Introduction to Meta-analysis for Medical Students
What a meta-analysis does, how to read a forest plot, fixed-effect vs random-effects models, heterogeneity (I²) and publication bias, explained simply.
A meta-analysis is a statistical method for combining the results of several studies that address the same question. When done within a systematic review, it gives a single, more precise estimate of an effect and shows how consistent the studies are.
When is a meta-analysis appropriate?
Only when the studies are similar enough in participants, interventions, comparisons and outcomes to make a combined estimate meaningful. Pooling very different studies produces a precise-looking number that answers no real question. A systematic review without a meta-analysis is perfectly acceptable when pooling is not sensible.
Effect measures
| Outcome type | Common effect measure |
|---|---|
| Binary (yes/no) | Risk ratio (RR), odds ratio (OR), risk difference |
| Continuous, same scale | Mean difference (MD) |
| Continuous, different scales | Standardised mean difference (SMD) |
| Time to event | Hazard ratio (HR) |
How studies are combined
Each study's result is given a weight, usually based on its precision (larger, more precise studies get more weight). The weighted results are combined into a pooled estimate with a 95% confidence interval.
- Fixed-effect model: assumes all studies estimate one common true effect.
- Random-effects model: assumes the true effect varies between studies and estimates the average. It usually gives wider confidence intervals when studies disagree.
Reading a forest plot
- Each line is one study: the square is its estimate (bigger square, more weight) and the horizontal line is its 95% CI.
- The vertical line is "no effect" (1 for ratios, 0 for differences).
- The diamond at the bottom is the pooled estimate; its width is the pooled CI.
- If the diamond does not cross the line of no effect, the pooled result is statistically significant at the chosen level.
Heterogeneity
Heterogeneity is variation in results beyond what chance would explain. It is commonly described with the I² statistic, the percentage of variation due to real differences rather than chance. The Cochrane Handbook gives rough, overlapping guides: 0–40% might not be important, 30–60% may be moderate, 50–90% may be substantial and 75–100% considerable. Always look for clinical or methodological reasons for heterogeneity, using subgroup or sensitivity analyses planned in your protocol.
Publication bias
Studies with positive results are more likely to be published, which can exaggerate pooled effects. Funnel plots and related tests can help explore this, but they are unreliable with few studies; the Cochrane Handbook advises they are generally used only when at least 10 studies are included.
Software
Common options include Cochrane's RevMan and the R packages meta and metafor. Stata also has meta-analysis commands.
Common mistakes
- Pooling studies with different designs or outcomes without justification.
- Ignoring high heterogeneity and reporting only the pooled estimate.
- Double-counting participants from multiple reports of the same study.
- Drawing strong conclusions from a few small, high-risk-of-bias studies.
Key takeaways
- Meta-analysis combines similar studies into a weighted pooled estimate.
- Choose effect measures by outcome type; choose fixed or random effects thoughtfully.
- Read the forest plot: squares, lines, the line of no effect and the diamond.
- Report and explore heterogeneity; be cautious with publication-bias tests.
References
- Higgins JPT, Thomas J, Chandler J, et al., eds. Cochrane Handbook for Systematic Reviews of Interventions. Cochrane. training.cochrane.org/handbook
- Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ. 2003;327(7414):557–560. doi:10.1136/bmj.327.7414.557
- Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. doi:10.1136/bmj.n71
This article is for education. Always follow the requirements of your university, guide, ethics committee and target journal.