A Clinical Tool to Predict the Microvascular Invasion Risk in Patients with Hepatocellular Carcinoma.

Abstract

Microvascular invasion (MVI) plays an important role in tumor progression. The aim of this study is to establish and validate an effective hematological nomogram for MVI prediction in hepatocellular carcinoma (HCC). A retrospective study was performed in a primary cohort that includes 1306 patients clinicopathologically diagnosed with HCC, and a validation cohort contained 563 continuous patients. Univariate logistic regression was used to assess the association between variables included both clinicopathologic factors and coagulation parameters (prothrombin time, activated partial thromboplastin time, fibrinogen, and thrombin time [TT]) and MVI. Multiple logistic regression was used to construct a prediction nomogram. We tested the accuracy of the nomogram by discrimination and calibration, and then plotted decision curves to assess the benefits of the nomogram-assisted decisions in a clinical context. In the two cohorts, patients without MVI had the longest overall survival (OS), compared the OS with MVI. The multivariate analysis indicated that age, sex, tumor node metastasis (TNM) stage, aspartate aminotransferase, alpha fetoprotein, C-reactive protein, and TT were identified as significant independent predictors of MVI of HCC patients. The Hosmer-Lemeshow test showed good point estimate associated P value between predicted risk and observed risk across the deciles. Moreover, the calibration performance of the nomogram risk scores in each decile of the primary cohort was within 5 percentage points of the mean predicted risk score, and in the validation cohort, the observed risk in 90% decile was within 5 percentage points of the mean predicted risk score. A noninvasive and easy-to-use nomogram was established and may be used to predict preoperative MVI in HCC.