How to assess the evidence when a statistical synthesis is not possible

Article type
Authors
Santesso N1, Thomson H2, Brennan S3, Schunemann H4, Mustafa R5
1Cochrane Canada, GRADEing Methods Group, McMaster University
2Cochrane Public Health, University of Glasgow
3Cochrane Australia
4GRADEing Methods Group, McMaster University
5University of Kansas Medical Center
Abstract
Background:
In over 20% of Cochrane Reviews, it is not possible to conduct a meta-analysis, and instead the quantitative data are synthesised narratively. After synthesising the evidence, it is also helpful to readers and users of systematic reviews if that evidence is assessed and presented clearly. When there is a meta-analysis, authors use GRADE to assess the certainty and present the information in 'Summary of findings' tables. GRADE can also be used when a statistical synthesis is not possible. The 'Summary of findings' table, and the software GRADEpro used to create the 'Summary of findings' table, can also include outcomes with narrative syntheses.

Objectives:
1) Provide guidance about how to apply the GRADE domains where there is no statistical estimate of effect
2) Provide guidance on the production of 'Summary of findings' tables for outcomes that have been synthesized narratively
3) Give participants an opportunity to practice applying the guidance to examples
4) Obtain feedback from participants about the current guidance and challenges

Description:
We will briefly describe the GRADE domains to assess the certainty of evidence for an outcome and how the domains are used when evidence is synthesised narratively. We will explain risk of bias, imprecision, indirectness, inconsistency, publication bias, and other GRADE domains. Participants will spend most of the workshop applying the concepts in small groups.