Enhancing Randomized Controlled Trials: A Bayesian Divide-and-Conquer Approach for Borrowing External Control Data

The use of real-world data (RWD) has generated increasing interest as a means to complement randomized controlled trials (RCT) for ethical or feasibility considerations. In this paper, by dividing the estimation of the treatment effects into strata, we propose a Bayesian divide-and-conquer approach to improve the estimation of an overall treatment effect for RCTs in the presence of external control data, also commonly known as hybrid control trials. Specifically, we extend the borrowing-by-parts power prior with novel plausibility indexes to better control borrowing, especially in the presence of temporal effects. We also propose a new metric to quantify the amount of data being borrowed. A simulation study demonstrates that the proposed method is robust to violations of the model’s assumptions and to the choice of weights used to combine the stratum-specific effects. Illustrative data used in preparation of this article were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database.