Citations

Abstract Optimal management of anticoagulation for patients with aortic mechanical heart valves (MHVs) receiving vitamin K antagonists (VKAs) undergoing gastrointestinal endoscopic procedures is clinically challenging. Risk of bleeding from an unanticipated polypectomy or biopsy further complicates this issue. Current guidelines on use of bridging with low-molecular-weight heparin (LMWH) for these procedures are based on low-quality evidence. We conducted a subanalysis of a retrospective, multicenter observational study of adult patients with aortic MHVs receiving VKAs who underwent elective gastrointestinal endoscopies (colonoscopy or upper endoscopy). We included patients who underwent the procedure between July 1, 2020, and July 1, 2023, at five centers in Canada and the United States Patients with procedures performed on separate days within a 30-day period were excluded. Efficacy and safety outcomes included thromboembolic events, major bleeding (MB), and clinically relevant non-major bleeding (CRNMB) in the 30 days post procedure. A total of 192 endoscopic procedures were analyzed. Median (interquartile range [IQR]) patient age was 66 years (58.2–72.8) and 22% were female. Warfarin was interrupted for 185 procedures (96%). Among these, 119 (64%) received LMWH bridging: 33 (17%) pre-procedure only, 10 (5%) post-procedure only, and 76 (40%) both pre- and post-procedure. Post-procedure LMWH was initiated a median (IQR) of 1 day (1–1) and discontinued 5 days post-procedure (3–8). MB occurred in two patients (1%), CRNMB in two patients (1%), and one death (0.5%) was identified; no thromboembolic events occurred. Prior to propensity score matching (PSM), post-procedure LMWH with or without pre procedure LMWH was associated with the composite outcome of MB and CRNMB (P = 0.034), although not with MB or CRNMB individually. After stepwise regression and PSM analyses, no significant differences in bleeding outcomes were observed between patients who received LMWH bridging and those who did not. Our findings indicate that although adverse thrombotic events were rare and use of post-procedure LMWH bridging was significantly associated with a small absolute increased risk of bleeding, it was not significant after PSM. Larger prospective studies are needed to better inform antithrombotic management guidelines for patients with aortic MHVs undergoing gastrointestinal endoscopy.

Abstract We aim to compare the benefits, harms, and costs of multiple plastic stents (MPS) versus covered self-expandable metallic stents (CSEMS) in the treatment of benign biliary strictures (BBS). A comprehensive database search through June 6, 2024 identified randomized controlled trials (RCT) involving adult patients with BBS comparing MPS to CSEMS in terms of stricture resolution, recurrence, technical success, number of required procedures and stents, treatment duration, all-cause mortality, adverse events, and stent migration. We utilized the Cochrane Risk of Bias (RoB) tool2 for study quality assessment. Meta-analysis following the random effects model was performed. After screening 365 references, we included 7 RCTs (high RoB) involving 594 patients. Most reported stricture etiology is chronic pancreatitis (52.4%), followed by orthotopic liver transplant (40.5%). Compared with MPS, CSEMS achieved similar efficacy and safety. However, CSEMS required fewer procedures (mean difference [MD]: −1.84 [−2.56 to −1.11]; I2 = 76%) and stents (MD: −6.04 [−9.55 to −2.52]; I2 = 93%), amounting for shorter duration of treatment (MD: −119.02 days [−137.41 to −100.62]; I2 = 96%) and savings in costs between 41 and 57% with absolute savings of 4,019$ to 9,192$ favoring CSEMS. All outcomes had a low to very low certainty of evidence (CoE), except for technical success and stricture resolution (moderate CoE). Our analysis suggests that CSEMSs are comparable to MPSs in terms of efficacy and safety. However, it shows that CSEMSs are more cost-effective. Sample size calculations from our analyses can guide the design of future studies to validate the safety profiles and determine the feasibility of the two modalities.

Abstract Coronary artery bypass graft (CABG) surgery is one of the most commonly performed surgeries in the United States. It remains the gold standard of coronary artery revascularization despite the introduction of percutaneous coronary interventions and advancements in medical therapy. First conceptualized in the 1900s, CABG underwent decades of trials, experimentation, innovations, and even failures. Despite the challenges, CABG continued to innovate, overcoming the complexities of coronary artery disease and improving patient outcomes. This manuscript aims to discuss the beginnings, evolution, and future of CABG.

