What Is a Meta-Analysis? A Plain-English Guide

What is a meta-analysis?

A meta-analysis is a set of statistical tools used to combine the results of multiple studies that address the same research question. The goal is simple: pool the data so you get one overall estimate of an effect, with more precision than any individual study could offer on its own¹.

The term covers a family of methods rather than a single technique. Different research questions need different criteria for selecting studies, different ways of organising the data, and different statistical approaches to synthesising it¹. A meta-analysis comparing two treatments, for example, works differently from one comparing many treatments at once.

How it relates to a systematic review

A meta-analysis is the statistical step that often follows a systematic review. The systematic review finds all relevant studies, screens them for quality and extracts the data. The meta-analysis then does the number-crunching on that data².

This is why the two terms are often mentioned together. A meta-analysis cannot rescue a sloppy search or a biased selection of studies. Its output is only as trustworthy as the review process that gathered the evidence in the first place².

Why meta-analyses sit at the top of the evidence hierarchy

In the study design hierarchy, meta-analysis produces evidence at the highest level¹. The reason is straightforward: it draws on more data than any single study. Where one randomised trial might involve a few hundred people, a meta-analysis can pool thousands across many trials, which reduces the influence of chance and gives a more stable estimate.

This is also why the newsroom weights meta-analyses most heavily when we assess claims. A single small study can produce a striking result by luck. A well-conducted meta-analysis asks what the whole body of evidence says. That is closer to the truth.

What makes a meta-analysis trustworthy?

Not all meta-analyses are equal. Two things matter most.

First, the quality of the underlying review. If the search for studies was incomplete or the selection criteria were biased, the pooled result will inherit those flaws².

Second, the transparency of the reporting. In 2009, an international group of researchers developed PRISMA, a 27-item checklist and four-phase flow diagram designed to improve how systematic reviews and meta-analyses are reported². The checklist exists because reviews of published meta-analyses found that key information was often missing or poorly described, which diminished their value to clinicians and policymakers².

When you read a meta-analysis, look for signs that the authors followed a reporting standard like PRISMA. If the methods section is vague about how studies were found and selected, treat the conclusions with caution.

Different types of meta-analysis

Meta-analysis is not one method but a broad term for a set of statistical tools¹. The most common form is pairwise meta-analysis, which compares two treatment groups directly. But when researchers want to compare multiple treatments at once, they use network meta-analysis, which can rank several options against each other¹.

Presenting the results of these more complex analyses is itself a challenge. Researchers have developed graphical methods, such as ranking plots, to show which treatments are likely to be most effective across a network of trials³. These displays help readers interpret results that would be difficult to grasp from a single number.

What are the limits?

A meta-analysis is only as good as the studies it pools. If the individual studies are small, poorly designed or measure outcomes differently, combining them can produce a precise-looking but misleading result. Statisticians emphasise that planning and conducting a meta-analysis requires close collaboration with experienced methodologists¹.

The evidence hierarchy is a guide, not a guarantee. A well-run randomised trial can sometimes tell you more than a badly conducted meta-analysis. The method matters as much as the label.

How to read a meta-analysis

When you encounter a meta-analysis, ask three questions:

1. Was the search systematic? Did the authors describe how they found studies, and would a reader be able to repeat that search?²
2. Was the reporting transparent? Did they follow a checklist like PRISMA, and do they explain their statistical assumptions?^1,2
3. Does the pooled result match the individual studies? If a few large studies dominate the pool, the overall estimate may reflect them more than the wider evidence.

If the answer to any of these is unclear, the meta-analysis deserves less weight than its position in the hierarchy might suggest.

For more on how to weigh research claims, see our explainer on why so many wellness claims rest on a single small study and what correlation is not causation means.