Maintaining Scientific Integrity: The Many Faces of Fraud

Last year, I had the privilege of hosting the University of Oxford’s Fostering Accountability for the Integrity of Research Studies conference (FAIRS; see https://www.sjcfairsmeeting.com/). This brought together journal editors, researchers, publishers, university officers of research integrity, independent ‘sleuths’ and others. I co-chaired this meeting with Professor Dorothy Bishop, whose own credentials in the field of integrity are second to none (https://deevybee.blogspot.com/2024/11/why-i-have-resigned-from-royal-society.html). This issue of Anesthesia & Analgesia includes a “high impact bundle” of papers arising as outputs from that conference.

My first contact with scientific misconduct was many years ago, as handling editor for the journal Anaesthesia for Dr John Carlisle’s analysis of the work of Fujii.^1,2 The original submission was extraordinary in several respects. First, it was huge; akin to a doctoral thesis. Second, it originated from Torbay, which is a small seaside town in England, a non-university center. Carlisle himself had published little before then, his only main output being a Cochrane review on nausea and vomiting (itself later withdrawn due to contamination by data from Fujii that he himself discovered).³

Carlisle’s submission was really in two parts. The “trivial” part was that Fujii’s results differed from those of other groups.² Of itself this was uninteresting; differences can arise as much due to novelty as much as due to misconduct. However, the important part was his second paper, a robust analysis that Fujii could not have sampled randomly in his trials, which in turn meant that the integrity of the data were in question.² I explained the methods Carlisle used in an editorial: a similar approach had been employed many decades previously by the biologist JBS Haldane, in discovering another fraud at the time.⁴ In essence, Carlisle did not examine Fujii’s experimental results, but rather the demographic tables. For example, random sampling should yield male:female ratios that are equal with a variance predictable from binomial modeling (eg, we can predict the range of proportions of heads versus tails when tossing a fair coin). Yet, like tossing an unfair coin, Fujii’s ratios for this (and other continuous variables like height, weight, etc) were highly deviant and therefore nonrandom. In turn, this led to several editors working together to build on Carlisle’s model to try develop a generic approach to identify nonrandom—and therefore likely fraudulent—sampling.⁵

Since then, I have had the unpleasant experience of witnessing more directly a wide range of forms of scientific misconduct, and also behaviors that arguably fall just short of misconduct. Reflecting this range of data manipulation that is now recognized, many articles in this issue go beyond the analysis of numerical data pioneered by Carlisle (and updated here by Brown and Carlisle).⁶ Thus, Bik⁷ outlines fascinating work in detecting the growing problem of digital image manipulation. Byrne⁸ discusses nonverifiable nucleotides. More generally, Szabo⁹ discusses the challenges to reproducibility; the group of Marcus and Oransky of Retraction Watch discuss the weakness of sanctions, which mean deterrence is limited.¹⁰ Dr James Rathmell, our colleague Editor-in-Chief of Anesthesiology who attended FAIRS offers a personal reflection¹¹ and along with other editor-in-chiefs across anesthesia and pain journals, we publish a consensus statement on scientific integrity.¹²

All this is apposite where recently we issued a retraction (see https://pubmed.ncbi.nlm.nih.gov/42228940/)—the second of my time as Editor-in-Chief. The first was a case of image manipulation (https://pubmed.ncbi.nlm.nih.gov/41115411/) which aligns with points made in Bik’s⁷ article. This (apart from one retraction in 2021; see https://pubmed.ncbi.nlm.nih.gov/33464758/) has been the first set of retractions in A&A after 6 years, when a series of Boldt’s papers were retracted as a result of my predecessor Editor-in-Chief Steve Shafer’s efforts, as outlined by Kale et al.¹⁰ This latest retraction concerns identical data being published in a table of results for several variables, across two different journal articles, when the number of patients differed within the tables. This is mathematically impossible. The authors’ final explanation was that a junior team member had simply “copied and pasted” and that although (in their words) the papers were completely unconnected and not duplicate or salami publications, this was a mere oversight. We leave it readers to judge for themselves whether this is scientific misconduct or a lesser transgression. Regardless, the publication of such tables makes the data in both papers unsound.

