JAMA Internal Medicine won't retract a 2000 opioid study. Its stated test is narrower than COPE's.

JAMA Internal Medicine has again declined to retract a 26-year-old study on oxycodone for osteoarthritis pain. The refusal is not the interesting part. The reason the journal gave is, because it rests on a definition of what a retraction is for, and that definition is narrower than the one in the guidance the field generally works from.

What happened

In late July a group of eight clinicians, initially including two of the paper's own authors, wrote to editor in chief Sharon K. Inouye asking the journal to retract a 2000 study of 133 patients, published when the journal was called Archives of Internal Medicine. The paper concluded that around-the-clock controlled-release oxycodone was safe and effective for osteoarthritis pain, and that withdrawal symptoms were rare.

The petition rests on court records made public in 2007. The Purdue Frederick Company and three of its executives pleaded guilty that year to misbranding OxyContin, with the company agreeing to pay more than $600 million. Part of what the company had to acknowledge was that it had falsely claimed OxyContin was less likely to cause withdrawal symptoms than other pain medications, when no medical research supported the claim. The petitioners say the settlement documents show Purdue knew, a year before the paper appeared, that eight participants in this trial had possible opioid withdrawal symptoms.

Nothing about the sponsorship was concealed. The paper disclosed Purdue's funding and identified one coauthor as a company employee. The dispute is about data the sponsor did not hand over.

The journal has said no three times, in November 2025, in December 2025, and now. A spokesperson, Jen Zeis, told Retraction Watch, which reported the story on 13 August 2026: "We saw no evidence of fabrication or falsification in the published article." Executive managing editor Annette Flanagin wrote in an August 4 email that "without agreement of all of the surviving coauthors to retract the article, we cannot move forward." One of the two author signatories, rheumatologist Roy Fleischmann, has since withdrawn his support, telling Retraction Watch he "signed the letter in haste without looking at all the facts" and now believes the manuscript was accurate. The lead author, Sanford H. Roth, died in 2024.

Two tests, and the distance between them

The journal's position amounts to two conditions: evidence of fabrication or falsification, and agreement from every surviving coauthor. Both are worth holding against what the field's own guidance says, not because a journal is obliged to follow it, but because the gap is instructive.

On the threshold, the Committee on Publication Ethics retraction guidelines, updated to version 3 in September 2025, put it this way: "Editors can decide to retract a publication if they no longer have confidence in the results and conclusions reported in the paper." The same document states that unreliable content "can result from honest error, naive mistakes, or research or publication misconduct," and that the purpose of retraction "is to correct the literature and ensure its integrity, not to punish the authors." Nothing in that formulation requires proving anyone fabricated anything. It asks whether the conclusions can still be relied on.

On author agreement, the International Committee of Medical Journal Editors is explicit in its recommendations: "Ideally, the authors of the retraction should be the same as those of the article, but if they are unwilling or unable the editor may under certain circumstances accept retractions by other responsible persons, or the editor may be the sole author of the retraction or expression of concern." COPE's 2025 revision went further in the same direction, strengthening the emphasis on retraction being the editor's decision rather than the publisher's.

Analysis

Two different questions are being answered here, and they are being treated as one.

"Was this paper fraudulent?" and "can a reader still rely on this paper?" are separable, and the second is the one a researcher citing the work actually needs answered. A paper can be written in complete good faith from the data the authors were given, and still reach a conclusion that the withheld data contradict. Under a fabrication-or-falsification test, that paper stays. Under a confidence-in-the-conclusions test, it is at least arguable.

There is also a third option that the reporting does not mention being considered. Both COPE and ICMJE describe the expression of concern as the instrument for exactly this situation: serious doubt about a publication where retraction is not, or not yet, the right answer. It flags the paper to every future reader without requiring the journal to make a finding of misconduct or to secure unanimous author consent. Whatever the merits of the retraction argument, the case for leaving a paper of this history entirely unmarked is harder to make than the case against retracting it.

What makes this more than a procedural disagreement is that the paper is still in circulation. Retraction Watch reports it has been cited 247 times according to Clarivate Web of Science, and the petitioners note it appears in at least two clinical practice guidelines. A citation count is not just a scoreboard here. It is a measure of how far a conclusion has travelled since the record that complicates it became public in 2007.

What to watch

Whether JAMA Internal Medicine issues an expression of concern as an intermediate step, and whether the all-surviving-coauthors condition is a stated policy or a case-specific judgment. The distinction matters well beyond this paper: a unanimity requirement is, in practice, a veto handed to anyone with a reason to keep a finding standing.