Multi-Arm Equivalence Tests in Cluster-Randomized Design

This technical report presents methodologies for computing power and sample size for multiple equivalence tests comparing treatment means to a control mean within a cluster-randomized design. It discusses the challenges associated with cluster-randomized trials, particularly the inefficiency in sample size due to the grouping of subjects. The report outlines the application of Schuirmann’s TOST procedure and the adjustment methods proposed by Campbell and Walters (2014) for sample size estimation. It details the structure of the equivalence tests, including the establishment of upper and lower equivalence limits for the differences between treatment and control means. Additionally, it explains the Bonferroni adjustment for type I error rates when conducting multiple comparisons. The report also provides technical details on power calculations and the design effect, emphasizing the importance of appropriate sample size determination in the context of cluster-randomized trials. An example scenario is included to illustrate the application of these methodologies in practice.