A Cluster Randomization Trial is one in which intact social units, or groups of individuals, are randomized to different intervention groups. The increasing popularity of this design in epidemiology and public health research has led to an extensive body of methodology on the subject. A short general introduction about cluster randomization designs will be given discussing the statistical issues that must be addressed when clusters instead of individuals are the unit of observation. The clustering of subjects within these units implies that standard approaches to sample size estimation and statistical analyses cannot be directly applied. As an application example the design and analysis of an epidemiological field trial will be presented. The Brazilian BCG-REVAC Trial against tuberculosis was a large cluster randomization trial which was performed to estimate the effect on the incidence rate of tuberculosis of a dose of BCG (Bacillus Calmette Guerin) given to school children in a population with high coverage of neonatal BCG vaccination. The study started in 1996 and was a pair-matched, stratified cluster randomization trial without placebo; schools were the unit of randomization. The recruited study population consisted of ca. 350.000 children aged 7-14 years enrolled in 763 state schools from the cities of Salvador (Bahia) and Manaus (Amazon), Brazil.

Wir freuen uns auf Ihr Kommen.

Andrea Berghold, Ernst Stadlober