New cochrane risk of bias tool for cluster randomised tools

Background - Bias has long been recognised as an issue that can seriously compromise the validity of study trial results. Assessing bias from trial reports is often not easy. The Cochrane risk of bias tool is well established tool for facilitating this. This tool has recently been updated and as part of the new tool there is a specific adaptation for cluster randomised trials which takes account of the design of these trials and specific issues for assessing bias in these trials.

Aim - The aim of the tool is to support reviewers of trial reports to assess bias in cluster randomised trials. In this talk we describe the aspects of bias which are particular to cluster randomised trials or operate differently in these trials from the way they operate in individually randomised trials.

Methods - A multi-disciplinary group of researchers including statisticians, other triallists, those leading the development of the new Cochrane risk of bias tool and experts in cluster randomised trials met over a period of a year to discuss the five different bias domains (bias arising from the randomization process, bias due to deviations from intended interventions, bias due to missing outcome data, bias in measurement of the outcome, bias in selection of the reported result) that are part of the new Cochrane risk of bias tool and how they relate to bias in cluster randomised trials.

Results - Given the extent of the differences between assessing risk of bias in individually randomised trials and in cluster randomised trials, the group developed an adapted Cochrane risk of bias tool for cluster randomised trials. Differences occur in relation to assessing allocation concealment; appropriate assessment of bias in relation to blinding of participants and assessors; and ensuring missing clusters are considered in addition to missing values from participants. We also added an additional domain (bias arising from the timing of identification and recruitment of individual participants in relation to timing of randomization) to cover the bias that may occur when individual participants in a cluster randomised trial are recruited after randomisation.

Conclusions - Assessing bias in cluster randomised trials is not the same as assessing bias in individually randomised trials. Authors and peer reviewers should be aware of key elements to include in trial reports to provide evidence that their trials are protected from bias. Systematic reviewers should use the Cochrane risk of bias tool adapted for cluster randomised trials to assess these trials for bias