interior landscape is installed are all waiting rooms regardless of space type. Furthermore, based on the physical environment of interior landscape space, it was found that there is a significant difference in the area and floor height of the waiting rooms. In terms of the floor height of the waiting rooms, there is a difference between the ground and underground subway station, and the floor height for the ground subway station is designed to be high to stimulate the inflow of natural light. Therefore, the representative space of the subway interior landscape model was selected to be the waiting room. The interior landscape models were established after separating the ground subway station, where natural lighting is enabled, and underground subway station, where natural lighting is not introduced, to evaluate the GI of different physical environments . To test and analyze the psychophysiological responses of research subjects with the presence of natural light inflow and a different GI, the area and floor height of the waiting rooms were maintained at 4958.94 m 2 and 3 m, respectively. The sunlight was set at 6500 K, which is the peak daylight hour; illumination intensity of subway station space was 500 lx ; and the walls, floor, and ceiling were given an achromatic color to easily recognize the presence of an interior landscape in the space. To simulate a subway waiting room for evaluating the GI, Auto Cad 2020, SketchUp 2019, and Twin motion 2020 programs were used, and a waiting room area of 4958.94 m 2 (room height of 3 m and volume of 14,876.81 m 3 ) was used, considering the area mean and median values of the Busan subway waiting room. To simulate the user walking from the subway station entrance to the square, passage, and ticket gate, a 30 s (30 s) video (considering the brainwave testing time) was used. Initially, the entire waiting room was the interior landscape space; however, because the video used in the experiment was limited to 30 s, the interior landscape of the waiting room space used was 1982.11 m 2 (5946.33 m 3 ) (excluding staircases and workspace). This granted users a realistic experience. During the experiments, subjects were informed that the interior landscape was applied to the entire subway waiting room, including four models that altered the interior landscape. The video was within 20° from the horizontal line at a point 1.5 m above ground based on the GI measurement to ensure subjects have a clear view of the interior landscape. The resolution of the video was 1920 × 1080 (2K Full HD). To view the GI, four representative spaces, namely, the entrance, square, passage, and ticket gate on the way from the waiting room to the platform, were selected as sample positions. The angle was adjusted to observe the interior landscape arrangement (to avoid obstruction by pillars), based on a previous study that investigated the interior GI. For the interior landscape design, the garden style, identified as the representative interior landscaping technique for the subway station in a previous study , was selected as the standard interior landscape model. The experiments were conducted in a mixed style, placing the pillars and walls at the center (planter style) and modularizing the interior landscape design to minimize pedestrian interference when using the subway station space. A previous study claimed that a two-column plantation is more effective than a one-column plantation when trees are planted in the garden-style; therefore, the two-column plantation was utilized. According to the results of a previous study , the landscape formed of a flat plane is preferred in the planter style (pillar/wall); the plantation was either used or excluded depending on the GI using a 2 m wall width and one pillar as standard. Sculpture and topography libraries in the Twinmotion program were used for the plants used in the interior landscape models. Five different types of plants with soft leaf shapes were selected according to findings from previous studies among green foliage plants widely utilized in case studies, which are identified to improve the psychological and physiological health of humans compared with plants of other colors . The volume of plants applied to the interior landscape models of underground subway station space was 236.66 m 3 (3.98% of the subway station area (volume)) for a GI of 10%, 648 m 3 (10.90%) for a GI of 15%, and 721.45 m 3 (12.13%) for a GI of 20%: for ground subway stations, 235.13 m 3 (3.95%) for a GI of 10%, 646.22 m 3 (10.87%) for a GI of 15%, and 721.40 m 3 (12.13%) for a GI of 20% . When organizing the interior landscape models for underground/ground subway station space, physical environment differences were observed for the windows and passages towards the platform (stair wall), and a permissible difference in the standard GI by each interior landscape design was accommodated.
2.2.1. Brainwave Measurement Tool An electroencephalogram (EEG) is a test for clinically detecting brain diseases, which has been used to diagnose epilepsy, sleep disturbance, comas, and encephalopathy in medical fields and perform research on neuroscience, cognition, physiology, and psychology in academic fields . Accordingly, various studies have been conducted to quantify inner responses to external stimulus as biosignals, including those on verifying effects of forest and gardens on reducing stress , identifying a difference in color emotions , and examining physiological effects of the fragrance of plants . In terms of the frequency range of brainwaves, α wave is in the range of 8–13 Hz and denotes a comfortable state such as an emotional break or rest . A high β wave (H.β) is in the range of 20–30 Hz and denotes awareness due to stress, such as tension and anxiety, and it occurs continuously . The prefrontal cortex reflects asymmetric characteristics because of the functional differentiation of the left brain and right brain . In this study, an analysis was conducted, emphasizing the absolute power values of α wave and high β wave to identify the asymmetric characteristics of the prefrontal cortex for the interior landscape model with different GI. The left and right asymmetric analyses of brainwaves were conducted by calculating the difference obtained after subtracting the left prefrontal (Fp1) frequency value from the right prefrontal (Fp2) frequency value of the absolute power value. As the α wave is inversely proportional to the level of brain activation , if the prefrontal α wave asymmetric value is positive , it implies that the left prefrontal cortex has been activated, which is related to a positive emotional state. However, if the prefrontal α wave asymmetric value is negative (−), it means the right prefrontal cortex has been activated, which is related to a negative emotional state . The effect of interior landscape installation and GI on prefrontal brainwave activities while watching the interior landscape model video for 8 min was investigated. The Neuro Harmony system, manufactured by Panaxto sand NeuroSpec 2.4 programs, was used to measure the EEG of the prefrontal cortex. The Neuro Harmony System showed a correlation of 0.916 ( p < 0.001) for left and right α, β, and θ waves with Grass System (USA), a well-known brainwave measuring instrument; thus, its reliability has been verified . The corresponding equipment was a two-channel brainwave measuring instrument and three dry electrodes plated with pure gold to measure the real-time brainwave activity of the left and right brains. This method has simple measurement due to its hairband form, and it is agreeable to the research participants. Fp1 (left) and Fp2 (right) of the prefrontal cortex were set as active electrodes, Fpz (center) was set as the ground electrode, and the earlobes were set as reference electrodes. In terms of the frequency range of brainwaves, α wave is in the range of 8–13 Hz and measures a comfortable state such as an emotional break or rest , and high β wave (H.β) is in the range of 20–30 Hz and measures awareness due to stress, such as tension and anxiety, and it continuously occurs . The occipital lobe is the area that primarily acquires visual information from the cerebrum, and the prefrontal lobe is the cortex, which analyzes color information obtained and is involved in a decision-making process. For this reason, this study used EEG values of the prefrontal lobe for analysis . 2.2.2. Questionnaire Development The preference for interior landscape models according to the GI of subway stations was investigated by preparing a questionnaire on the interior landscape model preference, attention restoration effect, and space image. In that process, the questions were reconstructed based on previous studies on the GI and interior landscaping, as well as guidelines for designing urban railway stations, amenities, and transfer facilities . The questionnaire comprised 118 questions (GI of 0%: 14 questions each and GI of 10%, 15%, and 20%: 15 questions each) on the preference for interior landscape models. For the interior landscape model preference eight questions (one question each), level of recognition of the interior landscape area six questions (GI of 10%, 15% and 20%: one question each), attention restoration effect 56 questions (seven questions each), and space image questions 48 questions (six questions each), and a 5-point Likert scale was employed. Subjects were asked to form questionnaires after watching videos on 3D landscape models of the subway interior according to the GI level. To analyze their preference for 3D landscape models of the subway interior, they watched videos on each 3D landscape model and expressed their preference for the corresponding interior environment. In terms of their perception of interior landscape area, they provided responses on the area of interior landscape installation that they perceived for each 3D landscape model. With regard to space images, the SD measurement method was utilized to analyze their perception of space images of each 3D landscape model. The answers for interior landscape model preferences were: “not preferred at all (1 point)”, “not preferred (2 points)”, “moderate (3 points)”, “preferred (4 points)”, and “strongly preferred (5 points).” The answers for the level of recognition of the interior landscape area were: “too low (1 point)”, “low (2 points)”, “suitable (3 points)”, “high (4 points)”, and “very high (5 points).” The answers for interior landscape model recognition level of the interior landscape area were: “strongly disagree (1 point)”, “disagree (2 points)”, “normal (3 points)”, “agree (4 points)”, “strongly agree (5 points)”. The attention restoration effect is subjectively measured based on the level of attention capacity, which is estimated to be restored when a person belongs to or looks at a certain environment . Previous studies measured the attention restoration effect by showing slides on the natural or urban landscape to research subjects and asking them about the possibility of attention restoration . The attention restoration effect can be measured based on recovery scales. The recovery scales consisted of seven items (coming to rest, renew energy, become myself again, lose all tension, order my thoughts again, put everything behind me, regain the ability to concentrate) . The original recovery scales were based on a 7-point scale ranging from “not at all (1 point)” to “the very much, for the affective states; very well, for the behavioral items (7 points).” However, in this study, a 5-point Likert scale was utilized to facilitate a more convenient comparison of preferences of research subjects for 3D landscape models of the subway interior, their perception of the area, and their space image evaluation. The answers for the attention restoration effect were, “not at all (1 point)”, “no (2 points)”, “moderate (3 points)”, “yes (4 points)”, and “absolutely (5 points).” The attention restoration points were calculated by summing up the points of seven attention restoration questions and calculating the average. Regarding questions on the space images, the SD (semantic differential) scale was used for the analysis. Contents of design guidelines for urban railway stations transfer facilities convenience facilities and previous studies were referred to. Six pairs of words describing the image, namely “dark–bright”, “uncomfortable–comfortable”, “not beautiful–beautiful”, “unpleasant–pleasant”, “unharmonious–harmonious”, “un-environmentally friendly–environmentally friendly” were employed. According to the SD measurement method, opposite adjectives were marked at both ends: “strongly”, scored “1 point and 5 points, “disagree” scored 2 points and 4 points, and “normal” scored 3 points. Before the correspondence analysis, preference, level of interior landscape area recognition, attention restoration effect, and space image measured as continuous variables (interval ratio scale) were recoded to discrete variables (nominal scale). As such, “not preferred at all (1 point)” and “not preferred (2 points)” were recoded to “not preferred GI (1)”, “moderate (3 points)” was recoded to “moderate GI (2)” and “preferred (4 points)” and “strongly preferred (5 points)” were recoded to “preferred GI (3).” For the level of interior landscape area recognition, “too low (1 point)” and “low (2 points)” were recoded to “small landscape area (1)”, “suitable (3 points)” was recoded to “suitable landscape area (2)”, and “high (4 points)” and “very high (5 points)” were recoded to “large landscape area (3).” For attention restoration effect, based on the total score (35 points), “7–20 points” was recoded to “low attention restoration effect (1)”, “21–27 points” was recoded to “moderate attention restoration effect (2)”, and “28–35 points” was recoded to “high attention restoration effect (3).” For space image, “1–2 points” were recoded to negative space image (e.g., “dark” for “bright–dark”), “moderate (3 points)” was recoded to “moderate (2)”, and “4–5 points” was recoded to positive space image (e.g., “bright” for “bright–dark”). To identify the relevance of interior landscape models providing a positive effect to research subjects among these, only the items with positive content among psychological response (“preferred GI (3)”, “suitable landscape area (2)”, “high attention restoration effect (3)”, “bright (3)”, “comfortable (3)”, “beautiful (3)”, “pleasant (3)”, “harmonious (3)”, and “environmentally friendly (3)”) were selected for analysis.
An electroencephalogram (EEG) is a test for clinically detecting brain diseases, which has been used to diagnose epilepsy, sleep disturbance, comas, and encephalopathy in medical fields and perform research on neuroscience, cognition, physiology, and psychology in academic fields . Accordingly, various studies have been conducted to quantify inner responses to external stimulus as biosignals, including those on verifying effects of forest and gardens on reducing stress , identifying a difference in color emotions , and examining physiological effects of the fragrance of plants . In terms of the frequency range of brainwaves, α wave is in the range of 8–13 Hz and denotes a comfortable state such as an emotional break or rest . A high β wave (H.β) is in the range of 20–30 Hz and denotes awareness due to stress, such as tension and anxiety, and it occurs continuously . The prefrontal cortex reflects asymmetric characteristics because of the functional differentiation of the left brain and right brain . In this study, an analysis was conducted, emphasizing the absolute power values of α wave and high β wave to identify the asymmetric characteristics of the prefrontal cortex for the interior landscape model with different GI. The left and right asymmetric analyses of brainwaves were conducted by calculating the difference obtained after subtracting the left prefrontal (Fp1) frequency value from the right prefrontal (Fp2) frequency value of the absolute power value. As the α wave is inversely proportional to the level of brain activation , if the prefrontal α wave asymmetric value is positive , it implies that the left prefrontal cortex has been activated, which is related to a positive emotional state. However, if the prefrontal α wave asymmetric value is negative (−), it means the right prefrontal cortex has been activated, which is related to a negative emotional state . The effect of interior landscape installation and GI on prefrontal brainwave activities while watching the interior landscape model video for 8 min was investigated. The Neuro Harmony system, manufactured by Panaxto sand NeuroSpec 2.4 programs, was used to measure the EEG of the prefrontal cortex. The Neuro Harmony System showed a correlation of 0.916 ( p < 0.001) for left and right α, β, and θ waves with Grass System (USA), a well-known brainwave measuring instrument; thus, its reliability has been verified . The corresponding equipment was a two-channel brainwave measuring instrument and three dry electrodes plated with pure gold to measure the real-time brainwave activity of the left and right brains. This method has simple measurement due to its hairband form, and it is agreeable to the research participants. Fp1 (left) and Fp2 (right) of the prefrontal cortex were set as active electrodes, Fpz (center) was set as the ground electrode, and the earlobes were set as reference electrodes. In terms of the frequency range of brainwaves, α wave is in the range of 8–13 Hz and measures a comfortable state such as an emotional break or rest , and high β wave (H.β) is in the range of 20–30 Hz and measures awareness due to stress, such as tension and anxiety, and it continuously occurs . The occipital lobe is the area that primarily acquires visual information from the cerebrum, and the prefrontal lobe is the cortex, which analyzes color information obtained and is involved in a decision-making process. For this reason, this study used EEG values of the prefrontal lobe for analysis .
The preference for interior landscape models according to the GI of subway stations was investigated by preparing a questionnaire on the interior landscape model preference, attention restoration effect, and space image. In that process, the questions were reconstructed based on previous studies on the GI and interior landscaping, as well as guidelines for designing urban railway stations, amenities, and transfer facilities . The questionnaire comprised 118 questions (GI of 0%: 14 questions each and GI of 10%, 15%, and 20%: 15 questions each) on the preference for interior landscape models. For the interior landscape model preference eight questions (one question each), level of recognition of the interior landscape area six questions (GI of 10%, 15% and 20%: one question each), attention restoration effect 56 questions (seven questions each), and space image questions 48 questions (six questions each), and a 5-point Likert scale was employed. Subjects were asked to form questionnaires after watching videos on 3D landscape models of the subway interior according to the GI level. To analyze their preference for 3D landscape models of the subway interior, they watched videos on each 3D landscape model and expressed their preference for the corresponding interior environment. In terms of their perception of interior landscape area, they provided responses on the area of interior landscape installation that they perceived for each 3D landscape model. With regard to space images, the SD measurement method was utilized to analyze their perception of space images of each 3D landscape model. The answers for interior landscape model preferences were: “not preferred at all (1 point)”, “not preferred (2 points)”, “moderate (3 points)”, “preferred (4 points)”, and “strongly preferred (5 points).” The answers for the level of recognition of the interior landscape area were: “too low (1 point)”, “low (2 points)”, “suitable (3 points)”, “high (4 points)”, and “very high (5 points).” The answers for interior landscape model recognition level of the interior landscape area were: “strongly disagree (1 point)”, “disagree (2 points)”, “normal (3 points)”, “agree (4 points)”, “strongly agree (5 points)”. The attention restoration effect is subjectively measured based on the level of attention capacity, which is estimated to be restored when a person belongs to or looks at a certain environment . Previous studies measured the attention restoration effect by showing slides on the natural or urban landscape to research subjects and asking them about the possibility of attention restoration . The attention restoration effect can be measured based on recovery scales. The recovery scales consisted of seven items (coming to rest, renew energy, become myself again, lose all tension, order my thoughts again, put everything behind me, regain the ability to concentrate) . The original recovery scales were based on a 7-point scale ranging from “not at all (1 point)” to “the very much, for the affective states; very well, for the behavioral items (7 points).” However, in this study, a 5-point Likert scale was utilized to facilitate a more convenient comparison of preferences of research subjects for 3D landscape models of the subway interior, their perception of the area, and their space image evaluation. The answers for the attention restoration effect were, “not at all (1 point)”, “no (2 points)”, “moderate (3 points)”, “yes (4 points)”, and “absolutely (5 points).” The attention restoration points were calculated by summing up the points of seven attention restoration questions and calculating the average. Regarding questions on the space images, the SD (semantic differential) scale was used for the analysis. Contents of design guidelines for urban railway stations transfer facilities convenience facilities and previous studies were referred to. Six pairs of words describing the image, namely “dark–bright”, “uncomfortable–comfortable”, “not beautiful–beautiful”, “unpleasant–pleasant”, “unharmonious–harmonious”, “un-environmentally friendly–environmentally friendly” were employed. According to the SD measurement method, opposite adjectives were marked at both ends: “strongly”, scored “1 point and 5 points, “disagree” scored 2 points and 4 points, and “normal” scored 3 points. Before the correspondence analysis, preference, level of interior landscape area recognition, attention restoration effect, and space image measured as continuous variables (interval ratio scale) were recoded to discrete variables (nominal scale). As such, “not preferred at all (1 point)” and “not preferred (2 points)” were recoded to “not preferred GI (1)”, “moderate (3 points)” was recoded to “moderate GI (2)” and “preferred (4 points)” and “strongly preferred (5 points)” were recoded to “preferred GI (3).” For the level of interior landscape area recognition, “too low (1 point)” and “low (2 points)” were recoded to “small landscape area (1)”, “suitable (3 points)” was recoded to “suitable landscape area (2)”, and “high (4 points)” and “very high (5 points)” were recoded to “large landscape area (3).” For attention restoration effect, based on the total score (35 points), “7–20 points” was recoded to “low attention restoration effect (1)”, “21–27 points” was recoded to “moderate attention restoration effect (2)”, and “28–35 points” was recoded to “high attention restoration effect (3).” For space image, “1–2 points” were recoded to negative space image (e.g., “dark” for “bright–dark”), “moderate (3 points)” was recoded to “moderate (2)”, and “4–5 points” was recoded to positive space image (e.g., “bright” for “bright–dark”). To identify the relevance of interior landscape models providing a positive effect to research subjects among these, only the items with positive content among psychological response (“preferred GI (3)”, “suitable landscape area (2)”, “high attention restoration effect (3)”, “bright (3)”, “comfortable (3)”, “beautiful (3)”, “pleasant (3)”, “harmonious (3)”, and “environmentally friendly (3)”) were selected for analysis.
The experiments were conducted after approval by the bioethics committee of Pusan National University in South Korea. According to the 2015 South Korea census, people in their 20 s (30%) and 30 s (25%) mainly used the subway and transit systems for work and school commutes. Thus, these age groups were selected as the study subjects, with 60 people participating in the study (30 men and 30 women) . The experiments were conducted for approximately six days—from 6 November to 13 November 2020. In terms of statistical calculation, the sample mean distribution approximates normal distribution as the sample size of population increases, following central limit theorem; thus, if the sample size exceeds 30, it approaches a normal distribution pattern regardless of the population distribution. Therefore, in this study, the number of samples of one group was set to 30 to ensure reliability. Prior to the EEG experiment, research subjects were informed of this research, research participation methods, and the formation of the agreement. Approximately 10 min were required to complete these procedures. The brainwave measurement was conducted on the same group as the psychological response investigation. Participants’ brainwaves were measured eight times while observing the interior landscape models: once with closed eyes and seven times during the break time. Furthermore, each participant’s brainwaves were measured for 480 s, with 30 s allocated for a single measurement. Then, a survey on the eight 3D landscape models implemented was conducted for approximately 10–15 min. In total, 18–23 min was required, including the amount of time required for conducting the brainwave measurement experiment and survey. The study subjects were selected based on those who consented to the experiments and did not have brain diseases or mental illness, ophthalmologic diseases or a visual impairment, heart-related diseases, did not take medicine for underlying conditions such as diabetes, and had normal blood pressure. Prior to experiments, the research objective and methodology, experimental contents, and required time were explained to the selected research subjects. Alcohol consumption was prohibited for two days before the experiment, and sufficient sleep was required the day before the experiment. Caffeinated drinks and smoking, which could affect the autonomic nervous system and sensitivity, were prohibited for 2.5 h before the experiment to control the effect of these actions on experimental results . On the day of the experiment, color blindness and color weakness were identified using the “Korean color test table”, and the experimental procedure was explained along with instructions on opening and closing eyes. Moreover, mobile phones and jewelry were restricted to prevent the interruption of electromagnetic waves. Brainwaves were measured after excluding conditions that could cause artifact by awakening the subjects’ attention, such as yawning or blinking, to minimize the movement in a comfortable position . An experimental booth with a width of 900 mm, length of 900 mm, and height of 1335 mm was installed . The monitor was fixed at eye level, 1 m from the participant. A 27-inch monitor was used, and the resolution was set at 1920 × 1080. With regards to the psychological effects on test subjects, the external light was blocked using a blackout curtain, and fluorescent lights installed on the ceiling were turned on . The effect of noise from the chair where the test subject sat was minimized by setting the criterion as below 40 dB, and an interior temperature of 22 ± 3 °C and humidity of 50 ± 5% RH was fixed. The same laboratory facilities and environmental conditions were provided to all the participants. The EEG for interior landscape models with different GI was measured by establishing a stable state baseline with participants closing their eyes for 30 s. Then, the EEG was measured when stimulating interior landscape models with different GI for 30 s each. A break (30 s for each) was given after watching each of the eight interior landscape models to prevent a lack of concentration and tedium. The video with a GI of 0% was first presented to compare the effect of the presence of an interior landscape, and then the models with a GI of 10%, 15%, and 20% were randomly presented . Approximately 10–15 min was required to complete the survey. The brainwave signal was first converted to the power value through fast Fourier transform (FFT), which interprets complex waves into simple waves before converting it to a text file (txt). One-way analysis of variance (ANOVA) was performed using SPSS Statistics 25, and the significant values of each GI were deduced based on significance probability. Frequency analysis, χ 2 test and T -test, one-way ANOVA, and MDS (Multidimensional Scaling) correspondence analysis utilizing SPSS Statistics 25 were employed on the data. When the one-way ANOVA is performed, a post-hoc test analysis is required according to the test result for homogeneity of variances. In this study, Duncan was used to perform a post-hoc test analysis when the assumption of homogeneity of variance was satisfied based on the p -value of the F value being 0.05 or higher in the test result for homogeneity of variances. On the other hand, the Games–Howell test was used to perform a post-hoc test analysis when the assumption of homogeneity of variance was not satisfied based on the p -value of the F value being 0.05 or below in the test result for homogeneity of variances. When the homogeneity of variance is assumed, Tukey’s HSD test, the Schaffe method, and the Duncan multiple range test are more frequently preferred for the multiple comparison procedures . Among these methods, the Duncan multiple range test compares the ends of the mean of k group(s) adjacent to each other in stages. It shows a stronger ability for separating groups than Tukey’s HSD test and the Schaffe method. In addition, it controls the error rate related to test sets, although it performs comparison based on phased comparison order such as the Student–Newman–Keul method . The purpose of the ANOVA in-depth analysis is to see if there is a statistically significant difference between groups and which group on a given issue is significant. Therefore, a Duncan post-hoc test is used to evaluate the actual cluster level .
3.1. Physiological Response to Different GI for the Interior Landscape of Subway Station Space 3.1.1. α Wave Asymmetry with Different GI for the Interior Landscape of Subway Station Space One-way ANOVA was conducted to investigate the α wave asymmetry of interior landscape models with a different GI of subway station space and closed eyes was showed 0.45 (SD = 0.98). There was a difference in the α wave asymmetry value depending on the presence of installed interior landscape, suggesting that the installation of interior landscapes could facilitate a psychologically stable state compared to space without an interior landscape. The test results for homogeneity of variances showed that the p -value of the F value was 0.05 or higher ( p -value of underground subway station = 0.65, p -value of ground subway station = 0.83) and that the assumption of homogeneity of variance was satisfied. For this reason, Duncan was utilized to conduct a post-hoc test analysis. Meanwhile, the underground space with a GI of 0% had the lowest α wave asymmetry value of 0.47 (SD = 0.90), while the ground GI of 0% had an α wave asymmetry value of 0.64 (SD = 0.8). These differing values indicate that for an environment with no interior landscape model installed, the α wave asymmetry value on the ground where natural light could be introduced was higher than that of the underground. These results are similar to previous studies , which claim that among subjects exposed to roses, the high-frequency component of heart rate variability was significantly higher than in controls. The results of brainwave change with different leaf shapes, sizes, and ear-types of foliage plants in this study were in line with the results of a previous study . An existing study reported no difference in physiological responses according to the GI (5%, 20%, 50%, and 80%) of indoor spaces. Similar to the aforementioned study, this study found no significant difference in models according to the GI of 10%, 15%, and 20%. Therefore, it was concluded that the presence of plants rather than the quantitative increase in plants could positively affect mental relaxation and psychological stability. 3.1.2. High β Wave Asymmetry with Different GI for the Interior Landscape of Subway Station Space One-way ANOVA conducted to investigate the high β wave asymmetry with a different GI of subway station space was 0.07 (SD = 0.24) with closed eyes; the underground GI of 0%, 10%, 15%, and 20% were 0.08, 0.12, 0.11, and 0.09, respectively; and the ground GI of 0%, 10%, 15%, and 20% were 0.03, 0.09, 0.09, and 0.10, respectively, indicating the installation of an interior landscape did not significantly impact high β waves. Such findings are due to the intensity of illumination of artificial lighting being set at 500 lx, which is higher than the mean intensity of illumination of existing subway station space (355.89 lx) , to accord with the criteria of plant vegetation environment while building interior landscape models of subway stations. Therefore, the space with a brighter image than the actual subway station space was established, and stress was not expected to be induced, even with the interior landscape model with a GI of 0%. 3.2. Psychological Response to Different GI for the Interior Landscape of Subway Station Space 3.2.1. Preference on Subway Interior Landscape Model with Different GI One-way ANOVA was performed to investigate the preference of the subway interior landscape model with a different GI . The underground GI of 15% and ground GI of 10% was highest at 4.48 points (SD = 0.70) and 4.50 points (SD = 0.62), respectively, followed by the underground GI of 10% with 4.05 points (SD = 0.67), the ground GI of 15% with 4.08 points (SD = 0.67), the underground GI of 20% with 3.67 points (SD = 0.73), and the ground GI of 20% with 3.57 points (SD = 0.79). The test result for homogeneity of variances showed that the p -value of the F value was 0.05 or below ( p -value of underground subway station = 0.03, p -value of ground subway station = 0.031) and that the assumption of homogeneity of variance was not satisfied. For this reason, Games–Howell test was used to conduct a post-hoc test analysis . Meanwhile, the preferences of GI of 0% in the underground and ground were 2.40 points (SD = 0.89) and 2.92 points (SD = 0.94), respectively, indicating the preference for a subway station space with an interior landscape rather than for that with no interior landscape. The preference points of GI of 20% in the underground and ground were 3.67 points (SD = 0.73) and 3.57 points (SD = 0.79), respectively, suggesting that the implicit improvement of a GI is not preferred. These results are similar to those of previous studies , which claim that harmonious natural environmental elements including plant leaves become with other elements of whole space considering arrangement and shape has a more significant impact on the preference when introducing plants. Compared with the preferred level of the GI, users believe the interior landscape area with a GI of 15% is suitable; however, for ground subway stations, the model with a GI of 10% is preferred because of the effect and security of the inflow of natural light. Therefore, as ground subway stations are advantageous for securing natural light due to high floor heights, future studies on the required interior GI are needed. A cross-analysis was performed to investigate the recognition level of the interior landscape area of the subway interior landscape model with a different GI . As a result, the underground GI of 15% and the ground GI of 10% received the highest response of “the interior landscape area is adequate” with 70.0% and 66.7%, respectively. Meanwhile, 53.3% and 55.0% of the respondents indicated that “the interior landscape area is adequate” for the underground GI of 10% and the ground GI of 15%, respectively. Of the respondents, 35.5% answered that “the interior landscape area is small” for the underground GI of 10%, while 43.3% of the respondents responded that “the interior landscape area is large” for the ground GI of 15%; this indicates that the recognition level of the interior landscape area differs depending on the availability of natural sunlight. On the other hand, 76.6% and 86.7% of the respondents considered the interior landscape area as “large” and “very large”, respectively, in the case of GI of 20% in the underground and the ground, suggesting that 20% of GI in subway station spaces is considered excessive in terms of the plant density. 3.2.2. Attention Restoration Effect of Subway Interior Landscape Models with Different GI The one-way ANOVA scores of attention restoration effects for an underground and ground GI of 0% were 2.39 points (SD = 0.57) and 2.65 points (SD = 0.70), respectively . Such results indicate that the attention restoration effect of subway station space with interior landscape is higher than that with no interior landscape. The result of the test for homogeneity of variances showed that the p -value of the F value was 0.05 or below ( p -value of underground subway station = 0.09, p -value of ground subway station = 0.38) and that the assumption of homogeneity of variance was not satisfied. For this reason, the Games–Howell test was used to conduct a post-hoc test analysis . The underground GI of 15% showed the highest attention restoration effect of 4.42 points (SD = 0.72), whereas ground GI of 10%, 15%, and 20% were 4.22 (SD = 0.68), 4.39 (SD = 0.78), and 4.24 points (SD = 0.71), respectively. These results suggest that for the ground GI, only a 10% GI could lead to an attention restoration effect of 15–20% GI. 3.2.3. Space Image of Subway Interior Landscape Model with Different GI The one-way ANOVA was conducted to evaluate space images of 3D landscape models of the subway interior according to the GI ( , and ). The test results for homogeneity of variances according to space image item showed that the p -value of the F value was 0.05 or below and that the assumption of homogeneity of variance was not satisfied. For this reason, the Games–Howell test was used to conduct a post-hoc test analysis ( and ). As for the “dark–bright” image in underground subway stations, the GI of 15% was 4.75 points, being the highest. However, a GI of 10% and 15% were the most comfortable space images, with 4.20 (SD = 0.99) and 4.52 points (SD = 0.93), respectively, for “uncomfortable–comfortable.” The GI of 15% was 4.45 points (SD = 0.87) for “not beautiful–beautiful”, suggesting that it was the most beautiful space image. The GI of 10% and 15% were 4.42 points (SD = 0.87) and 4.67 points (SD = 0.66), respectively, for “unpleasant–pleasant”, indicating that they were the most pleasant space images. The GI of 10% and 15% resulted in 4.03 (SD = 1.29) and 4.42 points (SD = 0.98), respectively, for “unharmonious–harmonious”, indicating they were the most harmonious space images. The GI of 15% was found to be the most environmentally-friendly space image for “un-environmentally friendly–environmentally friendly.” Regarding ground subway stations, the GI of 10% and 15% were the brightest space images with 4.68 (SD = 0.65) and 4.50 points (SD = 0.87), respectively, for “dark–bright.” The GI of 10% and 15% were the most positive images for “uncomfortable–comfortable” (4.43 points (SD = 0.87) and 4.55 points (SD = 0.67), respectively), “not beautiful–beautiful” (4.32 points (SD = 1.07) and 4.20 points (SD = 0.18), respectively) and “unpleasant–pleasant” (4.75 points (SD = 0.44) and 4.42 points (SD = 0.98), respectively). The GI of 10% and 15% were the most harmonious and environmentally friendly space images in terms of “unharmonious–harmonious” (4.45 points (SD = 0.93) and 4.80 points (SD = 0.51), respectively) and “un-environmentally friendly–environmentally friendly” (4.27 points (SD = 1.15) and 4.32 points (SD = 0.91), respectively). In summary, for the underground subway station with no natural light introduced, the space image with a GI of 15% was most positively evaluated. Further, the ground subway station where natural light could be introduced, the model with a GI of 10% was most positively evaluated. These results agree with the results of a previous study , which concluded that the preference of GI is affected by comfort properties. Meanwhile, as the space with a GI of 15% was recognized as a more “environmentally friendly” space compared to that with a GI of 20% if it exceeds the optimum level, the effect of an “environmentally friendly” space image could be reduced. Therefore, considering the characteristics of a subway station interior space, unlike the exterior space, the interior landscape space should be established by considering the GI evaluated as pleasant and environmentally friendly by users. 3.2.4. Correspondence Analysis on Subway Interior Landscape Models with Different GI As shown in the correspondence analysis table , this analysis was found to be statistically significant ( p < 0.001). The one-dimensional R 2 value was 41.6%, 2D R 2 value was 35.4%, and 2D cumulative R 2 value was 81.5%. Because a high relationship R 2 value is considered when the 2D R 2 is over 70% in correspondence analysis , the relationship between dimensions was significant. As presented in the positional map , which shows obtained results, “GI of 0% in the underground and ground”, “underground GI of 10%” and “ground GI of 15%”, “underground GI of 15%” and “ground GI of 10%”, and “GI of 20% in the underground and ground” were classified as the same group. In terms of “underground GI of 0%” and “ground GI of 0%”, the difference between models was larger than that of other groups. Furthermore, because “underground GI of 15%” and “ground GI of 10%” were classified as a similar group, space is recognized differently depending on the inflow of natural light. “GI of 20% in the underground and ground” were reclassified as “environmentally friendly” spaces providing “high attention restoration effect”: however, they had relatively low scores for “preference”, “suitable landscape area”, “pleasantness”, and “harmonious.” The “models of the underground and ground GI of 10–15%” were reclassified as “harmonious”, “beautiful”, “comfortable”, and “pleasant” spaces. Further, the “underground GI of 15%” and “ground GI of 10%” models had suitable landscape area and preferred it.
