Implementation of ‘Risk of bias’ (RoB) tables and ‘Summary of findings’ (SoF) tables in Cochrane reviews: a pilot study

Article type
Authors
Pregno S, Oxman A
Abstract
Background: ‘Risk of bias’ (RoB) and ‘Summary of findings’ (SoF) tables are new additions to Cochrane reviews introduced with RevMan 5. RoB tables are intended to improve assessments of risk of bias. SoF tables are intended to make key findings of reviews more accessible, including magnitude of effect and quality of evidence for each main outcome. Objectives: To identify challenges and solutions in the implementation of RoB and SoF tables in a sample of reviews. Methods: All five review groups that piloted RevMan 5 agreed to participate and identified three reviews due to be updated or completed. One of us (SP) supported review authors in preparing RoB and SoF tables. She consulted with authors to identify main comparisons for which SoF tables were needed, main outcomes to be included in RoB and SoF tables, and appropriate grouping of outcomes in RoB tables. She used RevMan 5 to create RoB tables and GRADEpro to create SoF tables, which were sent to authors to check, edit and approve. The RoB and SoF tables were then checked by at least one review group editor and a methodologist. Results: Information needed for RoB and SoF tables is often lacking in reviews. Preparation of RoB and SoF tables, if not planned in the protocol phase of a review, requires collecting additional information from included studies. Other challenges include deciding on the most important outcomes when there are large numbers of outcomes or outcomes are measured at multiple time points, deciding how to group outcomes in RoB tables, and summarising the evidence in SoF table for outcomes for which meta-analysis was not undertaken. Reviews of complex interventions, such as methodology reviews and EPOC reviews pose additional challenges. SoF tables may need to be adapted for reviews where meta-analysis was not possible and reviews of complex interventions. Conclusions: Central or editorial group support can help to produce RoB and SoF tables, but this requires interaction with review authors, who are ultimately responsible for these tables. We identified a number of challenges and solutions for implementing RoB and SoF tables across a range of reviews.