Abstract:
|
A bioequivalence (BE) study with crossover design has widely been used in formulation and manufacturing comparison. The sample size is driven by the intra-subject variance, which is usually estimated from previous studies or publication. When little information on variance is available for an unblinded study, a two-stage group sequential design with sample size re-estimation protects from unexpected high variance. Potvin and et al. (2008) gave recommendation on decision trees and operating characteristics with true variances. However, her recommendation still depends on the true variance assumption, which in practice may be challenging to make.
A simulation study was conducted to show that 36 subjects at the first stage can be a good option in practice to cover a true coefficient of variance (CV) up to 25% with 80% power and control 5% type I error rate. This sample size at the first stage ensures a reliable estimate of the variance for purposes of re-estimating the final sample size required, and a good chance of stopping early for success if the CV ?25%, which generally covers intra-subject variances observed in these early stage trials. The operating characteristics of this
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.