Guides for researchers
The library explains the concepts that come up most often when planning, running and reporting an evidence synthesis. Each guide gives a direct answer first, then the detail, with references to the methodological literature. Use the search box to filter the guides, or browse by topic.
Start here
- What is meta-analysis?A statistical method for combining results from separate studies, explained with a worked example.
- Meta-analysis vs systematic reviewHow the two relate, and when one is done without the other.
- Choosing a review typeA decision guide to systematic, scoping, rapid and umbrella reviews.
- Meta-analysis & evidence-synthesis glossaryTerms used in meta-analysis and evidence synthesis, cross-linked.
- Frequently asked questionsCommon questions about services, methods and the process.
Models and heterogeneity
- Fixed-effect vs random-effects modelsThe assumptions behind each model and how to choose.
- What is heterogeneity in meta-analysis?Clinical, methodological and statistical heterogeneity, and what to do about it.
- What is I²?What I² measures, and why it is not an absolute measure of difference.
- What is tau-squared (τ²)?The between-study variance and the estimators used for it.
- Prediction intervals in meta-analysisThe range of true effects expected in a new study.
- Sensitivity analysis in meta-analysisTesting whether a conclusion survives reasonable changes in method.
- Subgroup analysis in meta-analysisComparing pooled estimates across categories, and the pitfalls.
- How many studies do you need?Why there is no fixed minimum, and what few studies limit.
- Handling zero-event studiesContinuity corrections, Peto and Mantel-Haenszel methods for sparse data.
- Missing data in meta-analysisFinding, deriving and imputing missing results, and testing the effect.
Reading results
- How to interpret a forest plotThe anatomy of a forest plot and how to interpret the pooled result.
- What is a funnel plot?What asymmetry can and cannot show.
- What is publication bias?Small-study effects, selection models and assessment tools.
- Egger's testThe regression test for funnel-plot asymmetry and its limits.
- Trim-and-fill methodHow the adjustment works, its assumptions and criticisms.
- Small-study effectsCauses of small-study effects beyond publication bias.
Effect sizes
- Effect sizes in meta-analysisChoosing among standardized mean differences, ratios and correlations.
- Odds ratio vs risk ratioInterpretation of each, and why they differ for common outcomes.
- Standardized mean difference (Hedges' g, Cohen's d)Hedges' g and Cohen's d: computation and interpretation.
- Pooling hazard ratiosExtracting and pooling hazard ratios from survival outcomes.
- Correlation-based meta-analysisFisher's z, reliability corrections and use in psychology and management.
Reporting and registration
- PRISMA 2020 statementThe reporting checklist and flow diagram for systematic reviews.
- PRISMA extensions (ScR, DTA, NMA, IPD, P)Which PRISMA extension suits which kind of review.
- MOOSE reporting for observational meta-analysesReporting meta-analyses of observational studies.
- EQUATOR network & reporting guidelinesFinding the reporting guideline a journal expects.
- How to register a review in PROSPEROWho can register a review, and how.
- The Cochrane handbook for systematic reviews of interventionsThe structure of the Handbook and how to cite it.
- Reporting a meta-analysis: section-by-sectionWriting up a meta-analysis section by section.
Searching, screening and extraction
- PICO and related question frameworksPICO and related frameworks for framing the question.
- Building a systematic review search strategyBoolean operators, subject headings and search filters.
- Which databases to searchWhich databases to search for which field.
- Grey literature in systematic reviewsSources, risks and documentation of unpublished material.
- Deduplication and study screeningDe-duplication, dual screening and flow documentation.
- Data extraction in systematic reviewsExtraction forms, piloting and independent extraction.
Appraisal and certainty
- GRADE: rating certainty of evidenceRating the certainty of evidence, domain by domain.
- Risk-of-bias tools: RoB 2, ROBINS-I, NOS, QUADAS-2RoB 2, ROBINS-I, Newcastle-Ottawa and QUADAS-2: which tool for which design.
Network and Bayesian methods
- Transitivity & inconsistency in network meta-analysisChecking the assumptions behind network meta-analysis.
- SUCRA and treatment rankingsTreatment rankings and the cautions that go with them.
- Choosing priors in Bayesian meta-analysisChoosing priors, and the heterogeneity prior in particular.
Software and publishing
- Software for meta-analysisThe software commonly used for meta-analysis and how they compare.
- Reproducible meta-analysisCode, data sharing and version control for reproducible work.
- Responding to reviewers on a meta-analysisTypical statistical queries and how to structure a response.
- Choosing a journal for an evidence-synthesis paperScope, review types accepted and checks for predatory journals.
- Best-evidence synthesisAn alternative to pooling, and when it is used.
A suggested reading order
If you are new to the subject, this sequence builds from the basic idea to the points that most often matter in a manuscript.
What is meta-analysis?
The idea, the steps and a worked example.
Fixed-effect vs random-effects models
The two models and the assumption behind each.
What is heterogeneity in meta-analysis?
Why studies differ and how that is measured.
How to interpret a forest plot
How the standard figure is read.
What is publication bias?
The main threat to a pooled result.
PRISMA 2020 statement
How a review is reported so that readers can judge it.
Frequently asked questions
Where should a beginner start?
With the guide on what meta-analysis is, which includes a worked example, followed by the guides on fixed-effect and random-effects models, heterogeneity and forest plots. The suggested reading order on this page lists them in sequence.
Are the guides based on published sources?
Yes. Each guide gives its sources, which are mainly the Cochrane Handbook, the PRISMA, MOOSE and GRADE publications, and peer-reviewed methodological papers.
Can I use these guides instead of expert help?
The guides explain concepts and standards, but a specific project depends on its question and its studies. They are written to help you plan and to understand advice, not to replace advice on a particular design.