When One Study Appears in Several Papers

Research evidence is organized into articles, but patients are enrolled in studies. That difference becomes important when one trial or cohort produces several publications; multiple reports can be scientifically useful, yet they can also make one dataset appear to provide several independent confirmations. The ethical and analytic question is not simply, "Have these authors written about the data before?" It is, "Can editors and readers see that the reports come from the same study, and does each report add a justified contribution?"

The unit-of-evidence problem#

Consider a randomized trial with one primary outcome and several secondary outcomes. The investigators publish the main results, a longer-term follow-up, a harms analysis, and a quality-of-life analysis. Four articles exist, but there is still one randomized allocation and one underlying participant group.

If a review treats those papers as four trials, the sample size is inflated and uncertainty looks smaller than it is, and if favorable outcomes are separated into visible papers while unfavorable findings remain buried or unpublished, the field's impression can shift further. The integrity problem arises from obscured dependence, selective emphasis, or needless repetition, not merely from the number of outputs.

This is why systematic reviews distinguish reports from studies. One study may have many reports; one report may even describe several studies. Evidence synthesis must reconstruct that relationship.

Duplicate and redundant publication#

The International Committee of Medical Journal Editors defines duplicate publication around substantial overlap with already published work and the absence of a clear, visible reference. "Redundant publication" is often used more broadly for an article that repeats data, results, or analysis without adding enough to justify another paper.

The practical harms are concrete:

• a meta-analysis may count the same participants more than once;
• readers may mistake repetition for replication;
• positive findings may receive disproportionate visibility;
• peer reviewers and editors spend effort evaluating work whose history was withheld;
• copyright or journal-license commitments may conflict.

Disclosure changes the process. Authors should alert an editor to closely related manuscripts and provide them for comparison, and the editor can then decide whether the new question, analysis, audience, and amount of overlap warrant publication. Hidden overlap removes that informed judgment.

Fragmentation and the "minimum publishable unit"#

Salami slicing describes dividing a coherent study into several narrow articles mainly to increase the number of publications, but the label is easy to apply and harder to define, because large studies often support legitimate secondary analyses.

A useful test is whether each paper has a distinct, prespecified question that can be understood without concealing the wider study. Separate reports may be appropriate when an outcomes paper, methods paper, economic evaluation, and long-term follow-up require different expertise and space. They become problematic when outcomes that belong together are separated, the same hypothesis is repeated with small variations, or each paper presents itself as an unrelated confirmation.

Fragmentation can hide tradeoffs. If benefits appear in one paper and harms in another, reading either alone leaves you with an incomplete clinical picture. It can also complicate corrections because an error in the shared dataset may affect an entire family of papers.

Text reuse is a different question#

Reusing data and reusing wording are related but not identical issues. A methods description may need to repeat technical language so readers can reproduce the work. Reusing substantial discussion, results, or interpretation while presenting them as new is more concerning.

Editors consider how much text is repeated, where it appears, whether the prior source is cited, and whether copyright permits reuse. A similarity score alone cannot judge scientific redundancy: two papers can use fresh wording while double-reporting the same result, and two legitimate reports can share a necessary methods description while answering different questions. For you as a reader, overlap in sample size, recruitment dates, institutions, interventions, registration numbers, and baseline characteristics is often more informative than overlap in prose.

What does not automatically count as duplicate publication#

Current editorial recommendations recognize several important boundaries.

Preprints: Posting a manuscript on a recognized preprint server is not automatically treated as duplicate journal publication, although the preprint should be disclosed and later citations should identify it as such.

Conference material: An abstract or poster is generally not a complete prior publication, but a detailed public release can affect an editor's assessment. Authors should be transparent.

Trial registries and regulatory reports: Structured results posted to meet registration or regulatory obligations serve transparency and are not simply a second journal article. The later manuscript should still tell you where the prior results can be found.

Acceptable secondary publication: A translation or republication for a different audience can be legitimate when both editors approve, the first publication has priority, the second faithfully reflects the original, and the title and notice make the relationship explicit.

Public-database analyses: Different groups may analyze the same open dataset using different methods. Shared data do not make the papers duplicates if the questions and analyses are meaningfully distinct and the source is clear. All of these categories share one feature: you are not misled about what is original, what is repeated, and why another report exists.

How systematic reviewers find linked reports#

A study family may reveal itself through a trial registration number, protocol citation, grant number, investigator names, recruitment window, site list, intervention details, and nearly identical baseline totals. No single marker is perfect. Author lists change, sample sizes differ after exclusions, and follow-up papers may use new titles.

Reviewers therefore build a study-level record. They link every report, compare outcomes and time points, and choose the most complete or appropriate source for each result. When two papers report the same outcome at the same time point, the value should enter a meta-analysis once, and when one reports short-term and another long-term results, both may contribute to separate analyses.

Uncertainty should be documented rather than guessed away. Contacting authors may clarify whether cohorts overlap. Sensitivity analyses can test whether including a questionable report changes the conclusion.

How readers can recognize a study family#

When several papers seem to support the same conclusion, look beyond titles.

1. Do the recruitment dates and locations match?
2. Are baseline participant counts and characteristics unusually similar?
3. Is the intervention described in the same distinctive way?
4. Do the papers share a registration number or protocol?
5. Does each paper cite the others and explain what is new?
6. Are different outcomes being reported from the same randomized groups?

Transparent reports make these links easy. Evasive reports change author order or terminology without acknowledging the common source. Lack of a cross-reference is not proof of intent, but it is a reason to map the underlying study more carefully.

Corrections should follow the dataset#

If an error affects shared data, correcting only one article may leave the literature inconsistent. Editors and institutions should identify related reports and assess each one. A retraction of one paper does not automatically invalidate every other paper from the study, but readers need to know whether the reason, such as fabricated records or a flawed randomization, reaches the common foundation. The same study-level view that prevents double counting also improves post-publication repair, because the record is more accurate when linked outputs are treated as a connected set.