select search filters
briefings
roundups & rapid reactions
before the headlines
Fiona fox's blog
A randomised control trial (RCT) published by Annals of Internal Medicine assesses mask use as a public health measure to help prevent SARS-CoV-2 infection in Danish mask wearers.
Dr Julian Tang, Honorary Associate Professor/ Clinical Virologist, Respiratory Sciences, University of Leicester, said:
“This study may well be representative of the compliance and impact of mask use in this poorly compliant Danish population, but correspondingly, if this study had been performed in a Southeast/East Asian population (e.g. Taiwan, South Korea, Hong Kong, Vietnam, Thailand, China), where masking frequency is much higher, the benefits would also be shown to be much higher.
“During the ongoing COVID-19 pandemic, the compliance with universal masking in these Southeast/East Asian countries continues to be very high relative to Western countries – with clear benefits for virus control.
“The relatively lower apparent benefits in Western countries are mostly due to a ‘late start’ (after major virus outbreaks had already seeded many of the population) and earlier, poorer compliance and resistance (‘masks don’t work, don’t wear them’ – now they do) – though fortunately this has improved now.
“Also, given the relatively low (<50% compliance), a possible 50% protective effect seems pretty good in this otherwise poorly compliant Danish population. So this would likely be much higher – perhaps 70-90% in a much more compliant Southeast/East Asian population.
“This study does not change my view that earlier universal masking would have prevented a lot more cases/deaths from COVID-19 in the UK, Europe and the Americas if it had been adopted earlier and consistently by everyone – in addition to other interventions, like social distancing and hand-washing.”
Prof Trish Greenhalgh FMedSci, Professor of Primary Health Care Services, University of Oxford and Dr Deepti Gurdasani, Senior Lecturer in Machine Learning, QMUL, said:
“We are concerned about aspects of the DANMASK-19 trial and the inaccurate way it is being reported in some press reports.
“DANMASK-19 was an unblinded RCT of surgical mask wearing in 6000 people, to test whether masks protect the wearer (the authors did not look at the more important question of source control – whether masks protect other people). Notably, the authors themselves comment: ‘The findings, however, should not be used to conclude that a recommendation for everyone to wear masks in the community would not be effective in reducing SARS-CoV-2 infections, because the trial did not test the role of masks in source control of SARS-CoV-2 infection.’ It is very troubling, therefore, that the press and some policymakers are already interpreting the study as meaning that mask mandates should be abandoned. The study does not support that conclusion.
“Whilst the study involved random allocation of participants (and therefore counts as a randomised controlled trial or RCT), this does not necessarily make it high-quality science. On the basis of the published protocol and paper, there appears to have been no involvement of a clinical trials unit (the usual way of quality-checking a RCT), no data monitoring group (again, to be expected in a high-quality RCT), and (most concerning of all) no formal ethical approval. The claim that such approval ‘was not needed’ when blood tests and swabs were taken from 6000 people is very surprising. The gold standard criteria for RCTs are set out in the CONSORT statement. Most leading medical journals will not publish a RCT unless the CONSORT criteria are met (indeed, the Annals of Internal Medicine which published the DANMASK-19 trial, appears elsewhere to have signed up to the CONSORT standards). The DANMASK-19 authors did not mention the CONSORT standards in their paper.
“The DANMASK-19 trial, which recruited people mainly from media advertisements, involved only ‘DIY’ tests which were not checked by a professional (would people’s reading of the test be influenced by the fact that they felt protected or unprotected? – possibly; would lay people be as good at interpreting the test result as a health professional? – unlikely). The intervention group were sent a kit and asked to wear a mask when went out (fewer than half the people in this group complied fully); the controls were asked to not use masks. Everyone in the trial was also asked to follow social distancing and the incidence of COVID-19 was low at the time, so it is unsurprising that very few infections occurred in either group. Participants self-administered an antibody test the day the intervention began and the day it ended, and additionally if they became symptomatic during it.
“The researchers used a composite outcome: a positive swab or positive antibody test for SARS-CoV-2 or hospital-diagnosed COVID-19. This composite outcome was reached in 42/3030 people (1·8%) in the mask group and 53/2294 (2·1%) in the control group. This difference was not statistically significant. However, both antigen (swab) tests and antibody tests were problematic. The number of positive swab tests (the best test of infection) was *zero* in mask group compared to 5 in the no-mask group. Given the very low incidence of COVID-19 at the time and the too-short intervention period, these findings are consistent with a real effect of masks in protecting the wearer (the lack of statistical significance may be due to what is known as a type 2 error).
