Liver volume-based prediction model stratifies risks for hepatocellular carcinoma in chronic hepatitis B patients on surveillance

Abstract

BACKGROUND AND AIM: The aim of this study was to determine whether dynamic computed tomography (CT)-measured liver volume predicts the risk of hepatocellular carcinoma (HCC) when the CT scans do not reveal evidence of HCC in chronic hepatitis B (CHB) patients on surveillance. METHODS: This retrospective multicentre cohort study included 1,246 patients who received entecavir and regular HCC surveillance in three tertiary referral centres in South Korea. Liver volumes were measured on portal venous phase CT images. A nomogram was developed based on Cox independent predictors and externally validated. Time-dependent receiver operating characteristic (ROC) analysis was performed for comparison with previous prediction models. RESULTS: Patients who received dynamic CT studies during surveillance had significantly higher risk for HCC compared to patients without CT studies (hazard ratio [HR] = 3.1: p < 0.001). Expected/measured liver volume ratio was an independent predictor of HCC (HR = 4.2: p = 0.002) in addition to age, sex and cirrhosis. The nomogram based on the four predictors discriminated risks for HCC (HR = 4.1 and 6.0 in derivation and validation cohort, respectively, for volume score > 150: p < 0.001). Time-dependent ROC analysis confirmed better performance of the volume score compared to HCC prediction models with conventional predictors (integrated area under curve = 0.758 vs. 0.661-0.712: p < 0.05). CONCLUSIONS: CT-measured liver volume is an independent predictor of future HCC, and nomogram-based liver volume score may stratify the risks of HCC in CHB patients who showed negative CT findings for HCC during surveillance.