The dashboard PubPeer posted on November 9 carries a figure the journals have never put in front of readers in this form: 20,784 papers drew comments on the post-publication review site before they were retracted, and the median wait between the first comment and the retraction was 10 months, according to Retraction Watch. In the case where the money at stake was largest, the people raising the questions spent part of that wait as defendants in a defamation suit.
The dashboard goes further. More than 2,500 papers carried comments a median of 11 months before an expression of concern, and about 12,000 were discussed on the site before a correction. The Retraction Watch Database has logged 59,843 retractions since PubPeer opened in October 2012, so by that rough yardstick about a third of the period’s retractions had a public thread first. The dashboard cannot say whether a comment set off any of them, because it records the order of events, and the reasons for each retraction sit in journal case files that the dashboard does not reach.
A preprint posted to MetaArXiv tried to get at causation by asking researchers directly. Wytske Hepkema of Radboud University and colleagues in the NanoBubbles group emailed nearly 100,000 people: more than 10,000 authors of papers with PubPeer comments and nearly 85,000 who had cited such papers. Fewer than 1 percent answered, leaving 80 authors of commented papers and 720 citing authors. Most in both groups said the comments did not affect their own study’s results. About four in five of the 80 knew their paper had been discussed, over half of the 80 said they had asked their journal for an erratum, and only 10 to 15 percent of those who knew about the comments had replied on PubPeer itself. More than half of the citing authors called the criticism irrelevant to their work. Hepkema said the team was most surprised by “the argument that even when they said the study they cited was fraudulent, that it had no consequences for their own article.”
The authors concede that 800 self-selected respondents out of nearly 100,000 cannot support general conclusions, and José Luis Ortega of Spain’s National Research Council warned that citing authors “could perceive that citing erroneous studies is something more acceptable or that it does not require any action.” PubPeer cofounder Boris Barbour suggested authors who stay silent may hope “the issue will be quietly forgotten.” Retraction Watch disclosed that its cofounder Ivan Oransky sits as a volunteer on the PubPeer Foundation’s board.
A survey can only report what researchers say about the comments. Cassava Sciences offers the comparison the survey could not make, because there the questions were tested by a journal, a university, a regulator and two federal courts.
The Cassava file
In August 2021, Labaton Sucharow, a law firm known for shareholder suits, filed a citizen petition asking the FDA to halt trials of simufilam, Cassava’s Alzheimer’s drug, Retraction Watch reported at the time. The petition cited “many dozens of examples of protein bands that appear to have been duplicated and/or misrepresented” in papers by Hoau-Yan Wang, a City University of New York neuroscientist and Cassava adviser, and Lindsay Burns, a Cassava executive, and alleged that the researchers had “misrepresented preclinical and clinical research results for more than 15 years.” Once the letter became public, PubPeer commenters, the image analyst Elisabeth Bik among them, began posting on the papers. Cassava called the allegations “false and misleading.”
- 2021FDA petition filed, PubPeer threads open
- 2022PLOS ONE retracts five Wang papers
- 2023CUNY committee report leaks
- 2024SEC settlement and federal indictment
- 2025indictment dismissed with prejudice
The money ran in both directions. The critics included short sellers, who profit when a stock falls; Reuters coverage of their later lawsuit describes them as doctors and short-sellers. Cassava’s shares had risen more than 20-fold, to above $146, between January and July 2021. Taxpayers had a stake too: federal prosecutors would allege that Wang obtained about $16 million in National Institutes of Health funding through grant applications containing “materially false, fraudulent, and misleading statements.”
The journal moved first. In March 2022, seven months after the threads opened and faster than PubPeer’s median, PLOS ONE retracted five Wang papers, stating that “pixel patterns in background areas of blot images” across multiple panels “appear more similar than would be expected for data obtained in independent experiments.” Wang attributed the repeated background features to “scanner artifacts.” The notice records that he “did not agree with the retraction and stands by the article’s findings.” Burns was a coauthor on two of the five.
Federal regulators came next. On September 26, 2024, the Securities and Exchange Commission announced settled charges alleging that Wang, partly unblinded to Phase 2 trial data, used that information to identify about a third of enrolled patients and manipulate results so the drug appeared to sharply improve total and phosphorylated tau. The agency alleged that Cassava reported a subset of data Burns selected by hand and did not disclose that the full dataset showed no measurable improvement in episodic memory. The settlement terms were a $40 million civil penalty for Cassava, $175,000 and a three-year officer-and-director bar for founder Remi Barbier, and $85,000 and a five-year bar for Burns, with a $50,000 penalty for Wang. None admitted or denied the allegations. Simufilam failed its Phase 3 trial late that year, and Cassava said it would end the program by June 30, 2025, with its shares by then trading below $2.
