Cluster randomized trials in general (family) practice research

Cluster randomized trials are increasingly common in general practice (family medicine). This paper will consider the design and analysis of such trials and emphasize the similarities and differences with trials in education, heath promotion and public health. Issues discussed are the estimation and range of values of the intra-cluster correlation coefficient found in general practice, and the associated sample size problems. There are problems with widely varying numbers of subjects per cluster, which leads to planning and analysis difficulties. Ethical issues in these trials, and considerations such as the principle of intention to treat are also considered. An example of the type of analysis available for a continuous outcome variable is given, and the available software is summarized briefly.