In a report released on Thursday, the National Academies of Sciences, Engineering, and Medicine urged the scientific community to respond more aggressively to claims of errors or manipulated data in published research, saying that papers containing findings that can no longer be trusted should be more swiftly corrected or retracted.
The report encouraged journals to complete editorial review of potentially problematic studies within 12 weeks, or within six months in more complex cases. The authors also stressed that study authors need to hold onto original data, adding that being unable to produce these records when concerns about a finding’s authenticity are raised could lead to a retraction. And the expert panel wrote that it is no longer “prudent” for researchers to argue against retraction of their work by asserting that flawed findings in one part of a study don’t impact the paper’s overall conclusions. That’s because issues with any part of a study can cast doubt on other findings.
The authors called on universities and other research institutions to respond more quickly and completely to research integrity inquiries, in part by making findings from internal inquiries available to journals, funding organizations, and the broader public. And the National Academies report suggested that funding organizations tie grant awards to researchers and institutions agreeing to abide by strict standards for the conduct of research and addressing errors.
“While the committee’s recommendations might appear aspirational, we are certain that the funding agencies, the public officials who legislate financial and regulatory support for science, researchers, and, most importantly, the public must be assured that the record of scientific knowledge is intact and reliable,” wrote University of California, Berkeley cell biologist Randy Scheckman, chair of the committee that produced the report.
These and other recommendations in the 157-page National Academies report come at a time when concerns about the reliability of the scientific record have drawn increasing public attention, contributing to distrust in scientific institutions. High-profile investigations at leading research centers such as Stanford and Dana-Farber Cancer Institute have in recent years shined an uncomfortable spotlight on how studies published in top journals can contain doctored data. Meanwhile, the number of scientific articles published each year has risen from 1.3 million in 1995 to 6.7 million in 2025, increasing the number of studies that potentially require corrective action.
Retractions have become more common in recent years, with the report noting that more than 5,000 studies were retracted worldwide in 2025. But the report alleges that this rate is lower than expected given current estimates of the proportion of problematic articles, with fewer than 0.22% of studies from any given year retracted. Data from Retraction Watch indicate that, on average, retractions happen nearly 2.4 years after a study is published.
Correcting the record more often and more quickly, the authors write, won’t just require clearer policies from funding organizations, journals, and universities. It will take a cultural shift in science toward viewing correction and retraction as essential tools in maintaining the reliability of the record, not as a professional failure. They encouraged stakeholders to think of the consequences of not taking corrective action on studies, from policymaking and public health decisions made based on faulty data to researchers wasting time and taxpayer dollars when they attempt to follow up on a flawed finding.
“Serious consequences for the scientific enterprise and for society in general result either from failing to correct the scientific record or from failing to do so in a timely way,” the panel wrote. “These consequences make changing the process essential.”
While many corrections and retractions result from honest errors, the Trump administration has cited cases of alleged research misconduct as justification for its focus on “gold standard” science. In the new report, experts noted that these incidents can weaken public trust in science.
“It is critical that the scientific ecosystem continues to prioritize maintaining the integrity of the scientific record so that it remains trusted and trustworthy,” the authors wrote. “Even if these cases are rare, just one such incident erodes the public’s trust.”
That recognition spurred the National Academies, a 163-year-old nonprofit created by Congress to advise the nation on issues of science and technology, to assemble a team of university, journal, and academic leaders to explore what should be done. They included Holden Thorp, editor-in-chief of the Science family of journals; Schekman, a Nobel Prize-winning researcher and former editor-in-chief of the Proceedings of the National Academy of Sciences; and Mark Barnes, a lawyer with the firm Ropes & Gray and former associate provost for research at Harvard University. The group was tasked with identifying current practices and suggesting new standards for stakeholders.
The authors lay out what they frame as an “ideal timeline” for addressing concerns with published studies. That includes four weeks for an editor to determine whether a concern is potentially credible, meaning that it is specific, makes sense, and merits a closer look. The committee then allots another four weeks for contacting authors and determining whether their response adequately resolves a concern before initiating a full review. They note that an initial notice, such as an editorial expression of concern, should be added to a study within 12 weeks.
It’s a well-intentioned timeline, says Lisa Rasmussen, a bioethicist at the University of North Carolina at Charlotte. But Rasmussen, who is also editor-in-chief of the journal Accountability in Research, says that editors like herself are already stretched thin with work that is often poorly compensated. She notes that, in the past five years, the number of studies submitted to her own journal has increased almost tenfold. Rasmussen and a co-editor-in-chief often message each other with drowning emojis to sum up how they are doing.
“I think under the best of circumstances, it’s very reasonable. That’s certainly the kind of time frame that I try to abide by,” she said. “I think it’s just going to be increasingly difficult for people to handle the volume of concerns [and] queries that they’re receiving in the time frame.”
Rasmussen noted that much of the rise in submissions took place after large language models powered by artificial intelligence, such as ChatGPT, became readily available. The report touches on the impact of AI on research integrity, noting that these tools have simultaneously allowed researchers to manipulate or fake data while giving journals and universities new tools to spot suspect results. But to Rasmussen, the report’s discussion on AI was surprisingly brief given the technology’s disruptive potential.
“AI is being used to review. AI is being used to critique. AI is being used to fabricate data. So it’s like jet fuel for all of the stuff they talk about in the report,” she said. “But it’s almost as though the report was written pre-ChatGPT.”
Source: www.statnews.com



