To the Editor:
Preprint servers allow researchers to post manuscripts before publication in peer-reviewed journals. As of January 2022, 41 public preprint servers accepted medicine/science submissions.1 We sought to analyze characteristics of dermatology manuscripts in preprint servers and assess preprint publication policies in top dermatology journals.
Thirty-five biology/health sciences preprint servers1 were searched (March 3 to March 24, 2021) with keywords dermatology, skin, and cutaneous. Preprint server, preprint post date, location, metrics, journal, impact factor (IF), and journal publication date were recorded. Preprint policies of the top 20 dermatology journals—determined by impact factor of the journal (https://www.scimagojr.com/)—were reviewed. Two-tailed t tests and χ2 tests were performed (P<.05).
A total of 1420 articles were posted to 11 preprint servers between June 20, 2007, and February 15, 2021 (Table 1); 377 (27%) were published in peer-reviewed journals, with 350 (93%) of those published within 1 year of preprint post. Preprints were published in 203 journals with a mean IF of 6.2. Growth in preprint posts by year (2007-2020) was exponential (R2=0.78)(Figure). On average, preprints were viewed 424 times (Table 2), with published preprints viewed more often than unpublished preprints (596 vs 362 views)(P<.001). Only 23 of 786 (3%) preprints with comments enabled had feedback. Among the top 20 dermatology journals, 18 (90%) allowed preprints, 1 (5%) evaluated case by case, and 1 (5%) prohibited preprints.
Distribution of dermatology preprint articles posted by year. One dermatology preprint was posted in 2007; this data point has been excluded.
Our study showed exponential growth in dermatology preprints, a low proportion published in peer-reviewed journals with high IFs, and a substantial number of page views for both published and unpublished preprints. Very few preprints had feedback. We found that most of the top 20 dermatology journals accept preprints. An analysis of 61 dermatology articles in medRxiv found only 51% (31/61) of articles were subsequently published.2 The low rate of publication may be due to the quality of preprints that do not meet criteria to be published following peer review.
Preprint servers are fairly novel, with a majority launched within the last 5 years.1 The goal of preprints is to claim conception of an idea, solicit feedback prior to submission for peer review, and expedite research distribution.3 Because preprints are uploaded without peer review, manuscripts may lack quality and accuracy. An analysis of 57 of thelargest preprint servers found that few provided guidelines on authorship, image manipulation, or reporting of study limitations; however, most preprint servers do perform some screening.4 medRxiv requires full scientific research reports and absence of obscenity, plagiarism, and patient identifiers. In its first year, medRxiv rejected 34% of 176 submissios; reasons were not disclosed.5
The low rate of on-site comments suggests that preprint servers may not be effective for obtaining feedback to improve dermatology manuscripts prior to journal submission. Almost all of the top 20 dermatologyjournals accept preprints. Therefore, dermatologists may use these preprint servers to assert project ideas and disseminate research quickly and freely but may not receive constructive criticism.
Our study is subject to several limitations. Although our search was extensive, it is possible manuscripts were missed. Article metrics also were not available on all servers, and we could not account for accepted articles that were not yet indexed.
There has been a surge in posting of dermatology preprints in recent years. Preprints have not been peer reviewed, and data should be corroborated before incorporating new diagnostics or treatments into clinical practice. Utilization of preprint servers by dermatologists is increasing, but because the impact is still unknown, further studies on accuracy and reliability of preprints are warranted.