**Objective** The Retraction Watch Database reported in 2018
that conflict of interest (COI) was not a common sole reason
for a retraction.1
This study assesses the characteristics of
articles in the Retraction Watch Database that have COI as a
reason for retraction.

**Design** All retractions from Retraction Watch Database that
had COI as a reason for retraction were analyzed. Using
Scopus Application Programming Interface, data on the types
of articles retracted, fields of retracted articles, number of
authors and institutions, time between the publication of the
article and the retraction, first corresponding author’s
country, and other reasons for retraction were collected. The
Retraction Watch Database was also manually searched to
identify articles with expression of concern, errata, or
corrigendum notices for COI. The linguistic content of notices
was compared using the Linguistic Inquiry and Word Count
program, and t-distributed stochastic neighbor embedding
technique (t-SNE) was used to visualize high-dimensional
data by placing each data point in a 2-dimensional map.2

**Results** Of 194 articles tagged with COI as a reason for
retraction from 1991 to 2021, COI was the only reason for
retraction for 26 (13%). The most common additional reasons
for retraction were fake peer review (n = 34), investigation by
journal/publisher (n = 29), concerns/issues about authorship
(n = 24), and withdrawal (n = 24). Detailed explanation of
financial COI was found in only 17 retraction notices (9%); 35
notices (18%) included a detailed description of undisclosed
or misdisclosed employment relationships. Of 48 expressions
of concern, errata, corrigendum notices, or undisclosed or
concerning COI declarations were the only reason for 16
(33%). The most common additional reasons for a notice
were lack of information provided on methods, study data,
and process of institutional ethical approval for the published
article (n = 14), as well as concerns over the completeness of
acknowledged contributions (n = 5). The median (IQR) time
from article publication to retraction was 414 (224-1170)
days. Russia and China (38 articles each) were the most
common origins of the retracted articles. In logistic
regression, the predictive linguistic attributes for retraction
notices were higher analytic score (odds ratio [OR], 1.18; 95%
CI, 1.07-1.30) and lower clout score (OR, 0.93; 95% CI,
0.90-0.95). The t-SNE visualizing technique found no linear
separation between retraction and correction or expression of
concern points, indicating that similar language is used across
different publication notices.

**Conclusions** Nondisclosed or incorrect disclosure of COI
may be a sole reason for article correction, and evident COI
may be a sole reason for article retraction. Retraction notices
often do not describe in detail the conflicts involved. Higher
analytic tone of the retraction notices suggested that editors
used more logical reasoning when elaborating retractions.
Higher scores on clout for nonretraction notices suggest that
it may have been easier for editors to express more confidence
in writing these corrections.

**References**
1. Oransky I, Marcus A. Why do medical journals keep taking
authors at their word? Accessed January 29, 2022. https://
www.statnews.com/2018/09/14/jose-baselga-doctors-
industry-conflict/
2. van der Maaten LJP, Hinton GE. Visualizing data using
t-SNE. J Machine Learning Res. 2008;9:2579-2605.

**Conflict of Interest Disclosures** Ana Marušić is a member of
the Peer Review Congress Advisory Board but was not involved in
the review or decision for this abstract. No other disclosures were
reported.

**Funding/Support** This study was funded by the Croatian Science
Foundation “Professionalism in Health–Decision Making in
Practice and Research” (ProDem) under grant agreement IP-2019-
04-4882.

**Role of the Funder/Sponsor** The funder had no role in the
study design, data collection, and analysis; decision to publish; or
preparation of the manuscript.

**Acknowledgments** We thank Retraction Watch for providing
access to their retraction database.

![](https://assets.underline.io/uploads/markdown_image/1/image/85b6390bafb6d4e7c6a886e695969c98.png)

Analysis of Articles Retracted Because of Conflicts of Interest in the Retraction Watch Database

