How Do You Know Whether Two Papers Use the Same Participants?

Are These Papers Really Based on Independent Groups of People?

You have established that two papers are distinct publications, but another question remains: did they analyze different people?

This can be surprisingly difficult to determine. Researchers may publish several analyses from the same trial, cohort, survey, registry, administrative database, school sample, or other participant pool. A later paper may use the full original sample, a subgroup, participants with complete data, or people who remained at follow-up. Two papers can therefore have different sample sizes without representing independent populations.

The distinction matters whenever your interpretation assumes independence. If the same individuals contribute data more than once, treating those observations as though they came from separate groups can give that participant pool disproportionate influence in an evidence synthesis.

Reconstruct Where the Participants Came From

How to Trace Participants Across Publications

Same Study and Same Participants Are Related but Different Questions

Determining whether publications arise from the same study and determining whether they analyze the same participants are closely related tasks, but they are not interchangeable.

A single longitudinal study, for example, might enroll 1,000 participants at baseline. One paper could analyze all 1,000, another could examine 400 participants with a particular characteristic, and a later follow-up might contain 720 of the original participants. The publications belong to the same broader study, yet their analytic samples are not identical.

Conversely, separate analyses or nominally separate studies can draw from the same registry, administrative database, cohort, or other source population. The relevant question is therefore participant provenance: where did these observations actually come from?

If you first need to establish the relationship between the investigations themselves, compare the characteristics used to identify papers from the same underlying study.

Start With Study and Dataset Identifiers

A shared trial registration number, cohort name, study acronym, registry identifier, protocol, ethics identifier, grant number, or named dataset can provide strong evidence that the participant pools are related.

These identifiers do not always prove that the analytic samples are identical. Two papers using the same cohort may select different waves, age groups, geographic areas, or subsets. A shared identifier tells you where to look next.

Compare Recruitment Sites and Source Populations

Ask where the participants came from. Were they recruited from the same hospitals, schools, universities, communities, clinics, registries, panels, databases, or geographic areas?

The specificity of the match matters. Two studies both described as being conducted in the United States provide little evidence of participant overlap. Two papers drawing patients from the same three specialist clinics during the same period deserve much closer examination.

For database studies, identify the actual data source rather than stopping at a generic description such as “electronic health records.” The same named national registry, insurance database, longitudinal cohort, or institutional repository can support many publications.

Recruitment Dates Are Often Crucial

Compare the periods during which participants entered the study or dataset. Identical or substantially overlapping recruitment windows, combined with the same setting and eligibility criteria, can be a strong signal of shared participants.

Suppose one paper includes patients recruited at Hospital A from January 2020 through December 2022, while another uses patients from the same hospital from January through December 2021. The second population could be contained within the first. That does not prove overlap, but it creates a plausible mechanism for it.

Non-overlapping recruitment periods, by contrast, can sometimes establish independence even when the research team, institution, and procedures are the same.

Do Not Compare Only the Final Sample Size

The number appearing in a paper's results section may represent only the final analytic sample. Researchers may begin with a larger cohort and exclude participants because of missing data, eligibility restrictions, loss to follow-up, availability of specimens, completion of a particular instrument, or requirements of a secondary analysis.

Trace the sample backward when possible:

Two papers reporting 800 and 620 participants could therefore have complete overlap for those 620 people, partial overlap, or no overlap at all. The numbers alone cannot tell you which.

Compare Baseline Characteristics as a Participant Fingerprint

When explicit identifiers are unavailable, baseline characteristics can provide useful corroborating evidence. Compare age, sex or gender distribution, group sizes, diagnoses, disease severity, socioeconomic characteristics, geographic distribution, baseline outcome values, and other variables relevant to the population.

Highly similar values become more informative when several distinctive characteristics match simultaneously. Still, baseline characteristics are aggregate summaries, not individual identifiers. Similar populations can produce similar descriptive statistics by chance or design.

Read the Methods for Subset Language

Authors often reveal the relationship indirectly. Look for statements indicating that the analysis used a subset, subsample, secondary analysis, nested sample, follow-up cohort, respondents with complete data, biomarker subsample, or participants from selected sites.

Phrases such as “participants were drawn from,” “among participants enrolled in,” or “we analyzed data from” can be more informative than the title or abstract. Follow any citation attached to the description of the parent study.

Recognize Full, Partial, and Uncertain Overlap

Participant overlap is not an all-or-nothing phenomenon.

The distinction between these relationships becomes particularly important when evaluating an overlapping study population.

Shared Databases Require Particular Caution

Large cohorts, registries, electronic health records, claims databases, and public datasets can generate many publications. Papers based on the same dataset should not automatically be considered duplicate studies, because researchers may ask different questions and construct different samples. Yet neither should they automatically be treated as independent populations.

Published methodological work on evidence synthesis has highlighted participant double-counting when multiple studies use the same database or overlapping analytic periods. The practical problem is especially difficult when publications provide insufficient detail to reconstruct the exact overlap.

When the same named dataset appears repeatedly, record the dataset, sampling frame, dates, inclusion criteria, exclusions, geographic restrictions, and analytic sample for each paper. This also helps identify when one dataset produces many papers that appear to be independent evidence.

Sometimes You Cannot Determine the Exact Number of Shared Participants

Published reports usually provide aggregate information. Unless authors explicitly describe the overlap or provide identifiers that permit linkage, you may be able to conclude only that overlap is probable or possible rather than calculate exactly how many individuals are shared.

Do not convert suspicion into an invented percentage. Record what can be established, what remains uncertain, and why. If the distinction materially affects the synthesis, additional reports, protocols, registries, supplementary materials, or correspondence with investigators may clarify the relationship.

Working Out Whether Two Samples Actually Overlap

Common Mistakes When Assessing Participant Overlap

Map Participant Provenance Before Assuming Independence

For every potentially related paper, record enough information to reconstruct the sample: parent study or dataset, sites, recruitment period, original cohort size, inclusion and exclusion criteria, analytic sample size, follow-up wave, and any subgroup restrictions.

This information should remain attached to the study records during synthesis. Cochrane guidance emphasizes that multiple reports from the same study need to be linked and that multiplicity arising from the same participants creates statistical dependence. The exact analytical response depends on the synthesis being performed, but recognizing the dependence comes first.

The same reasoning becomes especially important when follow-up papers use only part of an original sample or when you need to use companion papers without double-counting participants.

What to Check Before Calling Two Samples Independent

Questions About Shared Participants Across Papers

Follow the People, Not Just the Papers

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.