Perspective

In 2002, a paper was published by a group of investigators at the University of MinnesotaReference Jiang¹ describing the potential for mesenchymal stem cells to differentiate into a range of mature cell types. This study would be cited more than 4,400 times in the following decades, but in 2005 concerns were raised that some of the data appeared to have been reused from an earlier paper where they were described as representing different conditions. An inquiry led by the University of Minnesota concluded that there was no misconduct related to this paper, only “flawed” data, errors, and “inaccuracies.”Reference Aldhous² Additional concerns were raised in 2019 showing that a pair of images with overlapping features and altered fluorescence had been used to represent different conditions and another image had internal areas that appeared to be clones.Reference Bik³ In June of 2024, more than two decades after publication, the study was retracted becoming probably the most cited paper to ever be retracted.Reference Kincaid⁴

Another highly influential paper was published by another group of researchers at the University of Minnesota in Nature in 2006.Reference Lesné⁵ In this case, a specific assembly of toxic peptide β-amyloid peptides, termed by the authors Aβ*56, was shown to induce memory impairment in laboratory animals and hypothesized to be a part of the neurobiology of Alzheimer’s disease. In 2022, concerns were raised on PubPeer that several Western blots in this paper seemed to contain cloned areas.⁶ Two blots representing key experiments appeared to be duplicates of each other, although some features were altered between the uses. In May 2024, the senior author of the study commented on PubPeer, “Although I had no knowledge of any image manipulations in the published paper until it was brought to my attention two years ago, it is clear that several of the figures in Lesné et al. (2006) have been manipulated.”⁷ The paper was ultimately retracted, joining the ranks of the most-cited papers to be retracted.Reference Lesné⁸ In two other papers from this group of authors, corrections were issued in which the authors admitted images were “processed inappropriately.”Reference Lesné⁹ Despite these admissions, an inquiry led by the University of Minnesota strikingly determined that no research misconduct had occurred,Reference Piller¹⁰ an echo of the similar conclusion related to the 2002 University of Minnesota paper.

In contrast, when allegations of data manipulation and fabrication were raised in November of 2022 regarding several papers from the laboratory of Marc Tessier Lavigne, then president of Stanford University, a panel of external experts were convened by the Stanford Board of Trustees to evaluate the allegations. The panel consisted of independent academic leaders with the scientific expertise to address the allegations supported by forensic and legal experts. Through weekly meetings, the panel completed their evaluation in approximately eight months and released a public report of their findings, including recommendations of specific papers requiring retraction or correction.¹¹ The rigor, transparency and efficiency of this investigation led to definitive corrective action, improved the scientific record, and in our view, created confidence in the fairness of the process. Other examples of institutionally led investigations into concerns about research publications include ones from the Ohio State University,Reference Van Noorden¹² Northeastern University,Reference Chawla¹³ University of Louisville,Reference Bik¹⁴ University of Texas/MD Anderson,Reference McCook¹⁵ Cornell University,Reference Servick¹⁶ Centre National de la Recherche Scientifique (France),Reference Sohn¹⁷ and University of Leiden (Netherlands),Reference Marcus¹⁸ all of whom published reports or statements about their investigatory findings.

When cases involving concerns of data manipulation or fabrication arise in an academic setting, the concerns are usually brought to the attention of a funding/regulatory body like the National Institutes of Health or the Office of Research Integrity at the US Department of Health and Human Services, the editorial personnel at a journal or publisher, the investigators themselves and/or their academic leadership.Reference Caron¹⁹ All of these routes will generally be referred to the originating laboratory’s parent institution through an institutional research integrity office (RIO). If such allegations appear to contain substance, the RIO will usually proceed to an inquiry phase in which preliminary data is gathered. If serious concerns are found, the inquiry phase will be followed by an investigation phase. At the beginning of the investigation, a committee is formed, usually consisting of professors and administrators from the same institution as the accused party. This investigation phase might involve interviews with the respondent — the person who is accused of misconduct — or sequestration of lab notebooks, hard drives, email correspondence etc.²⁰

