Drug Safety

Objectives. To present DrugSet, a validated R Shiny application supporting the construction medicinal products codelists based on the Anatomical Therapeutic Chemical (ATC) system and their mapping to Clinical Practice Research Datalink (CPRD) Aurum prodcodes within a single interactive workflow. Materials and Methods. DrugSet comprises four modules: data preparation, ATC-based hierarchical code selection, string-based CPRD Aurum prodcodes mapping, and codelist export. Validation was conducted against World Health Organization (WHO) ATC reference codelists and manually curated prodcodes mappings across three drug classes: metformin, beta-blocking agents, and topical salicylic acid. Sensitivity, specificity, and Positive Predictive Values (PPV) were calculated for ATC codelist generation. Agreement proportions (overlapping against total identified codes) were calculated for prodcodes mapping. Time needed for codelist construction using DrugSet was recorded and compared to manual approaches. Results. DrugSet ATC codelist generation against WHO manual reference achieved 100% sensitivity, specificity, and PPV across all medicinal products. Prodcodes mapping agreement ranged from 89.2% to 98.3% with discrepancies due to missing data in the prodcodes input vocabulary. DrugSet completed codelist construction in 9 minutes compared to 3 hours and 10 minutes manually, across all medicinal products classes. Discussion. DrugSet provides a unified workflow that runs directly on ATC and source CPRD Aurum vocabulary files. The reduction in codelist construction time and export of the generated codelists supports reproducibility in pharmacoepidemiologic studies where codelist creation can represent a significant proportion of study setup time. Conclusion. DrugSet is an open-source, validated tool that improves accuracy, and efficiency of codelist construction for medicinal products based on ATC codes towards CPRD Aurum prodcodes.

Background: HIV infection has long been associated with increased incidence of severe cutaneous adverse reactions (SCAR). It remains unknown whether this increased incidence is a direct biological result of HIV infection, differences in drug exposure, or other demographic factors. Objective: To evaluate the association between HIV and SCAR and determine whether this relationship persists after adjusting for demographic factors and structured drug exposure. Methods: We analyzed reports from the FDA Adverse Event Reporting System (FAERS) from 2013-2023. SCAR outcomes included Stevens-Johnson syndrome/toxic epidermal necrolysis (SJS/TEN), drug reaction with eosinophilia and systemic symptoms (DRESS), acute generalized exanthematous pustulosis (AGEP), and generalized bullous fixed drug eruption (GBFDE). HIV status was determined using antiretroviral exposure, indication text, and machine-learning imputation. Logistic regression models were constructed sequentially: unadjusted, demographic-adjusted, and fully adjusted with drug principal components to account for polypharmacy. Drug-level disproportionality and HIV-drug interaction analyses were also performed. Results: In unadjusted models, HIV was strongly associated with SCAR (OR ~2.0-2.7). Adjustment for demographics attenuated this association, and further adjustment for drug exposure reduced the effect to near null for overall SCAR and DRESS. A modest residual association persisted for SJS/TEN (OR ~1.3). Disproportionality analyses demonstrated enrichment of specific high-risk drugs in PLWH. Interaction modeling revealed drug-specific amplification of SCAR risk in HIV, notably for carbamazepine and clarithromycin, whereas other drugs showed minimal interaction. Conclusion: The association between HIV and SCAR is largely explained by differences in drug exposure and demographic factors. Residual risk is drug-specific rather than uniform, supporting a model in which HIV modifies susceptibility to select drug triggers rather than acting as a global risk factor. Further prospective and retrospective studies are required to quantify associations.

