What is Meta-Analysis?
Meta-analysis is a statistical technique that combines the results of multiple independent studies addressing the same research question. By pooling data, meta-analysis increases the sample size and statistical power, providing more precise estimates of effect size than any single study alone.
Meta-analysis is typically conducted as part of a systematic review, though not all systematic reviews include a meta-analysis. The decision depends on the clinical and methodological heterogeneity of the included studies.
Understanding Forest Plots
A forest plot is the hallmark visualization of a meta-analysis. It displays the effect estimate and confidence interval for each individual study, as well as the pooled (overall) effect. Key elements include:
- Individual study lines: A square (point estimate) and horizontal line (confidence interval) for each study
- Square size: Proportional to the study weight — larger squares indicate studies contributing more to the pooled estimate
- Diamond: The pooled effect estimate at the bottom; its width represents the confidence interval
- Vertical line: The line of no effect (1.0 for ratio measures, 0 for difference measures)
Heterogeneity: I-squared and Cochran Q
Heterogeneity refers to variability in study results beyond what would be expected by chance alone. Two commonly used statistics for measuring heterogeneity are:
- Cochran's Q test: A chi-squared test that assesses whether observed differences in results are compatible with chance alone. A significant p-value (typically <0.10) suggests heterogeneity.
- I-squared (I²): Describes the percentage of variability in effect estimates due to heterogeneity rather than chance. Values of 25%, 50%, and 75% are considered low, moderate, and high heterogeneity respectively.
Fixed Effects vs. Random Effects Models
The choice between fixed-effects and random-effects models is crucial:
- Fixed-effects model: Assumes all studies estimate the same true effect. Appropriate when studies are clinically and methodologically similar with low heterogeneity.
- Random-effects model: Assumes that the true effect varies between studies. Gives more weight to smaller studies. Appropriate when heterogeneity is expected (which is the case in most real-world scenarios).
Publication Bias and Funnel Plots
Publication bias occurs when studies with statistically significant results are more likely to be published than those with null results. A funnel plot visualizes this by plotting each study's effect estimate against its standard error. In the absence of bias, the plot should look like an inverted funnel — symmetric around the pooled estimate. Asymmetry may suggest publication bias.
Formal tests for funnel plot asymmetry include Egger's test and Begg's test. Trim-and-fill analysis can estimate the number of potentially missing studies.
Subgroup Analysis and Meta-Regression
When heterogeneity is high, subgroup analysis can help identify sources of variability. Common subgrouping variables include study design, geographic region, participant age, intervention dose, and follow-up duration. Meta-regression extends this by modeling the relationship between study-level covariates and effect size.
Sensitivity Analysis
Sensitivity analysis tests the robustness of the pooled estimate by systematically removing one study at a time (leave-one-out analysis), restricting analysis to studies with low risk of bias, comparing fixed and random effects models, and using different effect measures or statistical methods.
Software for Meta-Analysis
Several software options are available for conducting meta-analysis: RevMan (Cochrane's free tool), R (meta, metafor packages), Stata (metan, metafor commands), Comprehensive Meta-Analysis (CMA), and JASP. At Utkarsh Research Network, we use R and Stata for advanced meta-analytical techniques and can provide all standard outputs including forest plots, funnel plots, and sensitivity analyses.
