Assessment of retracted papers, and their retraction notices, from a cancer journal associated with “paper mills”

We identified 23 retractions in CB&R, all original articles whose authors had an affiliation in China (i.e., 100%), and 52/73 (71%) of those affiliations were associated with hospitals (Table 1). This finding closely mirrors a trend observed by Zhao et al. (2021). None of the RNs are pay-walled, and three of the RNs have a correction. Four of the RNs explicitly indicate paper mills as a reason for retraction, but the vast majority of papers were retracted for integrity-based issues related to raw data, data manipulation, or breach of data-related policies (Table 1). In Scopus, the retracted status of 17% of the 23 retracted papers is not indicated. For another four papers, the RN is available only when viewing the full description of the document, which might be confusing for Scopus users.

In addition to the association between paper mill-derived cancer research and affiliations that are Chinese hospitals, the use of web-based emails is also characteristic (Else & Van Noorden, 2021; Liu & Chen, 2021; Teixeira da Silva, 2021; Zhao et al., 2021). Corresponding authors’ emails were missing from 4/23 papers on the publisher’s website, one paper had three corresponding authors, but none had institutional emails, 12/21 (57%) emails were @163.com (5/21 were @126.com; 4/21 were @sina.com), and 9/21 (43%) emails had prefixes or names that were unrelated to the corresponding author’s name (Suppl. Table 1).

The CB&R instructions for authors (IFA) made no specific mention of email-related limitations (CB&R, 2022). Another cancer journal, Tumor Biology, published by IOS Press, explicitly prohibited the use of web-based emails for submission, stating: “The journal does not accept submissions from authors using nondescript, anonymous, email addresses (e.g., yahoo.com, gmail.com, 163.com, rediffmail.com, sina.com, 126.com, hotmail.com, etc.).” (Tumor Biology, 2022). No ORCID (Open Researcher and Contributor ID) for any corresponding author was identified in all 23 retracted papers, even though the CB&R IFA indicated that ORCID is mandatory for corresponding authors: “All submitting authors are required to complete their submissions using an ORCID identifier” (CB&R, 2022). The existence of blank or disposable (one-time use) ORCIDs, as are sometimes observed in paper mill-derived papers, degrade trust in the validity of author-based identifiers (Teixeira da Silva, 2022b).

According to Web of Science Core Collection Citations, the 23 retracted papers accumulated (until March 23, 2023) 287 citations (253 according to Scopus and 365 according to Google Scholar). At the same time, these papers continue to receive citations after the retraction: 54 citations according to Web of Science Core Collection, 54 according to Scopus, and 77 according to Google Scholar. Some retracted cancer papers continue to accrue positive (i.e., supporting) citations, although these tend to wane after the second year, and dip to near-zero citations after about a decade (Hamilton, 2019). Just over 55% of 12,231 Web of Science-indexed retracted papers published between 1981 and 2020 were cited at least once, about a quarter of which were self-citations (Sharma, 2021). Sharma (2021) also noted that Cancer Research had accrued 106 retractions that were cited on average 46 times. The reasons why authors cite retracted papers are not always clear (Dal-Ré & Ayuso, 2021), although the citation of older retracted papers tends to wane over time (Hsiao & Schneider, 2022).

We believe that more detailed information in the RNs would have made them more informative. Missing information from a RN, incomplete RNs, or RNs that are opaque or unclear do not benefit academia, and are a poor instrument of information (Teixeira da Silva & Vuong, 2022; Xu & Hu, 2022). In our view, a complete and informative RN – irrespective of its size – would have all of the following components to be both informative and transparent, holding the relevant parties accountable: 1) global appreciation of each concern and how it impacts the integrity of the paper, or its data; 2) an understanding of how the concerns came to light; 3) dates of any and all communication between all relevant parties; 4) outcomes of queries and – where possible – perspectives and responses of authors and editors.