Salami Slicing and Duplicate Publication Explained

Publishing several papers from one study is often legitimate. What turns it into a problem is overlap that editors and readers cannot see.

Explainer · 10 min read · Directive Publications Editorial Team

The terms that get mixed up

These words describe different problems, and the one that applies decides what an editor asks you to fix.

The ICMJE Recommendations define duplicate publication as "publication of a paper that overlaps substantially with one already published, without clear, visible reference to the previous publication." Note the last clause. Overlap itself is not the offence; overlap the reader cannot see is.

Why medical journals take it seriously

In clinical research, duplicate papers do more than waste reviewer time. The ICMJE section on overlapping publications warns that duplicate publication of original research "can result in inadvertent double-counting of data or inappropriate weighting of the results of a single study, which distorts the available evidence."

That distortion has been measured. In a 1997 BMJ case study, Tramèr and colleagues searched systematically for randomised trials of ondansetron for postoperative emesis. They found that 17% of the full trial reports and 28% of the patient data were duplicated, none with a clear cross-reference. Trials reporting a greater treatment effect were more likely to be duplicated, and including the duplicates in a meta-analysis overestimated the drug's antiemetic efficacy by 23%.

Salami slicing does the same damage more quietly. If a reviewer pooling studies cannot tell that three papers share one patient sample, that sample can be counted three times.

Where splitting a study crosses the line

Large studies legitimately produce more than one paper: a trial may report its primary outcome first and long-term follow-up later. The problem is splitting results that belong together, or presenting overlapping samples as independent.

Ask four questions about each planned paper:

1. Does it answer a distinct question? A separate hypothesis, not the same question applied to another outcome from the same participants at the same time point.
2. Would the findings mean more together? If a reader needs both papers to interpret either one, they belong in one paper.
3. Was the split planned? Outcomes prespecified in the protocol or registration are easier to defend than divisions made after the results came in.
4. Could a systematic reviewer tell? Each paper should cite the others, state that the participants overlap, and use the same registration number or dataset identifier.

A worked example. One randomised trial of a rehabilitation programme measures exercise capacity, quality of life and hospital readmission at the same follow-up visit. Three papers, one per outcome, each describing "a randomised trial" without mentioning the others, is salami slicing: the outcomes come from the same participants at the same time and are needed together to judge the programme. Reporting them together, then publishing a prespecified long-term follow-up that cites the first paper, is not.

COPE's flowcharts for redundant publication in a submitted manuscript give "salami publication" as their example of a failure to refer to previous papers that the editor believes was deliberate, and attach to it the step of considering whether to inform the authors' institutions. What triggers that is the missing cross-reference, not the number of papers.

Secondary papers that journals accept

These kinds of overlap are generally acceptable when handled openly. Policies vary, so check the target journal's instructions for authors.

• Secondary analyses of a trial. ICMJE asks authors to cite the primary publication, state clearly that the paper contains secondary analyses, and use the same trial registration number and a unique, persistent dataset identifier. If several papers were planned from the start, the original registration number can be used when all the outcomes were defined in the original registration.
• Extended follow-up and subgroup analyses. COPE's flowcharts group these with minor overlap and legitimate re-analysis rather than major redundancy. That is not a clearance: on that branch the editor still writes to the authors and asks for the missing reference to the original. Supply it yourself and the question never arises.
• A full paper after a preliminary report. ICMJE does not treat an earlier meeting abstract or poster, a letter to the editor, or a preprint as a bar to a later complete report. Tell the journal about each one.
• Results posted in a trial registry. ICMJE does not count this as prior publication when the posting is limited to a brief structured abstract or tables.
• Acceptable secondary publication, such as a translation for a different readership. Both editors must approve, the secondary version must faithfully reflect and cite the original, and its title must identify it as a secondary publication.
• Separate analyses of the same database. Papers from one dataset should add substantially to each other and cite earlier publications from it.

Self-plagiarism and recycled text

Self-plagiarism usually means reusing your own published wording without acknowledgement. The concern is text rather than data.

Reused methods descriptions are the common case, and COPE's flowcharts give methods as their example of legitimate text recycling, on the minor-overlap branch rather than with major redundancy. That is not approval: the minor branch still ends with the editor asking for a reference to the original and checking whether copyright permission is needed. Depending on the licence, the earlier wording may not be yours to reuse. Practical rules:

• Cite the earlier paper wherever you reuse its methods text ("as described previously").
• Write the introduction and discussion afresh. Recycled framing starts to suggest recycled findings.
• Never reuse results text, tables or figures without making the relationship explicit. That is data overlap, not text recycling.

Whether self-plagiarism counts as formal research misconduct depends on jurisdiction and institutional policy, but journals can act on it either way.

How editors spot overlap

Redundant publication usually comes to light through a few routes:

• Similarity and image checks. COPE's flowcharts note that redundancy may be detected by text-matching or image analysis software. Your own earlier papers appear in that report like anyone else's, and a high match against yourself prompts the editor to ask which sections overlap and why. Our guide to reading a similarity report explains what the numbers mean.
• Reviewers and readers. People in the field recognise a familiar cohort, setting or results table.
• Systematic reviewers. Comparing sample sizes, recruitment periods and sites across studies surfaces papers built on the same participants.

On the major-overlap branch, COPE's flowcharts describe evidence that authors have sought to hide redundancy: paraphrasing, changing the title, reordering, adding or removing authors, and not citing previous papers. Reordering there means author order, not shuffled sections. Each makes a paper harder to match to its predecessor, so an editor is unlikely to read them as cosmetic. If you have a real reason for one — an author genuinely joined the second study — say so before anyone asks.

What happens when overlap is found

COPE's flowcharts are a short decision tree, not a rulebook, but they show the shape of what follows. The editor writes to the corresponding author and copies in all co-authors. Where the overlap is major, that letter encloses the authors' signed statement that the work had not been published elsewhere, plus evidence of the duplication. The declaration you sign at submission is the document that comes back to you.

ICMJE is direct about it: authors who attempt duplicate publication without notification "should expect at least prompt rejection," and a published duplicate "might warrant retraction with or without the author's explanation or approval." If a notice is issued against your work, our guide to correcting the published record explains how corrections and retractions work.

How to disclose related work

Disclosure separates a legitimate secondary paper from a redundant one. ICMJE asks that the cover letter or submission form include "a full statement to the editor about all submissions and previous reports that might be regarded as redundant publication of the same or very similar work," that such work is referenced in the new paper, and that copies are sent with the submission.

1. Every related paper, preprint, abstract and manuscript under review is listed in the cover letter
2. Copies of related unpublished manuscripts are attached
3. The letter says what this paper adds that the others do not
4. Related publications are cited in the methods, with a statement that the participants overlap
5. The trial registration number or dataset identifier matches the related papers

A sentence you can adapt:

For the rest of the letter, see what editors look for in a cover letter.

The rule that covers every case

When you cannot decide whether two papers are too close, apply this test: write the sentence that would appear in a systematic review's table of included studies. If you can write "these two reports answer distinct questions in samples that overlap as follows" and it is true and checkable from the papers themselves, publish both and put that sentence in the methods. If writing it honestly would make one paper look redundant, you have your answer before an editor gives it to you. If the sentence is simply hard to write, that is the point to send both manuscripts and ask. An editor told about overlap in advance is deciding where a paper belongs; an editor who finds it afterwards is deciding whether it should stand.

The Directive Publications For Authors page gathers the author guidelines and publication ethics policies in one place, so you can check them before writing the disclosure section of your cover letter.