Drug repurposing offers a practical strategy to identify new therapeutic uses for approved drugs, potentially reducing the time and cost associated with conventional drug development. We present a novel three-stage drug repurposing pipeline that integrates knowledge graph-based gene prediction, network-based drug-disease association analysis, and systematic classification of candidate drugs by therapeutic class. The pipeline integrates DGLinker to predict novel disease-associated genes, SAveRUNNER to identify drug repurposing candidates, and ATC Category Enrichment Analysis (ATCEA) to prioritise candidates by pharmacological class. We benchmarked the pipeline across twelve diseases using DrugBank and MEDI2-HPS as validation resources. Utilising DGLinker-expanded disease-gene sets as input increased the number of predicted repurposed drugs, while overall discriminative performance remained stable across diseases (AUROC 0.71-0.77). Application of ATCEA consistently improved precision, F1-score, and specificity, while reducing recall, reflecting a conservative prioritisation strategy that contracts the candidate space while retaining pharmacologically coherent drug-disease candidates. We further applied the pipeline to amyotrophic lateral sclerosis (ALS), a neurodegenerative disease with limited therapeutic options, and performed a deeper literature-based validation of the results. Incorporation of DGLinker-predicted genes substantially increased the number of significant candidate drugs and uncovered enriched ATC categories not identified using known ALS genes alone, including antidepressants and antipsychotics. Moreover, several drugs with supporting evidence available in the literature were identified only when DGLinker-predicted genes were used. Overall, 77 candidate drugs were prioritised within significantly enriched ATC categories, several of which are supported by previously published studies. To provide exploratory real-world support for these findings, we further evaluated candidate drugs in a longitudinal electronic health record (EHR) dataset of 2361 patients with ALS from King's College Hospital. Although the number of evaluable drugs was limited due to sample size, the EHR analysis provided additional clinically relevant context for selected prioritised drugs and pharmacological classes. Our pipeline demonstrates potential to accelerate drug repurposing by integrating complementary computational approaches to each step of the process, providing an end-to-end framework that showed robust performance across benchmarking experiments and use cases.

Importance: Adverse drug events in older adults are a substantial public health burden, yet spontaneous reporting systems detect them poorly owing to underreporting and the lack of a defined population. These limitations are of particular concern for older adults, who are underrepresented in pre-approval trials yet at elevated risk owing to polypharmacy, multimorbidity, and age-related changes in drug metabolism. Objective: To develop and apply an active, claims-based pharmacovigilance framework using sequential target trial emulation to detect adverse drug event signals in older adults, with atorvastatin as the initial application. Methods: Using Medicare fee-for-service claims (2017-2019), we studied statin-naive beneficiaries aged 65 years or older following myocardial or cerebral infarction. We emulated up to 14 daily sequential trials from the discharge date, classifying patients as initiating atorvastatin (A1), initiating a different medication (A2), or no new medication (A0); the primary contrast was A1 versus A2. For each trial, incident outcomes were ascertained and classified into 552 outcomes based on the Clinical Classifications Software Refined categories. Per-protocol effects were estimated over a 6-month follow-up period using Fine-Gray regression models weighted by the inverse probability of treatment and censoring, treating death as a competing risk, with the false discovery rate controlled via the Benjamini-Hochberg procedure. A signal was declared when the q-value was 0.10 or lower and the subdistribution hazard ratio (sHR) was 1.20 or greater in any prespecified analytic stratum (sensitivity analyses used thresholds of q 0.20 or lower and sHR 1.20 or greater). Results: Of 70,130 eligible patients, 39,948 initiated atorvastatin (A1) and 19,182 initiated another new medication (A2); after weighting, baseline characteristics were closely balanced. After excluding outcomes with sparse cell counts, 295 outcomes were analyzed; five met the primary signal detection criteria: valve disorders (sHR 1.71, 1.20 to 2.43); sprains and strains (sHR 1.79, 1.26 to 2.54); general sensation/perception symptoms (sHR 1.23, 95 percent CI 1.11 to 1.36); abnormal findings without diagnosis (sHR 1.55, 1.18 to 2.05); and prediabetes (sHR 1.71, 1.24 to 2.36). In the sensitivity analysis, we additionally detected posthemorrhagic anemia, hemorrhagic stroke, varicose veins, and other circulatory and skin conditions. Conclusions: An active, claims-based framework using sequential target trial emulation detected both expected and previously unrecognized adverse drug event signals following atorvastatin initiation in older adults, offering a systematic alternative to passive surveillance that can be extended to other commonly prescribed medications.

Background: Sphingosine 1-phosphate receptor (S1PR) modulators can cause transient, dose-related negative chronotropic effects. Mocravimod is an oral S1PR modulator that is developed as a maintenance therapy in allogenic haematopoietic cell transplantation (allo-HCT). This phase I study evaluated whether two dose-titration regimens attenuate early bradycardia when initiating mocravimod while preserving pharmacokinetic (PK) and pharmacodynamic (PD) activity. Patients and methods: In this randomized, double-blind, placebo-controlled, parallel-group study, healthy adults received once-daily oral mocravimod using either dose titration (DT) regimen DT1 (0.3-2.0 mg with 4-day stepwise escalation) or regimen DT2 (0.5 mg to Day 14, 1.2 mg Days 15-18, then 2 mg), a fixed 2 mg regimen, or placebo for 21 days. The primary endpoint was the number of bradycardia episodes on treatment initiation and dose-escalation days derived from 24-hour Holter monitoring; PK of mocravimod and mocravimod-phosphate (whole blood) and PD effects (absolute lymphocyte count [ALC]) were assessed. Results: Fifty-six participants were randomized and 53 completed the study. Both titration regimens resulted in fewer bradycardia episodes than fixed initiation at 2 mg during the first week of treatment. Differences between titration and fixed dosing were no longer evident after Day 9, consistent with tolerance development. PK profiles were consistent with prior phase I data. By Day 21, DT1 achieved exposures close to the fixed 2 mg regimen, whereas DT2 yielded lower exposures, reflecting slower escalation. Peripheral lymphopenia developed in all active treatment groups and was comparable between regimens by Day 21, returning toward baseline by study end. Safety was similar between titration regimens and placebo, with similar distribution and incidence of adverse events. No serious adverse events occurred. Conclusion: Two practical titration regimens mitigated the early negative chronotropic effect observed with fixed-dose initiation of mocravimod at 2 mg once daily. Importantly, titration preserved the expected PK and PD profile, supporting dose escalation as an effective initiation strategy to improve early cardiac tolerability.

