Infighting among a group of prominent Alzheimer’s disease researchers has led to the withdrawal of a preprint they co-authored, which suggested a new Alzheimer’s drug markedly increases the risk of death. One scientist involved in the work charges the senior author failed to seek the go-ahead from his co-authors before posting an edited version of the article to a preprint server.
The 22 October preprint garnered attention from researchers and the media, including Science, for its dire picture of the risks of lecanemab, a therapy recently approved for early Alzheimer’s. The drug’s modest benefits and sometimes-severe side effects are the subject of intense disagreement among clinicians. But the quarrel also highlights the unique challenges of preprints, which allow scientists to upload study results prior to peer review and receive input from colleagues, but may draw attention to preliminary findings and can’t be easily purged from the public record. Some of the authors say the experience taught them lessons.
“Old men are used to publishing in journals,” says co-author Robert Howard, a professor of old age psychiatry at University College London who has no previous experience with posting preprints. Howard is one of five authors who asked the paper’s senior author, neurologist Alberto Espay of the University of Cincinnati (UC), to remove their names from the preprint because of doubts about the evidence it presents. He says he had previously viewed posting a preprint as akin to circulating a journal submission among editors and peer reviewers. There, it might be overhauled or even abandoned because of feedback received. “We kind of assumed that we could apply the same structure and principles and behavior to the preprint,” he says.
The story begins several months back, when Espay decided to learn more about lecanemab from a database called the Food and Drug Administration Adverse Events Reporting System (FAERS). FAERS collects and makes available anonymized data drawn from reports by doctors, patients, and others about potential drug side effects. Although the database can offer clues for future study, it’s far from definitive. Among other issues, the reports do not prove a particular side effect was caused by a particular drug, and reporting is voluntary.
Espay has long expressed concerns about the safety of the drug category to which lecanemab belongs, called antiamyloid monoclonal antibodies. The drugs, which sweep away sticky brain deposits that are widely believed to contribute to Alzheimer’s, have a well-documented risk of brain swelling and bleeding, and a handful of people died from taking lecanemab during clinical trials. “We desperately need” data on the risk of death from the drug in the real world, says Eric Widera, a geriatrician at the University of California San Francisco and a co-author on the preprint.
Espay and three colleagues collected reports from FAERS that recorded 25 deaths in people taking lecanemab. They also tried to estimate the total number of patients on the drug, as well as the mortality rate from Alzheimer’s in people with the disease. Their analysis suggested a 2.6 times increased risk of death with the drug. They described the finding in a 600-word letter and invited Howard, Widera, and three other researchers to offer feedback and join as co-authors. All agreed. The letter was submitted to JAMA and then JAMA Network Open. Both journals rejected it.
At that point, Espay suggested to his co-authors that they “consider adding this to a preprint server,” according to a 30 September email he shared with Science. He wanted the information out there and didn’t see a clear alternative, he says. He can’t recall whether anyone responded, and on 22 October he posted the preprint to the site Research Square. “It was largely the same manuscript sent to JAMA,” he says.
When Espay sent the preprint link to his co-authors, discord erupted.
“He submitted without telling me he was even submitting,” charges Lon Schneider, who directs the California Alzheimer’s Disease Center at the Keck School of Medicine of the University of Southern California. Schneider, who saw Espay’s email but did not interpret it to mean a preprint posting was imminent, was particularly incensed by changes Espay made to the original letter but didn’t share in advance. These included softening language on the study’s limitations, suggesting the data support decisions by regulators in Europe not to approve the drug, and saying the “unfavorable mortality signal” warrants “considerations of accelerated withdrawal” from the market. (Schneider shared the originally submitted version with Science for comparison.)
Among the preprint’s co-authors, Schneider is in a unique position: Although he has worries about the drug’s safety, he is also co-leading a clinical trial that gives lecanemab to people with a highly inherited form of Alzheimer’s. He’s concerned about side effects but says, “It’s incongruous that I or anybody else would lead such a study while saying the drug should be removed from the market.” On 24 or 25 October, Schneider says he asked Espay to have the preprint taken down or at least remove Schneider’s name as a co-author.
Meanwhile, the preprint was getting media attention. The New York Times included it in an article on risks from antiamyloid antibodies. The Daily Mail covered it. And so did Science, which spoke with Espay for a piece last week on the struggles of clinicians to integrate lecanemab into patient care. The coverage startled several co-authors. “I don’t think that preprints should be heavily publicized,” says Widera, for whom this was also his first preprint experience. “They are a work in progress.”
Schneider and other co-authors also began to get messages from colleagues critiquing the work. Some, for example, worried that deaths of trial participants might have been counted twice in FAERS, Widera says. Schneider and several other co-authors became concerned that the data in the preprint were too weak to support the claims it was making. Co-author Karl Herrup, a neurobiologist at the University of Pittsburgh, has grave concerns about the antibody treatments but says it’s crucial to get safety data right before making public claims. The Alzheimer’s field “is in a very precarious place right now,” he says. In the days that followed, he, Widera, and two other co-authors joined Schneider in asking that their names be removed. (Schneider says he did not contact the site himself because that’s normally the role of the corresponding author.)
On 31 October, about a week after Schneider first voiced complaints, Espay asked a fellow UC co-author, neurologist Abhimanyu Mahajan, to contact Research Square about having the article taken off its website. A response came the next day: Because preprints, like published papers, are assigned a digital object identifier, a number that enters them into the scientific record, they can’t simply be purged from it.
That’s a typical policy, says Richard Sever at Cold Spring Harbor Laboratory, who co-founded the preprint servers bioRxiv and medRxiv. Even a withdrawn preprint remains in the record, like a retracted journal article, he says, and “you can’t snap your fingers and make it go away.” Preprint servers aim to document updates or changes to a paper, but keep the original one in the public domain.
A staffer from Springer Nature, which owns Research Square, told Mahajan that in submitting the manuscript, Espay had agreed “that the contents will be permanently shared as a public preprint on our site.” He was encouraged instead to submit a revision, on which he could also change the list of authors.
Espay tried that on 4 November, crafting a different article that examined causes of death among those taking lecanemab, rather than assessing the death rates themselves. It reported that 40% of the deaths were due to brain swelling or bleeding, and the rest to presumably unrelated causes such as cancer or heart disease. The revised preprint also removed five co-authors, as they’d requested. But Espay says Research Square would not replace the original preprint with the new analysis.
The saga ended yesterday, when Research Square officially marked the preprint as “withdrawn,” though it remains accessible, as the platform’s guidance notes, if not searchable on the site. The full text and author list remains, with a note that the authors no longer stand by its conclusions.
The experience has left the participants with mixed feelings. Schneider is still outraged that he was listed as an author on a preprint whose contents he does not endorse. The article has racked up more than 4300 views.
Espay is sympathetic to Schneider’s critiques and wishes he’d handled things differently. “I jumped the gun without him,” he concedes.
But he adds that his abiding concern about lecanemab and related drugs made him reluctant to immediately disavow the preprint or replace its conclusions with something more ambiguous. He’s certain “there are more deaths than there should be,” whatever the precise increase. And in a sense, the preprint sphere worked as it was supposed to, he adds: “This paper underwent the most public and stern peer review”—the scrutiny of peers online—“and was ‘rejected.’” He plans to continue his efforts to study harms from antiamyloid drugs.
In the meantime, several of the co-authors may be steering clear of preprints for the time being. Asked whether he’d sign on to another anytime soon, Widera is unequivocal. “Heck no,” he says. “Heck no.”
