|
Abstract:
|
Unblinded sample size re-estimation (ubSSR) is useful in clinical trial design when the observed interim effect size is still clinical meaningful but smaller than the originally assumed one for sample size calculation. Based on the interim conditional power, the sample size can be re-estimated to ensure adequate power at the final analysis if the interim result is promising. We extend the unblinded sample size re-estimation method to ordinal data, such as a Likert scale, and utilize a weighted final test to maintain the type I error rate. We will investigate impact of ubSSR on the type I error control for secondary endpoints. We will show that the type I error for secondary endpoints is usually inflated and that, however, a sequential closed testing procedure could alleviate such inflation. To avoid back-calculation of the interim effect size, single step sample size increase procedures with a fixed sample size increase is usually considered in practice. We propose a procedure of step-wise sample size increase with a random component. We conduct extensive simulation studies to investigate the operating characteristics of the proposed method and compare with alternative methods.
|
