By Zeenat Diwan
This is the fifth post in our series about how science is communicated and the consequences thereof.
“Science has the potential to address some of the most important problems in society and for that to happen, scientists have to be trusted by society and they have to be able to trust each others' work. If we are seen as just another special interest group that are doing whatever it takes to advance our careers and that the work is not necessarily reliable, it's tremendously damaging for all of society because we need to be able to rely on science.”
—Ferric Fang quoted by Jha (2012)
As aptly described by the quote above, the issue of research misconduct is one that concerns both scientists and society. The Office of Research Integrity, which is part of the US Department of Health and Human Services, defines research misconduct as “fabrication, falsification, or plagiarism in proposing, performing, or reviewing research, or in reporting research results.” All three acts undermine the core values of research integrity, increase mistrust in scientific work and ultimately hinder progress in society.
We’ve all come across examples of plagiarism in our daily lives. As scientists and writers, we are rigorously trained in correctly citing information sources, while as an audience, we sometimes hear stories about instances of plagiarism in the news. But what about data fabrication or falsification? What do they look like? Surprisingly, these are topics that are not discussed as widely as plagiarism, but happen just as often. This blog is thus aimed toward providing a better understanding of these two arms of research misconduct, what they entail, their consequences on scientists and society, and efforts to combat these issues.
What are some examples of research misconduct and how prevalent is this issue?
One of the most common examples of data fabrication or falsification is image manipulation, wherein an image has been duplicated to create a non-existing piece of data, or part of an image has been glued on to another to falsify data. The most frequently manipulated images in biomedical research are for Western Blot (a technique used to detect presence or level of a protein in biological samples), FACS or Fluorescence Activated Cell Sorting (used to sort a specific cell type from a mixed population), tumor tissues, bacterial colonies plated on agar, etc.
The past decade has seen a rise in research misconduct studies. For example, a 2016 study assessed the pervasiveness of inappropriate image duplication. The authors visually screened over 20,000 biomedical papers published by 40 different journals between 1995 and 2014 and found around 4% of these papers to have at least one inappropriate image duplication, and at least half of those (2% of total) pointed to intentional manipulation. In a 2018 article, image integrity analyst Jana Christopher found that around 2–3% of formally accepted manuscripts at FEBS Press regularly have image integrity issues and have their acceptance revoked. In 2021, several research integrity sleuths assembled a Google spreadsheet containing a list of 42 alleged paper mills. A “paper mill” is essentially a company that spews out fake papers that scientists can buy authorships in for exorbitant amounts of money. The list encompasses more than 7000 problematic papers. That’s disturbing, to say the least.
Although a majority of the above-mentioned instances of misconduct were reported for articles in biomedical journals, editors at a forum held by London-based advisory committee COPE (Committee on Publication Ethics) have revealed that they’ve also seen paper mills in other scientific disciplines such as computer sciences, engineering, humanities and social sciences.
Why do scientists commit research misconduct?
There’s fierce competition amongst scientists to publish in order to climb the next step of the ever-shrinking academic ladder. Moreover, it is widely acknowledged that positive results have a higher chance of getting published and in better journals compared to negative ones. These factors may ultimately tempt some researchers to take the easier route to academic success – by manipulating or fabricating data. However, despite this thinking, the ’pressure to publish’ factor was shown by a 2018 study to not significantly be associated with instances of misconduct. Interestingly, the 2016 study by Bik et al. found the presence of repeat offenders in nearly 40% of published articles, indicating that individual researchers are more likely to continue committing fraud if they get away with it the first few times. The Retraction Watch database provides further evidence for this. Additionally, the widespread availability of image modification software and lack of ethical awareness among many researchers adds to the increasing pervasiveness of data manipulation.
How do you detect fabrication or falsification?
The most common method for detecting fabrication or falsification is by manual analysis of manuscripts. I spoke with Dr. Elisabeth Bik, a Science Integrity Consultant at Harbers Bik LLC, whose sharp eyes started her on a career in research integrity in 2019. She’s an expert in detecting image manipulations and has been at the forefront of reporting cases of research misconduct as a science sleuth. She says, “Until two years ago, I’ve used just my eyes to look for [image] duplications. . . . I’m starting now to use software to help me detect these duplications”.
