To our knowledge, this is the first investigation focusing on retracted articles in dentistry and its relevance is evident in the face of the rising number of retractions, which represents a challenge to be faced by the scientific community. Perhaps the most relevant result of this descriptive study is that the proportion of retracted papers in dentistry is low compared to the biomedical literature indexed in PubMed, as revealed by our exploratory findings. We could suggest some reasons for this: there is a time-lag of at least 3 years in PubMed notices of retraction [20]; the retraction of papers in this field of knowledge is a recent occurrence when compared with the biomedical literature as a whole; inappropriate or fraudulent data can be very difficult to detect [23]; dental journals have been slow to endorse well-recognised reporting guidelines [24] that could facilitate the identification of ethical problems in the manuscripts. Since there is lacking information on this topic in the field of dentistry, it is difficult to discuss the findings of our study within the specific context of the dental literature. Hence, comparison across studies and recommendations derived from our findings were presented and discussed within the perspective of the biomedical literature and should be considered in a broader standpoint of ethics and integrity in scientific research.
In the present study, the most observed reasons for retraction, from highest to lowest occurrence, were: redundant publication, plagiarism, misconduct, overlap and honest error. On the other hand, other study that identified 395 retractions in English language publications indexed in MEDLINE© between 1982 and 2002 showed that 61.8% were due to unintentional errors and 27.1% to misconduct [25]. Wager and Williams [4] observed that 28% of retractions were due to honest errors, 17% to redundant publication and 16% to plagiarism. More recently, an analysis of articles retracted in the scholarly literature in 2012 and 2013 demonstrated that the most frequent reasons for retraction were mistakes, plagiarism, and duplicate submission [19]. However, an analysis of 2047 retracted articles, as of May 3, 2012, revealed that 67.4% of them were due to misconduct and 21.3% to error [5]. For these authors [5], misconduct comprised fraud (confirmed or suspected), duplicate (redundant) publication and plagiarism. The lack of uniformity in the definition of ethical issues in relation to the retraction of articles [4] makes more consistent comparisons difficult.
We observed that redundant publication and plagiarism were more reported in journals with cites/doc under 2.0, whilst overlap was more frequent in journals with cites/doc under 1.0. The most observed reason for retraction in journals with cites/doc of 2.0 and above was scientific misconduct. These results are in accordance with the ethical issues found in 2047 retracted notices reported in PubMed [5]. Perhaps journals with smaller cites/doc do not routinely use software to detect plagiarism, redundant publication or overlap. In fact, software such as iThenticate does not effectively detect plagiarism. It detects similarities among texts, so skill is needed to interpret the report [26]. By contrast, journals with higher cite per doc indexes would prevent replication of text early in the process. However, reviewers, editors and the research community would find scientific misconduct more difficult to identify. Interestingly, the higher occurrence of retractions was observed in journals with cites/doc between 1.0 and 2.0.
The issues of redundancy, plagiarism and overlap deserve more discussion, if we consider that publishing a paper is very relevant for authors’ professional advancement. Sometimes, the results of a study may have different implications and outcomes might deserve being reported in several papers [23]. For those cases to be ethically acceptable, authors should disclose the links between that specific report and a wider study, at submission and in citations [23]; if not, that could be seen as “salami” publication. Redundant publications, however, are not advisable unless justified by authors and agreed by the editors of both journals. Although, when authors use a methodology that is quite similar to another study already published, clear cross-referencing is advisable to prevent software to identify repeated text as plagiarism. Actually, editors usually suspect plagiarism when a manuscript receives a 35% or higher score for similarity in iThenticate Plagiarism Detection Software [26]. Additionally, there is no agreement among editors regarding accepting papers that were previously presented at a scientific meeting or on the website of an academic institution. Therefore, it is important that authors disclose any prior reporting during the submission process [23].
Another aspect observed in our study was the notable occurrence of unclear retraction notices, which was also reported in a study of all retractions published in 2008 in PubMed under the publication type “retraction of publication” [10]. Besides being unclear, some retractions failed to distinguish between error and misconduct. In addition, retractions can be difficult to find and are not always available on the websites and in the databases of journals [9]. Moreover, attention was drawn to the need to use well-defined criteria, such as the COPE retraction guidelines, which would be important in avoiding improper and unwarranted retractions [2]. As the retraction topic is a relatively recent discussion for the biomedical literature [5], the idea that the retraction process should be extensively clarified among publishers, editors, authors and institutions [2, 8] should also be highlighted in the field of dentistry.
Our investigation also found that most of the retractions resulted from laboratory studies, followed by case reports and review articles. Similar findings were observed in other studies [4, 19], and one group of authors claimed that there is a greater propensity for error in the handling of results in experimental studies [19]. Regardless of the nature of the study, ethics and integrity must be maintained by the researchers, because, as already stated, patient care is at risk [20].
Data extracted in our study showed that retractions do not have country borders regarding the origins of authors and of journals. In the same vein, other studies have found that errors in reported research are a global issue [5, 13]. However, a survey of the general biomedical literature showed that the origin of the authors was related to the reason for retraction: the rates of retractions due to fraud was higher in the United States, German, Japan and China, whilst plagiarism was more common reason in India and China [5].
The time interval between the publication of the original article and the retraction notice varied considerably, ranging from 1 month to 37.6 years. Another study observed a mean time of 2.8 years in 2012, which reduced to 2.2 years in 2013 [19]. In addition, it was reported a gradual upward trend in the time for retraction over time [5]. Interestingly, this is not influenced by the journals’ impact factor [5, 6].
This study has strengths when compared to previous investigations in the biomedical literature. We based our specific search on three databases: SCImago, Retraction Watch and PubMed. SCImago is developed from Scopus, which is a database that indexes a larger number of journals than PubMed, Web of Science and Google Scholar [27]. We also attempted to look for other terms for identification of retraction notices such as “withdrawn” and “retracted” in the title of the publication, and, when the retraction notice was incomplete, we tried to identify the details of the retraction on the Retraction Watch website and other online reports.
One limitation of this study is that we did not follow up the articles retracted. How readers should deal with retracted papers is a debatable issue, especially when considering articles with conclusions of relevant scientific value [5, 10]. Some authors advocate that “improperly obtained” data, even if considered as “valuable” data should not be published, preventing future unethical research [28]. In this context, the decision to continue citing retracted papers is also questionable, since some reasons that lead to the retraction do not necessarily involve the validity of the methodology and then the results might somehow be useful to the scientific community.
We acknowledge that ‘to err is human’ and it is crucial to have the debate about research ethics in dental science reporting disseminated extensively to the different people involved. The ultimate purpose of this paper is to make editors, authors and overall readership aware of the ethical issues related to scientific investigations in dentistry, based on the understanding of the ideal retraction notice as a historical document [29]. Also, it is important to avoid stigmatising authors who made genuine mistakes along with those who have committed misconduct; that justifies the need for clear and standardised retraction notices [4]. Do the present findings represent the tip of the iceberg of ethically questionable publications in dentistry, or have the vast majority of articles in this area been ruled by ethics? The methodology used in this study does not allow for a response. However, our findings, although seemingly mild, and the lack of standardisation in retractions reporting, strongly suggest that researchers in the dentistry field should pay more attention to scientific integrity in their academic routines.