3.1.1. α Wave Asymmetry with Different GI for the Interior Landscape of Subway Station Space One-way ANOVA was conducted to investigate the α wave asymmetry of interior landscape models with a different GI of subway station space and closed eyes was showed 0.45 (SD = 0.98). There was a difference in the α wave asymmetry value depending on the presence of installed interior landscape, suggesting that the installation of interior landscapes could facilitate a psychologically stable state compared to space without an interior landscape. The test results for homogeneity of variances showed that the p -value of the F value was 0.05 or higher ( p -value of underground subway station = 0.65, p -value of ground subway station = 0.83) and that the assumption of homogeneity of variance was satisfied. For this reason, Duncan was utilized to conduct a post-hoc test analysis. Meanwhile, the underground space with a GI of 0% had the lowest α wave asymmetry value of 0.47 (SD = 0.90), while the ground GI of 0% had an α wave asymmetry value of 0.64 (SD = 0.8). These differing values indicate that for an environment with no interior landscape model installed, the α wave asymmetry value on the ground where natural light could be introduced was higher than that of the underground. These results are similar to previous studies , which claim that among subjects exposed to roses, the high-frequency component of heart rate variability was significantly higher than in controls. The results of brainwave change with different leaf shapes, sizes, and ear-types of foliage plants in this study were in line with the results of a previous study . An existing study reported no difference in physiological responses according to the GI (5%, 20%, 50%, and 80%) of indoor spaces. Similar to the aforementioned study, this study found no significant difference in models according to the GI of 10%, 15%, and 20%. Therefore, it was concluded that the presence of plants rather than the quantitative increase in plants could positively affect mental relaxation and psychological stability. 3.1.2. High β Wave Asymmetry with Different GI for the Interior Landscape of Subway Station Space One-way ANOVA conducted to investigate the high β wave asymmetry with a different GI of subway station space was 0.07 (SD = 0.24) with closed eyes; the underground GI of 0%, 10%, 15%, and 20% were 0.08, 0.12, 0.11, and 0.09, respectively; and the ground GI of 0%, 10%, 15%, and 20% were 0.03, 0.09, 0.09, and 0.10, respectively, indicating the installation of an interior landscape did not significantly impact high β waves. Such findings are due to the intensity of illumination of artificial lighting being set at 500 lx, which is higher than the mean intensity of illumination of existing subway station space (355.89 lx) , to accord with the criteria of plant vegetation environment while building interior landscape models of subway stations. Therefore, the space with a brighter image than the actual subway station space was established, and stress was not expected to be induced, even with the interior landscape model with a GI of 0%.
One-way ANOVA was conducted to investigate the α wave asymmetry of interior landscape models with a different GI of subway station space and closed eyes was showed 0.45 (SD = 0.98). There was a difference in the α wave asymmetry value depending on the presence of installed interior landscape, suggesting that the installation of interior landscapes could facilitate a psychologically stable state compared to space without an interior landscape. The test results for homogeneity of variances showed that the p -value of the F value was 0.05 or higher ( p -value of underground subway station = 0.65, p -value of ground subway station = 0.83) and that the assumption of homogeneity of variance was satisfied. For this reason, Duncan was utilized to conduct a post-hoc test analysis. Meanwhile, the underground space with a GI of 0% had the lowest α wave asymmetry value of 0.47 (SD = 0.90), while the ground GI of 0% had an α wave asymmetry value of 0.64 (SD = 0.8). These differing values indicate that for an environment with no interior landscape model installed, the α wave asymmetry value on the ground where natural light could be introduced was higher than that of the underground. These results are similar to previous studies , which claim that among subjects exposed to roses, the high-frequency component of heart rate variability was significantly higher than in controls. The results of brainwave change with different leaf shapes, sizes, and ear-types of foliage plants in this study were in line with the results of a previous study . An existing study reported no difference in physiological responses according to the GI (5%, 20%, 50%, and 80%) of indoor spaces. Similar to the aforementioned study, this study found no significant difference in models according to the GI of 10%, 15%, and 20%. Therefore, it was concluded that the presence of plants rather than the quantitative increase in plants could positively affect mental relaxation and psychological stability.
One-way ANOVA conducted to investigate the high β wave asymmetry with a different GI of subway station space was 0.07 (SD = 0.24) with closed eyes; the underground GI of 0%, 10%, 15%, and 20% were 0.08, 0.12, 0.11, and 0.09, respectively; and the ground GI of 0%, 10%, 15%, and 20% were 0.03, 0.09, 0.09, and 0.10, respectively, indicating the installation of an interior landscape did not significantly impact high β waves. Such findings are due to the intensity of illumination of artificial lighting being set at 500 lx, which is higher than the mean intensity of illumination of existing subway station space (355.89 lx) , to accord with the criteria of plant vegetation environment while building interior landscape models of subway stations. Therefore, the space with a brighter image than the actual subway station space was established, and stress was not expected to be induced, even with the interior landscape model with a GI of 0%.
3.2.1. Preference on Subway Interior Landscape Model with Different GI One-way ANOVA was performed to investigate the preference of the subway interior landscape model with a different GI . The underground GI of 15% and ground GI of 10% was highest at 4.48 points (SD = 0.70) and 4.50 points (SD = 0.62), respectively, followed by the underground GI of 10% with 4.05 points (SD = 0.67), the ground GI of 15% with 4.08 points (SD = 0.67), the underground GI of 20% with 3.67 points (SD = 0.73), and the ground GI of 20% with 3.57 points (SD = 0.79). The test result for homogeneity of variances showed that the p -value of the F value was 0.05 or below ( p -value of underground subway station = 0.03, p -value of ground subway station = 0.031) and that the assumption of homogeneity of variance was not satisfied. For this reason, Games–Howell test was used to conduct a post-hoc test analysis . Meanwhile, the preferences of GI of 0% in the underground and ground were 2.40 points (SD = 0.89) and 2.92 points (SD = 0.94), respectively, indicating the preference for a subway station space with an interior landscape rather than for that with no interior landscape. The preference points of GI of 20% in the underground and ground were 3.67 points (SD = 0.73) and 3.57 points (SD = 0.79), respectively, suggesting that the implicit improvement of a GI is not preferred. These results are similar to those of previous studies , which claim that harmonious natural environmental elements including plant leaves become with other elements of whole space considering arrangement and shape has a more significant impact on the preference when introducing plants. Compared with the preferred level of the GI, users believe the interior landscape area with a GI of 15% is suitable; however, for ground subway stations, the model with a GI of 10% is preferred because of the effect and security of the inflow of natural light. Therefore, as ground subway stations are advantageous for securing natural light due to high floor heights, future studies on the required interior GI are needed. A cross-analysis was performed to investigate the recognition level of the interior landscape area of the subway interior landscape model with a different GI . As a result, the underground GI of 15% and the ground GI of 10% received the highest response of “the interior landscape area is adequate” with 70.0% and 66.7%, respectively. Meanwhile, 53.3% and 55.0% of the respondents indicated that “the interior landscape area is adequate” for the underground GI of 10% and the ground GI of 15%, respectively. Of the respondents, 35.5% answered that “the interior landscape area is small” for the underground GI of 10%, while 43.3% of the respondents responded that “the interior landscape area is large” for the ground GI of 15%; this indicates that the recognition level of the interior landscape area differs depending on the availability of natural sunlight. On the other hand, 76.6% and 86.7% of the respondents considered the interior landscape area as “large” and “very large”, respectively, in the case of GI of 20% in the underground and the ground, suggesting that 20% of GI in subway station spaces is considered excessive in terms of the plant density. 3.2.2. Attention Restoration Effect of Subway Interior Landscape Models with Different GI The one-way ANOVA scores of attention restoration effects for an underground and ground GI of 0% were 2.39 points (SD = 0.57) and 2.65 points (SD = 0.70), respectively . Such results indicate that the attention restoration effect of subway station space with interior landscape is higher than that with no interior landscape. The result of the test for homogeneity of variances showed that the p -value of the F value was 0.05 or below ( p -value of underground subway station = 0.09, p -value of ground subway station = 0.38) and that the assumption of homogeneity of variance was not satisfied. For this reason, the Games–Howell test was used to conduct a post-hoc test analysis . The underground GI of 15% showed the highest attention restoration effect of 4.42 points (SD = 0.72), whereas ground GI of 10%, 15%, and 20% were 4.22 (SD = 0.68), 4.39 (SD = 0.78), and 4.24 points (SD = 0.71), respectively. These results suggest that for the ground GI, only a 10% GI could lead to an attention restoration effect of 15–20% GI. 3.2.3. Space Image of Subway Interior Landscape Model with Different GI The one-way ANOVA was conducted to evaluate space images of 3D landscape models of the subway interior according to the GI ( , and ). The test results for homogeneity of variances according to space image item showed that the p -value of the F value was 0.05 or below and that the assumption of homogeneity of variance was not satisfied. For this reason, the Games–Howell test was used to conduct a post-hoc test analysis ( and ). As for the “dark–bright” image in underground subway stations, the GI of 15% was 4.75 points, being the highest. However, a GI of 10% and 15% were the most comfortable space images, with 4.20 (SD = 0.99) and 4.52 points (SD = 0.93), respectively, for “uncomfortable–comfortable.” The GI of 15% was 4.45 points (SD = 0.87) for “not beautiful–beautiful”, suggesting that it was the most beautiful space image. The GI of 10% and 15% were 4.42 points (SD = 0.87) and 4.67 points (SD = 0.66), respectively, for “unpleasant–pleasant”, indicating that they were the most pleasant space images. The GI of 10% and 15% resulted in 4.03 (SD = 1.29) and 4.42 points (SD = 0.98), respectively, for “unharmonious–harmonious”, indicating they were the most harmonious space images. The GI of 15% was found to be the most environmentally-friendly space image for “un-environmentally friendly–environmentally friendly.” Regarding ground subway stations, the GI of 10% and 15% were the brightest space images with 4.68 (SD = 0.65) and 4.50 points (SD = 0.87), respectively, for “dark–bright.” The GI of 10% and 15% were the most positive images for “uncomfortable–comfortable” (4.43 points (SD = 0.87) and 4.55 points (SD = 0.67), respectively), “not beautiful–beautiful” (4.32 points (SD = 1.07) and 4.20 points (SD = 0.18), respectively) and “unpleasant–pleasant” (4.75 points (SD = 0.44) and 4.42 points (SD = 0.98), respectively). The GI of 10% and 15% were the most harmonious and environmentally friendly space images in terms of “unharmonious–harmonious” (4.45 points (SD = 0.93) and 4.80 points (SD = 0.51), respectively) and “un-environmentally friendly–environmentally friendly” (4.27 points (SD = 1.15) and 4.32 points (SD = 0.91), respectively). In summary, for the underground subway station with no natural light introduced, the space image with a GI of 15% was most positively evaluated. Further, the ground subway station where natural light could be introduced, the model with a GI of 10% was most positively evaluated. These results agree with the results of a previous study , which concluded that the preference of GI is affected by comfort properties. Meanwhile, as the space with a GI of 15% was recognized as a more “environmentally friendly” space compared to that with a GI of 20% if it exceeds the optimum level, the effect of an “environmentally friendly” space image could be reduced. Therefore, considering the characteristics of a subway station interior space, unlike the exterior space, the interior landscape space should be established by considering the GI evaluated as pleasant and environmentally friendly by users. 3.2.4. Correspondence Analysis on Subway Interior Landscape Models with Different GI As shown in the correspondence analysis table , this analysis was found to be statistically significant ( p < 0.001). The one-dimensional R 2 value was 41.6%, 2D R 2 value was 35.4%, and 2D cumulative R 2 value was 81.5%. Because a high relationship R 2 value is considered when the 2D R 2 is over 70% in correspondence analysis , the relationship between dimensions was significant. As presented in the positional map , which shows obtained results, “GI of 0% in the underground and ground”, “underground GI of 10%” and “ground GI of 15%”, “underground GI of 15%” and “ground GI of 10%”, and “GI of 20% in the underground and ground” were classified as the same group. In terms of “underground GI of 0%” and “ground GI of 0%”, the difference between models was larger than that of other groups. Furthermore, because “underground GI of 15%” and “ground GI of 10%” were classified as a similar group, space is recognized differently depending on the inflow of natural light. “GI of 20% in the underground and ground” were reclassified as “environmentally friendly” spaces providing “high attention restoration effect”: however, they had relatively low scores for “preference”, “suitable landscape area”, “pleasantness”, and “harmonious.” The “models of the underground and ground GI of 10–15%” were reclassified as “harmonious”, “beautiful”, “comfortable”, and “pleasant” spaces. Further, the “underground GI of 15%” and “ground GI of 10%” models had suitable landscape area and preferred it.
One-way ANOVA was performed to investigate the preference of the subway interior landscape model with a different GI . The underground GI of 15% and ground GI of 10% was highest at 4.48 points (SD = 0.70) and 4.50 points (SD = 0.62), respectively, followed by the underground GI of 10% with 4.05 points (SD = 0.67), the ground GI of 15% with 4.08 points (SD = 0.67), the underground GI of 20% with 3.67 points (SD = 0.73), and the ground GI of 20% with 3.57 points (SD = 0.79). The test result for homogeneity of variances showed that the p -value of the F value was 0.05 or below ( p -value of underground subway station = 0.03, p -value of ground subway station = 0.031) and that the assumption of homogeneity of variance was not satisfied. For this reason, Games–Howell test was used to conduct a post-hoc test analysis . Meanwhile, the preferences of GI of 0% in the underground and ground were 2.40 points (SD = 0.89) and 2.92 points (SD = 0.94), respectively, indicating the preference for a subway station space with an interior landscape rather than for that with no interior landscape. The preference points of GI of 20% in the underground and ground were 3.67 points (SD = 0.73) and 3.57 points (SD = 0.79), respectively, suggesting that the implicit improvement of a GI is not preferred. These results are similar to those of previous studies , which claim that harmonious natural environmental elements including plant leaves become with other elements of whole space considering arrangement and shape has a more significant impact on the preference when introducing plants. Compared with the preferred level of the GI, users believe the interior landscape area with a GI of 15% is suitable; however, for ground subway stations, the model with a GI of 10% is preferred because of the effect and security of the inflow of natural light. Therefore, as ground subway stations are advantageous for securing natural light due to high floor heights, future studies on the required interior GI are needed. A cross-analysis was performed to investigate the recognition level of the interior landscape area of the subway interior landscape model with a different GI . As a result, the underground GI of 15% and the ground GI of 10% received the highest response of “the interior landscape area is adequate” with 70.0% and 66.7%, respectively. Meanwhile, 53.3% and 55.0% of the respondents indicated that “the interior landscape area is adequate” for the underground GI of 10% and the ground GI of 15%, respectively. Of the respondents, 35.5% answered that “the interior landscape area is small” for the underground GI of 10%, while 43.3% of the respondents responded that “the interior landscape area is large” for the ground GI of 15%; this indicates that the recognition level of the interior landscape area differs depending on the availability of natural sunlight. On the other hand, 76.6% and 86.7% of the respondents considered the interior landscape area as “large” and “very large”, respectively, in the case of GI of 20% in the underground and the ground, suggesting that 20% of GI in subway station spaces is considered excessive in terms of the plant density.
The one-way ANOVA scores of attention restoration effects for an underground and ground GI of 0% were 2.39 points (SD = 0.57) and 2.65 points (SD = 0.70), respectively . Such results indicate that the attention restoration effect of subway station space with interior landscape is higher than that with no interior landscape. The result of the test for homogeneity of variances showed that the p -value of the F value was 0.05 or below ( p -value of underground subway station = 0.09, p -value of ground subway station = 0.38) and that the assumption of homogeneity of variance was not satisfied. For this reason, the Games–Howell test was used to conduct a post-hoc test analysis . The underground GI of 15% showed the highest attention restoration effect of 4.42 points (SD = 0.72), whereas ground GI of 10%, 15%, and 20% were 4.22 (SD = 0.68), 4.39 (SD = 0.78), and 4.24 points (SD = 0.71), respectively. These results suggest that for the ground GI, only a 10% GI could lead to an attention restoration effect of 15–20% GI.
The one-way ANOVA was conducted to evaluate space images of 3D landscape models of the subway interior according to the GI ( , and ). The test results for homogeneity of variances according to space image item showed that the p -value of the F value was 0.05 or below and that the assumption of homogeneity of variance was not satisfied. For this reason, the Games–Howell test was used to conduct a post-hoc test analysis ( and ). As for the “dark–bright” image in underground subway stations, the GI of 15% was 4.75 points, being the highest. However, a GI of 10% and 15% were the most comfortable space images, with 4.20 (SD = 0.99) and 4.52 points (SD = 0.93), respectively, for “uncomfortable–comfortable.” The GI of 15% was 4.45 points (SD = 0.87) for “not beautiful–beautiful”, suggesting that it was the most beautiful space image. The GI of 10% and 15% were 4.42 points (SD = 0.87) and 4.67 points (SD = 0.66), respectively, for “unpleasant–pleasant”, indicating that they were the most pleasant space images. The GI of 10% and 15% resulted in 4.03 (SD = 1.29) and 4.42 points (SD = 0.98), respectively, for “unharmonious–harmonious”, indicating they were the most harmonious space images. The GI of 15% was found to be the most environmentally-friendly space image for “un-environmentally friendly–environmentally friendly.” Regarding ground subway stations, the GI of 10% and 15% were the brightest space images with 4.68 (SD = 0.65) and 4.50 points (SD = 0.87), respectively, for “dark–bright.” The GI of 10% and 15% were the most positive images for “uncomfortable–comfortable” (4.43 points (SD = 0.87) and 4.55 points (SD = 0.67), respectively), “not beautiful–beautiful” (4.32 points (SD = 1.07) and 4.20 points (SD = 0.18), respectively) and “unpleasant–pleasant” (4.75 points (SD = 0.44) and 4.42 points (SD = 0.98), respectively). The GI of 10% and 15% were the most harmonious and environmentally friendly space images in terms of “unharmonious–harmonious” (4.45 points (SD = 0.93) and 4.80 points (SD = 0.51), respectively) and “un-environmentally friendly–environmentally friendly” (4.27 points (SD = 1.15) and 4.32 points (SD = 0.91), respectively). In summary, for the underground subway station with no natural light introduced, the space image with a GI of 15% was most positively evaluated. Further, the ground subway station where natural light could be introduced, the model with a GI of 10% was most positively evaluated. These results agree with the results of a previous study , which concluded that the preference of GI is affected by comfort properties. Meanwhile, as the space with a GI of 15% was recognized as a more “environmentally friendly” space compared to that with a GI of 20% if it exceeds the optimum level, the effect of an “environmentally friendly” space image could be reduced. Therefore, considering the characteristics of a subway station interior space, unlike the exterior space, the interior landscape space should be established by considering the GI evaluated as pleasant and environmentally friendly by users.
As shown in the correspondence analysis table , this analysis was found to be statistically significant ( p < 0.001). The one-dimensional R 2 value was 41.6%, 2D R 2 value was 35.4%, and 2D cumulative R 2 value was 81.5%. Because a high relationship R 2 value is considered when the 2D R 2 is over 70% in correspondence analysis , the relationship between dimensions was significant. As presented in the positional map , which shows obtained results, “GI of 0% in the underground and ground”, “underground GI of 10%” and “ground GI of 15%”, “underground GI of 15%” and “ground GI of 10%”, and “GI of 20% in the underground and ground” were classified as the same group. In terms of “underground GI of 0%” and “ground GI of 0%”, the difference between models was larger than that of other groups. Furthermore, because “underground GI of 15%” and “ground GI of 10%” were classified as a similar group, space is recognized differently depending on the inflow of natural light. “GI of 20% in the underground and ground” were reclassified as “environmentally friendly” spaces providing “high attention restoration effect”: however, they had relatively low scores for “preference”, “suitable landscape area”, “pleasantness”, and “harmonious.” The “models of the underground and ground GI of 10–15%” were reclassified as “harmonious”, “beautiful”, “comfortable”, and “pleasant” spaces. Further, the “underground GI of 15%” and “ground GI of 10%” models had suitable landscape area and preferred it.
This study provided reference data to prepare an interior landscape strategy for developing a subway station space by constructing a 3D interior landscape model, which considers the characteristics of subway station spaces and analyzes the psychological and physiological response of users about GI. Subway stations were classified as ground subway stations where natural light can be introduced and underground subway stations where natural light cannot be introduced to design interior landscapes considering the physical environments of subway stations. Based on the research results, the optimum GI range for planning a pleasant and healthy interior landscape in a subway station space is summarized as follows. First, based on the physiological response to the different GI of subway station spaces, the prefrontal α wave asymmetry value increased when the GI was 0% and 10–20%, regardless of the inflow of natural light (underground/ground), thereby suggesting that the installation of interior lighting evokes positive emotions. Therefore, introducing an interior landscape in subway station space is necessary because the introduction of plants positively affects mental relaxation and psychological stability. Moreover, because psychophysiological stability and satisfaction is ensured by improving the GI in subway station spaces, it is expected to result in positive impacts such as environmental improvement and activation of subway station spaces. Second, based on the psychological response with a different GI of the subway station space, the underground GI of 15% and the ground GI of 10% had the highest preference in the interior landscape model and the most suitable level of interior landscape installation area recognition. Furthermore, they were the most positively evaluated in all the space image adjective items. Such findings indicate that the GI of 10–15% is appropriate in interior landscape design for subway station spaces. In addition, based on detailed analysis using multidimensional scaling, the models of the underground GI of 15% and the ground GI of 10% were similar. They were the most preferred interior landscape models with the most suitable interior landscape area. This analysis shows that when plants are introduced in an indoor space, the degree of harmony between natural environmental elements (including plant leaves) with other elements in the space regarding arrangement, shape, and space characteristics has a more significant effect on the preference. Therefore, interior landscape planning must accommodate the space characteristics when plants are introduced in indoor spaces. Therefore, for underground subway stations with no natural lighting, a GI of 15% was the most suitable. Meanwhile, for the ground subway stations with natural lighting, the interior landscape with a GI of 10% was sufficient. Across the subway stations in Busan, the value of GI where an interior landscape is installed varies from a minimum of 1.03% to a maximum of 28.94% . Furthermore, efforts to promote the implementation of and improve the interior landscape have been recently undertaken. Therefore, there is a pressing need for introducing interior landscapes with GIs falling within an appropriate range in consideration of the efficient use of subway station spaces, psychological satisfaction of the users, and cost of introduction and maintenance of the interior landscape. According to a comparative analysis of the results of the present study and the previous related studies, a study that examined the physiological responses of users showed no difference in heart rate variability and EEG based on GI , which is similar to the EEG findings obtained in this study. It was indicated that the interior space with landscape induced users to feel more comfortable than the spaces without an interior landscape. Pertaining to psychological responses, in the study of , which examined the GI preference by setting up a laboratory having an area of 1.5 m × 1.5 m, the preference for a GI of 50% was found to be high. In another study that investigated GI preference in workspace , the GI of 10–20% was the most preferred, and when the GI is 20%, the attention restoration effect is high. Meanwhile, in this study, GI preference and attention restoration effect were the highest when the GI of 15% for underground subway stations and GI of 10% for ground subway stations. These findings indicate that different results are obtained depending on the target space of research. Taken together, it is necessary to implement landscaping in indoor spaces for a positive emotional effect on users, with an emphasis on the interior landscape design considering the functional aspects of the space and the characteristics of the physical environment. In this study, a differential GI response analysis was conducted using a 3D model according to the movement of users, such that the user experienced all the aspects of the interior landscape introduced in the subway station space. Our findings could serve to improve the existing GI measurement methods using 2D images and laboratory-based experiments. Furthermore, it can be argued that the proposed method is more effective in interior landscape design, taking into account the physical environment characteristics of subway stations. This study has significance based on the following reasons. First, it implemented 3D landscape models of the subway interior by classifying subway stations as underground and ground subway stations according to the presence of natural light inflow in consideration of the physical properties of subway station spaces. Second, it evaluated responses of subway station users by implementing 3D landscape models of the subway interior instead of performing flat (2D) measurements as conducted in existing studies. Third, it divided subjects according to the GI as physiological responses and psychological responses to perform a more precise analysis and establish a direction for enhancing indoor environments of subway station spaces. Because the GI preference varies according to the presence of natural light inflow, it is important to develop an interior landscape plan by considering the characteristics of subway station spaces. In this regard, it is expected that the result of this study can provide data for developing an interior landscape plan for subway station spaces and enhancing such interior landscapes. The limitations of this research are as follows. First, because the GI can be planned by utilizing various interior landscape techniques other than garden type/planter and wall type/pillar types, future studies on various interior landscape techniques are needed when introducing interior landscapes. Second, research considering demographic characteristics should be conducted by sampling and investigating research subjects of all ages. Third, evaluation of the GI needs to be continuously performed by selecting various public space targets other than subway station spaces as research and expanding the research range. Therefore, further research on applying various interior landscape formation methods and accumulating research data with more research participants is required. It is anticipated that these processes will contribute to developing a more specific and feasible plan for implementing interior landscape in subway station spaces.