When research misconduct allegations involve externally funded work, the RIO may be required to follow standards under applicable funding agency requirements.²¹ Partly due to regulatory requirements and applicable laws, but also as a practical matter, these investigations generally proceed slowly, often occurring over a period of years. Investigatory committees must be formed, lawyers consulted, facts gathered, interviews conducted, reports drafted, regulatory authorities consulted — all these things take time and resources. Certainly, in many cases, institutional bureaucracy contributes to the slow progress of these investigations. Because of the slow speed at which the institutional investigations progress, it is difficult to determine whether investigations are ongoing, whether they occurred at all, or whether a conclusion was ever reached. Eventually, most institutional investigations into allegations of misconduct are dismissed or never produce any public report at all.²² In some cases, like the first two examples described above, an institutional determination that no corrective action is required is eventually contradicted by other authorities (in both described cases, the journal editors ultimately elected to retract the paper).

There are several reasons why institutional investigations of legitimate misconduct allegations might never take place, take years to complete, or never result in a public report. Because an institutional finding of misconduct could result in reputational damage or in the institution being required to return financial resources to funding agencies, there is a conflict of interest in allowing institutions to investigate their own faculty.Reference Resnik, Hosseini and Rasmussen²³ In addition, university staff or faculty accused of research misconduct might threaten legal action.

Although protecting the legal and civil rights of the participants is critical, and while the confidential nature of the investigations may assist in preserving evidence and obtaining unbiased witness testimony, the current system of slow investigations and undisclosed outcomes does not serve the needs of the broader scientific community, nor those of taxpayers or patients. The failure to address misconduct might inflict harm on other researchers or even clinical trial participants, potentially directly or indirectly leading to public health and safety risks. Not investigating or publicizing findings of misconduct will lead to wasted effort and money spent by other scientists trying to replicate fabricated results and might even lead to patients fruitlessly participating in clinical trials based on faulty research. It would serve the scientific community and beyond if the results of misconduct information were to be made public.

Importantly, in the United States, there is no federal requirement that the results of an investigation must remain private. The US Health and Human Services Office of Research Integrity explains it this way:

Publicly funded research institutions owe a duty to the broader scientific community and to patients, and (we argue) to taxpayers too. These institutions have a responsibility to disseminate information from investigations where it may (1) improve the scientific record, (2) aid with research integrity education, (3) deter misconduct, (4) inform the public and medical community about research related to treatments, and (5) serve as an aid to help assure the public that institutions who are entrusted with tax dollars are actually worthy of that trust.

To address these concerns, we have the following recommendations that could be incorporated into a reformulation of best practices related to research integrity investigations:

1. 1. Credible reports of a violation or potential violation of research integrity of substantial magnitude should be investigated by an independent investigative body. “Substantial magnitude” specifically refers to potential violations occurring as a pattern over an extended time period, involving multiple publications, grants or other professional output and/or involving multiple individuals or potential violations of large scientific, medical, or financial consequence. “Independent” means the majority of the members, including the panel chair, are not affiliated with the same institution as the target of the investigation. The cost of the independent panels would be a well-justified investment in building public trust in the scientific enterprise and could be offset by mandating a more efficient process and by more rigorous enforcement actions against grants funding corrupted scientific output.

2. 2. The investigative team should at a minimum include scientific experts within the relevant domain, as well as members with expertise in data forensics, for example, people with specific experience with image manipulation,Reference Bik²⁵ or data analysis and statistics.Reference Simonsohn²⁶ The scientific domain expert(s) recruited for this process should be reasonably viewed as peers to the individual or individuals under investigation. The investigators should be given the authority to determine the scope of the investigation, which should generally be broader than the allegations.

