Goldenring, James R. MD, PhD
Scientific investigation is portrayed as a noble pursuit. Yet this endeavor comes with a competitive edge that is sharp and unforgiving. We have all read the shocking and dismaying litany of cases of blatant falsification that have been motivated, one would assume, by the pressures of competition.1 In response to these ethical breaches, journals have become more explicit in their guidelines for ethical practice, most of which are philosophically obvious.2–4 These guidelines include requirements for authorship inclusion, a definition of plagiarism, and procedures for the appropriate manipulation of digital images. The guidelines from the National Institutes of Health (NIH) define misconduct as “fabrication, falsification or plagiarism in proposing, performing or reviewing research, or in reporting research results.”5 Importantly, the regulations also specify that “research misconduct does not include honest error or differences of opinion.”5 Although a number of high-profile publications have tried to assess the incidence of fraud in science,6,7 few, if any, data exist addressing how often accusations of fraud are themselves inaccurate and how able the system is to acknowledge innocence.8
Recent government documents reveal some insights into the incidence of innocence among those who are referred to the highest level of inquiry at the Office of Research Integrity (ORI) at the NIH.9 Of interest, among allegations reviewed by the ORI between 1994 and 2003, findings of no misconduct were reached in over 50% of the cases. Few, if any, data address this process at lower levels, such as at the journal editorial review stage. One recent investigation reviewed a total of 49 cases referred to the Committee on Publications Ethics where outcomes were available. Sixteen of the accused were exonerated and another 17 received only reprimands for poor procedures without evidence of intent to deceive.10 Only a small percentage of the total number of cases were finally adjudicated as true instances of misconduct. Of great concern, the average time for resolution of accusations was greater than one year. These findings, therefore, suggest that although the majority of accusations may not lead to judgments against investigators, the time to resolve these accusations can be interminable. In light of these findings, I share here a cautionary tale of false accusation from my own recent experience.
A Personal Case Study for Consideration
I recently received a formal letter from an associate editor at a prominent journal detailing the fate of a paper that my coauthors and I had submitted for publication. The editor's letter informed me that the reviews would be coming the following day but that he wanted to notify me that the paper had been rejected outright. Moreover, although the paper's first reviewer did not furnish a scientific review, he had communicated to the associate editor in the confidential comments section an accusation that we had falsified data in two of our figures. Specifically, this reviewer alleged that we had used the same image twice to show colocalization. In the figures, we had shown the individual channels in grayscale and the overlay in pseudocolors. The associate editor informed me that this charge was viewed as a very serious matter and that the accusation had been communicated to the editor-in-chief of the journal and could result in “serious” repercussions. I wrote back to the associate editor indicating that these charges were unfounded and that we could immediately send the original figure files for his consideration. However, he informed me that the matter was now out of his hands and that further correspondence would have to be with the editor-in-chief.
Because the letter I received from the journal was technically a public document, it served as a formal accusation of misconduct. As our studies were supported by federal funds, I had to inform my direct superiors immediately. Thus, five minutes after receiving the accusation, I forwarded the e-mail to the chairman of my department. He then immediately forwarded the e-mail to the associate dean of faculty. The following morning, my postdoc and I met with the associate dean and his assistant. They impounded all of my postdoc's notebooks, and we created a disc containing all of the files associated with the manuscript, including the original Adobe figure files as well as the original microscope files. These images were incontrovertible proof of our innocence, because they were locked image files from the confocal microscope. The notebooks and the digital data were all sequestered in the dean's office. While the dean was present, we sent the original Adobe files and the original microscope files to the editor-in-chief of the journal. We received a curt reply thanking us for our rapid assistance in resolving the matter and promising timely adjudication of the matter. Despite these assurances, the journal did not confirm our innocence for another four weeks.
What the Scientist Needs to Know in Accusations of Misconduct
After the resolution of our case, I realized that most faculty members do not understand the important elements of the system necessary for response to formal allegations of misconduct. I detailed our experience here to frame a discussion of the appropriate response to such accusations of misconduct and the successful establishment of innocence. From the outset, scientists must understand that a formal letter of accusation from a journal editorial board or society is a legal document. Because my work was funded by the NIH, I was obligated to report the accusations at once to my immediate superior and the university. Such accusations do have legal implications that can lead to a range of outcomes, including loss of federal funding and even loss of academic position. Immediately informing the department chairman and the appropriate dean is the critical first step in protecting the accused faculty member. Although this process is somewhat demeaning, it clearly is the most important part in protecting a scientist's evidence and allowing the university to support his or her innocence. The quickest way to the truth is rapid and full disclosure to the authorities of the university. Impounding of data books and image files establishes a faculty member's intention to be completely forthcoming with all pertinent items and eliminates any future accusations of data tampering.
