ABSTRACT
ANDREW P. GRIEVE CIBA-GEIGY Pharmaceuticals, Horsham, West Sussex, England
The essential feature of a crossover design in clinical trials is that each patient receives more than one of the treatments in the study. In the simplest design, the two-treatment two-period crossover, with treatments A and B, patients are randomly allocated to one of the treatment sequences A -+ B and B -+ A. Such a design, or similar, more complicated designs with more than two treatments and/or periods, are attractive to clinicians due largely to an intuitive feeling that comparing different treatments on the same patient is likely to be more efficient than comparing the treatments on different patients. This intuitive feeling has essentially two components. First, the patient is his/her own control; this has the effect of increasing the precision of treatment comparisons since these are made within patients rather than between patients. Second, patients can express preferences for one or more treatments. Without doubt the first component is the more important since it has important ethical and economic consequences. Ethical, in that a clinician will wish to minimize .the number of patients who re-
ceive less efficacious treatments; economic, in that the use of fewer patients will reduce costs. In essence the argument in favor of crossover designs is that to obtain a given treatment-comparison precision, a within-patient comparison will require fewer patients than a between-patient comparison; they are therefore more ethical and less costly.
