SAM-HC: a Bayesian nonparametric construction of hybrid control for randomized clinical trials using external data.

Abstract

In situations where high-quality external data are available or when it is challenging to recruit participants to the control arm of a randomized and controlled clinical trial (eg rare or pediatric diseases), it is desirable to borrow information from external data to augment the control arm. 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 the Shared Atoms Model (SAM) 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. The degree of information borrowing is confined by the sample size and degree of similarity in outcomes. Simulation studies demonstrate the robustness of the new method, and an application to an Atopic Dermatitis dataset shows improved treatment effect estimation.