Background Severe Clostridioides difficile infection (CDI) morbidity and mortality disproportionately affect Black and Hispanic patients in the United States. Antibiotic exposure is the primary modifiable risk factor for CDI, and clindamycin is among the agents most strongly associated with related harm. Characterizing inequities in prescribing is critical. Dentistry is a major source of clindamycin prescriptions. Academic dental clinics serve diverse patient populations and provide an ideal setting to evaluate prescribing across racial and ethnic groups. We therefore examined antibiotic use and cumulative clindamycin exposure as measures of CDI-associated risk. Methods We conducted a retrospective study of electronic health records from 5 US academic dental institutions from 2021 through 2023. We analyzed 552,428 encounters among 132,770 patients with documented race/ethnicity to estimate adjusted odds of receiving any oral antibiotic and clindamycin by race/ethnicity. Secondary outcomes evaluated total antibiotic exposure among dental provider-prescribed antibiotics, focusing on higher-than-standard cumulative dosing of clindamycin (>8400 mg) and amoxicillin (>10,500 mg). Results Oral antibiotic prescribing occurred in 1.9% of encounters. Compared with White patients, Black, Hispanic, and Other race patients had slightly lower adjusted odds of receiving any oral antibiotic, while Black patients had greater odds of receiving a higher-than-standard cumulative clindamycin dose when clindamycin was prescribed (adjusted odds ratio, 2.19; 95% confidence interval, 1.25-3.82). Conclusion Racial and ethnic inequities in dental antibiotic prescribing extended beyond antibiotic receipt to cumulative clindamycin exposure. Although CDI outcomes were not directly measured, these prescribing differences may have implications for disparities in CDI-associated harm and warrant further investigation.

Background: Aspirin-free P2Y12-inhibitor monotherapy after percutaneous coronary intervention (PCI) is an alternative to dual antiplatelet therapy (DAPT), but the evidence rests largely on full-dose ticagrelor in acute coronary syndrome and on designs retaining a DAPT run-in; East-Asian patients may not require the same antithrombotic intensity. We compared standard DAPT, DAPT with half-dose ticagrelor, and aspirin-free half-dose ticagrelor monotherapy initiated on the day of PCI in chronic coronary syndrome (CCS). Methods: Sixty-one East-Asian patients with CCS scheduled for elective PCI were randomized 1:1:1 to Control (aspirin plus clopidogrel), Experimental A (aspirin plus ticagrelor 45 mg twice daily), or Experimental B (ticagrelor 45 mg monotherapy, aspirin discontinued at day 2). DAPT arms continued for six months; Experimental B continued indefinitely. P2Y12 reaction units (PRU) were measured at baseline and at a median of 17 days. Results: PRU reduction was three-fold greater in both ticagrelor arms than in Control ({Delta}PRU -188 and -181 versus -60.5; P<0.001), with no difference between ticagrelor arms (P=0.772). At 12 months, major adverse cardiovascular events (MACE) and clinically relevant bleeding each occurred in 1 of 17 Experimental B patients (5.9%) and in neither other arm. One Experimental A patient crossed over for ticagrelor-induced dyspnea; no stent thrombosis or cardiac death occurred. Conclusions: In East-Asian patients with CCS, half-dose ticagrelor produced markedly greater platelet inhibition than standard DAPT, with an identical effect whether given with or without aspirin. It merits evaluation in an adequately powered randomized trial. Clinical Trial Registration. URL: https://www.clinicaltrials.gov; Unique Identifier: NCT07622056

