|
Abstract:
|
Determination of noninferiority margin in an active control noninferiority trial has been a much-debated problem. The margin must not be larger than the assured effect size of the control to guard against a false positive claim. On the other hand, it cannot be too small, leading to impractical sample sizes. In view of the controversies in margin determination, we propose a unified approach for testing retention of a prespecified fraction of the control effect (d0, 0< d0< 1) using rate difference, odds ratio, and log of odds ratio. It turns out that for each retention rate procedure, there is a fixed margin counterpart. We derive a formula to express noninferiority margin as a function of d0. The sample size formula is provided and verified with simulation. We also provide a general guideline for selecting a proper d0. The methodology is illustrated with an approved drug.
|
