Analysis of two-treatment, two-period crossover trials in emergency medicine

Abstract

In an AB/BA crossover trial, patients are randomly assigned to receive either treastment A in the first period followed by treatment B in the second period or treatment B in the first period followed by treatment A in the second period. The crossover trial allows for a within-patient comparison between treatments because each patient serves as his or her own control subject, removes the interpatient variability from the comparison between treatments, and can provide unbiased estimates for the differences between treatments. When applied inappropriately, crossover designs have serious problems that might adversely influence and invalidate their results. The primary concern is the residual carryover effect of a treatment in subsequent treatment periods. Rather than depending on a statistical procedure to eliminate the possibility of the presence of carryover effects, it is more important that the crossover design be used only in those situations in which the likelihood of a carryover effect is exceptionally small. Even though the AB/BA crossover trial appeals to the physician researcher, it is surprisingly difficult to take advantage of this design. The primary objective of this article is to introduce readers and trialists to some of the issues surrounding crossover trials. Researchers who use this design should explicitly examine the assumptions about crossover effects and the adequacy of the lead-in washout period and the between-period washout period and clearly indicate that the results of the study are conditional on the acceptance of those conditions.