When Your Published Paper Comes Under Scrutiny: A Practical Guide for Authors

The email arrives with the usual journal header. The editor-in-chief writes politely to say that a question has been raised about the gel panels in Figure 3 of your 2023 paper in the Journal of Infectious Diseases. They would appreciate your response, along with the original unprocessed images, within three weeks. Or you check PubPeer and find that an anonymous commenter has posted side-by-side screenshots of two figures from different studies on your lab's publication list, asking whether a specific region appears in both. Or an associate editor at a hematology journal writes to say that a reviewer of your current submission noticed an overlap with a paper you published two years earlier and would like a clarification before the review continues.

None of these situations automatically means your paper will be retracted, your career damaged, or your institution notified of misconduct. The majority of post-publication integrity concerns resolve through clarification, documentation, and in many cases a minor correction. But how you respond in the first 48 to 72 hours, and whether you understand the process you are now inside, shapes the trajectory more than anything else. Graduate training in medicine and biomedical research teaches experimental design, statistical methods, and clinical reasoning. It rarely prepares researchers for the moment when something in the published record is questioned.

This guide covers the full arc: how concerns reach journals, what the three possible editorial outcomes mean, what changed when COPE updated its retraction guidelines in September 2025, what the investigation process looks like from the author's side, and what practical decisions you can make now to reduce the risk that a small problem becomes a larger one.

How Concerns About Published Papers Actually Get Raised

The most common trigger is not a formal misconduct investigation. It is a reader noticing something. Peer review covers the period from submission to acceptance. After that, your paper sits in public view indefinitely, read by colleagues, competitors, methodologists, statisticians, and, increasingly, by automated systems running similarity checks on figure panels and pixel arrays. The window during which errors or anomalies can surface does not close at publication. For high-visibility papers, it never really closes at all.

PubPeer is the best-known venue for these observations. Founded in 2012, the platform allows researchers to attach comments to any published paper identified by its DOI or PubMed ID. Comments can be posted anonymously or under a real name and include attached images, with moderation focused on scientific content rather than personal criticism. A substantial share of PubPeer discussion concerns apparent image duplication or figure manipulation alongside statistical anomalies, inconsistent reported values, and questions about data provenance. When a comment is posted, corresponding authors receive an email notification using the address listed in the published paper. Researchers who have changed institutions and whose contact details in older papers are outdated sometimes miss these notifications entirely and learn about the comment from a colleague or a literature search.

Beyond PubPeer, post-publication concerns reach editors through several routes. Direct emails from readers or competing researchers. Automated screening by publishers during post-publication audit cycles (Springer Nature, Wiley, and Elsevier all run image-integrity software on recently published papers, and periodically on back-catalog articles as well). Institutional research-integrity offices that initiate contact with journals after a student or staff member raises a concern internally. Occasionally, a statistical anomaly is flagged by an external researcher running a formal analysis of the kind that Statcheck produces for numerical reporting, or that analogous tools are beginning to produce for clinical literature.

Most post-publication concerns do not reach a formal investigation. The majority resolve when the corresponding author provides original data files, explains a legitimate artifact of figure preparation, or agrees to submit a correction. The cases that escalate are generally the ones where the initial journal contact goes unanswered, where the explanation is inconsistent with the available documentation, or where the concern touches something that cannot be resolved with the data the authors can produce.

Correction, Expression of Concern, Retraction: Understanding Each Outcome

When an editorial team decides a published paper needs action, they have three instruments. Understanding what each one means, and what it does not mean, matters both for interpreting the signal and for deciding how to respond.

A correction addresses a bounded error that does not materially undermine the paper's conclusions. The term "corrigendum" conventionally signals that the error originated with the authors; "erratum" signals a publisher-introduced error. Examples include a table header labeled incorrectly, a unit stated wrongly in a supplementary figure, or a statistical value transcribed incorrectly in the results text while the underlying analysis was correct. Corrections do not imply misconduct. They are routine at every journal and in every field. The major publishers link the correction to the original article and keep both in the online record. Receiving a request to issue a correction is not an accusation. It is a request to fix something that is wrong.

An expression of concern is an interim signal, not a final verdict. COPE defines it as appropriate when there are major and credible concerns about the reliability of a publication but either the evidence does not yet meet the bar for retraction or an investigation is still underway. It may be the final recorded outcome for an article that will not be updated further, or it may serve as a placeholder until a correction or retraction is issued. If you receive notification that an expression of concern has been or is about to be published about your paper, an investigation is likely in progress or imminent. This is the moment for a careful, factual response, not a defensive one.

Retraction is reserved for papers where the conclusions cannot be trusted, where the integrity of the data is sufficiently in doubt that the record must be marked. A retracted paper is not deleted from PubMed, Scopus, or the journal's own site. It remains indexed and accessible with a visible retraction notice, so researchers who have cited it can update their understanding. The notice is supposed to state the reason for retraction clearly, whether the retraction was agreed by the authors or issued over their objection, and whether it resulted from error or misconduct. These distinctions matter. A retraction for honest error is not the same as one for data fabrication, and both the journal and the secondary record (Retraction Watch, PubMed notices, institutional records) are supposed to reflect that.

