Matching-Assisted Power Prior for Incorporating Real-World Data in Randomized Clinical Trial Analysis.

Leveraging external data information to supplement randomized clinical trials has been a popular topic in recent years, especially for medical device and drug discovery. In rare diseases, it is very challenging to recruit patients and run a large-scale randomized trial. To take advantage of real-world data from historical trials on the same disease, we can run a small hybrid trial and borrow historical controls to increase the power. But the borrowing needs to be conducted in a statistically principled manner. Bayesian power prior methods and propensity score adjustments have been discussed in the literature. In this paper, we propose a matching-assisted power prior approach to better mitigate observed bias when incorporating external data. A subset of comparable external subjects is selected by groups through template matching, and different weights are assigned to these groups based on their similarity to the current study population. Power priors are then implemented to incorporate the information into Bayesian inference. Unlike conventional power prior methods, which discount all control patients similarly, matching pre-selects good controls, hence improved the quality of external data being borrowed. We compare its performance with the existing propensity score-integrated power prior approach through simulation studies and illustrate the implementation using data from a real acupuncture clinical trial.