In this issue is also one of the first outputs from our Editorial Fellow scheme,¹³ where we support Fellows to undertake a mini-project in publishing. This one concerns the relationship of text overlap (iThenticate) scores with chance of acceptance for a manuscript, and with geographic origin of the work. The paper reports that a crude cutoff of about 20% of text overlap is highly predictive or rejection (something that should help authors minimize extent of any copied text); and also reports that certain countries, notably China, Turkey, and Egypt have very high text overlap scores. Text overlap can be, but is not universally, misconduct. Rather it is more often a case of authors struggling to overcome language barriers and perhaps seeking shortcuts. However, high text overlap simply yields inarticulate, low-quality text which is something authors need to appreciate.

So we know the “what”: that is, the types of fraud and misconduct being perpetrated across the publishing modalities, from images, to nucleotide sequences, to data integrity. We also know the “how”; that is, the mechanisms employed to conduct the offence, from data or image manipulation to plagiarism to duplicate publication or lack of ethics etc. What we know less about is the “why”; that is, what is the motivation for highly trained, qualified, and intelligent medical professionals engaging in activity that most reasonable people regard as at best unethical and at worst criminal?

One reason, touched upon in the articles in this issue is the simple “gain vs pain” equation. For misguided researchers, the gains of publishing are high: academic promotion, peer recognition, monetary reward through research grants, industry sponsorship, academic bonus schemes, etc. However, the risk of getting caught is low—and even if caught the penalties remain trivial. It is only in extreme circumstances such as the cases of Fujii, Boldt, and Reuben that formal institutional or criminal processes come into play.¹⁰ A cursory review of the Retraction Watch database (https://retractionwatch.com/) reveals that since 2020, there are about half a dozen “household names” in anesthesia (not counting other less well known colleagues)—that is, leading researchers who continue to play significant roles on editorial boards, as chairs of department and grant review panels—who have had major papers retracted. Of course, retraction is not an automatic sign of dishonesty; it can be triggered by authors themselves, reflecting proper adherence to standards. Nevertheless, close examination of some these cases might lead some to conclude shortcomings in integrity. The supplementary material of the paper in this issue by Haney et al¹⁴ describes misconduct by a researcher who is currently a head of academic department at a university in an English-speaking country. I have personally witnessed heads of department and senior university officials lie barefaced to their colleagues, in the full knowledge that we knew they were being dishonest but had no hard evidence to take formal action. If some academic leaders fail to set proper standards, it is not surprising that others learn that misconduct offers more gain than pain.

However, unhealthy environments require certain character types for bad behaviors to be perpetuated. Research suggests character traits include^15,16: a tendency to be detached from conventional morality, unemotional, and manipulative, prioritizing personal achievement above other things; narcissism (an excessive need for admiration, a sense of entitlement, and a focus on self-enhancement); psychopathy (diminished empathy, remorse, and impulsiveness), coupled with high ambition and arrogance. The combination of “bad apples” (character) in “bad barrels” (environment) are the ultimate prerequisites for fraud.¹⁷ However, as Tijdink et al¹⁶ found, these personality traits overlap with those exhibited by academic leaders. Of course these studies are really associations and cannot conclude a causal effect.

Solutions are challenging. Articles in this review consider ways of increasing the “pain,”^10,11,14 coupled with increasing the chances of discovery.^5–8,13 However, with so many stakeholders involved (journals, publishers, professional regulators, universities, and other employers) it is difficult to agree a single, co-ordinated response, especially when—as must be their right—anyone accused may resort to legal action. Ultimately, the onus is on all of us to be vigilant and work to change the culture.¹⁸ Locally, we should speak up when we encounter unethical or unprofessional behavior.¹⁹ We can become involved in ethics committees and professional standards committees.²⁰ Within journals we can offer our services as peer reviewers (ideally progressing to editors) and serve to maintain high standards. Or if the interest takes us, we can ourselves become sleuths: diligently analyzing published literature for flaws and inconsistencies. We are all part of the same community and we can and should all play our part in upholding standards of scientific integrity.