Abstract Most clinical large language model (LLM) benchmarks rely on clean, concise vignettes that do not reflect the noisy, long-form documentation typical of real clinical records. How LLM performance degrades under realistic chart conditions remains poorly characterised. Here we test whether structured retrieval workflows protect National Institutes of Health Stroke Scale (NIHSS) scoring accuracy under systematic context stress. Using 100 de-identified acute stroke cases and a fully crossed 4 x 4 x 3 x 3 condition matrix (144 conditions per case), we vary context acquisition method, document length, distractor load and critical-information position across four Gemini models (57,047 retained runs). Structured retrieval reduces mean absolute error (MAE) from 4.58 to 2.96 points relative to non-agentic baselines (mean gain 1.62 MAE points; 95% CI 1.57–1.67; 35% relative reduction), with consistent gains across all 36 stress combinations. Lower-cost models show disproportionately larger gains (2.76 versus 0.45 MAE points). Tool-retrieved pipelines outperform retrieval-augmented generation in 33 of 36 combinations. These findings indicate that retrieval architecture, rather than model scale alone, is a tractable lever for robust, equitable clinical LLM deployment.

ABSTRACT Braak and others have proposed that Lewy body pathology (LBP) in Parkinson’s disease (PD) may arise not only in the brain but alternatively from an initial site in the gastrointestinal (GI) tract with subsequent passage to the central nervous system CNS through the vagus nerve or other routes. We tested this hypothesis by using both immunohistochemistry (IHC) and RT QuIC a form of alpha synuclein seed amplification assay (SAA) to detect alpha synuclein LBP in samples from selected brain regions and 10 GI tract sites taken from autopsies of 50 PD subjects and 128 elderly subjects without parkinsonism or dementia including 34 with IHC identified CNS incidental Lewy body disease (ILBD) and 94 with no Lewy body IHC pathology detected (NLB). A positive SAA or IHC result was restricted to the GI tract in only 2 subjects while LBP by either SAA or IHC was restricted to the brain in 11 subjects. To fairly compare GI-only with brain-only synucleinopathy, however, we would have to do SAA on brain samples from all ILBD and NLB cases in at least 4 critical brain regions: olfactory bulb, medulla, pons, and amygdala. Further SAA of brain regions is estimated, based on the proportional results to date, to potentially identify 21 additional brain-only LBP subjects, for a total of 32, if it were done on all of the NLB subjects. From this brain-only LBP is estimated to be 16 times more common than GI-only LBP. To assess the clinical impact of SAA-positive GI sites we found that the number of positive sites per subject is significantly correlated with UPDRS motor score and SCOPA-AUT GI related scores including those for salivation, straining, constipation, and bowel movement.

Abstract Introduction Evaluate large language models (LLMs) for scoring medical student essays, and compare various prompting techniques and models. Methods OpenAI GPT scored 51 medical student reflection essays (15 real, 36 fabricated) using a previously-reported 6-point rubric (April-May 2025). We compared 29 prompt-model conditions by systematically varying the LLM prompts (including the persona, scoring rubric, few-shot learning [exemplars], chain-of-thought reasoning, and temperature), fine-tuning, and model (including GPT-4.1, GPT-4.1-mini, GPT-o4-mini, and GPT-4-Turbo). Outcomes were accuracy (compared with human raters, measured using single-score intraclass correlation coefficient [ICC] and mean absolute difference [MAD; zero indicates perfect agreement]), within-condition reproducibility, and cost. Results Across all conditions, it took mean (SD) 3.73 (3.12) seconds to score 1 essay. The cost to score 100 essays was USD $0.04 for GPT-4.1-mini, $0.21 for GPT-4.1, $0.57 for GPT-4.1 with 3 exemplars, and $2.00 for fine-tuned GPT-4.1. When the one-time cost of fine-tuning was amortized across 10,000 essays, the cost for fine-tuned GPT-4.1 was $0.20 per 100. Accuracy was “almost perfect” (ICC >0.80) for 28/29 conditions (97%). Fine-tuned models were more accurate than non-fine-tuned models (MAD difference –0.24 [95% CI, –0.34, –0.14]). Conditions with exemplars were more accurate than those without (MAD difference –0.44 [CI, –0.57, –0.31]). Accuracy progressively decreased as 6, 3, 1, and 0 rubric levels were explicitly defined in the prompt (P<.001). Contrary to hypotheses, accuracies for chain-of-thought prompts and variations in temperature and persona were not significantly different from the baseline prompt. Reproducibility ICC was >0.80 for 28/29 conditions (97%). Discussion Automated LLM essay scoring demonstrated near-perfect accuracy and reproducibility for most prompt-model conditions. Fine-tuned models and prompts with exemplars had higher accuracy but higher cost. Fine-tuned models had lower per-essay costs for larger essay volumes. For smaller volumes, non-fine-tuned GPT-4.1 provided excellent results at moderate cost. GPT-4.1-mini provided very good results at low cost.

