BMJ  2003;327 (4 October), doi:10.1136/bmj.327.7418.0-c

Cluster trials are vulnerable to bias

Cluster trials, in which groups of patients rather than individuals are randomised, are vulnerable to bias. Puffer and colleagues (p 785) reviewed a sample of cluster trials published in the BMJ, Lancet, and New England Journal of Medicine over the past five years and found that some trials have methodological weaknesses that could make their results unsound. For example, of the 36 trials studied only 15 described having used a secure method of allocating clusters. The authors state that cluster trials are an important method of evaluating some healthcare interventions and they can be difficult to do; therefore, sources of bias in cluster trials must be recognised and prevented at the planning stage.

Related Article

Evidence for risk of bias in cluster randomised trials: review of recent trials published in three general medical journals

Suezann Puffer, David Torgerson, and Judith Watson
BMJ 2003 327: 785-789. [Abstract] [Full Text] [PDF]