Preprint Review Version 2 Preserved in Portico This version is not peer-reviewed

Face Masks Against COVID-19: An Evidence Review

Version 1 : Received: 10 April 2020 / Approved: 12 April 2020 / Online: 12 April 2020 (17:41:10 CEST)
Version 2 : Received: 12 May 2020 / Approved: 13 May 2020 / Online: 13 May 2020 (13:16:04 CEST)
Version 3 : Received: 10 July 2020 / Approved: 12 July 2020 / Online: 12 July 2020 (16:11:50 CEST)
Version 4 : Received: 30 October 2020 / Approved: 2 November 2020 / Online: 2 November 2020 (10:18:00 CET)

A peer-reviewed article of this Preprint also exists.

Journal reference: PNAS 2020
DOI: 10.1073/pnas.2014564118


The science around the use of masks by the general public to impede COVID-19 transmission is advancing rapidly. Policymakers need guidance on how masks should be used by the general population to combat the COVID-19 pandemic. Here,we develop an analytical framework to examine an overlooked aspect of mask usage: masks as source-control targeting egress from the wearer with benefits at the population-level, rather than as PPE used for ingress control for health-care workers with focus on individual outcomes. We consider and synthesize the relevant literature to inform multiple areas: 1) transmission characteristics of COVID-19, 2) filtering characteristics and efficacy of masks, 3) estimated population impacts of widespread community mask use, and 4) sociological considerations for policies concerning mask-wearing. A primary route of transmission of COVID-19 is likely via respiratory droplets, and is known to be transmissible from presymptomatic and asymptomatic individuals. Reducing disease spread requires two things: first, limit contacts of infected individuals via physical distancing and other measures, and second, reduce the transmission probability per contact. The preponderance of evidence indicates that mask wearing reduces the transmissibility per contact by reducing transmission of infected droplets in both laboratory and clinical contexts. Public mask wearing is most effective at reducing spread of the virus when compliance is high. The decreased transmissibility could substantially reduce the death toll and economic impact while the cost of the intervention is low. Given the current shortages of medical masks we recommend the adoption of public cloth mask wearing, as an effective form of source control for now, in conjunction with existing hygiene, distancing, and contact tracing strategies. We recommend that public officials and governments strongly encourage the use of widespread face masks in public, including the use of appropriate regulation.


COVID-19; SARS-CoV-2; masks; pandemic


MEDICINE & PHARMACOLOGY, General Medical Research

Comments (17)

Comment 1
Received: 13 May 2020
Commenter: Jeremy Howard
Commenter's Conflict of Interests: Author
Comment: Added new references that have been added in last month. Added more detail on Bae et al. Renamed some sections. Moved some paragraphs between sections. Fixed some notation issues. Added paragraph about face shields and other coverings. Fixed some grammar and style issues.
+ Respond to this comment
Comment 2
Received: 18 May 2020
Commenter: JN
The commenter has declared there is no conflict of interests.
Comment: I don't think anyone really questions whether a new CLEAN mask applied correctly COULD offer some protection.

But what happens after an hour or 2 hours of using a cloth mask by an infected person? Haven't we now introduced a new vector for transmission, that is probably spraying more virus into the air, possibly even smaller virus particles that have evaporated and are now traveling further? In addition, you only need to spend 5 minutes watching how people are using masks to realize that they are actually touching this surface more than they might normally even touch their face.

This paper also fails to mention the outcome of the Seongman Bae study, that ALL of the masks showed more contamination on the OUTSIDE of the masks. And further, concluded that "both surgical and cotton masks seem to be ineffective in preventing the dissemination of SARS–CoV-2 from the coughs of patients with COVID-19 to the environment and external mask surface".
+ Respond to this comment
Response 1 to Comment 2
Received: 20 May 2020
Commenter: Stephen Lowell
The commenter has declared there is no conflict of interests.
Comment: If you read the actual study (and the revealing comments between researchers) you will find this is not at all comparable to the current situation and this study is far from saying anything definitive about the lack of effectiveness of masks. First, this is a study looking at people coughing from 20 cm away. Nobody is saying it is safe to let a sick person cough in your face at a close distance just because they are wearing a mask. Masks are in addition to social distancing, not in place of it. In short, the Seongman Bae study did not aim to verify whether surgical or cloth masks are able to shorten the trajectory of droplets emitted through coughing, speaking or breathing. Second, and more to your point, in the study itself, compared with the control condition, the petri dish viral load was less with a cloth mask for all patients, and in half, was not detectable at all. Such reductions do help at the population level. So, the data from the study actually support mask-wearing, coupled with other measures, as a precautionary strategy.
Response 2 to Comment 2
Received: 29 May 2020
Commenter: Zac G
The commenter has declared there is no conflict of interests.
Comment: A virus cannot multiply outside the body like a bacteria can. The guidelines regarding face mask handling originate from the WHO and seem to be the general guidelines meant to also protect against spread of bacteria.

