METHODS: This retrospective analysis included clinical data from patients with gastric precancerous lesions who underwent ESD at Wuhan University People\'s Hospital between November 2016 and June 2022. An XGBoost model was built to predict delayed bleeding after ESD using risk factors identified by univariable and multivariate logistic regression analysis. The model was evaluated using receiver operating characteristic curves (ROC), and SHapely Additive exPlanations (SHAP) analysis was used to interpret the model.
RESULTS: Seven factors were statistically associated with delayed postoperative bleeding in gastric precancerous lesions after ESD: age, low-grade intraepithelial neoplasia, hypertension, lesion size ≥ 40 mm, operative time ≥ 120 min, female, and nonuse of hemoclips. A risk prediction model was established. In the training cohort, the model achieved an AUC of 0.97 (0.96-0.98), a sensitivity of 0.90, a specificity of 0.94, and an F1 score of 0.91. In the validation cohort, the AUC was 0.94(0.90-0.98), with a sensitivity of 0.85, a specificity of 0.89, and an F1 score of 0.85. In the test cohort, the AUC was 0.94 (0.89-0.99), the sensitivity was 0.80, the specificity was 0.92, and the F1 score was 0.84, indicating strong predictive capability.
CONCLUSIONS: In this study, an XGBoost prediction model for assessing the risk of delayed postoperative bleeding after ESD in patients with gastric precancerous lesions was developed and validated. This model can be applied in clinical practice to effectively predict the risk of post-ESD bleeding for individual patients.
方法:回顾性分析2016年11月至2022年6月在武汉大学人民医院接受ESD治疗的胃癌前病变患者的临床资料。使用单变量和多变量逻辑回归分析确定的危险因素,建立XGBoost模型来预测ESD后延迟出血。使用受试者工作特征曲线(ROC)评估模型,并使用形状加性扩张(SHAP)分析来解释该模型。
结果:7个因素与胃癌前病变ESD术后迟发性出血相关:年龄,低度上皮内瘤变,高血压,病变大小≥40mm,手术时间≥120min,女性,和不使用hemoclips。建立了风险预测模型。在训练组中,该模型的AUC为0.97(0.96-0.98),敏感性为0.90,特异性为0.94,F1评分为0.91.在验证队列中,AUC为0.94(0.90-0.98),敏感性为0.85,特异性为0.89,F1评分为0.85。在测试队列中,AUC为0.94(0.89-0.99),敏感性为0.80,特异性为0.92,F1评分为0.84,提示预测能力强.
结论:在这项研究中,建立并验证了XGBoost预测模型,用于评估胃癌前病变患者ESD术后延迟出血的风险.该模型可应用于临床,有效预测个别患者ESD后出血的风险。