A year before the paper ran, the company that paid for it already had the number in hand. Purdue’s own internal analysis of its osteoarthritis study, later made public in a 2007 federal criminal case, listed 8 patients with recorded symptoms that “may possibly have been related to opioid withdrawal.” The study itself, published in 2000 in Archives of Internal Medicine, told readers the opposite: that around-the-clock controlled-release oxycodone was safe for arthritis pain, and that “withdrawal syndrome was not reported as an adverse event for any patient during scheduled respites.” Eight cases in the sponsor’s files. In print, withdrawal was “rare.”
In July 2026 a group of clinicians asked the journal, now JAMA Internal Medicine, to retract the paper. On August 13 the journal said no, and the reasoning it gave is narrower than the rule it operates under. Spokesperson Jen Zeis told Retraction Watch, “We saw no evidence of fabrication or falsification in the published article.” Annette Flanagin, JAMA’s executive managing editor, put it in writing: “There is no evidence of fabrication or falsification in this article that warrants retraction.”
That answer resolves a question the petition never asked. The retraction guidelines of the Committee on Publication Ethics, the standard the JAMA Network’s own journals cite, tell editors to retract when they have “clear evidence that the findings are unreliable.” COPE lists the causes in the alternative: unreliability can come from major error just as readily as from fabrication or falsification, because the purpose of retraction is to correct the literature, not to punish authors. Falsification is one road to a retraction. It was never the only one. By replying “no falsification,” the editors answered the easy question and left the one on the table.
Here is what the documents establish. The sponsor was disclosed on the paper’s face: “This study was sponsored by Purdue Pharma LP, Norwalk, Conn.” One of the eight listed authors, Peter G. Lacouture, was a Purdue employee. In May 2007 the Purdue Frederick Company pleaded guilty in the Western District of Virginia to fraudulently marketing OxyContin with intent to mislead, and it and its executives agreed to pay the government more than $600 million, a record reconstructed in detail through the company’s own documents by STAT. The withheld withdrawal count surfaced in that same criminal case. The marketing fraud is adjudicated. The gap between the internal tally and the printed one sits in the record that case produced.
What the documents do not settle is whether the academic co-authors knew the funder was holding a fuller count when the manuscript went out. That question is open, and under COPE it is also beside the point: a retraction turns on whether the published findings are reliable, not on whether every author is culpable. A safety claim contradicted by data the sponsor possessed before publication is the textbook definition of findings that are unreliable, whoever is to blame for the gap. The journal folded a question about the record into a question about intent, then answered the one that was easier to clear.
The journal offered a second reason: the authors do not agree. The petition, led by Andrew Kolodny of Brandeis University’s Opioid Policy Research Collaborative, was signed at first by two of the study’s own authors, Roy Fleischmann and Barry Bockow. Fleischmann then withdrew, saying he had “signed the letter in haste” and now considered the manuscript “100% accurate.” Absent that consensus, the journal declined to act. But author agreement appears nowhere in the COPE grounds. The decision rests with editors precisely so that a paper’s reliability does not become a vote among the people whose names are on it. Reading a single author’s reversal as decisive is a choice the guidelines do not require.
Follow the incentive on the journal’s side, because it is not hidden. The paper has been cited 247 times, by Clarivate’s Web of Science count. Retracting it now would not just flag one twenty-six-year-old article. It would concede that the flagship of American internal medicine kept a sponsor’s safety claim in print for years after the contradicting data became public, and it would invite the obvious question of what else in the opioid-era literature carries the same defect. A “no falsification” letter closes that door. Leaving the paper in place, with nothing attached, keeps the citation live.
A live citation is not a neutral thing. The paper remains fully citable, a study the journal still stands behind, reachable as standing evidence that around-the-clock oxycodone was a tolerable long-term therapy. That is the claim the sponsor was convicted of pushing by fraud, and the literature is the last place it stands unqualified. The journal is its custodian.
The records that would close the gap are not in a courthouse. They are on the editors’ desk: the 2007 plea, the internal withdrawal count, the paper’s own text, and the COPE guidance the JAMA Network cites. The editors have all four, and as of their August 13 answer they have posted no correction, no expression of concern, no editor’s note. The retraction request has been answered. The reliability question it raised has not. It waits with JAMA Internal Medicine, which has the documents, has the standard, and so far has declined to hold the second up to the first.
Sources
- Retraction Watch – JAMA journal won’t retract oxycodone paper despite author request, citing no ‘falsification’ (2026)
- Roth et al., “Around-the-Clock, Controlled-Release Oxycodone Therapy for Osteoarthritis-Related Pain,” Archives of Internal Medicine (2000)
- Committee on Publication Ethics – Retraction Guidelines
- STAT – The history of OxyContin, told through Purdue Pharma documents (2019)