Unlike a bacteria, a virus is unable to multiply outside the host. The viral load on the outside of the mask is low compared to the viral load in saliva that is expelled from the mouth during speaking (especially susceptible are fricatives 'f', 's', 'v' and 'z' as well as sounds like 'p' and 't'). During a conversation at close distance and/or in noisy environments, some saliva may be transferred between persons as it is expelled from a speaking person and enters the mouth of a second person.

It seems reasonable to assume that fresh saliva being passed from an infected persons' mouth to a non-infected persons' mouth must have a much higher chance of spreading the virus than any mechanism by which viruses on the outside of a face mask can spread. It is difficult to imagine how a viable viral load would be able to transfer from the outside of a mask and cause a new infection. During every transfer from one surface to the next, the number of viruses on a surface decreases exponentially at minimum due to dilution alone. Additional decay of viruses that are damage during the transfer from one surface to the next increases the rate of decay even further.

Given the (speculated) low chances of the virus spreading through face mask or through surfaces in general, it seems awkward that the WHO is putting so much emphasis on face mask handling. Although this is appropriate for a hospital setting where health-care workers are potentially exposed to dangerous and resistant bacteria, this handling seems unnecessary and even counterproductive for face mask wearing, as it is used to argue against the wearing of face masks by the general public by some national health organizations.
Response 3 to Comment 2
Received: 5 June 2020
Commenter: Sampo Smolander
The commenter has declared there is no conflict of interests.
Comment: The Seongman Bae et al. 2020 study has now been retracted:
Comment 3
Received: 19 May 2020
The commenter has declared there is no conflict of interests.
Comment: I could not download the article, but when it comes to single-layered cloth masks I find no logic in wearing them before validating in real-life scenarios.
The only real-life scenario studied so far by C Raina MacIntyre et al. in a Vietnamese hospital showed that cloth masks were significantly less effective than controls (normal practice which involves surgical masks) in preventing infection. The models all systematic reviews published so far indicate a lack of significance for wearing masks in public. When you have over 11 studies (B. J. COWLING et al. 2010) show lack of significance and suddenly one says there is significance for wearing a cloth mask I would get suspicious. The only significant difference found in the above mentioned systematic review was for a controlled experimental setup where volunteers where asked to cough towards a petri dish and showed that nothing was on the petri dish compared to volunteers not wearing masks. The study did not investigate the leaked particles from the sides of the mask which in a real-life scenario might be even more scary than those coming directly out of the subject.
There are many conditions that need to be taken into account when considering cloth masks, like humidity, the concentration of viral droplets in the environment, whether the purpose is to protect others or protect from others, whether the subject is touching the mask, whether there are viruses on the surface around the subject. If the protection is less than 5% better than a control, then it should not be adopted in my opinion.
+ Respond to this comment
Response 1 to Comment 3
Received: 25 May 2020
Commenter: Paul V Sullivan
The commenter has declared there is no conflict of interests.
Comment: Face Mask Use and Control of Respiratory Virus Transmission in Households - MacIntyre, CR et al - Emerging Infectiuos Diseases - Feb-2009

The authors state that masks are effective, but in households the compliance was less than 50%. Doesn't help if you don't wear it.

