Methods and Applications of Statistics in Clinical Trials, Volume 1: Concepts, Principles, Trials, and Designs

Chapter 18

Cluster Randomization

Neil Klar and Allan Donner

18.1 Introduction

Randomized trials in which the unit of randomization is a community, worksite, school, or family are becoming increasingly common for the evaluation of lifestyle interventions for the prevention of disease. This form of treatment assignment is referred to as cluster randomization or group randomization. Reasons for adopting cluster randomization are diverse, but they include administrative convenience, a desire to reduce the effect of treatment contamination, and the need to avoid ethical issues that might otherwise arise.

Data from cluster randomization trials are characterized by between-cluster variation, which is equivalent to saying that responses of cluster members tend to be correlated. Dependencies among cluster members typical of such designs must be considered when determining sample size and in the subsequent data analyses. Failure to adjust standard statistical methods for within-cluster dependencies will result in underpowered studies with spuriously elevated type I errors.

These statistical features of cluster randomization were not brought to wide attention in the health research community until the now famous article by Cornfield [1]. However the 1980s saw a dramatic increase in the development of methods for analyzing correlated outcome data in general [2] and methods for the design and analysis of cluster randomized trials in particular [3, 4]. Books summarizing this research have also ...