August 2010

Carryover Bias in Crossover Trials

Author Affiliations

Author Affiliation: Department of Epidemiology, University of Washington, Seattle.


Copyright 2010 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2010

Arch Pediatr Adolesc Med. 2010;164(8):703-705. doi:10.1001/archpediatrics.2010.126

An interesting randomized crossover trial appears in this issue of the Archives.1 Crossover trials are common; of 526 randomized trials published in December 2000, 127 (24%) were crossover studies.2 The crossover trial design can be an appealing choice for outcomes that can be measured repeatedly over time, such as pulmonary function in asthma patients, seizure frequency, or systolic blood pressure. Initially, a randomized crossover trial mimics a parallel-group trial by randomly assigning subjects to treatment or placebo and measuring their outcomes after some interval. After stopping treatment and placebo for a washout period, the crossover trial resumes by giving placebo to those previously under treatment and treatment to those previously taking placebo. The trial continues until the outcome is measured at the end of a second period. This is a 2-period, 2-treatment crossover trial, the simplest version of the crossover design. Crossover trials can be expanded to more than 2 treatment periods and more than 2 treatments.

First Page Preview View Large
First page PDF preview
First page PDF preview