Adaptive blinded sample size adjustment for comparing two normal means-a mostly Bayesian approach

What is it about?

Sample sizes in clinical trials are usually chosen to provide a pre-specified level of statistical power. When the data arise from 2 normal populations (one for each treatment group), the power depends on an assumption about the mean treatment difference. Often, this assumption is uncertain, so that the power afforded by the sample size is also uncertain. In such cases, a blinded sample size adjustment (BSSA) can be conducted after some data have been collected. Standard BSSA methods provide evidence about the within-treatment variance; however, they do not provide evidence about the treatment difference. This paper proposes a BSSA method that provides evidence about both the variance and the treatment difference. The paper also demonstrates the superiority of the method in selecting a sample size that will provide the desired power.

Featured Image

Why is it important?

The new blinded sample size adjustment method, unlike previously developed methods, informs researchers about both the mean treatment difference and the within-treatment variance, improving researchers' ability to select the optimal sample size.