Adaptive Tests of Significance Using Permutations of Residuals with R and SAS

CHAPTER 8

MULTICENTER AND CROSS-OVER TRIALS

This chapter concerns the analysis of data from two important study designs that are often used in the development of new drugs. We will demonstrate how adaptive methods can be used in both multicenter trials and cross-over trials to increase the power of the tests.

In a multicenter clinical trial subjects are randomized to treatments within each center, and the subjects at one center may differ from the subjects at another center. Consequently, we will have a treatment effect and a center effect, but we are only interested in the treatment effect. Because subjects are randomized within center we will need to modify our permutation test methods so that the residuals are permuted within each center. We will see that this can easily be accomplished with existing software and that the resulting test is more powerful than the traditional test for non-normal error distributions.

Cross-over trials are also used to compare two treatments, but the approach is to give both treatments to each individual in the study. If we denote one treatment by A and the other treatment by B, a cross-over design gives some of the subjects treatment A first and treatment B second while it gives the other subjects treatment B first and treatment A second. These designs are appealing because this test of treatment equality may be much more powerful than one based on a design that gives some subjects treatment A and other subjects treatment B. Because, in the cross-over ...