# What changed between the studies in the pooled result?

Draft • Research checked 4 October 2026 • AI-assisted DomDNA editorial content. No independent clinical review or publication is claimed. General education, not individual medical advice.

A meta-analysis can turn several study estimates into one prominent number. That number may be useful, but it does not make the studies identical.

Participants, comparison groups, measurements, follow-up periods and implementation can differ. When reading a pooled result, the useful question is what was combined and whether the combined answer matches the claim you are considering.

## Start with the ingredients

Imagine a hypothetical review of reminder messages. One study measures appointment attendance after a text. Another measures completion of a questionnaire after an email. A third measures daily app use after repeated notifications.

All involve reminders, but the outcomes and settings differ. Combining them into “reminders improve engagement” can conceal practical distinctions. A result about returning a questionnaire does not directly answer whether patients attend appointments.

These are invented studies. They illustrate why a shared topic label is not sufficient evidence that estimates should answer one common question.

[Cochrane's analysis guidance](https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-10) discusses study differences and heterogeneity when synthesising results. The design of the synthesis matters as much as the existence of several studies.

## Statistical heterogeneity is one part of the picture

Researchers use statistics to describe variation in study estimates beyond what a sampling model would expect. The [original Higgins and Thompson methods paper](https://pubmed.ncbi.nlm.nih.gov/12111919/?dopt=Abstract) concerns quantifying that heterogeneity.

A statistic such as I-squared is not the percentage of studies that are wrong. It does not directly tell you the proportion of participants who benefited, and it cannot replace reading the study characteristics.

Conversely, a low heterogeneity statistic does not prove that all studies asked the same clinically relevant question. Similar-looking estimates can come from different measurements, and limited information can make heterogeneity difficult to assess.

## A pooled average can hide useful differences

Suppose the fictional review includes reminders delivered once and reminders delivered repeatedly. Perhaps the average estimate is favourable. You still need to know whether the average meaningfully describes either implementation.

Looking at subgroups may help explore a difference, but a subgroup pattern is not automatically a confirmed explanation. It can be influenced by other differences between studies, the number of comparisons examined and the amount of evidence available.

A random-effects model allows for variation in underlying effects under its assumptions. It does not repair biased studies, make incompatible outcomes interchangeable or guarantee that the average applies to a new setting.

These distinctions are especially important when a promotional summary presents the largest pooled number without the uncertainty or the range of included methods.

## Reporting makes the synthesis inspectable

The [PRISMA 2020 statement](https://www.bmj.com/content/372/bmj.n71) supports transparent reporting of how reviews identify, select and synthesise evidence. A reported checklist is not a certificate that every design decision was appropriate.

For a reader, the search strategy, inclusion criteria and study table are useful because they reveal the boundaries of the result. A review of short trials in one population should not silently become a promise about long-term outcomes elsewhere.

Also check whether the authors pooled the studies at all. A systematic review can present a structured narrative when a single summary estimate would be misleading. The absence of a meta-analysis is not automatically a weakness.

## What you can do with this

Before copying the pooled estimate, choose two included studies and compare their population, outcome and follow-up. This small exercise can reveal whether the headline hides a substantial change in the question.

Then write the pooled result with its scope attached. For the fictional reminders review, a careful note might say that the synthesis covered several forms of recorded engagement, rather than claiming a specific improvement in appointment attendance.

If the accessible summary lacks the study table or detailed methods, record that limit. Do not infer uniformity from the tidy appearance of a forest plot.

The next useful study may be one that closely matches your question, not necessarily another broad synthesis. A pooled estimate offers a way to organise evidence. Its practical meaning still depends on the studies that went into it.

Reading those differences protects against both extremes: dismissing a review because its studies vary, or trusting a single average as though the variation disappeared.

For an educational starting point, the [DomDNA lifestyle quiz](https://domdna.com/quiz) records lifestyle answers, not DNA analysis or a clinical diagnosis.