基于up-to-seven标准肝细胞癌肝移植术后生存预测模型的建立和验证

目的 建立并验证基于up-to-seven（Up7）标准肝细胞癌（HCC）肝移植术后患者的长期生存Cox回归预测模型，以辅助制定临床决策。方法 回顾性分析251例符合Up7标准的HCC肝移植术后患者的临床和随访资料。采用逐步回归向前法进行多因素Cox回归分析，获得HCC肝移植术后患者长期生存的独立预测因素，并建立长期生存Cox回归预测模型。使用R 3.4.3软件获得预测模型评分，采用生存决策树方式确定模型的截断值。绘制预测模型在其他肝移植标准[上海复旦标准、加利福尼亚大学旧金山分校（UCSF）标准、意大利米兰（Milan）标准]下HCC肝移植术后患者的Kaplan-Meier生存曲线，并采用log-rank检验分析组间差异。采用受试者工作特征（ROC）曲线检验预测模型的预测效能。结果 多因素Cox回归分析显示，甲胎蛋白（AFP）、总胆红素（T-Bil）、微血管侵犯（MVI）、肿瘤最大径（Diameter）是HCC肝移植术后患者长期生存的独立预测因素，据此建立的长期生存Cox回归预测模型命名为ATMD（AFP，T-Bil，MVI，Diameter）模型：h（t，x）=h0（t）exp[0.284×肿瘤最大径（cm）+0.773×MVI（是=1；否=0）+0.404×lg AFP（ng/mL）+0.003×T-Bil（μmol/L）]，根据判别生存树设定ATMD模型截断值为1.44，评分>1.44为高危组，≤1.44为低危组。符合Up7标准的高危组和低危组患者分别为87例和164例，符合上海复旦标准的分别为33例和144例，符合UCSF标准的分别为29例和134例，符合Milan标准的分别为29例和131例。Kaplan-Meier生存曲线分析显示，在Up7标准、上海复旦标准、UCSF标准和Milan标准下ATMD模型高危组和低危组患者累积生存率差异均有统计学意义（P<0.001，P=0.008，P<0.001，P=0.001），ATMD模型预测的HCC肝移植术后3年生存的ROC曲线下面积分别是76.63%、75.87%、73.32%和69.41%。结论 ATMD模型对于符合Up7标准、上海复旦标准、UCSF标准和Milan标准的HCC肝移植术后生存情况有良好的预测能力，对符合以上标准的HCC肝移植患者的术前决策和术后风险评估有重要意义。
Objective To establish a long-term survival prediction model for hepatocellular carcinoma (HCC) patients after liver transplantation based on up-to-seven (Up7) criteria, and to validate the prediction model in different liver transplantation criteria, so as to assist clinical decision-making for the treatment of HCC. Methods We retrospectively analyzed the clinical and follow-up data of 251 HCC patients who underwent liver transplantation with Up7 criteria. Stepwise regression method was used to conduct multivariate Cox regression analysis to obtain the independent predictors of long-term survival after HCC liver transplantation, and to establish the survival Cox regression prediction model. R 3.4.3 software was used to score the prediction model, and the decision tree technique was used to determine the cut-off value. The Kaplan-Meier survival curve of the HCC patients after liver transplantation was drawn to validate the prediction model in different criteria (Shanghai Fudan criteria, University of California, San Francisco[UCSF] criteria and Italy Milan criteria), and the difference between groups was analyzed by log-rank test. The receiver operating characteristic (ROC) curve was used to test the predictive effectiveness of the model. Results Multivariate Cox regression analysis suggested that α-fetoprotein (AFP), total bilirubin (T-Bil), microvascular invasion (MVI) and tumor maximal diameter (Diameter) were the independent predictors of long-term survival of HCC liver transplant recipients after liver transplantion. We built the ATMD (AFP, T-Bil, MVI, Diameter)