Retracted Publications in the Biomedical Literature from 2012-2018: An Overview

Background: The number of retracted publications in biomedical literature has increased dramatically in recent years. Many of these retractions are due to a faked review process and a large percentage of the retractions come from Chinese authors. The purpose of the current study was to determine the characteristics of retracted articles from 2012 to 2018 using the Medline database. Methods: The Medline database was searched to identify retracted publications from the year 2012 through 2017. Reasons for retraction were collected through PubMed, the Google search engine, and journal websites. Relevant data was collected for each retracted article and included the time from publication to retraction, the study type, the journal impact factor, and the institute of origin. Trends in characteristics related to retraction were determined. Results: Data from 4,043 retracted studies were included in the analysis. The most common reasons for retraction were fraud/suspected fraud (993), error (911), plagiarism (554), duplicate publication (547), and invalid peer review (384). The largest number of retracted articles was from China. Other countries with a large number of retractions were USA, Japan, India, Iran, Korea, Germany, and Italy. A faked review process, was responsible for a large number of retractions from China and Iran and from several journals. Conclusions: Misconduct, including fraud/suspected fraud, duplicate publication, and plagiarism, was the major cause for publication retraction. A faked review process caused the largest number of retractions in some countries. China has become the top country of origin for retracted publications.

Introduction review process, authorship disputes, and fraud/suspected fraud (Chen et al. 2018). Chinese authors contribute approximately one third of the retracted articles published in open access journals (Wang et al. 2018). Notably, some authors have retracted many articles. For example, the Japanese author, Yoshitaka Fujii, has retracted almost 200 publications since 2012 (Amarcus41 2012), the Korean author, Hyung-In Moon has retracted approximately 30 publications due to compromised review processes (Oransky 2012b), and Joachim Boldt from German who had 88 articles retracted in 2011 still had at least seven retractions after 2012 (Palus 2015). Information on the article's authors and institutions, country of origin, first author, journal name, year of publication, and year of retraction was collected. The time interval between publication and retraction was calculated according to the date on which each article was first available. The PubMed database, journal websites, and Google were used to search for retraction notices and reasons for retraction. Two authors independently analyzed retraction notices. Discrepancies were resolved through discussion. Cause of retraction was classified as fraud/suspected fraud, error, plagiarism, duplicate publication, authorship dispute, faked peer review process, unknown, and others (10). The number of citations of retracted publications was gathered from the Web of Science. Journal impact factors were based on the 2015 edition of Journal Citation Reports Science Edition. Journals without an impact factor were assigned a value of 0. Data analyses were conducted using SPSS 18 software. Data are presented as mean ± standard deviation (SD), number, and percentage. Chi-square tests were used to compare categorical variables. Differences in parametric data between two groups was calculated using Student's t tests.
Spearman's rank correlation tests were performed to evaluate the association between groups.
Statistical significance was set at p < 0.05.

Results
As shown in Figure 1, most retracted articles could be detected by searching "Retracted Publication" or "Retraction of Publication" in Medline. However, there were 102 (2.5%) retracted articles without a corresponding retraction notice and 13.8% of retractions were found only through searching "Retraction of Publication" (Figure 1). There were 4,043 retracted papers listed on Medline included in this study, with a one-year maximum of 804 in 2016 ( Figure 2).

Causes
There were 4,203 causes for retractions listed, with the most common reasons being fraud/suspected fraud, errors, plagiarism, duplicate publication, and a faked review process ( Figure 2). The causes of 354 retracted publications were unclear. Causes for retractions varied over years ( Figure 2) and a faked review process caused the largest number of retractions since 2015. The interval between publication and retraction was longer when retraction was due to fraud/suspected fraud and error than other causes ( Table 1).
Retractions from China have been dramatically increasing over the last several years and have surpassed the number of retractions from the USA since 2015 (Figure 3). Reasons for retraction varied by country of origin (p < 0.001) ( Figure 5). Error and fraud/suspected fraud were the main causes for retraction in the USA, Japan, Canada, and Spain, while a faked review process was the main reason for retraction in China, Iran and Korea. Duplicate publication and plagiarism were responsible for many retractions from China, Iran, Korea, India, and Italy ( Figure 5).

