Objective: To investigate the risk factors of microvascular invasion (MVI) in China liver cancer staging system stage Ⅰa (CNLC Ⅰa) hepatocellular carcinoma (HCC), and develop a nomogram for predicting MVI based on clinical and radiographic data. Methods: This retrospective study focused on CNLC Ⅰa HCC patients who underwent radical resection at the Cancer Hospital, Chinese Academy of Medical Sciences from January 2016 to December 2020. Patients' clinical characteristics and laboratory test results and pre-surgery gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid (Gd-EOB-DTPA)-enhanced magnetic resonance imaging results were collected. The clinical and radiographic risk factors for MVI were identified by univariate and multivariate logistic regression analyses and used for the construction of the predictive nomogram. The nomogram model was then internally validated, and its performance was assessed. Results: A total of 104 patients were divided into the MVI-positive group (n=28) and the MVI-negative group (n=76). Multivariate logistic regression analysis at the P<0.1 level identified serum alpha-ferroprotein >7 ng/ml, total bilirubin >21 μmol/L, prothrombin time >12.5 s, non-smooth margin, and incomplete or absent capsule as risk factors of MVI, based on which a nomogram model was built. The model achieved an area under the curve (AUC) value of 0.867 (95% confidence interval, 0.791-0.944) in the internal validation. The sensitivity and specificity of the nomogram model were 0.786 and 0.829, respectively, with the prediction curve nearly overlapping the ideal curve. Based on the Hosmer-Lemeshow test, the predicted and real results were not significantly different (P=0.956). Conclusions: The probability of MVI of CNLC Ⅰa HCC can be objectively predicted by the monogram model that quantifies the clinical and radiographic risk factors. The model can also help clinicians select individualized surgical plans to improve the long-term prognosis of patients.
目的： 探讨中国肝癌临床分期(CNLC)为Ⅰa期的肝细胞癌(HCC)微血管侵犯的危险因素，构建预测微血管侵犯的列线图模型。 方法： 选取2016—2020年在中国医学科学院肿瘤医院行根治性手术切除的Ⅰa期HCC患者104例，收集患者的临床特征、实验室检查指标和术前肝胆特异性对比剂增强磁共振影像学特征。采用单因素和多因素logistic回归分析确定微血管侵犯的危险因素，构建列线图预测模型，并对列线图预测模型进行内部验证。 结果： 根据组织病理学诊断结果将患者分为微血管侵犯阳性组(28例)和微血管侵犯阴性组(76例)。单因素和多因素logistic回归分析显示，甲胎蛋白(OR＝3.3，95% CI：0.9~12.6)、总胆红素(OR＝6.8，95% CI：1.3~36.8)、凝血酶原时间(OR＝0.2，95% CI：0.0~0.7)、肝胆期肿瘤边缘(OR＝4.6，95% CI：1.1~18.8)、肿瘤包膜(OR＝0.3，95% CI：0.1~1.1)是微血管侵犯的影响因素(均P<0.1)，在此基础上构建HCC微血管侵犯列线图预测模型。内部验证结果显示，该模型受试者工作特性曲线的曲线下面积为0.867(95% CI：0.791~0.944)，灵敏度为0.786，特异度为0.829。校准曲线显示，列线图模型的预测曲线接近理想曲线，一致性较好。Hosmer－Lemeshow检验显示，列线图模型的预测结果与实际发生情况差异无统计学意义(P＝0.956)。 结论： 结合临床和影像学特征的列线图预测模型可用于预测CNLCⅠa期HCC微血管侵犯状态，为患者的个体化治疗策略提供指导。.
Keywords: Carcinoma, Hepatocellular; Magnetic resonance imaging; Microvascular invasion; Nomogram.