Meta-Analysis

Definition

Meta-analysis is a quantitative method for synthesising results from multiple independent studies addressing the same research question. It combines effect-size estimates across studies to produce a pooled estimate with greater precision and statistical power than any single study alone. It also allows examination of heterogeneity, publication bias, and moderator effects.

Why It Matters

Individual studies often produce conflicting or under-powered results. Meta-analysis provides the highest level of evidence in evidence-based medicine and policy evaluation by systematically aggregating findings. It is the standard tool behind clinical guidelines, systematic reviews, and many influential policy recommendations. Understanding meta-analysis helps researchers critically evaluate review papers and design studies that will be useful for future syntheses.

Example

Twenty randomised controlled trials have examined whether aspirin reduces the risk of heart attack, with individual results ranging from strong benefit to no effect. A meta-analysis pools the hazard ratios from all twenty studies, yielding a precise pooled estimate of 0.82 with a narrow 95% confidence interval of [0.76, 0.89]. This indicates a consistent and statistically significant 18% risk reduction across the literature.

Related Terms

• Forest Plot
• Funnel Plot
• Effect Size
• Statistical Power
• Confidence Interval
• Random Sampling

Software Notes

• SPSS: SPSS does not have a built-in meta-analysis module. Researchers typically use Comprehensive Meta-Analysis (CMA), RevMan, or export data to R.
• R: metafor::rma(yi, vi, data = df) fits a random-effects meta-analysis. meta::metagen() provides comprehensive output including forest plots. Use funnel() and trimfill() for publication-bias diagnostics.
• Stata: meta set (Stata 17+) for modern meta-analysis syntax. metan effect se in older versions. metafunnel, metabias, and metatrim assess publication bias.