Background: Clinical outcomes of switching versus continuing direct oral anticoagulant (DOAC) among atrial fibrillation (AF) patients who experienced an ischemic stroke despite receiving DOAC therapy are uncertain. Methods: We included patients with AF who were hospitalized for ischemic stroke (index stroke) between January 1, 2016, and June 30, 2022, while receiving DOAC therapy and who resumed DOAC within 90 days after discharge in the Merative MarketScan Commercial and Medicare databases. Patients were classified as DOAC-switched or DOAC-continued according to whether the DOAC agent changed or remained the same after the index stroke; secondary analyses considered individual DOACs. The primary outcome was recurrent ischemic stroke; secondary outcomes included major bleeding and a composite outcome (bleeding or ischemic stroke). Propensity score-based overlap weighting and weighted Cox models were used to estimate adjusted hazard ratios (aHRs). Results: A total of 1175 patients were eligible for the study, of whom 970 (82.6%) continued and 205 (17.4%) switched DOAC therapy. Comparing DOAC-switched to DOAC-continued was not significantly associated with recurrent ischemic stroke (aHR, 1.20; 95% CI, 0.63-2.30), major bleeding (aHR, 0.60; 95% CI, 0.21-1.72), or the composite outcome (aHR, 0.98; 95% CI, 0.56-1.70). However, among patients who received apixaban before stroke, switching to rivaroxaban was associated with a higher risk of recurrent ischemic stroke (aHR, 2.70; 95% CI, 1.05-6.95). Conclusions: Overall, switching DOAC therapy after ischemic stroke was not associated with improved clinical outcomes. Switching from apixaban to rivaroxaban, however, could increase risk of recurrent ischemic stroke.

Background: Long COVID is a syndrome characterized by symptoms and conditions across all biological systems. This breadth of Long COVID phenotypes impedes efforts to identify the mechanistic pathways of Long COVID. Low serotonin may play a role in long-term sequelae of COVID-19, and selective serotonin reuptake inhibitors (SSRIs) may prevent these sequelae. Evaluation of the relationship between SSRIs and distinct categories of symptoms and conditions associated with Long COVID can highlight the mechanistic pathways that drive these relationships. Methods: We evaluated electronic health record data from a retrospective cohort of patients in the National Clinical Cohort Collaborative with comorbid depression and COVID-19 between October 2021 and February 2024. We estimated the relationship between SSRI prescription (versus no SSRI prescription) during acute COVID-19 and the one-year cumulative incidence of Long COVID-related conditions and symptoms across 14 human phenotype ontology categories. We applied Super Learner and targeted maximum likelihood estimation to estimate risk ratios while adjusting for confounders of interest and correcting for false discoveries from repeated testing. Results: We evaluated EHR data from 542,938 patients. We found that patients who were prescribed SSRIs during COVID-19 had a significantly lower risk of symptoms and conditions related to gastrointestinal factors (adjusted risk ratio (aRR) 0.95, 95% CI 0.92, 0.97), general health (aRR 0.91, 95% CI 0.88, 0.95), headaches (aRR 0.96, 95% CI 0.92, 0.99) and skin (aRR 0.92, 95% CI 0.87, 0.98). Discussion: We found that the prescription of SSRIs during acute COVID-19 was associated with a significantly lower risk of post-COVID sequelae related to gastrointestinal, headache-related, skin-related, and general symptoms and conditions, compared with no SSRI prescription. These findings highlight the role of serotonin in Long COVID and specific sequelae that may be reduced by SSRIs.

