目的：基于放射学特征和临床实验室数据构建肝细胞癌(hepatocellular carcinoma，HCC)患者发生微血管侵犯(mi- crovascular invasion，MVI)的预测模型，探讨HCC患者术前甲胎蛋白(alpha-fetoprotein，AFP)和γ -谷氨酰转肽酶(gamma-gluta- myl transferase，GGT)的临床应用价值。方法：回顾性地收集南京医科大学第一附属医院479例HCC手术切除患者的临床数据，使用Logistics回归分析确定与MVI显著相关的变量，建立模型并绘制列线图。使用受试者工作特征曲线下面积(area un- der curve，AUC)、校准曲线和决策曲线分析(decision curve analysis，DCA)来评估模型的预测性能。通过灵敏度、特异度和准确度分析对模型预测性能进行比较。应用Kaplan-Meier曲线来评估患者的生存概率。结果：在多因素分析中，肿瘤大小、肿瘤边界不清楚、肿瘤外生性生长、AFP和GGT是HCC患者发生MVI的独立危险因素，用于构建模型。有AFP和GGT的模型A比没有AFP和GGT的模型B表现出更好的预测性能，AUC分别为0.902(0.872~0.932)和0.876(0.842~0.909)。分析评估模型临床净收益的DCA显示出模型A略好于模型B，都具有良好的临床应用价值。模型在队列中显示出良好的识别能力，预测的概率与实际观测值具有较好的一致性。在生存分析中发现，AFP和GGT升高的患者预后明显差于未升高的患者。结论：建立了HCC 患者发生MVI的预测模型，研究证实了血清AFP和GGT的临床应用价值。
Objective：The current study aims to establish a risk model for microvascular invasion(MVI)in patients with hepatocellular carcinoma(HCC)，which is based on radiological characteristics and clinical laboratory data，and to explore the clinical application of preoperative alpha -fetoprotein(AFP)and gamma -glutamyl transferase(GGT)in HCC patients. Methods：Clinical data from 479 HCC patients with surgically resection at the First Affiliated Hospital of Nanjing Medical University were retrospectively collected to investigate the effects of potentially relevant factors on MVI，and to assess the benefits of serum AFP and GGT on improving model predictive performance. Univariate and multivariate logistic regression analyses were used to select variables for the final prediction model，and the predictive performance of the model was assessed by area under the receiver operating characteristic curve(AUC)，calibration curve，and decision curve analysis(DCA). Predictive performance was compared by sensitivity，specificity and accuracy analyses. Kaplan - Meier curves were applied to assess the probability of patient survival. Results：In a multifactorial analysis，tumor size，borders of the tumor，exophytic growth，AFP and GGT were independent predictors for the development of MVI in HCC patients. Two models were constructed with or without AFP and GGT，respectively. In both cohorts，model A with AFP and GGT showed better predictive performance than model B without AFP and GGT，with AUCs of 0.902（0.872-0.932）and 0.876（0.842-0.909）， respectively. The two models showed good discrimination，and the predicted probabilities were in good agreement with the observed probabilities. Analysis of the DCA，which assesses the net clinical benefit of the models，showed that model A was slightly better than model B，both with good clinical application. In the subgroup analysis，AFP and GGT could well predict the prognosis of HCC patients. Conclusion：A predictive model for the development of MVI in patients with HCC was developed，and the study confirmed the clinical predictive value of serum AFP and GGT.