What is a meta-analysis?
A meta-analysis takes the effect sizes from several studies and computes a weighted average based on each study's size and precision, producing one combined result. Gene Glass named the method in 1976. It typically follows a systematic review, which uses a predefined process to find every relevant study and assess its quality.
How do you read a forest plot?
Results are usually shown as a forest plot. Each row is one study's estimate with its 95% confidence interval, and the diamond at the bottom is the pooled result. For hazard ratios or relative risks the "no difference" line is 1; for mean differences it's 0. If a confidence interval doesn't cross that line, the difference is considered statistically significant.
Models and heterogeneity
- Fixed-effect model: assumes every study estimates the same true effect.
- Random-effects model: allows the true effect to vary between studies and accounts for that variance.
- I²: expresses how much between-study variation exceeds chance, from 0 to 100%; roughly 25% is low, 50% moderate and 75% high (Higgins et al., 2003).
Variations and limitations
Individual-patient-data meta-analyses pool raw data from every participant rather than summary numbers, making them more precise. Network meta-analyses compare treatments that were never tested head-to-head through a shared comparator. But pooling weak studies yields weak results ("garbage in, garbage out"), and publication bias—where positive results are more likely to be published—can inflate effects. That's why funnel plots, sensitivity analyses and evidence-grading systems like GRADE are used alongside. For the study designs being pooled, see randomized controlled trials and cohort studies.
FAQ
- Is a meta-analysis the same as a systematic review?
- No. A systematic review is the structured search and appraisal of studies; a meta-analysis is the statistical pooling that often follows it. Not every systematic review includes a meta-analysis.
Sources reviewed: 2026-10-25
References
- Glass, G. V. (1976). Primary, secondary, and meta-analysis of research. Educational Researcher, 5(10), 3–8.
- Higgins, J. P. T., Thompson, S. G., Deeks, J. J., & Altman, D. G. (2003). Measuring inconsistency in meta-analyses. BMJ, 327, 557–560.
- Page, M. J., et al. (2021). The PRISMA 2020 statement. BMJ, 372, n71.