**Objective** To examine the association of having editorial staff members affiliated with low- and middle-income countries with publications from low- and middle-income countries in leading biomedical journals. It was hypothesized that greater representation from low- and middle-income countries among editorial staff would be associated with more published articles from low- and middle-income countries. 
 **Design** A cross-sectional study was conducted of biomedical journals in 2020 in fields whose content covers the largest disease burden globally. To obtain editorial staff country affiliations, webpages of the 5 leading journals in each of the following fields were reviewed: general medicine, pediatrics, surgery, obstetrics and gynecology, cancer, cardiovascular diseases, infectious diseases, psychiatry, and nutrition. Original research articles in each journal were reviewed through MEDLINE. The study country of each original research article (ie, where the study was conducted) was determined by searching the article title, abstract, keyword, and medical subject heading using EndNote, and by 2 authors reviewing the full text of each article to assign a study country. Editorial staff country affiliations and study country location(s) were classified according to the World Bank income brackets and regions. Descriptive statistics were used to describe the proportion of editorial staff affiliated with each income bracket and region. Spearman ρ was used to assess the relation between the proportion of editorial staff affiliated with low- and middle-income countries and published articles reporting work conducted in these countries. 
 **Results** There were 3819 editorial staff in the 45 included journals: 3637 (95.2%) were affiliated with high-income countries, 140 (3.7%) with upper−middle income countries, 37 (1.0%) with lower−middle income countries, and 5 (0.1%) with low-income countries. Every editor in chief (n = 48; 100.0%) was affiliated with a high-income country. Of the 459 associate editors, 445 (96.9%) were affiliated with highincome countries, 10 (2.2%) with upper−middle income countries, 4 (0.9%) with lower−middle income countries, and 0 with low-income countries. Editorial staff were most commonly affiliated with North American (2120 [55.5%]) and European and Central Asian countries (1256 [32.9%]). Of the 10,096 original research articles included, 7857 (77.8%) reported research conducted in high-income countries, 1562 (15.5%) in upper−middle income countries, 507 (5.0%) in lower−middle income countries, and 170 (1.7%) in lowincome countries. Greater editorial staff representation correlated moderately with more published articles reporting research conducted in low- and middle-income countries among all articles (Spearman ρ, 0.51; P < .001) and among articles reporting multicountry studies (Spearman ρ, 0.42; P = .005) (Figure 2). 
 ![](https://assets.underline.io/uploads/markdown_image/1/image/14edf99e92065c3e95ce0751d301a2f3.png) 
 **Conclusions** The inclusion of editorial staff affiliated with low- and middle-income countries may be an approach to promoting the publication of research conducted in low- and middle-income countries.
1Precision Vaccines Program, Boston Children’s Hospital, Boston, MA, USA, chris.rees@emory.edu; 2Division of Pediatric Emergency Medicine, Emory University School of Medicine, Atlanta, GA, USA; 3Children’s Healthcare of Atlanta, Atlanta, GA, USA; 4School of Public Health and Social Sciences, Muhimbili University of Health and Allied Sciences, Dar es Salaam, Tanzania; 5Division of Infectious Diseases, Department of Paediatrics, The Hospital for Sick Children, Toronto, ON, Canada; 6Department of Physiology and Nutrition, St. John’s Medical College, Bengaluru, India; 7Center for Nutrition, Division of Gastroenterology, Hepatology and Nutrition, Boston Children’s Hospital, Boston, MA, USA; 8Department of Nutrition, Harvard T. H. Chan School of Public Health, Boston, MA, USA 
 **Conflict of Interest Disclosures** None reported. 
 **Funding/Support** Christopher P. Duggan was supported in part by the National Institutes of Health (K24DK104676 and 2P30 DK040561). 
 **Role of the Funder/Sponsor** The funder had no role in the study design or in the collection, analysis, or interpretation of the data. The funder did not write the report and had no role in the decision to submit the abstract for presentation. 
 **Acknowledgments** Gandolina Melhem and Chris A. Rees are co−first authors. We thank Chloe Rotman, at Boston Children’s Hospital, for her assistance in developing the PubMed query as well as acquiring the full-text articles reviewed in this study.

![](https://assets.underline.io/uploads/markdown_image/1/image/bf5305ad43d9e2c5cd859aec4fec7415.png)

Association Between International Editorial Staff and International Publications in Leading Biomedical Journals

