The question of how effective mask-wearing is at preventing the spread of Covid appears, at first glance, to have been settled. According to a meta-analysis published in the British Medical Journal (BMJ) on 18 November, use of face coverings reduces infection rates by 53 per cent. So too, and to the exact same degree, does regular hand-washing. Social distancing was reported to have a more modest 25 per cent protective effect.
Proponents of non-pharmaceutical interventions (NPIs) – all the stuff that helps control a pandemic without involving vaccines or drugs – will seize on these findings as definitive. Individual studies into any medical question invariably give a range of results. Meta-analyses, however – where the results of numerous individual studies are blended together – are regarded as authoritative, smoothing out these fluctuations and giving a far “truer” picture.
But although meta-analysis has become a byword for the highest quality evidence, in this instance it has been used somewhat tenuously. Yes, the BMJ paper did synthesise data from multiple studies so it was, strictly speaking, a meta-analysis. But a brief look beyond the headlines belies the term.
The first problem the authors had was a paucity of material. They identified 650 papers worthy of a detailed look, of which 72 ultimately met the inclusion criteria. Of these, only eight ended up being of sufficient relevance to be enjoined in the meta-analysis and only one of them was a randomised controlled trial (RCT). RCTs are the most powerful tool medical science has for investigating interventions. In essence, you take a large group of participants, divide them at random so that the two subgroups are similarly varied in terms of demographics, medical history, etc, then one group is subjected to the intervention (in this case, mask-wearing) while the other acts as the control. You compare the outcomes – Covid infections – and see if one group differs from the other.
The seven additional studies included in the meta-analysis were observational: they looked at differences in Covid rates between people in the everyday world and tried to explain them in relation to how much the subjects practised various NPIs. This kind of research is much more susceptible to confounding. That 53 per cent reduction in Covid from hand-washing? The longer the pandemic has gone on, the less important hand hygiene has actually proved to be. It probably comes out so strongly in this meta-analysis because people who do a lot of hand-sanitising are also more likely to behave in various other ways that coincidentally reduce infection risk.
The BMJ paper may do more harm than good. What RCT evidence there is for mask-wearing shows a protective effect, but it looks likely to be much lower than the quoted 53 per cent – potentially as low as 10 per cent. That’s still well worth having, but overstating it so starkly only fuels NPI-sceptics’ cynicism.
The most important implication of the study is the light it shines on global research priorities. The pandemic has presented virtually limitless opportunities to study what things work and how well. The bulk of the world’s spending on Covid-related research has gone on pharmaceutical interventions. Only around 4 per cent has funded efforts to establish the relative importance of different NPIs. The pandemic afforded a huge laboratory in which to research important questions that would better equip us to handle such events in the future. We squandered it.
This bias in favour of technical, pharmaceutical – and therefore industrial, profitable – solutions in health runs through modern medicine. I can tell you in bespoke detail how much taking a daily statin might reduce your risk of heart disease. I cannot tell you exactly how much better you might be for taking a daily walk. Every nuance of our Covid vaccines’ effectiveness and rare downsides is minutely detailed. Yet we still do not know how useful face masks – or indeed something as simple as better ventilation – might prove to be.
This article appears in the 24 Nov 2021 issue of the New Statesman, The Agent of Chaos