Guidance on the conduct of narrative synthesis in systematic reviews: a comparison of guidance-led narrative synthesis versus meta-analysis

Article type
Authors
Rodgers M, Popay J, Roberts H, Sowden A, Petticrew M, Arai L, Britten N
Abstract
Background: A commonly used method to synthesise research in the context of systematic reviews is that of narrative synthesis (NS), the defining characteristic of which is the adoption of a textual approach to the process of synthesis. This may occur either alongside or instead of statistical meta-analysis (e.g. where clinical or statistical heterogeneity precludes pooling). However, there is currently no consensus on the constituent elements of NS and the conditions for establishing trustworthiness are frequently absent.

Objectives: The reported work involved the development of evidence-based guidance on the conduct of narrative approaches and the practical application of the guidance to a re-synthesis of studies included in an existing Cochrane review.

Methods: We undertook a review of methods texts to develop the guidance. We then applied the guidance to NS of 11 studies
previously synthesised in a Cochrane meta-analysis (investigating the effects of interventions for promoting smoke alarm ownership and function), and compared the findings of the two approaches. The NS procedure was blinded to the findings of the original Cochrane review.

Results: Eleven of the 18 'tools and techniques' described in the guidance were considered relevant and were applied. Both the NS and the Cochrane syntheses concluded that educational interventions resulted in only modest increases in smoke alarm ownership and function and that there were insufficient injury/burn prevention data. The strength of conclusions for subgroup analyses differed between approaches.

Conclusions: The guidance provides a useful framework for the conduct of NS, particularly for increasing transparency. The availability of point estimates in subgroup and sensitivity analyses may allow the meta-analyst greater confidence to draw 'firm' conclusions about moderators of effects, whereas the detailed scrutiny of studies allowed by NS may provide insights into additional implications for research.