As one of the hottest artificial intelligence technologies currently, ChatGPT, as one of the hottest artificial intelligence technologies today, plays a significant role in advancing the field of evidence-based medicine, particularly in expanding the sources of original evidence, enhancing the efficiency of evidence acquisition, aiding in shared decision-making between doctors and patients, and promoting education in evidence-based medicine and public science education. Presently, ChatGPT is in its "technological budding phase" and it is crucial to be wary of the risks it brings, such as "evidence contamination", algorithmic black boxes, security vulnerabilities, and the digital divide. To balance the positive effects and potential risks of ChatGPT in the realm of evidence-based medicine, we offer countermeasures and suggestions from the perspectives of ChatGPT's ethical standards, evidence sources, expert verification, and usage norms.
ObjectiveTo evaluate the predictive performance of a nomogram model integrating preoperative CT body composition parameters, clinical laboratory indicators, and postoperative pathological TNM (pTNM) staging for overall survival in patients with gastric cancer after radical gastrectomy. MethodsClinical data were retrospectively collected from patients who underwent radical gastrectomy for gastric cancer at the First Affiliated Hospital of Chengdu Medical College from January 2017 to January 2025, including general information, abdominal CT-derived body composition parameters within 1 week before surgery, blood routine test results, tumor markers, and postoperative pTNM staging data. Univariate and multivariate Cox proportional-hazards regression were used to analyze the relation between each indicator and overall survival (variables with P<0.10 in univariate analysis were included in the multivariate analysis). A nomogram was constructed to predict the 1-, 2-, and 3-year overall survival probabilities. The C-index and time-dependent area under the curve (AUC) were used to evaluate model discrimination. Calibration curves were plotted to assess calibration performance, and decision-curve analysis (DCA) was performed to evaluate clinical net benefit. Based on the total risk score derived from the nomogram, the optimal cut-off value was determined using X-tile software, and patients were divided into high- and low-risk groups. Kaplan-Meier survival curves were plotted to compare overall survival differences between the two groups. ResultsA total of 191 eligible gastric cancer patients were enrolled in this study. Multivariate Cox proportional-hazards regression analysis demonstrated that carcinoembryonic antigen (CEA) >5 μg/L (HR=2.696, P=0.001), lymphocyte count <1.1×109/L (HR=2.027, P=0.013), pTNM stage Ⅲ–Ⅳ (HR=3.607, P<0.001), and increase in standardized subcutaneous fat density (HR=1.685, P<0.001) were independent risk factors for shortened overall survival. Collinearity was diagnosed, and variance inflation factors of <10 were observed for all four variables. An AUC of 0.762 [95%CI (0.686, 0.838)] for discriminating overall survival was achieved by the combination of these four indicators. Based on the nomogram risk score constructed from the four indicators, the optimal cutoff value of 5.723 was determined using X-tile software, and patients were divided into a high-risk group (n=59) and a low-risk group (n=132). The overall survival of patients in the high-risk group was poorer than that in the low-risk group (χ2=67.58, P<0.001). A C-index of 0.758 [95%CI (0.689, 0.823)] was calculated for the nomogram model, and after internal validation using the Bootstrap method (1 000 resamples), a C-index of 0.749 [95%CI (0.675, 0.825)] was obtained. The AUCs (95%CIs) for predicting 1-, 2-, and 3-year overall survival probabilities were calculated as 0.788 (0.679, 0.891), 0.811 (0.733, 0.893), and 0.856 (0.790, 0.919), respectively. Calibration curves showed that the predicted 1-, 2-, and 3-year survival probabilities generally approximated the ideal curves. DCA demonstrated that, within a threshold probability range of 5%–50%, clinical net benefit was provided for patients by the nomogram model. ConclusionsThe results of this study indicate that increased standardized subcutaneous fat density quantified by preoperative CT, elevated preoperative CEA levels, decreased preoperative lymphocyte count, and advanced postoperative pTNM stage are independent risk factors for poor overall survival in patients after radical gastrectomy for gastric cancer. The nomogram model constructed based on these factors exhibits good discriminative ability in predicting 1-, 2-, and 3-year overall survival probabilities, and DCA confirms that it offers a positive net benefit in clinical decision-making.