Meta-analysis: ask which results were combined

If a headline presents one impressive number “across studies,” start by finding which studies and which result from each study contributed to it. You do not have to reproduce the calculation to ask what the number is about.

A headline about reusable folders

In a fictional example, a report combines results from several comparisons of reusable event folders. Nadiya reads the headline as a verdict on the best folder for every school project. Her family opens the report before choosing supplies for a community display.

One included study concerns adults carrying documents indoors. Another concerns folders kept in outdoor stalls. The family asks what the combined result measures and how these settings were handled. They do not decide that the studies must be combined—or must not be combined—just from those short descriptions.

Nadiya's child wants a folder that opens easily. Her partner wants one that fits an existing bag. If the report's measured outcome is the number of torn corners after an event, the family has identified a different question from either preference. They can choose a folder without claiming the combined result answered every practical concern.

No folder findings or numbers in this scene come from a real study. The example illustrates where to look in a report, not how to perform a statistical analysis.

Follow the number back to its ingredients

You might look for answers to these questions:

• What outcome does the combined number describe?
• What is the unit: minutes, a proportion, a score, or another measure?
• How many studies contribute to this particular result?
• Which people, materials, and settings did those studies include?
• What uncertainty and limitations accompany the combined estimate?
• Does the report explain differences among the study results?

Keep the count for the particular result beside it. If the headline gives the total number of studies in a larger review, check whether all of them contributed to the number you are reading. Leave that connection unconfirmed until you find it.

Combining and selecting are different reading tasks

Use systematic review to investigate how the authors searched for and selected the evidence. Use meta-analysis to investigate the numerical combination. Ask both questions when a paper uses both labels, rather than assuming the calculation explains how the studies were chosen.

Sample size offers another useful check: which count refers to studies, which to participants, and which to a particular analysis? In your own notes, avoid turning all those counts into a single large “n.” If you cannot tell what was counted, quote no number until you can describe it accurately in your own words.

A result need not settle a family preference

A parent can be interested in a pooled finding while a child asks a question that it does not address. A sibling does not have to defend their preference with research. Partners can say, “This is one piece of information; our size and access requirements are separate.”

This page provides a route through a claim, not a shortcut for certifying a meta-analysis or applying clinical results to an individual. It is reasonable to stop with a clearly named uncertainty.

Source

[1] Deeks JJ, Higgins JP, Altman DG, McKenzie JE, Veroniki AA, editors. Chapter 10: Analysing data and undertaking meta-analyses. Cochrane Handbook for Systematic Reviews of Interventions, version 6.5. Cochrane; 2024. Chapter updated November 2024; accessed September 5, 2026. First key point supports [1]. No source figures, formulas, or study results reproduced.

Related guides

• Systematic review
• Sample size
• Primary source
• Control group