Abstract:
|
In recent years, both EMEA and FDA have been encouraging to leverage historical data in the design and analysis of clinical trials, and therefore various statistical methods have been proposed. In this presentation, we used time-to-event data from two identically designed Phase III studies for patients with Relapsing Remitting Multiple Sclerosis (RRMS) to assess the feasibility and performance of Bayesian historical control borrowing method with a robust mixed prior. In the analysis, we used the control group of either study to serve as the historical control for the other one. After identifying the baseline characteristics accounting for variation between two control groups, Bayesian historical borrowing model with full borrow provided similar estimate of the treatment effect in either study and achieved the same power as using new control only with the same total sample size. In the simulated design, after checking the similarity between two control groups with covariates adjusted in the interim analysis, we fully borrowed the historical control data and can achieve the same power as using new control only.
|