What Can We Learn from the Retraction of Medical Articles

Abstract

Article retractions happen across all types of research and can happen to any researcher at any institution. Retracted articles can have negative consequences, particularly when they are used in meta-analyses or clinical guidelines. Peer review is an important though insufficient means of safeguarding academic integrity, and thus all researchers, including trainees, are responsible for helping to ensure integrity in research.

The medical literature provides essential information for clinical care and scientific research. Most readers assume the published articles represent the best efforts of the journals and the authors. However, flawed research and publications do occur and can have important negative effects. In recent years, there has been concern over increased rates of article retractions [1]. High-profile reactions, such as those occurring during the COVID-19 pandemic, have heightened concerns over the quality of medical and scientific research [2]. This has been emphasized in ongoing discussions about research integrity, including the need to appropriately train the next generation of researchers about ethical conduct in science. Trainees in medical education and graduate medical education programs frequently engage in research [3, 4]. While research ethics are often discussed in graduate science courses, they should also be taught to trainees in medical and graduate medical education programs as well. This review of retracted medical articles provides 10 important lessons learned from retractions which can assist medical educators in instructing their trainees about research ethics and best practices:

1. Retractions Occur Periodically in Most Medical Journals, Even in Journals with Excellent Reputations.

   For example, The Lancet retracted an article on chloroquine and hydroxychloroquine treatment in COVID-19 infections in 2020 [5]. The Lancet also retracted a 2010 paper claiming a link between the MMR vaccine and autism [6]. In fact, some higher-impact journals have been shown to retract papers more frequently than others, perhaps due to increased scrutiny or competition after acceptance against stiff competition for journal space [7]. Therefore, journal prestige is no guarantee that a study is free from error or even fraud [8, 9]. In addition, the reputation of the author or institution does not provide certainty about the quality of the work. Scientific misconduct and retractions have involved Harvard faculty, society presidents, medical school deans, and world-renowned experts [10, 11]. While impact factors and reputation can serve as heuristics of quality, they are clearly no guarantee.

2. Retractions Occur in Most Types of Medical Articles, Including Clinical Trials.

   Clinical trials are retracted at the same rate as other medical articles; a 2011 study found that one in 6070 clinical trials was retracted between 2000 and 2010 [12]. This should be particularly concerning for physicians, whose current practices depend on evidence-based medicine and often rely heavily on these trials.

3. The Timeframe for the Retraction is Typically Months to Years.

   Time to retraction can vary widely, but it usually takes about 2 years [13, 14]. In some instances, retraction may take far longer. And while some studies were retracted rather quickly during the pandemic (some within days of publication) [15], this is more the exception than the rule. This occurs in part because the investigation of research integrity and misconduct often involves both the journal and the authors’ academic institutions [8]. A journal may issue an expression of concern while an article is investigated [16]. Some investigations have the cooperation of the authors, but others do not.

4. Retracted Articles May Continue to be Cited Long After their Retractions.

   This may seem paradoxical, since a retraction is intended to limit the spread of misinformation. However, multiple studies have reported that retracted articles continue to be cited after their retractions [17]. Some of these citations probably discuss the retraction process itself; in other cases, the retracted papers’ results continued to be viewed positively and worth citing [18]. Articles that cite now retracted articles are also rarely updated. This includes systematic reviews and meta-analyses [19].

   One study noted that 20.3% of the nearly 1400 retracted papers reviewed were cited in pharmacy systematic reviews [20]. Another study, for example, examined the influence of 12 retracted clinical trials, which continued to be cited in 68 reviews, guidelines, and clinical trials, including systematic reviews and meta-analyses. The authors determined that 13 guidelines and other published review articles would likely have changed their conclusion based on the subsequent retraction for the above-mentioned studies, with only one of the 68 citing publications having taken any corrective action [21]. It is difficult to measure the true impact that retracted research has on patient care, and in some cases, it is unlikely that the recommendations from one faulty trial have had a significant impact. However, the mere risk of harm to patients should raise concern by physicians and researchers.

5. The Reasons for these Retractions are Varied.

   These can include simple mistakes, such as duplicate publication, faulty statistical analysis, or more serious errors, such as plagiarism, misconduct, fraud, and data manipulation, including the use of databases with uncertain validity [22]. In many cases, the reason for retraction remains unknown. Not all cases are due to fraud. Dr. Linda Buck, a Nobel Laurette, retracted two papers after her own lab’s investigation revealed unreproducible findings and inconsistent figures. The common denominator was the misconduct of a former postdoctoral fellow who had since left the lab [23]. Indeed, a retraction does not necessarily mean that misconduct was present, though it is often the case. In some instances, the authors will voluntarily request an article be retracted [24]. Nevertheless, the long list of retracted articles highlights the vigilance that researchers must have during their work with colleagues and the importance of correction of any errors in their work.

6. Research Misconduct is Disturbingly Common.

   A 2009 survey of scientists found that 1.97% reported having manipulated data (including outright fabricating or falsifying results) at least once. Other problematic research behaviors, such as changing study design or results due to pressure to gain research funds (15.5%), reports drafted by pharmaceutical or contract organizations (29.6%), and premature termination of a study by a company (33.7%), were much higher. Up to 72% of respondents reported colleagues engaging in questionable or dishonest practices [25]. Other studies have noted that research misconduct, including fraud, is not infrequent among researchers [26]. These include questionable practices and forms of misconduct that may be less obvious to the researcher, such as excluding less favorable results, “rounding off” p values, or ending a study prematurely after the desired results are obtained [27].

7. Journals Cannot Function as the Ultimate Safeguard for Research.

   While journal publication may be seen as certification that the research is “correct,” this is inaccurate. True, peer reviewers can scrutinize submitted manuscripts for erroneous conclusions and poor methodology. Some journals even employ statistical reviewers, who look for data and analytical anomalies [28]. And the more recent push for more transparent science, including public data availability via online repositories [29], helps circumvent challenges faced by curious readers in obtaining data from study authors [30]. All these measures help identify and correct problematic research. Nevertheless, reviewers and publishers cannot possibly replicate studies. And while more efforts are being used to scrutinize data, in many cases, it is the scientific community that has sounded the alarm on problematic articles. For example, researcher Nick Wise began noticing articles with so-called tortured phrases, or wording that is indicative of computer generated text, and subsequently identified hundreds of problematic papers that were subsequently retracted [31].

8. Retractions Have Definite Negative Consequences.

   The information in retracted articles may lead to mistakes and harm in medical care. They cost money, both in research money lost and in money and time spent on investigating misconduct [32]. These articles damage research careers, even in instances in which the misconduct was apparently isolated. They also undermine confidence in medical research and medical literature by both healthcare workers and the public. This retracted research could cause harm, but determining the effect of faulty research on patient outcomes is quite difficult. The COOPERATE trial, for example, which claimed to show the benefits of angiotensin receptor inhibitor and angiotensin II receptor blocker medications in patients with non-diabetic renal disease, was ultimately found to have serious faults, including lack of ethical approval and unverifiable data, and was subsequently retracted [33]. However, as a result of this study, it is estimated that over 100,000 patients received combination ACEi-ARB treatment [34]. Another example comes from fabricated research on the use of hydroxyethyl-starch (HES) colloids for resuscitation, in which the exclusion of several studies from one now disgraced author changed the results of a meta-analysis to show greater risk [35]. Hydroxyethyl-starch subsequently fell out of favor (the European Medicines Agency suspended its use in 2022 [36]) and the author in question lost his professorship and has since had 100 of his articles retracted [37]. Although one of the more “high profile” instances of academic misconduct, perhaps the most important lesson here is that anyone can conceivably become engaged in academic misconduct, with potentially disastrous results.

9. Medical Education Programs Must Provide Training for Residents to Limit the Development and Submission of Misleading Medical Information.

   There are several ways that this can be accomplished. All projects should be organized and planned based on relevant medical questions after reviewing the available medical literature. Experienced authors and investigators associated with medical education programs should provide advice and feedback throughout the development of a possible publication. The authors should review their manuscripts for significant similarity to other published literature using software programs, such as iThenticate. Authors should make certain that none of the references used in their manuscripts have been retracted; programs such as the reference manager Zotero (which is integrated with Retraction Watch) or Scite.ai can assist with this. Consent from patients, from institutional review boards and from organizations which maintain medical databases, is essential. Even case reports, which historically did not require consent so long as they ensured anonymity, are now frequently required by journals to have signed consent from the patient. The number of authors should be limited to only individuals who have made significant contributions to the project. Trainees should be aware of less obvious forms of research misconduct, including so-called self-plagiarism, or deliberately copying verbatim text from one’s previously published study to be used in another. Authors should be thoughtful about and critical of the studies they cite. All authors should review the final manuscripts and be able to express questions and concerns; even if only one author is responsible for research misconduct, all authors can be affected by retraction. Statisticians should be involved with study designs where possible to ensure rigorous methods and appropriate sample population, and that study conclusions are consistent with the data. The Committee on Publication Ethics (COPE) guidelines could be reviewed by residency programs; often, the most striking messages come not from research guidelines but case studies of research fraud and misconduct, which show the unsettling extent to which research ethics can be ignored (COPE offers a library of such cases).

10. Honesty and Trust a Crucial to the Scientific Process.

    While the scientific and publishing community can incorporate many safeguards to prevent and detect fraud, there is simply no way to ensure that all research is valid. Confidence from both the scientific community and the general public is also built upon trust in researchers. As one journal founder said, “scientific publishing is based on the absolute trust [in] the authors. Without such trust, the scholarly publishing will be groundless and meaningless” [38]. While this is certainly more abstract than plagiarism checkers or expert reviewers, it is essential for scientific inquiry and something that should be understood by students, trainees, and faculty.

    All medical professionals and the public have an interest in and can contribute to the validity and quality of medical knowledge. This process starts with study design and project completion, includes peer review for journals, and ends with careful review of published articles. Professionals at all levels and other interested individuals can participate in one or more of these steps. And all should see themselves as playing a role in safeguarding scientific integrity.

Declarations

Competing Interests

The authors declare no competing interests.