“The intervention period was only 30 days. This is far too short given the incubation period of COVID-19, which can be up to 14 days. This means that a) people who developed a positive swab in the first 14 days of the study may have been infected before the study began, and b) people infected in the last 14 days of the study may not have developed a positive swab till after the study ended!
“The majority of composite outcomes (90 of 95) were based on antibody tests, rather than swab positivity. The antibody tests used (Zhuhai Livzon Diagnostic Inc) were highly problematic and very likely inaccurate in the context of the trial. They are no longer recommended by the US Food and Drug Administration, and indeed are on the FDA’s removed list of tests. The documents submitted to the FDA show evidence of validation only at 17 days or more after symptoms, and have been shown to have 0% sensitivity in the first 8 days after symptoms based on the current literature. This means that most of the antibody positives in the trial were very likely related to infection prior to the onset of the study, not to infection during it. A letter published in The Lancet raises further concerns and advises against using this particular test.
“In the authors’ ‘intention-to-treat’ analysis, 20% of those with positive antibody tests did not have any test done at baseline (so were not confirmed to be either positive or negative), but if tested positive during or after the study, were assumed to have become positive during the course of the study. A sensitivity analysis was done excluding these 20%, but the science of including them in the first place is highly questionable. With 20% of participants excluded, the sample does not meet the required size in the power calculation.
“The sample size calculation assumed that anything less than a halving of risk to the wearer was not clinically significant, but actually there may be much cumulative benefit to a community if transmission is reduced by a much smaller percentage, since pandemic spreads exponentially. Numerous modelling studies have shown that there would be a major reduction in community transmission with a much less than 50% efficacy of mask. For example:
“Additionally, if: the low degree of full compliance with mask use (47% only); the impact of using antibody tests that have low sensitivity in early infection (even the sensitivity after 17 days of symptoms would only be 80% ); the misclassification of outcomes due to a large proportion of positive tests reflecting infection occurring prior to intervention and negative tests in early infection in the latter part of the study; and ~20% tests not having baseline confirmation of negativity are considered, the power to detect even a 50% difference would be modest, and far lower than the 80% described in the study.
“If baseline risk of SARS-CoV-2 infection is 2%, to detect (say) a 20% reduction in risk would require a total sample size of ~35,000, and this would be conditional on high compliance, and no attrition during follow up.
“A number of other scholars have published concerns:
Dr Simon Clarke, Associate Professor of Cellular Microbiology at the University of Reading, said:
“This is a well-designed and carefully presented study. It provides very good evidence confirming what many people suspected: that wearing a facemask in public, while others around you don’t wear masks, does little or nothing to reduce your risk of being infected by the coronavirus. In fact, it might even slightly increase your risk of being infected.
“The people who took part in this trial were given a supply of free, high-grade disposable facemasks, and information on how to wear them, yet almost as many in this group caught COVID as those in a control group who did not wear masks.
“What the authors acknowledge is that the study doesn’t show if wearing facemasks protected others from infection by the wearers, if the wearers had coronavirus without realising it. Other studies have shown that facemasks can prevent transmission from mask wearers to others, which is why most countries have moved to policies to encourage mask wearing. Mass public facemask wearing will only work properly at preventing the spread of the disease if there is near universal compliance.
“Taken together, all the evidence shows that it is important for health authorities not to over-stress the effectiveness of facemasks as a way to protect wearers. If people think that wearing a mask means they are reducing their risk of being infected, they are very much mistaken. In the UK, campaigns that show how wearing a mask protects other people are important, so that the public do not become complacent about their vulnerability to infection from others.”
Prof James Naismith FRS FMedSci, Director of the Rosalind Franklin Institute, and University of Oxford, said:
“This paper is an attempt at a controlled trial of the benefits of mask wearing. We have to keep the following things in our mind. One, measuring the benefits of mask wearing will require many many thousands of people undertaking a controlled trial. Two, if we suspect masks confer a benefit, deliberately forcing humans not to wear masks (given mask wearing is a tiny burden) to answer a scientific question has ethical questions. We do not do experiments on humans that pose risk without good reasons. We do not do controlled real life trials of many safety measures, but instead rely on two things – observational data and plausible experimental models.