_ SURPASS-HF: Safety and Utility of Remote Pulmonary Artery Sensor Shared-management in Heart Failure --Background-- Insulin-dependent diabetics self-titrate therapy to self-obtained glucose values as standard of care, yet heart failure (HF) patients with implanted pulmonary artery (PA) pressure sensors never see their own readings; clinicians interpret and execute every dose change - a model that does not scale to a ~200-patient HF panel. To our knowledge, SURPASS-HF is the first prospective feasibility study applying the insulin-titration paradigm to PA-pressure-guided HF care: patients executing a prescribed loop-diuretic sliding scale, supported by ARTHUR, a domain-trained large language model, with clinician confirmation of every adjustment. --Methods-- Non-randomized, prospective, single-arm, single-center 90-day feasibility study (January 14-April 14, 2026; 60.1 patient-months). Twenty-one adults with implanted PA sensors enrolled (intention-to-treat, ITT); 19 completed full follow-up (per-protocol, PP). Regimens and individual PA diastolic (PAD) targets were explicitly prescribed; when daily pressures met published serial-reading thresholds, the software prepared the pre-determined adjustment, the clinician confirmed it, and the patient executed it. ARTHUR reinforced dose ceilings, prompted surveillance labs, and escalated edge cases for review. Pre-specified outcomes: adverse events, escalations, time in optimal PA range (TIR-PAP, +/- 5 mmHg of goal), reading adherence, provider overrides, and paired delta_PAD (first vs last 7-day windows). Confidence intervals are descriptive; the study was not powered for significance. --Results-- Mean age was 69+/-11 years, 52% women, mean baseline PAD 14.8 mmHg. No pre-specified safety event (KDIGO >or=1 AKI, hyperkalemia, hyponatremia, symptomatic hypotension) was detected (0/8 post-adjustment draws in 5/21 patients; exact 95% CI 0-37%); laboratory ascertainment was sparse, so a meaningful harm rate cannot be excluded. Seventeen of 19 PP patients (89%) required no protocol-triggered escalation; 4 escalations occurred in 2 patients. TIR-PAP was 88.4% (ITT)/91.3% (PP); reading adherence 92.1%; 53 provider alerts (0.88/patient-month) all resolved (median 24 h) with no overrides. delta_PAD was -0.89 mmHg (ITT; 95% CI -2.60 to +0.82) in a cohort already at goal at baseline. Two non-cardiac hospitalizations occurred. --Conclusions-- LLM-mediated, clinician-confirmed patient execution of a published deterministic PA-pressure-guided diuretic algorithm was feasible over 90 days, with high time-in-range and adherence and no detected safety events. Findings from this prospective, single arm, non-randomized, small cohort are descriptive. The study was not designed or powered to demonstrate evidence of a treatment effect; a randomized, well powered prospective comparison study against provider-led PA-pressure management is the next ideal step.

Introduction: Pediatric nicotine exposures remain an important and preventable public health issue, particularly with the rapid expansion of electronic nicotine delivery systems. This study compared demographic characteristics, exposure circumstances, and clinical outcomes between pediatric cases involving nicotine devices and bottled liquids reported to U.S. poison centers. Method: This retrospective cohort study analyzed National Poison Data System cases from 2011-2022 involving children aged less than 6 years exposed to nicotine devices or bottled liquids. Analyses were limited to cases with definitive medical outcomes. The primary outcome was defined as a moderate or major clinical effect or death. Odds ratios with 95% confidence intervals were calculated, with a secondary analysis restricted to route-concordant exposures. Results: The final cohort included 15,497 cases: 10,168 device exposures and 5,329 liquid exposures. Demographic characteristics were similar between groups. Device exposures more frequently involved inhalation, while ingestion predominated overall. Clinical effects were typically mild and transient, with vomiting and coughing most commonly reported. The primary outcome occurred in 1.9% of device cases and 2.0% of liquid cases (OR = 1.05; 95% CI 0.82-1.34). A secondary analysis restricted to inhalation-only device exposures and ingestion-only liquid exposures similarly found no significant difference in clinically important outcomes (OR = 1.38; 95% CI 0.92-2.12). Two deaths occurred, one in each group. Conclusion: These findings suggest that, despite differences in formulation and route of exposure, nicotine devices and bottled liquids produce broadly similar clinical toxicity profiles in young children. Prevention strategies should address all household nicotine products rather than focusing on specific delivery systems.

Effective postoperative pain control is essential following laparoscopic cholecystectomy, yet the analgesic value of a standardised 150 mg preoperative dose of pregabalin has not been clearly established. This systematic review and meta-analysis synthesised evidence from seven randomised controlled trials published between 2008 and 2025 to evaluate the efficacy and safety of pregabalin when administered before surgery. Four trials reported 24-hour postoperative pain scores, and pooled analysis demonstrated that pregabalin significantly reduced pain compared with control (SMD = 0.80 lower; 95% CI, 1.42 to 0.18 lower; p = 0.01), although statistical heterogeneity was high (I-squared = 81%). Pregabalin also produced notable reductions in opioid consumption, including fentanyl (SMD = 1.24 lower; p = 0.002) and tramadol (SMD = 4.21 lower; p = 0.002), again with considerable variability across studies. Sedation was slightly increased but did not reach statistical significance, and there were no significant differences in postoperative nausea, vomiting, or headache. Sensitivity analyses supported the stability of these findings. Overall, the results indicate that a single 150 mg preoperative dose of pregabalin meaningfully reduces postoperative pain and opioid requirements following laparoscopic cholecystectomy while maintaining an acceptable safety profile, supporting its use as part of a multimodal analgesic strategy.

