Abstract:
|
In order to improve the efficiency of making regulatory claims and to increase the likelihood of demonstrating activity of targeted agents, many confirmatory studies are now being designed with multiple patient populations, such as marker-positive subpopulation and full population. A few recently proposed multiplicity adjustment procedures are suitable for these studies to ensure a strong control of family-wise error rate, including Bonferroni-based graphical test (Bretz et.al 2009), weighted parametric test procedure (Bretz et.al 2011), weighted Simes procedure (Bretz et.al 2011), feedback method (Zhao et.al 2010) and parametric chain procedure (Millen et.al 2011). We investigated the statistical properties of these five methods and also evaluated their performances theoretically and empirically through extensive simulations. Based on the results, recommendations are made from statistical and practical perspective on the proper selection of multiply adjustment strategy to achieve power optimization, depending on the prevalence of marker-positive subpopulation, weighted importance of the hypotheses in different populations and prior knowledge on the treatment effect.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.