“Observational data suggest widespread mask wearing confers a small but meaningful benefit in reducing the spread of virus in the public. The same logic applies to hand washing – we don’t assign people to hand washing or non-hand washing groups, making sure they are matched for age, income, health etc, for a week then judge the effectiveness of hand washing. Observational studies show hand washing reduces viral spread. In experimental simulations, mask wearing reduces the spray of droplets that contain virus – it is therefore plausible that it could help reduce the chance of infecting someone else. Hand washing reduces the amount of virus on the hand – it is therefore plausible it will help reduce spread of the virus.
“This study measured the benefit of wearing the mask to the person with the mask on. Observational data and models suggest the likely benefit of simple masks (as opposed to respirators) is mainly to others, not the wearer. This study did not test this question.
“The credible interval of their data ranged from an increase of 23% to a reduction of 46% of infection to the mask wearer. The study design did not account for differences in behaviour – for example, did you always wear or not wear the mask when you said you did; as a trial participant, were you more careful with hand washing. The study lists a number of other important limitations. It concludes wearing a mask does not cut the infection rate of mask wearers by 50%. I would have been stunned if it had.
“It is highly unlikely that definitive ‘proof’ can ever be obtained for mask wearing. I think the resources that this would require would be better spent on researching new medicines, vaccines and treatments. With over 500 people announced as dead from COVID-19 over the last 28 days in the UK yesterday, there are much more important things for us to be getting on with. We should focus on reducing these tragedies.
“I will continue to wear a mask to protect others, and I would ask those who can to do so too.
“If I am wrong, the cost to me has been virtually zero. If I am right, then maybe someone avoided a serious illness since I have no way to way to know if I am or have been asymptomatically infected.”
Prof Paul Hunter, Professor in Medicine, UEA, said:
“The results of the DANMASK-19 randomised controlled trial on face mask use is a good study of the potential value of wearing a face mask to protect the wearer. It was not designed to show whether potentially infected individuals wearing facemasks would reduce the overall transmission of the infection.
“The DANMASK-19 study was a well-designed community study. Its main strength was the objective outcome measures used: ‘a positive result on an oropharyngeal/nasal swab test for SARS-CoV-2, development of a positive SARS-CoV-2 antibody test result (IgM or IgG) during the study period, or a hospital-based diagnosis of SARS-CoV-2 infection or COVID-19’.
“Swabbing and blood tests at one month would pick up most but not all infections, but this is unlikely to have biased the results and they are less likely to be biased than self-reported symptoms without a diagnosis confirmation.
“The findings were that infections were indeed reduced in the people wearing masks compared to those not wearing masks (1.8% v 2.1%), but this was not statistically significant. From the DANMASK-19 study the protective effect is 15%, but in the range 57% to -19% (i.e. no protection or some increased risk). This finding is in line with our own systematic review published in March, where we estimated the value of wearing masks as primary prevention was about 6% but in the range 20% to -19%. Adding this study to our own review would not materially affect our conclusions.
“As has been found with previous randomised trials of face masks compliance for influenza, face mask use compliance in the intervention group was not complete, with 46% of participants wearing the mask as recommended, 47% predominantly as recommended, and 7% not as recommended. So it could be argued that had everyone used their masks fully as recommended then the protective effect would have been greater, though in the real world people often do not wear masks as recommended so this study is likely to be a good proxy for mask use in the general public.
“This study does not add any evidence to whether infected individuals wearing masks protect others who are not infected. In our earlier review, we found many fewer epidemiological studies investigating this and what we found was that yes, there was some evidence, but that that was also not particularly strong.
“So this new paper strengthens the evidence base that wearing face masks can probably protect the wearer from infection by others, but their benefits are not great and are not an alternative to social distancing. The advice remains to continue to socially distance, and if you have to be in crowded enclosed spaces where you cannot keep your distance from others, then wear a face covering. But do not assume that if you are wearing a face covering you are necessarily protected.”
Dr Julii Brainard, Senior Research Associate in Modelling Public Health Threats, University of East Anglia, said:
“This is a well-run trial with enough participants to have high confidence in the results – therefore the statistical analysis was adequately powered and inherently adjusted for possible confounders, unlike most studies that try to make conclusions about mask-wearing and catching respiratory disease.
