A machine learning model to predict liver-related outcomes after the functional cure of chronic hepatitis B

초록

Background and aims: The risk of hepatocellular carcinoma (HCC) and hepatic decompensation persists after hepatitis B surface antigen (HBsAg) seroclearance. This study aimed to develop and validate a machine learning model to predict the risk of liver-related outcomes (LROs) following HBsAg seroclearance. Method: A total of 2, 046 consecutive patients who achieved HBsAg seroclearance between 2000 and 2022 were enrolled from 6 centers in South Korea: the training and validation cohorts consisted of 944 and 1, 102 patients, respectively. A new model (designated as PLAN-C) was developed using variables based on the results of multivariable Cox analysis and a gradient-boosting machine algorithm. The primary outcome was the development of any LRO, including HCC, cirrhosis-related complications, and liver-related death. Results: During a median follow-up of 55.2 (interquartile range = 30.1–92.3) months, 123 LROs were confirmed (1.1%/person-year). The PLAN-C was constructed using 6 variables: age, sex, diabetes, alcohol consumption, cirrhosis, and platelet count. Compared to previous HCC prediction models, the PLAN-C showed significantly superior predictive accuracy in both the training (c-index: 0.85 vs. 0.63–0.70, all P < 0.001) and the validation (0.84 vs. 0.61–0.81; all P < 0.05 except for CU-HCC, P = 0.09) cohorts. The calibration plots demonstrated a close correlation between the predicted and observed risks of LRO (Hosmer-Lemeshow test P > 0.05 in both cohorts). When entire patients were divided into 3 groups according to the risk predicted by PLAN-C, the low-risk group had a significantly lower 5-year incidence of LRO (0.8%), compared to the intermediate-risk (4.0%) and high-risk (22.8%) groups (both P < 0.001). Conclusion: This novel machine learning model consisting of 6 variables provides reliable risk prediction of LRO after HBsAg seroclearance that can be used for personalized surveillance.