Hypertension is the main risk factor for cardiovascular disease (CVD), affecting 20–40% of Latin American (LATAM) adults, and responsible for more than two million deaths annually due to CVD. The different ethnic, economic, geographic, and cultural characteristics of the LATAM population influence the high prevalence of all cardiovascular risk factors (CVRF), particularly metabolic disturbances such as type 2 diabetes (DM2), obesity and the metabolic syndrome. Their main determinants in LATAM includes environment, food quality, social inequity, low education, political aspects, contextual behaviour, and genetics. The prevalence of overweight and obesity in LATAM increased during the last four decades reaching figures of 10–20% in childhood, 30–40% in adolescence, and 60–70% in adults. Many studies in the region have reported the extremely low rates of awareness, treatment, and control of CVRF in the general population of LATAM, particularly in patients with metabolic disorders, and the consequent high cardiovascular morbidity and mortality. This 2026 LATAM consensus is developed by a large group of experts from different LATAM countries, the USA and Europe, representing areas of internal medicine, cardiology, nephrology, endocrinology, geriatrics, paediatrics, pharmacology, and epidemiology. A careful search for novel studies in LATAM, together with new evidence that has emerged since the 2019 LATAM consensus, support the statements and recommendations in the current report. This update aims to provide clear and useful recommendations for health professionals to improve awareness, treatment, and control of hypertension and associated CVRF in the region.

Abstract Despite effective control of inflammation in rheumatoid arthritis (RA), persistent and severe pain remains a challenge. Dysregulated pain processing in the central nervous system may underlie this disconnect. The PainDETECT questionnaire, originally developed to identify neuropathic pain, may capture features of central sensitization. We hypothesized that neuropathic-like pain at diagnosis predicts worse pain prognosis in early RA, independent of baseline pain severity and inflammation. We also explored brain activation patterns using neuroimaging. In this prospective cohort study, adults with newly diagnosed RA completed the PainDETECT questionnaire at baseline. Bodily pain was measured using the SF-36 Bodily Pain Subscale at baseline, 3, 6, and 12 months. C-reactive protein (CRP) was monitored. Linear mixed-effects models examined associations between PainDETECT, as a continuous measure, and bodily pain and CRP over time, adjusting for baseline outcome values. A neuroimaging substudy (n = 27) used task-based functional MRI to assess brain responses to evoked pressure pain. Among 158 participants (mean age 58.4 years; 35% male), 24.7% had likely neuropathic-like pain (PainDETECT ≥19) at baseline. Higher baseline PainDETECT scores were associated with worse bodily pain throughout follow-up in fully adjusted models (β = −0.43; 95% CI: −0.84 to −0.02; P = 0.043), independent of socio-demographics and comorbidities. No association was found between PainDETECT and CRP. On neuroimaging, higher PainDETECT scores correlated with increased activation in the left insula, dorsal anterior cingulate cortex, and left amygdala during evoked wrist pain. Neuropathic-like pain at diagnosis predicts a worse pain prognosis in early RA, independently of inflammation. Early identification may support more targeted and effective pain management.