Meta-synthesis is the umbrella term for a family of methods that systematically combine the findings of qualitative studies into a new, integrated interpretation. Where a quantitative review pools numbers, a meta-synthesis works with themes, concepts, and theories drawn from interview, focus group, and observational research, and its ambition is not to average findings but to reinterpret them together, producing understanding that goes beyond any single study. The family includes meta-ethnography, thematic synthesis, meta-aggregation, framework synthesis, and grounded formal theory, and its outputs increasingly sit alongside effect estimates in guidelines to explain how and why interventions work, and how people experience them.
Meta-synthesis versus meta-analysis: interpretation versus pooling
The contrast with quantitative synthesis runs deeper than the type of data. A meta-analysis seeks aggregation: it treats each study as an estimate of a common quantity and combines those estimates with statistical pooling of study results into a single weighted answer, simplifying by design. A meta-synthesis seeks interpretation: it deliberately exploits the variation and richness across studies, asking what the findings mean when read against one another. Divergent findings are not noise to be modelled, as statistical heterogeneity is in a quantitative pool, but analytic material in their own right. This also changes what counts as rigour. A meta-analysis defends its weighting and model choices; a meta-synthesis defends an audit trail from the words of the original papers to the concepts of the new interpretation, so a reader can trace every claim back to its source.
The main families of qualitative synthesis
The methods in the family sit on a continuum from aggregative to interpretive. Aggregative approaches assemble and summarise findings with minimal transformation, prizing dependability and directness for practice. Interpretive approaches treat the included studies as data for fresh theory-building, prizing conceptual innovation. Where a team stands on that continuum shapes every downstream choice, from how many studies are manageable to how searching and appraisal are handled, so the method should be named and justified in the protocol rather than decided quietly during analysis. The five approaches below are the ones review teams meet most often, ordered roughly from most to least interpretive.
Meta-ethnography and its seven phases
Meta-ethnography, developed by Noblit and Hare in 1988, is the oldest and most interpretive member of the family. It proceeds through seven phases: getting started; deciding what is relevant; reading the studies; determining how the studies are related; translating the studies into one another; synthesising the translations; and expressing the synthesis. The distinctive move is reciprocal translation, in which the concepts of one study are systematically expressed in the terms of another to see whether they are about the same thing. Where studies conflict, a refutational synthesis examines the contradiction itself, and a line of argument synthesis builds the translated concepts into a fresh overarching explanation. It yields the richest theory but demands the most analytic skill.
Thematic synthesis
Thematic synthesis, formalised by Thomas and Harden in 2008, is the most widely used and most accessible approach. Reviewers code the findings sections of included papers line by line, organise the codes into descriptive themes that stay close to the original studies, and then generate analytical themes that answer the review question and go beyond what any primary author said. The mechanics resemble primary qualitative analysis, and the craft involved is close to what we describe in coding data in qualitative reviews, applied to published findings rather than raw transcripts.
Meta-aggregation, framework synthesis, and grounded formal theory
Meta-aggregation, associated with JBI, formerly the Joanna Briggs Institute, deliberately mirrors the logic of a quantitative review: findings are extracted with supporting illustrations, grouped into categories, and aggregated into synthesised statements ready to inform practice, with minimal reinterpretation. Framework synthesis starts from an existing conceptual framework and codes study findings against it, allowing the framework to be revised as the data push back; it suits policy questions with a mature theoretical base. Grounded formal theory applies grounded theory procedures across studies to build mid-range theory. Choosing among them is a genuine methodological decision, driven by the question, the state of theory, and how interpretive the team can credibly be, not a matter of taste; the options sit within the broader landscape of review types and their uses.
First, second, and third order constructs
The family shares one crucial piece of vocabulary. First order constructs are the participants’ own interpretations of their experience, the quotations in a paper. Second order constructs are the original authors’ interpretations of those accounts, the themes and concepts in their findings sections. Third order constructs are the synthesist’s interpretations of the second order constructs, the new concepts the review itself creates. Keeping the three levels distinct is what protects a synthesis from simply relaunching one study’s framework as if it were a finding of the whole literature, and reviewers are expected to show, usually in tables, how each third order construct was built from identifiable second order material.
A concrete example makes the ladder clearer. In a synthesis of medicine-taking, a participant saying “I stop the tablets when I feel like myself again” is a first order construct. The original authors’ theme of “strategic non-adherence” is a second order construct. The synthesis team’s concept of resistance as rational testing, built by translating that theme across a dozen studies of different conditions, is a third order construct, and it is the level at which a meta-synthesis makes its distinctive contribution to knowledge.
