A Bayesian Hierarchical Meta-Analysis Framework for Estimating Mechanistic Model Parameters from Published Summary Statistics

Recent advances in Bayesian adaptive clinical trial designs increasingly integrate pharmacokinetic (PK) and pharmacodynamic (PD) data to support early-phase dose finding and optimization. These designs usually use noninformative priors for certain mechanistic PK/PD model parameters, largely due to the limited availability of parameter-level information in conventional clinical trial reporting which typically provides only aggregate PK summaries. Motivated by ponsegromab, a subcutaneously (SC) administered monoclonal antibody (mAb) currently investigated in phase I and II studies for cancer cachexia, we develop a three-level Bayesian hierarchical meta-analysis model to infer essential PK model parameters for SC-administered mAbs. The model takes as input conventional summary PK statistics, such as area under the concentration-time curve (AUC) and maximum serum concentration ( C max ). The PK parameters of interest include absorption rate constant ( K a ), clearance (CL) and bioavailability (F), among others. Under plausible assumptions about the PK model mechanism, the proposed statistical framework enables reconstruction of dynamic PK profiles including absorption, elimination, half-life, and duration of the drug action. The estimated PK parameters can in turn be used to enable estimation of drug concentration-time profiles under various dose/schedule settings via conventional pharmacometrics simulations, as well as incorporation in the form of prior distributions to potentially improve existing or novel PK/PD-based early phase dose-finding/optimization trial designs.