Post date: Nov 8, 2014 11:18:19 PM
12:15pm - 1:30pm SLAM open house
(food and drinks provided)
1:30pm-2:45pm Short Course "Clinical Trial Adaptations before Breaking the Blind - Part I & Part II"
Michael Proschan, National Institute of Allergy and Infectious Diseases, NIH
The planning of clinical trials requires knowledge of parameters used in sample size/power calculations. We do our best to estimate these parameter from prior data, but it would be nice to be able to use data from the most relevant source, the trial itself. Doing this in a way that preserves the blind is important in protecting the integrity of a trial. After the trial begins but before breaking the blind, we might notice other issues such as difficulty in measuring the primary endpoint. Is it permissible to change the primary endpoint under such circumstances? This talk focuses on when blinded adaptations are and are not valid. We show the importance of permutation tests in adaptive settings, and show that the validity of adaptive t-tests can be surmised from the validity of adaptive permutation tests. (slides)(slides for Part II)