A LARGE new analysis has confirmed that the GOLDEN model, a predictive algorithm originally developed for patients on nucleos(t)ide analogue therapy, also performs strongly in forecasting hepatitis B surface antigen (HBsAg) loss or steep decline among individuals receiving interferon-α–based treatment.
Drawing on data from the EXCEL randomised controlled trial and the long-term, real-world Search-B cohort, the study presents the strongest evidence to date that GOLDEN can guide more personalised management strategies for chronic hepatitis B (CHB).
Predictive Accuracy for Clinically Meaningful Declines
In the EXCEL trial, which followed treatment-naive, hepatitis B e antigen–positive patients for a median of 18 months, the model achieved an area under the curve (AUC) of 0.820 for identifying those who would reach qHBsAg levels below 100 IU/mL. Its performance was even more pronounced in the larger Search-B cohort, where long-term follow-up of more than five years enabled assessment of complete HBsAg loss. Here, the model reached an AUC of 0.964, maintaining similarly high accuracy across demographic and clinical subgroups.
Differentiating High- and Low-Probability Responders
When patients were stratified according to GOLDEN model predictions, the distinctions were striking. Individuals in the favourable category demonstrated substantially higher cumulative incidences of both qHBsAg <100 IU/mL and complete HBsAg loss compared with those labelled unfavourable.
They also exhibited significantly lower baseline qHBsAg values and faster antigen decline rates over time. Importantly, multivariable analyses confirmed that a favourable GOLDEN classification and lower qHBsAg at enrolment independently predicted successful antigen clearance.
Implications for Personalised Hepatitis B Therapy
These findings underscore the potential role of GOLDEN as a clinical decision-support tool for interferon-based CHB management, helping clinicians better identify patients most likely to benefit from immunomodulatory therapy. Given the ongoing global effort to achieve functional cure in chronic hepatitis B, reliable predictive models are increasingly crucial for optimising therapeutic pathways and avoiding unnecessary treatment exposure.
By demonstrating consistent and high-level predictive accuracy across both controlled and real-world settings, this study positions the GOLDEN model as a robust and clinically actionable tool for forecasting HBsAg loss in interferon-treated patients. Its integration into routine practice could enhance patient selection and refine personalised treatment strategies, supporting progress toward functional cure targets.
Reference
Liao X et al. External Validation of the GOLDEN Model for Predicting HBsAg Loss in Noncirrhotic Chronic Hepatitis B Patients With Interferon-Alpha-Based Therapy. Clinical Gastroenterology and Hepatology. 2025;DOI:10.1016/j.cgh.2025.06.005.
