Citations

Abstract Systematic reviews condense clinical trials into evidence tables, yet clinicians can interrogate these tables only through database queries, and many questions concern attributes that the table does not record, such as a drug’s target class or a harmonised endpoint. Here we introduce FD-SCoPE, a language-model framework that answers both kinds of question, exposes the query, the selected trials and the derivation rule behind every answer, and learns from expert corrections. On an oncology evidence table of 159 immune checkpoint inhibitor trial records, FD-SCoPE completed all 140 clinician-style tasks (alternatives, 90.7–97.9%). For questions needing derived attributes it retrieved 99.3% of relevant trial records at a positive predictive value of 89.8% and outperformed four alternative approaches (derived-value F1 77.7% versus 64.8–73.4%). Corrections on 299 questions, simulated from reference answers, raised F1 on 1,201 unseen questions from 77.9% to 84.9%. Language models coupled with executable queries, verified programs and expert feedback can give clinicians auditable access to trial evidence.

Abstract Background Femoral stem collars are intended to improve axial stability and load transfer following total hip arthroplasty (THA). However, substantial variability exists in collar geometry across implant systems, and how collar size and collar–calcar interaction influences femoral remodeling and the risk of mechanical complications remains incompletely understood. The purpose of this study was to determine radiographic calcar remodeling over time and to compare reoperation rates between femoral stems with constant-length collars (CLC) and size-specific collars (SSC), with the goal of better defining an optimal collar geometry. Methods A retrospective cohort study was performed using an institutional total joint registry to identify patients undergoing primary THA with either a CLC stem or an SSC stem. Radiographic measurements of collar overhang and collar-calcar gap were obtained at immediate postoperative, 3-month, and ≥2-year follow-up. Logistic regression models evaluated predictors of collar-calcar gap formation and closure. Receiver operating characteristic analysis identified a threshold for predictive gap resolution. A numerical optimization model was used to estimate the medial-to-lateral collar length that minimized absolute collar-calcar mismatch over time. Results Collar-calcar gaps were present in 12% of cases postoperatively, with resolution in 87% at final follow-up. Initial gap magnitude was the primary predictor of gap closure (OR 0.30 per mm, p < 0.001), with ROC analysis identifying an optimal cutoff of 1.54 mm for gap resolution. Progressive medial collar overhang increased over time. Optimization modeling demonstrated that the collar length minimizing absolute collar-calcar mismatch decreased from 7.5 mm at implantation to approximately 5.3 mm at final follow-up. Reoperation rates did not differ between implant designs, however occurred disproportionately in smaller CLC femoral stem sizes. Conclusion Proximal femoral remodeling following THA results in predictable thinning of the femoral calcar and progressive medial collar overhang over time. These findings suggest that collar geometry, particularly in smaller femora, should be considered relative to patient anatomy rather than implant design philosophy alone.

