What Is a Meta-Analysis?

A meta-analysis takes the idea of literature reviews a step further: instead of just discussing the studies side by side, it statistically combines their actual numerical results into one larger, more powerful analysis.

This is more important than you might think. A single study might have a small sample size, or results that could plausibly be explained by chance, or by some quirk specific to that particular group of subjects. When you combine data from dozens of independent studies all asking the same underlying question, patterns that were too subtle to reliably detect in any one study on its own often become clear. That's why meta-analyses sit near the very top of what researchers call the "evidence pyramid," a rough hierarchy of how much confidence different kinds of evidence deserve.

A closely related term you'll often see alongside it is a systematic review, which is a rigorous, methodical search and evaluation of all available studies on a topic, following a predefined process. A meta-analysis is frequently, though not always, part of a systematic review.

One challenge every meta-analysis has to grapple with is publication bias — the well-documented tendency for studies with exciting, positive results to get published far more often than studies that find nothing interesting, which sit in file drawers unpublished. If a meta-analysis only pulls from published papers, it can end up systematically overestimating an effect, simply because the "nothing happened here" studies never made it into the pool in the first place. Researchers try to catch this using a tool called a funnel plot, which can reveal a telltale lopsided pattern when negative or null results are missing from the published record.