Patient numbers needed in prophylactic migraine trials.

Based on a multicenter double-blind crossover prophylactic migraine trial (n = 96), the relative effectiveness of the crossover design versus group comparison was evaluated. If a 'time effect' is taken into account in the statistical model, the crossover design was estimated to be 10-11 times more effective than the group comparison. With more traditional statistical methods using paired and unpaired t tests, the crossover design was estimated to be 4 times more powerful than the group comparison. Based on the t test, sample size tables for numbers of patients needed for detecting effect differences from 10 to 50% with powers from 0.6 to 0.9 in a crossover and a group comparison are presented. These calculations are based on only one trial, and it is recommended that similar estimations of sample variability should be done in other previous trials.