Abstract:
|
In recent years, several methods about borrowing historical control data have been developed to improve the estimate on the current control in randomized clinical trials. In general, when the historical data and current control are consistent, the borrowing of historical data will result in increased power and reduced type I error. However, when the two types of data are inconsistent, it may result in biased estimate, reduced power and/or inflation of type I error. The inconsistency between the current and the historical control data may be partially due to a systematic variation in the known baseline prognostic factors. Therefore, we propose Bayesian hierarchical modeling method that incorporates patient-level baseline covariates to mitigate the inconsistency and to the enhance exchangeability assumption between the current and historical data. In this presentation, we will show the simulation results using covariate-adjusted dynamic borrowing method. We will also describe the application of this method to clinical trial design on amyotrophic lateral sclerosis (ALS), which is a rare devastating neurodegenerative disease with no cure.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.