基于SEER数据库构建肝细胞癌患者预后预测新模式

Construction of a new prognostic model for hepatocellular carcinoma patients based on the SEER database

目的研究肝细胞癌患者总生存期的影响因素,构建新的预测模式,为患者的个体化和精准化治疗方案提供理论支持和参考。方法在SEER数据库中获取2011—2015年被诊断为肝细胞癌的患者数据,运用单因素和多因素Cox回归分析筛选预后因素并构建预测模型。采集兰州大学第二医院280例肝细胞癌患者样本数据,作为外部验证队列。应用一致性指数、受试者操作特征曲线、校准曲线、决策曲线评估模型效能并与传统TNM分期进行对比,计算分数对患者进行危险分层并绘制小提琴图。结果回归分析提示年龄、性别、N分期、M分期、甲胎蛋白表达情况、手术情况、放射治疗、化学治疗是影响肝细胞癌患者总生存期的独立预后因素。训练组的一致性指数为0.739(95%CI:[0.730,0.748]),1、3、5年患者总生存期受试者操作特征曲线下面积分别为0.806、0.811、0.812;内部验证组的一致性指数为0.732(95%CI:[0.718,0.746]),1、3、5年患者总生存期受试者操作特征曲线下面积分别为0.787、0.803、0.810;外部验证组的一致性指数为0.719(95%CI:[0.684,0.753]),1、3、5年患者总生存期受试者操作特征曲线下面积分别为0.794、0.841、0.899;传统TNM模型的一致性指数为0.704(95%CI:[0.691,0.714]),1、3、5年患者总生存期受试者操作特征曲线下面积分别为0.716、0.687、0.677。决策曲线显示列线图较传统TNM分期预测效能更好。结论本研究构建的列线图能够用于预测肝细胞癌患者的总生存期,且效果优于传统TNM分期,基于新的预测模型的危险分层系统在区分患者生存期时更精确,对肝细胞癌患者的个体化治疗具有临床参考价值。

Objective To construct a new predictive model to study the factors influencing the overall survival of patients with hepatocellular carcinoma,thereby providing theoretical support and reference for individualized and precise treatment plans for patients.Methods Data of patients diagnosed with hepatocellular carcinoma from 2011 to 2015 were obtained from the SEER database.Univariate and multivariate Cox regression analysis was used to identify prognostic factors and develop prediction models.A validation cohort consisting of sample data from 280 hepatocellular carcinoma patients at the Second Hospital of Lanzhou University was collected for external validation.The model efficacy was evaluated via the consistency index,subject work characteristic curve,calibration curve and decision curve,which were compared with traditional TNM staging.Risk stratification scores calculated and violin plots drawn.Result Regression analysis revealed that age,sex,N stage,M stage,alpha-fetoprotein expression,operation,radiotherapy and chemotherapy were independent prognostic factors affecting the overall survival of patients with hepatocellular carcinoma.The training group's concordance index was 0.739(95%CI:[0.730,0.748]),the area under curve for 1,3,and 5 years were 0.806,0.811,and 0.812 respectively,and the internal validation group's concordance index was 0.732(95%CI:[718,0.746])and the area under curve for 1,3,and 5 years were 0.787,0.803,and 0.810,respectively.The consistency index of the external verification group was 0.719(95%CI:[0.684,0.753])and the area under the curve for the 1,3,and 5years periods were 0.794,0.841,and 0.899 respectively.Additionally,the consistency index of the traditional TNM model was found to be 0.704(95%CI:[0.691,0.714])and the areas under the curve for years 1,3,and 5years were determined to be 0.716,0.687,and 0.677 correspondingly.The decision curve analysis curve revealed that the nomogram had a superior prediction efficiency than the traditional TNM staging.Conclusion The nomogram developed in this study can predict overall survival in hepatocellular carcinoma patients and outperform the traditional TNM staging.Moreover,the risk stratification system based on the new predictive model demonstrates a greater precision in distinguishing patients'survival periods,thereby having clinical reference value for the individualized treatment of hepatocellular carcinoma patients.