Evidence-based medicine (EBM) represents a paradigm of providing patient care grounded in the most current and rigorously evaluated research. Recent advances in large language models (LLMs) offer a potential solution to transform EBM by automating labor-intensive tasks and thereby improving the efficiency of clinical decision-making. This study explores integrating LLMs into the key stages in EBM, evaluating their ability across evidence retrieval (PICO extraction, biomedical question answering), synthesis (summarizing randomized controlled trials), and dissemination (medical text simplification). We conducted a comparative analysis of seven LLMs, including both proprietary and open-source models, as well as those fine-tuned on medical corpora. Specifically, we benchmarked the performance of various LLMs on each EBM task under zero-shot settings as baselines, and employed prompting techniques, including in-context learning, chain-of-thought reasoning, and knowledge-guided prompting to enhance their capabilities. Our extensive experiments revealed the strengths of LLMs, such as remarkable understanding capabilities even in zero-shot settings, strong summarization skills, and effective knowledge transfer via prompting. Promoting strategies such as knowledge-guided prompting proved highly effective (e.g., improving the performance of GPT-4 by 13.10% over zero-shot in PICO extraction). However, the experiments also showed limitations, with LLM performance falling well below state-of-the-art baselines like PubMedBERT in handling named entity recognition tasks. Moreover, human evaluation revealed persisting challenges with factual inconsistencies and domain inaccuracies, underscoring the need for rigorous quality control before clinical application. This study provides insights into enhancing EBM using LLMs while highlighting critical areas for further research. The code is publicly available on Github.

Inflammatory bowel disease (IBD) is a global disease that is evolving with increasing incidence. However, there are few works on computationally assisted diagnosis of IBD based on pathological images. Therefore, based on the UK and Chinese IBD diagnostic guidelines, our study established an artificial intelligence-assisted diagnostic system for histologic grading of inflammatory activity in ulcerative colitis (UC). We proposed an efficient deep-learning (DL) method for grading inflammatory activity in whole-slide images (WSIs) of UC pathology. Our model was constructed using 603 UC WSIs from Nanjing Drum Tower Hospital for model train set and internal test set. We collected 212 UC WSIs from Zhujiang Hospital as an external test set. Initially, the pre-trained ResNet50 model on the ImageNet dataset was employed to extract image patch features from UC patients. Subsequently, a multi-instance learning (MIL) approach with embedded self-attention was utilized to aggregate tissue image patch features, representing the entire WSI. Finally, the model was trained based on the aggregated features and WSI annotations provided by senior gastrointestinal pathologists to predict the level of inflammatory activity in UC WSIs. In the task of distinguishing the presence or absence of inflammatory activity, the Area Under Curve (AUC) value in the internal test set is 0.863 (95% confidence interval [CI] 0.829, 0.898), with a sensitivity of 0.913 (95% [CI] 0.866, 0.961), and specificity of 0.816 (95% [CI] 0.771, 0.861). The AUC in the external test set is 0.947 (95% confidence interval [CI] 0.939, 0.955), with a sensitivity of 0.889 (905% [CI] 0.837, 0.940), and specificity of 0.858 (95% [CI] 0.777, 0.939). For distinguishing different levels of inflammatory activity in UC, the average Macro-AUC in the internal test set and the external test set are 0.827 (95% [CI] 0.803, 0.850) and 0.908 (95% [CI] 0.882, 0.935). the average Micro-AUC in the internal test set and the external test set are 0.816 (95% [CI] 0.792, 0.840) and 0.898 (95% [CI] 0.869, 0.926). Comparative analysis with diagnoses made by pathologists at different expertise levels revealed that the algorithm reached a proficiency comparable to the pathologist with 5 years of experience. Furthermore, our algorithm performed superior to other MIL algorithms.

