Synthesis when meta-analysis is not possible (Part I): approaches to grouping, presentation and synthesis

Article type
Authors
McKenzie J1, Brennan S2, Thomson H3, Johnston R4, Ryan R5
1Monash University
2Cochrane Australia
3Cochrane Public Health, University of Glasgow
4Cochrane Musculoskeletal Review Group, Monash University
5Cochrane Consumers and Communication Review Group, La Trobe University
Abstract
Background:
This workshop will be based on new guidance in the Cochrane Handbook for Systematic Reviews of Interventions and is the first of a two-part workshop. There are many circumstances that may preclude the use of meta-analysis of effect estimates. For example, meta-analysis is not possible when data are sparse (e.g. when there is only one study available for a particular comparison and outcome), or when there is incomplete information reported about the intervention effect estimates (e.g. missing standard errors), or inconsistency in the reported effect metrics across studies.

Objectives:
In this workshop we will:

1) consider options for grouping studies to facilitate synthesis; and
2) provide guidance on presentation and synthesis methods that can be used when meta-analysis of effect estimates is not possible (referred to as 'narrative synthesis').

Description:
In this workshop we will use a mix of practical exercises, interactive examples and presentation to:

1) discuss scenarios that may preclude meta-analysis;
2) explore options for structuring reviews to facilitate syntheses (by grouping interventions and outcomes using taxonomies and frameworks);
3) present other synthesis and presentation methods, along with their advantages and disadvantages and guidance on when to use which approach; and
4) provide suggestions for describing the results of the synthesis.

Participants will work through an example, examining approaches to synthesising data on patient satisfaction with care. Planning for circumstances that may preclude meta-analysis can ensure that review authors make the best use of the available data and produce more useful syntheses for decision makers.