Framework synthesis of 82 systematic reviews suggests narrative syntheses and meta-analyses use different approaches to argumentation

Article type
Year
Authors
Melendez-Torres G1, O'Mara-Eves A2, Thomas J2, Brunton G2, Caird J2, Petticrew M3
1University of Warwick, United Kingdom
2UCL Institute of Education, United Kingdom
3London School of Hygiene and Tropical Medicine, United Kingdom
Abstract
Background and objective: Opinion is divided about the role of narrative synthesis in systematic reviews. On one hand, narrative synthesis and meta-analysis play complementary roles, and depending on user needs, one method may be preferable to the other. Though both methods ostensibly aim to present a picture of intervention effectiveness, it remains unclear and unexamined if narrative syntheses answer the same questions in the same ways as meta-analyses do. That is to say, it may not be meaningful to compare one against the other on the same terms. Using Toulmin’s argumentation theory, we analysed the texts of systematic reviews to explore differences in the modes of reasoning embedded in reports of narrative synthesis as compared to reports of meta-analysis.

Methods: We used a sample of 106 systematic reviews on workplace health promotion interventions published in English after 1995 that were collected as part of an overview of reviews, and used framework synthesis and grounded theory methods to analyse 82 of the reviews that specifically addressed intervention effectiveness.

Results: Two core categories, or ‘modes of reasoning’, emerged to frame the contrast between narrative synthesis and meta-analysis: practical-configurational reasoning in narrative synthesis (‘What is going on here? What picture emerges?’) and inferential-predictive reasoning in meta-analysis (‘Does it work, and how well? Will it work again?’). Modes of reasoning examined quality and consistency of the included evidence differently. Meta-analyses clearly distinguished between claims and warrants, or arguments bridging data and claims, whereas narrative syntheses often presented joint warrant-claims.

Discussion: Systematic reviewers are likely to be addressing research questions in different ways when using these different approaches to synthesis. These findings provide an alternative perspective on the role of narrative synthesis as ‘second-best’ to meta-analysis. They complement existing guidance on narrative synthesis by highlighting modes of reasoning used, and suggest how meta-analysis deploys narrative ‘tools’ in ways that are not explicitly stated in public.