Narrative Synthesis and SWiM: When Meta-Analysis Isn’t Possible

Narrative synthesis is the set of textual, tabular, and graphical methods used to summarize findings across studies in a systematic review when a statistical pooling of effect estimates — a meta-analysis — is not appropriate or not possible. Reviewers turn to narrative synthesis when studies are too methodologically or clinically heterogeneous to pool, when included studies report incompatible outcome measures, when too few comparable studies exist, or when the review question calls for synthesizing very different study designs side by side. This isn’t a fallback of lower rigor; it is a distinct, reportable method with its own reporting standard: SWiM (Synthesis without Meta-analysis), published in The BMJ in 2020.

This guide covers when narrative synthesis is the right choice instead of meta-analysis, what the SWiM reporting guideline actually requires, the specific methods reviewers use under a narrative-synthesis approach, and how the whole thing sits inside PRISMA 2020 reporting. Last verified 2026-08-16 against the SWiM guideline (Campbell et al., BMJ 2020;368:l6890, doi:10.1136/bmj.l6890).

Meta-Analysis vs. Narrative Synthesis: How to Decide

The decision isn’t a matter of preference. Statistical and clinical heterogeneity across the included studies, and how comparable their outcome data actually are, drive the choice.

Situation	Meta-analysis	Narrative synthesis
Studies report the same outcome on a comparable, poolable metric	Appropriate	Not required
High clinical/methodological heterogeneity across populations, interventions, or designs	Pooling would obscure or mislead	Appropriate
Outcomes measured with incompatible instruments/timepoints that can’t be standardized	Not possible without a synthetic metric	Appropriate
Very few comparable studies (e.g., 1–2 per outcome)	Statistically unstable or not calculable	Appropriate
Mixed study designs (e.g., RCTs alongside cohort studies) addressing one question	Usually pooled separately, if at all	Often the only option for the combined picture
Sufficient homogeneous studies exist, but the review team wants a plain-language summary too	Primary method	Reported alongside, not instead of

See Systematic Review vs. Meta-Analysis for the broader relationship between the review and the pooling step. A systematic review does not require a meta-analysis to be a valid systematic review — PRISMA 2020 and the Cochrane Handbook both treat narrative synthesis as a first-class synthesis method, not a consolation prize when pooling fails.

The SWiM Reporting Guideline

Before 2020, narrative synthesis was the least transparently reported part of most systematic reviews: methods sections often said only that findings were “narratively synthesized,” with no account of how studies were grouped, what synthesis method was actually applied, or how heterogeneity and certainty were handled. SWiM (Synthesis without Meta-analysis), developed by Mhairi Campbell and colleagues and published as a reporting guideline in The BMJ (2020;368:l6890), was built specifically to close that gap, the same way the PRISMA statement standardized reporting for the review as a whole.

SWiM sets out reporting items covering these areas, and reviewers are expected to report each explicitly rather than describe the synthesis only as “narrative”:

• Grouping studies for synthesis — the rationale for how studies were grouped (by outcome, population, intervention type, or another logic) before synthesis.
• Standardized metric — whether and how outcome data were transformed onto a common metric to make studies comparable, even without pooling.
• Synthesis method — which specific method was used (see the table below) and why, described in enough detail to be reproducible.
• Heterogeneity — how variation in results across studies was investigated and reported, since narrative synthesis doesn’t get to skip this just because it isn’t calculating I².
• Certainty of the body of evidence — an assessment of how much confidence the synthesized findings deserve, typically using GRADE or an equivalent framework.
• Reporting results — presenting the actual synthesis output clearly, commonly through summary tables, effect-direction plots, or harvest plots rather than prose alone.
• Limitations — naming the limitations of the synthesis method itself, not just the limitations of the included studies.

PRISMA 2020’s own synthesis-reporting items (item 13, methods and results of synthesis) point reviewers to SWiM specifically for reviews that don’t meta-analyze — the two guidelines are designed to interlock, not compete.