Journals
Most retracted articles were published in journals listed in the Science Citation Index (SCI; n = 3,503) and had an average impact factor of 3.98 (SD = 5.55). Impact factor was significantly correlated with the interval between publication and retraction (rho = 0.248, p < 0.001) and with the number of retracted publications due to errors (rho = 0.235, p < 0.001) and fraud/suspected fraud (rho = 0.211, p < 0.001). Journals with more than 15 retracted articles are listed in Table 2. A large number of retractions from some journals were due to different causes as shown in Table 3. In the USA, Japan, and Canada, many retractions were from journals with a high impact factor; while retractions from China, India and Iran were mostly from journals with an impact factor less than 5 ( Figure 6). Most retraced articles published in Tumour Biology were from Chinese (n = 127) and Iranian (n = 15) authors. The USA had more retractions from popular journals such as Nature (n = 10), Science (n = 9), and Cell (n = 8).

Authors with Many Retractions
Some authors had multiple retractions. Fujii Yoshitaka (n = 148) and Kato Shigeaki (n = 32) from Japan, and Staple Diederik (n = 28) from the Netherlands, were responsible for many retracted publications due to fraud. Moon Hyung-In from Korea was responsible for 28 retracted articles due to a faked review process. Some retracted articles were cited numerous times (Table 4).
Fraud/suspected fraud were the leading reasons for the most cited retractions.

Discussion
The current study explored the characteristics of retracted articles between 2012 and 2017.
Incidences of misconduct, including fraud/suspected fraud, duplicate publication, plagiarism, and a faked review process, were the most frequent causes of retracted publications. A faked review process was found to be the most frequent cause for retraction in some countries and in several journals. In some cases, the interval between publication and retraction was lengthy, and as such, these articles have been widely cited. China has become the leader in the number of retracted publications.
Misconduct was the most frequent cause for retraction, with fraud/suspected fraud as the leading cause. These results are like those from previous studies. Fang et al. (2012) found that 67.4% of retractions in PubMed prior to 2012 were attributed to misconduct, including fraud/suspected fraud (43.4%), duplicate publication (14.2%), and plagiarism (9.8%). Another study investigating retracted publications between 2013 and 2015, found that 63% of all retractions were from biomedical, medicine or clinical science literature and that 47% of retracted articles were due to misconduct (Ribeiro and Vasconcelos.2018). Error and fraud/suspected fraud were the most frequent reasons for retraction in the USA, Japan, Germany, the UK, and Canada, while plagiarism and duplicate publications were the leading causes for retraction in China, India, and Italy. In addition, the largest number of retractions due to a faked review process came from China, Iran and Korea. The difference in retraction characteristics between countries has been previously reported Stretton et al. 2012). For example, it has been suggested that non-English-speaking countries that are relatively new to research and have a low average income, retract more articles due to plagiarism and duplication, while developed countries, with a history of longstanding research, retract more articles due to fraud/suspected fraud Stretton et al. 2012).
The current study found that a journal's impact factor was significantly correlated with the number of retracted publications due to errors and fraud/suspected fraud. Moreover, the length of time from publication to retraction was longer when retraction was due to fraud/suspected fraud and when the article was published in a journal with a high impactor, ultimately leading to more citations. These results are consistent with previous studies Wang et al. 2018).
A faked review process leaded to 9.1% of all retractions. In these cases, authors fabricated a suggested reviewer email address and reviewed their own papers for rapid acceptance (Rees 2015;Hvistendahl 2015). Since 2012, many retractions due to a fabricated peer review have come from authors in China, Iran and Korea (Chen et al. 2018;Wang et al. 2018;Qi et al. 2016). Hyung-in Moon, a South Korean researcher, had 28 articles retracted due to a faked review process in 2012 (Oransky 2012b). Since 2014, Chinese and Irian authors have contributed a large number of retractions due to a faked review process (Chen et al. 2018;Wang et al. 2018;Qi et al. 2016). It has been argued that in addition to the authors and the editing companies, editors should be held partially responsible for these cases of misconduct (Rees 2015;Hvistendahl 2015;Hindawi Publishing 2015). In recent years, researchers have been overwhelmed by requests to review articles and therefore, it is difficult for journal editors to find qualified reviewers. One recent survey revealed that 52.9% of editors list finding reviewers as the hardest part of their job (Wellington 2018; Goldman 2017). Some editors have reported sending 15 or more requests in order to secure two or more reviewers (Wellington

