Ⅱ/Ⅲ期胃癌术后患者预后预测模型的建立

Establishment of a prognosis prediction model for postoperative patients with stageⅡ/Ⅲgastric cancer

目的:建立预测Ⅱ/Ⅲ期胃癌患者术后总生存期(OS)和无病生存期(DFS)的列线图模型。方法:回顾性收集2012年1月至2017年12月在医院接受根治性手术的535名Ⅱ/Ⅲ期胃癌患者资料,随机分为训练集(n=428)和验证集(n=107)。胃癌诊断由组织病理学确定,并由第8版AJCC-TNM分期系统确定临床分期。研究终点为OS和DFS。在训练集中,单因素和多因素Cox回归分析用于确定预后的独立预测因素。基于这些因素,绘制Ⅱ/Ⅲ期胃癌术后OS和DFS的列线图。采用校准曲线和一致性指数(C指数)评估模型性能。结果:年龄、浸润深度、淋巴结转移状态、分化程度、人血白蛋白/纤维蛋白原比值和癌胚抗原为OS的预测因素。浸润深度、淋巴结转移状态、分化程度、血小板/淋巴细胞比值和癌胚抗原为DFS的预测因素。由此构建的OS列线图和DFS列线图表现出良好一致性和区分度,C指数HR(95%CI)分别为0.726(0.688~0.763)和0.720(0.682~0.758),优于传统第8版TNM分期系统[HR(95%CI)分别为0.650(0.616~0.683)和0.649(0.615~0.683)](P<0.001)。并通过验证集进行了内部验证,OS列线图和DFS列线图具有良好的临床预测性能。结论:本研究建立的列线图模型可以有效预测Ⅱ/Ⅲ期胃腺癌患者根治术后的OS和DFS。

Aim:To establish a nomogram model to predict the prognosis of postoperative patients with stageⅡ/Ⅲgastric cancer.Methods:A total of 535 patients with stageⅡ/Ⅲgastric cancer who accepted radical gastrectomy at the First Affiliated Hospital of Zhengzhou University from January 2012 to December 2017,were included in this retrospective study.These patients were randomly allocated into a training cohort(n=428)and a validation cohort(n=107).Univariate and multivariate analysis were used to determine independent prognostic factors.Nomograms for predicting the overall survival(OS)and disease-free survival(DFS)of the patients were developed based on the prognostic factors.The concordance index(C-index)and calibration curves were used to evaluate the predictive accuracy of the nomogram models.Results:Multivariate analysis indicated that age,differentiation grade,the depth of tumor invasion,lymph node metastasis,albumin-to-fibrinogen ratio(AFR)and carcinoembryonic antigen(CEA)were independent predictive factors for OS.Differentiation grade,the depth of tumor invasion,lymph node metastasis,platelet-to-lymphocyte ratio(PLR)and CEA were independent predictive factors for DFS.In the training cohort,nomograms for predicting OS and DFS showed better discrimination than the 8th AJCC TNM stage system alone[C-index:0.726(0.688-0.763)vs 0.650(0.616-0.683),0.720(0.682-0.758)vs 0.649(0.615-0.683),P<0.001].A similar conclusion was drawn for the validation cohort,calibration curves showed a good discrimination of the model.Conclusions:Nomograms predicting OS and DFS have been established and show good predictive performance.