Abstract Background Our institution has a cTnT biomarker panel consisting of an initial cTnT, followed by a 2-hour and if necessary, a 6-hour collection. Accurate cTnT results are important for rapid diagnosis of acute coronary syndrome and to avoid prolonged stays in an emergency room. We compared cTnT results from Roche (Roche, Indianapolis, IN) cobas e 801 (801), e 601 (601), and two e 402 (402) instruments. Methods Comparisons (N=91) were completed using cTnT results from a single 601 instrument as the reference and comparing to the results from two 402 analyzers. Additional comparisons (N=45) were completed using cTnT results from an 801 as reference compared to results from the 402 instruments. Residual plasma separator tubes (PST) from clinically ordered cTnT testing were used. To ensure sample quality all plasma samples were aliquoted and recentrifuged before testing on each analyzer. Analysis was completed on all analyzers within 1 hour of the first comparison measurement. Acceptance limits were set as Passing-Bablok slope 1.00 ±0.10 and bias ±5 ng/L (for cTnT values <100 ng/L) or ±5% (for cTnT values =100 ng/L) when compared to reference 601 or 801. Student’s T-Test was performed to assess statistical significance between the observed mean bias between Pure instruments compared to the reference instrument, significance was defined as p =0.05. Results When compared to 601, results from the first 402 instrument demonstrated a slope of 1.11 with 54% (49/91) of samples within our acceptance criteria and a mean difference of 115 ng/L (Figure 1). Results from the second 402 instrument had a slope of 1.11 with 53% (48/91) within acceptance limits with a mean difference of 121 ng/L (p = 0.53, no difference between 402 instruments). For samples with cTnT <100 ng/L, 88% (36/41) and 88% (36/41) were within acceptance. In cTnT samples =100 ng/L, the mean percent bias was observed as 12% on both 402 analyzers. When using results from the 801 instrument as reference, results from the first 402 instrument had a slope of 0.954 with 77% (33/43) within acceptable limits and a mean difference of -61 ng/L. Results from the second 402 instrument had a slope of 0.978 with 79% (34/43) within acceptable limits and a mean difference of -39 ng/L (p = 0.33, no difference between 402 instruments). In samples with cTnT <100, comparisons to the 801 showed 82% (14/17) and 94% (16/17) within acceptance criteria and a mean difference of 2 ng/L on each 402 instrument. For cTnT samples with concentrations =100 ng/L, the mean percent bias was -2% on both 402 instruments. Conclusion We observed proportional bias on the 402 instruments compared to both 601 and 801 instruments, with large and clinically significant differences particularly when comparing to the 601 at higher cTnT concentrations. To avoid a delta being caused by observed systematic bias we will complete all cTnT testing within a panel on one platform.

Abstract Background Lung transplant (LT) recipients are at high risk for severe respiratory syncytial virus (RSV) infection, but predictors of worse outcomes remain poorly defined. Methods Retrospective multicenter cohort study of adult LT recipients hospitalized with PCR-confirmed RSV infection between January 2016 and December 2019. Clinical characteristics, treatments, and outcomes within 90 days were analyzed. Logistic regression was used to identify predictors of outcomes, while Cox proportional hazards models assessed risk factors for mortality. Results 155 LT recipients were included. Median age was 61 years, RSV occurred a median of 14 months post-transplant. Risk factors for hypoxic respiratory failure were baseline oxygen use [OR 4.8, (1.29–23.93)] and lower respiratory tract symptoms [OR 3.74, (1.49–10.18)]. Older age increased the risk of mechanical ventilation [OR 1.06, (1.02–1.12)], while oral ribavirin was associated with reduced risk [OR 0.34, (0.14–0.82)]; this association was attenuated when oral and inhaled ribavirin were combined into a single exposure variable [OR 0.77, (0.22–2.66)]; oral ribavirin was not associated with reduced 90-day mortality [HR 5.27, (0.68–40.83)]. Secondary bacterial or fungal infections occurred in 33.5%, associated with prior mechanical ventilation. Lung function declined significantly post-RSV infection, with nearly half experiencing >10% decrease in FEV1. Ninety-day all-cause mortality was 7.7%, with mechanical ventilation independently associated with death [HR 7.79, (2.47–24.60)]. Conclusions RSV infection in LT recipients is associated with high morbidity, graft function decline, and mortality. The association between older age and severe outcomes suggests the importance of preventive strategies in this high-risk population.

Abstract Provided herein are novel aminocyclohexenyl-6,6-bicyclic compounds as 5-HT2A receptor agonists, pharmaceutical compositions, use of such compounds in treating depression, anxiety, substance abuse and headaches, and processes for preparing such compounds.

