Methods and when to use them
Evidence synthesis is a family of methods and not a single technique. The right one depends on the question, on the kinds of studies that exist, and on the data they report. Quantitative methods combine numerical results and are covered first. Review designs, which structure the identification and summarizing of evidence whether or not results are pooled, follow.
Not every method suits every field or dataset, and each page says what the method assumes, what data it needs and where it fails. The guides in the resource library explain the underlying concepts.
Quantitative synthesis
- Meta-analysisThe statistical combination of results from separate studies.
- Pairwise meta-analysisTwo-group comparisons pooled across studies.
- Random-effects meta-analysisAllows the true effect to vary between studies.
- Fixed-effect meta-analysisAssumes one common true effect.
- Network meta-analysisSeveral interventions compared through connected trials.
- Bayesian meta-analysisPrior distributions combined with study data.
- Meta-regressionEffect sizes related to study characteristics.
- Dose-response meta-analysisHow an outcome changes with the amount of an exposure.
- Diagnostic accuracy meta-analysisPooled sensitivity and specificity of tests.
- Prognostic meta-analysisFactors or models that predict outcomes.
- Prevalence meta-analysisPooled proportions of a condition in a population.
- Incidence meta-analysisPooled rates of new events over time.
- Individual participant data meta-analysisParticipant-level data combined across studies.
- Longitudinal meta-analysisOutcomes measured repeatedly over time.
- Multivariate meta-analysisCorrelated outcomes analyzed jointly.
- Cumulative meta-analysisResults updated as each study appears.
- Component network meta-analysisEffects of the components of complex interventions.
- Comparative effectiveness researchHead-to-head evidence on alternatives in practice.
Review designs
- Systematic reviewStructured identification, appraisal and synthesis of all relevant studies.
- Scoping reviewMaps the extent and nature of the evidence without pooling effects.
- Umbrella reviewSynthesizes evidence from existing systematic reviews.
- Rapid reviewStreamlines review methods to meet a time limit, with stated trade-offs.
- Integrative reviewCombines studies of different designs, common in nursing.
- Narrative reviewA non-systematic overview, with a higher risk of bias.
- Qualitative evidence synthesisIntegrates findings from qualitative studies, including meta-synthesis.
- Mixed-methods reviewIntegrates qualitative and quantitative evidence.
- Realist reviewExplains how and why programs work in context.
- Living systematic reviewUpdated continually as new evidence appears.
Choosing a method
Most questions in health and social research map to a small number of methods. The table gives a first orientation.
| Your question | Method to consider |
|---|---|
| Does an intervention work compared with a control? | Pairwise meta-analysis |
| Which of several treatments works best? | Network meta-analysis |
| How accurate is a diagnostic test? | Diagnostic accuracy meta-analysis |
| Which factors predict an outcome? | Prognostic meta-analysis |
| How common is a condition? | Prevalence meta-analysis or incidence meta-analysis |
| Does risk change with the amount of an exposure? | Dose-response meta-analysis |
| Do effects depend on study characteristics? | Meta-regression |
| Are there few studies, with useful prior evidence? | Bayesian meta-analysis |
| Are the studies too different to pool? | Scoping review, narrative review or qualitative evidence synthesis |
| Do you need to map what evidence exists? | Scoping review or umbrella review |
Frequently asked questions
How do I choose between a systematic review and a meta-analysis?
They answer different needs. A systematic review is the structured process of finding and appraising studies, and a meta-analysis is the statistical step that combines their results. Most meta-analyses sit inside a systematic review, and a review can be reported without pooling when the studies are too different.
Can every method be used in every research field?
No. Meta-analysis is well established in clinical medicine, psychology and education, used in parts of the social, management and ecological sciences, and rare elsewhere. Each method page explains where it applies.
What if the studies are too different to pool?
Then pooling would hide more than it shows. A structured narrative synthesis, a scoping review or a qualitative synthesis may be more appropriate, and the scoping stage of a project is where that decision is made.