PAM-HC: A Bayesian Nonparametric Construction of Hybrid Control for Randomized Clinical Trials Using External Data

Abstract

It is highly desirable to borrow information from external data to augment a control arm in a randomized clinical trial, especially in settings where the sample size for the control arm is limited. However, a main challenge in borrowing information from external data is to accommodate potential heterogeneous subpopulations across the external and trial data. We apply a Bayesian nonparametric model called Plaid Atoms Model (PAM) to identify overlapping and unique subpopulations across datasets, with which we restrict the information borrowing to the common subpopulations. This forms a hybrid control (HC) that leads to more precise estimation of treatment effects Simulation studies demonstrate the robustness of the new method, and an application to an Atopic Dermatitis dataset shows improved treatment effect estimation.