基于SEER数据库构建并验证ⅢB期非小细胞肺癌患者预后模型

Constructing and validating a prognostic model for patients with stage IIIB non-small cell lung cancer based on SEER database

目的探讨ⅢB期非小细胞肺癌(NSCLC)患者的预后,并构建临床可广泛应用的预后预测工具。方法应用监测、流行病学和最终结果(Surveillance,Epidemiology,and End Result,SEER)数据库回顾性收集ⅢB期非小细胞肺癌患者临床信息,将患者以7∶3比例随机分为训练集和验证集。另收集2005至2019年就诊于山东大学附属山东省立医院胸外科的ⅢB期非小细胞肺癌患者信息作为次要验证集(SPH验证集)。在训练集中进行单因素和多因素Cox比例风险回归分析以构建列线图和网络工具。应用R软件进行ROC曲线、C指数和DCA曲线分析以评估列线图预后模型以及临床获益。结果研究共纳入7903例患者(其中训练集5532例,验证集2371例)。预后列线图中最终纳入10个变量,包括性别、年龄、肿瘤原发部位、肿瘤分化程度、肿瘤组织学类型、TNM分期(第八版ⅢB期具体亚组)、是否行区域淋巴结清扫、手术范围、是否行放疗和是否行化疗。在训练集和验证集中,5年AUC值分别为0.724和0.708。训练集、验证集和SPH验证集的C指数分别为0.661、0.649和0.685。校准图显示模型具有良好的预测能力。相较于单纯放化疗,ⅢB期患者接受手术联合化疗后预后最佳(P<0.001)。此外,研究可视化预后模型并构建交互式网络工具。结论成功构建预后列线图以及交互式网络工具。

Objective To investigate the prognosis of patients with stage IIIB non-small cell lung cancer(NSCLC)and to construct a prognostic tool that can be widely used in clinical practice.Methods Clinical data of patients with stage IIIB NSCLC were retrospectively collected from the Surveillance,Epidemiology and End Result(SEER)databases,and randomly grouped into the training set and validation set with a 7∶3 ratio.Information of patients with stage IIIB NSCLC treated in the Department of Thoracic Surgery of Shandong Provincial Hospital Affiliated to Shandong University during 2005 and 2019 was collected as a secondary validation set(SPH validation set).Univariate and multivariate Cox proportional hazards regression analyses were conducted in the training set to construct a nomogram and web tool.Receiver operating characteristic(ROC)curve,concordance index(C-index),and decision curve analyses were performed using R software to evaluate the nomogram and assess the clinical benefits.Results A total of 7,903 patients were involved,including 5,532 in the training set and 2,371 in the validation set.Altogether 10 variables were included in the nomogram,including gender,age,primary site of tumor,grade of tumor,histology of tumor,tumor-node-metastasis stage specific subgroups ofⅢB stage in the eighth edition,dissection of regional lymph node,extent of surgery,radiotherapy and chemotherapy.The area under the ROC curve(AUC)of 5-year overall survival(OS)was 0.724 and 0.708 in the training and validation sets,respectively.The C-index was 0.661,0.649 and 0.685 in the training,validation and SPH validation sets,respectively.The calibration plot showed that the model had good predictive power.Surgery with chemotherapy was identified as the optimal treatment strategy compared to chemotherapy or radiotherapy alone(P<0.001).An interactive web tool was also constructed.Conclusion A prognostic nomogram was successfully constructed and an interactive web tool to predict patients’survival was built.