COPE's September 2025 Update: What Changed and Why It Matters

COPE replaced its 2019 retraction guidelines with a substantially revised version in September 2025. The update was reported by Retraction Watch at the time of release and was framed explicitly as a response to problems that had become far more common in the six years since the previous version: industrial-scale paper-mill fraud, batch retractions involving hundreds of papers originating from a single source, and cases where the fraud was committed by a third party (a commercial agency hired to prepare manuscripts) without the named authors' knowledge or, in some cases, with authors who were entirely fictitious.

For authors, the most practically significant change is the updated guidance on third-party misconduct. The 2025 revision acknowledges that retraction may be appropriate even when the named authors did not personally commit fraud, if the paper cannot be trusted regardless of who was responsible. This is a meaningful shift from the older framing that concentrated primarily on author culpability. Authors who used commercial manuscript preparation services, even without fraudulent intent, now carry greater exposure under the current guidelines if the agency introduced manipulated data, fabricated citations, or recycled content from other sources. Journals are not required to investigate and exonerate each author individually before issuing a retraction when the paper itself cannot be verified.

The 2025 update also strengthened guidance on investigation timelines. The revised guidelines explicitly state that journals should not allow inquiries to proceed indefinitely, and that an expression of concern is the appropriate interim instrument when an investigation is taking longer than expected. This means an expression of concern appearing while an investigation continues does not indicate a conclusion has been reached. It indicates the journal is proceeding according to current best practice.

COPE is a membership organization, not a regulatory body. Journals that are COPE members commit to following its guidance, and most major biomedical publishers are members: JAMA, the BMJ, The Lancet, the New England Journal of Medicine, PLOS, and the major journals under Elsevier, Springer Nature, and Wiley. When one of these journals opens an integrity inquiry, the COPE framework governs how the editor is expected to handle it, which gives authors a reasonably predictable set of steps to anticipate.

What the Investigation Looks Like from the Author's Side

When a journal opens a formal investigation, the process is more procedural than dramatic. You will typically receive a formal email from the editor-in-chief or an associate editor describing the specific concern, identifying which figures, tables, or passages are at issue, and requesting a response with supporting documentation by a stated deadline. The timeline varies by journal and by the complexity of the concern, but a first response is usually requested within two to four weeks.

What the editor is looking for in a first response is the original data and files underlying the specific items in question. Not a narrative defense. Not an explanation of how difficult the study was to conduct. Not a request to identify who raised the concern (COPE guidelines instruct journals to keep the identity of complainants confidential, and the editor may be genuinely unable to tell you). If you cannot produce the original unprocessed files, say so clearly and explain why: data were stored on a server that has since been decommissioned, a co-author who held primary custody has moved institutions, export-control rules at a collaborating hospital prevent remote access. An acknowledged, explained absence of original data reads very differently from an unexplained one.

If the concern involves co-authors, the investigation will typically contact them separately or ask the corresponding author to collect their responses. If there is a genuine disagreement among co-authors about what happened, say so explicitly rather than presenting a unified account that does not reflect the actual situation. Editors who handle many such inquiries are practiced at detecting inconsistencies across separately submitted responses.

The institution at which the work was conducted may be contacted separately by the journal, or the journal may defer to the institution's research-integrity process and wait for its findings. At major research universities and medical schools, the research integrity office will conduct its own interview-based process, which can take months. If your institution becomes involved, contact its legal counsel or research integrity guidance resources. They are not adversaries. Their job is to document the process correctly, which protects you as well as the institution, and to conduct the investigation in a way that is defensible if it is later challenged.

How to Write a Response to a Journal Integrity Inquiry

The tone that works best is matter-of-fact. You are providing a factual account of what happened, with documentation. Defensive language raises more questions than it answers. Aggressive language about the source of the concern is counterproductive. A lengthy explanation of your general reputation or your laboratory's productivity is irrelevant to the specific question the editor asked. So are attestations from colleagues who were not directly involved in producing the figures or data at issue.

A first response to a journal integrity inquiry should include the original data files or an honest explanation of their unavailability. It should describe the figure-preparation workflow that produced the contested image in enough detail that the editor can follow it independently. It should include any relevant correspondence with image-processing software, analysis logs, or co-authors. And it should directly answer each specific question the editor asked, not a broader version of the question you would have preferred to answer.

If you identify an error in the process of writing your response, say so. A correction proposed by the author, before the editor reaches the same conclusion, almost always results in a less severe outcome than the same correction extracted through a prolonged investigation. This is not a negotiating tactic. It is how the process is designed to work. COPE guidance treats author-initiated corrections as a positive sign of research integrity, not as an admission that requires further penalty.

The Case for Moving First: Self-Reporting Your Own Errors

Many researchers who discover an error in their own published paper instinctively wait to see if anyone notices. This is understandable. It is also strategically wrong in most situations.

