Preprints: straight from lab to web

There's been an incredible boom in science related to the coronavirus, with more than 10,000 peer reviewed articles on the subject since February. Teams are rushing to publish as quickly as possible. And the quickest way to do so is not via a peer-reviewed journal, but rather through what’s called a ‘preprint server’; these are online platforms that have been the subject of global scientific attention and frantic collaboration. Theo Bloom is one of the founders of these preprint servers, called medRxiv - she’s also executive editor of the British Medical Journal - and Chris Smith and Phil Sansom asked her how it works...

Theo - Yes, we heard that scientists write up their results into papers to share with one another, because science is a collective enterprise and we all build on what everyone else has done before. But typically the process of peer review and preparing things for publication in a journal can take several weeks, up to months. And in this pandemic, we all want to know things much sooner than that. Now some scientists have always shared material with a few colleagues at the same time as they send it to a journal. But with the internet, we have the ability to share with absolutely everyone. And so preprint servers are a place to post that draft article that you've got ready for a journal, but put it out to the rest of the world too where it can have an awful lot of people looking at it, commenting, and saying whether they think the study's been well done and so on.

Phil - Presumably that's been something that's used quite widely in this pandemic to study the coronavirus - what's been happening?

Theo - Yeah, so preprint servers have existed in the physical sciences for tens of years, and they're only much more recently in the life sciences, in the last few years they've been growing in importance. And medRxiv, the clinical preprint server that I work with, only launched last June. We wanted to be very careful about preprints which could affect how people take their medications, cause public health panics if we weren't careful, and so on. So we happened to launch just a little while before the pandemic came along. But really from the moments in December/January when Chinese researchers were starting to see unusual cases of pneumonia, and worked out that they came from a previously unknown virus, they have been sharing their results; and now the rest of the world too, as rapidly as they can, and that has meant in preprint form.

Phil - What kind of scale are we talking about? How many preprints are you getting? How much in a day, for example?

Theo - We've currently posted around two or three thousand preprints on medRxiv, and we're seeing submissions of 70/80 a day at the moment. You know... "what's the best mode of treatment? What's the best models for how the disease is progressing? What has worked in countries that have seen reduced rates?" And so on.

Chris - Theo, do you get oversight of what goes on to medRxiv? So if I wrote a paper and it was complete bunkum, could I just put it there and you would put it out; or would you actually have some kind of ability to pull things that were quite clearly misleading?

Theo - There's two answers to that. One is that things are screened before they go up. This is not as rigourous as the peer review we've heard about, it's not about 'is it right' and 'is it replicable'. It's simply 'is this ethical, reasonable, and unlikely to cause harm to individuals or the population'. But even so that's a multistage process that takes a handful of days before things are posted; and on very rare occasions where something gets hugely criticised after it is posted, it can be taken down. But of course we know with, you know, in the internet in general it's very hard to remove something; so we'd rather it was right before we post it, rather than having to try and remove it afterwards.

Chris - It's interesting you raise the point about the fact there is discussion and critique, because that's kind of a strength of this, in the sense that researchers can almost have a preliminary run at seeing how - it's almost like a testing the waters - how their paper's likely to be received, and they can perhaps take some of that feedback and respond to it. Some people are saying they're concerned though when they put things into these preprint servers that someone might come along and steal their ideas, and then gazzump them scientifically speaking and publish the real deal before they do, by just replicating the work themselves.

Theo - I mean there certainly is that fear, but one of the things about posting it publicly with a date stamp is that the world can see that you did it first. And I think journals are increasingly recognising that that's an important feature. Scientists were realising this even before the pandemic, when they rely on things like showing what work they're doing to get jobs and promotions and grants from the funding bodies who fund their research; that being able to say, "look, I've done this work and completed it in January of this year", even if they haven't yet got a fully peer reviewed journal article out. So in fact in some ways it does give them protection from being scooped by someone else.

Chris - We've included in this programme in the past a number of items of things which have come from medRxiv and other preprint servers. We have been at pains to emphasize that this is non-peer reviewed yet. The thing is: is there not a danger that you will end up with things passing straight through and ending up in print, in press, in podcasts, in television, radio programs; and therefore disseminating the wrong message, if it's subsequently found that they are flawed and there hasn't been that normal editorial process that would have stopped that, the peer review process? So there's a danger of misinformation proliferating off the back of these preprints?

Theo - That's clearly a risk. And what I would say is twofold to that. One is, there's an awful lot of things published in journals that turn out to be wrong. And for that reason, if no other, most journalists and others will look for more than one claim of something before deciding that it's right. It's one of those things we're seeing a number of different reports from different locations, for example, saying that people lose their sense of taste and smell when they have a coronavirus infection. That started to come out as little reports of a few people in Italy and a few other people in China, and gradually it becomes, "actually, we're seeing that everywhere and it's something that can be added to trackers of symptoms," for example. So I would say it's always the case that a single result should probably not be relied on, but it's also an obligation to those who are desperate to share knowledge, to try and temper what they're saying with "how much can we rely on this particular finding".