中老年Ⅱ~Ⅲ期胃癌患者预后预测列线图模型的构建和验证

Establishment and verification of a prognostic Nomogram for middle-aged and elderly patients with stage Ⅱ to Ⅲ gastric cancer

目的 构建并验证中老年Ⅱ~Ⅲ期胃癌患者总生存期(OS)的列线图预测模型。方法 回顾性分析2012年3月1日—2022年12月1日在扬州大学附属医院、江苏省苏北人民医院、扬州市中医院就诊的中老年Ⅱ~Ⅲ期胃癌患者的临床、病理及随访资料。基于单因素和多因素Cox回归分析,明确中老年Ⅱ~Ⅲ期胃癌患者OS的独立危险因素,并进一步构建和验证列线图预测模型。通过受试者工作特征(ROC)曲线和校准曲线评估模型的诊断效能,通过决策曲线分析(DCA)评估模型的临床效应。结果 共纳入382例患者,其中282例归为训练集,100例归为验证集。单因素和多因素Cox回归分析提示,胃癌家族史、侵犯脉管、侵犯神经、T分期、N分期是中老年Ⅱ~Ⅲ期胃癌患者OS的独立危险因素(P<0.05)。基于上述变量构建预后列线图,训练集和验证集中模型的一致性指数分别为0.667(95%CI:0.601~0.726)和0.708(95%CI:0.622~0.766);ROC曲线提示模型具有良好的预测准确性;校准曲线显示模型预测值与实际值具有良好的一致性;DCA表明模型具有良好的临床应用价值和潜力。结论 本研究基于真实世界大数据构建的中老年Ⅱ~Ⅲ期胃癌患者1、3、5年OS的列线图模型具有理想的预测效果,有助于临床医师有效评估患者预后。

Objective To construct and validate a Nomogram prediction model for overall survival(OS) in middle-aged and elderly patients with stage Ⅱ to Ⅲ gastric cancer.Methods The clinical,pathological,and follow-up data of middle-aged and elderly patients with stage Ⅱ to Ⅲ gastric cancer in the Affiliated Hospital of Yangzhou University,Northern Jiangsu People' s Hospital,and Yangzhou City Hospital of Traditional Chinese Medicine from March 1,2012 to December 1,2022 were retrospectively analyzed.Based on univariate and multivariate Cox regression analyses,the independent risk factors for OS in middle-aged and elderly patients with stage Ⅱ to Ⅲ gastric cancer were identified,and a Nomogram prediction model was further constructed and validated.The diagnostic performance of the model was evaluated by the receiver operating characteristic(ROC) curve and calibration curve,and the clinical effect of the model was assessed by decision curve analysis(DCA).Results A total of 382 patients were included.A total of 282 cases were as training sets and 100 cases were as validation sets.Univariate and multivariate Cox regression analyses indicated that family history of gastric cancer,vascular invasion,nerve invasion,T stage,and N stage were independent risk factors for OS in middle-aged and elderly patients with stage Ⅱ to Ⅲ gastric cancer(P <0.05).A prognostic Nomogram was constructed based on these variables,and the concordance index of the model in the training and validation sets was 0.667(95% CI,0.601 to 0.726) and 0.708(95% CI,0.622 to0.766) respectively.The ROC curve indicated that the model had good predictive accuracy.The calibration curve showed that the predicted value of the model was in good agreement with the actual value.DC A demonstrated that the model had good clinical application and potential values.Conclusion The Nomogram model for 1-,3-and 5-year OS in middle-aged and elderly patients with stage Ⅱ to Ⅲ gastric cancer constructed based on real-world big data in this study has an ideal predictive effect,which can help clinicians effectively assess patients' prognosis.