**Objectives** To investigate women’s representation among peer reviewers of medical journals overall and according to the gender of the editor in chief and women’s representation in the editorial board. 
 **Design** This cross-sectional study included journals of the BMJ Publishing Group that reported the names of their peer reviewers in 2020. For each journal, the gender of the editor in chief, deputy editors, and associate editors was determined based on photographs and pronouns available on the journal website or professional affiliations. The package genderizeR in R was used to predict the gender of the peer reviewers based on given names, which were extracted from full names and assigned as woman or man. 
 **Results** Overall, this study included 47 of the 74 journals in the BMJ Publishing Group because data were not publicly available for the remaining journals (Table 3). Women accounted for 30.2% of the 42,539 peer reviewers, with marked variation ranging from 8% in the Journal of ISAKOS to 50% in Medical Humanities. Women represented 33.4% of the 555 editors, including 19.2% of the 52 editors in chief. There were 5 journals with more than 1 editor in chief, all of which had 2 men as editors in chief. There were 5 journals with no woman among the editors and 12 journals in which women’s representation as editors was 50% or greater. Among those 12 journals, 7 had a woman as editor in chief. There was a moderate positive correlation between the percentage of women as editors and as reviewers (Spearman correlation coefficient, 0.590; P < .001). The percentage of women as editors excluding editors in chief was higher when the editor in chief was a woman than a man (53.3% vs 29.2%; P < .001). The percentage of women as peer reviewers was also higher in journals with a woman as editor in chief compared with a man (32.0% vs 26.4%; P < .001). There was no significant correlation between women’s representation and the journal CiteScore (Spearman correlation coefficient, −0.288; P = .07) or impact factor (Spearman correlation coefficient, −0.343; P = .09). This study has some limitations, such as using binary prediction of gender based on given names, relying on data for accepted rather than invited reviewers, and not accounting for the number of reviews performed by each individual. 
 ![](https://assets.underline.io/uploads/markdown_image/1/image/2dfaac8ba7785221c799d813135c0ddd.png)![](https://assets.underline.io/uploads/markdown_image/1/image/2e2d56fb188047c3e0ee0b87f9d73c3c.png)
![](https://assets.underline.io/uploads/markdown_image/1/image/8125764c399777c8413734c0ee19b4f0.png)**Conclusions** Women account for less than 1 in 3 peer reviewers of medical journals. Women’s representation as peer reviewers is higher in journals with a higher percentage of women as editors or with a woman as editor in chief, suggesting that increasing women’s representation as editors may be one of many necessary steps toward gender equity. The gender gap among peer reviewers of medical journals mimics that previously found among authors and editors,1-3 thus emphasizing the need to address the persisting gender gap at all levels of the publishing system. 
 **References**
1.	Lundine J, Bourgeault IL, Clark J, et al. The gendered system of academic publishing. Lancet. 
2018;391(10132):1754-1756. doi:10.1016/S01406736(18)30950-4
2.	Hart KL, Perlis RH. Trends in proportion of women as authors of medical journal articles, 2008-2018. JAMA Intern Med. 2019;179(9):1285-1287. doi:10.1001/ jamainternmed.2019.0907
3.	Pinho-Gomes AC, Vassallo A, Thompson K, et al. Representation of women among editors in chief of leading medical journals. JAMA Netw Open. 2021;4(9):e2123026. doi:10.1001/jamanetworkopen.2021.23026 
 **Conflict of Interest Disclosures** Ana-Catarina PinhoGomes reported receiving funding through an Academic Clinical 
Fellowship by the National Institute for Health Research. Mark Woodward is a consultant for Amgen, Kyowa Kirin, and Freeline and reported receiving grants from the National Health and Medical Research Council of Australia. Sanne A. E. Peters reported receiving support through a UK Medical Research Council Skills Development Fellowship. No other disclosures were reported.

![](https://assets.underline.io/uploads/markdown_image/1/image/1b9a73ff68953942366e96aec9a60f58.png)

