In 2013, Dr. Ramón Estruch and his team published a report on the PREDIMED study. By all accounts, the paper was well-received— it was published in the highly-regarded New England Journal of Medicine, and as of December 2020, it boasted 2735 citations in Web of Science–indexed journals. There’s just one issue: the paper was retracted in 2018 because of statistical anomalies and failure to properly randomize subjects. Worst of all, 816 of its 2735 citations (a whopping 33%!) came after the article was retracted.
This paper sits at the top of Retraction Watch’s December 2020 list of the most highly cited retracted articles, but unfortunately, this is far from an isolated incident. Retraction Watch’s database indicates that 4583 articles were retracted in 2021. The publication of this post, in fact, was spurred by a recent, far-reaching wave of retraction notices, including approximately 500 papers from Hindawi and Wiley, 500 papers from the Institute of Physics, 100 papers from PLOS ONE, and 50 papers from Elsevier. All of these notices were posted between August 3 and October 18 of 2022— in other words, about 1150 retractions in the span of less than 3 months.
In previous posts on this blog, we’ve discussed how many published articles and preprints can report false or misleading results. Now, we’ll talk about the typical end result when this is caught: retraction.
What are ‘retractions’ of published research?
In essence, a retraction is a formal declaration and removal of a manuscript previously published by a journal. This can come in the form of a complete retraction or as a republication, in which the problems identified in the manuscript are corrected (this is the case for Dr. Estruch’s study, though some reviewers still call for retraction of the republished manuscript). The Committee of Publication Ethics (COPE) provides widely-followed comprehensive guidelines that outline best practices for retractions, from identification to announcement. Notably, a retracted article remains available online for public viewing with a notice attached.
Retractions are most commonly issued directly by a paper’s authors but can also be issued by the publisher or journal editors. Reasons for retraction fall in one of three general categories: honest error, misconduct, or unclear/undistinguishable reasons.
Honest errors
Authors are often forced to produce papers within unrealistic time frames, and unfortunately, this can lead to errors. Honest errors include unintentional inclusion of incorrect data or inappropriate analysis techniques, inadvertent duplicate publication, or unreproducible results. However, retraction due to honest error faces an intense stigma of association with misconduct, often resulting in a researcher losing their status, respect, funding, or job.
Research misconduct
Research misconduct primarily encompasses intentional falsification or fabrication of data, analyses, or findings. It also includes plagiarism (sometimes including self-plagiarism), manipulation of scientific images, ethical misconduct, intentional duplicate publication, and inclusion of overstated or intentionally misleading conclusions. There are two particular types of research misconduct that have been gaining more attention over recent months:
- Peer review fraud. The combination of these problems has led to the inception of peer review rings, or coordinated networks of reviewers that authors can recommend who are guaranteed to provide fast, positive reviews regardless of the paper’s quality or potential ethical shortcomings. Alternatively, some fraudulent peer review can come in the form of reviewers providing feedback under a false name, sometimes even reviewing their own work.
- Paper mills. This is a short-hand term to describe a commercial entity that churns out falsified or manipulated research for profit. Researchers can hire a paper mill to produce a manuscript that they can publish under their own name (an unethical practice called ghostwriting), thereby increasing their authorship credits. This problem has become so troubling that a US congressional hearing was held in July 2022 to discuss the matter.
Unclear reasons for retraction
Sometimes, there’s just not enough evidence to distinguish honest error from misconduct. This catchall category can include authorship disputes, copyright issues, and nondisclosure of competing interests.
Key takeaways for researchers, administrators, and publication professionals
- Unfortunately, retractions won’t fix the problem of research misconduct. They’re an important step in the process of weeding out misconduct, but their retroactive nature leads to journals staying a step behind scam artists in their pursuit of a quick buck. Some publication specialists have begun devising AI that can detect potentially fraudulent papers, with the intention of proactively evaluating manuscripts before publication, but this technology is far from implementation.
- An important step for academia as a whole to take is to be more conscious about the stigmatization of retractions. Authors who notice errors in their own work are forced to choose between scientific integrity and the potential maelstrom of career-ruining backlash. By promoting a distinction between intentional and unintentional misconduct, we can reduce the hesitancy authors face when debating whether to voluntarily retract their own study.
- Be careful when citing studies. Thoroughly evaluate your sources. Recently, conversations have begun to mention retracted literature’s epistemic cost, or the proportion of research (especially review articles) that requires reevaluation or revision because it hinges on retracted literature. Even if the paper you’re citing hasn’t been retracted, it may rely on research that has been. Public opinion generally agrees that a retracted paper can still be cited if necessary, but it should always be accompanied by a clear note regarding the citation’s retraction.
