Large language models (LLMs) have demonstrated remarkable capabilities in natural language understanding and reasoning. However, their real-world applicability in high-stakes medical assessments remains underexplored, particularly in non-English contexts. This study aims to evaluate the performance of DeepSeek-R1 and ChatGPT-4o on the Chinese National Medical Licensing Examination (NMLE), a comprehensive benchmark of medical knowledge and clinical reasoning. We evaluated the performance of ChatGPT-4o and DeepSeek-R1 on the Chinese National Medical Licensing Examination (2019–2021) using question-level binary accuracy (correct = 1, incorrect = 0) as the outcome. A generalized linear mixed model (GLMM) with a binomial distribution and logit link was used to examine fixed effects of model type, year, and subject unit, including their interactions, while accounting for random intercepts across questions. Post hoc pairwise comparisons were conducted to assess differences across model–year interactions. DeepSeek-R1 significantly outperformed ChatGPT-4o overall (β = − 1.829, p < 0.001). Temporal analysis revealed a significant decline in ChatGPT-4o’s accuracy from 2019 to 2021 (p < 0.05), whereas DeepSeek-R1 appeared to maintain a more stable performance. Subject-wise, Unit 3 showed the highest accuracy (β = 0.344, p = 0.001) compared to Unit 1. A significant interaction in 2020 (β = − 0.567, p = 0.009) indicated an amplified performance gap between the two models. These results highlight the importance of model selection and domain adaptation. Further investigation is needed to account for potential confounding factors, such as variations in question difficulty or language biases over time, which could also influence these trends. This longitudinal evaluation highlights the potential and limitations of LLMs in medical licensing contexts. While current models demonstrate promising results, further fine-tuning is necessary for clinical applicability. The NMLE offers a robust benchmark for future development of trustworthy AI-assisted medical decision support tools in non-English settings.
Background: Lenvatinib, a type of tyrosine kinase inhibitor (TKI), has become a standard molecularly targeted therapy for liver cancer. Lenvatinib targets multiple receptors and inhibits the kinase activity of vascular endothelial growth factor receptors, fibroblast growth factor receptors and platelet-derived growth factor receptor alpha. Numerous case reports and clinical trials have shown that Lenvatinib intake can increase the incidence of cardiovascular diseases, the most of which is hypertension. However, the mechanism by which Lenvatinib causes various vascular diseases is still unclear. In our hospital, we attended to a patient who had suffered from hypertension, acute coronary artery dissection and carotid artery stenosis after Lenvatinib therapy due to primary liver carcinoma in his fifties. Considering he had no risk factor and all these conditions were concerned with endothelial function, we proposed a hypothesis that Lenvatinib can directly influence the functions of endothelium, which plays an important role in maintaining vascular homeostasis, causing all kinds of vascular conditions. Methods: We collected the patient's medical history and performed carotid endarterectomy for him. After being reviewed and approved by the hospital ethics committee, we performed immunofluorescence staining on the pathological specimen obtained during surgery and a specimen from another patient with carotid artery stenosis who did not take Lenvatinib, and compared the expression levels of some important endothelial function markers. Both patients had similar medical histories except history of liver cancer, intake of Lenvatinib and onset of acute coronary artery dissection. Results: The plaques from both patients had intact CD31-positive endothelial layers. The levels of zonula occludens-1 (ZO-1), which is a critical component of tight junctions between cells, were comparable in both patients’ endothelium. However, the expression of endothelial nitric oxide synthase (eNOS), which is crucial for regulating vasodilation and maintaining vascular permeability, was markedly lower in the endothelium of the patient who took Lenvatinib. Conclusions: Lenvatinib intake can disrupt the endothelial function by reducing the expression level of eNOS, leading to a series of vascular diseases. This finding provides a new direction for detailed basic research and clinical prevention exploration in vascular adverse reaction of Lenvatinib and other TKIs.
