Cluster randomised trials in Cochrane reviews of infectious diseases

Article type
Authors
Donegan S, Faragher B, Garner P
Abstract
Background: In infectious diseases, particularly in developing countries, cluster randomised trials help evaluate public health interventions. In reviews, cluster randomised trials requires more complex analysis than individual randomised trials. Sometimes authors are unaware that a trial is cluster randomised and therefore the trial is incorrectly included in the review as though it is an individual randomised trial. Furthermore, a cluster randomised trial must account for clustering in the analysis and some trial’s analyses do not make this adjustment. Authors have difficulty determining when a trial’s analysis has been cluster adjusted. In either event, the confidence intervals for the estimated treatment effects are overly narrow. Objectives: To locate Cochrane reviews of the Infectious Diseases Group that include cluster randomised trials, source the included trial reports, and verify the trials were correctly classified as individual or cluster randomised (adjusted or not adjusted). We also aim to overview the review’s methods and to investigate how cluster randomised trials are reported in order to produce reporting guidelines. Methods: Reviews of the Cochrane Infectious Diseases Group that included cluster randomised trials were eligible. For eligible reviews, we obtained the trial reports to verify that trials were correctly classified as individual or cluster randomised (adjusted or not adjusted) and whether information was reported to allow authors to adjust the trial’s results for clustering. Also, we extracted data regarding the cluster trials such as: number of cluster trials, trial’s analysis methods, author’s analysis, methods reported and interpretation. The data extracted were summarised. Results: The results of the verification of trial classification and a summary of the methods and reporting of cluster randomised trials in Cochrane reviews will be presented. Conclusions: Recommendations to improve the reporting of cluster randomised trials in future reviews will be proposed.