Common Narrative Synthesis Methods

Method	What it does	When reviewers use it
Structured summary by outcome or grouping	Organizes findings under each outcome or pre-specified subgroup, describing direction and consistency of effect across studies	Default approach when outcomes can’t be pooled but are conceptually related
Vote counting based on direction of effect	Counts how many studies found a positive, negative, or null effect, without weighting by effect size or precision	When effect sizes can’t be standardized across studies; explicitly discouraged as a stand-alone method by SWiM unless clearly labeled and its limitations reported
Combining statistical significance (e.g., via p-value combination)	Summarizes whether the body of evidence, taken together, reaches statistical significance	Rarely recommended on its own; SWiM asks reviewers to justify it explicitly given known limitations
Summarizing/tabulating effect estimates	Presents each study’s effect estimate and confidence interval in a structured table without a pooled estimate	When individual effect estimates are informative but not statistically poolable
Harvest plots / effect direction plots	Visual grid showing the direction and sometimes the strength of effect for each study across outcomes or subgroups	Large numbers of heterogeneous studies where a table alone is hard to scan

The 2006 ESRC Methods Programme guidance by Popay and colleagues (“Guidance on the conduct of narrative synthesis in systematic reviews”) remains the widely cited methodological foundation for several of these techniques; SWiM builds on that guidance by turning it into checklist items a reviewer can be held to and a reader can audit.

Worked Example (Illustrative)

The following is an illustrative composite, not a real published review, constructed only to show how the SWiM items map onto an actual write-up.

Suppose a review team is synthesizing 14 studies on a workplace intervention, where five report a continuous productivity score, six report a binary “improved/not improved” outcome, and three report only qualitative descriptions of change. Meta-analysis of the full set isn’t possible — the outcome measures aren’t compatible. Under SWiM, the team would report: (1) that studies were grouped into three outcome-type strata before synthesis; (2) that binary outcomes were converted to a standardized “improved” rate for comparability where possible; (3) that the synthesis method was a structured summary by outcome group, supplemented with an effect-direction plot for the continuous and binary strata; (4) how consistency of direction was examined as a proxy for heterogeneity in the absence of a pooled statistic; (5) a GRADE-based certainty rating per outcome group; (6) results presented as one summary table per stratum plus the effect-direction plot; and (7) an explicit limitation noting that the qualitative-only studies could not be integrated into the effect-direction analysis and were reported separately.

Common Pitfalls

• Describing the method only as “narrative synthesis” with no further detail. This is the exact gap SWiM was built to close; reviewers should name the specific method (structured summary, vote counting, harvest plot, etc.), not just the umbrella term.
• Using vote counting as if it were a quantitative result. Vote counting by direction of effect ignores sample size and effect magnitude; SWiM asks reviewers to flag this limitation explicitly if the method is used at all.
• Skipping heterogeneity and certainty reporting because there’s no I² to calculate. Both are still required under SWiM — they’re just assessed narratively rather than statistically.
• Treating narrative synthesis as a downgrade rather than a distinct method. A well-reported narrative synthesis under SWiM is a legitimate, citable, PRISMA-compatible synthesis approach, not an admission that the review fell short.

Frequently Asked Questions

Is narrative synthesis less rigorous than meta-analysis?

No — it’s a different method suited to different evidence conditions, not a lesser version of meta-analysis. A poorly reported narrative synthesis is less trustworthy than a well-reported one, exactly as with meta-analysis; SWiM exists to make narrative synthesis auditable to the same standard.

Does PRISMA 2020 require a specific synthesis method?

No. PRISMA 2020’s synthesis items (item 13) require reviewers to report their methods and results in enough detail to be reproducible, whichever synthesis approach was used, and point to SWiM specifically for reviews that don’t meta-analyze.

What’s the difference between narrative synthesis and a narrative review?

A narrative synthesis is a synthesis method used inside a systematic review, following a pre-specified, reproducible search and eligibility process. A narrative review is a different, broader review type that typically lacks a systematic search protocol altogether — the word “narrative” describes the write-up in both cases, but the underlying rigor and reproducibility of the search process differ.

How many studies do I need before I can meta-analyze instead?

There’s no fixed threshold in PRISMA or SWiM. The decision depends on whether the studies report a comparable outcome metric and whether pooling would produce a meaningful, non-misleading estimate — not on hitting a specific study count.

Can a single review use both meta-analysis and narrative synthesis?

Yes, and this is common: outcomes with enough comparable studies are meta-analyzed, while outcomes that can’t be pooled are narratively synthesized under SWiM within the same review.