“This experiment was run in a situation where many social distancing measures were already being imposed, and yet most people were not wearing masks. Therefore this study informs what protection masks can add to that situation, not what protection masks by themselves may offer in absence of any other adjustments to social contact. Also, a study like this can only comment on the recruited individuals and what happened to them, not whether any of them were stopped from transmitting the infection to others who were not tested and not monitored. That said, the study found that mask-wearers had somewhat fewer infections than the controls who were not asked to wear masks.
“The findings are very similar to what emerged when we assessed earlier research on mask wearing to prevent influenza-like illness: that mask wearing appears to have a small protective effect to the wearers. The magnitude of the protective effect and its statistical significance are not at the thresholds that would normally be required to make a recommendation in favour of mask-wearing. This is the thorny difficulty of being in a pandemic situation: some actions may not help a lot, but small gains in health protection that can be achieved with negligible possible harms shouldn’t be ignored. It’s wrong for anyone to proclaim that this study shows that “masks don’t work”. This kind of experiment can’t prove that.
“In terms of what this means for policy: a trial like this can tell us what happened to individuals, but not what happens to average chains of transmission in the wider community. Mask-wearing mandates are imposed to try to cut total spread, not merely to protect individuals. We can infer from experiments like this that prevalence is likely to be lowered because slightly more transmission chains were interrupted when masks were worn. A small protective effect for individuals may have larger consequences for reducing harms to everyone when multiplied up over the whole population.”
Prof Ashley Woodcock FMedSci, Professor of Respiratory Medicine at the University of Manchester, said:
“This is a very valuable community study. The paper is very clear, the analysis correct and the interpretation appropriate.
“The investigators randomised over 6000 subjects to either wear masks ‘outside the home among others’ (each subject provided with 50 masks to use) or to wear no masks. The intervention was in addition to social distancing and hygiene recommendations which were in place in Denmark during the months of the study in 2020. About half of the mask group actually wore the masks as recommended (that is what happened in pragmatic studies!).
“At one month, there was no significant difference between the groups in terms of infection rates or immunological response. This suggests that in the real world, an overall public health recommendation to wear masks outdoors (which was adhered to by half of the participants in a trial… it would be less in the real world) would not protect individuals from acquiring SARS-CoV-2 infection.
“It’s worth noting that this study was still relatively small/short and only powered to look for a large change in efficacy (50% or greater). This shows the difficulty in real world studies of pragmatic interventions.
“The results make common sense when we consider what we know about coronaviruses. They are spread primarily by droplets during sneezing and coughing, and by direct or indirect contact with people/surfaces.
“Masks do reduce droplet spread, and this trial does not exclude an effect of a mask stopping someone carrying SARS-CoV-2 infecting others.
“Masks are much less effective in reducing exposure in aerosols (small particles which stay airborne); and there is still uncertainty as to the role of aerosols in spreading live virus.
“The risk of over-reliance on a potentially ineffective public health intervention such as wearing masks outdoors is that individuals lose focus on effective hygiene measures and physical separation.”
Dr Michael Head, Senior Research Fellow in Global Health, University of Southampton, said:
“The usefulness of face-masks is a very difficult thing to measure in trials. For example, in this trial compliance with the intervention was low, and also social distancing guidance was in place. This means it’s hard to discern the impact of facemasks in addition to the impact of all the other interventions. We don’t really get any useful new knowledge from this study, with the authors acknowledging the difficulties in drawing conclusions stating that ‘ the most important limitation is that the findings are inconclusive’.
“Overall, this study should not change guidance around wearing of face-masks, where the overall evidence base tells us that they are a useful contributory factor in reducing transmission.”
Prof KK Cheng, Professor of Public Health and Primary Care, University of Birmingham, said:
“In this pandemic, the main role of a mask is based on the principle of ‘I wear a mask to protect you and you wear a mask to protect me’, i.e. source control. However, this study only examined the protection for the wearers.
“Even in the context of their own objective, the authors acknowledged many limitations. These included low compliance, with only 46% of participants wearing the mask as recommended. Also, the incidence rates of COVID-19 during the two-month period of the study were not high, especially during the second month. The study was therefore powered to detect a 50% reduction in risk among the wearers, which was ambitious in an environment where few others were wearing masks.
“To answer the question of whether everyone wearing masks would reduce SARS-CoV-2 infections in the community, one would require randomising whole communities in a cluster design, with mass masking introduced in the intervention communities and no mask wearing in the control communities. One then compares the rates of infections between the two groups of communities. Such a design poses many difficulties, making a proper study nigh impossible.