The above article is freely downloadable.
Comment 4
Received: 21 May 2020
The commenter has declared there is no conflict of interests.
Comment: With supply of surgical masks slowly being restored (soon there should be no more shortage) recommendation of surgical masks over cloth masks should be made, given that they use of surgical masks by public should in no way affect supply to healthcare workers.
+ Respond to this comment
Comment 5
Received: 28 May 2020
The commenter has declared there is no conflict of interests.
Comment: There’s a short paragraph on risk compensation behavior which merely cites past instances where risk compensation was not enough to offset a measure (eg seat belts). What was the thought process for including only those cases? How about cases where risk compensation might have been enough to offset? For instance on artificially sweetened beverages for weight control, Borges MC, Louzada ML, de Sá TH, Laverty AA, Parra DC, et al. (2017) Artificially Sweetened Beverages and the Response to the Global Obesity Crisis

The paragraph then ends listing countries with mask policies that had a “successful” response. There’s no indication as to why masks were a net positive (and risk compensation was not an offset), other than a statement from the government in those countries. What’s the reasoning for even including that argument there? It doesn’t seem to be well grounded.
+ Respond to this comment
Comment 6
Received: 1 June 2020
Commenter: FvhG
The commenter has declared there is no conflict of interests.
Comment: Could the authors please clarify the graph shown in Figure 1?

Given an R_0 of 2.4, a reproduction number R_e of 1 requires a 58% decrease of the reproduction number (1 - (1/2.4)). The graph however shows that a ~40% adherence rate combined with a 100% mask efficacy reduces the R_e to 1. This seems incorrect given an R_0 of 2.4. Absent any other factors, it is expected that an R_0 requires an adherence rate of 58% in order to obtain an R_e equal to 1.
+ Respond to this comment
Response 1 to Comment 6
Received: 19 July 2020
Commenter: Manfred Schopp
The commenter has declared there is no conflict of interests.
Comment: Facemasks have double effect:

With 100% mask efficacy, 35.5% usage reduces infections due to reduced emission by 35.5% and again by 35.5% due to reduced intake.

Hence formula to get from 2.4 to 1 is x= 1 - sqrt(1/2.4)
Comment 7
Received: 7 June 2020
Commenter: Bruce L. Daniel
The commenter has declared there is no conflict of interests.
Comment: There are two kinds of masks/respirators. Those with valves and those without. Valves (which are even present on some n-95 masks) prevent unfiltered ingress, but allow totally unfiltered egress of exhaled air, droplets and nuclei. Given the importance of preventing egress highlighted by this article, the use of masks with valves should be discouraged.
+ Respond to this comment
Comment 8
Received: 25 June 2020
The commenter has declared there is no conflict of interests.
Comment: This article does not address the major counterpoint to its conclusion about masks for the general public, which would be the study by Jingyi Xiao, Eunice Y. C. Shiu, Huizhi Gao, Jessica Y. Wong, Min W. Fong, Sukhyun Ryu, and Benjamin J. Cowling. This study comes to the conclusion, after studying 10 relevant randomized controlled trials (RTC's) that "Ten RCTs were included in the meta-analysis, and there was no evidence that face masks are effective in reducing transmission of laboratory-confirmed influenza. [with the caveat that 'A study suggested that surgical and N95 (respirator) masks were effective in preventing the spread of influenza']" Page 26 of the WHO publication "Non-pharmaceuticalpublic health measures for mitigating the risk and impact of epidemic and pandemic influenza".

The study is available in two forms, a summarized form from the CDC and a full article from the WHO:
The SARS-CoV-2 virus that causes the respiratory disease COVID-19 is for all intents and purposes an Influenza respiratory disease in that it functions and behaves as previous influenza respiratory diseases (with relevant specific deviations and strains which influenza's exhibit). The prevalent detrimental portion of COVID-19, as with other respiratory diseases, being the pneumonia after the virus has run its course. See:
+ Respond to this comment
Response 1 to Comment 8
Received: 28 June 2020
Commenter: FvhG
The commenter has declared there is no conflict of interests.
Comment: The WHO-published study cited by you (titled "Non-pharmaceutical public health measures for mitigating the risk and impact of epidemic and pandemic influenza") recommends face masks wearing by the public in case of a high severity pandemic (see page 3).

The studies that measure effectiveness of face mask wearing which are included in their meta-analysis are mainly in household settings. The study notes that "Some studies reported that low compliance in face mask use could reduce their effectiveness." (page 26). The low or unknown mask wearing compliance in these studies prevent any conclusion to be drawn. The study therefore concludes that "there was no evidence that face masks are effective in reducing transmission of laboratory-confirmed influenza."

The study recommends face mask wearing under circumstances currently present as the result of SARS-CoV-2 on the basis of mechanistic plausibility:

"Face masks worn by asymptomatic people are conditionally recommended in severe epidemics or pandemics, to reduce transmission in the community. Although there is no evidence that this is effective in reducing transmission, there is mechanistic plausibility for the potential effectiveness of this measure.