Women's Representation Among Peer Reviewers of Medical Journals

**Objective** Retracted clinical trials may be
influential in systematic reviews and
guidelines. Fanelli et al 1 examined the impact of
retractions on 50 meta-analyses and found
little influence but suggested that the impact
was likely variable and context specific. Using
retracted clinical trial reports from one
research group, 2 responses of editors and
authors to notifications that systematic reviews
and guidelines they had published had cited
the retracted trials were evaluated. 
 **Design** Between November 2019 and January
2020, for 27 retracted trials (published 1997 to
2012, retracted 2016 to 2019) in osteoporosis
and neurology, searches were conducted in
Web of Science, Scopus, and researchers’ files
for systematic reviews and clinical guidelines
that cited the trials as evidence. Citing
publications that acknowledged that they had
cited retracted work were excluded. For each
citing publication, 2 researchers independently
coded the likely impact of removing the
retracted trial reports, including findings likely
to change (yes, no, or uncertain) and size of change
(substantial, moderate, or minor). In a factorial design with 4
groups, authors of citing publications were randomized to
receive up to 3 emails (if no reply) to the contact author and
journal editor vs contact author only and, for citing
publications with 2 or more authors, to an email to the
contact author only vs up to 3 authors (selected from among
contact, first, second, and final authors). Emails giving details
of the retracted trials were sent monthly September to
October 2020. Follow-up was 1 year for outcomes, assessed
by replies to emails and notices in the public domain,
including any reply (yes or no), time for a reply from first
author, time for any reply, and action taken. Comparisons
were undertaken using χ² tests. Email replies were analyzed
for content by 3 researchers. 
 **Results** A total of 88 citing publications (published 2003 to 2020) were identified; 2 were corrected before emails were sent. Authors/editors were emailed about 86 citing publications. A total of 45 of 88 citing publications (51%) had findings coded likely to change if the retracted trials were removed, and 39 of these (87%) were likely substantial impacts. Replies were received for 44 of 86 citing publications (51%). Emailing 3 authors was more likely to elicit a reply than emailing the contact author alone (26 of 42 [62%] vs 16 of 40 [40%]; P = .03), but including the editor did not increase replies (23 of 44 [52%] with editor vs 21 of 44 [48%] without editor; P = .66). Including more authors and/or the editor, whether findings were coded likely to change and the size of the likely change, had no effect on published corrective action. One year after emails were sent, only 9 publications had been corrected. 
 **Conclusions** Retracted trials impact systematic reviews and guidelines. Emailing more authors to notify them of the retractions yielded more replies but did not increase corrections. Email alerts to authors and editors are inadequate to correct the impact of retracted publications in citing systematic reviews and guidelines. Publications with retracted citations should be marked until authors resolve concerns. 
 **References**
1.	Fanelli D, Wong J, Moher D. What difference might retractions make? an estimate of the potential epistemic cost of retractions on meta-analyses. Account Res. Published online July 14, 2021. doi:10.1080/08989621.2021.1947810 
 2.	Bolland MJ, Avenell A, Gamble GD, Grey A. Systematic review and statistical analysis of the integrity of 33 randomized controlled trials. Neurology. Published online November 9, 2016. doi:10.1212/WNL.0000000000003387 
 **Conflict of Interest Disclosures** None reported. 
 **Funding/Support** Mark J. Bolland is the recipient of an HRC Clinical Practitioner Fellowship. The Health Services Research Unit is funded by the Chief Scientist Office of the Scottish Government Health and Social Care Directorates. 
 ![](https://assets.underline.io/uploads/markdown_image/1/image/c24532dbbbf609fe14bad507ca0cd465.png)

Effect of Alerting Authors of Systematic Reviews and Guidelines That Research They Cited Had Been Retracted: A Randomized Controlled Trial

**Objective** Paper mills represent an offer for on-demand writing of fraudulent academic manuscripts for a fee. There is evidence of increasing infiltration of the academic literature with fraudulent papers originating from paper mills. The majority of known paper mills originate from China and less is known about their operations in other countries, namely Iran and Russia, where paper mills also operate.1 This study attempted to shed light on the activity of paper mills using the experience of a Russian mill to detect papers originating from the Russian-based paper mill International Publisher LLC and to identify a set of factors associated with fraudulent papers. 
 **Design** Offers published during 2019-2021 were collected from the 123mi.ru website. The details of the offer as the title/ topic of the paper, number of coauthors, country/region of a journal, date of publication, and in some cases country of the author and abstract were analyzed to identify auctioned papers. Many fraudulent papers were detected because they were published with identical or closely worded titles. The correctness of detection was confirmed by matches of other offer details. 
 **Results** A total of 1009 offers were published during the study period, and the study identified at least 436 papers (43%) published in 154 journals potentially linked to the 
paper mill as of mid-March 2022; 22 of the papers appeared in 3 hijacked journals. More than 800 scholars from at least 39 countries purchased coauthorships from this paper mill.
Fraudulent papers were published in predatory journals, journals of reputable publishers, and hijacked journals that represent cybercriminal publishers who clone titles and other metadata of legitimate journals (**Figure 1**). The list of identified papers can be accessed via a shared spreadsheet.2 This study found collaboration anomalies in questionable papers in which authors did not have common research interests, specialized in different disciplines, or were affiliated with different universities that did not focus research on the paper’s subject or might not specialize in the topic of the paper. There was also evidence of data fabrication in these papers. The study detected dishonest collaboration with journals (purchase of an entire issue) or editors who were listed as coauthors of fraudulent papers. This paper mill also applied a “one paper–one journal” principle, ie, submission of a problematic paper to an individual legitimate journal only once. 
 **Figure 1. Number of Articles per Individual Journal**
![](https://assets.underline.io/uploads/markdown_image/1/image/36ccf052cd29559d1116f599c5519033.png)
![](https://assets.underline.io/uploads/markdown_image/1/image/b3aaf14cc6092b68003954157b9a862e.png)
 **Conclusions** The production of this paper mill was difficult to detect due to individually tailored papers being submitted to 154 journals. Journals themselves had no opportunity to notice irregularities from 1 paper. Detection of irregularities can require regular upgrades of the system of detection of fraudulent papers by publishers and COPE (Committee on Publication Ethics), which is based on tracing similarity patterns among manuscripts. Because the study analyzed a single paper mill, it is likely that the number of papers with forged authorship is much higher. 
 **References**
1.	Else H, Van Noorden R. The fight against fake-paper factories that churn out sham science. Nature. 2021;591:516519. doi:10.1038/d41586-021-00733-5
2.	Abalkina A. Paper mill “International Publisher.” https://docs.google.com/spreadsheets/d/1vzjtRPX7kd2KczdtKONEpRZb2F-4lj5Sd9jL6DfbiBk/edit?usp=sharing 
 **Conflict of Interest Disclosures** None reported. 
 ![](https://assets.underline.io/uploads/markdown_image/1/image/2b4a2d896142a5537f9159199f35eb9b.png)

