Did You Accidentally Count the Same Participants More Than Once?

Are Those Really Two Studies, or the Same Participants Appearing Twice?

You identify two eligible papers. They have different titles, different publication years, and perhaps even somewhat different author lists. It is tempting to enter them as two separate studies.

But both papers may report data from the same underlying participants.

A single study can generate a protocol, conference abstract, primary results paper, secondary analysis, subgroup analysis, long-term follow-up, and several outcome-specific publications. Different studies can also draw overlapping samples from the same cohort or administrative database. If those reports are treated as independent evidence when they are not, some participants can influence your synthesis more than once.

This is more than a housekeeping problem. Double-counting can distort sample sizes, precision, effect estimates, and the apparent amount of independent evidence.

Your Unit of Evidence Is Usually the Study, Not the Paper

Why One Sample Can Produce Many Apparently Different Pieces of Evidence

A publication is not necessarily a study

Systematic searches retrieve reports, but reviews generally seek evidence from underlying studies. Cochrane explicitly distinguishes the two: one study may be described in several articles, abstracts, registry entries, or other reports, and those reports need to be linked so that the study rather than the publication becomes the unit of interest.

This distinction sounds straightforward until you encounter real publication histories. One article may report the primary outcome, another the secondary outcomes, another a particular subgroup, and another longer follow-up. All four papers can be useful without representing four independent participant samples.

Removing duplicate database records is therefore only the first step. Reference-management software may identify two identical citations, but it cannot safely assume that two differently titled papers are independent studies.

Multiple reports are not always obvious duplicates

Reports from the same study may not share identical author lists, sample sizes, outcomes, or publication dates. Some may not clearly cross-reference one another. Cochrane recommends comparing features such as trial identifiers, authors, location, intervention details, participant numbers and baseline characteristics, and recruitment dates when trying to determine whether reports belong to the same study.

Do not throw away secondary reports once you identify them

Discovering that two papers belong to one study does not mean keeping one and deleting the rest. Secondary reports may contain information absent from the primary publication, including additional outcomes, longer follow-up, subgroup information, adverse events, or methodological details.

Cochrane recommends collating reports belonging to the same study and using information across them. When reports provide conflicting results or information, reviewers need a transparent rule for deciding which source supplies particular data.

Overlapping participants can occur without duplicate publication

A more difficult situation arises when two genuinely different studies or analyses include some of the same people. This can occur when researchers repeatedly analyze a long-running cohort, national survey, registry, electronic health-record system, claims database, school system, or other shared data source.

Imagine one paper analyzing participants from a cohort between 2018 and 2021 and another using the same cohort between 2020 and 2023. They are not necessarily duplicate publications, but their participant samples may overlap. Treating the effect estimates as completely independent can therefore overstate how much independent information you have.

Check cohort names, source databases, institutions, recruitment periods, eligibility criteria, geographic coverage, and sample characteristics. If overlap remains uncertain and could materially affect the synthesis, contacting authors may be necessary.

Multiple outcomes from the same participants are also dependent

Multiplicity can occur within a single study. Researchers may report several measures of the same outcome, multiple time points, alternative analyses, or several related outcomes based on the same participants. These effect estimates are statistically dependent because they come from the same people. Cochrane recommends prespecifying how such multiplicity will be handled or using synthesis methods that account for the dependency.

Choosing whichever measure produces the strongest result after seeing the data creates another problem: selective inclusion. Decisions about which eligible result represents a study should therefore be based on defensible rules that do not depend on the direction or statistical significance of the result.

Multi-arm studies create a classic double-counting problem

Suppose a trial compares two intervention groups with one shared control group. If you enter Intervention A versus Control and Intervention B versus Control as though they were completely independent comparisons in the same conventional meta-analysis, the control participants appear twice.

Cochrane identifies this as a unit-of-analysis error because the comparisons are correlated through the shared group. Possible solutions depend on the synthesis and may include combining relevant groups, splitting a shared group appropriately, using separate comparisons, or applying methods that explicitly account for the dependency.

Overviews of reviews can double-count entire primary studies

The same problem moves up one level when synthesizing systematic reviews. Two included reviews may contain many of the same primary studies. Treating both reviews as wholly independent evidence can repeatedly count the same underlying study data and give some evidence disproportionate influence. Cochrane identifies overlapping primary studies as a specific methodological issue in overviews of reviews.

This is why the number of reviews, papers, or effect estimates is not automatically the number of independent pieces of evidence.

Double-counting can create false confidence

Repeated observations do not create the same information as genuinely independent observations. When dependency is ignored, calculated precision can be exaggerated and some studies or participant groups can exert disproportionate influence.

Duplicate publication has historically been shown to bias meta-analytic findings when repeated data were inadvertently treated as independent. Cochrane consequently requires multiple reports of the same study to be collated rather than counted as separate studies.

When Four Papers Are Actually One Study

The four publications remain useful sources. What changes is their interpretation. They represent several windows onto one study rather than four independent replications.

Common Ways Participants Get Counted More Than Once

Track Studies and Participants, Not Just Citations

As your review develops, maintain a distinction between reports and underlying studies. When several records appear related, investigate before assigning them separate study identities.

This issue becomes particularly easy to miss when searching registries, conference materials, and other non-journal sources, because a single study may appear in several forms. Wider searching is valuable, but every additional report needs to be linked back to the underlying study.

Before You Treat Every Paper as Independent Evidence

Questions About Duplicate Studies and Overlapping Samples

Count Studies and Independent Information, Not PDFs

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.