基于SEER数据库混合型肝癌列线图预测模型的构建及其应用价值

Construction and application value of nomogram predictive model for mixed-type liver cancer based on SEER database

目的探讨基于美国国家癌症研究所监测、流行病学和最终结果(SEER)数据库混合型肝癌(CHC)预后危险因素及列线图预测模型构建,并探讨其应用价值。方法采用回顾性队列研究方法。收集2004年1月至2019年12月SEER数据库208例CHC患者的临床病理资料;男150例,女58例;年龄≥60岁121例,<60岁87例。208例CHC患者按照随机数字表法按7∶3分为训练集145例和验证集63例。使用COX比例风险模型构建列线图预测模型,验证并比较列线图预测模型和第8版美国癌症联合委员会(AJCC)分期预测模型的预测效能。观察指标:(1)CHC患者生存情况。(2)CHC患者预后影响因素分析。(3)列线图预测模型的构建和效能评价。正态分布的计量资料以■±s表示,偏态分布的计量资料以M(范围)表示。计数资料以绝对数表示,组间比较采用χ^(2)检验。等级资料比较采用秩和检验。采用COX回归模型进行单因素及多因素分析。采用Kaplan⁃Meier法计算生存率,采用Log⁃Rank检验进行生存分析。采用一致性指数(C⁃index)、受试者工作特征曲线下面积评价预测模型的预测效能;校准曲线验证预测模型的准确性;决策曲线评价预测模型的临床效能。结果(1)CHC患者生存情况。145例训练集CHC患者生存时间为15.0(1.0~166.0)个月,1年总生存率为52.7%。63例测试集患者生存时间为11.5(1.0~176.0)个月,1年总生存率为48.3%。(2)CHC患者预后影响因素分析。多因素分析结果显示:T分期、手术治疗、肿瘤分化程度、化疗是CHC患者预后的独立影响因素(P<0.05)。(3)列线图预测模型的构建和效能评价。根据多因素分析结果,构建列线图预测模型。训练集和验证集列线图预测模型的C⁃index分别为0.758和0.742;训练集和验证集AJCC分期预测模型的C⁃index分别为0.622和0.662。训练集和验证集CHC患者列线图预测模型的1、3、5年总生存率受试者工作特征曲线下面积分别为0.830、0.861、0.887和0.798、0.813、0.844;训练集和验证集CHC患者AJCC分期预测模型的1、3、5年总生存率受试者工作特征曲线下面积分别为0.668、0.699、0.747和0.719、0.760、0.796。校准曲线结果显示:1年总生存率校准曲线与理想的斜率为1的直线拟合良好,2种预测模型的预测结果与实际结果具有较高的一致性。决策曲线结果显示:列线图预测模型比AJCC分期预测模型具有更好的预测效能。结论T分期、手术治疗、肿瘤分化程度、化疗是CHC患者预后的独立影响因素,基于SEER数据库构建的列线图预测模型可更精准地预测和评估CHC患者的预后。

Objective To investigate the prognostic factors of combined hepatocellular-cholangiocarcinoma(CHC)based on Surveillance,Epidemiology,and End Results(SEER),and to construct and investigate the application value of a nomogram predictive model for CHC.Methods The retrospective cohort study was conducted.The clinicopathological data of 208 patients with CHC entered into SEER database from January 2004 to December 2019 were analyzed.There were 150 males and 58 females,121 cases with age≥60 years,87 cases with age<60 years.All the 208 patients with CHC were divided into a training set and a validation set in a 7∶3 ratio based on the random table method.The COX proportional hazard model was used to construct the nomogram predictive model,which was validated and compared the predictive performance with the eighth edition of the American Joint Committee on Cancer(AJCC).Observation indicators:(1)survival of patients with CHC;(2)prognostic factors analysis of patients with CHC;(3)construction and performance evaluation of the nomogram predictive model.Measurement data with normal distribution were represented as Mean±SD,and measurement data with skewed distribution were represented as M(range).Count data were described as absolute numbers,and comparison between groups was analyzed using the chi-square test.Comparison of ordinal data was analyzed using the rank sum test.The COX regression model was used for univariate and multivariate analyses.The Kaplan-Meier method was used to calculate survival rate.The Log-Rank test was used for survival analysis.The prognostic efficacy of the predictive model was evaluated using the consistency index(C-index)and the area under curve(AUC)of receiver operating characteristic(ROC)curve.The calibration curve was used to validate the accuraly of predictive model and decision curve was used to evaluate the clinical utility of predictive model.Results(1)Survival of patients with CHC.The overall survival time of 145 patients with CHC in the training set was 15.0(range,1.0−166.0)months,with 1-year survival rate as 52.7%.The overall survival time of 63 patients with CHC in the validation set was 11.5(range,1.0−176.0)months,with 1-year survival rate as 48.3%.(2)Prognostic factors analysis of patients with CHC.Results of multivariate analysis showed that T staging,surgery,degree of tumor differentiation and chemotherapy were independent factors for prognosis of patients with CHC(P<0.05).(3)Construction and performance evaluation of the nomogram model.The nomogram predictive model was constructed based on results of multiveriate analysis.The C-index of nomogram predictive model in the training set and the validation set was 0.758 and 0.742,respectively.The C-index of the AJCC predictive model in the training set and the validation set was 0.622 and 0.662,respectively.The AUC of ROC of the nomogram predictive model for 1-,3-,5-year overall survival rates was 0.830,0.861,0.881 for CHC patients in the training set and 0.798,0.813,0.844 for CHC patients in the validation set.The AUS of ROC of AJCC predictive model for 1-,3-,5-year overall survival rates was 0.668,0.699,0.747 for CHC patients in the training set and 0.719,0.760,0.796 for CHC patients in the validation set.Results of calibration curve showed that 1-year overall survival rate calibration curve fitted well with the ideal straight line with a slope of 1,and the predicted results of the two prediction models were highly consistent with the actual results.Results of decision curve showed that nomogram predictive model had a better prediction efficiency than the AJCC predictive model.Conclusion T staging,surgery,degree of tumor differentiation,chemotherapy are independent factors for prognosis of patients with CHC,and the nomogram constructed based on SEER database can more accurately evaluate the prognosis of CHC patients.