Volume 22, No. 1 - 2001 

[Table of Contents]

Book Review

Design and Analysis of Cluster Randomization Trials in Health Research

By Allan Donner and Neil Klar
London (England): Arnold Publishers, 2000;
x + 151 pp; ISBN 0 340 69153 0; (hardcover)

The increasing popularity of the cluster randomization design among health researchers over the past two decades has led to an extensive body of methodology and a growing literature that cuts across several disciplines in the statistical, social and medical sciences. This book is the first to provide a unified and systematic treatment of the topic. It may be used as a reference source for investigators in the planning or analysis stages of a study or as a textbook for a graduate level course in research methodology.

The book includes fairly non-technical chapters summarizing key issues of study design, data analysis and reporting as well as more technical material describing extensions of standard regression models (e.g. generalized estimating equations approach, multilevel models) that are needed to account for the variance inflation due to clustering.

The book also provides intriguing discussions of the historical development of cluster randomized trials and summarizes the unique ethical challenges of cluster randomization. For example, in community randomized trials it is typically not possible to obtain informed consent from all individuals who may be affected by the intervention prior to random assignment. This would be the case, for example, in trials evaluating innovative methods of water treatment for the prevention of infectious diseases or in trials evaluating smoking cessation interventions using mass media.

This is a well-written book, which includes data and worked examples illustrating methods of sample size estimation and data analysis.

A challenge facing the authors is that the methodology for cluster randomization trials is undergoing very rapid development. For instance, since publication of the book, special issues of two leading journals^1,2 have been devoted to cluster randomization. It is to be hoped that a second edition is being considered in which the authors can discuss some of these newer developments. For example, as the number of trials adopting cluster randomization has increased, meta-analyses of trials using various units of allocation are starting to appear in the literature. However, investigators have very little guidance, as yet, on how best to conduct such meta-analyses.

Purchasers of the book are entitled to a 25% discount on ACluster, a computer software compatible with Windows 3.1, 95, 98 and NT that implements many of the sample size and analysis formulas presented in the book. Information concerning the software can be obtained at <http://www.arnoldpublishers.com/support/cluster/>

References

1.  Campbell MJ, Donner A, Elbourne D. Design and analysis of cluster randomized trials. Editorial. Stat Med 2001;20(3):329-30.

2.  Donner A, Klar N. Cluster randomization. Editorial. Stat Methods Med Res 2000;9:79-80.

Yang Mao
Chief
Environment Risk and Case Surveillance
Cancer Division
Centre for Chronic Disease Prevention and Control
Public Health Agency of Canada
Health Canada
Address Locator: 0601C
Ottawa, Ontario K1A 0L2

[ Previous] [Table of Contents] [Next]