Abstract Selective prediction in clinical AI conventionally defers low-confidence predictions to human clinicians. But confidence captures only one dimension of prediction quality, and the patients it flags may not be the ones who most need human review. We propose reliability-gated selective prediction, where a composite of five dimensions—confidence, data sufficiency, cohort similarity, temporal stability, and uncertainty consistency—determines whether the model predicts or defers. Using 84,013 ICU admissions from MIMIC-IV and an XGBoost ensemble across three tasks (sepsis, mortality, vasopressor need), we find that reliability-based and confidence-based deferral policies select almost entirely disjoint patient subsets (Jaccard < 0.001 for all tasks). Confidence defers low-probability cases; reliability defers patients with sparse data and distributional unfamiliarity—precisely the failure modes invisible to confidence. The reliability gate achieves 1.4–2.4× lower Brier scores at every coverage level on temporal holdout. At 80% coverage, the number needed to review (NNR) is 2.7–3.9 for the reliability gate versus 32–2,085 for the confidence gate. A two-dimension composite (confidence and uncertainty consistency) matches the full five-dimension composite, providing a minimal deployment configuration. These findings replicate on MIMIC-III (46,187 patients, 2001–2012): Jaccard <0.001, Brier reduction 2.4–2.7×, and NNR 3–4. The reliability gate outperforms a conformal selective prediction baseline on Brier score across all tasks while providing per-dimension diagnostic information. These results define reliability as a deployment primitive for clinical AI: a multidimensional assessment of prediction trustworthiness that is not reducible to model confidence and generalizes across database versions and time periods.

Background and Objective: This study aimed to investigate glaucoma progression following secondary intraocular lens (SIOL) implantation by surgical approach. Patients and Methods: This was a single-center retrospective study of 440 eyes that received SIOL implantation, including scleral-sutured, scleral-fixated, anterior chamber, and sulcus-fixated. Glaucoma progression was defined as new diagnosis, procedure for intraocular pressure (IOP) control, or IOP elevation ≥ 24 mmHg or ≥ 5 mmHg above baseline requiring additional medications. Risk factors were assessed using multivariable analysis. Results: Mean age at surgery was 71.3 years and 29.3% of patients had prior glaucoma. Overall, 33.2% developed glaucoma progression. Prior glaucoma (hazard ratio 2.08 [95% CI 1.47–2.94], P < .001) and cystoid macular edema (1.63 [1.06–2.52], P = .026) were associated with glaucoma progression on multivariable analysis. There was no difference in glaucoma progression between scleral-sutured, scleral-fixated, anterior chamber, or sulcus-fixated implantation techniques. Conclusion: Our results suggest that surgeons should use their preferred SIOL implantation technique but closely monitor postoperative IOP.

Hypertrophic cardiomyopathy is a common heritable heart disease. The disease is genetically, anatomically, and clinically heterogeneous. Hypertrophic cardiomyopathy is inherited in an autosomal dominant pattern but has allelic heterogeneity, incomplete penetrance, and variable expressivity. Beyond family screening, clinical evaluation for hypertrophic cardiomyopathy is often prompted by identification of symptoms or detection of a heart murmur, additionally, abnormal electrocardiogram or imaging results performed for another reason can provide evidence. Evaluation should include obtaining a patient’s history, a physical examination, cardiac imaging, and, often, genetic testing. Prognosis is variable; major causes of morbidity and mortality are caused by sudden cardiac death, heart failure, and atrial fibrillation with associated stroke risk. An understanding of the pathogenesis at the genetic, cellular, and individual levels helps clinicians to implement appropriate screening, recognise clinical manifestations, and improve the treatment of patients with hypertrophic cardiomyopathy. The understanding of this condition is evolving, with recent advances and insights into its pathogenesis, refinement of risk stratification, and use of artificial intelligence in screening, diagnosis, and prognostication.

Abstract Cellular senescence is a central hallmark of aging, yet its measurement in humans remains invasive, low-throughput and tissue-specific, precluding population-scale study. We developed Methylation-associated Gene Expression (MaGE) predictors: whole-blood DNA methylation proxies of p14 ARF , p16 INK4A and p21 CIP1 expression, plus a composite score, providing the first scalable measure of senescence-marker expression. Deployed across Generation Scotland (n=18,859), MaGE yielded 45 Bonferroni-significant associations spanning 18 incident diseases and all-cause mortality. The two strongest associations recapitulated the established tissue specificity of p21 CIP1 and p16 INK4A activation, with MaGE-p21 associating with incident alcoholic liver disease and MaGE-p16 with pulmonary fibrosis. In a separate study, MaGE tracked disease severity and treatment response in Crohn’s disease. MaGE is a scalable, interpretable biomarker of senescence that enables its study at population scale and offers a route to patient stratification in senolytic trials.