In simple terms
A meta-analysis pools the results of several studies that asked the same question. It treats them as one larger body of evidence.
The output is a single summary figure, reported with a confidence interval that shows how much uncertainty is left.
How it works
A meta-analysis normally sits inside a systematic review. Researchers look for every eligible study, judge its quality, and pull the same measure out of each one.
The studies are then averaged, but not equally. Each result is weighted by how precise it is, so large and carefully run studies count for more.
Analysts also test heterogeneity, meaning how far the studies disagree. The I2 statistic puts a number on that spread.
A fixed-effect model assumes every study measures one shared true effect. A random-effects model assumes the true effect varies between settings, and widens the confidence interval to admit it.
Guidance is clear that some sets of studies should not be pooled at all, for example when the participants, treatments or outcomes are too different.
Why it matters
Single studies are often too small to settle a question. Pooling them raises precision and can turn a contradictory literature into a usable answer.
Meta-analyses therefore feed straight into clinical guidelines, drug regulation and public health advice in many countries.
They also inherit the weaknesses of what goes in. If studies that found nothing were never published, the pool tilts toward positive results, a problem called publication bias.
Where you’ll see it
- Cochrane reviews, which combine trials in health care and health policy.
- Headlines that open with a line such as “a review of 87 trials found”.
- National clinical guidelines and regulators assessing a new drug.
- Psychology, education and economics, where single effects are often small.
Example
Suppose twelve small trials of the same blood pressure drug each end inconclusive. A meta-analysis pools all twelve and reports one average effect, with limits tight enough to act on.
Often confused with
A systematic review is the search and appraisal step. A meta-analysis is the optional statistical step that combines the numbers, and many systematic reviews stop short of it.
Key facts
- A meta-analysis is the statistical combination of results from two or more separate studies, calculated as a weighted average in which more precise studies count for more.1
- Heterogeneity is the variation in results across studies, and the I2 statistic measures how much of that variation exceeds chance.1
- A fixed-effect model assumes one shared true effect, while a random-effects model lets the true effect vary and widens the confidence interval.1
- Meta-analysis is a technique used within some systematic reviews, not all of them.2
- Publication bias arises when studies with statistically significant results are more likely to be published, and registering trials at their start is one recommended safeguard.3
Related concepts
In the news
Quick checkWhat is the difference between a systematic review and a meta-analysis?Show answer
A systematic review searches for and appraises the relevant studies. A meta-analysis is the optional statistical step that combines their results into one estimate.
Sources
- Cochrane. Chapter 10: Analysing data and undertaking meta-analyses, Cochrane Handbook for Systematic Reviews of Interventions. Undated (accessed 18 September 2026)
- Cochrane Library. About Cochrane Reviews. Undated (accessed 18 September 2026)
- Viechtbauer, W., Psychometrika 72(2). Publication bias in meta-analysis: Prevention, assessment and adjustments. 2007 (accessed 18 September 2026)
Editorially reviewed by Specialty Digest Editorial TeamLast reviewed September 21, 2026Researched and drafted with AI assistanceReport an issue