|
Precise in-field molecular diagnostics of crop diseases by smartphone-based mutation-resolved pathogenic RNA analysis | eab3e0fc-c8e9-4200-9e05-091ff4a67f72 | 10356797 | Pathology[mh] | A sustainable food supply is burdened by the increasing global population . Crop diseases caused by phytopathogens, mainly fungal pathogens, devastate crops and put food supplies at risk – . To combat crop pathogens, pesticides are overused, which has damaged the environment and soil health. However, despite two million tonnes of pesticides, the Food and Agriculture Organization reports an annual yield reduction of wheat and rice by 21.5% and 30.0%, respectively . Crop fungal disease management is therefore crucial for food security, but remains challenging. Molecular diagnostic tools that can be routinely deployed in the field to provide the information about clonal lineage, virulence and drug-resistance of pathogens, should enable us to detect an infection at the early stage, precisely estimate the disease risk, and decide which fungicide to use, and when – . These tools are essential for crop disease control, but are currently unavailable. Conventional phenotypic methods such as symptom observation, foliage machine vision and imaging approaches—including hyperspectral, thermographic, and infrared thermometric imaging technologies—have been successively employed to achieve non-invasive and in-field measurements of crop infection – . In most cases, however, these methods cannot identify the pathogen. Molecular diagnostics based on genotypic or immunological methods can offer information about pathogeny , . Nucleic acid-based methods, in principle, can cover the detection of any species of pathogens – . Besides, genetic information can illuminate the virulence and drug resistance features of phytopathogens , , which is vital for guiding the adoption of the optimal treatment strategy for particular diseases. An available low-cost in-field genotypic method could fill the gaps between molecular diagnostics and conventional crop disease monitoring. However, nucleic acid-based methods are currently complex, primarily due to the requirement for a nucleic acid amplification process. Nucleic acid assays developed to date are mostly restricted to labs, and are not applicable in the field. Here, we demonstrate an in-field, rapid and cost-effective screening strategy for wheat pathogens with information about viability and drug-resistance using a nucleic acid amplification-free, gene mutation-resolved and smartphone-integrated genetic assay. The smartphone-based diagnostic tool has great potential advancing precision crop disease management. Assay principle The fungal RNA sequence is recognized based on a simple nucleic acid reaction, termed toehold-mediated strand displacement (TMSD), requiring a double-stranded DNA probe (the DProbe) (Fig. and Supplementary Fig. ). The DProbe is a hybrid of the cis strand and the trans strand, and is designed with terminal domains called the forward and reverse toehold (shown in pink and green, respectively). The forward toehold is the domain in the cis strand that hybridizes with the fungal RNA and not with the trans strand. Conversely, the reverse toehold is the domain that hybridizes with the trans strand and not with the fungal RNA. There is one cytosine-Ag(I)-cytosine artificial base pair in the reverse toehold (Supplementary Fig. ). In the presence of fungal RNA, the TMSD reaction is initiated and facilitated via hybridization between the fungal RNA and the forward toehold, but is hindered by the disruption of the reverse toehold hybridization. Thus, the reaction can be precisely controlled via tuning the toehold domain. Fungal RNA fuels the TMSD reaction to release the trans strand and the Ag(I) ion from the DProbe. The presence of fungal RNAs is then indicated by urease-based ammonia production and a pH indicator, phenol red (which turns from yellow to red over the pH range 6.8 to 8.2). Urease catalyses the hydrolysis of urea to yield carbamate and ammonia (Supplementary Fig. ); the production of ammonia contributes to a notable pH increase (to pH > 8.2), where upon phenol red acquires its red colour. Simultaneously, urease is highly sensitive to trace quantities of Ag (I) ion ; Ag(I) ion has a strong inhibitory effect on urease, causing phenol red to become yellow (pH <6.8). The presence of the fungal RNA releases Ag(I) ions to block urease, and can thus be visualized by the colour change of phenol red. Puccinia striiformis ( Pst ) can cause a destructive disease of wheat, stripe rust, and significantly reduce wheat production worldwide , . We synthesized four short non-overlapping RNA sequences (22 nucleotides, nt) on the internal transcribed spacer (ITS) of Pst , to optimize the binding site in the ITS based on colour change of the assay (Supplementary Fig. and Supplementary Table ). The DProbe targeting the optimized binding site 3 was then used to test the assay principle. Urease catalysis and Ag(I) ion-induced urease inhibition were verified by measuring the absorbance of phenol red (Fig. ). The TMSD reaction was demonstrated using electrophoretic and fluorescence analysis (Fig. , Supplementary Fig. and Supplementary Table ). For detecting long fungal RNAs from Pst , we found that lengthening the forward toehold from 8 nt to 15 nt facilitated the TMSD reaction, and improved assay’s response to Pst (Supplementary Fig. ). We then evaluated the assay performance for detecting Pst . A dilution test of Pst showed that the assay can detect 1.0 ng/μL Pst without a nucleic acid amplification procedure (Fig. ) (Welch’s t test: *P < 0.05). Neither extraction of nucleic acid from Pst infected wheat leaves nor mixing Pst RNAs with wheat leaf RNAs significantly affected the absorbance of the assay compared to only Pst RNAs (Supplementary Fig. ), and the presence of wheat RNA matrix did not compromise the sensitivity for detecting Pst (Supplementary Fig. ). Congeneric pathogenic fungi can cause similar infection symptoms and have been overlapping geographical distribution. We tested whether the assay could discriminate among three congeneric rust fungi, Pst, Puccinia graminis ( Pgt ) and Puccinia triticina ( Pt ). The ITS sequences of Pgt and Pt are similar to that of Pst (Fig. ). Three DProbes were designed to target the ITSs of Pst , Pgt and Pt . We mixed each set of RNAs extracted from these pathogenic fungi, and each could be detected using its corresponding DProbe without cross-interference. Besides rust fungi, Blumeria graminis ( Bgt ), Fusarium graminearum ( Fg ) and Rhizoctonia cerealis ( Rc ) responsible for yield reduction of wheat , they cause the diseases powdery mildew, fusarium head blight and wheat sharp eyespot, respectively. By aiming for the ITS sequence of the six fungi (Supplementary Table ), the assay allowed visual detection of each species (Supplementary Fig. ). Multiplex detection of fungal pathogens using colorimetric papers To facilitate in-field detection, we integrated each assay reaction using a paper folding strategy termed origami. Wax printing was utilized to prepare the origami paper with defined detection spots via the formation of hydrophobic barrier . The procedures are shown in Supplementary Fig. . Reagents including DProbe, enzyme, and pH indicator were then loaded on each page of the paper (Fig. ). The folding procedure sequentially initiated the TMSD reaction and colour reporting reaction, and yielded a folded paper to be imaged by the camera on a smartphone. We developed an algorithm to locate the detection spot and defined a green pixel ratio (GPR) value to record the colour response of phenol red towards fungal RNAs. GPR value was defined as the portion of pixels with positive grey value in the green channel within the detection spot. The image processing method is illustrated in Supplementary Fig. . We found that the origami paper with a large pore size or surfactant modification yielded improved sensitivity, colour consistency and uniformity of the colorimetric response via the paper-folding strategy (Supplementary Fig. , Supplementary Fig. , Supplementary Note ). In addition, the robustness of the detection of Pst using the origami papers related to batches of origami papers and operators has been tested (Supplementary Fig. ). Colour calibration has been demonstrated to be able to improve the robustness of the detection using different smartphones (Supplementary Fig. ). By designing the origami paper with eight sample-loading sites, six pathogenic fungi ( Pst, Pgt, Pt, Bgt, Fg and Rc ) can be in parallel detected (Fig. ). The result showed that all six fungi could be distinguished without cross-interference using one origami paper loaded with their congenetic probes (Fig. ). The colorimetric assay using the origami papers yielded a comparable sensitivity compared to that proceeded in solution, and allowed to detect as low as 1 ng/μL Pst (Supplementary Fig. ). We further used the assay to detect fungal pathogens in infected wheat samples. The infection of Fusarium culmorum that causes fusarium crown rot can be indicated via the colour change of the origami papers (Supplementary Fig. ). To further evaluate the accuracy of the assay, thirty-two wheat leaf samples collected from different fields in China were tested (Supplementary Figs. , ). The samples were analysed in parallel using quantitative PCR (qPCR) and the colorimetric paper-based assay. The assay showed an agreement of 95.3% positive prediction, and 91.0% negative prediction when compared to qPCR in the test of the infection of Pst (Fig. , Supplementary Table , ). Besides pathogenic fungi, we explored the assay to detect a plant pathogenic bacterium, Pseudomonas syringae (Supplementary Fig. ) and a crop pathogenic virus, barley stripe mosaic virus (Supplementary Fig. ), and demonstrated that the assay is capable to diagnose the infection of not only fungi but also viruses and bacteria, showing the potential for the broad applicability for in-field detection of crop diseases. Early diagnosis of fungal infection Early detection of crop infection can dramatically alleviate crop yield reduction and reduce fungicide use. We tested the assay for early diagnosis of Pst infection. Figure shows the observed phenotype of wheat leaves for during two weeks after infection. Spores are visible by the naked eye on day 10, demonstrating that wheat infection can be recognised after 10-day-infection by symptom observation. Pst in infected wheat leaves was stained using wheat germ agglutinin . Histological observation also showed the rapid spread of Pst in the leaves by10 days after inoculation. Nucleic acid was extracted from the infected leaves, and analysed by the colorimetric assay and qPCR. RNA extracted from the wheat samples on day 3 triggered a sharp drop of the absorbance signal in the colorimetric assay (Fig. ) (Welch’s t test: *P < 0.05). Pst infection can thus be identified after 3-day infection using the colorimetric assay. The qPCR result yielded positive detection of Pst from the 3-day-infected wheat samples (Welch’s t test: ** *P < 0.001) (Fig. , Supplementary Fig. ). It was estimated that 64.29 ng/μL Pst was present in the 3-day-infected wheat samples (Supplementary Fig. ). The paper-based strategy was further applied for parallel detection of Pst -infected wheat samples ( n = 6). Pst infection was ascertained after 3 days using the colorimetric paper (Fig. ) (Welch’s t test: **P < 0.01). The paper-based assay using optimized origami paper yielded an increased GPR value for testing Pst from the wheat sample infected for 3 days and a reduced variation in duplicate detection (Supplementary Fig. ), facilitating the robustness of positive detection of early infection by Pst . In addition, we inoculated wheat leaves with Pst in a dilution series. The colorimetric assay allowed to detect low-level Pst that did not cause observable infection symptom in the wheat leaves infected for 14 days (Supplementary Fig. ). Collectively, the colorimetric assay showed a capacity for early detection of Pst infection comparable to qPCR, and advanced the identification of an infection by 7 days compared to symptom observation. Viable fungus detection facilitates prediction of disease occurrence and severity Disease cycles comprise four stages: dormancy, reproduction, dispersal and pathogenesis . Pathogens may be present on residues left in the field, in soil, and on weeds and tools. Importantly, only viable pathogenic fungi in dormancy will be activated in favourable conditions and enter the reproduction stage. Therefore, it is essential to devise strategies to distinguish viable fungi in total fungal samples to prevent disease circulation. Current DNA-targeting methods, such as qPCR, usually fail to distinguish viable pathogens from dead ones for the long-term persistence of DNA in dead microbial cells. RNA, however, is rapidly degraded in dead cells, and so, methods that target RNAs, such as the colorimetric assay, should specifically detect viable pathogens , . We therefore tested the feasibility of viable fungus detection using Pst mixtures containing viable and dead spores, in which the portion of viable spores was 0%, 0.1%, 1%, 10% and 100% (Fig. ). Nucleic acid was extracted from equal quantities of these Pst mixtures, and analysed using the colorimetric assay and qPCR. The results showed that absorbance signals attenuated gradually with the increase of viable fungi in both the absence and presence of wheat RNA matrix (Fig. , Supplementary Fig. ), indicating that the colorimetric assay can reliably indicate the amount of viable fungus. In contrast, there was no significant difference among Pst mixtures using qPCR (Fig. , Supplementary Fig. ). Pst mixtures with different proportions of viable spores were utilized to inoculate wheat. Due to the low abundance of Pst , the Ct values for qPCR measurement of Pst infection at 0 day were all close to 40 (Supplementary Fig. ). Fourteen days after the infection, leaves inoculated with Pst mixtures containing 0% and 0.1% viable fungi showed no obvious symptoms, but when the content of viable fungi increased to 1%, the leaves turned yellow and were loaded with observable spore piles (Fig. ). Measurements using qPCR and the colorimetric assay indicated that the amount of both total and viable Pst fungus in infected leaves increased in accordance with the proportion of viable fungus in the Pst mixture used for inoculation (Fig. and Supplementary Fig. ). This result indicated that the occurrence and severity of stripe rust were related specifically to the quantity of viable fungus, rather than the total quantity of all fungus. Detection of viable fungal pathogens is highly important, given that a large proportion of fungal cells may die through winter and summer. These fungi cannot cause effective crop infection. Because it can detect viable fungi, the colorimetric assay allows for a more precise prediction of disease occurrence and severity compared to methods that cannot distinguish viable pathogens from dead ones. We also investigated the defence response of wheat towards dead Pst . Dead spores slightly increased the expression of two pathogenesis-related genes, PR1 and PR2 (Supplementary Fig. ) , , but pre-inoculation with dead spores did not significantly change Pst biomass after 12, 24 or 48 h of infection with viable spores (Supplementary Fig. , Supplementary Note ). The result indicates that dead pathogenic fungi do not elicit a defence response that can efficiently inhibit pathogen infection. Identification of fungicide-resistant isolates Fungicides are intensively used to prevent and treat crop diseases. The emergence of fungicide-resistance has become a severe issue. Identifying fungicide-resistant isolates can instruct us to choose an effective fungicide or other treatment strategies . Mutations leading to conformational changes in the drug target site are the main cause of fungicide-resistance in pathogenic fungi. CYP51, one of the cytochrome P450 monooxygenases, acts on fungal invasive growth, hypha formation and virulence. Inhibitors that target CYP51 serve as key antifungal agents , . The point mutation Y134F in CYP51 was found to be associated with a significant degree of triadimefon resistance in Pst isolates . Via competitive hybridization to hinder binding with non-target RNAs, an assay based on TMSD showed promise for detecting single-nucleotide mutations (SNMs) in RNAs. The principle of TMSD for identifying SNMs is illustrated in Supplementary Note . To maximize discrimination between the mutated RNA (F134 RNA) and the non-target wild type RNA (Y134 RNA), with a single-nucleotide difference, the TMSD reaction should be optimized by tuning the toehold length of the DProbe, which allows to block non-specific hybridization induced by the wild RNA while the DProbe/mutated RNA hybrid still forms. We designed DProbes targeting the Y134F mutation with a fixed 7-nt reverse toehold and forward toeholds ranging from 5 nt to 25 nt (Supplementary Fig. ). Fluorescence analysis of the TMSD reaction using a fluorophore and quencher-modified DProbe showed that DProbes with forward toehold lengths of 5, 9, 13, and 17 -nt can effectively distinguish F134 RNA from Y134 RNA. Long forward toeholds (21 nt and 25 nt) yielded a low discrimination capacity for the mutation because both F134 RNA and Y134 RNA can efficiently fuel the TMSD reaction. In contrast, shorter toehold length led to a lower displacement efficiency for both F134 RNA and Y134 RNA (Supplementary Fig. ). Electrophoretic analysis confirmed these results (Supplementary Fig. ). Based on the colorimetric reaction of the assay (Fig. ), the highest ratio of the absorbance of F134 RNA to that of Y134 RNA was achieved using the DProbe with a 17-nt forward toehold. Using this optimized DProbe, a dilution experiment with the mutated isolate indicated that, in either the absence or the presence of wheat RNA matrix, the assay could detect as little as 0.1% mutated isolate in a background of 99.9% wild type (Fig. , Supplementary Fig. ). The assay can thus detect low-abundance single-nucleotide mutations and should therefore be useful for finding rare fungicide-resistant isolates. We collected eight Pst -infected wheat samples from field sites in China (Fig. ). Using the above assay, we found that four isolates (XY-14, GZ-01, GY-07 and GY-09) from these samples were Y134F mutation-free, and four isolates (TS-05, TS-07, TS-02 and GD-04) had the Y134F mutation (Fig. ), which indicated the possibility of triadimefon resistance in the latter four isolates. Pst isolates from the eight wheat leaves were collected and sequenced. XY-14, GZ-01, GY-07 and GY-09 were homozygous wild type; and TS-05, TS-07, TS-02 and GD-04 were heterozygous Y134F mutants (Fig. ). Responses of eight Pst isolates to triadimefon showed that TS-05, TS-07, TS-02 and GD-04 had a much higher EC 50 towards triadimefon than other isolates (Fig. , Supplementary Fig. , and Supplementary Table ). These results demonstrate the feasibility of the colorimetric assay for identifying mutation-carrying fungicide-resistant isolates. Smartphone-assisted in-field diagnosis To explore the ‘sample-to-answer’ in-field use of the assay, we integrated the assay with rapid nucleic acid extraction strategies, smartphone attachment and a mobile application program (app) (Fig. , Supplementary Fig. ). Nucleic acid extraction is a bottleneck for on-site or in-field nucleic acid tests, due to the complex setups involved that include centrifugation and chemical reagents. First, we explored a simple strategy using quartz sand to grind wheat leaf samples and release nucleic acid (Supplementary Fig. ), and found that the presence of Pst can be detected due to the colour change compared to samples without Pst . However, pigments in leaves can induce colour deviation in the papers, and consequently the quantification of GPR values was not stable. As an alternative, we then synthesized a microneedle (MN) patch for nucleic acid extraction (Supplementary Figs. , ). The MN patch, made of polyvinyl alcohol, has been used to extract DNA from leaf tissues . The patch can penetrate the wheat leaf and break cell walls to isolate nucleic acids. Via a rapid swelling through absorption of water molecules, the patch absorbs and concentrates nucleic acids on the needle tips. MN patch extraction yielded high-purity nucleic acid, but extracted less nucleic acid, due to the relatively low sample volume, than the Trizol-based method (Supplementary Fig. ). Nevertheless, the use of an MN patch accomplishes nucleic acid extraction within 1 min via a single press action; no instruments or chemical reagents are needed, thus facilitating nucleic acid tests for in-field diagnostics for crop diseases. By designing a smartphone attachment, the colorimetric detection of fungal RNAs could be completed by three simple pull operations, and the risk of contamination was avoided (Supplementary Fig. ). Furthermore, an app was designed for the graphical user guidance and the display of diagnostic results (Supplementary Fig. ). Notably, the app can also provide treatment advice based on the diagnostic results (Supplementary Fig. , Supplementary Table ), which is an essential feature for effective crop disease control. The fungal RNA sequence is recognized based on a simple nucleic acid reaction, termed toehold-mediated strand displacement (TMSD), requiring a double-stranded DNA probe (the DProbe) (Fig. and Supplementary Fig. ). The DProbe is a hybrid of the cis strand and the trans strand, and is designed with terminal domains called the forward and reverse toehold (shown in pink and green, respectively). The forward toehold is the domain in the cis strand that hybridizes with the fungal RNA and not with the trans strand. Conversely, the reverse toehold is the domain that hybridizes with the trans strand and not with the fungal RNA. There is one cytosine-Ag(I)-cytosine artificial base pair in the reverse toehold (Supplementary Fig. ). In the presence of fungal RNA, the TMSD reaction is initiated and facilitated via hybridization between the fungal RNA and the forward toehold, but is hindered by the disruption of the reverse toehold hybridization. Thus, the reaction can be precisely controlled via tuning the toehold domain. Fungal RNA fuels the TMSD reaction to release the trans strand and the Ag(I) ion from the DProbe. The presence of fungal RNAs is then indicated by urease-based ammonia production and a pH indicator, phenol red (which turns from yellow to red over the pH range 6.8 to 8.2). Urease catalyses the hydrolysis of urea to yield carbamate and ammonia (Supplementary Fig. ); the production of ammonia contributes to a notable pH increase (to pH > 8.2), where upon phenol red acquires its red colour. Simultaneously, urease is highly sensitive to trace quantities of Ag (I) ion ; Ag(I) ion has a strong inhibitory effect on urease, causing phenol red to become yellow (pH <6.8). The presence of the fungal RNA releases Ag(I) ions to block urease, and can thus be visualized by the colour change of phenol red. Puccinia striiformis ( Pst ) can cause a destructive disease of wheat, stripe rust, and significantly reduce wheat production worldwide , . We synthesized four short non-overlapping RNA sequences (22 nucleotides, nt) on the internal transcribed spacer (ITS) of Pst , to optimize the binding site in the ITS based on colour change of the assay (Supplementary Fig. and Supplementary Table ). The DProbe targeting the optimized binding site 3 was then used to test the assay principle. Urease catalysis and Ag(I) ion-induced urease inhibition were verified by measuring the absorbance of phenol red (Fig. ). The TMSD reaction was demonstrated using electrophoretic and fluorescence analysis (Fig. , Supplementary Fig. and Supplementary Table ). For detecting long fungal RNAs from Pst , we found that lengthening the forward toehold from 8 nt to 15 nt facilitated the TMSD reaction, and improved assay’s response to Pst (Supplementary Fig. ). We then evaluated the assay performance for detecting Pst . A dilution test of Pst showed that the assay can detect 1.0 ng/μL Pst without a nucleic acid amplification procedure (Fig. ) (Welch’s t test: *P < 0.05). Neither extraction of nucleic acid from Pst infected wheat leaves nor mixing Pst RNAs with wheat leaf RNAs significantly affected the absorbance of the assay compared to only Pst RNAs (Supplementary Fig. ), and the presence of wheat RNA matrix did not compromise the sensitivity for detecting Pst (Supplementary Fig. ). Congeneric pathogenic fungi can cause similar infection symptoms and have been overlapping geographical distribution. We tested whether the assay could discriminate among three congeneric rust fungi, Pst, Puccinia graminis ( Pgt ) and Puccinia triticina ( Pt ). The ITS sequences of Pgt and Pt are similar to that of Pst (Fig. ). Three DProbes were designed to target the ITSs of Pst , Pgt and Pt . We mixed each set of RNAs extracted from these pathogenic fungi, and each could be detected using its corresponding DProbe without cross-interference. Besides rust fungi, Blumeria graminis ( Bgt ), Fusarium graminearum ( Fg ) and Rhizoctonia cerealis ( Rc ) responsible for yield reduction of wheat , they cause the diseases powdery mildew, fusarium head blight and wheat sharp eyespot, respectively. By aiming for the ITS sequence of the six fungi (Supplementary Table ), the assay allowed visual detection of each species (Supplementary Fig. ). To facilitate in-field detection, we integrated each assay reaction using a paper folding strategy termed origami. Wax printing was utilized to prepare the origami paper with defined detection spots via the formation of hydrophobic barrier . The procedures are shown in Supplementary Fig. . Reagents including DProbe, enzyme, and pH indicator were then loaded on each page of the paper (Fig. ). The folding procedure sequentially initiated the TMSD reaction and colour reporting reaction, and yielded a folded paper to be imaged by the camera on a smartphone. We developed an algorithm to locate the detection spot and defined a green pixel ratio (GPR) value to record the colour response of phenol red towards fungal RNAs. GPR value was defined as the portion of pixels with positive grey value in the green channel within the detection spot. The image processing method is illustrated in Supplementary Fig. . We found that the origami paper with a large pore size or surfactant modification yielded improved sensitivity, colour consistency and uniformity of the colorimetric response via the paper-folding strategy (Supplementary Fig. , Supplementary Fig. , Supplementary Note ). In addition, the robustness of the detection of Pst using the origami papers related to batches of origami papers and operators has been tested (Supplementary Fig. ). Colour calibration has been demonstrated to be able to improve the robustness of the detection using different smartphones (Supplementary Fig. ). By designing the origami paper with eight sample-loading sites, six pathogenic fungi ( Pst, Pgt, Pt, Bgt, Fg and Rc ) can be in parallel detected (Fig. ). The result showed that all six fungi could be distinguished without cross-interference using one origami paper loaded with their congenetic probes (Fig. ). The colorimetric assay using the origami papers yielded a comparable sensitivity compared to that proceeded in solution, and allowed to detect as low as 1 ng/μL Pst (Supplementary Fig. ). We further used the assay to detect fungal pathogens in infected wheat samples. The infection of Fusarium culmorum that causes fusarium crown rot can be indicated via the colour change of the origami papers (Supplementary Fig. ). To further evaluate the accuracy of the assay, thirty-two wheat leaf samples collected from different fields in China were tested (Supplementary Figs. , ). The samples were analysed in parallel using quantitative PCR (qPCR) and the colorimetric paper-based assay. The assay showed an agreement of 95.3% positive prediction, and 91.0% negative prediction when compared to qPCR in the test of the infection of Pst (Fig. , Supplementary Table , ). Besides pathogenic fungi, we explored the assay to detect a plant pathogenic bacterium, Pseudomonas syringae (Supplementary Fig. ) and a crop pathogenic virus, barley stripe mosaic virus (Supplementary Fig. ), and demonstrated that the assay is capable to diagnose the infection of not only fungi but also viruses and bacteria, showing the potential for the broad applicability for in-field detection of crop diseases. Early detection of crop infection can dramatically alleviate crop yield reduction and reduce fungicide use. We tested the assay for early diagnosis of Pst infection. Figure shows the observed phenotype of wheat leaves for during two weeks after infection. Spores are visible by the naked eye on day 10, demonstrating that wheat infection can be recognised after 10-day-infection by symptom observation. Pst in infected wheat leaves was stained using wheat germ agglutinin . Histological observation also showed the rapid spread of Pst in the leaves by10 days after inoculation. Nucleic acid was extracted from the infected leaves, and analysed by the colorimetric assay and qPCR. RNA extracted from the wheat samples on day 3 triggered a sharp drop of the absorbance signal in the colorimetric assay (Fig. ) (Welch’s t test: *P < 0.05). Pst infection can thus be identified after 3-day infection using the colorimetric assay. The qPCR result yielded positive detection of Pst from the 3-day-infected wheat samples (Welch’s t test: ** *P < 0.001) (Fig. , Supplementary Fig. ). It was estimated that 64.29 ng/μL Pst was present in the 3-day-infected wheat samples (Supplementary Fig. ). The paper-based strategy was further applied for parallel detection of Pst -infected wheat samples ( n = 6). Pst infection was ascertained after 3 days using the colorimetric paper (Fig. ) (Welch’s t test: **P < 0.01). The paper-based assay using optimized origami paper yielded an increased GPR value for testing Pst from the wheat sample infected for 3 days and a reduced variation in duplicate detection (Supplementary Fig. ), facilitating the robustness of positive detection of early infection by Pst . In addition, we inoculated wheat leaves with Pst in a dilution series. The colorimetric assay allowed to detect low-level Pst that did not cause observable infection symptom in the wheat leaves infected for 14 days (Supplementary Fig. ). Collectively, the colorimetric assay showed a capacity for early detection of Pst infection comparable to qPCR, and advanced the identification of an infection by 7 days compared to symptom observation. Disease cycles comprise four stages: dormancy, reproduction, dispersal and pathogenesis . Pathogens may be present on residues left in the field, in soil, and on weeds and tools. Importantly, only viable pathogenic fungi in dormancy will be activated in favourable conditions and enter the reproduction stage. Therefore, it is essential to devise strategies to distinguish viable fungi in total fungal samples to prevent disease circulation. Current DNA-targeting methods, such as qPCR, usually fail to distinguish viable pathogens from dead ones for the long-term persistence of DNA in dead microbial cells. RNA, however, is rapidly degraded in dead cells, and so, methods that target RNAs, such as the colorimetric assay, should specifically detect viable pathogens , . We therefore tested the feasibility of viable fungus detection using Pst mixtures containing viable and dead spores, in which the portion of viable spores was 0%, 0.1%, 1%, 10% and 100% (Fig. ). Nucleic acid was extracted from equal quantities of these Pst mixtures, and analysed using the colorimetric assay and qPCR. The results showed that absorbance signals attenuated gradually with the increase of viable fungi in both the absence and presence of wheat RNA matrix (Fig. , Supplementary Fig. ), indicating that the colorimetric assay can reliably indicate the amount of viable fungus. In contrast, there was no significant difference among Pst mixtures using qPCR (Fig. , Supplementary Fig. ). Pst mixtures with different proportions of viable spores were utilized to inoculate wheat. Due to the low abundance of Pst , the Ct values for qPCR measurement of Pst infection at 0 day were all close to 40 (Supplementary Fig. ). Fourteen days after the infection, leaves inoculated with Pst mixtures containing 0% and 0.1% viable fungi showed no obvious symptoms, but when the content of viable fungi increased to 1%, the leaves turned yellow and were loaded with observable spore piles (Fig. ). Measurements using qPCR and the colorimetric assay indicated that the amount of both total and viable Pst fungus in infected leaves increased in accordance with the proportion of viable fungus in the Pst mixture used for inoculation (Fig. and Supplementary Fig. ). This result indicated that the occurrence and severity of stripe rust were related specifically to the quantity of viable fungus, rather than the total quantity of all fungus. Detection of viable fungal pathogens is highly important, given that a large proportion of fungal cells may die through winter and summer. These fungi cannot cause effective crop infection. Because it can detect viable fungi, the colorimetric assay allows for a more precise prediction of disease occurrence and severity compared to methods that cannot distinguish viable pathogens from dead ones. We also investigated the defence response of wheat towards dead Pst . Dead spores slightly increased the expression of two pathogenesis-related genes, PR1 and PR2 (Supplementary Fig. ) , , but pre-inoculation with dead spores did not significantly change Pst biomass after 12, 24 or 48 h of infection with viable spores (Supplementary Fig. , Supplementary Note ). The result indicates that dead pathogenic fungi do not elicit a defence response that can efficiently inhibit pathogen infection. Fungicides are intensively used to prevent and treat crop diseases. The emergence of fungicide-resistance has become a severe issue. Identifying fungicide-resistant isolates can instruct us to choose an effective fungicide or other treatment strategies . Mutations leading to conformational changes in the drug target site are the main cause of fungicide-resistance in pathogenic fungi. CYP51, one of the cytochrome P450 monooxygenases, acts on fungal invasive growth, hypha formation and virulence. Inhibitors that target CYP51 serve as key antifungal agents , . The point mutation Y134F in CYP51 was found to be associated with a significant degree of triadimefon resistance in Pst isolates . Via competitive hybridization to hinder binding with non-target RNAs, an assay based on TMSD showed promise for detecting single-nucleotide mutations (SNMs) in RNAs. The principle of TMSD for identifying SNMs is illustrated in Supplementary Note . To maximize discrimination between the mutated RNA (F134 RNA) and the non-target wild type RNA (Y134 RNA), with a single-nucleotide difference, the TMSD reaction should be optimized by tuning the toehold length of the DProbe, which allows to block non-specific hybridization induced by the wild RNA while the DProbe/mutated RNA hybrid still forms. We designed DProbes targeting the Y134F mutation with a fixed 7-nt reverse toehold and forward toeholds ranging from 5 nt to 25 nt (Supplementary Fig. ). Fluorescence analysis of the TMSD reaction using a fluorophore and quencher-modified DProbe showed that DProbes with forward toehold lengths of 5, 9, 13, and 17 -nt can effectively distinguish F134 RNA from Y134 RNA. Long forward toeholds (21 nt and 25 nt) yielded a low discrimination capacity for the mutation because both F134 RNA and Y134 RNA can efficiently fuel the TMSD reaction. In contrast, shorter toehold length