Some such software that Dr. Bik uses are Forensically and Imagetwin. But as helpful as these software might be, they can come with their own set of limitations. As Dr. Bik notes, Forensically, a free online photo forensics tool, is “very good in finding pixel-level duplication in images, but you can only feed it one image [at a time]”. Imagetwin, an artificial intelligence (AI)-based software, is slightly more advanced as it can scan an entire PDF. “It will find all the photographic images and compare them to each other”, explains Dr. Bik. However, she notes that using the software to detect inappropriate image duplications in a time series of images can be challenging as it will more often flag appropriate duplications. Proofig is another AI-based software that is comparable to Imagetwin in its detection capabilities and limitations. Consequently, Dr. Bik believes that such software “still need a human with some experience and brain to interpret these [results] . . .Currently it’s not at that stage yet where you can completely rely on it”.
What are the consequences of research misconduct?
Consequences for the scientific community and society:
Advances in research undeniably depend on pre-existing scientific literature. However, if this literature is being constantly contaminated by fabricated and falsified data, the basis for many of researchers’ hypotheses becomes shaky and they inadvertently end up following incorrect leads. This not only undermines their hard work, but also causes loss of billions of dollars in research funding that could’ve been spent on genuine research and led to new discoveries or therapeutics.
Ultimately, it’s the society that suffers the definitive aftermath. For example, Joachim Boldt, a German anesthesiologist and research leader, published at least 7 studies that claimed the effectiveness of a compound called hetastarch, used to stabilize blood pressure in patients. The studies were found to be misleading and later retracted. A meta-analysis of existing clinical trials on hetastarch found that excluding Boldt’s studies revealed a modest but significant increase in mortality and acute kidney injury in patients given the compound. Boldt’s studies have, thus, not only misled the scientific literature, but may have also affected patient outcomes. Another such example is that of British physician Andrew Wakefield, who authored the 1998 paper claiming increased risk of autism in children given the MMR vaccine (used against measles, mumps and rubella). The study seeded public doubt in the MMR vaccine, causing vaccination rates in the United Kingdom to drop in the mid-2000s. More importantly, the study fueled the global anti-vaccine fire that continues to this day and most likely costs thousands of people their lives each year. Dr. Bik sullenly notes, “[The study] is a prime example where misconduct has led to probably kids being sick with measles. Measles are on the rise again in the US and most of these kids are either unvaccinated or not properly vaccinated . . . So it’s a new problem and pretty unnecessary”.
In yet another example, several instances of image manipulation were discovered in a landmark 2006 study on Alzheimer’s research authored by Sylvain Lesné, a researcher at the University of Minnesota. So far, no other research groups have been able to reproduce Lesné’s findings. The study is one of the most cited Alzheimer’s studies of the century and has led to years of follow-up work by other research groups. The discovery that the work has been tampered with implies wastage of millions of dollars in federal funding, misleading of a significant line of Alzheimer’s research, and years of wasted work by researchers building their hypotheses off Lesné’s work.
Consequences for researchers involved:
In most instances where journals catch fakery in individual papers, the result isn’t a retraction, but instead a painfully long process of communication between the editors and authors that ends up in just an ‘Expression of Concern’ or a correction without any explanation from authors as to the reason for the fabricated data. Amongst the papers that Dr. Bik reported to journals in 2015 with instances of inappropriate image duplication, she notes that by 2020 “two-thirds of these papers had not been corrected or retracted.” This forgiving attitude of journal editors and organizations causes further insult to the scientific establishment by propagating the idea that there is no punishment for faking science.
However, there exist rare cases wherein researchers face significant penalties for their actions. Andrew Wakefield was stripped of his medical license in 2010 after his paper was found to contain a significant volume of fabricated data. Joachim Boldt has had 164 of his publications retracted and faced a criminal investigation for research forgery that ultimately cost him his professorship and medical license. Canadian endocrinologist and former Professor of Medicine, Sophie Jamal, received a lifetime funding ban from Canadian science funding agencies and had her medical license revoked after an investigation revealed significant data fabrication in her publications showing a positive effect of the compound nitrogylcerin in the treatment of osteoporosis.
In a rare case scenario, Dong-Pyou Han, a former assistant professor at Iowa State University, US, was sentenced to nearly five years in prison for fabricating data in clinical trials for an experimental HIV vaccine. Han was also ordered to return $7.2 million in funding expenses to the National Institutes of Health.
