1
What is a meta-analysis?
A meta-analysis combines numerical results from multiple independent studies into a single pooled estimate. Instead of one study at a time, it gives you the best overall estimate by weighing each study by its precision and sample size.
Part of a systematic reviewA meta-analysis is usually done within a systematic review but is not always required. You can have a systematic review without a meta-analysis if studies are too different to pool.
What is a forest plot?⌄
A forest plot shows each study as a horizontal line. The square is the effect estimate, the line is the confidence interval. The diamond at the bottom is the pooled estimate. If a line crosses the vertical no-effect line, that study alone is not statistically significant.
What is heterogeneity and why does it matter?⌄
Heterogeneity is variation between studies beyond chance. Measured by I-squared. Above 75% suggests substantial heterogeneity and means you should be cautious about pooling results from very different studies.
2
Step by step guide
1
Complete your systematic review first
A meta-analysis requires a complete systematic search. Identify all relevant studies before deciding which to pool.
2
Assess study comparability
Only pool studies similar in population, intervention, comparison, and outcome.
3
Extract numerical data
For each study extract the effect size, sample size, and measure of variance.
4
Choose your statistical model
Fixed effects assumes studies estimate the same true effect. Random effects assumes true effects vary. Random effects usually preferred.
5
Run the analysis
RevMan (free from Cochrane) is standard software. Enter data, run analysis, generate forest plot.
6
Report heterogeneity
Report I2 statistic, Cochrane Q statistic, and p-value. If I2 is high conduct subgroup analyses.
🎬
Video walkthrough coming soon
A complete video guide for visual and auditory learners.
Coming soon3
Checklist
Click each item to mark complete.
Systematic review completed first
Studies assessed for comparability
Numerical data extracted from all studies
Statistical model chosen with justification
RevMan or equivalent software used
Forest plot generated and checked
I2 and Q statistic reported
Subgroup analysis if high heterogeneity
Funnel plot for publication bias
PRISMA extension for meta-analyses followed
Test your understanding
5 questions on the key concepts.
1. What does a meta-analysis do?
2. What does the diamond on a forest plot represent?
3. What does I-squared measure?
4. Which free software is standard for meta-analysis?
5. Can you do a meta-analysis without a systematic review?