BACKGROUND: Despite effectiveness as a once-weekly antimalarial prophylaxis, mefloquine has fallen out of favor due to its neuropsychiatric side effects. While possible genetic susceptibilities have been identified in preliminary studies, pharmacogenomic testing guidance is not available for mefloquine. METHODS: Volunteers with a history of mefloquine exposure were recruited to a cross-sectional case-control study. Pharmacogenomic analysis was performed on 7 candidate genes of interest with 16 missense variants including ORM1, MTHFR, MDR1, PYK2, HT2A, ADA, and ADORA2A. RESULTS: Fifty participants enrolled including those who had mefloquine exposure and chronic adverse effects (AEs) lasting 6 months or longer (n = 23); with subsequent AEs less than 6 months (n = 12); no AEs (n = 8); and a control group with a history of post-traumatic stress disorder (PTSD) but no mefloquine exposure (n = 7). Psychometric testing showed that mefloquine users with AEs lasting 6 months or more and PTSD patients who had not used mefloquine reported more evidence of sleep impairment, balance and equilibrium disorders, and lower levels of psychological well-being than mefloquine users who reported without AEs or with AEs but lasted less than 6 months. The ADORA2A gene was found to carry a higher burden of variation among volunteers exposed to mefloquine with AEs compared to those who did not. The variant rs141942830 within ADORA2A was observed to be higher among cases compared to the reference allele frequency listed in the gnomAD database but was found to not be significantly enriched. In addition, MTFHR gene was found to be enriched for variation in volunteers with long-term side effects compared to those with short-term or no side effects. CONCLUSIONS: Volunteers who reported long-term adverse events after exposure to mefloquine had excess rare variation within the ADORA2A gene compared to those without adverse events and those with short term adverse events. The ADORA2A rs141942830 was identified as a new variant of interest, as it was elevated but not significantly enriched among cases of long-term AEs, compared to the population frequency reported by gnomAD. These non-silent variants may serve as mediators to alternate pathways for signal transduction or drug metabolism.

Background: Antimicrobial resistance poses a major threat to global public health. Large language models (LLMs) offer new possibilities for optimizing antibiotic prescribing decisions, but the capabilities of general-purpose versus domain-specific medical LLMs under different prompting strategies remain to be clarified. Methods: This double-blind, randomized-sequence evaluation used a 2X2 factorial design comparing four AI conditions-the domain-specific model MedGo and the general-purpose model DeepSeek V3.5, each under standard direct prompting and chain-of-thought (CoT) prompting-alongside real physician prescriptions across 59 complex inpatient infection cases. Five parallel regimens were generated per case and independently evaluated by three senior clinicians (1-5 comprehensive score and five domain sub-scores). ChatGPT 5.2 was additionally assessed as an automated evaluation tool. Results: Score ranking: real physicians > MedGo-CoT > DeepSeek-CoT > MedGo> DeepSeek (Friedman test, p<0.001). In base mode, MedGo significantly outperformed DeepSeek (Holm-adjusted p=0.040). CoT improved both models (Holm-adjusted p<0.001 for DeepSeek; p=0.024 for MedGo) and reduced score dispersion. MedGo-CoT significantly outperformed DeepSeek-CoT in individualized adjustment (adjusted p<0.001) and dosing precision (adjusted p=0.005). ChatGPT-expert correlation was negligible (overall Kendall {tau}=0.153, p=0.003; subgroup {tau}=0.06-0.20, all p>0.05). Conclusions: Domain-specific medical LLMs enhanced by CoT approach the antibiotic decision-making level of real physicians, with advantages in individualization and dosing precision. However, notable deficiencies persist in antimicrobial stewardship ecological awareness and automated evaluation reliability, underscoring the continued indispensability of senior clinical expertise.

Background: Whether a patient's pain improved after emergency department (ED) treatment is read from the record to benchmark EDs, compare drugs, and label research outcomes. It is interpretable only if a post-treatment score is recorded, appropriately timed, and chosen by a fixed rule; its stability across these choices is unknown. Methods: Retrospective measurement study of adult headache visits in a de-identified ED database (MIMIC-IV-ED, 2011-2019). Among treated visits, we quantified reassessment completeness by time window, estimated meaningful relief (a reduction of at least 2 points) under score-selection rules and missing-data assumptions, tested whether reassessment was predictable at treatment, and compared headache with other painful presentations. Results: Among 19,501 visits (15,273 patients), 13,682 (70.2%) were treated. A post-treatment pain score appeared at any time for 77.1% (95% CI, 76.4 to 77.8), but within 2 hours of the analgesic for only 47.9% and within 1 hour for 27.5%. Meaningful relief was 66.9% using the first post-treatment score but 81.0% and 83.4% using the last or lowest score; it was 67.5% under inverse-probability weighting and could be bounded only between 51.8% and 74.4%. Whether a score was recorded was weakly predictable at treatment (area under the curve, 0.566) and unrelated to baseline pain. Completeness was similar across headache strata and comparator painful presentations. In an independent ED (MC-MED, a different EHR), the score-selection effect replicated: relief rose from 71.1% (first) to 80.6% (last) and 83.4% (lowest). Conclusions: Documented pain relief after ED headache treatment was not a stable outcome: it varied with the reassessment window and score-selection rule, was not point-identified for unreassessed patients, and behaved like other painful ED presentations. Programs and research that use documented relief should prespecify the reassessment window, score-selection rule, completeness denominator, and a missing-data range, and favor protocol-timed reassessment.

