The Comparison of Success Rates in Cross‐Over Trials in the Presence of an Order Effect

S ummary The most commonly used methods for testing the significance of differences between treatments in a cross‐over clinical trial with a binary response utilize only those patients responding differently to the two treatments. A method of analysis is proposed in which the data from all patients are utilized. Simulations suggest that in the usual experimental setting the proposed test is more powerful than those currently in use.