All Times EDT
Keywords: Bayesian augmented control, asymmetrical randomization, entropy balancing, power prior
In pediatric or orphan diseases, there are many instances where it is unfeasible to conduct randomized and controlled clinical trials due to difficulty of enrollment. One solution to reduce the sample size or expedite the trial timeline is to complement the current trial with real-world data. Several propensity score-based methods have been developed to create defined groups of patients that are controlled for confounding based on a set of measured covariates at baseline. However, balance checking on the measured covariates and tweaks to the propensity score models are usually inevitable to achieve the joint balance across all covariates. To mitigate this iterative procedure, we utilize the entropy balancing weighting technique which focuses on balancing the covariates of subjects between the experimental and control groups directly and augments the current trial with the external control data via a power prior. The finite-sample properties of the proposed method are assessed via simulations, comparing with other methods such as covariate-balancing propensity score (CBPS) and propensity score matching (PSM) and weighting (PSW).