Hepatocellular carcinoma (HCC) still occurs in patients with hepatitis C who achieved sustained virologic response (SVR) after direct-acting antiviral therapy. We developed and validated an AI-assisted HCC prediction model using longitudinal data in patients with chronic hepatitis C who had achieved SVR. A total of 1,984 HCV patients who achieved SVR from ten hospitals in South Korea were included in the derivation cohort. External validation cohorts were recruited nationwide at 29 university-affiliated hospitals. Machine learning models were trained with parameters at baseline, one year after treatment, and both time points, respectively, and their performance was assessed via Harrell’s c-index and 5-year AUROC. Age, platelet, AST, ALT, bilirubin, and albumin at baseline and one year after treatment were significant predictors of HCC risk. A random forest model trained with longitudinal inputs presented a superior performance in comparison to machine learning models with parameters at a single time point provided, yielding Harrell’s c-index of 0.886/0.796 and 5-year AUROC of 0.903/0.820 for internal/external validation cohorts. Furthermore, compared with previous HCC risk scores, the new model exhibited markedly enhanced discriminatory capability. The findings suggest that a machine learning based model may serve as a useful tool for predicting HCC after SVR in chronic hepatitis C patients.
Lee et al. (2026) studied this question.