A disposable surgical mask is recommended to be worn at all times by symptomatic individuals when in contact with other individuals. Although there is no evidence that this is effective in reducing transmission, there is mechanistic plausibility for the potential effectiveness of this measure." (page 14).

The WHO-published study is therefore not a counterpoint to the conclusion of this studies, because:
  • Mere absence of statistically significant evidence is not invalidating this studies;
  • The WHO-published study recommends the wearing of face masks by the public during a high severity pandemic;
  • The WHO-published study predates this studies and more empirical evidence whether face masks are effective has become available;
  • The WHO-published study is about influenza which has differing characteristics compared to SARS-CoV-2.
Response 2 to Comment 8
Received: 30 June 2020
The commenter has declared there is no conflict of interests.
Comment: Thank you for your comments.

The point I was making was that this article uses much of the studies that the study the WHO and CDC article, listed above in my original comment, would have also reviewed, and found them either to be overly bias (page 11 of WHO publication) or did not provide actual evidence during the Randomized Controlled Trials (RCT), that masks actually had the desired effect. Here they recognized the RCT's as the best actual evidence, as opposed to observational (which was lower than RCT but higher than computational model) or computer models - since these last two can be more prone to assumptions and bias that can lead to improper conclusions. These studies reviewed by the meta-study that underlies the WHO and CDC articles were from the dates 1946–July 27, 2018. While this pre-print article does include some newer publications, it appears it would be strengthened at minimum, if it also applied the same criteria to it's sources which are newer (addressing possible bias and evidence in the studies used), and to address the areas of older sources which would also be part of the underlying study in the WHO and CDC publications listed above, but were found to not actually support RCT evidence that masks had a significant impact on influenza.

I believe this is particularly important, since the point of this pre-print is explicitly geared toward influencing public policy. It also relies on studies that could easily be attributed to other, although probably less likely, conclusions. For instance, it talks about a case study of airline passengers coming from China. Yet, the underlying articles does not lead to the actual study. For instance, since the flight is coming from China, and some areas of China having been exposed to SARS-Cov-2 since Aug 2019, could it be that most of the passengers and crew were from that region and already had antibodies (had herd immunity been reached), or where was the COVID-19 passenger in their infectious stage, etc.. The second case study actually dealt with Influenza A, but made the statement "We recommend a more comprehensive intervention study to accurately estimate this effect." In regards to mask effectiveness as a recent meta-study, I can find similar articles, from authoritative publication sources, that point out opposite conclusions (that masks are somewhat effective in certain aspects, but have no clinical evidence that they actually have a significant impact on influenza), such as, as an example (although it is classified commentary, it does much of the meta-study review with citations):
  • .

  • This underscores the need to actually deal with any study bias or assumptions, and using a similar approach that the study that underlies the CDC and WHO publicans used, could help alleviate similar studies that have come to opposite conclusions (or at least not as strong of a conclusion as to the effectiveness of masks).

    In addition, my other underlying point to this pre-print, since it explicitly is geared toward influencing public policy, is to ask the fundamental question: Is suppression of the virus actually the best course of action for those who are not fragile, or who are not protecting someone at home, or in a nursing home, who is fragile and has comorbidities? Up until now, the reason given for suppression and flattening the curve have relied upon a number of key assumptions, mainly that a vaccine was needed, and secondly that people would flawlessly perform until the vaccine was completed. Neither of these are reasonable in reality - humans are social and there will be contacts (particularly within households), and vaccines are also difficult to develop successfully for coronaviruses (not to mention anything of the actual approval process). There are other factors that should be considered - what about cross immunity as a form of immunization? What about natural herd immunity? What about the fact that COVID-19 as a virus is not really the detrimental part, it is the pneumonia that follows it (which with quick identification can be combated with antibiotics just as it would be with any influenza respiratory disease)? As Dr. Knut M. Wittkowski has pointed out in the links I provided above, suppression only keeps the respiratory disease in the population longer, widens the actual impact, and prevents it from being destroyed through natural herd immunity (or possibly by using a cross-immunity with SARS-Cov[-1]). While this is outside of the scope of the pre-print at first glance, since this pre-print is explicitly is geared toward influencing public policy, I believe it would be much stronger if it actually approached the potential assumption and bias topics of the underlying studies used and even delved into why it believes suppression, rather than quickly fostering natural herd immunity while attempting to protect the vulnerable, is the desired goal.

