The essential foundation
Understanding this project
Meta-analysis combines compatible quantitative results statistically. The key challenge is not producing a pooled number but making defensible decisions about effect measures, model choice, heterogeneity, study quality and interpretation.
Meta analysis can summarise quantitative evidence, but it is not an automatic requirement of a systematic review. Researchers must understand the effect measure, the populations being compared, the variability between studies and the assumptions of the statistical model. We focus on interpretation and methodological defensibility rather than producing an attractive forest plot without a sound rationale.
A decision specific to this project
Pooling decision
First inspect outcome definitions, follow up periods and interventions for comparability. Then consider effect measures and model assumptions.
What to prepare
An explicit pooling rationale, effect measure selection and uncertainty interpretation.
Methodological boundary: Statistical significance is not interchangeable with clinical significance.
Where difficulties begin
When Studies Cannot Be Combined Reliably
I do not know whether the included studies can be pooled.
I am unsure which effect size to use.
My heterogeneity is high.
I need help interpreting forest plots and pooled estimates.
I need to explain sensitivity analyses clearly.
A practical approach
Assessing Whether Quantitative Pooling Is Defensible
Compare the clinical question, intervention definitions, outcome measures and study designs. Statistical techniques cannot remove fundamental differences in what the studies measure. Understand mean differences, standardised mean differences, risk ratios and odds ratios. The choice depends on the outcome and data available, not the visual appearance of the results.
Decide whether pooling is justified
Compare the clinical question, intervention definitions, outcome measures and study designs. Statistical techniques cannot remove fundamental differences in what the studies measure.
Choose an appropriate effect measure
Understand mean differences, standardised mean differences, risk ratios and odds ratios. The choice depends on the outcome and data available, not the visual appearance of the results.
Interpret heterogeneity and uncertainty
Assess clinical and methodological diversity alongside statistical heterogeneity. Confidence intervals and sensitivity analyses help explain uncertainty, while model choice requires explicit justification.
Communicate results responsibly
Present effect estimates with confidence intervals, study weights and limitations. Explain the strength of the evidence without overstating significance or clinical importance.
Putting the ideas into context
Comparing Medication-Adherence Outcomes
Suppose studies examine educational interventions to improve medication adherence but measure adherence using different instruments. A researcher needs to assess whether standardisation is appropriate and whether the interventions are sufficiently comparable before calculating a combined effect.
Illustrative example only; check your programme requirements, actual evidence and any necessary permissions.
What support looks like for this topic
How we support your research
Suitability Assessment
Determine whether the studies, outcomes and designs are sufficiently compatible for quantitative pooling.
Effect Measures
Choose and calculate effect measures that fit the outcome type and included studies.
Model Selection
Justify fixed- or random-effects approaches and the assumptions behind the model.
Heterogeneity
Assess variation across studies and investigate plausible sources where appropriate.
Sensitivity & Bias
Explore robustness, influential studies and publication-bias considerations where feasible.
Interpretation
Explain the pooled result in context rather than relying on statistical significance alone.
Good research practice
A practical note for this stage
If the studies differ or the results are uncertain, we can work through the possible analysis choices and how to explain what the findings actually mean.
Explore the subject further
Related specialist areas
Focused guidance for the next decision in your research.
Question formulation, protocol, searching, screening, appraisal, synthesis and PRISMA reporting.
Map the breadth of evidence, concepts, study characteristics and research gaps.
Effect sizes, pooling, heterogeneity, sensitivity analysis, forest plots and interpretation.
PICO concepts, keywords, controlled vocabulary and searches across CINAHL, PubMed and other databases.
Eligibility criteria, title/abstract screening, full-text decisions and structured extraction.
JBI, CASP, RoB 2 and other design-appropriate appraisal approaches.
Questions researchers ask
Frequently asked questions
Can I request only one stage of a review?
Yes. The architecture supports modular requests such as search strategy, screening, data extraction, appraisal, synthesis or manuscript support.
Do all systematic reviews need a meta-analysis?
No. Meta-analysis is appropriate only when the included studies are sufficiently compatible and the review question supports quantitative pooling.
Can a scoping review be used instead?
Yes, when the objective is to map the breadth and characteristics of evidence rather than estimate a pooled effect.
Further reading
Research standards and reliable resources
These external resources can help you explore relevant methodological guidance. Always follow the requirements of your programme and verify the version of any standard you use.