Retracted publications and their citation in dental literature: A systematic review

Abstract Objectives The present systematic review aimed to perform an in‐depth analysis of the different features of retracted publications in the dental field. Material and methods This review has been recorded in the PROSPERO database (CRD42017075634). Two independent reviewers performed an electronic search (Pubmed, Retraction Watch) for retracted articles in dental literature up to December 31, 2018. Results 180 retracted papers were identified, the first published in 2001. Retractions increased by 47% in the last four‐year period (2014–2018), when compared with 2009–2013 (94 and 64 retracted publications, respectively). Author misconduct was the most common reason for retraction (65.0%), followed by honest scientific errors (12.2%) and publisher‐related issues (10.6%). The majority of retracted research was conducted in Asia (55.6%), with 49 papers written in India (27.2%). 552 researchers (89%) are listed as authors in only one retracted article, while 10 researchers (1.6%) are present in five or more retracted publications. Retracted articles were cited 530 times after retraction: the great majority of these citations (89.6%) did not consider the existence of the retraction notice and treated data from retracted articles as reliable. Conclusions Retractions in dental literature have constantly increased in recent years, with the majority of them due to misconduct and fraud. The publication of unreliable research has many negative consequences. Studies derived from such material are designed on potentially incorrect bases, waste funds and resources, and most importantly, increase risk of incorrect treatment for patients. Citation of retracted papers represents a major issue for the scientific community.


| INTRODUCTION
The publication of unreliable medical research has many negative consequences. Studies deriving from such material are designed on potentially incorrect bases, waste funds and resources, and most importantly, increase risk of incorrect therapy for patients.
The retraction of a scientific article may be decided by the journal editor or be requested directly by the author when validity of the research and its findings is seriously compromised. Reasons for retraction are usually related to authors' behaviour, either fraudulent (e.g. misconduct, plagiarism, intentional duplicate publication) or honest (e.g. methodological errors), and can sometimes be related to publisher issues (e.g. accidental duplicate publication).
Even if retractions are uncommon in medical literature, their number has significantly increased in recent years (Cokol, Ozbay, & Rodriguez-Esteban, 2008;Steen, 2011). This could be due to tighter and more accurate control by scientific journals over acceptance and publication processes, including strict peer review and use of specific software to identify plagiarism. This tendency has also been confirmed in the field of dentistry. A recently published analysis showed that 57% of retractions in dental literature were performed after 2012 (Faggion Jr, Ware, Bakas, & Wasiak, 2018).
Unfortunately, articles are often cited even after their retraction. In both medicine (Budd, Sievert, Schultz, & Scoville, 1999;Neale, Dailey, & Abrams, 2010) and in dentistry (Faggion Jr et al., 2018) approximately 60% of retracted papers continue to be cited in literature. In these cases, however, citations may be appropriate (indicating the presence of a retraction in the manuscript and bibliographic references, and possibly discussing the questionable value of the reported data) or inappropriate (treating the findings of the retracted publication as reliable).
Therefore, the aim of the present systematic review is to perform an in-depth analysis of the different features of retracted publications in the dental field, with particular attention paid to the presence and appropriateness of the citations received after retraction.

| Search strategy
The present review has been recorded in the PROSPERO database (www. crd.york.ac.uk/PROSPERO) with registration number CRD42017075634.
Methods for conducting this analysis are derived from previous reviews on retracted articles in medicine and from the Cochrane Handbook for Systematic Reviews (Higgins et al., 2019).
An electronic search was conducted on Pubmed Central (PubMed, www.ncbi.nlm.nih.gov/pubmed) by two independent authors (A.R. and T.L.), selecting articles published from database inception up to the latest access on December 31, 2018. In addition, the Retraction Watch website (www.retractiondatabase.org) was carefully browsed for retracted studies in the dental and oro-maxillofacial field. No language restriction was applied in order to limit selection bias.

| Search
Search in the selected electronic database was performed by using the following algorithms:

| Selection of studies
Two blinded authors (C.S. and F.B.) independently performed eligibility assessment of the studies. Inter-examiner reliability in the study selection process was assessed using the Cohen k-test assuming a threshold value of 0.61 to indicate substantial agreement (Landis & Koch, 1977). Eventual conflicts were resolved by discussing each article until consensus was reached. When necessary, an attempt to contact the Editors and the Authors of the included studies was made in order to retrieve any missing information or to clarify specific items.
The following inclusion criteria were used: 1 any topic related to dental sciences, oral surgery and oral pathology; 2 all levels of scientific evidence; 3 basic science, animal and human studies.
The following exclusion criteria were applied: 1 topic unrelated to dental research; 2 use of the words "retraction" or "retracted" with a different meaning from the one considered by the present review; 3 duplicate papers.

