A Messy Withdrawal of a Sloppy Preprint
A preprint of ivermectin to treat Covid-19 gets pulled — but does it, really?
A preprint touting ivermectin as a treatment for Covid-19 — inevitably trumpeted on right-wing media, by anti-vaxxers, and by conspiracy theorists generally — has been withdrawn from Research Square, according to a report from the Guardian.
Or has it?
You may recall that Research Square assigns a new DOI to each version of a preprint, as I noted in a post earlier this year outlining duplication and redundancy in the preprint space. This particular preprint has four versions. Version 4 consists solely of the text “Research Square has withdrawn this preprint.” Apparently, that’s how Research Square’s system is able to stack a withdrawal notice atop prior versions. Meanwhile, Versions 1-3 — versions of the actual preprint — are still available, now topped with a prominent statement that the preprint has been withdrawn, but with full-text, PDFs, and DOIs still working normally.
Version 1 was posted on November 13, 2020, more than eight months ago, and the preprint is badged as having passed the “Prescreen” for ResearchSquare. This badge gives the false impression that the paper was screened for more than just the most cursory of information.
Over the past eight months, the preprint has been included in two meta-analyses of ivermectin, and cited more than 30 times, although the quality of these citations remains unclear. One in JAMA simply notes that trials like this haven’t yet passed peer-review, for example.
As Jack Lawrence, a medical student and journalist for a site called Grftr News (“Grifter Analysis and Review” — love it) notes, the plagiarism within the paper was easy to spot because the authors stole text and then modified it using a thesaurus, changing “severe acute respiratory syndrome” to “extreme intense respiratory syndrome” on one occasion.
Lawrence’s full analysis of the paper is a little overheated, but worth reviewing. Another analysis by Nick Brown, an expert in fraudulent papers, is also available. He identifies 79 fraudulent records in the dataset — duplicates that appear to have been manipulated to appear unique. Other experts have also debunked the paper and found glaring errors.
The concerns here are multiple, yet again. How did a paper this shoddy get through “Prescreen” and remain up for eight months before being flagged? Why are the earlier versions, each with its own DOI, still available? If preprints are working as intended, why does this keep happening?
The rapid uptake and distribution of preprints remains a concern, as well. For instance, Europe PMC links to Version 1 of the preprint, and frames the text of that version in its site. There is no indication the paper has been withdrawn.
I have questions in to Research Square, and will provide updates if warranted. But once again, we have a preprint server being used to push bad science to support right-wing conspiracies, a sloppy withdrawal process, and amplification of claims that haven’t passed peer-review.
Happy Friday . . .