What a Meta-Analysis Actually Does to the Evidence

28 October 2026

A meta-analysis is a statistical technique that combines the results of several separate studies into a single, more precise estimate. Done well, it can turn a handful of noisy, inconclusive trials into a clear answer. Done badly, it just adds together the same mistakes and hands them back with a more confident-sounding number attached.

How Pooling Studies Works

Every study that measures an effect, whether it is a drug lowering blood pressure or a habit changing sleep quality, comes with a margin of error. Small studies especially tend to produce results that bounce around, sometimes showing a big effect, sometimes none at all, just because of who happened to be included.

A meta-analysis takes several of these studies and combines their results into one weighted average. Larger, more precise studies get more say in the final number than smaller, shakier ones. The idea is that random ups and downs across different studies will tend to cancel out, leaving a clearer signal than any single study could offer on its own.

When Pooling Clarifies The Picture

Pooling works best when the studies being combined are genuinely asking the same question. That means similar populations, similar doses or interventions, similar ways of measuring the outcome, and similar study designs. Under those conditions, each study is like a separate, independent look at the same underlying truth, and averaging them reduces the noise.

Picture several modest trials testing the same blood pressure medication, each too small on its own to say anything with confidence. Individually they might point in slightly different directions purely by chance. Combined, if the underlying effect is real, the signal strengthens and the uncertainty shrinks, giving a clearer answer than any one trial could provide.

When Pooling Just Amplifies Noise

The trouble starts when the studies are not actually measuring the same thing, or when they share a common flaw. If every study used a biased way of measuring the outcome, or failed to account for the same confounding factor, averaging them does not cancel that bias out. It just produces a single, more official looking number that carries the same mistake all the way through.

This is sometimes summed up as garbage in, garbage out. A meta-analysis can take ten flawed studies and produce an estimate with a narrow, reassuring looking range around it, which makes the result look more certain, not less wrong. Confidence in the number goes up while the accuracy of the number does not.

Publication bias makes this worse. Studies that find a positive, exciting result are more likely to be written up and published than studies that find nothing. If a meta-analysis only draws on the published literature, it is pooling a skewed sample of studies, and the pooled effect can end up larger than reality simply because the null results never made it into the pile.

The Forest Plot And What To Check

Meta-analyses are usually presented with a forest plot, a simple chart where each horizontal line represents one study, showing its estimated effect and the range of uncertainty around it. At the bottom sits a diamond shape, which is the pooled estimate from combining everything above it.

The useful habit is to look at the individual lines before looking at the diamond. If most of the study lines overlap with each other, clustering around a similar area, that is a good sign the studies agree and pooling is doing its job. If the lines are scattered all over the chart, barely overlapping, that is a signal that the studies may not be measuring the same thing, and the diamond at the bottom deserves more scepticism than its narrow width suggests.

Heterogeneity The Hidden Warning Sign

Heterogeneity is the technical word for how much studies differ from each other beyond what chance alone would explain. It can come from different patient groups, different doses, different definitions of the outcome, or different lengths of follow up. Some disagreement between studies is normal, but a lot of it is a warning sign.

When heterogeneity is high, the single pooled number starts to hide more than it reveals. It is often more honest at that point to ask why the studies disagree, perhaps the treatment works well in one group and not another, rather than to quote one average figure as if it applies to everyone. A trustworthy meta-analysis will usually explore this directly, breaking results down by subgroup rather than papering over the disagreement.

Reading A Meta-Analysis Yourself

Start by checking how many studies went in and how big they were. A meta-analysis built from two small trials is still a small amount of evidence, no matter how official the pooled number looks.

Check whether the included studies actually measured the same outcome in a similar way. Different definitions of success, different time frames, or different populations should make you treat the pooled result with more caution.

Look for any mention of publication bias or a funnel plot, a chart that checks whether small studies with unfavourable results are missing from the pile. Finally, see whether the authors discuss heterogeneity openly. A meta-analysis that acknowledges disagreement between its studies is usually more trustworthy than one that presents a single tidy number without comment.

Common questions

What's the difference between a meta-analysis and a systematic review?

A systematic review is the process of searching for and gathering all the relevant studies on a question using a transparent, repeatable method. A meta-analysis is the optional next step, where the numbers from those gathered studies are statistically combined into one pooled estimate.

Can a meta-analysis be wrong even if every individual study is well done?

Yes. If the well conducted studies are measuring subtly different things, or if the published studies are not a representative sample of all the research that was done, the pooled result can still mislead even though each piece going into it was sound.

What is publication bias and how does it affect meta-analyses?

Publication bias is the tendency for studies with striking or positive results to get published more often than studies finding nothing notable. A meta-analysis built only from published papers can end up overestimating an effect because the quieter, null results are underrepresented in what it is averaging.

Why do two meta-analyses on the same topic sometimes disagree?

They often used different criteria for which studies to include, covered different time periods, or handled disagreement between studies differently. Small decisions about inclusion and weighting can shift the pooled answer more than people expect.

How many studies does a meta-analysis need to be trustworthy?

There is no fixed number, since a handful of large, consistent, well matched trials can be more convincing than twenty small, mismatched ones. What matters more than the count is whether the included studies are genuinely comparable and whether the field's full body of evidence, not just the published slice of it, is represented.

Read next

• Observational Study vs Experiment: Why Diet Advice Flip-Flops

  Most nutrition headlines come from observational studies that track habits, not experiments that test them, which is why the advice keeps changing.

• What a Randomised Controlled Trial Actually Buys You

  A randomised controlled trial sorts people into groups by chance, which cancels out hidden differences and lets researchers credit any gap to the treatment.

• How P-Hacking Happens Without Anyone Lying

  P-hacking happens when researchers make ordinary, well-intentioned analysis choices that quietly raise the odds of a false positive, with no lie required.