Abstract
Pegylated-interferon-α (Peg-IFNα) is a treatment option for chronic hepatitis B virus (HBV) infection. To quantify treatment response variability, we conducted a model-based meta-analysis (MBMA) of hepatitis B surface antigen (HBsAg) loss, defined as a binary outcome based on HBsAg levels falling below the limit of detection, with Peg-IFNα-based regimens at end-of-treatment (EOT) and 24 weeks post-treatment. A systematic review of HBsAg loss in chronic HBV infection was performed, searching Embase, MEDLINE, and Cochrane (January 2000-July 2022). Studies reporting only per-protocol results were excluded; intent-to-treat (ITT) or modified ITT results were prioritized. Models described the proportion achieving HBsAg loss with respect to treatment regimens, exploring baseline clinical and demographic covariates. For the EOT model, 83 study-strata-arms (11,493 participants) were included; for the 24-week model, 58 study-strata-arms (4267 participants) were included. In both models, Peg-IFNα duration and baseline HBsAg significantly predicted HBsAg loss (p < 0.001); baseline hepatitis B e-antigen (HBeAg) was an additional predictor at EOT (p = 0.007). These covariates reduced between-trial variance by 58.1% (EOT) and 77.6% (24-week), highlighting their role in explaining heterogeneity. This MBMA supports clinical trial design by simulating outcomes with Peg-IFNα across diverse populations, optimizing trial parameters, estimating sample sizes, and informing enrichment strategies. Notably, these findings have been applied to calibrate and validate in silico trials, demonstrating utility in advancing computational approaches for HBV drug development. This approach enhances precision in predicting treatment outcomes and sets a precedent for leveraging MBMA in chronic hepatitis B research, paving the way for more effective strategies.