led to a lower displacement efficiency for both F134 RNA and Y134 RNA (Supplementary Fig. ). Electrophoretic analysis confirmed these results (Supplementary Fig. ). Based on the colorimetric reaction of the assay (Fig. ), the highest ratio of the absorbance of F134 RNA to that of Y134 RNA was achieved using the DProbe with a 17-nt forward toehold. Using this optimized DProbe, a dilution experiment with the mutated isolate indicated that, in either the absence or the presence of wheat RNA matrix, the assay could detect as little as 0.1% mutated isolate in a background of 99.9% wild type (Fig. , Supplementary Fig. ). The assay can thus detect low-abundance single-nucleotide mutations and should therefore be useful for finding rare fungicide-resistant isolates. We collected eight Pst -infected wheat samples from field sites in China (Fig. ). Using the above assay, we found that four isolates (XY-14, GZ-01, GY-07 and GY-09) from these samples were Y134F mutation-free, and four isolates (TS-05, TS-07, TS-02 and GD-04) had the Y134F mutation (Fig. ), which indicated the possibility of triadimefon resistance in the latter four isolates. Pst isolates from the eight wheat leaves were collected and sequenced. XY-14, GZ-01, GY-07 and GY-09 were homozygous wild type; and TS-05, TS-07, TS-02 and GD-04 were heterozygous Y134F mutants (Fig. ). Responses of eight Pst isolates to triadimefon showed that TS-05, TS-07, TS-02 and GD-04 had a much higher EC 50 towards triadimefon than other isolates (Fig. , Supplementary Fig. , and Supplementary Table ). These results demonstrate the feasibility of the colorimetric assay for identifying mutation-carrying fungicide-resistant isolates. To explore the ‘sample-to-answer’ in-field use of the assay, we integrated the assay with rapid nucleic acid extraction strategies, smartphone attachment and a mobile application program (app) (Fig. , Supplementary Fig. ). Nucleic acid extraction is a bottleneck for on-site or in-field nucleic acid tests, due to the complex setups involved that include centrifugation and chemical reagents. First, we explored a simple strategy using quartz sand to grind wheat leaf samples and release nucleic acid (Supplementary Fig. ), and found that the presence of Pst can be detected due to the colour change compared to samples without Pst . However, pigments in leaves can induce colour deviation in the papers, and consequently the quantification of GPR values was not stable. As an alternative, we then synthesized a microneedle (MN) patch for nucleic acid extraction (Supplementary Figs. , ). The MN patch, made of polyvinyl alcohol, has been used to extract DNA from leaf tissues . The patch can penetrate the wheat leaf and break cell walls to isolate nucleic acids. Via a rapid swelling through absorption of water molecules, the patch absorbs and concentrates nucleic acids on the needle tips. MN patch extraction yielded high-purity nucleic acid, but extracted less nucleic acid, due to the relatively low sample volume, than the Trizol-based method (Supplementary Fig. ). Nevertheless, the use of an MN patch accomplishes nucleic acid extraction within 1 min via a single press action; no instruments or chemical reagents are needed, thus facilitating nucleic acid tests for in-field diagnostics for crop diseases. By designing a smartphone attachment, the colorimetric detection of fungal RNAs could be completed by three simple pull operations, and the risk of contamination was avoided (Supplementary Fig. ). Furthermore, an app was designed for the graphical user guidance and the display of diagnostic results (Supplementary Fig. ). Notably, the app can also provide treatment advice based on the diagnostic results (Supplementary Fig. , Supplementary Table ), which is an essential feature for effective crop disease control. We have reported an in-field molecular diagnostic platform for crop diseases based on a cheap (~$0.30 per test), rapid (~10 min), multiplexed (6 pathogens or more) and mutation-resolved genetic assay. Compared to currently available genotypic methods, such as qPCR, the proposed assay is nucleic acid amplification-free and colour-readable, which simplifies operation and result readouts while eliminating the need for dedicated instruments, thus allowing in-field use. It is expected to be used by farmers, enabling them to detect crop infection at a very early stage by screening potential pathogens, and to obtain instructions about treatment protocols from smartphones. Crop disease diagnosis at the early stage can dramatically reduce fungicide use while alleviating crop yield reduction. Despite its lack of nucleic acid amplification, the assay is enough to achieve early detection of fungal infection. Using a metal ion-mediated urease catalysis reaction, the recognition of fungal RNAs is amplified to become a colour-readable signal, yielding a limit of detection of 1.0 ng/μL Pst . Due to this high sensitivity, the assay allowed to identify 3-day infection by Pst , advancing the identification by 7 days compared to symptom observation, by 3 days compared to hyperspectral imaging , and to a period comparable to the sensitive genotypic method, qPCR, for early detection of Pst infection. Besides, the diagnostic accuracy for wheat stripe rust using the colorimetric assay was an improvement of about ~6% compared to the reported hyperspectral imaging technique . Identification of the specific crop pathogen is critical for choosing which fungicide to use, but is practically unachievable by phenotypic methods such as foliage machine vision, or by hyperspectral or thermographic imaging approaches. Screening pathogens requires a multiplexed assay that can distinguish the genotypic, immunological or chemical differences among pathogens and encompass the detection of potential pathogens. The colorimetric paper was patterned for multiplexed detection of fungal RNAs, and we demonstrated the parallel detection of six prominent fungi that can infect wheat ( Pst, Pgt, Pt, Bgt, Fg and Rc ). The multiplexing capacity of the assay can be expanded by printing more reagent loading sites. In particular, given that the DProbe can be reprogrammed to recognise specific fungal RNAs, the assay can, in principle, diagnose any crop pathogen of interest, thus offers a programmable multiplexing capacity to cover all potential pathogens that infect the cultivated crop. Fungal pathogens can reside in soil and on weeds, but many of them cannot survive through summer and winter. Only viable ones will be activated in favourable conditions, to infect crops and cause plant diseases. Therefore, detection of viable pathogens is important for estimating infection and disease risk. By detecting fungal RNAs, the proposed assay allows to differentiate viable and dead fungi. In contrast, qPCR cannot distinguish dead from viable fungi. We showed that an increase in the proportion of viable Pst led to a more severe symptom of wheat stripe rust, indicating that the occurrence and severity of stripe rust correlates with the quantity of viable Pst , rather than total Pst . Although disease occurrence cannot be determined by pathogen presence alone, because the host plant and environment are also key factors , the detection of viable pathogens that are active, rather than pathogens, that are dead, should increase the prediction accuracy of disease occurrence and severity. For example, the estimated abundance of viable fungi in soil obtained via the assay may be highly valuable for predicting the occurrence of soil-borne crop diseases. Knowing the fungicide-resistance of fungal pathogens can further help us to choose the right fungicide or other intervention strategies. This is exemplified by the identification of antibiotic-resistant infectious diseases in humans, which contributes to reducing the mortality and morbidity rate of nosocomial infections . Drug resistance is usually caused by genetic mutations. Only molecular methods are potentially applicable for discriminating the subtle genetic difference involved. However, PCR lacks the sequence specificity to identify resistance caused by point mutations, while sequencing techniques are currently impractical in the field, and are also costly. By using a strand displacement reaction to identify fungal RNAs, the proposed assay allows to discriminate single-nucleotide mutations. We demonstrated its identification of a single-nucleotide mutation, Y134F, that is associated with resistance to the fungicide triadimefon. Moreover, the assay could detect down to 0.1% mutated Pst in an otherwise wild-type population, indicating its capacity to find rare fungicide-resistant pathogens. In particular, fungal pathogen screening, viable pathogen detection and drug-resistance identification are now achievable in-field by simply using a low-cost test paper and a smartphone, greatly advancing plant disease diagnostics. Nucleic acid extraction is simplified using an MN patch, allowing this step to be completed within 1 min. Simple pull operations allow the users to accomplish the colorimetric detection of fungal RNAs by means of a 3D-printed smartphone attachment, and the sample-to-result test for wheat diseases can be finished in 10 min in the field. The smartphone app can recommend intervention strategies based on the diagnostic results. Therefore, the end-users without any knowledge about phytopathology can be reliably guided to precisely treat the crop infection. Besides, the colorimetric assay only requires cheap reagents such as urease and non-chemically labelled DNA probes, yielding an inexpensive test covering six wheat fungal pathogens (estimated to be US $0.30). The low cost of tests and wide availability of smartphones should render the assay to be an abundantly available, regularly applicable in-field diagnostic tool for crop diseases. In comparison, a nanopore sequencing-based method, termed MARPLE has been utilized for high-throughput and point-of-care detection of strain-level fungal pathogens and fungicide resistance with single-nucleotide resolution, but involves the nucleic acid amplification process, the use of PCR instrument and MinION sequencer , thus increasing the assaying time (to be 48 h), complexity and the costs for the tests. We demonstrate a proof-of-concept for the in-field diagnosis of fungal pathogens and their fungicide resistance, the detection performance of the assay could be further improved. First, the colorimetric paper-based assay has been used for the qualitative and semi-quantitative measurement of different pathogens, yet it is, currently, not feasible for quantitative detection. Paper-based colorimetric readout yields colour nonuniformity due to the variation of reagent diffusion , and in most cases, it has not been used for target quantification. We explored the optimization and modification of paper substrates to alleviate the data nonuniformity. Besides, the development of image algorithms for colour calibration or rescaling and the use of deep learning approach show the promise to fuel paper-based colorimetric assays to be a quantitative assay. Second, although the assay has been used for determining the resistance of Pst towards triadimefon via detecting genetic mutations, gene mutation markers that can indicate drug resistance are still lacking, particularly for crop pathogens . Thus, the assay can currently identify only a very small proportion of fungicide-resistant fungi, but this coverage will be improved as new drug-resistance genetic markers are discovered. With the development of new fungicides, as well as the emergence of new resistant fungi, the tools to identify fungicide resistance information will become increasingly important for precision plant disease control. In addition, microneedle-based nucleic acid extraction is rapid and simple, but has been only tested with leaf samples, a tough microneedle will be needed to extract nucleic acids from the samples with relatively hard surfaces such as roots. The work advances molecular diagnostics for crop diseases via achieving in-field nucleic acid tests with single-nucleotide resolution. The application of the in-field diagnostic tool covering the detection of pathogenic fungi, viruses and bacteria is of potential to facilitate efficient crop disease management, reduce the use of pesticides, and contribute to sustainable agriculture via alleviating crop diseases in precise plant diagnostics. Oligonucleotides DNA oligonucleotides (Supplementary Table – , ) were ordered from Sangon (Shanghai, China). 6-carboxyfluorescein and Black Hole Quencher 1 modified DNA oligonucleotides, RNA oligonucleotides were purchased from Takara (Beijing, China). RNA oligonucleotides and chemically modified DNA oligonucleotides were purified via HPLC. DNA oligonucleotides were purified by PAGE. DNA and RNA oligonucleotides were dissolved in molecular biology grade water (cat. no. 46-000-CM, Corning) to prepare stock solutions with a concentration of 100 μM. DProbe preparation DProbes were synthesized by annealing 5 μL cis strand (3 μM), 5 μL trans strand (3 μM) and 5 μL AgNO 3 (3 μM) in 15 μL NaNO 3 (1 M) at 90 °C for 5 min, followed by incubation at 25 °C for 25 min, and kept at 4 °C prior to use. Labelled DProbes were used to verify the formation of the TMSD reaction, and the fluorescence was measured using a Synergy H1 microplate reader and analyzed using Gen5 CHS 3.08 system. The excitation wavelength was 480 nm, the emission wavelength was 520 nm. Pathogen isolates Puccinia striiformis ( Pst ) (CYR32), Puccinia triticina (Pt) (XJ-7), Puccinia graminis ( Pgt ) (HQM), Blumeria graminis ( Bgt ) (E09) and Fusarium culmorum ( F. culmorum ) were provided by Dr. Jie Zhao, College of Plant Protection of Northwest A&F University (Yangling, China). Rhizoctonia cerealis ( Rc ) (R0301) was provided by Dr. Li Wei, Institute of Plant Protection, Jiangsu Academy of Agricultural Sciences (Nanjing, China). Pseudomonas syringae pv. tomato DC3000 ( P. syringae pv. tomato DC3000) was kindly provided by Dr. Honghong Wu, Huazhong Agricultural University, (Wuhan, China). Barley stripe mosaic virus (BSMV) was provided by Dr. Qiang Xu, Sichuan Agricultural University (Chengdu, China). Fusarium graminearum ( Fg ) (CICC 2697) was purchased from the China Centre of Industrial Culture Collection. Wheat cultivation and fungal inoculation For Pst inoculation, wheat cultivar Suwon 11 was first grown in the greenhouse under 16 °C with 16-h light and 10 °C with 8-h dark. Once the wheat had grown to the two-leaf stage, the second leaf was inoculated with Pst spores and the plant was placed at 10 °C under saturated humidity with 24-h dark. The plant was then returned to normal conditions (16 °C with 16-h light and 10 °C with 8-h dark). A soil mixing method was used for F. culmorum inoculation . The mycelium blocks were placed on a millet culture medium and incubated at 25 °C for five days. When the surface of the millet was covered with mycelium, the cultivation was stopped and dried for later use. Sterile soil and millet culture medium were mixed evenly in a ratio of 125:1 to produce fungal soil. The mixture was transferred in plastic flower pots containing 200 g of fungal soil, plant ten healthy wheat seeds, and located in a greenhouse. The greenhouse temperature should be controlled at 25 ± 2 °C during the day and 20 ± 2 °C at night under natural light. Wheat samples were collected after 30-day infection for analysis. Barley cultivation and BMSV inoculation BMSV-negative wheat was grown in an environmental chamber at 16–14 °C with a 16 h/light: 8 h/dark photoperiods. When the plants had grown to the two to three-leaf stage, BSMV were inoculated onto the second leaves from the bottom of wheat . The seedlings were moisturized at saturated humidity for 24 h without light and then placed in an incubator at 25 °C for nine days before the phenotype was manifestation . Arabidopsis cultivation and P. syringae pv. tomato DC3000 inoculation Arabidopsis seeds were sown in soil mix and grown in growth chamber with the following settings: 22 °C for 10-h light and 20 °C for 14-h dark, 150 μmol m −2 s −1 light intensity, 70% relative humidity. Then, 4-weeks old Arabidopsis plants were infiltrated with solution containing P. syringae pv. tomato DC3000 (OD 600 0.001). After 3-h inoculation under room light condition, plants were transferred back to the growth chamber. Arabidopsis leaves were collected for test after another 3-day infection. Nucleic acid extraction Fungi were first placed in 2-mL microcentrifuge tubes, frozen in liquid nitrogen, and powdered using a SCIENTZ-48 tissue grinder (Ningbo Scientz Biotechnology). DNA was extracted using an Ezup Column Fungi Genomic DNA Purification Kit (cat. no. B518259-0050, Sangon Biotech). Briefly, 200 μL Buffer Digestion, 2 μL β-Mercaptoethanol and 20 μL Proteinase K were added and incubated at 56 °C for 1 h to lyse the cells. Subsequently, 100 μL Buffer PF was added and stored at −20 °C for 5 min, followed by centrifuging at 12,000 × g for 5 min at 4 °C. The supernatant was transferred to a new adsorption column, mixed with 200 μL Buffer BD, 200 μL ethyl alcohol, and centrifuged at 9500 × g for 1 min. Then, 500 μL PW solution and 500 μL Wash solution were added in turn, followed by centrifuging at 12,000 × g for 2 min at 4 °C. Finally, DNA was resuspended in 50 μL H 2 O and used for qPCR. RNA was extracted using TRIzol Reagent (cat. no.15596018, Thermo Fisher Scientific). 1 mL of which was added into the ground fungi and incubated at room temperature for 5 min. Then, 200 μL chloroform was added and vibrated for 15 s, followed by centrifuging at 12,000 × g for 15 min at 4 °C. The aqueous phase was transferred to 1.5 mL microcentrifuge tube. Next, 500 μL isoamyl alcohol was added in the tube, which was incubated at room temperature for 10 min and centrifuged at 12,000 × g for 5 min at 4 °C. The pellet was washed using 1 mL 70% ethanol, and suspended in 50 μL H 2 O. For the plant samples, leaves or roots (cut to be a diameter of about 3–4 cm) were frozen and powdered, and nucleic acid was extracted based on the protocols used for fungal samples. Rapid nucleic acid extraction using quartz sand Leaf samples (3–4 cm) were placed in 2-mL microcentrifuge tubes with 0.5 g quartz sand (8–16 meshes). The samples were ground using a plastic pestle in the presence of 100 μL extraction buffer (1 μL TCEP (100 mM), 1 μL Tris (2-carboxyethyl) phosphine (0.5 mM), 1 μL RNA Carrier and 97 μL H 2 O) for 5 min. Rapid nucleic acid extraction using MN patch MN was synthesized according to a published protocol . Briefly, an MN mold was first cleared in an ultrasonic bath for 5 min. Next, 1 mL 10% polyvinyl alcohol (PVA) solution was added to the mold, which was placed in a sealed vacuum chamber (600 mm Hg) for 20 min to draw the PVA solution into the needle cavities. The mold was then kept in the vacuum at 25 °C for 24 h. The MN patch contained an 11 × 11 microneedle array. The height and base of the needle were 600 µm and 300 µm, respectively. The spacing tip to tip was 600 µm. For the MN patch-based nucleic acid extraction, and a MN patch was placed on a leaf and pressed gently by hand for 10 s. The MN patch was peeled off and collected in 50 μL H 2 O for further analysis. Pathogenic RNA detection in solution A 2.5-μL sample was added to 10 μL prepared DProbe, and incubated at room temperature for 10 min. Then, 10 μL urease (10 nM, cat. no. U0017, EC 2.5.1.5; TCI, Tokyo, Japan), and 77.5 μL colorimetric mixture (10 μL phenol red (2.5 mM) ordered from Innochem (Beijing, China), 10 μL urea (5M, CON 2 H 4 ) and 55.5 μL H 2 O) were added. The mixture was measured by a Synergy H1 microplate reader to record the absorbance at 560 nm, and was also photographed by a HUAWEI P40 smartphone. Preparation of origami papers Filter paper was cut into A4 size (297 cm × 210 cm). A wax printer (Xerox ColourQube 8580/8880 N) was utilized to print the filter paper. The printed paper was heated at 170 °C for 10 s using a hot plate to melt the wax. The melted wax diffused through the paper, thus form hydrophobic pattern of hydrophilic reaction regions (3 mm in diameter). The reagents containing DProbe (5 μL, 100 nM), urease (10 nM) in 6 nM pullulan ((cat. no. A60187) purchased from Innochem (Beijing, China)), and colorimetric mixture (5 μL (1 μL phenol red (1.25 mM), 1 μL urea (2.5 M) and 3 μL H 2 O)) were dropped on each page of the printed paper to finish the preparation of the origami paper. Origami paper-based detection A 5-μL sample was loaded on Page 1 through the upper window of a smartphone attachment. The origami paper was then moved to the bottom of the chamber. Folding processes for the origami paper are achieved via control sticks (Page 2 to Page 1, Page 3 to Page 2, and Page 4 to Page 3). Origami papers were photographed via a smartphone, and the images were analysed using a smartphone app (termed ifDiag ). Histological observation and biomass measurement of Pst in infected wheat samples First, the inoculated wheat leaf was decolourized and made transparent using 95% ethanol and chloral hydrate . The samples were then treated in KOH (1 M) at 121 °C for 5 min and stained with 20 μg/mL wheat germ agglutinin Alexa-488 (Invitrogen) for 15 min. The samples were observed by fluorescence microscopy (Olympus BX63, excitation wavelength 450–480 nm, emission wavelength 515 nm). Fungal biomass measurement was based on qPCR, using total DNA extracted from infected wheat leaves at different day post inoculation. The ratio of total fungal DNA to total wheat DNA was assessed by normalizing the data to the wheat gene TaEF-1α and the Pst gene PstEF1 . The primers used for detecting TaEF-1α and PstEF1 are listed in Supplementary Table . Gel electrophoresis Gel electrophoresis was carried using 5% agarose gel stained with 1 × Gelred (Biotium, cat. no. 41001), and then was performed in 1 × TAE buffer at 150 V for about 60 min. After electrophoresis, the gel was visualized via the Gel Doc XR + system (BioRad, USA). qPCR and RT-qPCR detection qPCR was performed to analyse fungal DNA samples using the Platinum SYBR Green qPCR SuperMix-UDG (cat. no. 11744100, Thermo) on the CFX96 Thermal Cyclers (Bio-Rad). qPCR conditions were as follows: 50 °C for 5 min, 94 °C for 5 min, and then 45 cycles of 94 °C for 15 s, 60 °C for 15 s and 72 °C for 45 s. The qPCR mixture contained 2.5 μL fungal DNA sample, 10 μL of SYBR Green qPCR Supermix, 2 μL primers (1 μL each of 10 mM forward and reverse primer) and 5.5 μL H 2 O. Expression levels of pathogenesis-related genes were measured using RT-qPCR. Total RNA (2 μg) was used for reverse transcription with a RevertAid First Strand cDNA Synthesis Kit (MNI, K1622). Diluted cDNA (1:5; 2 μL) was utilized in the following qPCR procedure. The expression levels of all tested genes were normalized to PstEF1. qPCR primers are listed in Supplementary Table . Statistics and reproducibility Figures were created using Origin 2019. Two-sided Welch’s t test was used for all statistical comparisons and calculated by SPSS 25. The reproducibility of the results was assessed using a minimum of three independent experiments except Figs. c, , Supplementary Fig. , and Supplementary Fig. , which were conducted once. “n” in the legend indicates the number of independent experiments. Receiver operating characteristic (ROC) curves were generated using GraphPad Prism 8. Reporting summary Further information on research design is available in the linked to this article. DNA oligonucleotides (Supplementary Table – , ) were ordered from Sangon (Shanghai, China). 6-carboxyfluorescein and Black Hole Quencher 1 modified DNA oligonucleotides, RNA oligonucleotides were purchased from Takara (Beijing, China). RNA oligonucleotides and chemically modified DNA oligonucleotides were purified via HPLC. DNA oligonucleotides were purified by PAGE. DNA and RNA oligonucleotides were dissolved in molecular biology grade water (cat. no. 46-000-CM, Corning) to prepare stock solutions with a concentration of 100 μM. DProbes were synthesized by annealing 5 μL cis strand (3 μM), 5 μL trans strand (3 μM) and 5 μL AgNO 3 (3 μM) in 15 μL NaNO 3 (1 M) at 90 °C for 5 min, followed by incubation at 25 °C for 25 min, and kept at 4 °C prior to use. Labelled DProbes were used to verify the formation of the TMSD reaction, and the fluorescence was measured using a Synergy H1 microplate reader and analyzed using Gen5 CHS 3.08 system. The excitation wavelength was 480 nm, the emission wavelength was 520 nm. Puccinia striiformis ( Pst ) (CYR32), Puccinia triticina (Pt) (XJ-7), Puccinia graminis ( Pgt ) (HQM), Blumeria graminis ( Bgt ) (E09) and Fusarium culmorum ( F. culmorum ) were provided by Dr. Jie Zhao, College of Plant Protection of Northwest A&F University (Yangling, China). Rhizoctonia cerealis ( Rc ) (R0301) was provided by Dr. Li Wei, Institute of Plant Protection, Jiangsu Academy of Agricultural Sciences (Nanjing, China). Pseudomonas syringae pv. tomato DC3000 ( P. syringae pv. tomato DC3000) was kindly provided by Dr. Honghong Wu, Huazhong Agricultural University, (Wuhan, China). Barley stripe mosaic virus (BSMV) was provided by Dr. Qiang Xu, Sichuan Agricultural University (Chengdu, China). Fusarium graminearum ( Fg ) (CICC 2697) was purchased from the China Centre of Industrial Culture Collection. For Pst inoculation, wheat cultivar Suwon 11 was first grown in the greenhouse under 16 °C with 16-h light and 10 °C with 8-h dark. Once the wheat had grown to the two-leaf stage, the second leaf was inoculated with Pst spores and the plant was placed at 10 °C under saturated humidity with 24-h dark. The plant was then returned to normal conditions (16 °C with 16-h light and 10 °C with 8-h dark). A soil mixing method was used for F. culmorum inoculation . The mycelium blocks were placed on a millet culture medium and incubated at 25 °C for five days. When the surface of the millet was covered with mycelium, the cultivation was stopped and dried for later use. Sterile soil and millet culture medium were mixed evenly in a ratio of 125:1 to produce fungal soil. The mixture was transferred in plastic flower pots containing 200 g of fungal soil, plant ten healthy wheat seeds, and located in a greenhouse. The greenhouse temperature should be controlled at 25 ± 2 °C during the day and 20 ± 2 °C at night under natural light. Wheat samples were collected after 30-day infection for analysis. BMSV-negative wheat was grown in an environmental chamber at 16–14 °C with a 16 h/light: 8 h/dark photoperiods. When the plants had grown to the two to three-leaf stage, BSMV were inoculated onto the second leaves from the bottom of wheat . The seedlings were moisturized at saturated humidity for 24 h without light and then placed in an incubator at 25 °C for nine days before the phenotype was manifestation . P. syringae pv. tomato DC3000 inoculation Arabidopsis seeds were sown in soil mix and grown in growth chamber with the following settings: 22 °C for 10-h light and 20 °C for 14-h dark, 150 μmol m −2 s −1 light intensity, 70% relative humidity. Then, 4-weeks old Arabidopsis plants were infiltrated with solution containing P. syringae pv. tomato DC3000 (OD 600 0.001). After 3-h inoculation under room light condition, plants were transferred back to the growth chamber. Arabidopsis leaves were collected for test after another 3-day infection. Fungi were first placed in 2-mL microcentrifuge tubes, frozen in liquid nitrogen, and powdered using a SCIENTZ-48 tissue grinder (Ningbo Scientz Biotechnology). DNA was extracted using an Ezup Column Fungi Genomic DNA Purification Kit (cat. no. B518259-0050, Sangon Biotech). Briefly, 200 μL Buffer Digestion, 2 μL β-Mercaptoethanol and 20 μL Proteinase K were added and incubated at 56 °C for 1 h to lyse the cells. Subsequently, 100 μL Buffer PF was added and stored at −20 °C for 5 min, followed by centrifuging at 12,000 × g for 5 min at 4 °C. The supernatant was transferred to a new adsorption column, mixed with 200 μL Buffer BD, 200 μL ethyl alcohol, and centrifuged at 9500 × g for 1 min. Then, 500 μL PW solution and 500 μL Wash solution were added in turn, followed by centrifuging at 12,000 × g for 2 min at 4 °C. Finally, DNA was resuspended in 50 μL H 2 O and used for qPCR. RNA was extracted using TRIzol Reagent (cat. no.15596018, Thermo Fisher Scientific). 1 mL of which was added into the ground fungi and incubated at room temperature for 5 min. Then, 200 μL chloroform was added and vibrated for 15 s, followed by centrifuging at 12,000 × g for 15 min at 4 °C. The aqueous phase was transferred to 1.5 mL microcentrifuge tube. Next, 500 μL isoamyl alcohol was added in the tube, which was incubated at room temperature for 10 min and centrifuged at 12,000 × g for 5 min at 4 °C. The pellet was washed using 1 mL 70% ethanol, and suspended in 50 μL H 2 O. For the plant samples, leaves or roots (cut to be a diameter of about 3–4 cm) were frozen and powdered, and nucleic acid was extracted based on the protocols used for fungal samples. Leaf samples (3–4 cm) were placed in 2-mL microcentrifuge tubes with 0.5 g quartz sand (8–16 meshes). The samples were ground using a plastic pestle in the presence of 100 μL extraction buffer (1 μL TCEP (100 mM), 1 μL Tris (2-carboxyethyl) phosphine (0.5 mM), 1 μL RNA Carrier and 97 μL H 2 O) for 5 min. MN was synthesized according to a published protocol . Briefly, an MN mold was first cleared in an ultrasonic bath for 5 min. Next, 1 mL 10% polyvinyl alcohol (PVA) solution was added to the mold, which was placed in a sealed vacuum chamber (600 mm Hg) for 20 min to draw the PVA solution into the needle cavities. The mold was then kept in the vacuum at 25 °C for 24 h. The MN patch contained an 11 × 11 microneedle array. The height and base of the needle were 600 µm and 300 µm, respectively. The spacing tip to tip was 600 µm. For the MN patch-based nucleic acid extraction, and a MN patch was placed on a leaf and pressed gently by hand for 10 s. The MN patch was peeled off and collected in 50 μL H 2 O for further analysis. A 2.5-μL sample was added to 10 μL prepared DProbe, and incubated at room temperature for 10 min. Then, 10 μL urease (10 nM, cat. no. U0017, EC 2.5.1.5; TCI, Tokyo, Japan), and 77.5 μL colorimetric mixture (10 μL phenol red (2.5 mM) ordered from Innochem (Beijing, China), 10 μL urea (5M, CON 2 H 4 ) and 55.5 μL H 2 O) were added. The mixture was measured by a Synergy H1 microplate reader to record the absorbance at 560 nm, and was also photographed by a HUAWEI P40 smartphone. Filter paper was cut into A4 size (297 cm × 210 cm). A wax printer (Xerox ColourQube 8580/8880 N) was utilized to print the filter paper. The printed paper was heated at 170 °C for 10 s using a hot plate to melt the wax. The melted wax diffused through the paper, thus form hydrophobic pattern of hydrophilic reaction regions (3 mm in diameter). The reagents containing DProbe (5 μL, 100 nM), urease (10 nM) in 6 nM pullulan ((cat. no. A60187) purchased from Innochem (Beijing, China)), and colorimetric mixture (5 μL (1 μL phenol red (1.25 mM), 1 μL urea (2.5 M) and 3 μL H 2 O)) were dropped on each page of the printed paper to finish the preparation of the origami paper. A 5-μL sample was loaded on Page 1 through the upper window of a smartphone attachment. The origami paper was then moved to the bottom of the chamber. Folding processes for the origami paper are achieved via control sticks (Page 2 to Page 1, Page 3 to Page 2, and Page 4 to Page 3). Origami papers were photographed via a smartphone, and the images were analysed using a smartphone app (termed ifDiag ). Pst in infected wheat samples First, the inoculated wheat leaf was decolourized and made transparent using 95% ethanol and chloral hydrate . The samples were then treated in KOH (1 M) at 121 °C for 5 min and stained with 20 μg/mL wheat germ agglutinin Alexa-488 (Invitrogen) for 15 min. The samples were observed by fluorescence microscopy (Olympus BX63, excitation wavelength 450–480 nm, emission wavelength 515 nm). Fungal biomass measurement was based on qPCR, using total DNA extracted from infected wheat leaves at different day post inoculation. The ratio of total fungal DNA to total wheat DNA was assessed by normalizing the data to the wheat gene TaEF-1α and the Pst gene PstEF1 . The primers used for detecting TaEF-1α and PstEF1 are listed in Supplementary Table . Gel electrophoresis was carried using 5% agarose gel stained with 1 × Gelred (Biotium, cat. no. 41001), and then was performed in 1 × TAE buffer at 150 V for about 60 min. After electrophoresis, the gel was visualized via the Gel Doc XR + system (BioRad, USA). qPCR was performed to analyse fungal DNA samples using the Platinum SYBR Green qPCR SuperMix-UDG (cat. no. 11744100, Thermo) on the CFX96 Thermal Cyclers (Bio-Rad). qPCR conditions were as follows: 50 °C for 5 min, 94 °C for 5 min, and then 45 cycles of 94 °C for 15 s, 60 °C for 15 s and 72 °C for 45 s. The qPCR mixture contained 2.5 μL fungal DNA sample, 10 μL of SYBR Green qPCR Supermix, 2 μL primers (1 μL each of 10 mM forward and reverse primer) and 5.5 μL H 2 O. Expression levels of pathogenesis-related genes were measured using RT-qPCR. Total RNA (2 μg) was used for reverse transcription with a RevertAid First Strand cDNA Synthesis Kit (MNI, K1622). Diluted cDNA (1:5; 2 μL) was utilized in the following qPCR procedure. The expression levels of all tested genes were normalized to PstEF1. qPCR primers are listed in Supplementary Table . Figures were created using Origin 2019. Two-sided Welch’s t test was used for all statistical comparisons and calculated by SPSS 25. The reproducibility of the results was assessed using a minimum of three independent experiments except Figs. c, , Supplementary Fig. , and Supplementary Fig. , which were conducted once. “n” in the legend indicates the number of independent experiments. Receiver operating characteristic (ROC) curves were generated using GraphPad Prism 8. Further information on research design is available in the linked to this article. Supplementary Information Reporting Summary Source Data |
