A predictive model for carcinogenesis in patients with chronic hepatitis B undergoing entecavir therapy and its validation

We created a model to predict the development of liver carcinogenesis in patients with chronic hepatitis B (CHB) undergoing entecavir (ETV) therapy and to validate the accuracy using an independent dataset. A total of 328 CHB subjects were analyzed. Subjects were randomly assigned into 2 groups: the training group (n = 164) and the validation group (n = 164). Using data from the training group, we built a predictive model for liver carcinogenesis by performing univariate and multivariate analyses using variables associated with liver carcinogenesis. We subsequently assessed the applicability of the constructed model in the validation group. The median (range) follow-up periods in the training and the validation groups were 5.03 years (1.03–9.98) and 4.84 years (1.10–9.97), respectively. The proportion of hepatitis B virus-DNA at 24 weeks <1.9 log IU/mL in the training group was 70.7% (116/164), while that in the validation group was 71.3% (117/164). For the entire cohort (n = 328), the median alpha-fetoprotein (AFP) value at 24 weeks (3.45 ng/mL; range, 0.9–102.7 ng/mL) significantly decreased compared to the baseline values (5.55 ng/mL; range, 0.9–1039.5 ng/mL), while the median alanine aminotransferase (ALT) value at 24 weeks (24 IU/mL; range, 6–251 IU/mL) also significantly decreased compared to baseline values (57 IU/mL; range, 7–1450 IU/mL). During the observation period, hepatocellular carcinoma (HCC) developed in 15 (9.1%) patients in the training group and in 17 (10.4%) patients in the validation group. The 3- and 5-year cumulative HCC incidence rates in the entire cohort were 4.48% and 9.52%, respectively. In the multivariate analysis of the training group, age ≥54 years (P = 0.0273), ALT level at 24 weeks (P = 0.0456), and AFP at 24 weeks (P = 0.0485) were found to be significant predictors linked to HCC. Using these independent predictors, the risk for HCC development was well stratified in the validation group (overall significance, P < 0.0001). Similar results were observed in subgroup analyses of patients with or without cirrhosis and HBe antigen positivity. In conclusion, our predictive model was well verified; hence, it may be a promising model for the prediction of the development of liver carcinogenesis in CHB patients undergoing ETV therapy.