ABSTRACT
One of the fi rst decisions to be taken when designing a CRT relates to the choice and defi nition of the clusters that are to be randomised during the trial. In practice, published trials have used a wide variety of types and sizes of clusters. These have ranged from CRTs of families or households, with only a few individuals in each cluster, to CRTs of large geographic regions in which each cluster may include millions of individuals. While the general principles of CRT design may be similar for these widely varying situations, the practical aspects of implementing such trials are clearly very different.