At the center of these actions must be an understanding that cases of scientific misconduct are adjudicated in a manner quite distinct from the democratic judicial system. In cases of scientific misconduct, the accused is clearly guilty until proven innocent. Although there has been some recent discussion of this discrepancy in the medicolegal literature,8,11,12 the implications are generally unfamiliar to most members of the scientific community. The rules governing misconduct among scientists are derived not from true legal determinations but from standards of appropriate behavior set down as a general understanding by the scientific community as a whole. These guidelines include ethical practices related to the authenticity of data, avoidance of conflict of interest in peer review of manuscripts and grants, and equitable sharing of knowledge. These governing rules have evolved over centuries without specific legislation, and therefore the principles of behavior more closely resemble the requirements for membership in a guild. One could argue that the guild stature of the scientific community makes for the equitable application of rules outside the normal rules of legal process. Whereas physicians adopt a formal affirmation of conduct in the Hippocratic Oath, entrance into the community of scientists would seem to assume acceptance of a similar code of conduct and ethics that transcends the normal rules covering the assumption of guilt. Although no particular oath of conduct is taken by scientists, more recently these guild rules have been codified in guidelines from the NIH.5 The acceptance of NIH funding carries with it an agreement to adhere to parameters of ethical scientific conduct. Although these policies do establish that scientific misconduct should be separated from “honest error or differences of opinion,” they still confirm a general understanding of a burden of proof on the accused. Because NIH guidelines are issued by a federal agency, they can also have criminal implications. Thus, a scientist who fabricates his or her data is not only violating the guild obligations but is also defrauding the federal government, an offense subject to legal prosecution. Despite these serious implications, my own concern lies more with the relationships within the scientific community.
If scientists admit that an assumption of guilt in cases of suspected scientific misconduct is part of the guild credo, then we must also promulgate procedures that protect our ability to provide appropriate documentation of innocence. Most cases of data manipulation involve inappropriate or fraudulent alteration of digital images. At the time of manuscript submission, I always create a CD with all of the relevant text and image files, especially nonflattened Adobe Photoshop files. In the past, I built such a resource so that my coauthors and I could modify figures in the future if the need arose, but now I understand that these discs should also include the original microscope files for all of the images used in the figures. This record is the ultimate protection. As the burden of digital image archiving falls on the junior investigators in the lab, it was clearly fortuitous for me that my postdoc was so organized that she could easily retrieve the original microscope files in less than five minutes. In this age of digital images, most without detailed annotation, scientists must be more diligent in archiving digital data, both published and unpublished. This type of archived documentation is the best defense against any allegations of fraud.
Given the acceptance of the assumption of guilt as part of the structure for this guild-based code, scientists must therefore accept responsibility for rapid accessibility to original material and digital records. Indeed, the NIH guidelines specifically state that the accused researcher carries the burden of producing the requisite documentation to demonstrate his or her innocence.5 Principal investigators must develop procedures for accurate archiving of data and images so that any concerns of reviewers or even colleagues can be answered rapidly. This archiving system should include not only original figures but also original, full-size images of gels and autoradiographs, and original microscope image files with complete annotations.
The Appropriate Editorial Response to Concerns of Misconduct
As a critical ramification of our system of peer review, manuscript reviewers can level accusations while protected by the veil of anonymity. Therefore, in contrast to the rules of criminal proceedings, for peer-reviewed research the accused cannot confront his accuser. This critical part of our code of legal jurisprudence cannot be applied in a situation in which accusations are leveled by a protected reviewer. Without the guarantee of anonymity, many reviewers may be reluctant to effectively criticize manuscripts and grant proposals. Nevertheless, because of this anonymity, there is a certain risk of abuse. One would have to be exceedingly naïve to believe that some reviewers do not use their shielded status to delay the publication of articles by their competitors and/or the competitors of their colleagues. Although editors can try to avoid these conflicts of interest, ultimately the success of the system is subject to honorable intent, which I believe does pervade most scientific reviews. In our case, we thought that the reviewer was likely acting out of ignorance and overzealousness rather than malice. Still, the experience does reveal a disturbing mechanism that could be used to delay publication of a rival. A reviewer determined to delay publication could level an accusation from a point of anonymity, knowing full well that the victim would likely be cleared of charges eventually. Fraud seldom occurs without context, so a complete review of the paper is necessary for establishment of intent. Editors should register some concern if a reviewer delivers accusations without completing a scientific review, as was the case with the allegations against our manuscript.
Diligent editorial oversight is critical to identifying and discouraging acts of fraud or misconduct in scientific publishing. Both editors and reviewers must be free to raise appropriate concerns. Nevertheless, our own experience illustrates that the editors also must use procedures appropriate to the situation, especially as they pertain to concerns of the validity of pieces of data contained in figures. All journal submission guidelines contain a statement that authors may be asked to present original data for further examination. These requests commonly arise when a reviewer wants to see a complete gel image rather than one that is cropped, or when a reviewer is concerned with the inappropriate assembly of a gel or other digital image.4 In our case, rather than immediately turning the matter over to the editor-in-chief, the associate editor should have requested the original image files to demonstrate to the reviewer that no manipulation of data had occurred during the assembly of the figure, suggesting colocalization.