When an author contacts the journal unprompted to report an error and propose a correction, the editorial response is almost always to issue a corrigendum, note the author's transparency, and move on. The process typically takes two to four weeks. The correction is linked to the original article, both remain in the record, and the author's conduct is noted as cooperative.

When the same error surfaces through PubPeer or a reader complaint instead, the investigation that follows is more formal, takes longer, and requires the author to explain not just the error but also why they did not report it when they found it. The journal may have already moved to an expression of concern before the author has had a chance to respond. The difference in outcome between the two pathways, for the same underlying error, can be significant.

Self-reporting does not require certainty that something is wrong. If you are uncertain whether a figure panel was prepared correctly, contact the co-author who ran the relevant analysis and verify together. If you find that a reported value has a transcription error in the results text, write to the journal. The initial contact does not have to be adversarial or formal. Journals receive author-initiated correction requests routinely and handle the straightforward ones quietly, often without any notice that an inquiry was opened at all.

PubPeer: What Authors Need to Know Before Their Paper Appears There

PubPeer currently indexes over a million papers. A comment can be attached to your paper at any point after publication, including years after the fact. About 85 percent of PubPeer comments are posted anonymously. Author response rates are low, around 7 to 8 percent according to platform-level analyses. This is partly because many authors do not know the platform exists until they see their own paper flagged, and partly because the correct response is genuinely unclear the first time it happens.

There is no formal obligation to respond to a PubPeer comment. But research on how journals actually handle papers commented on PubPeer suggests that author responses, when factual and constructive, are associated with better outcomes than silence. Silence tends to be read as absence of a good explanation rather than as confidence in the original work. The platform has served as a significant early-warning channel for papers that were later retracted, with a February 2026 analysis of highly cited retracted articles (published on the BishopBlog research commentary site) finding that PubPeer comments frequently preceded formal editorial action by months.

If your paper receives a PubPeer comment, the first step is to read it carefully before deciding whether to respond publicly. If the concern is clearly based on a misunderstanding of your methodology, a brief public clarification serves the scientific record and you. If the concern touches something you need to investigate before you can answer honestly, contact the journal directly rather than posting a public statement that may need to be walked back later. Comments on PubPeer are permanent. Author responses are permanent. It is worth taking a day to think before posting.

One practical step worth taking before anything goes wrong: search for your own lab's papers on PubPeer periodically, especially any high-citation papers from the past five years. A comment may have been posted without your knowledge if your contact details in the published paper are outdated. Finding out proactively is better than discovering it through a journal inquiry.

What Stays in the Record Permanently

One property of the post-publication record that authors routinely underestimate is its permanence. A correction is linked to the original article and remains indexed alongside it. An expression of concern is published as a citable notice. A retraction notice appears next to the original article in PubMed and in the journal's own record. PubPeer comments do not disappear when a paper is corrected or retracted. Retraction Watch maintains a searchable database of retractions, organized by institution, author, and journal, that is widely read by hiring committees and grant reviewers.

This permanence is a feature of the system, not a failure of it. The scientific record is meant to be a running account of what was learned and what turned out to be wrong. A retraction notice is not an erasure. It is an annotation. The authors who emerge from integrity inquiries with the least reputational damage are those who treated them as documentation exercises, cooperated with the process, and let a correction stand when a correction was warranted.

The authors who come out worst are those who fought conclusions that were clearly correct, stonewalled documentation requests, or encouraged co-authors to provide inconsistent accounts. Editors and research integrity officers are experienced with all of these patterns, and the downstream notices describe them plainly.

Protecting Your Record Before Problems Arise

Most of the practical advice here concerns data management rather than legal strategy.

Keep original, unprocessed files for every published study for a minimum of ten years. Most funding agencies (NIH, MRC, Wellcome Trust, and most European funders) require data retention for at least five years after publication; leading journals and most research institutions recommend ten. Store the files in a format readable without proprietary software, and include enough metadata (date, software version, analyst initials, processing steps) that someone else could reconstruct what you did from the file alone.

Document figure-preparation decisions at the time you make them. If you adjusted contrast for a gel image because the background was uneven, note it in the working file. If you used a representative image selected from a larger panel, note which image was selected and why. The argument "this is standard practice in our field" is less effective in 2026 than it was a decade ago. Springer Nature's image integrity office has been issuing public guidance on acceptable image adjustments since 2023. PLOS journals have published detailed image-preparation policies since 2014. The BMJ requires authors to confirm at submission that images have not been altered beyond what those policies permit.

Read the journal's author guidelines before submission, including the section on corrections and post-publication ethics. Most instructions for authors include a paragraph describing how the journal handles integrity concerns. Knowing the process before you need it is the kind of preparation that makes a difficult situation manageable rather than alarming.

Publication does not end your responsibilities to a paper. It extends them. A paper that has been read by thousands of researchers, cited in clinical guidelines, or used to support a funding application carries obligations that persist long after the acceptance email. The post-publication period is when the scientific community does its most thorough reading. Being prepared for that scrutiny is not paranoia. It is professionalism.

Further Reading