Keywords: Clinical trial design, external data, historical data, propensity score method, Bayesian design
The use of external data in clinical trials provides a great benefit for efficiency but also poses practical challenges regarding the similarity between external and current study patient population as well as factors impacting treatment responses. In this presentation, both conventional and Bayesian designs that enable borrowing of historical data will be discussed in a randomized controlled trial setting. In the conventional design, we will propose a propensity score approach that incorporate an assessment of response consistency between natural and concurrent controls. In the Bayesian design, a conditional power prior approach will be used only if the difference in sample means between the concurrent control and the historical control falls within a pre-specified range. Details of the operational characteristics of the proposed designs will be presented.