Using baseline measurements in the two-period crossover clinical trial

Abstract

In a clinical trial designed to compare the efficacy of two treatments, use can be made of the two-period crossover design, in which each patient is randomly assigned to receive both treatments in one of two sequences. A restriction of its utilization involves the issue of carryover effect, otherwise known as sequence effect or treatment by period interaction. Carryover effects can sometimes be eliminated by using baseline measurements taken at the beginning of each period just prior to treatment. We examine this approach of using baseline measurements in this setting and find that there are many circumstances in which, in spite of eliminating carryover, the use of baseline measurements provides a less powerful test of treatment effect and a less precise estimate of the treatment difference.