Research Article: Post retraction citations among manuscripts reporting a radiology-imaging diagnostic method

Date Published: June 13, 2019

Publisher: Public Library of Science

Author(s): Sorana D. Bolboacă, Diana-Victoria Buhai, Maria Aluaș, Adriana E. Bulboacă, Wolfgang Glanzel.


Our study aimed to evaluate the trends of post retraction citations of articles reporting a radiology-imaging diagnostic method and to find if a different pattern exists between manuscripts reporting an ultrasound method and those reporting other radiology diagnostic methods. This study reviewed retractions stored in PubMed on the subject of radiology-imaging diagnosis to identify the motivation, time from publication to retraction, and citations before and after retraction. The PubMed database was searched on June 2017 to retrieve the retracted articles, and the Scopus database was screened to identify the post-retraction citations. The full text was screened to see the type of post-retraction citation (positive/negative) and whether the cited article appears or not as retracted. One hundred and two retractions were identified, representing 3.5% of the retracted articles indexed by PubMed, out of which 54 were included in the analysis. Half of the articles were retracted in the first 24 months after publication, and the number of post retraction citations was higher than the number of citations before retraction in 30 out of 54 cases (US methods: 9/20, other diagnostic methods 21/34, P-value = 0.2312). The plagiarism was the most common reason for retraction (31%), followed by repetitive publication (26%), and errors in data/manuscript (24%). In less than 2% of cases, the retracted articles appear as retracted in the text or reference list, while the negative citation is observed in 4.84% among manuscripts reporting an US diagnostic method and 0.32% among manuscripts reporting a diagnostic method other than US (P-value = 0.0004). No significant differences were observed when post retraction weighted citation index (WCI, no. of citations weighted by citation window) was compared to WCI prior retraction (P-value = 0.5972). In light of the reported results, we enumerated some recommendations that could potentially minimize the referral to retracted studies as valid.

Partial Text

Research is the primary source of knowledge in medicine and publication is the principal instrument for dissemination of research results. Furthermore, scientific productivity in academia is traditionally assessed based on the number of publications. Consequently, new scientific journals appear even if some of them are classified as ‘predatory’ (lacking integrity in the publishing process usually without an accurate peer review process, fake impact factor or editorial team, issues that could be difficult to detect) and the number of published scientific articles increases annually [1–3]. The retraction of an article is used to alert scientists to serious problems identified with a published article and follows the Committee on Publication Ethics retraction guideline [4]. According to this guideline [4], a retraction should be considered if clear evidence exists that ① the findings are unreliable because of misconduct (such as data fabrication, data falsification, etc.) or ② honest errors (such as experimental error, or miscalculations), ③ the findings have previously been published (redundant publication), ④ the text constitutes plagiarism or ⑤ the article reports unethical research. The retraction notice should be linked to the retracted article and clearly labeled by including the title and the authors in the retraction heading, designating the retraction, and specifying a clear reason for retraction [4,5].

PubMed database, involving PubMed Clinical Queries was searched to identify retracted articles using the following keywords: (Diagnosis/Broad[filter]) AND ((diagnostic or diagnosis) and (imaging or radiology)) AND ((Retracted Publication[sb] OR Retraction of Publication[sb]) AND Humans[Mesh] AND English[lang]))). No limit of years was imposed, and we only looked after publications in English that reported research on humans. The PubMed Clinical Queries was searched on June 15, 2017. The retrieved documents were first screened by reading the title and the abstract. The abstract was read from PubMed and, whenever available, the journal web page was screened to retrieve the abstract. Articles presenting a diagnostic-related technique such as radiography, radioscopy, ultrasonography (US, all types), computer tomography (CT), magnetic resonance imaging (MRI), positron emission tomography (PET), single photon emission computed tomography (SPECT), or scintigraphy were included in the analysis. The following data were collected for each publication included in the analysis: year and month of publication, the reported diagnostic imaging method, the journal, no. of authors, the authors’ affiliation country, month and year of retraction, the reason for retraction, number of citations until and after retraction. The Journal Citation Reports was used to retrieve the domain and the rank of the journal whenever applicable. The time expressed in months from publication to retraction and from investigation to retraction was calculated based on the collected data. The citation pattern is influenced by many factors, the age of the paper and the citation time windows being the deciding factors [27–29]. The weighted citation index (WCI) was defined as the ratio between the number of citations and the citation time window (no. of months/12). The WCI was calculated both prior to (WCIprior) and post (WCIpost) retraction to analyze the dynamic of citations.

A total of 2,897 articles indexed as a publication type of “retracted of publication” or “retraction of publication” published until June 15, 2017, on human species and in the English language were identified under the Diagnosis Clinical Study Categories, PubMed Clinical Queries. One hundred and two articles (3.5%) were retrieved for the searched keywords, and 54 were finally included in the analysis. Forty-eight articles were excluded and the reasons for elimination are given in Fig 1.

Post retraction citations proved to be a current practice among articles reporting radiology diagnostic techniques, and in some cases with more post retraction citations compared to citations prior to the retraction. Furthermore, in most cases, the notice of retraction did not appear either in the text or in the reference list, with a low percentage of negative citations.

Our study provides the first insights regarding the post retraction citations in radiology-imaging diagnostic methods scientific literature available in PubMed. The retractions are mainly due to plagiarism, repetitive publication and error in data. The post retraction citations are twice as frequent for the whole sample of analyzed manuscripts but are halved for those manuscripts that report an ultrasound diagnostic method. Concerning the citation window (WCI), there was no significant difference between the number of citations before (WCIprior) and after (WCIpost) retraction. In less than 2% of cases, the retracted manuscript appears as retracted in the reference list or in the text of the manuscript and in less than 5% the citation of a retracted manuscript is negative.