Searching: exhaustive retrieval or purposive sampling
Searching is where meta-synthesis debates its own identity. One camp argues for the exhaustive approach of effectiveness reviews, running sensitive strategies across the major bibliographic databases plus citation chasing, on the ground that transparency demands it. The other camp argues that an interpretive synthesis does not need every relevant study, it needs the right ones, and so defends purposive sampling of maximally informative papers and stopping at conceptual saturation, the point where new studies stop changing the developing interpretation. Qualitative research is also indexed erratically, so berry-picking techniques and reference chasing carry more weight than in quantitative reviewing. In practice most published syntheses run a comprehensive search and then, if the yield is unmanageable, sample from it using explicit, pre-stated criteria such as conceptual richness, which keeps both transparency and interpretive depth defensible.
Running the synthesis in practice
Whatever family you choose, the working sequence is broadly stable. The team fixes the question and the synthesis method in a protocol, runs and documents the search, and screens in duplicate, usually inside the same screening platforms used for quantitative reviews. Extraction then diverges from quantitative practice: reviewers lift the findings, the themes, concepts, and supporting quotations, rather than numbers, and record enough context about setting, population, and method to interpret them. The analysis itself is iterative. Codes and candidate themes are drafted, challenged in team meetings, tested against deviant cases, and revised, with memos recording why each decision was taken. Software matters mainly as bookkeeping: qualitative analysis packages and review platforms hold the coding tree and keep every construct linked to its source passages, which is what makes the audit trail real rather than rhetorical. Two disciplines separate strong syntheses from weak ones. The first is constant comparison, forcing every new study to interrogate the developing interpretation instead of being slotted into it. The second is reflexivity: the team records its own assumptions and how they shifted, because in an interpretive method the analysts are themselves an instrument, and an unexamined instrument is an uncalibrated one.
Appraising studies with CASP
Quality appraisal is contested too, but the dominant instrument is the CASP qualitative checklist, ten questions covering aims, methodology, design, recruitment, data collection, reflexivity, ethics, analysis, findings, and value. Unlike risk of bias scoring in trials, CASP results rarely drive exclusion, because a methodologically thin paper can still contribute a conceptually important insight. Most teams use appraisal to weight their confidence, run sensitivity analyses checking whether themes survive the removal of weaker studies, and feed the judgements forward into confidence rating. What matters is deciding the policy in the protocol, before results are known, exactly as you would when writing any systematic review.
GRADE-CERQual: how much confidence in each finding
GRADE-CERQual does for qualitative findings what the GRADE certainty framework does for effect estimates. Each review finding is rated on four components: methodological limitations of the contributing studies, drawn from the CASP appraisals; coherence, how well the data support the finding; adequacy, the richness and quantity of supporting data; and relevance, how directly the studies bear on the review question. The result is a confidence level of high, moderate, low, or very low for every finding, presented in a summary of qualitative findings table. Guideline panels increasingly require CERQual ratings before qualitative evidence can inform recommendations, which has quietly professionalised the whole field.
Reporting with ENTREQ, and when meta-synthesis is the right method
Reporting follows the ENTREQ statement, twenty-one items covering the synthesis methodology, searching, screening, appraisal, and how constructs were derived, playing the role that the PRISMA 2020 checklist plays for quantitative reviews. As for when to choose the method: a meta-synthesis is the right tool when the question is about experience, meaning, acceptability, or process, why people decline screening, how patients live with a condition, what makes an intervention workable in practice. It is the wrong tool for questions of effectiveness, which need trials and pooling, and it adds little when the literature is too thin or too descriptively shallow to interpret. The strongest evidence programmes now pair the two, a quantitative review establishing whether something works and a linked meta-synthesis explaining how, for whom, and under what conditions, and it is that explanatory power that has moved qualitative synthesis from the margins of evidence-based practice to its centre.
Choosing within the family follows from the same logic. If the aim is to generate theory from a conceptually rich literature, meta-ethnography or grounded formal theory fits. If the aim is practical guidance for a profession, meta-aggregation delivers findings in a form committees can act on. If a credible conceptual framework already exists, framework synthesis tests and extends it efficiently. If the team is newer to qualitative synthesis or the literature is descriptively oriented, thematic synthesis offers the most forgiving route to a rigorous result. Timelines are comparable to a conventional review, typically four to nine months, with the analysis phase, not the searching, consuming the largest share, because translation and interpretation cannot be rushed without producing exactly the shallow summary the method exists to avoid. The one mistake that outranks all others is treating the synthesis as a sorting exercise, filing quotes under headings and calling the headings findings. A meta-synthesis earns its place in the literature only when its third order constructs tell readers something no included study could have told them alone.