Systematic analysis of levels of evidence supporting American Academy of Ophthalmology Preferred Practice Pattern guidelines, 2012–2021 | 6f575744-d323-4f7c-a1d0-c58c291c716e | 10067168 | Ophthalmology[mh] | Over the past four decades, there has been an increased emphasis on evidence-based medicine across all medical specialties, particularly in the development of clinical practice guidelines. However, prior analyses of clinical guidelines in cardiology suggest that randomized controlled trials , typically considered the highest level of evidence, are not cited in a majority of clinical guidelines developed by key professional societies . Additionally, the proportion of recommendations citing the highest level of evidence has not increased significantly over time . Prior studies in ophthalmology have evaluated the types of evidence published in ophthalmology journals ; however, there is limited literature analyzing the levels of evidence present in ophthalmology clinical guidelines. The American Academy of Ophthalmology Preferred Practice Patterns guidelines designate recommended diagnostic and treatment approaches for various ophthalmic conditions and are typically revised every five years . The purpose of this systematic analysis of the American Academy of Ophthalmology Preferred Practice Pattern guidelines was to understand the evidence behind current guidelines, assess changes over time in the levels of evidence used to generate recommendations, and compare levels of evidence utilized across guidelines from different ophthalmology subspecialties. Review of Guidelines Current American Academy of Ophthalmology (AAO) Preferred Practice Pattern (PPP) guidelines were identified as those posted on the AAO website ( https://www.aao.org/guidelines-browse ) as of March 20, 2022. Only full-text PPP guidelines documents were included. Summary Benchmarks and PPP Clinical Questions were not included. Since PPP guidelines are typically valid for five years, the immediate predecessors of current guidelines were identified to assess changes over time. Prior guidelines were either identified on the AAO website ( https://www.aaojournal.org/content/preferred-practice-pattern ) or requested from the AAO if they were issued prior to 2015 . No human subjects, human-derived materials, or human medical records were involved in this study to necessitate review by an Institutional Review Board. The guidelines report levels of evidence (LOE) based on the Scottish Intercollegiate Guidelines Network (SIGN) scale (Table ) . The guidelines also report quality of evidence and strengths of recommendation defined by the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) scale (Table ) . Current PPPs report SIGN and GRADE ratings throughout the PPP main texts. Prior PPPs report the ratings in a centralized appendix. For each guideline, recommendations with reported LOE were abstracted by one of two reviewers (either A.S. or J.B.L.). Any statement with a reported LOE was considered a recommendation. Further details of LOE reporting are presented in Table . For each recommendation with LOE reported, we recorded the LOE, the quality of evidence, and the recommendation strength. Each recommendation was categorized as a recommendation for diagnosis, management, or both. We also recorded the subspecialty associated with each PPP guideline following the subspecialty categories listed on the AAO PPP website (cataract/anterior segment, comprehensive ophthalmology, cornea/external disease, glaucoma, neuro-ophthalmology, ocular pathology/oncology, oculoplastics/orbit, pediatric ophthalmology/strabismus, refractive management/intervention, retina/vitreous, and uveitis). An additional comprehensive review of all sentences in the Care Process section, where recommendations are typically located, of the 2021 edition of the Cataract in the Adult Eye PPP was performed . A single reviewer (A.S.) abstracted all sentences from the document. Two reviewers (A.S. and J.B.L) independently determined whether each sentence constituted a recommendation statement—a sentence was considered a recommendation statement if it addressed how patients should be diagnosed or managed clinically. For each recommendation statement, the two reviewers examined all references in the document associated with the statement to determine their LOE based on the SIGN scale and recorded the highest LOE (LOE I > LOE II > LOE III). When disagreement existed, the two reviewers had a discussion to reach a consensus decision. Further, a third reviewer (A.N.K.), a board-certified anterior segment ophthalmologist and cataract surgeon, validated all the findings. All current PPP guidelines were also reviewed to determine whether or not cost-effectiveness or cost/value factors were explicitly mentioned as part of the justification for each recommendation. Additional review of each guideline was performed to determine the presence of cost/value statements, broadly defined as any statement in which cost or value was mentioned, and whether such statements were used to: 1) report a gap in cost/value evidence; 2) highlight economic impact of disease or care; and 3) advocate for cost/value-related issues, consistent with a framework previously described in the cardiology literature . Data analysis We calculated the total number of recommendations with LOE reported for each current and prior PPP and calculated the change in number of recommendations over time. The median number of recommendations per guideline was determined. The number of recommendations with reported LOE were also summarized by subspecialty and care process category (diagnosis, management, or both). Additionally, the numbers and proportions of recommendations classified as LOE I, II, and III among all current and prior PPPs were determined. To further assess differences across subspecialties, we compared the number and proportion of LOE I, II, and III recommendations in current PPPs with those in prior PPPs by subspecialty. We also reported the quality of evidence according to the GRADE scale, stratified by LOE. For the current Cataract in the Adult Eye PPP, agreement between the two reviewers for whether a sentence constituted a recommendation statement was measured by percent agreement and kappa statistics. If references were provided in the document, the numbers and proportions of validated recommendation statements that were classified as LOE I, II, and III were determined. If recommendation statements also had LOE explicitly reported in the document, we compared the reported LOE with the study team-determined LOE. To evaluate the role of cost/value in PPPs, the proportion of current PPPs that had any recommendation supported by cost-effectiveness or cost/value considerations was determined. Further analysis was done to assess the proportions of current PPPs that contain statements addressing each of the specific areas related to cost/value considerations. Current American Academy of Ophthalmology (AAO) Preferred Practice Pattern (PPP) guidelines were identified as those posted on the AAO website ( https://www.aao.org/guidelines-browse ) as of March 20, 2022. Only full-text PPP guidelines documents were included. Summary Benchmarks and PPP Clinical Questions were not included. Since PPP guidelines are typically valid for five years, the immediate predecessors of current guidelines were identified to assess changes over time. Prior guidelines were either identified on the AAO website ( https://www.aaojournal.org/content/preferred-practice-pattern ) or requested from the AAO if they were issued prior to 2015 . No human subjects, human-derived materials, or human medical records were involved in this study to necessitate review by an Institutional Review Board. The guidelines report levels of evidence (LOE) based on the Scottish Intercollegiate Guidelines Network (SIGN) scale (Table ) . The guidelines also report quality of evidence and strengths of recommendation defined by the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) scale (Table ) . Current PPPs report SIGN and GRADE ratings throughout the PPP main texts. Prior PPPs report the ratings in a centralized appendix. For each guideline, recommendations with reported LOE were abstracted by one of two reviewers (either A.S. or J.B.L.). Any statement with a reported LOE was considered a recommendation. Further details of LOE reporting are presented in Table . For each recommendation with LOE reported, we recorded the LOE, the quality of evidence, and the recommendation strength. Each recommendation was categorized as a recommendation for diagnosis, management, or both. We also recorded the subspecialty associated with each PPP guideline following the subspecialty categories listed on the AAO PPP website (cataract/anterior segment, comprehensive ophthalmology, cornea/external disease, glaucoma, neuro-ophthalmology, ocular pathology/oncology, oculoplastics/orbit, pediatric ophthalmology/strabismus, refractive management/intervention, retina/vitreous, and uveitis). An additional comprehensive review of all sentences in the Care Process section, where recommendations are typically located, of the 2021 edition of the Cataract in the Adult Eye PPP was performed . A single reviewer (A.S.) abstracted all sentences from the document. Two reviewers (A.S. and J.B.L) independently determined whether each sentence constituted a recommendation statement—a sentence was considered a recommendation statement if it addressed how patients should be diagnosed or managed clinically. For each recommendation statement, the two reviewers examined all references in the document associated with the statement to determine their LOE based on the SIGN scale and recorded the highest LOE (LOE I > LOE II > LOE III). When disagreement existed, the two reviewers had a discussion to reach a consensus decision. Further, a third reviewer (A.N.K.), a board-certified anterior segment ophthalmologist and cataract surgeon, validated all the findings. All current PPP guidelines were also reviewed to determine whether or not cost-effectiveness or cost/value factors were explicitly mentioned as part of the justification for each recommendation. Additional review of each guideline was performed to determine the presence of cost/value statements, broadly defined as any statement in which cost or value was mentioned, and whether such statements were used to: 1) report a gap in cost/value evidence; 2) highlight economic impact of disease or care; and 3) advocate for cost/value-related issues, consistent with a framework previously described in the cardiology literature . We calculated the total number of recommendations with LOE reported for each current and prior PPP and calculated the change in number of recommendations over time. The median number of recommendations per guideline was determined. The number of recommendations with reported LOE were also summarized by subspecialty and care process category (diagnosis, management, or both). Additionally, the numbers and proportions of recommendations classified as LOE I, II, and III among all current and prior PPPs were determined. To further assess differences across subspecialties, we compared the number and proportion of LOE I, II, and III recommendations in current PPPs with those in prior PPPs by subspecialty. We also reported the quality of evidence according to the GRADE scale, stratified by LOE. For the current Cataract in the Adult Eye PPP, agreement between the two reviewers for whether a sentence constituted a recommendation statement was measured by percent agreement and kappa statistics. If references were provided in the document, the numbers and proportions of validated recommendation statements that were classified as LOE I, II, and III were determined. If recommendation statements also had LOE explicitly reported in the document, we compared the reported LOE with the study team-determined LOE. To evaluate the role of cost/value in PPPs, the proportion of current PPPs that had any recommendation supported by cost-effectiveness or cost/value considerations was determined. Further analysis was done to assess the proportions of current PPPs that contain statements addressing each of the specific areas related to cost/value considerations. Current PPP guidelines Overall, LOE from 24 current PPP guidelines published between 2017 and 2021 were abstracted. Across the 24 guidelines, the LOE (SIGN) was provided for 94 recommendations. The median number of recommendations with LOE per guideline was 1.5 (interquartile range [IQR]: 1.0—5.0). 83 (88%) recommendations had LOE I, 2 (2%) had LOE II, and 9 (10%) had LOE III. All LOE II and III recommendations were in the 2017 Refractive Errors and Refractive Surgery PPP. Among the 94 recommendations with LOE, the vast majority (98%) were for management. The remaining 2% were for diagnosis. The number of guidelines per subspecialty area ranged from 0 to 7 (Table ). The guidelines also reported recommendation strengths (GRADE) for 104 recommendations. 86 (83%) recommendations were strong recommendations, and 18 (17%) were discretionary recommendations. These recommendations included the 94 recommendations with reported LOE, of which 78 (83%) were strong recommendations, and 16 (17%) were discretionary recommendations. Importantly, aside from indication of recommendation strengths, the current guidelines did not clearly designate statements as recommendations and thus differentiate them from background or evidence synthesis information in the documents. The 2021 Cataract in the Adult Eye PPP was the guideline with the greatest number of recommendations with reported LOE ( n = 18). In our systematic review of this guideline, we identified 510 statements that could be considered recommendation statements. The two reviewers both identified 386 (76%) of these recommendation statements independently. Their overall percent agreement was 82%, and the kappa statistic was 0.64. Ninety-five additional statements were identified as recommendation statements by consensus after initially being included by only one reviewer. Twenty-nine statements were additionally considered recommendations by the third reviewer. We found that 267 (52%) recommendation statements did not have any reference associated with them. For the recommendations that did have references, we reviewed the references and categorized 92 (18%) recommendations as supported by LOE I, 95 (19%) as supported by LOE II, and 56 (11%) recommendations as supported by LOE III. For the 18 statements that had reported LOE, our categorization had 100% agreement with the reported LOE. Changes from prior to current guidelines Twenty-three current PPP guidelines had a prior edition (published between 2012 and 2016) available for comparison. The current Adult Strabismus PPP was published for the first time in 2019. Across the 23 prior PPPs, LOE were provided for 1254 recommendations, compared with 94 (93% decrease) in current editions. Table shows the changes in the number of recommendations with LOE by subspeciality and topic. Overall, the number of recommendations with LOE has decreased for a majority of specialties and topics. Among the 23 PPPs with both prior and current editions available, the number of recommendations with LOE I decreased from 114 to 83, the number of recommendations with LOE II decreased from 147 to 2, and the number of recommendations with LOE III decreased from 993 to 9. The proportion of LOE I recommendations rose from 9 to 88%, driven by a disproportionate decrease in reporting of evidence lower than LOE I. The median number of LOE I recommendations per PPP was 2 (IQR: 0–7), compared with a median of 1.5 LOE I recommendations (IQR: 1.0—5.0) per PPP in current guidelines. A subgroup analysis by subspecialty revealed similar findings (Fig. ). In addition to LOE based on the SIGN scale (LOE I, II and III), both current and prior PPP guidelines reported quality of evidence based on the GRADE scale (good quality, moderate quality, and insufficient quality). Figure shows the proportions of recommendations by level (SIGN) and quality (GRADE) of evidence in current and prior PPP guidelines. Among all recommendations with quality of evidence ratings, current guidelines rated the evidence for 61 (64.9%) recommendations as good quality, 22 (23.4%) as moderate quality, and 11 (11.7%) as insufficient quality. By contrast, prior guidelines rated 761 (57.2%) recommendations as good quality, 109 (8.2%) as moderate quality, and 461 (34.6%) as insufficient quality. An analysis of all recommendations with insufficient quality evidence in the current guidelines showed that though these recommendations had level I evidence ( n = 11), the evidence base either had a high risk of bias (i.e., LOE I- based on SIGN) or was not rated for risk of bias. By contrast, in prior guidelines, only 3 recommendations with insufficient quality evidence had level I evidence. The role of cost/value in PPP guidelines Among the 24 current PPP guidelines, 21 (88%) guidelines contained cost/value statements. A majority (75%) used cost/value statements to highlight the economic impact of disease or care, and 58% used cost/value statements to report gaps in cost/value evidence. None of them used cost/value considerations to justify specific recommendations or advocated for cost/value-related issues (Fig. ). Overall, LOE from 24 current PPP guidelines published between 2017 and 2021 were abstracted. Across the 24 guidelines, the LOE (SIGN) was provided for 94 recommendations. The median number of recommendations with LOE per guideline was 1.5 (interquartile range [IQR]: 1.0—5.0). 83 (88%) recommendations had LOE I, 2 (2%) had LOE II, and 9 (10%) had LOE III. All LOE II and III recommendations were in the 2017 Refractive Errors and Refractive Surgery PPP. Among the 94 recommendations with LOE, the vast majority (98%) were for management. The remaining 2% were for diagnosis. The number of guidelines per subspecialty area ranged from 0 to 7 (Table ). The guidelines also reported recommendation strengths (GRADE) for 104 recommendations. 86 (83%) recommendations were strong recommendations, and 18 (17%) were discretionary recommendations. These recommendations included the 94 recommendations with reported LOE, of which 78 (83%) were strong recommendations, and 16 (17%) were discretionary recommendations. Importantly, aside from indication of recommendation strengths, the current guidelines did not clearly designate statements as recommendations and thus differentiate them from background or evidence synthesis information in the documents. The 2021 Cataract in the Adult Eye PPP was the guideline with the greatest number of recommendations with reported LOE ( n = 18). In our systematic review of this guideline, we identified 510 statements that could be considered recommendation statements. The two reviewers both identified 386 (76%) of these recommendation statements independently. Their overall percent agreement was 82%, and the kappa statistic was 0.64. Ninety-five additional statements were identified as recommendation statements by consensus after initially being included by only one reviewer. Twenty-nine statements were additionally considered recommendations by the third reviewer. We found that 267 (52%) recommendation statements did not have any reference associated with them. For the recommendations that did have references, we reviewed the references and categorized 92 (18%) recommendations as supported by LOE I, 95 (19%) as supported by LOE II, and 56 (11%) recommendations as supported by LOE III. For the 18 statements that had reported LOE, our categorization had 100% agreement with the reported LOE. Twenty-three current PPP guidelines had a prior edition (published between 2012 and 2016) available for comparison. The current Adult Strabismus PPP was published for the first time in 2019. Across the 23 prior PPPs, LOE were provided for 1254 recommendations, compared with 94 (93% decrease) in current editions. Table shows the changes in the number of recommendations with LOE by subspeciality and topic. Overall, the number of recommendations with LOE has decreased for a majority of specialties and topics. Among the 23 PPPs with both prior and current editions available, the number of recommendations with LOE I decreased from 114 to 83, the number of recommendations with LOE II decreased from 147 to 2, and the number of recommendations with LOE III decreased from 993 to 9. The proportion of LOE I recommendations rose from 9 to 88%, driven by a disproportionate decrease in reporting of evidence lower than LOE I. The median number of LOE I recommendations per PPP was 2 (IQR: 0–7), compared with a median of 1.5 LOE I recommendations (IQR: 1.0—5.0) per PPP in current guidelines. A subgroup analysis by subspecialty revealed similar findings (Fig. ). In addition to LOE based on the SIGN scale (LOE I, II and III), both current and prior PPP guidelines reported quality of evidence based on the GRADE scale (good quality, moderate quality, and insufficient quality). Figure shows the proportions of recommendations by level (SIGN) and quality (GRADE) of evidence in current and prior PPP guidelines. Among all recommendations with quality of evidence ratings, current guidelines rated the evidence for 61 (64.9%) recommendations as good quality, 22 (23.4%) as moderate quality, and 11 (11.7%) as insufficient quality. By contrast, prior guidelines rated 761 (57.2%) recommendations as good quality, 109 (8.2%) as moderate quality, and 461 (34.6%) as insufficient quality. An analysis of all recommendations with insufficient quality evidence in the current guidelines showed that though these recommendations had level I evidence ( n = 11), the evidence base either had a high risk of bias (i.e., LOE I- based on SIGN) or was not rated for risk of bias. By contrast, in prior guidelines, only 3 recommendations with insufficient quality evidence had level I evidence. Among the 24 current PPP guidelines, 21 (88%) guidelines contained cost/value statements. A majority (75%) used cost/value statements to highlight the economic impact of disease or care, and 58% used cost/value statements to report gaps in cost/value evidence. None of them used cost/value considerations to justify specific recommendations or advocated for cost/value-related issues (Fig. ). This systematic analysis of LOE supporting AAO PPP guidelines evaluated the proportion of guidelines with a LOE listed, as well as changes in reporting patterns over time and across specialties. Overall, this study demonstrated that while current guidelines report LOE for substantially fewer recommendations, a much higher proportion of recommendations are supported by evidence from randomized controlled trials. Eighty-eight percent of current recommendations with reported LOE had LOE I. Subgroup analysis by subspecialty showed similar trends. These results suggest that while current AAO PPPs emphasize evidence from randomized controlled trials, LOE from other types of studies may not be formally rated or reported. Although analyses of guidelines have been performed in other specialties, such as cardiology , prior investigation of evidence supporting guidelines in ophthalmology is limited. A 2015 study examined the LOE of papers published in four major ophthalmology journals and concluded that lower LOE publications would continue to play a large role in guiding the field of ophthalmology . At first glance, the findings from our study do not appear to suggest this same trend among the reported LOE supporting AAO PPPs, as the vast majority of recommendations with reported LOE had the highest level of evidence. However, our comprehensive review of the 2021 Cataract in the Adult Eye PPP and independent rating of LOE of the citations show that 30% of recommendations rely on level II and III evidence (vs. 18% level I), but the LOE was simply not reported in the PPP. The majority (52%) of recommendations did not have any citations, consistent with a prior study investigating the relationship between findings from systematic reviews and the 2015 AAO PPP on interventions for age-related macular degeneration . The study found that only 1 out of 35 treatment recommendations in the PPP cited a reliable intervention systematic review . Our study complements the existing literature, highlighting that there may be areas to include additional supporting evidence in AAO PPPs. In evidence-based medicine, randomized controlled trials (RCTs) and systematic reviews/meta-analyses synthesizing their results are the pinnacle of evidence as randomization reduces bias and allows for investigation of causal relationships. A study conducted in 2019 found that only 2% of all publications in the field of ophthalmology were RCTs . In our study, while the proportion of LOE I (ie, meta-analysis, systematic reviews of RCTs, or RCTs) recommendations has increased from prior PPPs to current PPPs, this increase is primarily driven by a disproportionate underreporting of lower-level evidence. In fact, the number of LOE I recommendations has not increased. On the one hand, this trend suggests guideline authors may have attempted to highlight LOE I recommendations in the current PPPs. On the other hand, fully reporting both level I and lower-level evidence could help to expand the evidence base highlighted in ophthalmology guidelines. The Institute of Medicine’s landmark reports on clinical practice guidelines were the impetus for the initial development of many guidelines in effect today . In 2011, the Institute of Medicine recommended standards for developing trustworthy clinical practice guidelines . The standards state that for each recommendation in a guideline, “a rating of the level of confidence in the evidence underpinning the recommendation” should be provided . Our results suggest that substantial underreporting of LOE may exist in current PPPs, as the number of recommendations with reported LOE fell from 1254 in prior PPPs to 94 in current PPPs. These results suggest that there is significant opportunity to include level II and III evidence, which, despite risk of bias, is nonetheless often critically important data . Use of data sources such as insurance claims or multi-institutional registries can provide information about real-world clinical practice that cannot be generated by randomized clinical trials . Furthermore, there are clinical questions for which a randomized trial is infeasible, such as for rare conditions or for procedures where shams are not possible, and in these circumstances, lower levels of evidence ought to be weighted more heavily . Since many ophthalmologic diseases have a low incidence and a heavy reliance on surgical management in certain subspecialties, performing randomized controlled trials may be especially challenging . By acknowledging this and including varied levels of evidence in ophthalmology PPPs, authors may be able to more easily adopt the Institute of Medicine’s recommendation about LOE reporting in clinical practice guidelines. The same report from the Institute of Medicine also proposed that “recommendations should be articulated in a standardized form detailing precisely what the recommendation action is, and under what circumstances it should be performed” . Our findings demonstrated that aside from indication of recommendation strengths ( n = 104 recommendations across 24 PPPs), the current guidelines do not articulate recommendations in a standardized form. Without such standardization, our comprehensive review of the 2021 Cataract in the Adult Eye PPP identified 510 statements that addressed how patients should be diagnosed or managed clinically and thus could be considered recommendations. While the agreement between our two reviewers was good (k statistic = 0.64), this result suggests that interpretations of potential recommendation statements in the PPP can be variable. Standardized articulation of recommendations would help clinicians clearly identify recommended actions for clinical practice. For example, American Heart Association guidelines list all recommendations in visually distinctive boxes, which stand out from the surrounding text and include levels of evidence . Clearly articulating recommendations could also facilitate the creation and assessment of programs to improve the quality of care. As health care usage and expenditure continue to rise in the United States, value-based care has become an increasingly important concept . A recent systematic review found that between 75.7 and 101.2 billion was spent on low-value care in the United States . Clinical practice guidelines play an important role in shaping practice patterns and thus may be well-suited to promote high-value care. In this area, prior work in cardiology has evaluated cost and value considerations in contemporary heart failure clinical guidelines . The study concluded that although most contemporary heart failure guidelines contained cost/value statements, they were rarely used to support clinical guidance recommendations. In the ophthalmology guidelines, a majority (88%) of PPPs included cost/value statements. In particular, the high economic impact of disease or care was frequently highlighted (75% of PPPs). However, cost/value considerations have yet to be incorporated into the development of specific recommendations, representing an avenue for future work in ophthalmology guideline development. More than half of the PPPs also reported gaps in cost/value evidence—ongoing efforts in the field such as the IRIS® Registry, which includes performance metrics, may facilitate real-world evidence generation in this area and help to provide needed data for guideline development . The strengths of this study include analysis of all PPPs spanning a 10-year period, including all contemporary PPPs and their immediate predecessors. This thorough analysis allowed us to assess evolutions of PPPs over time and trends in all the specialties and topics that PPPs cover. Furthermore, we reported levels of evidence exactly as described in the guidelines. Additional strengths of this study included independent two-party grading, with validation by a board-certified anterior segment specialist, of levels of evidence for our review of the 2021 Cataract in the Adult Eye PPP. The limitations include the potential underreporting of LOE in current PPPs, which prohibits us from drawing conclusions about all evidence supporting PPPs. This is partially addressed by our comprehensive review of the 2021 Cataract in the Adult Eye PPP, including its references, which suggests substantial underreporting of LOE across all LOE and disproportionate underreporting of lower-level evidence. In conclusion, we performed a systematic analysis of reported LOE supporting AAO PPP guidelines. Compared with prior PPPs, current PPPs emphasize evidence from randomized controlled trials. While underreporting of LOE across all LOE exists, there appears to be a disproportionate underreporting of lower-level evidence. Future guideline development may consider clearly defining recommendations, explicitly reporting LOE associated with each recommendation, and integrating cost/value considerations in recommendations. |
FREEDA: An automated computational pipeline guides experimental testing of protein innovation | ec3df1f0-3838-46c0-a273-ebd3c67089ca | 10292211 | Anatomy[mh] | Purifying selection eliminates deleterious non-synonymous mutations, leading to conservation of amino acid sequence. In contrast, positive selection results in the accumulation of non-synonymous mutations that lead to functional innovation and adaptation (reviewed in ). Compelling examples of how positive selection has regulated protein function come from studying host–pathogen genetic conflicts. In these evolutionary arms races, positive selection leads to rapid accumulation of mutations in both viral proteins that help infect the host and host proteins that help evade the infection (reviewed in ; ). To experimentally test functional innovation, evolutionary biologists swap protein regions (or entire alleles) from closely related species that are suspected to have diverged due to positive selection. This approach generates an “evolutionary mismatch” between the divergent protein and the cellular environment, revealing which protein function might have evolved adaptively (reviewed in ). For example, swapping a region of the TRIM5 protein between human and rhesus monkey suggested that positive selection shaped its role in fighting species-specific retroviral infections . Remarkably, variation at even single residues under positive selection can lead to functional changes, as in the human MAVS (Mitochondrial Antiviral Signaling) protein that has evolved to evade infection with hepaciviruses . These examples illustrate that innovation-guided functional analyses can complement more traditional conservation-guided approaches in revealing regulation of protein function. Genetic conflicts, like those between host and pathogen, can result in recurrently changing selection pressure and recurrent adaptation of proteins regulating essential cellular processes. For example, pressure to maintain genome integrity at fertilization is thought to fuel a sexual conflict between paternal proteins that adapt to maximize the chance of fertilizing the egg and maternal proteins that adapt to prevent entry of more than one sperm (reviewed in ). Similarly, selfish genetic elements such as transposons constantly disrupt genome integrity, leading to intragenomic conflicts and recurrent adaptation of DNA packaging proteins (reviewed in ). Centromere DNA sequences have also been proposed to act as selfish elements, raising the possibility of intragenomic conflict with centromere-associated proteins. Centromeres are repetitive DNA regions that direct chromosome segregation in mitosis and meiosis by assembling kinetochores, multiprotein structures that connect to spindle microtubules. Despite their essential function, centromeric DNA and proteins evolve rapidly across taxa, suggesting an evolutionary pressure to recurrently innovate. The centromere drive hypothesis proposes that selfish centromeric DNA sequences achieve non-Mendelian segregation during asymmetric female meiosis, increasing their transmission to the egg. Fitness costs imposed by this selfish behavior would lead to recurrent adaptation of centromeric proteins to suppress the costs . While there is experimental evidence for selfish centromeric DNA, the impact of positive selection on centromeric protein function remains largely untested (reviewed in ). The scarcity of experimental studies of adaptive evolution in centromeric proteins, in contrast to our increasingly detailed understanding of their conserved functions , reflects the general focus of cell biology research on protein conservation rather than innovation. This discrepancy is due in part to challenges in designing experiments to infer functional consequences of positive selection, but also to the complexity of methods needed to distinguish positive selection from neutral evolution of protein-coding sequences (reviewed in ). A widely used method calculates the rate ratio of non-synonymous (dN) to synonymous (dS) substitutions per codon (dN/dS ratio; ; ; ) using multiple sequence alignment of closely related orthologs, which are homologous genes that arise when speciation occurs. This approach assumes that synonymous mutations are neutral, while deleterious non-synonymous mutations are purged by purifying selection. An enrichment of non-synonymous relative to synonymous substitutions within the alignment suggests recurrent adaptation to a constantly changing selection pressure (types of recurrent evolution are discussed in ). Well-established computational suites such as PAML (Phylogenetic Analysis by Maximum Likelihood; ) and HyPhy (Hypothesis Testing using Phylogenies; ) offer a number of tools that can reliably detect statistical signatures of positive selection but are seldom used by cell biologists because expertise in computational biology and molecular evolution is required to generate the input data, and the output is rarely provided in an intuitive visual format. Automated molecular evolution pipelines that incorporate the abovementioned tools have been developed (see for a non-exhaustive list), but their complexity and the need for user-provided input still render them inaccessible to experimental cell biologists with limited computational skills. Increasing this access requires a “one-click” application that (1) offers a simple graphical user interface, (2) fully automates input preparation, (3) finds orthologs despite the lack of genomic annotations, (4) reduces parameterization, and (5) provides intuitive visual representation of the output. Here, we present FREEDA (Finder of Rapidly Evolving Exons in Diverse Assemblies), a fully automated, end-to-end pipeline designed for cell biologists seeking to apply an evolutionary lens by testing for statistical evidence of positive selection in their favorite proteins. FREEDA provides the key functionalities listed above, including a unique ability to map positively selected residues onto any predicted protein structure. As a proof-of-principle, we first use FREEDA to map positive selection across centromeric proteins in rodents, as mice are currently the only experimentally tractable cell biological model system for centromere drive (reviewed in ). Guided by these computational analyses, we use the evolutionary mismatch approach to provide experimental evidence of functional innovation in the centromeric protein CENP-O.
