Despite disagreement among scientists, people in Britain will soon have to wear cloth face coverings on public transport. The policy could be extended to places such as shops where social distancing is difficult writes Trish Greenhalgh in The Guardian. They are already mandatory in Panama.

Passions run high on this issue. I have some sympathy with the person who demanded that scientists should be locked in a room and not let out till they agree the facts. Except that this debate is as much about values as facts.

One of my colleagues is a philosopher of science. On day one of his course, he sends his students out to “go and get a fact, and bring it back to class”. When they return, he invites them, one by one, to carefully present and defend their fact against critique from fellow students. How was it sourced? How was it measured? Does it stand firm now that it has been plucked from its context? No fact ever survives in the form that the student presents it. The lesson is that being rigorous and systematic in producing a fact does not make that fact absolute.

Take randomised controlled trials, for example. Surely it’s good science to take a sample of people during a pandemic, randomly allocate half of them to wear cloth face coverings, and look for a statistically significant difference in how many develop the disease?

Let’s unpack the quest for that particular fact. First, let’s design the intervention. As laboratory studies have shown, fabric materials are good at stopping large viral-laden droplets on their way out, but they’re not very good at stopping tiny airborne viral particles on the way in. So the intervention we need to test is something called source control: the effectiveness of my cloth mask in protecting you; yours in protecting me.

Let’s therefore test, in all members of the general public (population), the impact of cloth face coverings (intervention) compared to no coverings (control) in public places (setting) for preventing infection in other members of the public (outcome).

There have never been any such trials. Why not? Because such a trial would be impossible. Which people, exactly, would you randomise? Everyone who lives in a particular street? Everyone who catches a particular bus? To be included in a randomized trial, people have to give informed consent and agree to comply with the intervention. If only a small proportion agrees, how many streets or bus-loads or squares-full of people would you need to invite in order to get enough to generate your “fact”?

How would you test the hypothesis that my mask protects you? This is problematic because, instead of just following up with one individual who is already signed up to the study (me), you’d somehow need to identify a fraction of all the people I came into contact with as I went about my life, and test each of them to see if they developed the disease. But none of those other people (random street-dwellers or bus passengers or square-crossers) agreed to be in the trial in the first place. You simply can’t study them like laboratory animals (though someone has done a randomized controlled trial to show that face coverings are highly effective in reducing SARS-CoV-2 transmission in hamsters).

In order to “do science” on the effectiveness of face coverings in the real world, you would either have to define some kind of closed system to be your sampling frame, say a school or home environment or compromise and study the effect of mask-wearing on the wearer rather than on other people. And that’s what scientists have done.

I’ve reviewed all the published randomized trials on masks and face coverings for the general public in a peer-reviewed scientific paper. The trials consist of studies of mask-wearing in closed systems and of masks to protect the wearer, usually at specific mass events (notably, pilgrimages to the Hajj). Furthermore, all these studies were done some time ago when the prevailing diseases were colds and flu. None looked at prevention of Covid-19.

What did these not-terribly-relevant trials show? Masks seemed to provide some protection for the wearer, but effects were small and usually not statistically significant (perhaps because studies were just too small to produce a clear answer). Masks given in the home, for example, to protect the parents of a child with flu from catching it themselves, didn’t have much effect – but then they weren’t worn much (for a variety of reasons including “child did not like parent wearing a mask”).

These findings are facts of a sort. I’d love to discuss them with my philosopher colleague’s students. But the bottom line is we have no directly relevant facts from randomized controlled trials of source control in Covid-19. What now? Demand that policy should not change until the much-desired randomized trial exists? Or suck it and see? Here’s where values come in.

Drug testing
Randomized trials were developed to test drugs. As we know from thalidomide, new drugs can cause terrible harm. Scientists arguing for caution in the masking debate are almost all medically trained and view the principle “do no harm” (by which they mean, never give a new drug to any patient before it’s been tested in a randomized trial) as overriding.

In public health, however, interventions are usually introduced pragmatically and evaluated as before-and-after natural experiments. There are no randomized controlled trials in community settings, for example, of handwashing, social distancing, closing schools, quarantining, closing borders or contact tracing. Public health scientists value natural experiments highly because they view the trade-off between real-world achievability and experimental ideals as worthwhile.

Like dozens of other countries, England is about to have a natural experiment of face coverings in public places. Still, the scientists will continue to argue. If transmission rates of Covid-19 fall as predicted, public health experts – and I count myself among them – will say this has produced a “fact” that face coverings are effective source control. And triallists will say that in the absence of a trial (which they will continue to demand), this is not a fact because there’s no robust evidence that the association is causal. Whether we can agree on “the facts” or not, we’ll hopefully welcome the positive outcomes, as I predict that transmission of Covid-19 will fall and that any harms will be relatively minor and worth the trade-off.

Trish Greenhalgh is a professor of primary care health sciences at Oxford University