2018; Goldman 2017).
Tumor Biology has retracted the largest number of articles due to a faked review process, while It is not surprising that China has recently surpassed the USA in the number of retracted publications, as Chinese authors now produce the largest number of scientific publications (Tollefson 2018). In addition, there is a great amount of "publish or perish" pressure in China in recent years. Chinese researchers must publish articles in SCI journals to meet graduation requirements, achieve promotions, win awards, and gain funding support (Chen et al. 2014;jiang et al. 2016;He and Xia 2014;Liao et al. 2015). These incentives have greatly increased scientific misconduct. In addition, the punishment for these misconducts from the Chinese government and institutions is mild (Liao et al. 2015). Retractions from Chinese authors (also from Indian, Iranian and Korean authors) were more often published in journals with a low impact factor compared to authors from the USA, Japan, Germany, and Canada. This may partly be explained by the finding that only a small percentage of articles published in high-quality journals are from these developing countries (Chawla 2018). Iran has dramatically increased the production of scientific articles in the last decade (Chawla 2018) and the current study found that Iranian authors contributed 123 retracted articles, and that most of these retractions were due to peer review fraud.
Consistent with previous studies, the current study found that fraudulent articles can go undetected for many years. An extreme case of this is portrayed by the Japanese anesthesiologist, Yoshitaka Fujii who produced 144 articles that took more than 12 years to be retracted. Fujii's work, most of which were clinical trials, had enrolled patients and co-worked by other authors (Amarcus41 2012;Amarcus41 2013, 24). However, no one pointed out these misconducts in the long periods. Journals or publishers may be reluctant to retract flawed papers. For example, it took four years for the British Journal of Ophthalmology to withdraw one of Fujii's articles which had been flagged since 2012 (McCook 2016a). Fraudulent articles published in high impact journals are frequently cited prior to retraction. In these cases, the impact of the fraudulent data can be extensive. For example, Joachim Boldt had published numerous papers claiming the benefits of hydroxyethyl starch for fluid resuscitation (Zarychanski et al. 2013). However, further studies have shown controversial effects of hydroxyethyl starch, some reporting increased mortality and risk of kidney failure (Zarychanski et al. 2013). Recently, it has been reported that hydroxyethyl starch may be taken off the European market (Cohen 2018).
While the current paper reports important data on the characteristics of retracted publications, there are some limitations. First, causes for retractions were determined according to retraction notices. Some journals did not provide details of reasons for retractions. In addition, some notices of retractions were called "unhelpful retraction notices", which might be misleading (Oransky 2015).
Second, the Medline database was used for our search of retracted publications, and a previous study (Schmidt 2018) has shown that some retractions may not be well-labelled in the database. Third, fraudulent papers were in publication for many years and as such, some retractions may have been overlooked.
A variety of forms of misconduct were responsible for most retracted articles in the biomedical literature in the past six years. It had been argued that reforming scientific incentive programs may be an effective method to reduce the occurrence of misconduct Liao et al. 2017      The number of retractions in Medline due to different causes by year.

Figure 4
The number of retractions in Medline from different countries by year. Figure 5 The proportion of retractions due to different causes by country of origin. Reasons for retraction significantly differed across country of origin (p < 0.001).