What steps can we take to combat research misconduct?
Science publishing needs a rigorous and collaborative effort to weed out fabricated and falsified research. It thus becomes essential to address research misconduct using a two-pronged approach: strengthening the pre-publication review process and responsibly investigating studies flagged post-publication. In line with the former approach, journal editors are increasingly updating their policies to include a requirement for raw data during manuscript submission. Many editors have now revamped their review process by hiring data/image integrity specialists like Dr. Bik. There are also ongoing efforts to automate some of the screening process, with firms like LPIXEL in Japan, Proofig in Israel and Resis in Italy developing software that can spot image duplication or manipulation within or between manuscripts. The next step would be to increase the capabilities of such software to screen papers on a massive scale, which would ultimately require a large shared database of all existing publications, especially ones flagged for misconduct.
One such effort toward this goal was made by two longtime health journalists—Ivan Oransky and Adam Marcus. They assembled an enormous list of retracted papers and conference abstracts and released it in 2018 in the form of a searchable open-source database. This database is now the largest and most comprehensive of its kind and contains over 42,000 retractions dating back to the 1970s.
When it comes to post-publication quality control, journals have a history of ignoring most potential cases of misconduct flagged by sleuths or other readers. This culture is now changing but at a much slower rate than would be expected of scientific publishing entities. In recent years, journals including RSC Advances and FEBS Letters have retracted several of their published articles found to be a result of paper mills. In an attempt to aid journals’ efforts in addressing misconduct and keeping everyone in the loop, COPE has issued several clear guidelines to address post-publication critiques, suspected image manipulation, data fabrication, paper mill activity, etc. COPE also provides advice and guidance on the process of retraction, including what details the retraction notices should provide. Additionally, enlightening the scientific community that a retraction doesn’t always imply research misconduct, but can also result from honest mistakes, will encourage both authors and publishers to be more open and forthcoming about problematic papers.
What hurdles exist in combating misconduct?
Despite the multi-step approach to tackling research misconduct, many issues remain to be addressed. For example, when it comes to increasing awareness about acts of misconduct between journals or even more publicly, publishers are cautious about what author information they can share without the risk of being defamatory. Moreover, data-protection rules can hinder the sharing of author’s research or personal data between journals. As a result, if one journal rejects a manuscript that it suspects as being fabricated, the authors can just submit it to another.
Additionally, although the requirement for raw data submission was thought to be one of the ultimate tools against data fabrication pre-publication, publishers are now starting to realize that even raw data can be fabricated. Dr. Bik explains, “I’ve seen people send in photoshopped ‘originals’ where they just photoshopped something around it and they say ‘Oh yeah, but these are the original blots’. I’ve seen editors fall for that, like ‘Okay well the authors sent in the originals so we believe it.’ And I’m like ‘that’s photoshopped, come on!’”
Another more serious threat to research integrity is advanced AI, which was recently shown to be capable of generating fake scientific images that were indistinguishable from real ones even to the eyes of experts. Moreover, despite the development of effective algorithms that could detect AI-generated fake images, there is widespread fear that these detection methods will quickly become obsolete as deep learning technologies evolve faster.
Conclusion:
Contamination of scientific literature, whether by error or intention, hinders scientific as well as societal progress. Although research misconduct in all its forms is difficult to completely eradicate, there are concrete actions that can be taken to curb its spread. These include increased scrutiny of manuscripts pre-publication, making post-publication instances of misconduct more open, raising awareness about acceptable methods of image modification, and establishing research integrity policies that penalize researchers with proven instances of fraud. Moreover, destigmatizing retractions and corrections will go a long way in establishing trust in the peer review process. Although some of these actions are currently being instituted by certain journals or countries, a more extensive and concerted reform is still lacking. But all hope is not lost. As Dr. Bik notes, “Publishers are now more and more convinced that some fraction of their papers are going to be fraudulent and they need to be paying more attention to that. And I think they’re also taking a harder stance, so that’s slowly changing. But it is changing and I have good hopes in that direction.” Some researchers have taken it upon themselves to help facilitate this process by establishing platforms (e.g., COPE) to provide global guidelines on ethical publication practices, and databases (e.g., Retraction Watch) to keep the scientific community updated on retracted studies. The next few decades will be a defining era in scientific ethics and one can only hope that the outcome is in favor of research integrity.