We
determined that a sample size of 11,664 participants
would allow the observation of 87 patients
with the composite end point, which would give
the study 90% power to show the noninferiority
of fluticasone–salmeterol to fluticasone alone,
with the use of the log-rank test, at a one-sided
alpha level of 0.025, and to reject the null hypothesis
that the risk associated with fluticasone–
salmeterol, as compared with fluticasone alone,
would be greater than the noninferiority margin