Qi Sun, Yao Fu, Xiaobing Chen, Lin Li, Hongyan Wu, Yixuan Liu, Haojun Xu, Guoren Zhou, Xiangshan Fan, and Hongping Xia. Gut and Liver 2022;16:875-91. https://doi.org/10.5009/gnl210359

Background: KRAS is the most frequently mutated oncogene in human cancers, with KRAS^G12C being a prevalent driver mutation in 12–13% non-small cell lung cancer (NSCLC) cases. Despite breakthroughs in KRAS^G12C inhibitors such as sotorasib (AMG-510) and adagrasib (MRTX-849), clinical resistance remains a challenging issue, highlighting the need for deeper understanding of the molecular mechanisms underlying KRAS^G12C-driven oncogenic signaling in NSCLC. Previously, we identified RASON as a novel regulator of KRAS^G12D/V signaling in pancreatic cancer. Herein, we aim to explore the role of RASON in KRAS^G12C-driven NSCLC and its therapeutic potential. Methods: Immunohistochemistry analysis of NSCLC patient cohorts was performed to demonstrate the correlation between RASON expression and NSCLC progression. Immunoblotting was performed to evaluate the effects of RASON on KRAS^G12C downstream signaling. In vitro and in vivo assays including cell proliferation, sphere formation, tumor implantation and genetic mouse models were performed to determine the oncogenic role of RASON. RNA-seq analysis was utilized to identify the key signaling pathway regulated by RASON. Immunofluorescence, immunoprecipitation, nuclear magnetic resonance and biochemistry assays were used to validate the interaction between KRAS^G12C and RASON. Phagocytosis assay and flow cytometry were conducted to explore the effects of RASON on the tumor immune microenvironment. Pharmacological inhibition in subcutaneous xenograft model was used to determine the therapeutical potential of RASON. Results: RASON is overexpressed in NSCLC with KRAS^G12C mutation and correlates with poor patient prognosis. Genetic knockout of RASON significantly reduced lung tumor burden in LSL-KRAS^G12D; Trp53^R172H/+ mice. In KRAS^G12C-mutant lung cancer cell lines, RASON overexpression enhanced, while CRISPR-mediated knockout suppressed, both in vitro proliferation and in vivo tumor growth. Mechanistically, RASON directly binds KRAS^G12C, stabilizes it in the GTP-bound hyperactive state and promotes downstream signaling. RASON knockout significantly reduced CD47 expression, enhancing macrophage-mediated phagocytosis and anti-tumor immunity. Therapeutically, antisense oligonucleotides targeting RASON not only exhibited tumor-suppressive effects, but also synergized with the KRAS^G12C inhibitor AMG-510 to significantly enhance anti-tumor efficacy. Conclusion: This study reveals RASON as a key oncogenic regulator of KRAS^G12C signaling, driving lung tumorigenesis and progression, and identifies RASON as a promising therapeutic target for KRAS^G12C mutant non-small cell lung cancer.

Background: Readmission shortly after discharge is indicative of an increased disease severity for patients with ulcerative colitis (UC) and ineffectiveness to medical therapy, which may contribute to a dismal prognosis. Objectives: This study aimed to explore prognostic variables with a nomogram to predict unplanned UC-related readmission within 1 year after discharge. Design: A retrospective cohort study. Methods: Electronic medical records of all UC patients treated at our center between 1 January 2014 and 31 June 2021 were reviewed. A comprehensive analysis of various characteristics, such as demographics, comorbidities, medical history, follow-up appointments, admission endoscopy, histopathologic features, etc., was used to determine the primary end point, which was unplanned UC-related calendar year readmission. Results: We found that the unplanned UC-related readmission rate within 1 year was 20.8%. In multivariable cox analysis, the predictors of the Elixhauser comorbidity index [Hazard ratio (HR): 3.50, 95% confidence interval (CI): 1.93–6.37], regular follow-up (HR: 0.29, 95% CI: 0.16–0.53), any history of corticosteroid use (HR: 3.38, 95% CI: 1.83–6.27), seral level of C-reactive protein (HR: 1.01, 95% CI: 1.00–1.02), and the UC endoscopic index of severity (HR: 1.29, 95% CI: 1.05–1.57) independently predicted calendar year readmission after discharge. The established nomogram had a consistently high accuracy in predicting calendar year readmission in the training cohort, with a concordance index of 0.784, 0.825, and 0.837 at 13, 26, and 52 weeks, respectively, which was validated in both the internal and external validation cohorts. Therefore, UC patients were divided into clinically low-, high-, and extremely high-risk groups for readmission, based on the calculated score of 272.5 and 378. Conclusion: The established nomogram showed good discrimination and calibration powers in predicting calendar year readmission in high-risk UC patients, who may need intensive treatment and regular outpatient visits.