On August 13, 2026, Retraction Watch reported that JAMA Internal Medicine has declined to retract a 2000 study of around-the-clock, controlled-release oxycodone therapy for osteoarthritis-related pain — even after one of the study’s own co-investigators briefly joined a formal retraction request, and after litigation records showed the study’s sponsor, Purdue Frederick Company, knew of unreported opioid-withdrawal symptoms among trial participants before the paper was published.
The case is a useful, concrete illustration of a distinction research-integrity officers and clinical-research compliance staff already work with in the abstract: a journal’s retraction bar is built around fabrication and falsification, and a sponsor withholding relevant safety data from investigators and readers does not automatically clear that bar — even when it looks, to almost everyone else involved, like exactly the kind of thing a retraction exists to fix.
What the 2000 study said
The study, led by Sanford H. Roth with co-investigators including Roy Fleischmann and Barry Bockow, enrolled 133 patients and was published in Archives of Internal Medicine in 2000 (the journal that became JAMA Internal Medicine in 2013; DOI 10.1001/archinte.160.6.853). Per Retraction Watch’s reporting, the published paper stated that “withdrawal syndrome was not reported as an adverse event for any patient during scheduled respites” — a conclusion that minimized opioid-withdrawal risk in a trial designed to support oxycodone’s use for a common, non-cancer chronic-pain indication.
What surfaced through litigation, 19 years later
According to court settlement documents from Purdue Frederick’s 2007 federal criminal case, the company knew that eight trial participants had possible opioid-withdrawal symptoms a year before the paper was published — data that, on Retraction Watch’s account, was not reflected in the published conclusions. The documents came to light not through a journal audit or an institutional inquiry, but because Andrew Kolodny, medical director of the Opioid Policy Research Collaborative at Brandeis University, found them while researching opioid-related misinformation for a federally funded project and recognized their relevance to a still-cited, still-uncorrected paper.
The retraction request, and JAMA’s decision
Kolodny petitioned the journal for retraction in October 2025, then followed up with supporting documentation in December 2025. On July 29, 2026, a letter signed by eight clinicians pressed the case further; two of the study’s original co-investigators, Fleischmann and Bockow, were initially among the signatories. Fleischmann has since said publicly that he no longer supports retraction, stating the manuscript contains “no falsified data.”
JAMA Internal Medicine held its ground on that same basis. In a statement reported by Retraction Watch, the journal said: “We saw no evidence of fabrication or falsification in the published article.” Editor Annette Flanagin added that the journal would not proceed “without agreement of all of the surviving coauthors.”
Why “falsification” is such a narrow trigger
Retraction criteria used across scholarly publishing — codified most influentially in COPE’s retraction guidelines — center on a specific, author-attributable set of failures: data fabrication, falsification, plagiarism, duplicate publication, or research conducted without proper ethical approval. Those categories were built to answer “did the people who wrote this paper lie about what they did or found,” and they answer that question reasonably well.
What they answer far less cleanly is a different question: what happens when the people who funded the study held safety-relevant data that the people who wrote it apparently did not see, and that data never made it into the published record? If the trial’s investigators genuinely were not shown the withdrawal-symptom findings Purdue Frederick held, then no individual author falsified anything — the incompleteness originated with the sponsor, not the authorship. That is a real, and much greyer, category: not a hard “someone lied,” but a softer, harder-to-act-on “the data underlying this conclusion was less complete than the sponsor’s own records show it should have been.” A retraction bar built around author misconduct has comparatively little to say about sponsor-level omission, however consequential the omission turns out to be.
This is also why an outright retraction is not the only mechanism available to a journal facing a case like this — NISO CREC also covers expressions of concern for exactly the situation where evidence is credible but incomplete or contested rather than conclusively fabricated. JAMA’s public reasoning here addresses only why it declined the harder remedy; it does not resolve what, if anything, a softer one would require.
What this means for research administration and research-integrity offices
For institutions, the practical lesson is not “wait for the journal to act” — it is that a journal’s retraction decision answers a narrower question than an institution’s own obligations do. A few implications worth carrying into practice:
- A declined retraction is not a clean bill of health. “No evidence of fabrication or falsification” is a statement about the authorship’s conduct, not a finding that the published conclusions were adequately supported by everything the sponsor knew. Institutions citing, teaching from, or building on a paper flagged this way should treat the underlying safety conclusion as contested even where the retraction record stays clean.
- Sponsor data obligations exist upstream of authorship for a reason. ICH E6(R2) Section 5 and the broader modern architecture around trial transparency (results reporting to public registries, structured data-sharing agreements, safety-signal escalation duties) exist precisely to prevent a sponsor from holding safety-relevant data that never reaches the investigators writing up the results. A 2000-era trial predating most of that architecture is a reminder of what it was built to close, not evidence the gap is fully closed today.
- Sponsor conflict of interest in industry-funded research is a disclosure and governance problem as much as an authorship one. Research-integrity offices reviewing sponsored trials should be asking not only whether authors disclosed financial ties, but whether the data-sharing terms between sponsor and investigators gave the authors real visibility into everything the sponsor collected — adverse events included.
- An author’s own request for retraction is informative, not dispositive. This case shows both directions of that: a co-author’s initial support for retraction did not bind the journal to act, and his later reversal became part of the journal’s stated reason not to. Institutions should not treat a journal’s retraction decision as automatically tracking the balance of author opinion on their own paper.
- The gap between “falsification” and “incomplete due to what the sponsor withheld” is exactly the kind of finding a research-integrity office may need to pursue on its own footing — through its conflict-of-interest and data-integrity processes — even when a journal’s retraction criteria, applied narrowly and correctly to the authorship’s own conduct, come out negative.
For more on how editors actually work through a retraction decision once a credible allegation lands, see CASRAI’s COPE Flowcharts Explained, and for the mechanics of how a decision like JAMA’s gets recorded and surfaced to future readers, see Retraction vs. Redaction.
Frequently asked questions
Does a journal declining to retract a paper mean the underlying misconduct allegation was unfounded?
No. It means the specific evidence presented did not meet that journal’s retraction criteria as the editors read them — typically a finding of fabrication, falsification, plagiarism, duplicate publication, or a fundamental ethical-approval failure. In this case, JAMA’s stated basis was narrowly about the authorship’s conduct (“no evidence of fabrication or falsification”), not a broader finding about whether the sponsor’s undisclosed data would have changed the paper’s conclusions.
Who decides whether a paper gets retracted — the authors, the journal, or the institution?
The journal makes the final call, but editors weigh input from all of these: the original authors (individually and collectively, since COPE guidance generally expects agreement or at least notification of all surviving co-authors), any institutional investigation findings, and outside allegations. As this case shows, co-author opinion can shift during the process and the journal is not bound to follow it in either direction.
What can a research-integrity office do if a journal declines to retract a paper it has concerns about?
Options outside the journal’s own process include requesting an expression of concern as a lesser step than retraction, pursuing an internal conflict-of-interest or data-integrity review independent of the journal’s decision, and, where sponsor conduct in an active clinical relationship is at issue, escalating through the sponsor-oversight and data-sharing terms already in place for current trials rather than relying on the publication record alone.
Source: Retraction Watch, “JAMA journal won’t retract oxycodone paper despite author request, citing no ‘falsification'” (August 13, 2026).