Glucagon-like peptide-1 receptor agonists (GLP-1RAs) have been linked to heterogeneous, potentially pleiotropic effects across organ systems, motivating outcome-wide comparative risk profiling in real-world data. A central challenge in such analyses is \emph{residual bias} that remains after adjustment for observed confounders, which can distort effect estimates and mis-calibrate uncertainty. We present distributional diagnosis and calibration (DC), which uses panels of negative control outcomes (NCOs) to diagnose residual bias and calibrate uncertainty. DC evaluates null behavior via $p$-value uniformity and empirical coverage across NCOs, and uses the empirical distribution of NCO effect estimates to calibrate confidence intervals for prespecified primary outcomes. DC is modular: it can wrap around commonly used causal inference methods and operates directly on summary statistics, supporting collaborative research under data-sharing constraints. Using electronic health records from a large U.S. clinical research network (152.7 million patients), we compared GLP-1RAs with sodium--glucose cotransporter~2 inhibitors across 15 prespecified outcomes spanning cardiovascular, mental health, and genitourinary domains using four causal estimators. Across outcomes and methods, DC diagnostics revealed substantial and method-dependent residual systematic error. DC calibration attenuated systematic error signals observed in negative controls and yielded more stable, better-calibrated estimates for clinical outcomes, supporting DC as a practical strategy to strengthen the credibility of real-world comparative effectiveness research.

Background: Metformin is one of the most commonly prescribed medications for individuals with diabetes and may provide protection against long-term sequelae of COVID-19. Methods: We evaluated a retrospective cohort of individuals in the National Clinical Cohort Collaborative with type 2 diabetes mellitus and COVID-19 who were prescribed metformin or a dipeptidyl peptidase-4 inhibitor (DPP4i) at least 30 days before the onset of acute COVID-19 between October 1, 2021, and November 15, 2023. We compared the 12-month cumulative incidence of Long COVID diagnosis (ICD-10 U09.9: Post COVID-19 condition, unspecified), probable Long COVID (based on a model-derived phenotype), and mortality between individuals prescribed metformin vs. DPP4i. We applied Super Learner and targeted maximum likelihood estimation to obtain risk ratios while adjusting for covariates of interest. Results: In our sample of 53,332 individuals with type 2 diabetes and COVID-19, we found that metformin prescription was associated with a lower risk of all-cause mortality after COVID-19 (adjusted risk ratio [aRR] 0.61, 95% CI 0.51, 0.73). We also observed that metformin users, compared to DPP4i users, had a slightly lower risk of probable Long COVID (aRR 0.87, 95% CI 0.81, 0.94) but did not detect a significant relationship with Long COVID diagnosis (aRR 0.90, 95% CI 0.68, 1.20), although we observed similar point estimates across Long COVID outcomes. Conclusions: These findings support the hypothesis that metformin prescription during acute COVID-19 may be associated with lower mortality among adults with diabetes. These analyses also provide modest evidence of a protective association against Long COVID in adults with diabetes, although estimates were imprecise.