Overview of the FREEDA pipeline FREEDA is a stand-alone application with an intuitive graphical user interface (GUI) operating on UNIX systems (MacOS and Linux; Windows users, please see documentation). An overview and documentation of the pipeline are provided at https://ddudka9.github.io/freeda/ with a more detailed walkthrough in Materials and methods. FREEDA first downloads the reference genome of the user-selected species and prepares input data for the gene of interest by connecting to genomic, protein, and protein structure databases ( ; blue). Next, FREEDA downloads a preselected set of non-annotated genome assemblies related to the reference species, performs a BLAST (Basic Local Alignment Search Tool) search for orthologs of the gene of interest, and uses the reference species data to find orthologous sequences ( ; orange). Combining several well-established molecular evolution tools, FREEDA aligns all coding sequences, builds phylogenetic trees, determines the likelihood that positive selection has shaped the evolution of the gene, and estimates the probability that given residues have evolved under positive selection ( ; brown). Key results are displayed within the GUI and all results are saved into the “Results-current-date” folder generated in a location selected by the user (“Set directory”; ). These files include the BLAST output, nucleotide alignment, phylogenetic tree, protein alignment, residue mapping onto reference coding sequence, and residue mapping onto protein structure. The raw data and intermediate alignment files are saved in the “Raw_data” folder. Since FREEDA finds orthologs by downloading entire genomic assemblies, the user is advised to select an external data storage device (e.g., a hard drive) when setting the directory. A stable internet connection is also required to allow communication with various databases . Advantages over existing automated pipelines Several features distinguish FREEDA from currently available automated pipelines . First, FREEDA is fully automated, requiring only a gene name, and distributed as a self-contained application that does not require installation or compilation of any additional programs, except for a straightforward installation of the widely used protein structure viewer PyMOL (The PyMOL Molecular Graphics System, Version 2.0, Schrödinger, LLC) for MacOS users. Second, FREEDA uses a defined set of non-annotated genomic assemblies that ensure high statistical power of the analysis while absolving the users from manually curating their input. As new genomic assemblies become available, they will be incorporated into new FREEDA releases. Third, FREEDA automatically maps residues with the highest probabilities of having evolved under positive selection onto protein structure models by querying the AlphaFold database containing structure predictions for nearly all known proteins . Finally, by providing a simple GUI, FREEDA minimizes complexity compared to currently available pipelines, while offering a restrained number of advanced options . Therefore, the user may consider FREEDA as an entry point to performing the first molecular evolution analyses of their proteins of interest. FREEDA validation: Finding orthologous genes To demonstrate that FREEDA’s simplicity does not compromise its functionality, we first tested its ability to find orthologous sequences in non-annotated assemblies—a notoriously challenging task in the field. To remain unbiased, we randomly selected five genes from reference assemblies of rodents ( Murinae ; mouse), primates ( Simiiformes ; human), carnivores ( Carnivora ; dog), and birds ( Phasianidae ; chicken) and compared the FREEDA-identified orthologs to their annotations in 26 species available in the highly curated Ensembl database . While only 26/74 species used by FREEDA for these clades have Ensembl-annotated assemblies, we managed to analyze >120 orthologs. We confirmed the identities of all the FREEDA-identified orthologs by showing that they share (1) genomic location with Ensembl-annotated flanking genes and (2) an average of 99.9% (without insertions/deletions [indels]) or 90.3% (with indels) nucleotide sequence identity with Ensembl-annotated orthologs. The use of alternative exons and start codons explains the vast majority of sequence differences when indels are not excluded (see ). These unbiased analyses validate FREEDA’s ability to reliably detect orthologous genes. To ensure rigor in detecting orthologs, we tested if FREEDA can distinguish them from paralogs, which form by a duplication event and may evolve under different selective pressures. FREEDA demonstrated the ability to resolve ancient duplications by correctly distinguishing HERC5 orthologs from HERC6 paralogs present within a dataset of previously curated human immune genes used to validate the DGINN pipeline (Detect Genetic INNovations; ; see raw data of the analyzed dataset in additional online supplemental material: https://doi.org/10.5281/zenodo.7997737 ). Additionally, we tested if FREEDA could resolve tandem duplications (duplicated genes located side by side) and retro-duplications (intron-less mRNA that was reverse-transcribed and inserted back into the genome). Using the “Tandem duplication expected” option (see GUI; ), FREEDA successfully distinguished primate genes H4C1 and H4C2 , both encoding histone H4 and located merely 5 kb apart with 85% nucleotide sequence identity (additional supplementary materials). Visual examination of the nucleotide alignment revealed that one H4C2 ortholog (in Plecturocebus donacophilus ) lost the ancestral start codon and likely pseudogenized. In such cases, the user may choose to rerun the analysis using the “Exclude species” option (see GUI; ). We further used the “Duplication expected” option (see GUI; ) to show that FREEDA could correctly distinguish KIF4A , encoding a kinesin motor, from its retroduplicate KIF4B . These genes are an example of a recent duplication that occurred in a common ancestor of primates, retaining 96% nucleotide sequence identity ( ; additional supplementary materials). We also found that KIF4B , and not KIF4A , has likely evolved under positive selection, suggesting that a duplication event spurred adaptive evolution of this kinesin. While human KIF4A regulates chromosome segregation , cellular transport , and anti-viral response , KIF4B remains poorly studied. Together, these analyses demonstrate that FREEDA reliably finds orthologous sequences even when a gene has undergone duplication. FREEDA validation: Detecting statistical signatures of positive selection We then tested if FREEDA could accurately detect statistical signatures of positive selection in genes with known evolutionary histories. To do so, we used a dataset of 23 primate ( Simiiformes ) genes whose statistical signatures of positive selection (or lack thereof) have been previously defined. The dataset included 19 genes curated to validate the DGINN pipeline . Analyzing a set of 19 primate species, FREEDA found 18 orthologs with 98% coding sequence coverage (median values; ). Consistent with the literature, FREEDA found statistical signatures of positive selection in TRIM5 , MAVS , SAMHD1 , IFI16 , ZC3HAV1 , RSAD2 , GBP5, MX1 , APOBEC3F , and NBN . Although previous studies also reported that positive selection has likely shaped the evolution of BST2 (using nine primate species; ; ), FREEDA only found a weak statistical signature of positive selection in that gene (P = 0.0864; ), which likely stems from the lineage of New World monkeys . Of six genes whose evolutionary history is less clear, with results dependent on the method used , FREEDA found statistical evidence of positive selection in only one ( SERINC3 ; ), highlighting the stringency of the analysis. Of six genes whose adaptive evolution has been previously deemed unlikely, FREEDA detected a signature of positive selection in one, TREX1 , a nuclease that guards genome integrity . One of the residues with the highest probability of positive selection (serine at position 166 in human; probability = 0.97; ) is proximal to a primate-specific DNA-binding site (arginine at position 164 in humans; ), suggesting that adaptive evolution has shaped DNA recognition. Consistent with our finding, divergent DNA binding sites in TREX1 regulate DNA recognition . We suspect that differences in regions removed from the analysis are responsible for the disparity between published results and ours. Altogether, using previously curated datasets allowed us to objectively validate our pipeline and provided additional insight into the evolutionary history of these genes. To further validate accuracy of the pipeline at the level of single residues, we compared specific sites that have likely evolved under positive selection found by FREEDA to those previously mapped in MAVS , MX1 , SAMHD1 , and TRIM5 . Exact matching of probabilities for each residue was not expected due to differences in algorithms for aligning orthologous sequences (see Materials and methods for details). Nevertheless, FREEDA found statistical signatures of positive selection in all published sites, except for those located in regions removed from the alignment to ensure its high quality (five residues in each SAMHD1 and MX1 ; two residues in TRIM5 ; ; and ). Using MX1 as an example , FREEDA maps detected sites onto the reference coding sequence and onto structural prediction models generated by AlphaFold . Overall, these analyses demonstrate that FREEDA can retrieve expected sites with previously reported signatures of positive selection and showcase FREEDA’s key results visualization features. Finally, we tested if FREEDA can reliably detect statistical signatures of positive selection in rodent genomes ( Murinae ). As a test dataset, we selected 104 centromeric genes, 42 of which have been previously analyzed using a smaller number of species (up to 11; ). Analyzing a set of 19 Murinae species, FREEDA found 16 orthologs with 94% coding sequence coverage (median values; ). Consistent with our previous findings of pervasive evolutionary innovation across the rodent centromere , FREEDA found that 36/104 genes have likely evolved under positive selection ( and ). Corroborating our previous results, FREEDA detected statistical signatures of positive selection in genes encoding CENP-C, CENP-I, CENP-T, HJURP, INCENP, MIS18BP1, KNL1, and SGO2. In contrast, DSN1 and HEC1 did not show statistical signatures of positive selection. This discrepancy is likely due to a difference in coding sequence coverage between the analyses (higher in FREEDA) or it reflects a higher statistical power due to more orthologs (up to 19 used by FREEDA), which facilitates distinguishing between positive selection and relaxation of purifying selection. This high statistical power revealed several previously unknown targets of positive selection, including components of the fibrous corona, which helps capture microtubules (CENP-F, SPINDLY, ZWILCH, ROD, NUP85, NUP98, and ELYS; reviewed in ), microtubule motors (CENP-E, KIF2B, and KIF18A), and protein kinases (AURKC and HASPIN). To further validate our findings, we repeated the analyses with rat as reference species. Since the quality of the available rat genome annotation is lower than that of mouse, FREEDA was able to collect reliable input data for only 89/104 genes. As expected, we found statistical evidence (or lack thereof) of positive selection in almost exactly the same genes as when using mouse as reference (85/89 genes; see Discussion; ). Overall, these tests show that despite its simplicity for the user, FREEDA is a fully functional and dependable tool to detect statistical signatures of positive selection. Using FREEDA to derive evolution-guided hypotheses To test if FREEDA can help derive evolution-guided hypotheses, we leveraged its ability to map residues that have likely evolved under positive selection onto protein structures. We found statistical evidence of positive selection within ancient (retained across long evolutionary timescales) protein domains of centromeric proteins, suggesting that adaptive evolution shaped essential protein functions . For instance, we detected residues with high probability of having evolved under positive selection in the ancient Yippee domain of MIS18β (encoded by Oip5 in mouse), which participates in centromere chromatin assembly (reviewed in ). In addition to its divergent N- and C-termini, one of the most likely adaptive residues (arginine at position 76 in mouse; probability = 0.98) is located within the loop-forming CXXC motif of the Yippee domain , which is required for MIS18 complex assembly at centromeres . Similarly, we found strong statistical evidence of positive selection in one of the loops of an ancient protein kinase domain (reviewed in ) in the meiosis-specific Aurora kinase C (AURKC, asparagine at position 150 in mouse, probability = 0.98; ), which helps correct erroneous kinetochore-microtubule attachments . In contrast, we found no recurrent changes in the related AURKA and AURKB kinases . These data suggest that positive selection has uniquely tuned the kinase activity of the specialized meiotic Aurora kinase, consistent with previous reports of adaptive evolution of reproduction genes . Finally, we found statistical evidence that positive selection shaped evolution of the ancient double RWD domain (RING-WD-DEAD; ) of CENP-O, which regulates kinetochore–microtubule attachments by forming the CENP-OPQUR complex . RWD domains are prevalent structural modules that facilitate protein–protein interactions across the centromere . CENP-O shares a high structural similarity with its binding partner CENP-P, which also shows statistical signatures of positive selection within its double RWD domain . Furthermore, some of the residues that have evolved under positive selection with the highest probability are located in or near loops and turns flanking highly structured α-helices and β-sheets in CENP-O and -P C-terminal RWD domains ( and ). Based on these results, we propose that positive selection has regulated essential functions of centromeric proteins by acting on loops and turns of ancient domains, consistent with previous reports of frequent innovation of flexible regions in other proteins . Altogether, we demonstrate that FREEDA can help derive evolution-guided hypotheses by highlighting protein domains whose function has likely been shaped by adaptive evolution. Using FREEDA to infer molecular mechanisms regulated by positive selection Each of the proteins discussed above (MIS18β, AURKC, and CENP-OP) functions as part of a complex. To infer mechanisms regulated by positive selection in this context, we aligned FREEDA-annotated protein structure predictions of mouse proteins to experimentally solved structures of their orthologs in complex with binding partners (see Materials and methods for details). Two loops formed by CXXC motifs within the Yippee domain of MIS18β together give rise to a tetrahedral module whose four conserved cysteines bind a zinc ion , likely stabilizing protein conformation . Aligning mouse MIS18β to the crystal structure of the fission yeast MIS18 Yippee-like domain shows that the side chain of arginine at the positively selected position 76 in mouse likely faces the opening of the tetrahedral module. This finding is consistent with XX residues regulating the function of CXXC motifs in other proteins . Alternatively, R76 could mediate MIS18α and MIS18β heterodimerization . Therefore, we hypothesize that positive selection favored amino acid changes within the CXXC motif to modulate MIS18 complex stability. Consistent with functional innovation of CXXC motifs, we also found recurrently changing residues within the second CXXC motif of MIS18β (glycine at position 135 in mouse; ) and in the first CXXC motif of its binding partner MIS18α (serine at position 57 in mouse; ), albeit the probability that they have evolved under positive selection was lower (probabilities = 0.88 and 0.77, respectively). These data suggest that positive selection in the loops of the ancient Yippee domains regulated centromere assembly by modulating stability of the MIS18 complex. AURKB and AURKC kinase activity requires binding to a conserved domain of INCENP (INner-CENtromere Protein; reviewed in ). Aligning the Mus musculus (Mm) AURKC protein kinase domain and MmINCENP AURK-binding domain to the crystal structure of the orthologous human domains shows the side chain of positively selected asparagine at position 150 in mouse in close proximity to tyrosine at conserved position 827 in MmINCENP. This finding suggests modulation of INCENP binding and, therefore, kinase activity by positive selection . The rodent AURKC activation loop also contains a recurrently changing, albeit less likely adaptive residue (serine at position 156 in mouse; probability = 0.77; ) whose side chain reaches toward the AURKC ATP-binding site (marked by the inhibitor BRD-7880; ; ). These data suggest that positive selection in the loop of the ancient protein kinase domain of AURKC regulated meiotic functions by modulating kinase activity. Double RWD domains mediate the formation of CENP-OP heterodimers, allowing recruitment of the CENP-OPQUR complex to centromeres . Aligning the FREEDA-annotated CENP-O and -P C-terminal RWD domains to the experimentally solved human CENP-OPQUR complex suggests that positive selection shaped opposite sides of the CENP-OP heterodimer and therefore is unlikely to have impacted heterodimerization. In yeast, C-terminal RWD domains of CENP-O and -P orthologs bind to CENP-Q and -U orthologs to form the COMA complex . We were unable to reliably align mouse CENP-Q and -U to the human CENP-OPQUR complex, likely due to long unstructured regions in CENP-Q and -U, but the striking pattern of likely adaptive residues in C-terminal RWD domains facing the outside of the heterodimer suggests that positive selection regulated binding to nearby centromeric components . We find statistical signatures of positive selection in CENP-Q and -U in rodents , suggesting that positive selection regulated interactions between CENP-OPQUR complex components. Altogether, these analyses of multiple centromere proteins demonstrate how FREEDA-annotated structures can be used to generate hypotheses for how positive selection might have regulated essential protein functions. Experimental testing of functional protein innovation To test our hypothesis that loops and turns in ancient protein domains regulate their essential functions, we chose to focus on CENP-O because FREEDA suggests that positive selection operated on residues flanking α-helices and β-sheets of both rodent and primate C-terminal RWD domains , and centromere binding provides a straightforward functional assay. We used mouse oocytes for these experiments because they are an established model system for centromere drive, the most likely selective pressure sculpting evolution of centromeric proteins, and thus a natural context to probe for functional protein innovation. To create an evolutionary mismatch (see Introduction), we introduced GFP-tagged full-length mouse (control) or rat (divergent) CENP-O at similar expression levels . Mouse CENP-O localized to centromeres as expected, but rat CENP-O was nearly undetectable at mouse centromeres , indicating functional innovation in centromere binding. To test if the C-terminal RWD domain is responsible for that innovation, we compared three chimeric rat CENP-O constructs with different regions of mouse CENP-O: N-terminal (N-terminal tail and N-terminal helix), central (N-terminal RWD domain and central helix), or C-terminal (C-terminal RWD domain; and ). Only the mouse C-terminal RWD domain could rescue, albeit not fully, the localization of rat CENP-O to mouse centromeres. In an inverse experiment, a chimera of mouse CENP-O with the rat C-terminal RWD domain failed to localize to mouse centromeres ( and ). Together, these results demonstrate that mouse-specific innovation in the C-terminal RWD domain is required for CENP-O binding to mouse centromeres. Within this domain, 10 out of 13 residues that differ between mouse and rat are putatively adaptive (probability ≥ 0.5; ). Almost all (9/10) of these residues flank highly structured α-helices or β-sheets (±1 amino acid), consistent with our hypothesis that positive selection drives functional innovation of ancient domains in centromeric proteins by acting on their loops and turns. Swapping five of the most likely adaptive residues in the mouse C-terminal RWD domain to those found in rat did not, however, reduce centromere localization of mouse CENP-O . Similarly, swapping equivalent rat residues within the C-terminal RWD domain of rat CENP-O to mouse-specific ones (in addition to six mutations in other parts of the protein) did not restore its centromere localization . These analyses highlight the difficulty in attributing innovation to specific residues given the number of possible combinations as well as the potential for epistasis . Altogether, we show that our fully automated molecular evolution pipeline can guide experimental testing of functional protein innovation.
FREEDA is a stand-alone application with an intuitive graphical user interface (GUI) operating on UNIX systems (MacOS and Linux; Windows users, please see documentation). An overview and documentation of the pipeline are provided at https://ddudka9.github.io/freeda/ with a more detailed walkthrough in Materials and methods. FREEDA first downloads the reference genome of the user-selected species and prepares input data for the gene of interest by connecting to genomic, protein, and protein structure databases ( ; blue). Next, FREEDA downloads a preselected set of non-annotated genome assemblies related to the reference species, performs a BLAST (Basic Local Alignment Search Tool) search for orthologs of the gene of interest, and uses the reference species data to find orthologous sequences ( ; orange). Combining several well-established molecular evolution tools, FREEDA aligns all coding sequences, builds phylogenetic trees, determines the likelihood that positive selection has shaped the evolution of the gene, and estimates the probability that given residues have evolved under positive selection ( ; brown). Key results are displayed within the GUI and all results are saved into the “Results-current-date” folder generated in a location selected by the user (“Set directory”; ). These files include the BLAST output, nucleotide alignment, phylogenetic tree, protein alignment, residue mapping onto reference coding sequence, and residue mapping onto protein structure. The raw data and intermediate alignment files are saved in the “Raw_data” folder. Since FREEDA finds orthologs by downloading entire genomic assemblies, the user is advised to select an external data storage device (e.g., a hard drive) when setting the directory. A stable internet connection is also required to allow communication with various databases .
Several features distinguish FREEDA from currently available automated pipelines . First, FREEDA is fully automated, requiring only a gene name, and distributed as a self-contained application that does not require installation or compilation of any additional programs, except for a straightforward installation of the widely used protein structure viewer PyMOL (The PyMOL Molecular Graphics System, Version 2.0, Schrödinger, LLC) for MacOS users. Second, FREEDA uses a defined set of non-annotated genomic assemblies that ensure high statistical power of the analysis while absolving the users from manually curating their input. As new genomic assemblies become available, they will be incorporated into new FREEDA releases. Third, FREEDA automatically maps residues with the highest probabilities of having evolved under positive selection onto protein structure models by querying the AlphaFold database containing structure predictions for nearly all known proteins . Finally, by providing a simple GUI, FREEDA minimizes complexity compared to currently available pipelines, while offering a restrained number of advanced options . Therefore, the user may consider FREEDA as an entry point to performing the first molecular evolution analyses of their proteins of interest.
To demonstrate that FREEDA’s simplicity does not compromise its functionality, we first tested its ability to find orthologous sequences in non-annotated assemblies—a notoriously challenging task in the field. To remain unbiased, we randomly selected five genes from reference assemblies of rodents ( Murinae ; mouse), primates ( Simiiformes ; human), carnivores ( Carnivora ; dog), and birds ( Phasianidae ; chicken) and compared the FREEDA-identified orthologs to their annotations in 26 species available in the highly curated Ensembl database . While only 26/74 species used by FREEDA for these clades have Ensembl-annotated assemblies, we managed to analyze >120 orthologs. We confirmed the identities of all the FREEDA-identified orthologs by showing that they share (1) genomic location with Ensembl-annotated flanking genes and (2) an average of 99.9% (without insertions/deletions [indels]) or 90.3% (with indels) nucleotide sequence identity with Ensembl-annotated orthologs. The use of alternative exons and start codons explains the vast majority of sequence differences when indels are not excluded (see ). These unbiased analyses validate FREEDA’s ability to reliably detect orthologous genes. To ensure rigor in detecting orthologs, we tested if FREEDA can distinguish them from paralogs, which form by a duplication event and may evolve under different selective pressures. FREEDA demonstrated the ability to resolve ancient duplications by correctly distinguishing HERC5 orthologs from HERC6 paralogs present within a dataset of previously curated human immune genes used to validate the DGINN pipeline (Detect Genetic INNovations; ; see raw data of the analyzed dataset in additional online supplemental material: https://doi.org/10.5281/zenodo.7997737 ). Additionally, we tested if FREEDA could resolve tandem duplications (duplicated genes located side by side) and retro-duplications (intron-less mRNA that was reverse-transcribed and inserted back into the genome). Using the “Tandem duplication expected” option (see GUI; ), FREEDA successfully distinguished primate genes H4C1 and H4C2 , both encoding histone H4 and located merely 5 kb apart with 85% nucleotide sequence identity (additional supplementary materials). Visual examination of the nucleotide alignment revealed that one H4C2 ortholog (in Plecturocebus donacophilus ) lost the ancestral start codon and likely pseudogenized. In such cases, the user may choose to rerun the analysis using the “Exclude species” option (see GUI; ). We further used the “Duplication expected” option (see GUI; ) to show that FREEDA could correctly distinguish KIF4A , encoding a kinesin motor, from its retroduplicate KIF4B . These genes are an example of a recent duplication that occurred in a common ancestor of primates, retaining 96% nucleotide sequence identity ( ; additional supplementary materials). We also found that KIF4B , and not KIF4A , has likely evolved under positive selection, suggesting that a duplication event spurred adaptive evolution of this kinesin. While human KIF4A regulates chromosome segregation , cellular transport , and anti-viral response , KIF4B remains poorly studied. Together, these analyses demonstrate that FREEDA reliably finds orthologous sequences even when a gene has undergone duplication.
We then tested if FREEDA could accurately detect statistical signatures of positive selection in genes with known evolutionary histories. To do so, we used a dataset of 23 primate ( Simiiformes ) genes whose statistical signatures of positive selection (or lack thereof) have been previously defined. The dataset included 19 genes curated to validate the DGINN pipeline . Analyzing a set of 19 primate species, FREEDA found 18 orthologs with 98% coding sequence coverage (median values; ). Consistent with the literature, FREEDA found statistical signatures of positive selection in TRIM5 , MAVS , SAMHD1 , IFI16 , ZC3HAV1 , RSAD2 , GBP5, MX1 , APOBEC3F , and NBN . Although previous studies also reported that positive selection has likely shaped the evolution of BST2 (using nine primate species; ; ), FREEDA only found a weak statistical signature of positive selection in that gene (P = 0.0864; ), which likely stems from the lineage of New World monkeys . Of six genes whose evolutionary history is less clear, with results dependent on the method used , FREEDA found statistical evidence of positive selection in only one ( SERINC3 ; ), highlighting the stringency of the analysis. Of six genes whose adaptive evolution has been previously deemed unlikely, FREEDA detected a signature of positive selection in one, TREX1 , a nuclease that guards genome integrity . One of the residues with the highest probability of positive selection (serine at position 166 in human; probability = 0.97; ) is proximal to a primate-specific DNA-binding site (arginine at position 164 in humans; ), suggesting that adaptive evolution has shaped DNA recognition. Consistent with our finding, divergent DNA binding sites in TREX1 regulate DNA recognition . We suspect that differences in regions removed from the analysis are responsible for the disparity between published results and ours. Altogether, using previously curated datasets allowed us to objectively validate our pipeline and provided additional insight into the evolutionary history of these genes. To further validate accuracy of the pipeline at the level of single residues, we compared specific sites that have likely evolved under positive selection found by FREEDA to those previously mapped in MAVS , MX1 , SAMHD1 , and TRIM5 . Exact matching of probabilities for each residue was not expected due to differences in algorithms for aligning orthologous sequences (see Materials and methods for details). Nevertheless, FREEDA found statistical signatures of positive selection in all published sites, except for those located in regions removed from the alignment to ensure its high quality (five residues in each SAMHD1 and MX1 ; two residues in TRIM5 ; ; and ). Using MX1 as an example , FREEDA maps detected sites onto the reference coding sequence and onto structural prediction models generated by AlphaFold . Overall, these analyses demonstrate that FREEDA can retrieve expected sites with previously reported signatures of positive selection and showcase FREEDA’s key results visualization features. Finally, we tested if FREEDA can reliably detect statistical signatures of positive selection in rodent genomes ( Murinae ). As a test dataset, we selected 104 centromeric genes, 42 of which have been previously analyzed using a smaller number of species (up to 11; ). Analyzing a set of 19 Murinae species, FREEDA found 16 orthologs with 94% coding sequence coverage (median values; ). Consistent with our previous findings of pervasive evolutionary innovation across the rodent centromere , FREEDA found that 36/104 genes have likely evolved under positive selection ( and ). Corroborating our previous results, FREEDA detected statistical signatures of positive selection in genes encoding CENP-C, CENP-I, CENP-T, HJURP, INCENP, MIS18BP1, KNL1, and SGO2. In contrast, DSN1 and HEC1 did not show statistical signatures of positive selection. This discrepancy is likely due to a difference in coding sequence coverage between the analyses (higher in FREEDA) or it reflects a higher statistical power due to more orthologs (up to 19 used by FREEDA), which facilitates distinguishing between positive selection and relaxation of purifying selection. This high statistical power revealed several previously unknown targets of positive selection, including components of the fibrous corona, which helps capture microtubules (CENP-F, SPINDLY, ZWILCH, ROD, NUP85, NUP98, and ELYS; reviewed in ), microtubule motors (CENP-E, KIF2B, and KIF18A), and protein kinases (AURKC and HASPIN). To further validate our findings, we repeated the analyses with rat as reference species. Since the quality of the available rat genome annotation is lower than that of mouse, FREEDA was able to collect reliable input data for only 89/104 genes. As expected, we found statistical evidence (or lack thereof) of positive selection in almost exactly the same genes as when using mouse as reference (85/89 genes; see Discussion; ). Overall, these tests show that despite its simplicity for the user, FREEDA is a fully functional and dependable tool to detect statistical signatures of positive selection.
To test if FREEDA can help derive evolution-guided hypotheses, we leveraged its ability to map residues that have likely evolved under positive selection onto protein structures. We found statistical evidence of positive selection within ancient (retained across long evolutionary timescales) protein domains of centromeric proteins, suggesting that adaptive evolution shaped essential protein functions . For instance, we detected residues with high probability of having evolved under positive selection in the ancient Yippee domain of MIS18β (encoded by Oip5 in mouse), which participates in centromere chromatin assembly (reviewed in ). In addition to its divergent N- and C-termini, one of the most likely adaptive residues (arginine at position 76 in mouse; probability = 0.98) is located within the loop-forming CXXC motif of the Yippee domain , which is required for MIS18 complex assembly at centromeres . Similarly, we found strong statistical evidence of positive selection in one of the loops of an ancient protein kinase domain (reviewed in ) in the meiosis-specific Aurora kinase C (AURKC, asparagine at position 150 in mouse, probability = 0.98; ), which helps correct erroneous kinetochore-microtubule attachments . In contrast, we found no recurrent changes in the related AURKA and AURKB kinases . These data suggest that positive selection has uniquely tuned the kinase activity of the specialized meiotic Aurora kinase, consistent with previous reports of adaptive evolution of reproduction genes . Finally, we found statistical evidence that positive selection shaped evolution of the ancient double RWD domain (RING-WD-DEAD; ) of CENP-O, which regulates kinetochore–microtubule attachments by forming the CENP-OPQUR complex . RWD domains are prevalent structural modules that facilitate protein–protein interactions across the centromere . CENP-O shares a high structural similarity with its binding partner CENP-P, which also shows statistical signatures of positive selection within its double RWD domain . Furthermore, some of the residues that have evolved under positive selection with the highest probability are located in or near loops and turns flanking highly structured α-helices and β-sheets in CENP-O and -P C-terminal RWD domains ( and ). Based on these results, we propose that positive selection has regulated essential functions of centromeric proteins by acting on loops and turns of ancient domains, consistent with previous reports of frequent innovation of flexible regions in other proteins . Altogether, we demonstrate that FREEDA can help derive evolution-guided hypotheses by highlighting protein domains whose function has likely been shaped by adaptive evolution.
Each of the proteins discussed above (MIS18β, AURKC, and CENP-OP) functions as part of a complex. To infer mechanisms regulated by positive selection in this context, we aligned FREEDA-annotated protein structure predictions of mouse proteins to experimentally solved structures of their orthologs in complex with binding partners (see Materials and methods for details). Two loops formed by CXXC motifs within the Yippee domain of MIS18β together give rise to a tetrahedral module whose four conserved cysteines bind a zinc ion , likely stabilizing protein conformation . Aligning mouse MIS18β to the crystal structure of the fission yeast MIS18 Yippee-like domain shows that the side chain of arginine at the positively selected position 76 in mouse likely faces the opening of the tetrahedral module. This finding is consistent with XX residues regulating the function of CXXC motifs in other proteins . Alternatively, R76 could mediate MIS18α and MIS18β heterodimerization . Therefore, we hypothesize that positive selection favored amino acid changes within the CXXC motif to modulate MIS18 complex stability. Consistent with functional innovation of CXXC motifs, we also found recurrently changing residues within the second CXXC motif of MIS18β (glycine at position 135 in mouse; ) and in the first CXXC motif of its binding partner MIS18α (serine at position 57 in mouse; ), albeit the probability that they have evolved under positive selection was lower (probabilities = 0.88 and 0.77, respectively). These data suggest that positive selection in the loops of the ancient Yippee domains regulated centromere assembly by modulating stability of the MIS18 complex. AURKB and AURKC kinase activity requires binding to a conserved domain of INCENP (INner-CENtromere Protein; reviewed in ). Aligning the Mus musculus (Mm) AURKC protein kinase domain and MmINCENP AURK-binding domain to the crystal structure of the orthologous human domains shows the side chain of positively selected asparagine at position 150 in mouse in close proximity to tyrosine at conserved position 827 in MmINCENP. This finding suggests modulation of INCENP binding and, therefore, kinase activity by positive selection . The rodent AURKC activation loop also contains a recurrently changing, albeit less likely adaptive residue (serine at position 156 in mouse; probability = 0.77; ) whose side chain reaches toward the AURKC ATP-binding site (marked by the inhibitor BRD-7880; ; ). These data suggest that positive selection in the loop of the ancient protein kinase domain of AURKC regulated meiotic functions by modulating kinase activity. Double RWD domains mediate the formation of CENP-OP heterodimers, allowing recruitment of the CENP-OPQUR complex to centromeres . Aligning the FREEDA-annotated CENP-O and -P C-terminal RWD domains to the experimentally solved human CENP-OPQUR complex suggests that positive selection shaped opposite sides of the CENP-OP heterodimer and therefore is unlikely to have impacted heterodimerization. In yeast, C-terminal RWD domains of CENP-O and -P orthologs bind to CENP-Q and -U orthologs to form the COMA complex . We were unable to reliably align mouse CENP-Q and -U to the human CENP-OPQUR complex, likely due to long unstructured regions in CENP-Q and -U, but the striking pattern of likely adaptive residues in C-terminal RWD domains facing the outside of the heterodimer suggests that positive selection regulated binding to nearby centromeric components . We find statistical signatures of positive selection in CENP-Q and -U in rodents , suggesting that positive selection regulated interactions between CENP-OPQUR complex components. Altogether, these analyses of multiple centromere proteins demonstrate how FREEDA-annotated structures can be used to generate hypotheses for how positive selection might have regulated essential protein functions.
To test our hypothesis that loops and turns in ancient protein domains regulate their essential functions, we chose to focus on CENP-O because FREEDA suggests that positive selection operated on residues flanking α-helices and β-sheets of both rodent and primate C-terminal RWD domains , and centromere binding provides a straightforward functional assay. We used mouse oocytes for these experiments because they are an established model system for centromere drive, the most likely selective pressure sculpting evolution of centromeric proteins, and thus a natural context to probe for functional protein innovation. To create an evolutionary mismatch (see Introduction), we introduced GFP-tagged full-length mouse (control) or rat (divergent) CENP-O at similar expression levels . Mouse CENP-O localized to centromeres as expected, but rat CENP-O was nearly undetectable at mouse centromeres , indicating functional innovation in centromere binding. To test if the C-terminal RWD domain is responsible for that innovation, we compared three chimeric rat CENP-O constructs with different regions of mouse CENP-O: N-terminal (N-terminal tail and N-terminal helix), central (N-terminal RWD domain and central helix), or C-terminal (C-terminal RWD domain; and ). Only the mouse C-terminal RWD domain could rescue, albeit not fully, the localization of rat CENP-O to mouse centromeres. In an inverse experiment, a chimera of mouse CENP-O with the rat C-terminal RWD domain failed to localize to mouse centromeres ( and ). Together, these results demonstrate that mouse-specific innovation in the C-terminal RWD domain is required for CENP-O binding to mouse centromeres. Within this domain, 10 out of 13 residues that differ between mouse and rat are putatively adaptive (probability ≥ 0.5; ). Almost all (9/10) of these residues flank highly structured α-helices or β-sheets (±1 amino acid), consistent with our hypothesis that positive selection drives functional innovation of ancient domains in centromeric proteins by acting on their loops and turns. Swapping five of the most likely adaptive residues in the mouse C-terminal RWD domain to those found in rat did not, however, reduce centromere localization of mouse CENP-O . Similarly, swapping equivalent rat residues within the C-terminal RWD domain of rat CENP-O to mouse-specific ones (in addition to six mutations in other parts of the protein) did not restore its centromere localization . These analyses highlight the difficulty in attributing innovation to specific residues given the number of possible combinations as well as the potential for epistasis . Altogether, we show that our fully automated molecular evolution pipeline can guide experimental testing of functional protein innovation.
The motivation to develop FREEDA was to catalyze participation of the cell biology community in testing functional consequences of protein innovation. We demonstrate that detecting statistical signatures of positive selection, which implicates functional innovation, can be fully automated by compiling widely used bioinformatic and molecular evolution tools into a single pipeline . FREEDA’s simple and user-friendly GUI makes it a suitable entry point for experimentalists who may have limited programming skills . Moreover, by leveraging the ever-growing pool of newly sequenced but not yet annotated genomic assemblies, FREEDA bypasses the requirement for obtaining tissue samples and cloning the genes of interest to have sufficient numbers of orthologs from closely related species to detect signatures of positive selection. Nevertheless, as with any fully automated tool, FREEDA has limitations. First, by inferring orthologs based on the annotated reference sequence, rather than experimentally validated transcripts, FREEDA does not account for tissue-specific splicing, shifts in intron–exon boundaries, or the use of alternative exons (see ). Despite this caveat, using independently annotated rat sequences as reference led to the same result as using mouse annotations in 85/89 centromeric genes (note that relatively poorer rat genome annotation quality prevented reliable input generation for 15 genes; ). Nevertheless, isoforms that substantially differ from the reference coding sequence might interfere with the accurate detection of positive selection. As an example, annotated variants of the rat NUP37 nucleoporin substantially differ at their C-termini from the reference mouse NUP37 sequence, suggesting the use of alternative exons (additional supplementary materials), which likely led to inconsistent signals of positive selection (P = 0.510 with mouse as reference vs. P = 0.0499 with rat as reference; ). Second, to prioritize computational speed and reduce output complexity, FREEDA does not test for possible recombination events known to increase the probability of false positives when recombination rates are high . While estimated recombination frequencies are lower in vertebrates and insects as compared to yeast and protozoa, we cannot exclude the influence of recombination events . Third, while FREEDA can robustly resolve gene duplications present in the ancestor of a selected taxon (e.g., primates), caution is advised when analyzing lineage-specific genes. For example, primate MICA (MHC class I chain-related gene A) is known to have duplicated from MICB in the common ancestor of hominoids and Old World monkeys . Therefore, searching for MICA orthologs across the entire primate taxon yields MICB coding sequences in New World monkeys (additional supplementary materials). In case lineage-specificity is suspected, we suggest using the “Subgroup” option (currently supporting: hominoidea , catarrhini , caniformia , and melanogaster subgroup) and/or the “Exclude species” option . Fourth, FREEDA is designed to test for signatures of recurring (pervasive) positive selection acting on the entire taxon (e.g., primates) rather than episodic selection that may have led to adaptation in a specific branch (e.g., hominoids). Nevertheless, using the aforementioned “Subgroup” or “Exclude species” options allows narrowing of the phylogenetic window if needed. Finally, mapping FREEDA’s results onto protein structures is not yet fully supported for carnivores ( Carnivora ) and birds ( Phasianidae ). Our analyses of genes with known evolutionary histories demonstrate that FREEDA is reliable. Building on this validation, we provide the most detailed characterization of signatures of positive selection at rodent centromeres to date. Consistent with previous analyses , we infer pervasive evolutionary innovation in domains of centromeric proteins that do not directly touch DNA, such as RWD . Therefore, our data support the idea that fitness costs of centromere drive are suppressed by innovation in protein–protein interactions as well as protein–DNA interactions . Furthermore, by mapping regions that are likely under positive selection onto protein structures, we derive a hypothesis that positive selection acting on loops and turns of ancient domains impacts essential protein functions. For example, recurrent amino acid changes within the loops formed by CXXC motifs could regulate MIS18 complex formation. Similarly, recurrent changes in loops of the ancient AURKC protein kinase domain could modulate kinetochore–microtubule attachment dynamics, specifically in meiosis I. Both centromere assembly and microtubule detachment (in meiosis I) represent mechanisms potentially hijacked by selfish centromeres . These analyses provide a starting point for future experiments probing the functional impacts of innovation, including testing whether positive selection reduces fitness costs associated with centromere drive (reviewed in ). Previous experiments in fruit flies, using evolutionarily mismatches of the L1 loop within the ancient histone fold domain of Cid CENP-A , suggested functional innovation in a DNA-binding domain of a centromere protein . Here, we propose that positive selection in CENP-O may have regulated centromere binding via recurrent changes in loops and turns within an ancient C-terminal RWD domain , which does not interact with DNA . While we are unable to pinpoint the exact combination of residues responsible for functional innovation, the observation that most recurrently changing residues are within regions that flank the highly structured α-helices or β-sheets of that domain supports our hypothesis. CENP-O is expected to dock the CENP-OPQUR complex to centromeres , promoting kinetochore–microtubule attachment stability . Therefore, we propose that innovation in the CENP-O C-terminal RWD domain modulated interactions with other centromeric components (possibly CENP-Q and -U) necessary to form a stable CENP-OPQUR complex at centromeres , potentially stabilizing kinetochore–microtubules to counteract destabilizing activities associated with driving centromeres . Future work using centromere drive models (reviewed in ) will be needed to experimentally test this idea. Overall, we show how FREEDA can help derive evolutionary hypotheses and guide experimental testing of functional innovation, starting from just a gene name, making it a powerful tool for incorporating evolutionary analyses into cell biology research and generating new insights into essential cellular processes.
