CMStatistics 2018: Start Registration
View Submission - CMStatistics
Title: A new design test based on a bootstrap method for response-adaptive clinical trials Authors:  Alessandro Baldi Antognini - University of Bologna (Italy)
Marco Novelli - Novelli (Italy) [presenting]
Maroussa Zagoraiou - University of Bologna (Italy)
Abstract: The problem of testing hypothesis in sequential clinical trials for treatment comparisons managed via response-adaptive (RA) randomization is addressed. We propose a new bootstrap test which is more efficient and robust with respect to the classical tests proposed in the literature. In particular, through a suitably choice of the target, we introduce a new test statistic based on the treatment allocation proportion and its bootstrap estimate of the variance. We derive the theoretical properties of the suggested procedure in terms of power and ethical gain; moreover, its performance is illustrated through simulations, also compared with those of other tests suggested in the literature, showing a significant improvement from the viewpoint of inferential precision and ethical concerns as well.