While those findings piled up, Cassava pursued its critics in court. It sued Adrian Heilbut, Jesse Brodkin, Enea Milioris, David Bredt and Geoffrey Pitt for defamation, then dropped the case in August 2024, weeks after Wang’s indictment. The five countersued for malicious prosecution, and in March 2025 U.S. District Judge Jennifer Rochon let that suit proceed. She found the plaintiffs had plausibly alleged that Cassava brought the defamation case with actual malice to “harass and intimidate plaintiffs and to suppress criticism from the scientific community,” and that the company knew the scientific claims at issue were “uncorroborated.” Those are allegations allowed to go forward, not findings of liability.
The university that held the records
CUNY had the original material and moved slowest in public. Bik reported her concerns to the university in November 2021, according to the timeline she published. A CUNY committee finished in May 2023 and found evidence “highly suggestive of deliberate scientific misconduct” on 14 of 31 allegations. It could not assess the rest objectively because Wang did not provide original data or research records, and it described “long-standing and egregious misconduct in data management and record keeping.” CUNY sent the report to the federal Office of Research Integrity but never published it. The public learned of it when Science reported on a leaked copy on October 12, 2023. Fifteen days later, CUNY said it would pause the underlying inquiry while it reviewed its own process.
The university’s final word arrived in a May 2025 letter that, as summarized by Being Patient, said Wang had improperly manipulated images in several studies but that this did not meet CUNY’s threshold for a finding of scientific misconduct. Five months later, on October 17, 2025, the Justice Department dropped the criminal case against Wang with prejudice, after a jury had been seated. Prosecutors have not explained the decision publicly. Wang’s lawyers have said the government failed to disclose a CUNY report favorable to their client.
The documents do not show that PubPeer comments alone forced any of these outcomes; a law firm, an FDA petition, Bik’s direct reports and sustained press coverage all bore down on the same papers. What they do show is that the central image concerns raised in public in 2021 were upheld by the journal’s retraction notice, by CUNY’s own 2023 committee and, according to that summary, by CUNY’s 2025 letter, which acknowledged the manipulation even as it declined to call it misconduct. The institution with the notebooks took four years to reach that conclusion and never published either report, while the commenters were being sued.
The Hepkema survey asked whether PubPeer comments make a difference, and the journals hold the case files that could answer it: when each investigation opened, what triggered it, and what the editors weighed. PubPeer has put its timestamps on a public dashboard, while no publisher has released the matching dates from its own files.
The Wang record shows what is left undone. As of Bik’s November 2023 count, 42 of his papers carried PubPeer posts, with seven retracted and five under expressions of concern, which left at least 30 with no notice from their journals on threads that were by then more than two years old. The editors of those journals still have the authority to act on them. CUNY holds two reports on Wang that it has not published. The malicious-prosecution claims against Cassava, Barbier and Burns remain before Judge Rochon in the Southern District of New York, and no resolution has been reported.
Sources
- Retraction Watch: Do PubPeer comments make a difference? A small study tried to find out (Nov. 11, 2026)
- MetaArXiv preprint: Hepkema et al., survey of authors and citers of PubPeer-commented papers (2026)
- SEC press release 2024-151: Charges against Cassava Sciences, Remi Barbier, Lindsay Burns and Hoau-Yan Wang (Sept. 26, 2024)
- CourtListener: Heilbut v. Cassava Sciences, Inc., No. 1:24-cv-05948 (S.D.N.Y.)
- Claims Journal (Reuters): Cassava Sciences must face malicious prosecution lawsuit (Mar. 27, 2025)
- Being Patient: Cassava fraud case dismissed, closing a chapter in Alzheimer’s drug saga (Oct. 30, 2025)
- STAT: Cassava Sciences adviser Hoau-Yan Wang indicted (June 28, 2024)
- Science Integrity Digest: Investigation finds “egregious misconduct” by CUNY scientist (Nov. 12, 2023)
- Retraction Watch: Five studies linked to Cassava retracted (Mar. 30, 2022)
- Retraction Watch: Biotech’s data supporting Alzheimer’s trials under scrutiny (Aug. 26, 2021)