OBJECTIVE:Ablation is an important therapy for hepatocellular carcinoma (HCC). This study aims to explore the changes in pre- and short-term-post-ablative two-dimensional shear wave elastography (2D-SWE) measurements and assess the feasibility of using pre- and short-term-post-ablative indicators, including 2D-SWE measurements, to predict HCC progression after curative ablation. METHODS:A total of 161 treatment-naïve chronic hepatitis B-related HCC patients were divided into training (n = 107) and validation (n = 54) groups. Data, including 2D-SWE measurements (Emean), were collected pre-ablation, 1 mo post-ablation, and at other scheduled follow-ups. The independent predictor of progression-free survival (PFS) was analyzed using Cox regression analysis, and a prediction model was established. RESULTS:The median follow-up time was 12 mo (range 1-36 mo), and tumor progression occurred in 36 (22.4%) patients. Compared to the pre-ablation, the 1-mo-post-ablative Emean decreased in 40 (24.8%) patients, while increased in 71 (44.1%) patients, and did not significantly change in 50 (31.1%) patients. Cox analysis identified pre-ablative prothrombin time (PT; p = 0.014, HR = 1.54), 1-mo-post-ablative alpha-fetoprotein (AFP; p = 0.014, HR = 1.01), and categorized Emean change between pre-ablative and 1-mo-post-ablative period (with decrease as reference, not-significant-change: p = 0.096, HR = 0.38; increase: p = 0.014, HR = 0.22) as the independent predictor of PFS. The Cox model could significantly predict PFS in both training (p < 0.001, HR = 2.10) and validation (p = 0.009, HR = 1.17) groups. CONCLUSION:Longer pre-ablation PT, higher 1-mo post-ablation AFP, and decreased 1-mo post-ablation Emean were associated with shorter post-ablation PFS. Changes in 2D-SWE measurements may serve as a potential indicator of HCC progression after ablation.
Background and purpose: Percutaneous coronary intervention (PCI) for chronic total occlusions (CTO) remains challenging, with lower success rates compared to non-CTO procedures, particularly in the presence of calcified or fibrotic lesions. This study evaluates a novel intravascular piezoelectric wire system (IPWS) designed to facilitate guidewire advancement through these difficult CTO. Methods: We conducted a two-phase investigation: a preclinical feasibility and safety study in a porcine model, followed by a first-in-human clinical trial. The IPWS consists of a shockwave generator and a disposable guidewire that delivers high-amplitude mechanical pulses to facilitate penetration through calcified or fibrotic lesions. In the clinical phase, the system was used in 10 patients with CTO where standard antegrade wire escalation had failed. Results: In the preclinical study, the IPWS demonstrated excellent device compatibility with microcatheters and showed no significant vascular injury or thrombosis on histology. In the clinical trial, the IPWS achieved a 90% procedural success rate, successfully recanalizing nine of 10 CTO. No major adverse cardiac events (MACE) occurred during the 30-day follow-up. Mild complications, such as slow blood flow and marginal branch involvement, were transient and resolved without further intervention. Conclusions: The IPWS is a safe and feasible technology that may improve the success rate in complex CTO PCI. By facilitating wire passage through challenging lesions, the IPWS could reduce procedural complexity and risk. Further refinements are warranted, but this technology holds significant promise for advancing the treatment of CTO and other complex vascular interventions.
Background: To explore the value and significance of triple immunohistochemical detection of alpha-methylacyl-CoA racemase (AMACR), p63 and Ki-67 in the pathological diagnosis of prostate cancer. Methods: A total of 156 paraffin-archived prostate biopsy samples collected from our hospital from June 2023 to August 2024 were selected as research materials. The expression levels of AMACR, p63 and Ki-67 in puncture samples from 48 patients with prostate cancer, 32 patients with high-grade prostate intraepithelial neoplasia, 26 patients with low-grade prostate intraepithelial neoplasia and 50 patients with benign prostatic hyperplasia were detected via immunohistochemistry. Results: There were statistically significant differences in the positive expression rates of AMACR, p63 and Ki-67 among the different groups of samples. The positive expression rate of AMACR in the puncture samples of the prostate cancer group was 100%, with a high expression rate of 81.25%. The negative expression rate of p63 was 97.92%, which was significantly greater than that of the other three groups. The positive expression rate of Ki-67 was 81.25%, and the high expression rate was 54.17%. The AMACR is related to the long diameter of the tumour, TNM stage, degree of differentiation and Gleason score in patients with prostate cancer. The rates of high AMACR in patients with tumours with long diameters >= 1.5 cm, stage moderate to highly differentiated, and Gleason scores ranging from 8-10 were significantly higher than those in patients with long diameters <1.5 cm, stage I. The expression of Ki-67 is related to the long diameter of the tumour, degree of differentiation, degree of lymph node metastasis and Gleason score in patients with prostate cancer. The rates of high expression of Ki-67 in patients with a long tumor diameter of >= 1.5 cm, moderate and high differentiation, and lymph node metastasis and a Gleason score of 8-10 were significantly greater than those in patients with a long tumor diameter of <1.5 cm, poor differentiation, no lymph node metastasis, and a Gleason score of 2-7. p63 was not related to the clinical or pathological characteristics of patients with prostate cancer (all P>0.05). The sensitivity and negative predictive value of the combined diagnosis of AMACR-positive/p63-nega-tive/Ki-67-positive prostate cancer were both 100.00%, and the specificity was 81.82%. Conclusions: AMACR, p63 and Ki-67 in prostate biopsy samples can be used as promising biomarkers for the diagnosis or exclusion of prostate cancer. Combined detection can improve the accuracy of prostate cancer diagnosis.
