Library / research-methodology-guide
Understanding meta-analysis
Audience: patients-public, students-learners ยท Review state: published
What a meta-analysis does
A meta-analysis pools the numerical results of several studies that asked the same question, producing one combined estimate with a tighter confidence interval than any single study alone. It sits at the end of a systematic review, which first finds and appraises the studies under a pre-registered protocol.
The strength of the result depends entirely on what went in: a meta-analysis of weak, biased studies produces a precise but misleading number.
How the pooling works
Each study contributes an effect size and a measure of its precision. Studies with more participants or less variance carry more weight, so a large careful trial moves the combined estimate further than a small noisy one.
Two common models exist. A fixed-effect model assumes all studies estimate one true effect. A random-effects model assumes the true effect varies between settings and estimates its average. The choice changes the result and its uncertainty.
Heterogeneity and bias
Heterogeneity is variation between study results beyond what chance explains, often shown with the I-squared statistic. High heterogeneity warns that the pooled number may not describe any single setting well, and good analysts investigate its sources rather than ignore it.
Publication bias is the second threat: studies with positive results appear in print more often than negative ones, so the visible literature overstates effects. Funnel plots and related tests look for this asymmetry.
Reading one critically
Check whether the review behind the analysis had a published protocol and a real multi-database search. Look at which studies were included and excluded, whether heterogeneity was examined, and whether the conclusion matches the size and precision of the pooled effect.
A meta-analysis is a summary of a literature, not a verdict on reality. New trials can shift the pooled estimate, which is why living reviews update as evidence arrives.
Limitations
Educational methods overview, not a critique of any specific meta-analysis and not guidance for clinical decisions.