Methods and Applications of Statistics in Clinical Trials, Volume 1: Concepts, Principles, Trials, and Designs

Chapter 26

Crossover Design

Michael G. Kenward

26.1 Introduction

In the familiar parallel-group design, each subject is randomly allocated to one of the experimental interventions or treatments. In contrast, in a crossover design, each subject is randomized to a sequence of treatments, which is a special case of a repeated measures design. Despite this difference, the goal of a crossover study, in nearly all settings remains the same as that of a parallel-group study: to compare the effects of the single treatments not the effects of the sequences to which the subjects are randomized. Thus, in a crossover trial, subjects are not randomized to the interventions under comparison, and it has important implications for the design, analysis, and interpretation. The main advantage of a crossover design over the parallel group is the opportunity it provides to compare the effects of treatments within subjects. That is, any component of an individual’s response that is consistent across time is removed from the treatment comparison. The degree of such consistency is measured by the within-subject (or intra-class) correlation. As a consequence, the precision of the treatment effect from a crossover study will be larger than that from a parallel group study with the same numbers of subjects, and it will have the greatest advantage when the within-subject correlation is high. In this respect, the crossover study has much in common with a blocked or matched study, in which the subject is the ...