基于SEER数据库的鼻咽癌患者预后预测模型的构建及验证

Construction and verification of prognostic prediction model for nasopharyngeal carcinoma patients based on SEER database

目的利用监测、流行病学和最终结果(SEER)数据库分析鼻咽癌患者的独立预后因素,建立并验证预后预测模型,以评估患者预后并为个体化治疗提供依据。方法回顾性检索SEER数据库中2010-2018年诊断为鼻咽癌患者的病历信息,以7∶3比例分为SEER训练集和SEER验证集,比较SEER训练集和SEER验证集患者基线特征,通过单因素及多因素Cox比例风险回归模型确定鼻咽癌患者独立预后因素,使用R语言构建鼻咽癌患者1、3、5年总生存期(OS)可视化列线图,预测患者1、3、5年死亡风险,使用一致性指数(C-index)、受试者工作特征(ROC)曲线和校准曲线对模型的预测能力进行验证。结果共纳入1602例鼻咽癌患者。患者年龄、婚姻状况、临床分期、M分期、放疗情况是OS的独立预后因素(P<0.05)。根据得到的独立预后因素构建鼻咽癌患者OS列线图。SEER训练集中的C-index为0.717,1、3、5年生存率的ROC曲线下面积(AUC)分别为0.772、0.760、0.758;SEER验证集中的C-index为0.748,1、3、5年生存率的AUC分别为0.805、0.767、0.744。校准曲线与标准参考曲线重叠较多。结论基于SEER数据库建立的预测模型能较好地预测鼻咽癌患者预后。

Objective To analyze the independent prognostic factors of patients with nasopharyngeal carcinoma by using the Surveillance,Epidemiology and End Results(SEER)database,and to establish and verify the prognostic prediction model,so as to evaluate the prognosis of patients and provide the basis for individualized treatment.Methods The medical records of patients diagnosed with nasopharyngeal carcinoma from 2010 to 2018 in the SEER database was retrospectively retrieved.The patients were divided into SEER training set and SEER validation set at a ratio of 7∶3.The baseline characteristics of patients in the SEER training set and SEER validation set were compared.The independent prognostic factors of patients with nasopharyngeal carcinoma were determined by univariate and multivariate Cox proportional hazard regression model.The 1-,3-,and 5-year overall survival(OS)visual nomogram of patients with nasopharyngeal carcinoma was constructed using R language to predict the 1-,3-,and 5-year death risk of patients.The consistency index(C-index),receiver operating characteristic(ROC)curve and calibration curve were used to verify the predictive ability of the model.Results A total of 1602 patients with nasopharyngeal carcinoma were included.Age,Marital status,clinical Stage,M stage and radiotherapy were independent prognostic factors of OS(P<0.05).The OS nomogram of patients with nasopharyngeal carcinoma was constructed according to the obtained independent prognostic factors.The C-index in the SEER training set was 0.717,and the area under the ROC curve(AUC)of 1-,3-,and 5-year survival rates was 0.772,0.760 and 0.758,respectively.The C-index in the SEER validation set was 0.748,and the AUC of 1-,3-,and 5-year survival rates were 0.805,0.767,and 0.744,respectively.The calibration curve and the standard reference curve overlapped more.Conclusion The predictive model based on SEER database can better predict the prognosis of patients with nasopharyngeal carcinoma.