The One-Sentence Version
A meta-analysis is a study of studies: it statistically combines the results of multiple separate studies that asked the same question, producing a single, more precise estimate than any one study alone.
Systematic Review vs Meta-Analysis
The two terms travel together but mean different things. A systematic review is the method — a transparent, pre-planned search for every relevant study, followed by a quality appraisal [1]. A meta-analysis is the optional statistical step that pools those studies' numbers. You can have a systematic review without a meta-analysis, but a trustworthy meta-analysis almost always sits on top of a systematic review (see Systematic Review vs Narrative Review).
How Pooling Works
Each study is weighted — larger, more precise studies count for more — and the results are combined into a single summary effect, often shown on a forest plot. The pooled estimate comes with a confidence interval that's usually narrower than any single study's, because more data means more precision (see Relative Risk & Confidence Intervals).
Heterogeneity: When Studies Disagree
If the included studies point in very different directions, the pooled number can be misleading. Researchers measure this disagreement (often reported as *I²*). High heterogeneity is a warning that the studies may be too different — in dose, population, or design — to combine cleanly.
Garbage In, Garbage Out
A meta-analysis is only as good as its inputs. Pooling ten biased or tiny studies produces a precise-looking but unreliable answer. Publication bias — the tendency for positive results to get published while null results sit in a drawer — can also tilt the pooled estimate toward a benefit that isn't really there.
How to Read One Critically
- How many studies and participants were pooled?
- Were they randomized trials or weaker designs?
- Was heterogeneity high?
- Did the authors check for publication bias?
- Who funded the included studies?
Government resources like NCCIH's *Know the Science* walk through these checks in plain language [2].