“Overall, as the authors said, the results are inconclusive as far as their own study question is concerned. It is of even less relevance for policy-making in countries (e.g. the US) where the issue mandating mask for all is hotly debated.”
Dr James Gill, Honorary Clinical Lecturer, Warwick Medical School, and Locum GP, said:
“The concept of mask use is frequently debated. The prevailing wisdom is that masks save lives, hence the government advice of ‘Hands. Face. Space’.
“Science and medicine run on data. Throughout the early parts of the COVID-19 pandemic, expert opinion has driven policy, as research takes time to accurately and effectively generate data to provide a stronger, more robust guidance.
“The DANMASK-19 study has taken a remarkably simple concept – taking 6000 people, asking half to wear masks and the other half not to.
“Now there are several limitations to the study, not least that bias is huge – you can’t miss that you are wearing a mask! The bias point here is important, as the strongest studies have both the subject and the investigators “blinded”, so you don’t know who is in the control or in the test group.
“The study relied upon patients reporting findings of home test results if they developed symptoms. The teams attempted to reduce bias on reporting of symptoms by also testing each participant at the start of the trial via an antibody test, and again at the end. This allowed asymptomatic infections to also be detected
“The DANMASK-19 study has provided us with data on how effective masks are at controlling COVID-19 spread. Both groups unfortunately did have cases of COVID-19 infection, but thankfully in very small numbers. This is good for the patients, but hurts the data, in that the differences were not statistically significant – what that means is that it’s possible the results seen are due to chance, and that cannot be dismissed.
“What did the DANMASK study show? Given that we know the true answer will always exist in a range, the study has shown that mask wearing may be as effective as reducing 43% of infections – however 23% of infection may have been as a result of wearing the masks
“Having evidence that the masks are twice as likely to help as they are to harm is important. The DANMASK-19 study demonstrates that masks are part of the ‘Hands. Face. Space’ protocol, that none of the three components alone is sufficient to be relied upon, and that excess confidence on any of these three pieces of advice, at the detriment to others, is likely to result in an increased risk.
“Together, the ‘Hands. Face. Space’ approach is the best advice we have for people to protect themselves. The DANMASK-19 study strengthens the argument that all three components of this public health message are vital.”
Prof Babak Javid, Associate Professor of Experimental Medicine at the University of California, San Francisco, said:
“Unlike the situation at the beginning of the COVID pandemic, the majority of countries in the world now have some form of mask mandate to protect the population. There is a wealth of observational data to support mask use, but randomised controlled data on mask efficacy in a population setting is lacking in this context.
“Here, scientists in Denmark embarked on a randomised controlled trial of surgical mask use (or not) in the early phase of the first wave of the pandemic. All volunteers were asked to adhere to public health guidance (e.g. distancing, hygiene etc) and were then randomised to wearing surgical masks or not. The number of COVID-19 cases between the two groups after one month of randomisation was measured by antibody testing.
“The results were disappointing: mask usage was not associated with significant protection, although there was a great deal of uncertainty associated with the finding (ranging from c. 40% protection to 20% exacerbation).
“The authors are honest about the major limitations of their study: the study was conducted at a time when much of the country had some form of lockdown, limiting exposures outside of the household, and was not powered to detect smaller than 50% efficacy. The authors also didn’t measure mask compliance: either in the mask group (did they wear a mask when they were supposed to?) and in the unassigned group (did they wear a mask anyway?), and even small differences here could muddy interpretation of the results given the small number of COVID cases. Most importantly, the study could not measure the efficacy of arguably the most important aspect of how population mask wearing may influence outcomes in the pandemic: as ‘source control’ i.e. limit transmission of SARS-CoV2 by the wearer.
“Nonetheless, the authors should be congratulated on performing the study at all, at a time when most governments resisted mask mandates, but also didn’t engage in performing studies on potential efficacy. ”
‘Effectiveness of Adding a Mask Recommendation to Other Public Health Measures to Prevent SARS-CoV-2 Infection in Danish Mask Wearers’ by Henning Bundgaard et al. was published in Annals of Internal Medicine on Wednesday 18th November 2020.
DOI: 10.7326/M20-6817
All our previous output on this subject can be seen at this weblink:
www.sciencemediacentre.org/tag/COVID-19
Declared interests
Dr Michael Head: “No conflicts of interest to declare”
Dr James Gill: “I have a medical advisory role for Wise Protec.”
None others received.