High accuracy model for HBsAg loss based on longitudinal trajectories of serum qHBsAg throughout long-term antiviral therapy

OBJECTIVE: Hepatitis B surface antigen (HBsAg) loss is the optimal outcome for patients with chronic hepatitis B (CHB) but this rarely occurs with currently approved therapies. We aimed to develop and validate a prognostic model for HBsAg loss on treatment using longitudinal data from a large, prospectively followed, nationwide cohort. DESIGN: CHB patients receiving nucleos(t)ide analogues as antiviral treatment were enrolled from 50 centres in China. Quantitative HBsAg (qHBsAg) testing was prospectively performed biannually per protocol. Longitudinal discriminant analysis algorithm was used to estimate the incidence of HBsAg loss, by integrating clinical data of each patient collected during follow-up. RESULTS: IU/mL at entry were analysed. With a median follow-up of 65.6 (IQR 51.5-84.7) months, the 5-year cumulative incidence of HBsAg loss was 2.4%. A prediction model integrating all qHBsAg values of each patient during follow-up, designated GOLDEN model, was developed and validated. The AUCs of GOLDEN model were 0.981 (95% CI 0.974 to 0.987) and 0.979 (95% CI 0.974 to 0.983) in the training and external validation sets, respectively, and were significantly better than those of a single qHBsAg measurement. GOLDEN model identified 8.5%-10.4% of patients with a high probability of HBsAg loss (5-year cumulative incidence: 17.0%-29.1%) and was able to exclude 89.6%-91.5% of patients whose incidence of HBsAg loss is 0. Moreover, the GOLDEN model consistently showed excellent performance among various subgroups. CONCLUSION: The novel GOLDEN model, based on longitudinal qHBsAg data, accurately predicts HBsAg clearance, provides reliable estimates of functional hepatitis B virus (HBV) cure and may have the potential to stratify different subsets of patients for novel anti-HBV therapies.