    In regards to the bullet points:
    • "Mere absence of statistically significant evidence is not invalidating this studies;"
      • This is true, but as I mentioned above, many of the studies this pre-print meta-analysis are dealing with, in regard to RCT's, would have been part of the underlying study of the CDC and WHO publications, that found either significant bias or inconclusive evidence for RCT's (with observational and computational being rated lower in actual evidence than RCT's). The newer studies mentioned in this article are not analyzed by the pre-print using the same type of evidence / bias analysis, which at minimum could make this study stronger (as well as address why the two studies come to fairly different conclusions - ie. how strong is the evidence that is underlying this meta-study pre-print).
    • "The WHO-published study recommends the wearing of face masks by the public during a high severity pandemic;"
      • They did, but this, as you pointed out was based on mechanistic plausibility, not on actual evidence from RCT's. In other words, it does not hurt anything, and there is some reason to believe it might provide some suppression even though we do not have any RCT evidence that it actually does.
    • "The WHO-published study predates this studies and more empirical evidence whether face masks are effective has become available;"
      • This is true, but again, much of the studies it included would have also been a part of the underlying study meta-analysis that the CDC and WHO publications were based upon. You mentioned household studies, but the CDC publication acknowledges that of the 10 RCT studies available in this time period, 7 were household studies - but even then, this pre-print article is recommending as a public policy that masks be used mainly by the full population, not just on the basis of it-might-help, but that there is significant evidence that it-does-help (if that is the case, and the majority of the people it would impact would be in the "home" category, it seems that the fact that 7 of the 10 studies were "home" studies actually strengthens the point that the evidence of available RCT's show that masks are not effective in real life scenarios). But, with the caveats I mentioned above, I concede it is a fair point.
    • "The WHO-published study is about influenza which has differing characteristics compared to SARS-CoV-2."
      • Although the SARS's virus' are coronaviruses, they functions as an influenza respiratory disease would (as is pointed out by Dr. Wittkowski in the links I mentioned above), and for these purposes can be treated similarly. But, in regards to the mechanics of masks, the actual virus size of actual Coronavirus' are about 17nm to 90nm, whereas for Influenza it is about 80nm. These are way under the micrometers that masks will be able to block. As for the aerosol transmission aspect, of non-N95-or-above masks, they filter about .06 to .1 micrometers, and aerosols can be in the range of 10nm to 10 micrometers. So, in reality, for masks, the underlying study represented in the CDC and WHO publications that I linked to in my original comment would have the same mechanics for SARS related coronaviruses.
    Comment 9
    Received: 13 July 2020
    Commenter: Justin Nathaniel White
    The commenter has declared there is no conflict of interests.
    Comment: It seems as though point #6 has been glossed over. The case studies have been done with medical staff who know how to properly use and replace masks.

    There are many examples of the general public fidgiting with masks and not properly sealing masks allowing any potential benefit to be lost. Additionally, it creates a 'magic glove' syndrome where they may wear the masks for too long and not replace these.

    The report handwaves these as insignificant and claims these concerns will not affect the finding.

    While public information campaigns could prove useful, the uptake on similar public awareness re:seatbelts and drinking and driving bear looking at to see if these measures could be adopted in time to have any significant impact.
    + Respond to this comment
    Comment 10
    Received: 11 August 2020
    Commenter: Noam Peleg
    The commenter has declared there is no conflict of interests.
    Comment: Article presume the primary route of transmission of SARS-CoV-2 in droplets of size minimum of 5µm to 10 µm. Masks are efficient for this size of droplets but what if the virus is transmitted in aerosols of smaller size. To clear this point with evidence and know the efficacy of the masks we need a randomized controlled trial in SARS-CoV-2. Unfortunately, in the article there are no reference to such a trail.
    + Respond to this comment

    We encourage comments and feedback from a broad range of readers. See criteria for comments and our diversity statement.

    Leave a public comment
    Send a private comment to the author(s)
    Views 0
    Downloads 0
    Comments 17
    Metrics 0

    Notify me about updates to this article or when a peer-reviewed version is published.
    We use cookies on our website to ensure you get the best experience.
    Read more about our cookies here.