Resources and datasets FREEDA was written in Python and compiled into a stand-alone application using pyinstaller ( https://pyinstaller.org/en/stable/ ). Core packages used for the compilation were installed using standard package managers: pip ( https://pypi.org/project/pip/ ) and conda ( https://docs.conda.io/en/latest/ ). All selected genomic assemblies and Ensembl releases used to generate datasets are listed in . All data were collected using desktop computers: iMac (late 2015) with MacOS Monterey 12.6 (8 GB RAM; 4 CPU cores; 2.8 GHz Quad-Core Intel Core i5) and iMac (2017) with MacOS Ventura 13.2 (16 GB RAM, 2 CPU cores; 2.3 GHz Intel Core i5), or a laptop MacBook Pro (mid-2014) with MacOS Mojave 10.14.6 (16 GB RAM, 2 CPU cores; 3 GHz Intel Core i7). Input extraction and identification of potential orthologous exons FREEDA can be downloaded from an open-source repository: https://github.com/DDudka9/freeda/releases . When running the app for the first time, MacOS users will be prompted to download PyMOL, which renders a 3D result of the FREEDA analysis. To run the pipeline, the user needs to provide at least one gene name, select a reference species, and select a location where all the data will be stored. At least 100 GB of storage space is needed to analyze a single vertebrate taxon (e.g., primates), 20 GB is sufficient to analyze only flies, and 500 GB is recommended to analyze all taxons. Optionally, the user can (1) specify the coordinates of residues or domains of interest that will be labeled on the protein structure prediction, (2) customize the BLAST search and ortholog finding (see below), (3) exclude selected species from the analysis, and (4) narrow the analysis to a specific subgroup (e.g., catarrhini ). Advanced users can also specify the codon frequency model used (F3X4 or F3X4 and F61; see PAML manual for details: http://abacus.gene.ucl.ac.uk/software/pamlDOC.pdf ). The pipeline starts with downloading the reference genome (using NCBI Datasets ) and then retrieving all possible UniProt IDs for a protein encoded by the gene of interest and matching them to AlphaFold database entries . Next, FREEDA extracts protein sequence and coding sequence (using pyensembl package: https://github.com/openvax/pyensembl ) from the Ensembl database and exon sequences and gene sequence (using pybedtools ) from the downloaded reference genome. Visualization of residues that have likely evolved under positive selection requires that the protein sequence of the structural prediction and the protein sequence from the Ensembl database are identical (tested using the Biopython package; ). If the proteins are not identical, FREEDA performs the analysis without mapping the residues onto structure predictions. The first run triggers downloading of the selected reference genome (e.g., human), followed by the genomes of closely related species (e.g., Simiiformes ), and then building of local BLAST databases (using BLAST + applications; ). FREEDA queries these databases to find genomic coordinates of putative orthologous regions using tblastn algorithms (default identity threshold is set at 60% [or 30% for Drosophila ] but can be increased to 80% [or 60% for Drosophila ] by selecting the advanced option “Common domains expected”) and retrieves corresponding nucleotide sequences (using pybedtools) from downloaded related genomes. Overall, these features allow fully automated generation of the input data needed to find orthologous coding sequences in non-annotated genomes. Finding orthologous exons FREEDA performs a multiple sequence alignment of each region found during the BLAST search to the reference coding sequence, BLAST sequences stitched together, the genomic locus these sequences reside in (contig), and the reference gene sequence using MAFFT (Multiple Alignment using Fast Fourier Transform; ). Regions aligning to both the coding sequence and the reference gene sequence are considered putative exons. To determine if the putative exon is syntenic (resides in a homologous locus), the flanking sequence is compared with the introns of the reference gene separately at 5′ and 3′ ends. An exon is considered syntenic if at least one of the flanking regions is at minimum 75% (60% for Drosophila ) identical to the reference intron over a stretch of at least 50 bp (30 bp for Drosophila ). The identity is calculated as the Hamming distance (the number of different bases in a pair-wise comparison of two aligned sequences; ; divided by the sequence length). The putative exon is called as not syntenic and discarded from the analysis if none of the flanking regions reaches the identity threshold and the exon itself is <80% (70% for Drosophila ) identical to the reference exon over a stretch of the first 30 bp. Since introns are generally less conserved than exons, when the identity of a flanking region is uncertain (66–75%; 50–60% for Drosophila ), a longer sequence is compared. Lowered values for detecting synteny in Drosophila genomes are due to an observed high divergence of intragenic regions and high rates of indels within orthologous loci. To increase stringency in detecting synteny and reduce uncertainty resolving segmental gene duplications, the user can additionally select an advanced option “Duplication expected” that penalizes any exon that is syntenic only at one end (e.g., recent segmental duplication whose introns have not yet diverged significantly). Segmental duplications that lead to duplicated genes residing next to each other (tandem duplications) will not only have similar flanking regions but might also be difficult to align if residing on the same contig. To ensure robust analysis of tandem duplications, the “Tandem duplication expected” option limits the flanking region of each blast hit (leading to smaller contigs), decreasing the chance of tandemly duplicated genes residing on the same contig. In addition, to avoid retroduplications (mRNA-derived gene duplications; ), FREEDA always discards exons that are at least 80% (70% for Drosophila ) identical to the reference exon but lack the flanking regions (intron-less). To preserve intron–exon boundaries, each putative exon is given a number and directly aligned to a reference exon of the same number. Therefore, exons do not need to reside on the same contig to form a complete coding sequence, which is helpful when querying genomic assemblies with short contigs. Very small exons (microexons; reviewed in ) shorter than 18 bp cannot be reliably aligned and are discarded from the analysis. If the same putative exon is found on different contigs (e.g., due to a duplication), the contig containing fewer putative exons is discarded. If both contigs carry the same number of putative exons (likely due to heterozygosity of the orthologous locus or a very recent duplication), these are compared to corresponding reference exons, and the contig with a higher overall identity of exons is considered orthologous. Rare cases of mistakes in aligning intron–exon boundaries may lead to indels, which are resolved at later stages (e.g., the entire codon is removed in case of a 1-nt indel). To allow manual review, all the above-mentioned steps are logged and all the intermediate alignments are saved as raw data (“Raw_data” folder). Manual verification of detected orthologs Genomic location and nucleotide sequence identity of >120 FREEDA-identified orthologs representing 20 randomly selected genes (five per clade: Murinae , Simiiformes , Carnivora , Phasianidae ) were compared with their annotations found in the Ensembl database. Genomic location (contig number) logged in the “FREEDA-current-date.log” file was compared to that of the expected flanking genes that were also analyzed by FREEDA. Nucleotide sequence identity with Ensembl-annotated orthologous coding sequences was measured using pairwise alignment and MAFFT protocol designed to limit over-aligning errors . This approach generates large indels in the alignment and facilitates detection of alternative exons and start codons. Each alignment was visually inspected without manual curation. See detailed analysis and commentary in . Manual verification of known recent gene duplications The ability to distinguish tandem duplication ( H4C1 from H4C2 ) and recent retro-duplication ( KIF4A from KIF4B ) was tested by manual BLAST (blastn) of the nucleotide sequence of each ortholog identified by FREEDA (“GENE_raw_nucleotide_alignment.fasta” file) against the primate NCBI gene database ( Simiiformes ; taxid: 314294). For each gene, an orthologous sequence was always the highest scoring hit (by similarity) as opposed to a paralogous sequence. Exceptions were sister species Aotus nancymaae and Callithrix jacchus , whose identified KIF4B coding sequences were more similar to KIF4A than KIF4B . However, all the KIF4B orthologs detected by FREEDA were intron-less, consistent with KIF4B being a primate-specific retro-duplication of KIF4A , which FREEDA called correctly. Therefore, we are confident that FREEDA identified KIF4B orthologs for all species and not KIF4A paralogs (additional supplementary materials). Building the multiple sequence alignment and phylogenetic gene tree Detection of recurring amino acid substitutions requires a gapless, in-frame multiple sequence alignment. To avoid large gaps (suggesting incomplete coding sequences), FREEDA first removes entire coding sequences that are shorter than 90% compared with the reference sequence. To ensure high-quality alignment of the remaining sequences, we tested the commonly used aligners: MUSCLE , PRANK , MACSE , and MAFFT . We decided to use a modified MAFFT protocol designed to limit over-aligning errors as we found it both fast and accurate. To curate the alignment, FREEDA removes insertions that are defined as regions missing in the reference coding sequence and deletes stop codons (including premature ones). At this point, coding sequences that are <69% (60% for Drosophila ) identical to the reference sequence are discarded as likely too divergent to produce a reliable alignment (based on ; ) or misaligned. Additionally, remaining small gaps (deletions) and ambiguously aligned codons are removed from the alignment (using Gblocks; ; ). Note that MAFFT is not codon-aware, which allows aligning incomplete sequences (e.g., missing some of the exons or containing indels) that cannot be expected to have a length of multiplication of 3. Therefore, to ensure that the aligning process and alignment curation did not alter the open reading frame of the aligned sequences, FREEDA compares the identity of the translated reference sequence within the curated alignment to the original reference protein sequence from the Ensembl database. Once 100% identity is confirmed, FREEDA translates the orthologous sequences and checks if >10 contiguous non-synonymous substitutions compared with the reference sequence are present. FREEDA considers such rare cases as likely frameshift events and removes the entire sequences from the alignment. Final in-frame multiple nucleotide sequence alignment is then used to build a phylogenetic gene tree (using RAxML; ), which guides the widely used CODEML program from the PAML suite to infer the enrichment of recurrent non-synonymous substitutions suggestive of positive selection. We strongly urge the user to manually verify the final protein alignment (“Results-Current-Date/Results/Protein_alignments/GENE_protein_alignment.fasta” file), ensuring that there are no obvious misaligned regions before considering the results of the PAML analysis (e.g., using the free software Unipro UGENE; ). In case of apparent misalignments, we suggest simply rerunning the analysis using the “Exclude species” option . An example of a misaligned sequence can be found in the documentation. Detection of positive selection To detect statistical signatures of positive selection, FREEDA relies on the rate ratio of non-synonymous (dN) to synonymous (dS) substitutions (dN/dS > 1 suggests positive selection). However, most genes contain conserved regions that evolve under purifying selection (dN/dS < 1), which usually decreases gene-wide dN/dS below 1. Therefore, to find specific regions that likely evolved under positive selection, FREEDA uses “site models” of the CODEML program that allow for varying dN/dS between different codons. Each model describes a set of parameters (including dN/dS per codon; for details, see the official PAML guide, http://abacus.gene.ucl.ac.uk/software/pamlDOC.pdf , or a beginners guide; ) and either allows for sites (codons) with a dN/dS ratio of >1 (signature of positive selection; M8 and M2a models) or not (null hypothesis; M7 and M1a models). Using a maximum likelihood approach, CODEML then fits the parameters estimated from the data to each model. Significantly more likely fit (based on the likelihood ratio test) to the model that allows for codons with dN/dS > 1 indicates the presence of sites that have likely evolved under positive selection. Bayesian statistics (Bayes Empirical Bayes) are then used to estimate probabilities of positive selection acting on specific codons. FREEDA outputs the key results of the CODEML analysis within the GUI’s “Results window” (likelihood ratio test value for M7 vs. M8 comparison, P value, and number of codons with the highest probability to have evolved under positive selection). Additionally, the results of the M1a vs. M2a comparison and the identity of specific codons under positive selection are saved in an Excel sheet (“Results-Current-Date/Results/Results_sheet” folder). Visualization of residues under positive selection FREEDA maps the protein sequence of the reference species from the multiple sequence alignment to the expected reference protein sequence and, if appropriate, introduces gaps that represent residues excluded from the analysis (“GENE_protein_alignment.fasta” file). Based on that mapping, FREEDA provides both 2D and 3D visual representations of residues that likely evolved under positive selection. 2D bar graphs are provided in the “Results-Current-Date/Results/Graphs” folder. These graphs display the positions of recurrently changing residues (“Posterior mean omega,” top) and the probability of positive selection acting on each codon (“Prob. positive selection,” middle, probability 0.7–1.0; “High prob. positive selection,” bottom, probability 0.9–1.0). Codons excluded from the analysis are marked in gray. 3D representation of the most likely adaptive residues is found in the “Results-Current-Date/Results/Structures” folder, provided that the prediction model from the AlphaFold database matches the protein sequence extracted from the Ensembl database. For clarity, only residues with the highest probability (≥0.9) of having evolved under positive selection are mapped and their side chains are shown. The residues excluded from the analysis are colored in gray, and the N-terminal and C-terminal ends are labeled. Additionally, any domain annotation available in the Interpro database is automatically marked with a distinct color and labeled allowing quick visual identification. Manual alignment of structural prediction models FREEDA-annotated protein structure prediction models from AlphaFold designated by their UniProt entries (MmMIS18β—A2AQ14; MmAURKC—O88445; MmINCENP—Q9WU62; MmCENP-O—Q8K015; MmCENP-P—Q9CZ92) were aligned to PDB entries (SpMIS18— 5HJ0 ; HsAURKC— 6GR8 ; HsINCENP— 6GR8 ; HsCENP-OP— 7PB8 ) using an aligner module in PyMOL. Briefly, a FREEDA-annotated structure (e.g., “Aurkc_Mm.pse” found in “Results-Current-Date/Results/Structures”) was opened in PyMOL, the selected PDB entry was downloaded (e.g., “fetch 6GR8”), and the two structures were aligned (e.g., “align Aurkc_Mm, 6GR8”). The align module first aligns protein sequences and then superimposes their structures, returning RMSD (Root-Mean-Square-Deviation). Lower RMSD values indicate better alignment. All alignments presented here returned RMSD below 2 Å. Generation of CENP-O constructs All CENP-O coding sequences were cloned from testis or liver samples. The use of rat CENP-O to represent a divergent ortholog was motivated by the availability of a rat ( Rattus norvegicus ) tissue sample for cloning. Chimeric CENP-O constructs were designed based on their 3D structure (AlphaFold database). Tissue was mechanically homogenized and total mRNA was isolated using TRIzol reagent (15596026; Invitrogen); cDNA was prepared using reverse transcription (18080051; SuperScript III First-Strand Synthesis System), amplified using construct-specific PCR primers (KK2502; KAPA HiFi Hot Start plus dNTPs; Roche), and inserted into the pGEMHE plasmid backbone (638948; In-fusion kit; Takara). Primers were designed using SnapGene (Dotmatics) software and are listed in . Each CENP-O construct was tagged with GFP at the C-terminus, separated by a linker of five glycines. Site-directed mutagenesis was performed using the Quik-Change Multisite Directed Mutagenesis kit (200515; Agilent) to introduce R225G, C226T, T228A, N247G, and P261H mutations. Rat CENP-O with 11 mouse point mutations was synthesized (Genewiz). The identity of all constructs was confirmed using Sanger sequencing of the entire coding sequence, including the reporter gene. Oocyte isolation, microinjection, and in vitro maturation Mus musculus mice (CF-1 strain) were purchased from Envigo NSA stock # 033. Females were primed with 5 U of pregnant mare somatic gonadotropin (367222; Calbiochem) injected into the intraperitoneal cavity 44–48 h prior to oocyte collections to induce superovulation. The ovaries were isolated using M2 medium (M7167; Sigma-Aldrich) supplemented with 2.5 mM of the maturation-blocking phosphodiesterase 3 inhibitor milrinone (M4659; 2.5 mM; Sigma-Aldrich Milipore). Germinal-vesicle oocytes were collected, denuded mechanically from cumulus cells, and incubated for at least 1 h prior to microinjection on a hot plate (38°C) under mineral oil (9305; FUJIFILM Irvine Scientific). Oocytes were then microinjected with ∼5 pl of mRNAs in M2 medium with 2.5 mM milrinone and 3 mg/ml BSA at RT with a micromanipulator TransferMan NK 2 (Eppendorf) and a picoinjector (Medical Systems Corp). Oocytes were then incubated in 30–50 μl drops of Chatot-Ziomek-Bavister medium (MR019D; Thermo Fisher Scientific) under mineral oil (M5310; Sigma-Aldrich Milipore) at 37.8°C and 5% CO2 (Airgas) for 16 h to allow protein expression. The concentration of mRNA (15–30 ng/μl) used was selected to ensure similar cytoplasmic expression. mRNAs were synthesized using the T7 mScriptTM Standard mRNA Production System (C-MSC100625; CELLSCRIPT). Immunofluorescence imaging Oocyte maturation was induced in vitro by washing out milrinone 7.5 h before fixation. MI oocytes were fixed in freshly prepared 2% paraformaldehyde in PBS (pH 7.4) with 0.1% Triton X-100 for 20 min at RT, permeabilized in PBS with 0.1% Triton X-100 for 15 min at RT, placed in blocking solution (PBS with 0.3% BSA and 0.01% Tween-20) overnight at 4°C, incubated 1 h with primary antibody in blocking solution, washed three times for 15 min each, incubated 1 h with secondary antibody, washed three times for 15 min each, and mounted in Vectashield with DAPI (H-1200; Vector) to visualize chromosomes. Centromeres were labeled with CREST (human anti-human anti-centromere antibody, 1:200, HCT-0100; Immunovision) and an Alexa Fluor 594–conjugated goat anti-human secondary antibody (A-110014; Thermo Fisher Scientific). Confocal images were collected as 31 z-stacks at 0.5-μm intervals to visualize the entire meiotic spindle, using a confocal microscope (DMI4000B; Leica) equipped with a 63× 1.3 NA glycerol-immersion objective lens, an xy piezo Z stage (Applied Scientific Instrumentation), a spinning disk (Yokogawa Corporation of America), and an electron multiplier charge-coupled device camera (ImageEM C9100-13; Hamamatsu Photonics), controlled by MetaMorph software (Molecular Devices). Excitation was done with a Vortran Stradus VersaLase 4 laser module with 405-, 488-, 561-, and 639-nm lasers (Vortran Laser Technology). Panels of microscopic images were prepared using ImageJ (National Institutes of Health) free software. Automated image analysis Confocal images were analyzed using a custom-built Python-based automated program “Centrocalc” available here: https://github.com/DDudka9/Centrocalc . First, the program performs whole chromosome segmentation using the chromosome channel as a mask by grayscale dilation at a width of 16 pixels to ensure the centromeres are included. Then, a threshold is calculated using the Iterative Self-Organizing Data Analysis Technique method. If no chromosome channel is given, the entire cell is considered. Second, the program identifies centromeres using an approach based on . A difference of Gaussians algorithm is used to isolate spots of 150 nm. Third, a local maxima algorithm is used to identify centromeres. Up to 38 spots are chosen (the mouse 2n genome contains 40 chromosomes), separated by a minimum of two pixels by Chebyshev distance. The spots must be away from the edges of the image (20 pixels in x and y; 2 pixels in z). Fourth, 3D ellipsoid regions of interest (4 × 4 × 3 pixels) are drawn using the local maxima. Background regions of interest (ROIs) are drawn as volumes around the centromere ROIs (expanding the 3D ellipsoids by 1 pixel in all dimensions). Overlapping centromere ROIs are resolved by distance, where each pixel is assigned to the closest maxima. Fifth, centromere and background intensities are calculated as average grayscale pixel values and saved. ImageJ ROI files are also created to reference back to the original images. Modifying the Centrocalc source code can customize most of the described parameters. Statistical analysis Statistical analysis was performed using GraphPad Prism 9.3.1 (GraphPad Software). Statistical significance was assessed using a two-tailed T test (data distribution was normal under Kolmogorov–Smirnov and Shapiro–Wilk tests) or one-way ANOVA with Tukey’s multiple comparison test (data distribution was assumed normal but was not formally tested). Graphs display means with standard deviations. P values indicated on graphs are P ≥ 0.05, not significant (ns); P < 0.05, *; and P < 0.0001, ****. Online supplemental material Comparison of FREEDA with other pipelines is outlined in . Results of FREEDA validation are listed in (validation of ortholog detection based on Ensembl database), (accuracy in detecting positive selection using published datasets), , and (accuracy in detecting sites under positive selection based on published data). Results of positive selection detection in rodent centromere and kinetochore proteins are shown in . The list of all genomic assemblies used by the FREEDA pipeline and the core packages comprising the pipeline are listed in . lists all primers used to generate CENP-O constructs. , , , and show structural analyses of selected centromeric proteins. Experimental analyses of rodent CENP-O are shown in , , and . indicates sites that differ between mouse and rat CENP-O.
FREEDA was written in Python and compiled into a stand-alone application using pyinstaller ( https://pyinstaller.org/en/stable/ ). Core packages used for the compilation were installed using standard package managers: pip ( https://pypi.org/project/pip/ ) and conda ( https://docs.conda.io/en/latest/ ). All selected genomic assemblies and Ensembl releases used to generate datasets are listed in . All data were collected using desktop computers: iMac (late 2015) with MacOS Monterey 12.6 (8 GB RAM; 4 CPU cores; 2.8 GHz Quad-Core Intel Core i5) and iMac (2017) with MacOS Ventura 13.2 (16 GB RAM, 2 CPU cores; 2.3 GHz Intel Core i5), or a laptop MacBook Pro (mid-2014) with MacOS Mojave 10.14.6 (16 GB RAM, 2 CPU cores; 3 GHz Intel Core i7).
FREEDA can be downloaded from an open-source repository: https://github.com/DDudka9/freeda/releases . When running the app for the first time, MacOS users will be prompted to download PyMOL, which renders a 3D result of the FREEDA analysis. To run the pipeline, the user needs to provide at least one gene name, select a reference species, and select a location where all the data will be stored. At least 100 GB of storage space is needed to analyze a single vertebrate taxon (e.g., primates), 20 GB is sufficient to analyze only flies, and 500 GB is recommended to analyze all taxons. Optionally, the user can (1) specify the coordinates of residues or domains of interest that will be labeled on the protein structure prediction, (2) customize the BLAST search and ortholog finding (see below), (3) exclude selected species from the analysis, and (4) narrow the analysis to a specific subgroup (e.g., catarrhini ). Advanced users can also specify the codon frequency model used (F3X4 or F3X4 and F61; see PAML manual for details: http://abacus.gene.ucl.ac.uk/software/pamlDOC.pdf ). The pipeline starts with downloading the reference genome (using NCBI Datasets ) and then retrieving all possible UniProt IDs for a protein encoded by the gene of interest and matching them to AlphaFold database entries . Next, FREEDA extracts protein sequence and coding sequence (using pyensembl package: https://github.com/openvax/pyensembl ) from the Ensembl database and exon sequences and gene sequence (using pybedtools ) from the downloaded reference genome. Visualization of residues that have likely evolved under positive selection requires that the protein sequence of the structural prediction and the protein sequence from the Ensembl database are identical (tested using the Biopython package; ). If the proteins are not identical, FREEDA performs the analysis without mapping the residues onto structure predictions. The first run triggers downloading of the selected reference genome (e.g., human), followed by the genomes of closely related species (e.g., Simiiformes ), and then building of local BLAST databases (using BLAST + applications; ). FREEDA queries these databases to find genomic coordinates of putative orthologous regions using tblastn algorithms (default identity threshold is set at 60% [or 30% for Drosophila ] but can be increased to 80% [or 60% for Drosophila ] by selecting the advanced option “Common domains expected”) and retrieves corresponding nucleotide sequences (using pybedtools) from downloaded related genomes. Overall, these features allow fully automated generation of the input data needed to find orthologous coding sequences in non-annotated genomes.
FREEDA performs a multiple sequence alignment of each region found during the BLAST search to the reference coding sequence, BLAST sequences stitched together, the genomic locus these sequences reside in (contig), and the reference gene sequence using MAFFT (Multiple Alignment using Fast Fourier Transform; ). Regions aligning to both the coding sequence and the reference gene sequence are considered putative exons. To determine if the putative exon is syntenic (resides in a homologous locus), the flanking sequence is compared with the introns of the reference gene separately at 5′ and 3′ ends. An exon is considered syntenic if at least one of the flanking regions is at minimum 75% (60% for Drosophila ) identical to the reference intron over a stretch of at least 50 bp (30 bp for Drosophila ). The identity is calculated as the Hamming distance (the number of different bases in a pair-wise comparison of two aligned sequences; ; divided by the sequence length). The putative exon is called as not syntenic and discarded from the analysis if none of the flanking regions reaches the identity threshold and the exon itself is <80% (70% for Drosophila ) identical to the reference exon over a stretch of the first 30 bp. Since introns are generally less conserved than exons, when the identity of a flanking region is uncertain (66–75%; 50–60% for Drosophila ), a longer sequence is compared. Lowered values for detecting synteny in Drosophila genomes are due to an observed high divergence of intragenic regions and high rates of indels within orthologous loci. To increase stringency in detecting synteny and reduce uncertainty resolving segmental gene duplications, the user can additionally select an advanced option “Duplication expected” that penalizes any exon that is syntenic only at one end (e.g., recent segmental duplication whose introns have not yet diverged significantly). Segmental duplications that lead to duplicated genes residing next to each other (tandem duplications) will not only have similar flanking regions but might also be difficult to align if residing on the same contig. To ensure robust analysis of tandem duplications, the “Tandem duplication expected” option limits the flanking region of each blast hit (leading to smaller contigs), decreasing the chance of tandemly duplicated genes residing on the same contig. In addition, to avoid retroduplications (mRNA-derived gene duplications; ), FREEDA always discards exons that are at least 80% (70% for Drosophila ) identical to the reference exon but lack the flanking regions (intron-less). To preserve intron–exon boundaries, each putative exon is given a number and directly aligned to a reference exon of the same number. Therefore, exons do not need to reside on the same contig to form a complete coding sequence, which is helpful when querying genomic assemblies with short contigs. Very small exons (microexons; reviewed in ) shorter than 18 bp cannot be reliably aligned and are discarded from the analysis. If the same putative exon is found on different contigs (e.g., due to a duplication), the contig containing fewer putative exons is discarded. If both contigs carry the same number of putative exons (likely due to heterozygosity of the orthologous locus or a very recent duplication), these are compared to corresponding reference exons, and the contig with a higher overall identity of exons is considered orthologous. Rare cases of mistakes in aligning intron–exon boundaries may lead to indels, which are resolved at later stages (e.g., the entire codon is removed in case of a 1-nt indel). To allow manual review, all the above-mentioned steps are logged and all the intermediate alignments are saved as raw data (“Raw_data” folder).
Genomic location and nucleotide sequence identity of >120 FREEDA-identified orthologs representing 20 randomly selected genes (five per clade: Murinae , Simiiformes , Carnivora , Phasianidae ) were compared with their annotations found in the Ensembl database. Genomic location (contig number) logged in the “FREEDA-current-date.log” file was compared to that of the expected flanking genes that were also analyzed by FREEDA. Nucleotide sequence identity with Ensembl-annotated orthologous coding sequences was measured using pairwise alignment and MAFFT protocol designed to limit over-aligning errors . This approach generates large indels in the alignment and facilitates detection of alternative exons and start codons. Each alignment was visually inspected without manual curation. See detailed analysis and commentary in .
The ability to distinguish tandem duplication ( H4C1 from H4C2 ) and recent retro-duplication ( KIF4A from KIF4B ) was tested by manual BLAST (blastn) of the nucleotide sequence of each ortholog identified by FREEDA (“GENE_raw_nucleotide_alignment.fasta” file) against the primate NCBI gene database ( Simiiformes ; taxid: 314294). For each gene, an orthologous sequence was always the highest scoring hit (by similarity) as opposed to a paralogous sequence. Exceptions were sister species Aotus nancymaae and Callithrix jacchus , whose identified KIF4B coding sequences were more similar to KIF4A than KIF4B . However, all the KIF4B orthologs detected by FREEDA were intron-less, consistent with KIF4B being a primate-specific retro-duplication of KIF4A , which FREEDA called correctly. Therefore, we are confident that FREEDA identified KIF4B orthologs for all species and not KIF4A paralogs (additional supplementary materials).
Detection of recurring amino acid substitutions requires a gapless, in-frame multiple sequence alignment. To avoid large gaps (suggesting incomplete coding sequences), FREEDA first removes entire coding sequences that are shorter than 90% compared with the reference sequence. To ensure high-quality alignment of the remaining sequences, we tested the commonly used aligners: MUSCLE , PRANK , MACSE , and MAFFT . We decided to use a modified MAFFT protocol designed to limit over-aligning errors as we found it both fast and accurate. To curate the alignment, FREEDA removes insertions that are defined as regions missing in the reference coding sequence and deletes stop codons (including premature ones). At this point, coding sequences that are <69% (60% for Drosophila ) identical to the reference sequence are discarded as likely too divergent to produce a reliable alignment (based on ; ) or misaligned. Additionally, remaining small gaps (deletions) and ambiguously aligned codons are removed from the alignment (using Gblocks; ; ). Note that MAFFT is not codon-aware, which allows aligning incomplete sequences (e.g., missing some of the exons or containing indels) that cannot be expected to have a length of multiplication of 3. Therefore, to ensure that the aligning process and alignment curation did not alter the open reading frame of the aligned sequences, FREEDA compares the identity of the translated reference sequence within the curated alignment to the original reference protein sequence from the Ensembl database. Once 100% identity is confirmed, FREEDA translates the orthologous sequences and checks if >10 contiguous non-synonymous substitutions compared with the reference sequence are present. FREEDA considers such rare cases as likely frameshift events and removes the entire sequences from the alignment. Final in-frame multiple nucleotide sequence alignment is then used to build a phylogenetic gene tree (using RAxML; ), which guides the widely used CODEML program from the PAML suite to infer the enrichment of recurrent non-synonymous substitutions suggestive of positive selection. We strongly urge the user to manually verify the final protein alignment (“Results-Current-Date/Results/Protein_alignments/GENE_protein_alignment.fasta” file), ensuring that there are no obvious misaligned regions before considering the results of the PAML analysis (e.g., using the free software Unipro UGENE; ). In case of apparent misalignments, we suggest simply rerunning the analysis using the “Exclude species” option . An example of a misaligned sequence can be found in the documentation.
To detect statistical signatures of positive selection, FREEDA relies on the rate ratio of non-synonymous (dN) to synonymous (dS) substitutions (dN/dS > 1 suggests positive selection). However, most genes contain conserved regions that evolve under purifying selection (dN/dS < 1), which usually decreases gene-wide dN/dS below 1. Therefore, to find specific regions that likely evolved under positive selection, FREEDA uses “site models” of the CODEML program that allow for varying dN/dS between different codons. Each model describes a set of parameters (including dN/dS per codon; for details, see the official PAML guide, http://abacus.gene.ucl.ac.uk/software/pamlDOC.pdf , or a beginners guide; ) and either allows for sites (codons) with a dN/dS ratio of >1 (signature of positive selection; M8 and M2a models) or not (null hypothesis; M7 and M1a models). Using a maximum likelihood approach, CODEML then fits the parameters estimated from the data to each model. Significantly more likely fit (based on the likelihood ratio test) to the model that allows for codons with dN/dS > 1 indicates the presence of sites that have likely evolved under positive selection. Bayesian statistics (Bayes Empirical Bayes) are then used to estimate probabilities of positive selection acting on specific codons. FREEDA outputs the key results of the CODEML analysis within the GUI’s “Results window” (likelihood ratio test value for M7 vs. M8 comparison, P value, and number of codons with the highest probability to have evolved under positive selection). Additionally, the results of the M1a vs. M2a comparison and the identity of specific codons under positive selection are saved in an Excel sheet (“Results-Current-Date/Results/Results_sheet” folder).
FREEDA maps the protein sequence of the reference species from the multiple sequence alignment to the expected reference protein sequence and, if appropriate, introduces gaps that represent residues excluded from the analysis (“GENE_protein_alignment.fasta” file). Based on that mapping, FREEDA provides both 2D and 3D visual representations of residues that likely evolved under positive selection. 2D bar graphs are provided in the “Results-Current-Date/Results/Graphs” folder. These graphs display the positions of recurrently changing residues (“Posterior mean omega,” top) and the probability of positive selection acting on each codon (“Prob. positive selection,” middle, probability 0.7–1.0; “High prob. positive selection,” bottom, probability 0.9–1.0). Codons excluded from the analysis are marked in gray. 3D representation of the most likely adaptive residues is found in the “Results-Current-Date/Results/Structures” folder, provided that the prediction model from the AlphaFold database matches the protein sequence extracted from the Ensembl database. For clarity, only residues with the highest probability (≥0.9) of having evolved under positive selection are mapped and their side chains are shown. The residues excluded from the analysis are colored in gray, and the N-terminal and C-terminal ends are labeled. Additionally, any domain annotation available in the Interpro database is automatically marked with a distinct color and labeled allowing quick visual identification.
FREEDA-annotated protein structure prediction models from AlphaFold designated by their UniProt entries (MmMIS18β—A2AQ14; MmAURKC—O88445; MmINCENP—Q9WU62; MmCENP-O—Q8K015; MmCENP-P—Q9CZ92) were aligned to PDB entries (SpMIS18— 5HJ0 ; HsAURKC— 6GR8 ; HsINCENP— 6GR8 ; HsCENP-OP— 7PB8 ) using an aligner module in PyMOL. Briefly, a FREEDA-annotated structure (e.g., “Aurkc_Mm.pse” found in “Results-Current-Date/Results/Structures”) was opened in PyMOL, the selected PDB entry was downloaded (e.g., “fetch 6GR8”), and the two structures were aligned (e.g., “align Aurkc_Mm, 6GR8”). The align module first aligns protein sequences and then superimposes their structures, returning RMSD (Root-Mean-Square-Deviation). Lower RMSD values indicate better alignment. All alignments presented here returned RMSD below 2 Å.
All CENP-O coding sequences were cloned from testis or liver samples. The use of rat CENP-O to represent a divergent ortholog was motivated by the availability of a rat ( Rattus norvegicus ) tissue sample for cloning. Chimeric CENP-O constructs were designed based on their 3D structure (AlphaFold database). Tissue was mechanically homogenized and total mRNA was isolated using TRIzol reagent (15596026; Invitrogen); cDNA was prepared using reverse transcription (18080051; SuperScript III First-Strand Synthesis System), amplified using construct-specific PCR primers (KK2502; KAPA HiFi Hot Start plus dNTPs; Roche), and inserted into the pGEMHE plasmid backbone (638948; In-fusion kit; Takara). Primers were designed using SnapGene (Dotmatics) software and are listed in . Each CENP-O construct was tagged with GFP at the C-terminus, separated by a linker of five glycines. Site-directed mutagenesis was performed using the Quik-Change Multisite Directed Mutagenesis kit (200515; Agilent) to introduce R225G, C226T, T228A, N247G, and P261H mutations. Rat CENP-O with 11 mouse point mutations was synthesized (Genewiz). The identity of all constructs was confirmed using Sanger sequencing of the entire coding sequence, including the reporter gene.
Mus musculus mice (CF-1 strain) were purchased from Envigo NSA stock # 033. Females were primed with 5 U of pregnant mare somatic gonadotropin (367222; Calbiochem) injected into the intraperitoneal cavity 44–48 h prior to oocyte collections to induce superovulation. The ovaries were isolated using M2 medium (M7167; Sigma-Aldrich) supplemented with 2.5 mM of the maturation-blocking phosphodiesterase 3 inhibitor milrinone (M4659; 2.5 mM; Sigma-Aldrich Milipore). Germinal-vesicle oocytes were collected, denuded mechanically from cumulus cells, and incubated for at least 1 h prior to microinjection on a hot plate (38°C) under mineral oil (9305; FUJIFILM Irvine Scientific). Oocytes were then microinjected with ∼5 pl of mRNAs in M2 medium with 2.5 mM milrinone and 3 mg/ml BSA at RT with a micromanipulator TransferMan NK 2 (Eppendorf) and a picoinjector (Medical Systems Corp). Oocytes were then incubated in 30–50 μl drops of Chatot-Ziomek-Bavister medium (MR019D; Thermo Fisher Scientific) under mineral oil (M5310; Sigma-Aldrich Milipore) at 37.8°C and 5% CO2 (Airgas) for 16 h to allow protein expression. The concentration of mRNA (15–30 ng/μl) used was selected to ensure similar cytoplasmic expression. mRNAs were synthesized using the T7 mScriptTM Standard mRNA Production System (C-MSC100625; CELLSCRIPT).
Oocyte maturation was induced in vitro by washing out milrinone 7.5 h before fixation. MI oocytes were fixed in freshly prepared 2% paraformaldehyde in PBS (pH 7.4) with 0.1% Triton X-100 for 20 min at RT, permeabilized in PBS with 0.1% Triton X-100 for 15 min at RT, placed in blocking solution (PBS with 0.3% BSA and 0.01% Tween-20) overnight at 4°C, incubated 1 h with primary antibody in blocking solution, washed three times for 15 min each, incubated 1 h with secondary antibody, washed three times for 15 min each, and mounted in Vectashield with DAPI (H-1200; Vector) to visualize chromosomes. Centromeres were labeled with CREST (human anti-human anti-centromere antibody, 1:200, HCT-0100; Immunovision) and an Alexa Fluor 594–conjugated goat anti-human secondary antibody (A-110014; Thermo Fisher Scientific). Confocal images were collected as 31 z-stacks at 0.5-μm intervals to visualize the entire meiotic spindle, using a confocal microscope (DMI4000B; Leica) equipped with a 63× 1.3 NA glycerol-immersion objective lens, an xy piezo Z stage (Applied Scientific Instrumentation), a spinning disk (Yokogawa Corporation of America), and an electron multiplier charge-coupled device camera (ImageEM C9100-13; Hamamatsu Photonics), controlled by MetaMorph software (Molecular Devices). Excitation was done with a Vortran Stradus VersaLase 4 laser module with 405-, 488-, 561-, and 639-nm lasers (Vortran Laser Technology). Panels of microscopic images were prepared using ImageJ (National Institutes of Health) free software.
Confocal images were analyzed using a custom-built Python-based automated program “Centrocalc” available here: https://github.com/DDudka9/Centrocalc . First, the program performs whole chromosome segmentation using the chromosome channel as a mask by grayscale dilation at a width of 16 pixels to ensure the centromeres are included. Then, a threshold is calculated using the Iterative Self-Organizing Data Analysis Technique method. If no chromosome channel is given, the entire cell is considered. Second, the program identifies centromeres using an approach based on . A difference of Gaussians algorithm is used to isolate spots of 150 nm. Third, a local maxima algorithm is used to identify centromeres. Up to 38 spots are chosen (the mouse 2n genome contains 40 chromosomes), separated by a minimum of two pixels by Chebyshev distance. The spots must be away from the edges of the image (20 pixels in x and y; 2 pixels in z). Fourth, 3D ellipsoid regions of interest (4 × 4 × 3 pixels) are drawn using the local maxima. Background regions of interest (ROIs) are drawn as volumes around the centromere ROIs (expanding the 3D ellipsoids by 1 pixel in all dimensions). Overlapping centromere ROIs are resolved by distance, where each pixel is assigned to the closest maxima. Fifth, centromere and background intensities are calculated as average grayscale pixel values and saved. ImageJ ROI files are also created to reference back to the original images. Modifying the Centrocalc source code can customize most of the described parameters.
Statistical analysis was performed using GraphPad Prism 9.3.1 (GraphPad Software). Statistical significance was assessed using a two-tailed T test (data distribution was normal under Kolmogorov–Smirnov and Shapiro–Wilk tests) or one-way ANOVA with Tukey’s multiple comparison test (data distribution was assumed normal but was not formally tested). Graphs display means with standard deviations. P values indicated on graphs are P ≥ 0.05, not significant (ns); P < 0.05, *; and P < 0.0001, ****.
Comparison of FREEDA with other pipelines is outlined in . Results of FREEDA validation are listed in (validation of ortholog detection based on Ensembl database), (accuracy in detecting positive selection using published datasets), , and (accuracy in detecting sites under positive selection based on published data). Results of positive selection detection in rodent centromere and kinetochore proteins are shown in . The list of all genomic assemblies used by the FREEDA pipeline and the core packages comprising the pipeline are listed in . lists all primers used to generate CENP-O constructs. , , , and show structural analyses of selected centromeric proteins. Experimental analyses of rodent CENP-O are shown in , , and . indicates sites that differ between mouse and rat CENP-O.
Review History Click here for additional data file. Table S1 shows testing FREEDA accuracy in detecting orthologs. Click here for additional data file. Table S2 shows FREEDA results analyzing test datasets comprising 23 primate proteins. Click here for additional data file. Table S3 shows comparison of specific sites that likely evolved under positive selection published previously and those detected by FREEDA in selected genes. Click here for additional data file. Table S4 shows FREEDA results analyzing 104 centromeric proteins in rodents. Click here for additional data file. Table S5 lists genomic assemblies and core packages used by FREEDA to detect statistical signatures of positive selection. Click here for additional data file. Table S6 lists primers used for generating CENP-O constructs. Click here for additional data file.
|
Content Analysis of Official Public Health Communications in Ontario, Canada during the COVID-19 Pandemic | 63b32f42-1a15-4990-9f97-9fc8f108d29a | 10970262 | Health Communication[mh] | Since early 2020, countries and governments worldwide have engaged in numerous communication strategies to update the public on the rapidly changing situation relating to the Coronavirus Disease 19 (COVID-19) pandemic. These strategies must be effective to educate the public on their risk of disease and increase compliance with any public health measures that are put in place. Risk communication, defined by the World Health Organization as “the real-time exchange of information, advice, and opinions between experts or officials and people who face a threat” , is essential to allow the public to make informed decisions to mitigate personal risks and implement preventative measures. In Canada, recommendations for preventative public health measures can vary across the country, as each province or territory has partial jurisdiction over healthcare decisions. As such, it is important that the public is provided with accurate, timely, and transparent communications to allow for effective decision-making within each jurisdiction. Several frameworks have been previously created to guide effective risk communication by public health officials, including the Crisis and Emergency Risk Communication Emergency Risk Communication (CERC) framework , created by the Centers for Disease Control and Prevention (CDC), and the Health Canada and Public Health Agency of Canada Strategic Risk Communication Framework and Handbook . The Strategic Risk Communication Framework was developed to help federal government scientists and communicators conduct risk communication in a more systematic and effective manner . The framework emphasizes that communication is one of the most powerful influences on how people make decisions regarding health behaviors, and, therefore, providing comprehensive information to the public will allow them to make informed decisions regarding their health and wellbeing . Despite the existence of established risk communication frameworks, it is unclear whether these methods are routinely applied in practice. Several studies have highlighted gaps in the application of best practices for risk communication during the COVID-19 pandemic. For example, a study examining how closely COVID-19 communications delivered by Scott Morrison, the Prime Minister of Australia, followed the CERC framework found that some components of the CERC model were not included in these communications . Similarly, another study examining COVID-19 crisis communication on social media using constructs from the health belief model and extended parallel processing model found an overall low use of these constructs in captions and images . The effective implementation of risk communication frameworks is further complicated by factors such as trust, emotions, and divergence in perspectives, which can affect how the public views the message . Since the public judges the trustworthiness of communications based on the characteristics of their content, including consistency, repetition, timeliness, transparency, and uncertainty, these factors are often included as key elements within established risk communication frameworks . However, further research on whether these frameworks are used in practice, especially within a Canadian context, is warranted. The objective of this study was to determine whether the communications by the Ontario government contained key elements of effective risk communication, as defined by the guiding principles of the Health Canada and Public Health Agency of Canada Strategic Risk Communication Framework and Handbook . These guiding principles outlined five key elements that were essential for risk communication between the government and the public. By identifying the presence or absence of these key elements among the communications, we aimed to determine whether there could be improvements in how COVID-19 risk could be communicated to the public. Publicly available news releases from the Ontario government (Ministry of Health and the Office of the Premier) were obtained from the Ontario Newsroom (Government of Ontario, 2020–2022, https://news.ontario.ca/en (accessed on 8 November 2022)). Although the Ontario government does utilize other public engagement tools (e.g., Government of Ontario Announcements YouTube Channel), we obtained news releases from this source as it was where information was first released and upon which other statements were based, as well as the most extensive and frequently updated source available. All releases that included significant mention of COVID-19 (i.e., the main purpose of the communication was regarding COVID-19) between 25 January 2020 (the first news release regarding COVID-19) and 15 December 2022 (the final news release regarding COVID-19 from 2022) were included. News releases were copied verbatim from the website and imported into NVivo (NVivo 1.7.1, Lumivero, London, UK) for analysis. Directed content analysis of the news releases was conducted by the first author. Content analysis is a method of analyzing written, verbal, or visual communication and involves characterizing these communications into manageable codes for further analysis . A directed content analysis uses an existing framework to deductively apply a theory to new data . Since we wanted to determine whether elements of the Strategic Risk Communication Framework were present in the news releases, a directed content analysis was the most appropriate method to achieve this goal . The guiding principles of the Strategic Risk Communication Framework were used as five top-level codes: (1) decisions are evidence-based, tapping both social and natural sciences; (2) risk management and risk communication processes are transparent; (3) stakeholders are the focal point; (4) strategic risk communication is integral to integrated risk management; and (5) the strategic risk communication process requires continuous improvement through evaluation . The first author (MF) initially read earlier news releases to become familiar with the data, then read the Strategic Risk Communication Framework to gain familiarity and understanding of the key elements. Once MF had read through the entire dataset and felt comfortable with the content, they followed the content analysis process as described by Elo and Kyngas . MF completed line-by-line deductive coding, where codes were assigned to key sentences and phrases in the news releases that represented elements of the Strategic Risk Communication Framework. MF used the guiding principles of the Strategic Risk Communication Framework and created descriptions of these elements. They then read each news release, and if the text aligned with the description, codes were assigned (see ). While only one author (MF) coded the data, consistent with , the research team (MF and KS) continuously discussed the codes and data interpretation to increase reliability of the results. Within some of the top-level codes, the relevant content from the news releases was further characterized through inductive coding of sub-categories that better explained the data, as some top-level codes were broad and did not accurately represent the diversity of the content presented in the news releases. Sub-categories were not mutually exclusive, and quotes could be coded to multiple top-level codes and sub-categories. Sub-categories were formed inductively through reading the data, assigning key words and phrases based on the data. A total of 322 news releases (176 by the Office of the Premier and 174 by the Ministry of Health) were included in this study . In 2020, there was a high frequency of news releases each week, beginning in week 4 of the year. In 2021, the frequency of releases each week remained high, but was lower than in 2020. The spike in releases observed from weeks 10–17 corresponded to increased communications regarding vaccine availability for the public. Finally, 2022 showed a substantial decrease in news releases, with a maximum of 1–2 news releases per week, with many weeks containing no releases. All elements of the Strategic Risk Communication Framework were identified across the news releases . Overall, transparency, risk management, and stakeholders as the focal point were the most frequently identified elements across the news releases. Transparency in communications often consisted of content relating to action plans throughout the pandemic. When discussing stakeholders as the focal point, this often referred to public-centered content. Finally, many news releases discussed risk management by including statements regarding the decisions made by the government, as well as the acts, laws, and groups that were involved in those decisions. Although an important aspect of the Strategic Risk Communications Framework, the news releases did not often mention continuous improvements or corrections to earlier statements as the situation evolved. 3.1. Element 1: Decisions Are Evidence-Based, Tapping Both Social and Natural Sciences Many news releases at the beginning of the pandemic discussed evidence or data supporting the decisions made by the Ontario government; however, as the pandemic progressed, this occurred less frequently. Much of the language used when describing the data that formed the basis of these decisions was vague. For example, when the need for lockdowns (i.e., closing of businesses and increasing physical distancing measures) was communicated, the rationale was “based on trends of key public health indicators” (13 July 2020); however, these indicators were not further described. Modelling of COVID-19 trends was often used as a rationale to support both why lockdowns were necessary and why lockdowns should continue to occur. Additionally, when data were provided to rationalize decisions, they often focused on data within natural sciences, with social sciences rarely being mentioned. 3.2. Element 2: Risk Management and Risk Communication Processes Are Transparent Most news releases with content relating to this element focused on describing an action plan or description of the ongoing situation. Prior to the initial lockdown in March 2020, the news releases consisted of case descriptions that described the first few cases of COVID-19 in Ontario. Here, the public was provided general information on the patient, including their age, gender, location, and the treatment they were given. This content was discussed again briefly following identification of the alpha COVID-19 variant. There was also an increase in content describing vaccines beginning in November 2020, which coincided with the timing of vaccine distribution to eligible populations. Early in the pandemic (January 2020 onwards), there was diversity in messaging in the releases, which was evident through the differences in sub-category use . Over time, there was a shift in the content discussed in the news releases, including a lack of diversity in messaging. Furthermore, in several instances, news releases contained quotes from speakers (e.g., Premier Doug Ford) that were used to convince the public about decisions behind their risk management actions. Occasionally when quotes were used, the language was distinct from the language used in the rest of the news release, as it involved a shift in urgency. An example of this can be seen in the news release from 17 March 2020: “This is a decision that was not made lightly. COVID-19 constitutes a danger of major proportions. We are taking this extraordinary measure because we must offer our full support and every power possible to help our health care sector fight the spread of COVID-19.” 3.3. Element 3: Stakeholders Are the Focal Point Content in the news releases was frequently public-centered, with a large focus on healthcare and frontline workers at the beginning of the pandemic . However, as the pandemic progressed, content discussing stakeholders as the focal point diminished, with news releases from April to June 2022 not mentioning stakeholders at all. When stakeholders were mentioned, they were often accompanied by praising language, with phrases like “frontline heroes” and “Ontario spirit” being used frequently to describe healthcare workers and businesses. There was also a strong focus on content relating to protecting the elderly and other vulnerable populations, including Indigenous people, individuals with mental health concerns, people with disabilities, and people who were homeless. Early in the pandemic, the news releases were centered on actions the government was taking to protect these vulnerable populations, for example: “We will continue to take aggressive action to support our most vulnerable residents and their caregivers” (20 April 2020). However, over time, the language shifted that responsibility to the public, as demonstrated in a news release on 19 February 2021: “As the vaccination rollout continues, it remains critically important that all Ontarians stay at home as much as possible and continue following regional public health measures, restrictions, and advice to protect our most vulnerable populations and help stop the spread of COVID-19.” Similar to the top-level code regarding evidence-based decisions (Element 1), this code, specifically in the context of education, contained vague language that mentioned the decision but did not provide further information on how that decision was made. For example: “We are taking decisive and preventative action today to ensure students can safely return to learning in our schools” (12 April 2021). 3.4. Element 4: Strategic Risk Communication Is Integral to Integrated Risk Management Content included within this code focused on higher-level descriptions of risk management processes, for example: “With a majority of Ontario adults having received a first dose of the COVID-19 vaccine and over three million doses of the Moderna vaccine arriving in June, the province is continuing to accelerate its vaccine rollout by expanding eligibility for second doses ahead of schedule.” (17 June 2021). Earlier releases more often mentioned specific laws, acts, and groups, while later releases contained more statements of decisions made by the government . 3.5. Element 5: The Strategic Risk Communication Process Requires Continuous Improvement through Evaluation The presence of this element was the least evident across the news releases. Although it was identified in some news releases at the beginning of the pandemic, it was only used once after May 2021. An example of its use is seen in the news release on 30 May 2020: “Today, the Ontario government made amendments to the Retirement Homes Act, 2010 regulation, enabling the Retirement Homes Regulatory Authority (RHRA) to better support seniors living in retirement homes during the COVID-19 outbreak.” There were several instances where this element could have been used in the news releases, but was not, such as in instances where something was changed but was not acknowledged. For example, in a news release from 11 December 2020, the speaker says that “…while vaccines will not be mandated during phase three, people will be strongly encouraged to get vaccinated”. However, on 17 August 2021, it is then said that “…in response to evolving data around the transmissibility of the Delta variant and based on the recent experiences of other jurisdictions, the government in consultation with the Chief Medical Officer of Health, is taking action […] this includes making COVID-19 vaccination policies mandatory in high risk settings”. Many news releases at the beginning of the pandemic discussed evidence or data supporting the decisions made by the Ontario government; however, as the pandemic progressed, this occurred less frequently. Much of the language used when describing the data that formed the basis of these decisions was vague. For example, when the need for lockdowns (i.e., closing of businesses and increasing physical distancing measures) was communicated, the rationale was “based on trends of key public health indicators” (13 July 2020); however, these indicators were not further described. Modelling of COVID-19 trends was often used as a rationale to support both why lockdowns were necessary and why lockdowns should continue to occur. Additionally, when data were provided to rationalize decisions, they often focused on data within natural sciences, with social sciences rarely being mentioned. Most news releases with content relating to this element focused on describing an action plan or description of the ongoing situation. Prior to the initial lockdown in March 2020, the news releases consisted of case descriptions that described the first few cases of COVID-19 in Ontario. Here, the public was provided general information on the patient, including their age, gender, location, and the treatment they were given. This content was discussed again briefly following identification of the alpha COVID-19 variant. There was also an increase in content describing vaccines beginning in November 2020, which coincided with the timing of vaccine distribution to eligible populations. Early in the pandemic (January 2020 onwards), there was diversity in messaging in the releases, which was evident through the differences in sub-category use . Over time, there was a shift in the content discussed in the news releases, including a lack of diversity in messaging. Furthermore, in several instances, news releases contained quotes from speakers (e.g., Premier Doug Ford) that were used to convince the public about decisions behind their risk management actions. Occasionally when quotes were used, the language was distinct from the language used in the rest of the news release, as it involved a shift in urgency. An example of this can be seen in the news release from 17 March 2020: “This is a decision that was not made lightly. COVID-19 constitutes a danger of major proportions. We are taking this extraordinary measure because we must offer our full support and every power possible to help our health care sector fight the spread of COVID-19.” Content in the news releases was frequently public-centered, with a large focus on healthcare and frontline workers at the beginning of the pandemic . However, as the pandemic progressed, content discussing stakeholders as the focal point diminished, with news releases from April to June 2022 not mentioning stakeholders at all. When stakeholders were mentioned, they were often accompanied by praising language, with phrases like “frontline heroes” and “Ontario spirit” being used frequently to describe healthcare workers and businesses. There was also a strong focus on content relating to protecting the elderly and other vulnerable populations, including Indigenous people, individuals with mental health concerns, people with disabilities, and people who were homeless. Early in the pandemic, the news releases were centered on actions the government was taking to protect these vulnerable populations, for example: “We will continue to take aggressive action to support our most vulnerable residents and their caregivers” (20 April 2020). However, over time, the language shifted that responsibility to the public, as demonstrated in a news release on 19 February 2021: “As the vaccination rollout continues, it remains critically important that all Ontarians stay at home as much as possible and continue following regional public health measures, restrictions, and advice to protect our most vulnerable populations and help stop the spread of COVID-19.” Similar to the top-level code regarding evidence-based decisions (Element 1), this code, specifically in the context of education, contained vague language that mentioned the decision but did not provide further information on how that decision was made. For example: “We are taking decisive and preventative action today to ensure students can safely return to learning in our schools” (12 April 2021). Content included within this code focused on higher-level descriptions of risk management processes, for example: “With a majority of Ontario adults having received a first dose of the COVID-19 vaccine and over three million doses of the Moderna vaccine arriving in June, the province is continuing to accelerate its vaccine rollout by expanding eligibility for second doses ahead of schedule.” (17 June 2021). Earlier releases more often mentioned specific laws, acts, and groups, while later releases contained more statements of decisions made by the government . The presence of this element was the least evident across the news releases. Although it was identified in some news releases at the beginning of the pandemic, it was only used once after May 2021. An example of its use is seen in the news release on 30 May 2020: “Today, the Ontario government made amendments to the Retirement Homes Act, 2010 regulation, enabling the Retirement Homes Regulatory Authority (RHRA) to better support seniors living in retirement homes during the COVID-19 outbreak.” There were several instances where this element could have been used in the news releases, but was not, such as in instances where something was changed but was not acknowledged. For example, in a news release from 11 December 2020, the speaker says that “…while vaccines will not be mandated during phase three, people will be strongly encouraged to get vaccinated”. However, on 17 August 2021, it is then said that “…in response to evolving data around the transmissibility of the Delta variant and based on the recent experiences of other jurisdictions, the government in consultation with the Chief Medical Officer of Health, is taking action […] this includes making COVID-19 vaccination policies mandatory in high risk settings”. This study aimed to determine whether communications by the Ontario government contained key elements of effective risk communication as defined by the Health Canada and Public Health Agency of Canada Strategic Risk Communication Framework. We found that, in general, news releases at the beginning of the pandemic were more comprehensive in terms of risk communication and contained most elements of the Strategic Risk Communication Framework. Additionally, earlier releases were more diverse in their content, which provided richer descriptions on each risk communication element. However, over time, the news releases became less detailed and included fewer risk communication elements. It is important to note that we cannot say whether these news releases were created with the federal government’s risk communication framework in mind. However, we can comment on whether the releases contained the key elements of the framework, and subsequently, whether any improvements in public health communication can be made. 4.1. Content and Language Use over Time As the pandemic progressed, there were two notable changes in how content was portrayed across news releases. First, the news releases were most detailed at the beginning of the pandemic and contained several rich details relevant to COVID-19 (e.g., case descriptions). Second, the news releases near the beginning of the pandemic had language that displayed a sense of urgency for individuals to partake in public health measures, which was infrequently used as the pandemic progressed. Both findings may coincide with differences in how information was expressed to the public across various pandemic phases. For instance, a study evaluating global COVID-19 narratives found a quick shift in how information was communicated during “peak pandemic” compared to the “recovery phase” following the lessening of public health recommendations . This communication strategy is also consistent with the need for different types of information during different pandemic phases. At the beginning of the pandemic, information most relevant to the public included timely and accurate scientific information on case descriptions and transmission mechanisms . Information shared during later phases of the pandemic most often focused on vaccinations, treatments, and other actions the public could take to protect themselves and others. It should be noted that detailed pandemic information and positive risk communication has been associated with the uptake of protective behaviors by the public , so even as pandemic phases shift over time, detailed content should still be provided to encourage use of public health measures. 4.2. Language and Transparency Vague language was frequently used across news releases when referring to the data that supported public health decisions. Vague language may be viewed by the public as a lack of transparency, which can decrease trust between the public and the government and lead to a decline in adherence to public health measures. A study of public perceptions towards COVID-19 communications by health authorities in Quebec found that the lack of transparency regarding uncertainties and evolving scientific knowledge was the most frequently identified criticism, especially surrounding the rationale to justify the implementation of public health measures . In our study, a lack of transparency was further identified in instances where risk communication processes should have been evaluated and communicated (Element 5). For example, when discussing the need for mandatory vaccination policies, changes to the requirements were described but were not acknowledged. Perceived inconsistencies in messaging have been previously linked to distrust of messages and can undermine public trust in the government. For example, a study on early pandemic news coverage across Canadian media found that reporters often framed changing guidelines and lack of transparency as public health incompetence by authorities, which can damage how these changes are reported by the media . Since a lack of transparency in public health messaging has also been previously reported in other studies , changes to sharing information by health authorities, including increased transparency regarding data being used to make public health decisions, are recommended. 4.3. Responsibility, Officials, and the Public Our results indicate that, over time, the content described in the news releases shifted responsibilities for public health protections from the government (e.g., lockdowns) to the public (e.g., physical distancing). A similar shift from government restrictions to personal responsibility occurred in the United Kingdom in February 2022, which some researchers argued was unsustainable without the government sharing clear information about risks and providing safe environments for the public to engage in individual risk management . This shift in responsibility may be viewed negatively by the public. One qualitative study found that those interviewed believed it was the government’s responsibility to create mandates and enforce the health orders that are determined by public health officials . Further, these individuals believed there was a shared responsibility between the government and the public to enforce and follow public health guidelines . Therefore, this shift in responsibility for public health measures from the government to the public may create a social barrier between the two parties, and by extension, cause distrust of information presented by the government . While communications that invoke personal responsibility are often intended to encourage community engagement, communications that deliver imperative messages (e.g., “you should stay home”) are more effective at increasing adherence to public health measures . As such, a balance between communicating imperative messages and those that invoke personal responsibilities might be needed to ensure desired adherence to public health measures. 4.4. Limitations We approached our methodology (content analysis) from an interpretivist perspective, which acknowledges that each researcher inherently influences data interpretation and analysis, and as such, there may be differences between coders . Since only one person coded the data, differences may be expected if another researcher was to conduct the analysis. However, this choice was made to allow the single coder to become immersed in the data rather than verify accuracy across coders . Since our analysis relates to only one framework that is not heavily present in the previously published literature, this may limit the generalizability of our findings. However, the elements of the framework are not unique, and are present in many other risk communication frameworks, such as the CERC framework. Our study focused on only one source of public health communications (the news releases), so further research is recommended to investigate content of other sources and how they may influence perceptions of public health messaging. As the pandemic progressed, there were two notable changes in how content was portrayed across news releases. First, the news releases were most detailed at the beginning of the pandemic and contained several rich details relevant to COVID-19 (e.g., case descriptions). Second, the news releases near the beginning of the pandemic had language that displayed a sense of urgency for individuals to partake in public health measures, which was infrequently used as the pandemic progressed. Both findings may coincide with differences in how information was expressed to the public across various pandemic phases. For instance, a study evaluating global COVID-19 narratives found a quick shift in how information was communicated during “peak pandemic” compared to the “recovery phase” following the lessening of public health recommendations . This communication strategy is also consistent with the need for different types of information during different pandemic phases. At the beginning of the pandemic, information most relevant to the public included timely and accurate scientific information on case descriptions and transmission mechanisms . Information shared during later phases of the pandemic most often focused on vaccinations, treatments, and other actions the public could take to protect themselves and others. It should be noted that detailed pandemic information and positive risk communication has been associated with the uptake of protective behaviors by the public , so even as pandemic phases shift over time, detailed content should still be provided to encourage use of public health measures. Vague language was frequently used across news releases when referring to the data that supported public health decisions. Vague language may be viewed by the public as a lack of transparency, which can decrease trust between the public and the government and lead to a decline in adherence to public health measures. A study of public perceptions towards COVID-19 communications by health authorities in Quebec found that the lack of transparency regarding uncertainties and evolving scientific knowledge was the most frequently identified criticism, especially surrounding the rationale to justify the implementation of public health measures . In our study, a lack of transparency was further identified in instances where risk communication processes should have been evaluated and communicated (Element 5). For example, when discussing the need for mandatory vaccination policies, changes to the requirements were described but were not acknowledged. Perceived inconsistencies in messaging have been previously linked to distrust of messages and can undermine public trust in the government. For example, a study on early pandemic news coverage across Canadian media found that reporters often framed changing guidelines and lack of transparency as public health incompetence by authorities, which can damage how these changes are reported by the media . Since a lack of transparency in public health messaging has also been previously reported in other studies , changes to sharing information by health authorities, including increased transparency regarding data being used to make public health decisions, are recommended. Our results indicate that, over time, the content described in the news releases shifted responsibilities for public health protections from the government (e.g., lockdowns) to the public (e.g., physical distancing). A similar shift from government restrictions to personal responsibility occurred in the United Kingdom in February 2022, which some researchers argued was unsustainable without the government sharing clear information about risks and providing safe environments for the public to engage in individual risk management . This shift in responsibility may be viewed negatively by the public. One qualitative study found that those interviewed believed it was the government’s responsibility to create mandates and enforce the health orders that are determined by public health officials . Further, these individuals believed there was a shared responsibility between the government and the public to enforce and follow public health guidelines . Therefore, this shift in responsibility for public health measures from the government to the public may create a social barrier between the two parties, and by extension, cause distrust of information presented by the government . While communications that invoke personal responsibility are often intended to encourage community engagement, communications that deliver imperative messages (e.g., “you should stay home”) are more effective at increasing adherence to public health measures . As such, a balance between communicating imperative messages and those that invoke personal responsibilities might be needed to ensure desired adherence to public health measures. We approached our methodology (content analysis) from an interpretivist perspective, which acknowledges that each researcher inherently influences data interpretation and analysis, and as such, there may be differences between coders . Since only one person coded the data, differences may be expected if another researcher was to conduct the analysis. However, this choice was made to allow the single coder to become immersed in the data rather than verify accuracy across coders . Since our analysis relates to only one framework that is not heavily present in the previously published literature, this may limit the generalizability of our findings. However, the elements of the framework are not unique, and are present in many other risk communication frameworks, such as the CERC framework. Our study focused on only one source of public health communications (the news releases), so further research is recommended to investigate content of other sources and how they may influence perceptions of public health messaging. This study evaluated whether the COVID-19 communications by the Ontario government contained key elements of effective risk communication, as defined by the Strategic Risk Communication Framework, and whether improvements in risk communication could be made. Our findings identified several areas where risk communication could be improved, such as increasing transparency regarding evidence-based decisions and explaining rationales behind changes in decisions. Based on our findings, there are several practical recommendations for the Ontario government. We recommend increasing transparency regarding the evidence and rationale behind public health decisions to improve public trust, and, therefore, compliance in health measures. We further recommend that communications are detailed whenever possible and are frequently improved upon to effectively convey the level of risk to the public and promote protective measures throughout all pandemic phases. Adhering to best practices for risk communication will allow the public to make informed decisions about their health and take the necessary measures to stay safe. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.