A Springer Nature journal has retracted a paper on hepatitis C infection it had previously corrected for problematic data – but in between the editors declared the case closed.
The paper, “The interaction between microRNA-152 and DNA methyltransferase-1 as an epigenetic prognostic biomarker in HCV-induced liver cirrhosis and HCC patients,” was published in July 2019 in Cancer Gene Therapy. The authors were affiliated with Al-Azhar University, in Cairo, and the Ministry of Scientific Research, in Gizah, Egypt.
Some two years later, Alexander Magazinov – a data sleuth about whom we’ve written before – posted concerns on PubPeer about the article and three other papers by members of the group.
Those articles were published the Egyptian Journal of Medical Human Genetics; the Al-Azhar Journal of Pharmaceutical Sciences; and Gene Reports.
Rady El-Araby, the first author of two of the articles, responded to Magazinov, triggering a months-long exchange in which El-Araby tried to address the mounting questions about apparent flaws in the Cancer Gene Therapy paper.
Meanwhile, in February, the journal amended the article with the following correction:
After publication an error was found in Table 3 of the manuscript in the columns related to gender distribution. After precise review and return to the source of the research and the collected papers, the following was found: Female (1a) = 10 (55.6%); Male (1a) = 8 (44.4%); Total N of (1a) = 18. Female (2) = 4 (20.0%); Male (2) = 16 (80.0%); Total N of (2) = 20. These modifications do not affect the results or the discussion of the manuscript, as age was not an influencing factor in this study or associated with any other parameters in the study. The amended figure accompanies this correction.
But that notice didn’t satisfy Magazinov, who pointed out that the correction was “inconsistent with Figure 6.”
El-Araby thanked Magazinov for his comment, adding:
I am with you on this, but I confirm that the ROC curve is correct and the settings that were made are correct.
But everything in the matter was that the difference was in the gender column, where there were unwritten cases, but they were modified in the previous amendment, and all the rest of the parameters in the study were prepared exactly.
Magazinov then noted that several figures continued conflicting information about the number of data points and called on the authors to explain the discrepancies.
When El-Araby was unable to do so – and Magazinov pressed – the researcher posted the following white flag:
Dear Editor
Kindly find the corrected graph regarding coment 5#
Fig. 3 Correlation between DNMT-1 and miRNA-152 in Ia group
Which evidently led to this retraction statement, dated June 22:
The Editor-in-Chief has retracted this Article. After publication, concerns were raised regarding the numbers of patients in various groups in Table 3, which the Authors addressed via Correction [1]. However, further similar issues have been identified in Figs. 3–6, which affect the results and conclusions of this study. The numbers of data points in these figures do not match the group descriptions in the Materials and methods. Additionally, the Authors’ subsequent study [2] appears to report the same patient recruitment period from the same institution as well as the same inclusion and exclusion criteria, but contains a higher number of patients than this Article. The Authors have been unable to sufficiently explain these discrepancies.
Mahmoud A. Khalifa, Mona M. Zoheiry, Manal Y. Zahran, Mohamed I. Rady, Raafat A. Ibrahim, Mohamed D. El-Talkawy and Faiza M. Essawy do not agree to this retraction. Rady E. El-Araby has not explicitly stated whether they agree to this retraction notice.
El-Araby did not respond to a request for comment.
Despite the retraction, however, Magazinov wasn’t happy with the journal’s handling of the case. Emails between him and an editor at Nature show that the publisher appeared content to let the article remain corrected even after further problems surfaced:
That editor, Virginia Mercer, wrote on February 28:
I have discussed your comment with the Editor-in-Chief and the authors and we don’t agree there is an issue here or that anything else needs to be done.
The ROC curve is correct and the settings that were made are correct. The only difference was in the gender column, where there were unwritten cases, but they were modified in the correction, the rest of the parameters in the study were prepared correctly.
Here we are talking about the number of the samples not the interpretation/validity of the data.
The authors provided all the raw data so we don’t think there is anything of concern.
Clearly, something changed, as Magazinov pointed out to Mercer on June 26:
First of all, thank you for doing the right thing!
It appears, however, that the retraction notice does not report any additional problems beyond what I communicated to you or posted on PubPeer: https://pubpeer.com/publications/88655ADD5EEFE5A0FEF0B03367CB6D
In your previous email you stated that these problems do not warrant any action besides the previous correction. Could you please clarify what happened in the meantime that changed your judgement?
Even more important, let me remind that the just-retracted paper is only one of the 4 highly similar papers by highly overlapping groups of authors. Moreover, it has been reported as such, and the similarity has even been acknowledged in the retraction notice. Regarding these matters, have you communicated your findings to the other journals involved?
Magazinov told us that he found the retraction notice lacking detail:
The concern was not only about the number of patients in the experimental groups. There was also a question why the average age, the SD of age, and quite a lot of averages and SD of various biochemical indicators are the same across the four papers – at least in some pairs of papers the groups are not the same even by the number of subjects (and no matter how much they overlap). I don’t see this problem reflected in the notice.
In one of the groups there are two subjects with identical expression of something up to 5 significant digits, and two other subjects with identical expression of other something up to 5 significant digits. I agree, sometimes this happens, but it is rare enough to be an additional concern, provided that there are other problems with the data.
And he took issue with the way the publisher responded to his concerns:
I notified them of additional concerns that arose post-correction immediately, and they were very fast to dismiss those with “no further action is warranted.” Now they retract, and the retraction notice contains nothing additional to what I have raised. Meaning they changed their decision, and the change doesn’t appear to be based on any additional information. What made them rethink? An explanation would be very helpful, but I haven’t heard one.
Like Retraction Watch? You can make a tax-deductible contribution to support our work, follow us on Twitter, like us on Facebook, add us to your RSS reader, or subscribe to our daily digest. If you find a retraction that’s not in our database, you can let us know here. For comments or feedback, email us at [email protected].