| Statistical analysis
An independent examiner (R.D.L.) analysed all datasets with statistical software (Statistical Package for Social Sciences v.15, SPSS Inc., Chicago, IL). Descriptive statistics presented parametric continuous variables as mean ± SD, non-parametric continuous variables as median with interquartile ranges and discrete variables as counts or proportions. The association between retraction characteristics and journal IF was analysed with a univariate linear regression model entering reasons for retraction as a dummy variable, followed by Bonferroni post hoc test for multiple comparisons.

| RESULTS
Electronic search resulted in a total of 12,294 records (12,137 in Pubmed and 157 in Retraction Watch) and, after removing duplicates, a total of 12,151 publications were screened. 11,971 articles were excluded after examination of titles and abstracts and 180 papers were included in the final analysis (inter-reviewer agreement = 0.98).
The results of the electronic search are summarised in Figure 1. The list of retracted articles with their respective retraction notices is reported in the Supporting information (Table S1).  Table 1. Oral pathology is the subspecialty with the highest number of retractions (n = 50; 27.8%), followed by implantology (n = 31; 17.2%) and periodontology (n = 23; 12.8%). The complete list is presented in Table 2. In terms of geographical distribution, the majority of retracted research was conducted in Asia (n = 100; 55.6%), with 49 papers written in India (27.2%). Complete data for continent and country are listed in Table 3.
Plagiarism related reasons (duplication, overlapping with previous works) represent the main cause for retraction in journals with IF < 2, while unreliable and/or fabricated data related reasons prevail in journals with IF > 2 ( Figure 3). Nevertheless, no statistically significant associations were demonstrated between reasons for retraction and journal IF (p = .5496).

| Citations
The articles included in the present systematic review (n = 180) were cited 530 times after their retraction (mean 2.94). Citations in new

| DISCUSSION
The retraction of scientific publications has constantly increased in recent years (Steen, 2011;Steen, Casadevall, & Fang, 2013). However, it remains unclear if this trend is mainly the result of increasing intentional scientific misconduct or the result of improvements in the detection of unreliable articles (Cokol et al., 2008), due to the adoption by many journals of the guidelines suggested by the Committee on Publication Ethics (COPE) (2009). Regrettably, it has to be said that many cases of scientific misconduct may remain undetected. A metaanalysis of surveys conducted among scientists showed that approximately 2% admitted to having fabricated, falsified or modified data or results at least once, and 34% admitted other forms of intentional scientific misconduct, such as graph manipulation and unjustified removal of data from final analyses (Fanelli, 2009). The findings of the present systematic review confirm, in accordance with previous works, that dental literature is currently following the same trend (Faggion Jr et al., 2018;Nogueira, Gonçalves, Leles, Batista, & Costa, 2017). Retractions increased by 47% in the last four-year period Intentional misconduct resulted by far the most common reason for retraction (65.0%), followed by honest scientific errors (12.2%) and publisher-related issues (10.6%). These findings are in accordance with the results of recent reviews in both dental (Faggion Jr et al., 2018) and in other medical fields (Chambers, Michener, & Falcone, 2019;Rai & Sabharwal, 2017;Rosenkrantz, 2016). Certain characteristics of retracted articles seem to be related to journal IF, even if these associations do not reach statistical significance. Plagiarism related issues are the most common reason for retraction in journals with IF < 2, while inaccurate or falsified research is the prevailing motivation in journals with IF > 2. This is potentially due to the fact that high quality journals have more efficient tools to detect plagiarism at an early stage of manuscript evaluation and to perform more accurate post-production control.
The majority of retracted research was conducted in Asia Finally, it is highly desirable that editorial offices perform electronic screening of references citing retracted articles in all newly-F I G U R E 4 Appropriateness of post-retraction citations submitted articles, in order to reduce persistence of error in future studies. Cosentino and Veríssimo (2016) proposed that editorial offices utilise a database of retracted articles for cross-checking purposes to prevent citation of retracted publications.
The system of research evaluation currently adopted in many countries should be seriously reconsidered. Current methods based mainly on the number of works and citations tend to recognise quantity more than quality. Analogously, subject to the same methodologic criteria, the 17th century Dutch master Johannes Vermeer, recognised as having painted only 45 unique masterpieces, would have disappeared entirely from art history, being surpassed and obscured by numerically more prolific contemporary artists producing greater numbers of aesthetically negligible and historically irrelevant paintings.