Cluster randomized trials: Considerations for design and analysis

Scientists often use randomized controlled trials to compare a newly developed treatment to the existing one, or to a placebo. Patients are randomly assigned to a treatment, and they are compared with respect to the outcome of interest. The cluster randomized trial (CRT) is a type of randomized controlled trial in which the treatments are randomized at the group, rather than individual, level. The intracluster correlation (ICC) measures the degree of similarity between individuals within clusters. CRTs can be designed in several ways; it is essential that researchers carefully plan the study, from sample size calculations to ICC calculation to analysis, in order to get valid and meaningful results. In this article we review and discuss the considerations essential to conducting a successful CRT using both frequentist and Bayesian approaches, and we discuss recent trends in CRT analysis, including highlighting new methodology for both binary and continuous data.

Duke Scholars

Author Hrishikesh Chakraborty Biostatistics & Bioinformatics