It is becoming increasingly common for health researchers to consider randomizing intact social units (e.g. families, schools, communities) rather than independent individuals in experimental trials. Reasons are diverse, but include administrative convenience, a desire to reduce the effect of treatment contamination and the need to avoid ethical issues, which might otherwise arise. Dependencies among cluster members typical of such designs must be considered when determining sample size and analyzing the resulting data. Failure to adjust standard statistical methods for within‐cluster dependencies may result in severely underpowered studies and in spuriously elevated Type I error rates. The purpose of this article is to review the key issues in the design and analysis of cluster randomization trials. These ideas will be illustrated using data from several recently completed studies.