3. 3. Allegations from anonymous sources or individuals with conflicts of interest should not be discounted but should be verified by a non-conflicted investigative body and referred for independent investigation if there is merit to the complaint.

4. 4. A function that should remain the primary role of the institution is to assemble and preserve the evidence required by the appointed investigative body, ensure robust cooperation with the panel and assist with scheduling and coordination. To ensure adequate preservation of evidence, the collection of key records generally needs to occur at the beginning of the investigative process and be completed expeditiously. Evidence that tampering with or destroying lab notebooks, samples, files, films, blots, scans, storage media etc. has occurred after initial concerns have been raised should be considered misconduct.

5. 5. The investigation should occur efficiently so that necessary corrective actions are applied within a timeframe that will mitigate the harm of the integrity violation or alternatively, the accused party/parties may be publicly exonerated within a timeframe that protects their work and minimizes reputational harm. The timeframe should be publicly specified, enforceable and should rarely extend beyond one year. While current US RIO regulations often include timelines for investigations, in practice, our experience has been that these timelines are not meaningfully enforced, frequently leading to unreasonable delays. Consequently, independent investigations should have clear milestones, members of the investigative body should meet at regular intervals until their work is complete and they should have the authority to ensure the investigation progresses efficiently. This may require the authority to impose punitive measures, including levying fines or initiating the recovery of a portion of the implicated funding, if investigations are unreasonably delayed, to ensure that cooperation with the investigative panel is incentivized.

6. 6. The investigative body should produce a public report addressing the reliability of the involved scientific output and, when necessary, clearly indicate which published papers should be retracted or corrected. Correction should be reserved for errors of a limited nature and where the source data can clearly establish confidence in the conclusion. Examples of this type of error would include isolated publication of an incorrect representative image, or minor errors in data handling when the raw data is available for re-analysis. Intentional manipulation of data, even for only “cosmetic” purposes, should generally be addressed with retraction. Similarly, significantly corrupted papers, even if due to error rather than manipulation, should not be permitted to be corrected. Because many journals will not initiate corrective action without direction from an investigative body, recommendations on retraction or correction should be communicated to the relevant journals as soon as clear evidence supporting a decision is available — these recommendations do not need to be delayed by the completion of the investigation. Institutions should periodically check if journals addressed those papers as recommended.

7. 7. For potential research integrity violations that are not viewed as “substantial,” institutional officials should conduct diligent oversight to ensure the issue or issues are not part of a larger pattern. Institutions should have an integrity maintenance program which should involve active screening of research output, early intervention on anomalous data and regular public discussion to raise the profile of these issues and intentionally create a culture of integrity. Simple examples of concrete preventative measures would include active monitoring of forums like PubPeer where concerns about data integrity are frequently raised, and ensuring that screening tools, like ImageTwin or Proofig, are accessible to investigators and their departmental leaders to detect common problems.

8. 8. Mechanisms should exist to address institutions with multiple episodes of substantial violations of research integrity standards in externally funded research to systematically improve the research culture at the institution, reevaluate the adequacy of prior investigations and ensure institutional leaders are held accountable when frequent integrity violations occur. To incentivize cultural improvements in institutions with multiple substantial violations of research integrity standards, it may be necessary to pause eligibility for certain forms of research funding or suspend payments on existing grants until improvements are achieved. Independent investigative panels would be in an ideal position to address whether research integrity violations represent isolated behaviors or institutional culture deficiencies and recommend appropriate corrective actions to federal funders and enforcement agencies.

We believe the current model for addressing research integrity concerns does not adequately disincentivize data manipulation and creates an inherent conflict of interest. Delayed and anemic investigations that reach secret conclusions not only compromise public trust in science, they compromise the reliability of science and ultimately delay scientific progress. Complex diseases will only be solved through clinical studies which require a partnership between scientists and patients — a partnership where patients are asked to accept significant personal risk. The seriousness with which we address scientific misconduct in this domain needs to keep our responsibility to patients and to the public a central focus.