Background. Septal reduction therapy (SRT) is recommended in drug-refractory, symptomatic obstructive hypertrophic cardiomyopathy (oHCM). We evaluated whether aficamten, a novel cardiac myosin inhibitor, can reliably transition guideline-eligible SRT candidates to ineligibility, and the associated safety profile of aficamten in this group. Methods. We analyzed participants with oHCM enrolled in FOREST-HCM (NCT04848506), the long-term open-label extension study of aficamten, from 28 May 2021 to 9 May 2025. Results. Three hundred and fifteen patients were included, of whom 104 met 2024 ACC/AHA guideline criteria for SRT eligibility at baseline. The SRT-eligible cohort was predominantly female (57%), with mean resting and Valsalva left ventricular outflow tract (LVOT) gradients of 63 {+/-} 39 and 109 {+/-} 42 mmHg, and all were in New York Heart Association (NYHA) class III. All baseline SRT-eligible patients became SRT-ineligible with aficamten therapy during study follow-up over a median of 42 days (IQR: 17, 49), except for one participant who withdrew from the study to pursue SRT (total of 3 participants withdrew). After dose titration, 3/104 (2.9%) remained guideline-eligible; by week 72 no patients met eligibility criteria. At maintenance, resting and Valsalva LVOT gradients improved by a least-squares mean of ?41 mmHg ([95% CI ?44 to ?37]; P<0.0001) and ?56 mmHg ([95% CI ?62 to ?51]; P<0.0001), respectively. Relative to baseline, NT-proBNP improved by 77% (95% CI 74 ? 80%), high-sensitivity cardiac troponin I decreased by 38% (95% CI 30 ? 46%), KCCQ-CSS improved by a mean of 20.2 (SD 19.3) points, and 95.2% of SRT-eligible patients had improved by ?1 NYHA class. Overall, the safety profile was favorable, with 2 occurrences of left ventricular ejection fraction (LVEF) < 50% over 193.7 patient-years of follow-up (1 event per 100 patient-years), managed by down-titration. There were no baseline SRT-eligible patients who died or developed LVEF <40%. Conclusions. Aficamten resolved guideline eligibility for SRT in nearly all baseline-eligible patients, with rapid and durable improvements in hemodynamics, symptoms, biomarkers and health status sustained for up to 3.5 years. Instances of LVEF <50% were rare and without clinical sequelae. These data support aficamten as a safe and effective alternative to SRT in oHCM.

Motivation: Human liver microsome (HLM)-based metabolic stability assays are fundamental in early drug discovery, shaping pharmacokinetic profiles and oral bioavailability. However, these experimental assays are labor-intensive and time-consuming, limiting their application in large-scale virtual screening. Computational models can prioritize compounds at scale, yet most are classification-based, leaving quantitative and interpretable prediction of HLM half-life limited. Results: In this study, we developed a quantitative machine learning model for the direct prediction of HLM half-life (T1/2) by integrating 11,790 compounds combining in-house and curated public data. Among various combinations of molecular features and learning algorithms, the XGBoost model with RDKit 2D descriptors achieved the best predictive performance, with an RMSE of 0.507 and an R2 of 0.431 on an independent test set. Shapley Additive Explanations (SHAP) analysis identified lipophilicity and known metabolic soft-spot features as the primary contributors to the predictions. These results suggest that this quantitative approach provides a practical framework for defining metabolic stability margins, thereby supporting rapid Go/No-go decisions in preclinical drug discovery. Availability: The source code, data, and trained model are available at https://github.com/joshua-416/PredHLM.

Caregivers increasingly act as proxies, managing patients digital accounts and making complex end-of-life decisions. Greater dyadic engagement in advance care planning (ACP) improves patient and caregiver outcomes, yet empirical evidence linking formal digital proxy roles to ACP engagement remains limited. The study aims to quantify patterns of ACP engagement, digital proxy roles, and how these caregivers behaviors are associated among caregivers in Singapore. We conducted a cross-sectional survey among an online panel of nationally representative adults in Singapore to identify caregivers and assessed their lifetime engagement in formal proxy roles across legal, financial, and medical digital domains, along with ACP proxy behaviors. Formal digital proxies had institutional or joint access to digital financial accounts (for financial digital proxies) or digital patient health/caregiver accounts (for medical digital proxies). ACP engagement was measured using 13 proxy-related behaviors, such as discussing end-of-life care preferences. Multivariable regressions were performed. In total, we identified 276 caregivers, who assisted with instrumental activities daily living to another adult from 311 completed responses. Among caregivers (age 41.0{+/-}13.8, 46.2% female), 28.9% were legal proxies and 40.2% were formal digital proxies (31.5% financial; 29.0% medical). Overall engagement was modest (mean 3.97{+/-}4.54) despite most reported completing at least one behavior. Compared to non-proxies, medical (AME=3.722, 95%CI: 2.143-5.301) and financial digital proxies (AME=1.515, 95%CI: 0.121-2.910) reported significantly higher ACP engagement while legal proxy status did not. High-stakes discussions on life-sustaining treatment and health-state preferences showed low engagement. Formal digital proxy roles are positively associated with ACP engagement and may provide a strategic entry point for interventions. Persistent deficits in high-stakes ACP highlight limited readiness for complex end-of-life decisions and the need for targeted decision-support tools.