The covid-19 pandemic has led to increased interest in pre-prints among clinicians and the general public. But where does the balance lie between open sharing of research and minimizing the risk of confusion during a public health emergency?

bioRxiv and medRxiv were founded on the belief that early sharing of results as preprints can speed up biomedical research. Making articles available months to years before traditional publication allows experts to analyze, discuss, and build on findings immediately. This approach can accelerate understanding of basic biology and disease pathogenesis and, ultimately, the development of therapeutic and prevention strategies to improve human health. Never has this been more evident than amid the covid-19 pandemic, during which we have posted more than 15,000 pandemic related preprints so far, covering everything from structural studies of SARS-CoV-2 proteins to host immunological responses to covid-19 epidemiology, public health measures, and clinical trials.

We take responsibility for disseminating clinical information extremely seriously. Thus, we established medRxiv as a separate server that goes beyond other preprint servers in fostering transparency and promoting the responsible reporting of research by requiring clinical trial registration and declarations from authors about funding, competing interests, data availability, and ethical oversight. We also implemented dedicated screening processes involving health professionals, overseen by a leadership team that includes physicians and clinical research publishing professionals. There is a potentially huge benefit to early dissemination of the vast majority of medical findings, but we recognized from the outset that some articles would be better not disseminated as preprints because they describe work that might represent a biosecurity threat, endanger individuals, or threaten critical public health measures. Articles of this kind are declined, not because of the quality of the work or the likely accuracy of the findings, but because of the potential danger if the conclusions were to turn out to be wrong—for example, loss of public confidence in a life-saving immunization programme, as happened in the case of MMR vaccination. We recommend to the authors of such articles that their observations be disseminated after peer review by a journal, where the findings and implications can be thoroughly evaluated by experts prior to publication.

The covid-19 pandemic has presented bioRxiv and medRxiv with many challenges. We have received articles in unprecedented numbers and have made these available as fast as possible without compromising our screening processes. We have debated internally whether our posting criteria should remain the same during the pandemic. There has been intense public interest in covid-19 preprints, amplified by media coverage. We are acutely aware that although the primary aim of medRxiv is to encourage research sharing among researchers, its doors are open to all to read preprints once they have been posted. Should we therefore be more cautious about what we post? Or does the extraordinary public health emergency justify less caution? External advice has urged us in both opposing directions.

One adjustment has been made concerning dual-use research of concern (DURC). We have in the past declined work involving pathogens of pandemic potential (PPPs). But multiple labs are now performing work on pseudo- and live SARS-CoV-2 viruses and we believe that knowledge about viral variants gained from this work should be disseminated rapidly. All covid-19-related work will still be screened for DURC, however, and non-covid-19 studies of this kind will continue to be declined.

We have also implemented pandemic-specific criteria for work reporting potential covid-19 therapies. We continue to post clinical trials and in vitro studies, but we have received many articles predicting therapeutic efficacy based solely on computational work. These include strong claims for various substances, ranging from over-the-counter medicines to herbal remedies and occasionally dangerous chemicals. Preprint servers cannot conduct the detailed expert review of the computational analyses needed to establish veracity, so manuscripts making computational predictions of covid-19 therapies are accepted only if they also include in vitro or in vivo work (a higher bar). This restriction does not apply to non-covid-19 work, and computational submissions in non-therapeutic areas continue to be welcome at both bioRxiv and medRxiv.

The covid-19 pandemic has not altered our policy on studies that may cause people to behave in ways that compromise public health measures, which generally is to decline to post them as preprints. This approach does not mean we are making an assessment of the quality of the work or that we are not supportive of work that questions conventional wisdom. We are simply doing our best to tread carefully in the early days of clinical preprints, as we gain experience and bias our actions toward doing no harm. These policies are under constant evaluation and ultimately we do want to post as many of the manuscripts that we receive as we can.

A recent germane example illustrates these issues: covid-19 vaccines and adverse effects. We have received manuscripts on this topic, but given the risks of increased vaccine hesitancy, particularly early on in population vaccination efforts, we have tended to recommend their distribution only after the thorough expert peer review, evaluation, and context that a clinical journal, not a preprint server, can provide. These articles have major public health implications, but if wrong could cause harm through public confusion, particularly if magnified by disinformation campaigns. Such caution can be criticized as being at odds with our mission to accelerate knowledge, but we consider it part of our duty of responsible stewardship. Are we making the right call? We welcome feedback and debate on this and other issues as our servers evolve.

Richard Sever, Assistant Director, Cold Spring Harbor Laboratory Press

John Inglis, Executive Director, Cold Spring Harbor Laboratory Press

Theodora Bloom, Executive Editor, The BMJ

Claire Rawlinson, Publisher, BMJ Journals

Harlan Krumholz, Harold H. Hines, Jr. Professor of Medicine at Yale University and Yale New Haven Hospital.

Joseph S. Ross, Professor of Medicine and Public Health, Yale University

Competing interests: All authors are cofounders of medRxiv. Richard Sever and John Inglis are also co-Founders of bioRxiv. Richard Sever and John Inglis employed at Cold Spring Harbor Laboratory Press. Theodora Bloom and Claire Rawlinson are employees of BMJ. Joseph Ross is US Outreach and Research Editor at The BMJ.