COVID-19, Masks and Magical Thinking

Anthropologists have been interested in magical thinking since the earliest days of their discipline. This is the idea that you can influence the outcome of specific events by doing something that has no causal connection to them. The concept has been picked up in psychology and psychiatry, as something that is common in children but discarded in the course of development, and as something that may be a pathology in some people with mental health problems.

The bitter controversy over the use of masks or face coverings in community settings that has erupted in the USA and can also be seen in the UK and mainland Europe has many of the characteristics of the contest between magic and science. Advocates of masks have struggled to demonstrate a causal connection between face covering and the transmission of the SARS-COV-2 virus. Their critics might well be forgiven for claiming that mask mandates are based on magical thinking and questioning whether the power of the state should be used to enforce this. Surely we have moved on from the Salem witch trials?

Here in the UK, there is growing chorus of voices demanding that the Westminster government should reinstate a mask mandate in indoor spaces in England in response to a rising rate of positive PCR tests – the Scottish and Welsh governments have never relaxed theirs. In this clamor, little attention is being given to the publication, on 14 October, by the Health Security Agency (HSA) of an expert review of the evidence and a statement of the panel’s conclusions. This is the considered position of a blue-riband, interdisciplinary committee. As such, it is worth attending to, whether you are reading this from the UK or elsewhere. The government is, of course, free to take other advice and the review only covers literature up to 28 April.

The first thing to note is that the panel excluded any evidence from studies with other respiratory infections, unless this happened to be included in the systematic reviews they considered. The focus on COVID-19 might be justifiable in making the task more manageable and ensuring that any specific differences between this virus and others were not complicating the picture. However, it also overwrote the acknowledged weakness of that evidence base and its historic failure to demonstrate any clear or significant benefits from the use of surgical or cloth masks in community settings.

For this kind of question, randomized controlled trials (RCTs) are the best evidence. An authoritative overview of previous RCTs in 2020 by the respected Cochrane Collaboration concluded that wearing a face covering made little or no difference to the likelihood of infection with flu or any other flu-like illness. The HAS panel’s criteria only included one, inconclusive, trial from Denmark, published online in November 2020, and did not include the trial from Bangladesh, reported later in 2021. Another trial in Guinea-Bissau seems to have sunk without trace.

This means that the HSA panel started with a second-class evidence base. The language used in their analysis acknowledges this: observational studies ‘may be influenced by selection bias…and recall bias’; ecological studies ‘provide results at population level that may not apply at individual level’; laboratory studies ‘provide mechanistical evidence and do not always take into account real-world conditions’. The most relevant systematic review, by Kim and colleagues, found that the risk reduction from surgical and non-medical masks was not statistically significant, but this is a pre-print rather than a peer reviewed publication.

Nevertheless, the panel go on to state that ‘the evidence suggests that the use of face coverings within the community is effective in helping to reduce transmission’. They give two sources for this. One is a continuously updated review by a team from Oregon. When we go back to the source, however, we find that the original authors concluded that ‘the strength of evidence for any mask use versus nonuse in community settings remains low’. They did not change this conclusion in a later update, which was not available to the panel. Moreover, in their view, the literature was now moving so slowly that their updates, having shifted from monthly to bi-monthly could now be issued every six months. The other source is a previous review from Public Health England, which is markedly less critical in its language when referring to the evidence base: it draws robust conclusions from methods that the panel have acknowledge to be weak.

Having acknowledged the fragility of the evidence base, it is worth moving on to the accompanying statement, which is what most people will read. This declares, with high confidence, that ‘all types of face coverings are, to some extent, effective in reducing transmission in community settings’ and, also with high confidence, that laboratory studies show that cloth masks may be as effective as surgical masks.

We might be forgiven for questioning whether these firm statements are entirely consistent with the more qualified tone of the evidence review. We might also note the absence of any attempt to quantify the potential benefit: a 2 percent reduction in transmission might not justify the potential harms, while a 25 percent reduction might be worth having.

Would it have made any difference if the Bangladesh RCT had been included? This seems unlikely. That study has yet to be peer reviewed. However, the evaluation of masks is clouded by a battery of other interventions. These succeeded in increasing mask compliance but had a relatively small impact on infections: 8.62 percent of controls and 7.62 percent of controls reported COVID-like symptoms. This could be presented as a 10 percent reduction – but the absolute numbers are so small that we might want to ask whether the benefits exceeded the harms. The data cast considerable doubt on whether there was any benefit from cloth masks and the benefit from surgical masks in the community was minimal.

The authors, however, strove to defend the case for face masks. This is a common trope. Kim et al declare that their metanalysis does not show a statistically significant benefit from face masks in the community but nevertheless suggest that there is a trend in that direction. Now I may only be a sociologist but when I was taught quantitative methods, I was firmly instructed that this sort of comment was entirely inappropriate: either something was statistically significant or it was not. You did not try to wriggle around and preserve a hypothesis by claiming that the significance level in your study was almost there.

Something of the same seems to have happened with the HSA panel. “To some extent” might mean anything but allows everyone to sign on. Even I could have signed up to it. I am happy to concede that face coverings may have some benefit but that it is so small and uncertain that it is not justified as an intervention, particularly one backed by legal sanctions.

The state of the face mask debate is rather as if Galileo had published his account of the heliocentric universe and then included a paragraph at the end telling the reader to ignore all the evidence because the Church had declared that everything revolved around the Earth. In the absence of better-quality work – and we must ask why that research has not been done – some of the claims for face masks look much more like magical thinking than anything that demonstrates the sort of casual connection that might be recognizable as science. If the UK’s Health Security Agency and its chosen experts cannot identify a clear, evidence-based benefit from face masks in the community, who am I to disagree?