Bayesian Divide-and-Conquer Propensity Score Based Approaches for Leveraging Real World Data in Single Arm Clinical Trials

There has been a substantial rise in the usage of real-world data (RWD) to supplement trial data in the medical and statistical literature. Propensity score methods such as stratification have been used to balance baseline characteristics and prognostic factors between external patients and current trial patients to improve the estimation of the current trial's parameter of interest. This paper merges propensity score methodology and Bayesian inference to estimate a current trial's parameter of interest as follows: (i) match current patients and external patients by strata using the percentiles of the current patients' propensity scores, (ii) apply a prior within each stratum to leverage RWD to estimate the stratum-specific parameter of interest, and (iii) then use a weighted average scheme to combine the stratum-specific parameters to estimate the overall current trial's parameter of interest. In stage (ii), the three priors used are a double hierarchical prior, an extension of the robust mixture prior, and an extension of the power prior. An extensive simulation study is carried out to evaluate the performance of the proposed approaches.