[Analysis of results of a trial using the subject as its own control with a criterion of binary response]

The two-period cross-over design is often used in clinical trials in which subjects serve as their own controls. This procedure may produce a more powerful test, but also may produce bias due to carry-over effects. Three tests have been proposed for the analysis of two-period binary response cross-over trials (McNemar, Gart and Prescott). These tests and a simple test for the carry-over effect are presented in this paper. The contrasts used are similar for the three tests, but McNemar's test does not allow an order effect, contrary to the other methods. The Prescott's test is more powerful, but calculations are heavy and need computer assistance. Therefore, McNemar's test can be used as a first approximation.