肝细胞癌微血管侵犯术前预测列线图的建立与应用

The establishment and application of a preoperative predictive nomogram for hepatocellular carcinoma with microvascular invasion

目的建立术前预测肝细胞癌(HCC)微血管侵犯(MVI)的列线图模型并检验其预测效能。方法回顾分析2014年1月至2017年12月广西医科大学附属肿瘤医院收治的798例HCC患者资料,其中男性690例,女性108例,年龄(49.8±10.9)岁。按手术顺序分为模型组(579例)和验证组(219例)。单因素和多因素logistic回归分析模型组MVI的危险因素,并建立预测MVI的列线图模型,校准曲线和一致性指数(C-index)分析该模型预测的准确度。受试者工作特征(ROC)曲线分析列线图模型预测MVI的价值。结果术后病理结果显示,模型组579例HCC患者中,278例存在MVI,无MVI 301例;验证组219例中119例MVI,无MVI 100例。多因素分析,总胆红素>15 μmol/L(OR=1.519,95%CI:1.041~2.217)、碱性磷酸酶>60 U/L(OR=1.681,95%CI:1.059~2.670)、甲胎蛋白>200 ng/L(OR=2.192,95%CI:1.531~3.134)、肿瘤最大直径(OR=1.120,95%CI:1.057~1.187)是MVI的独立危险因素,据此建立预测列线图模型。模型组的C-index为0.680,验证组的C-index为0.773。模型组和验证组的标准曲线与校准预测曲线贴合良好,该模型得出的MVI的预测值与观察值符合度好。ROC曲线分析列线图模型预测MVI的效能偏低。结论成功建立了术前预测HCC患者MVI的列线图模型,该模型对HCC患者临床治疗决策有一定的指导作用。

Objective To establish a preoperative nomogram model in predicting microvascular invasion (MVI) and to test its predictive effectiveness in hepatocellular carcinoma (HCC). Methods This retrospective study was conducted on 798 patients with HCC, including 690 males and 108 females, aged (49.8±10.9) years old who underwent curative hepatectomy in the Guangxi Medical University Affiliated Tumor Hospital between January 2014 and December 2017 were retrospectively analyzed. The patients were divided into the model group (n=579) and the validation group (n=219) according to the periods of the operation time. Independent risk factors of MVI were identified by univariate and multivariate logistic regression analysis in the model group, and a nomogram model was established according to the independent risk factors. The accuracy of the nomogram model in predicting MVI was detected in the two groups by the computer consistency coefficient (C-index) and calibration graph method. The predictive value was evaluated by receiver operating characteristic curve. Results Histopathological diagnosis revealed 278 patients with MVI and no MVI in the 301 patients of HCC out of the 579 patients in the model group. In the validation group, there were 119 patients with MVI and 100 patients with no MVI out of the 219 patients. Total bilirubin >15 μmol/L(OR=1.519, 95%CI: 1.041~2.217), alkaline phosphatase >60 U/L(OR=1.681, 95%CI: 1.059~2.670), alpha-fetoprotein >200 ng/L (OR=2.192, 95%CI: 1.531~3.134) and tumor maximum diameter (OR=1.120, 95%CI: 1.057~1.187) were the independent risk factors of MVI on multivariate analysis. After establishment of the nomogram model using the independent risk factors, the C-indexes were 0.680 and 0.773 respectively in the model group and the validation group. In the calibration graph, the standard curve properly fitted with the predicting calibration curve. The predicted value of MVI obtained was in good agreement with the observed value. The ROC curve analysis nomogram model predicted the low performance of MVI. Conclusion The nomogram model in predicting MVI in patients with HCC was successfully established. The model offered certain guiding significance in the clinical treatment of HCC.