Publication and Collaboration Anomalies in Academic Papers Originating From a Russian-Based Paper Mill

As techniques for generating deepfake videos (in which a generative adversarial network is used to alter the human subject of a video) improve, so does the potential to use them for nefarious ends. It is therefore important that we identify those among us who are most susceptible to being fooled. Tidler and Catrambone (2021) presented results which suggest that one’s political orientation and affect detection ability are strong predictors of one’s ability to detect deepfakes. The current work is a re-analysis of their data, intended to reveal a more highly-resolved profile of a susceptible individual. The data (N = 169) were submitted to a k-means clustering algorithm which revealed a 2-cluster solution.

Younger, more progressive thinking individuals, with agnostic or atheistic religious beliefs, and greater adeptness at spotting deepfakes formed Cluster 1; and individuals with more conservative political leanings, greater religiosity, and less skill at detecting deepfakes formed Cluster 2.


Assessing Deepfake Video Detection Ability Using Unsupervised Machine Learning

In a ground war, achieving peace is challenging even when both sides suffer. Here we demonstrate that when humans jointly viewed a battlefield as perceptually grouped by colors, a factor irrelevant to the rules of land-war games, they waged fewer wars and accumulated more wealth due to the perceptual border’s constraint on their aggression. This facilitation of peace arose from perceptual grouping that serves as a visual common sense shared by humans, as a mismatched perceptual grouping between players failed to limit war. Moreover, it relies on the geopolitical principles that make ground wars destructive and disappeared when these principles were violated. Together, we show that perceptual grouping limits human ground war as it is a mutually acknowledged “rallying point” where expectations of peace converge.

Perceptual Grouping for War and Peace

A private learning scheme TextHide was recently proposed to protect the private text data during the training phase via so-called instance encoding. We propose a novel reconstruction attack to break TextHide by recovering the private training data, and thus unveil the privacy risks of instance encoding. We have experimentally validated the effectiveness of the reconstruction attack with two commonly-used datasets for sentence classification. Our attack would advance the development of privacy preserving machine learning in the context of natural language processing.

Reconstruction Attack on Instance Encoding for Language Understanding

The present study explores moral judgment in COVID-19 related moral dilemma situations involving allocation of ventilators with conflicting allocation principles. Utilitarian triage criteria like the chance of recovery or longer life expectancy are opposed to egalitarian procedures like random allocation and ‘first come, first served’. In the first part of the experiment, participants are presented with three hypothetical situations in which there are two patients admitted to a hospital in a critical state needing a ventilator but only one is available. The conditions about the patients are described and several triage procedures are suggested and rated by participants. Separately, participants rated their agreement with several triage principles. The result shows a clear preference for utilitarian allocation principles. The random allocation principle receives the lowest ratings. The ‘first come, first served’ correlates with the belief in fate score hinting that the egalitarian nature of this principle is questionable.

Underline digital video library

Analysis of Articles Retracted Because of Conflicts of Interest in the Retraction Watch Database

Association Between International Editorial Staff and International Publications in Leading Biomedical Journals

Women's Representation Among Peer Reviewers of Medical Journals

Effect of Alerting Authors of Systematic Reviews and Guidelines That Research They Cited Had Been Retracted: A Randomized Controlled Trial

Publication and Collaboration Anomalies in Academic Papers Originating From a Russian-Based Paper Mill

Assessing Deepfake Video Detection Ability Using Unsupervised Machine Learning

Perceptual Grouping for War and Peace

Reconstruction Attack on Instance Encoding for Language Understanding

Moral Judgments and Triage Principles related to COVID-19 Pandemic

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES