Research Appraisals
Evidence-based critical appraisals of the latest medical research, systematically evaluated using Oxford CEBM methodology.
Showing 5 appraisals
PloS one
Developing a natural language processing system using transformer-based models for adverse drug event detection in electronic health records
OBJECTIVE: To develop a transformer-based natural language processing (NLP) system for detecting adverse drug events (ADEs) from clinical notes in electronic health records (EHRs). MATERIALS AND METHODS: We fine-tuned BERT Short-Formers and Clinical-Longformer using the processed dataset from the 2018 National NLP Clinical Challenges (n2c2) shared task Track 2. Two data processing methods, window-based and split-based approaches, were compared to identify the optimal processing method. Model generalizability was evaluated on a dataset extracted from Vanderbilt University Medical Center (VUMC) EHRs. RESULTS: On the n2c2 dataset, the best 5-fold cross-validation AUPRC, micro F1, and macro F1 scores were 0.840 (Clinical-Longformer, 4-chunk split), 0.965 (BioBERT, 15-word window), and 0.852 (Clinical-Longformer, 10-chunk split). On the VUMC dataset, the best AUPRC, micro F1, and macro F1 scores were 0.536 (Clinical-Longformer, 6-chunk split), 0.966 (BERT-base-uncased, 6-chunk split), and 0.762 (Clinical-Longformer, 4-chunk split). DISCUSSION: Transformer-based models demonstrated strong performance for ADE detection, with split-based processing generally outperforming window-based methods. Clinical-Longformer combined with a practical split-based approach showed promise for real-world implementation. Beyond token limits, chunk size substantially influenced model performance, even when text length remained within limits. CONCLUSION: Our findings provide practical guidance for developing transformer-based ADE detection systems from clinical notes. The selection of both text preprocessing strategies and model architectures should be guided by note characteristics and practical considerations such as annotation burden.
2 July 2026
Read appraisal →PloS one
Probiotics, synbiotics and berberine in Type 2 diabetes mellitus: A systematic review, meta-analysis, and molecular dynamics simulation study
BACKGROUND: Type 2 diabetes mellitus is characterized by impaired regulation of blood glucose. Probiotics, synbiotics, and berberine (BBR) have been proposed as adjunctive interventions, but their overall effectiveness remains uncertain. OBJECTIVE: To evaluate the effects of these interventions on glycemic control and to explore a potential molecular interaction of BBR with a key carbohydrate-digesting enzyme. METHODS: A systematic review and meta-analysis of randomized controlled trials was conducted. Pooled effects were estimated for fasting plasma glucose (FPG) and glycated hemoglobin (HbA1c) using random-effects models. A subgroup analysis compared probiotics with placebo. An exploratory computational analysis examined the interaction of BBR with α-glucosidase. RESULTS: More than 30 trials involving over 2,000 participants were included. The pooled analysis showed significant but modest reductions in FPG (-0.71 mmol·L ⁻ ¹) and HbA1c (-0.19%), with substantial between-study variability. Probiotics alone also reduced FPG (approximately -0.80 mmol·L ⁻ ¹) and HbA1c (approximately -0.21%) compared with placebo. Computational analysis indicated weaker enzyme binding for BBR than for the reference inhibitor acarbose. CONCLUSIONS: Probiotics, synbiotics, and BBR provide statistically significant but clinically modest improvements in glycemic control. These findings support their use as adjunctive, rather than primary, therapeutic options and highlight the need for larger and longer trials with standardized interventions. SYSTEMATIC REVIEW REGISTRATION: https://www.crd.york.ac.uk/PROSPERO/view/CRD420251116387, identifier CRD420251116387.
30 May 2026
Read appraisal →Toxicology mechanisms and methods
Introducing AutoML framework for drug-drug interaction prediction: application of AutoGluon.
Drug-drug interactions (DDIs) remain a major concern in medication safety. Although advanced artificial intelligence methods such as deep learning have improved DDI prediction, their adoption is limited by the need for specialized expertise and complex model development. This study introduces the first application of the AutoGluon AutoML framework to DDI prediction using molecular features, aiming to automate and simplify model development. A curated subset of 100,000 drug pairs from DrugBank was used, employing three molecular representations: 2D molecular descriptors, 2048-bit Morgan fingerprints, and their combination. Models were trained using AutoGluon-Tabular with no manual hyperparameter tuning. The descriptor-only model achieved the best performance, with 84.4% test accuracy and an AUC of 0.916, outperforming fingerprint-based and hybrid models. Feature importance analysis identified key physicochemical and topological descriptors-such as drug-likeness, electrotopological indices, and hydrophobic surface area-as critical predictors of DDIs. These results demonstrate that AutoML can extract chemically meaningful patterns while reducing technical barriers. Overall, our results validate AutoGluon as a scalable approach to DDI prediction that provides chemically meaningful, feature-level interpretability, and lay the groundwork for future applications involving larger datasets and more complex chemical representations.
15 Apr 2026
Read appraisal →Vaccine
Impact of pharmacy-based interventions on vaccination coverage among vulnerable and underserved populations: a systematic review and meta-analysis
BACKGROUND: Vaccination coverage remains suboptimal among vulnerable and underserved adult populations. We investigated the effectiveness of pharmacy-based interventions in improving vaccine uptake in high-risk groups and the most effective strategies. METHODS: A systematic search (PubMed/MEDLINE, Embase, Scopus; 26 July 2025) identified experimental and observational studies evaluating pharmacy-led vaccination interventions targeting adults ≥65 years or individuals aged 18-64 years with chronic conditions or limited healthcare access. The primary outcome was vaccination coverage. Odds ratios (ORs) were pooled using fixed- and random-effects models. Sensitivity analyses were conducted by intervention type and setting. Risk of bias was assessed using ROBINS-I, RoB 2, and NOS tools. RESULTS: A total of 35 studies were included, encompassing over 1.3 million participants. In pre-post comparisons, pharmacy-based interventions nearly doubled vaccination uptake (random-effects OR = 1.88; 95% CI: 1.68-2.12). In case-control analyses, individuals exposed to pharmacist interventions were almost five times more likely to be vaccinated (random-effects OR = 4.87; 95% CI: 2.23-10.64). Multicomponent strategies and direct vaccine administration by pharmacists produced the most consistent improvements. CONCLUSIONS: Pharmacist-led vaccination interventions significantly increase vaccine uptake among vulnerable adults. The most effective approaches involved pharmacists included eligibility assessment, counselling, administration, and documentation, integrated within routine care workflows.
11 Apr 2026
Read appraisal →JMIR mHealth and uHealth
Trust and Use of Recommendations for Health Apps Among European Residents: Cross-Sectional Survey
BACKGROUND: There is growing recognition of the role of health apps in addressing health care system challenges, yet app quality and evidence vary widely, and consumers have little decision support at the point of download. OBJECTIVE: This study aimed to explore which sources of recommendations European residents use and trust when choosing health apps and whether residents support government review and rating of health apps. METHODS: We conducted a cross-sectional online survey (December 7, 2022, to February 16, 2023) in 26 languages targeting residents of the European Economic Area, the United Kingdom, and Ukraine. The survey contained 11 questions covering demographics, types of apps used, sources of advice used and trusted, and views on government review and rating. We included only fully completed responses (N=1228). Descriptive statistics are presented as counts and percentages. For subgroup analyses, we dichotomized trust responses ("I trust" vs other responses) and tested associations with gender (Fisher exact test), age group, and education level (chi-square test); an adjusted significance threshold was applied (P<.004) to account for multiple testing. We followed the CHERRIES (Checklist for Reporting Results of Internet E-Surveys) guidelines for reporting online surveys and STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) recommendations for observational studies. RESULTS: A total of 1228 respondents from 33 countries completed the survey; 1110 (90.4%) reported using one or more health apps. COVID-19 apps (763/1228, 62.1%) and activity apps (737/1228, 60.0%) were the most frequently used, whereas disease management (86/1228, 7.0%), diagnostic (77/1228, 6.3%), and treatment apps (70/1228, 5.7%) were the least used. Sources used to choose apps included family and friends (434/1228, 35.3%), health professionals (412/1228, 33.6%), and government or health authorities (358/1228, 29.2%). The most trusted sources were health professionals (987/1228, 80.4%), pharmacists (750/1228, 61.1%), and government or health authorities (736/1228, 59.9%). No statistically significant differences in trust by gender were observed (all P>.0038). Some differences by age and education were observed for select sources (eg, government or health authorities: χ21=8.546; P=.003; family and friends: χ2=19.133; P<.001). Overall, 1060 (86.3%) of 1228 respondents supported government review and rating of health apps (either directly or by commissioning another organization). CONCLUSIONS: In this large multilingual European survey, most respondents reported experience with health apps and placed greatest trust in health professionals, pharmacists, and government or health authorities, yet professional recommendations were used less often than informal sources. There is clear public support for government-led review and rating schemes to guide consumer choice. Efforts to make trustworthy, easy-to-find information available at the point of download and to support health care professionals in recommending high-quality apps could help bridge the gap between trusted and used sources.
10 Apr 2026
Read appraisal →