Recommendations for preprints

Dear Editor,

In a recent article¹, AMWA-EMWA-ISMPP provide a series of recommendations regarding preprints and peer review in medical publications. The authors provide useful advice regarding labelling of preprints, the need for pre-publication checks and the desirability of linking preprints and subsequently published journal articles, all of which are currently practiced by the leading clinical preprint server, medRxiv. Unfortunately, the article also includes several errors and misunderstandings, some of which lead to inappropriate recommendations.

The authors say that preprints are often not indexed by mainstream bibliographic services. This is not correct. Biomedical preprints are in fact indexed by arguably the most popular indexing service, Google Scholar, as well as other initiatives such as Scopus, Meta, and Europe PubMed Central (as the authors note). They also state that “only about a third to a half [of preprints] are ever fully published”. This is incorrect and not supported by the references cited^2,3. Analyses of bioRxiv and arXiv have shown that around 70% of preprints end up in peer-reviewed journals after a two-year window to account for the time delay to publication^2–5. Given that articles can sometimes take several years to appear in journals and that title changes mean that some preprints cannot be matched with the corresponding journal article, this figure is likely to be an underestimate. medRxiv has only been operational for two years, with the majority of preprints appearing only in the past year and so it is premature to calculate equivalent numbers for this server, but there is little reason to assume that under normal circumstances the fraction will be lower than in other disciplines. It is possible that this fraction will differ for COVID-19 preprints, because the changing course of the pandemic has meant reports rapidly become outdated and, as a consequence, authors may choose to submit fewer preprints for formal publication by journals. Any future calculations should take this into account.

The article recommends that preprints should not be cited in reference lists and instead treated like in-text personal communications. There are good arguments for differentiating preprint and journal citations, but it is essential that both are included in the reference list as this is essential for citation indexing by services such as Google Scholar. It is sometimes claimed that preprints should not be cited in reference lists because they are not peer-reviewed. This argument is flawed, however, because editorials, books, websites, code, meeting abstracts, posters, and other non-peer-reviewed material are already routinely cited. More importantly, citation itself should not be construed or counted as an endorsement of a work. A citation is simply a link to an object, and the author’s intent and context varies. Indeed, preprints emphasize the importance of careful scrutiny of references by both authors and readers rather than an assumption of validity.

The authors are rightly concerned about the dangers of poor-quality research being highlighted by the media. This is a critical issue but is far from unique to preprints. Indeed, in the context of COVID-19, papers suggesting the disease is associated with snakes⁶, 5G towers⁷, and comets⁸ all appeared in “peer-reviewed” journals not preprints. Such work has a veneer of legitimacy imparted by peer review. By contrast, preprints are clearly labelled as “not certified by peer-review” and “yet to be evaluated”. Medical writers have a critical role to play in helping readers navigate this complex information landscape. If the recommendations made to them by AMWA, EMWA and ISMPP are to be adopted, I strongly suggest they are revised to account for the issues raised here.