Design considerations in crossover trials with a single interim analysis and serial patient entry.

A two-stage sequential design is presented to facilitate a single interim analysis in crossover trials with serial patient entry. The interim analysis is based on a linear statistic that combines data from individuals observed for only one treatment period with data from those observed for both periods (Cook, R. J., 1995, Biometrics 51, 932-945). The final analysis is based on the usual test statistic used in crossover trials. The size of this procedure is controlled by partitioning the experimental type I error rate over the two analyses and deriving the appropriate critical values. We investigate the design implications of adopting this procedure over the usual analysis for crossover trials by examining the necessary sample size inflation factors to maintain power, and indicate the expected savings in terms of the number of responses required. Data from a study designed to compare two antiemetic therapies for previously untreated chemotherapy patients (Osaba, D., et al., 1986, Clinical and Investigative Medicine 9, 225-231) are used to illustrate the procedure.