Design of randomized controlled confirmatory trials using historical control data to augment sample size for concurrent controls

This paper deals with the methods to augment concurrent controls (CC) in a randomized controlled trial with available historical data in clinical studies. In their article, Matching with multiple control groups and adjusting for group differences, Stuart and Rubin proposed a matching method where the primary/local control and the secondary/non-local control are both included in the propensity score estimates. The authors discuss a similar approach taking the CC as the primary and the historical control as the secondary, and find that this approach does not save the sample size of the randomized trial compared to the traditional randomized design without supplementation of historical data. A new matching method that saves sample size is proposed, where propensity scores are estimated without the concurrent randomized control patients. A two-stage design is proposed, which allows one to examine the assumption of the new matching method before a commitment of using the matching method in the second stage. Previous clinical trials data is used as an example to illustrate the feasibility of the proposed methods. Simulation studies have been used to investigate operating characteristics of the proposed method.