Sample Size Determination: Power and Minimal Detectable Difference
The first presentation introduces and discusses two different ways to justify the size of a study, one is the usual power calculation, we are all familiar with, and the other approach is based on the minimal detectable difference. The reasons for introducing the concept of the minimal detectable difference is to align statistical significance with clinical relevance. The presentation then discusses two examples. The first example deals with the size of a cardiovascular outcome study where typically a realistic effect is close to a minimal clinically relevant difference. And the second example discusses the sample size calculation of a confirmatory trial in Alzheimer's disease where different issues occur, making again the concept of a minimal detectable difference important.
Non-inferiority assessments: Overview and issues
The second presentation discusses the assessment of non-inferiority in clinical trials. It presents some of the underlying methodology (high level), issues and also provides a number of examples. Major issues will be discussed like the choice of the population, the constancy assumption and the determination of the non-inferiority limit. All these issues point basically to the overarching issue of assay sensitivity in a non-inferiority trial. For the determination of the non-inferiority limit key approaches will be introduced, a basic definition based on clinical relevance and two retained effect methods, the 95-95% method and the synthesis method. Three examples will finally illustrate the importance of all these elements.
