Downstream retraction of preprinted research in the life and medical sciences

Retractions have been on the rise in the life and clinical sciences in the last decade, likely due to both broader accessibility of published scientific research and increased vigilance on the part of publishers. In this same period, there has been a greater than ten-fold increase in the posting of preprints by researchers in these fields. While this development has significantly accelerated the rate of research dissemination and has benefited early-career researchers eager to show productivity, it has also introduced challenges with respect to provenance tracking, version linking, and, ultimately, back-propagation of events such as corrigenda, expressions of concern, and retractions that occur on the journal-published version. The aim of this study was to understand the extent of this problem among preprint servers that routinely link their preprints to the corollary versions published in journals. To present a snapshot of the current state of downstream retractions of articles preprinted in three large preprint servers (Research Square, bioRxiv, and medRxiv), the DOIs of the journal-published versions linked to preprints were matched to entries in the Retraction Watch database. A total of 30 retractions were identified, representing only 0.01% of all content posted on these servers. Of these, 11 retractions were clearly noted by the preprint servers; however, the existence of a preprint was only acknowledged by the retracting journal in one case. The time from publication to retraction averaged 278 days, notably lower than the average for articles overall (839 days). In 70% of cases, retractions downstream of preprints were due – at least in part – to ethical or procedural misconduct. In 63% of cases, the nature of the retraction suggested that the conclusions were no longer reliable. Over time, the lack of propagation of critical information across the publication life cycle will pose a threat to the scholarly record and to scientific integrity. It is incumbent on preprint servers, publishers, and the systems that connect them to address these issues before their scale becomes untenable.