Abstract:
|
In clinical trials, in order to understand the overall treatment effect, data on several possibly correlated endpoints are collected and tested. When multiple hypotheses are tested simultaneously, there is an increased risk of Type I error. There are many methods that control the Type I error in such situations, however, not all of them account for the correlations between the endpoints and thus are less efficient. Our approach is to try and estimate the unknown correlations, and see if that can be used to build more powerful test procedures. We exploit the correlation between progression free survival (PFS) and overall survival (OS) using a joint distribution model assuming exponential time to event distributions. We study the robustness of these estimates when the time to event distributions do not satisfy the exponential distribution. We compare methods like Flexible Fixed Sequence (FFS) and Adaptive Alpha Allocation Approach (4A) that incorporate correlation between endpoints to adjust the significance level for testing the hypotheses against methods that do not take this correlation into consideration.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.