Cluster-Randomized Trials

Abstract

In some situations, it is preferable or necessary not to randomize individual patients but groups of patients, for example clinics, care homes, or households. These groups are commonly denoted as “clusters”. Two approaches for the analysis of cluster-randomized trials can be distinguished, and the chosen method has to be adequately incorporated in sample size calculation. In cluster-level analyses, summary measures are calculated for each cluster condensing the data to be included in the analysis to a single observation per cluster. In contrast, individual-level analyses use the individual outcomes of the subjects in the analysis. In this chapter, sample size calculation methods for both analysis approaches are provided. The principle of sample size calculation for cluster-randomized trials and its differences to individually randomized trials are described in detail for normally distributed outcomes. Furthermore, a condensed overview of related methods for other scale levels of the outcome is given.