Because requests for original data usually are handled through e-mail communication, no formal allegations are made and concerns can be addressed easily with the presentation of the original data files. The reviewer remains involved in the resolution if questions are approached in this manner. If such procedures had been followed in our case, a great deal of delay and emotional turpitude could have been avoided because we were able to provide the source files in question quickly. Bringing accusations such as ours to the editor-in-chief without an initial inquiry often elevates more trivial allegations to the level of serious infractions. In my experience as an associate editor for the American Journal of Physiology, I have found that questions of self-plagiarism or double publication are also best evaluated at the level of the associate editors. In both contexts, issues can often be resolved without entering into official proceedings initiated by formal accusation. Once an editor provides a letter on journal and/or society letterhead containing accusations of scientific misconduct, a formal and public claim has been made, triggering the internal report system that we activated at our university.
If our scientific guild requires that investigators be responsible for rapid accessibility to original data, then editors must assume the responsibility for rapid due process of all accusations. When an editor asks for the original files to clarify a reviewer's concerns, there is a distinct clarity in the process. The authors can expect that once the original data have been provided, the situation will receive rapid adjudication by the editor and reviewer. What makes the process untenable is when there is a lack of clarity in the process of review. For instance, in our case, it was not clear who was evaluating the original material we had provided or how authenticity was being established. Just as important, no clear timeline for resolution had been established.
Because the preponderance of accusations will likely be either erroneous or will uncover an unintentional mistake, editors must be dedicated to the rapid resolution of all concerns raised during the review process. In our case, no explanation was given as to why it took 30 days to determine that the image files were authentic. Nevertheless, some studies have suggested that many such reviews take up to a year or more to complete.9 This protracted timeline likely involves inadequacies on the author's side to produce the appropriate original documentation9 as well as inefficiencies in editorial channels to process and analyze disputed items. Still, best practice guidelines have been suggested for editorial review processes,13 and the incorporation of these strategies into the standard operating procedures for editorial boards may help expedite the evaluation of allegations of misconduct.
Due Diligence Is Required by All
Science remains an avocation in which an individual's reputation is his or her calling card. Unfortunately, the present system employs little beyond confidentiality to ensure that the reputation of the innocent is not sullied by the process. In my own case, the efforts of the dean's office were successful in maintaining this level of confidentiality. It is not clear whether this result is always the case. Some have characterized science as a competitive sport, and a contact sport at that. But science is at its essence a pursuit of transient truth. Great scientists must critically evaluate their hypotheses in light of new knowledge, whether it comes from their own labors or the labors of others. As investigators, we build our reputations among members of the scientific research guild by demonstrating our ability to criticize both ourselves and others without malice. We are participants in the scientific discourse, rather than combatants in pankration. I still cannot conceive of any process better than peer review. Peer review is successful as long as the reviewers are open to the flow of new ideas and new investigators. Despite my recent experience, I have not soured on the process; I do believe, though, that the integrity of the scientific research guild rules requires an increased measure of due diligence by all involved parties.
The author is grateful to Marialuisa Gallozzi and Dr. Vivan Siegel for their critical review of this manuscript.
1Saunders R, Savulescu J. Research ethics and lessons from Hwanggate: What can we learn from the Korean cloning fraud? J Med Ethics. 2008;34:214–221.
2Redman BK, Merz JF. Research misconduct policies of high impact biomedical journals. Account Res. 2006;13:247–258.
3Vollmer WM. Responsibilities of authorship. Chest. 2007;132:2042–2045.
4Rossner M, Yamada KM. What's in a picture? The temptation of image manipulation. J Cell Biol. 2004;166:11–15.
5Public Health Service policies on research misconduct. Fed Regist. 2005;70:48370–48400.
6Marshall E. Scientific misconduct. How prevalent is fraud? That's a million-dollar question. Science. 2000;290:1662–1663.
7Titus SL, Wells JA, Rhoades LJ. Repairing research integrity. Nature. 2008;453:980–982.
8Spece RG, Bernstein C. What is scientific misconduct, who has to (dis)prove it, and to what level of certainty? Med Law. 2007;26:493–510.
9Rhoades LJ. ORI Closed Investigations Into Misconduct Allegations Involving Research Supported by the Public Health Service: 1994–2003. Rockville, Md: Office of Research Integrity, Office of Public Health and Science, Department of Health and Human Services; 2004.
10Wager E. What do journal editors do when they suspect research misconduct? Med Law. 2007;26:535–544.
11Spece RG Jr, Marchalonis JJ. Fourth Amendment restrictions on scientific misconduct proceedings at public universities. Health Matrix Clevel. 2001;11:571–626.
12Spece RG Jr, Marchalonis J. What should the standard of proof be in scientific misconduct proceedings relating to public health service-funded research? Cell Mol Biol (Noisy-le-grand). 2003;49:565–577.
13Graf C, Wager E, Bowman A, Fiack S, Scott-Lichter D, Robinson A. Best Practice Guidelines on Publication Ethics: A publisher's perspective. Int J Clin Pract. January 2007:1–26.