"Face Masks Against COVID-19: An Evidence Review
Jeremy Howarda,c,1 , Austin Huangb , Zhiyuan Lik , Zeynep Tufekcim, Zdimal Vladimire , Helene-Mari van der Westhuizenf,g , Arne von Delfto,g , Amy Pricen , Lex Fridmand , Lei-Han Tangi,j , Viola Tangl , Gregory L. Watsonh , Christina E. Baxs , Reshama Shaikhq , Frederik Questierr , Danny Hernandezp , Larry F. Chun , Christina M. Ramirezh , and Anne W. Rimoint
a fast.ai, 101 Howard St, San Francisco, CA 94105, US; bWarren Alpert School of Medicine, Brown University, 222 Richmond St, Providence, RI 02903; cData Institute, University of San Francisco, 101 Howard St, San Francisco, CA 94105, US; dDepartment of Electrical Engineering & Computer Science, Massachusetts Institute of Technology, 77 Massachusetts Ave, Cambridge, MA 02139; e Institute of Chemical Process Fundamentals, Czech Academy of Sciences, Rozvojová 135, CZ-165 02 Praha 6, Czech Republic; fDepartment of Primary Health Care Sciences, Woodstock Road, University of Oxford, OX2 6GG, United Kingdom; gTB Proof, Cape Town, South Africa; hDepartment of Biostatistics, UCLA Fielding School of Public Health, 650 Charles E Young Drive, Los Angeles, CA 90095; iDepartment of Physics, Hong Kong Baptist University, Kowloon Tong, Hong Kong SAR, China; jComplex Systems Division, Beijing Computational Science Research Center, Haidian, Beijing 100193, China; kCenter for Quantitative Biology, Peking University, Haidian, Beijing 100871, China; lDepartment of Information Systems, Business Statistics and Operations Management, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong SAR, China; mUniversity of North Carolina at Chapel Hill; nSchool of Medicine Anesthesia Informatics and Media (AIM) Lab, Stanford University, 300 Pasteur Drive, Grant S268C, Stanford, CA 94305; oSchool of Public Health and Family Medicine, University of Cape Town, Anzio Road, Observatory, 7925, South Africa; pOpenAI, 3180 18th St, San Francisco, CA 94110; qData Umbrella, 345 West 145th St, New York, NY 10031; rVrije Universiteit Brussel, Pleinlaan 2, 1050 Brussels, Belgium; sUniversity of Pennsylvania, 3400 Civic Center Blvd, Philadelphia, PA 19104; tDepartment of Epidemiology, UCLA Fielding School of Public Health, 650 Charles E Young Drive, Los Angeles, CA 90095
This manuscript was compiled on April 10, 2020...
The science around the use of masks by the general public to impede COVID-19 transmission is advancing rapidly...
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...
"3. Filtering Capability of Masks
...Multiple studies show the filtration effects of cloth masks relative to surgical masks. Particle sizes for speech are on the order of 1 µm (20) while typical definitions of droplet size are 5 µm-10 µm (5). Generally available household materials had between a 49% and 86% filtration rate for 0.02 µm exhaled particles whereas surgical masks filtered 89% of those particles (21). In a laboratory setting, household materials had 3% to 60% filtration rate for particles in the relevant size range, finding them comparable to some surgical masks (22). In another laboratory setup, a tea cloth mask was found to filter 60% ofparticles between 0.02 µm to 1 µm, where surgical masks filtered 75% (23). Dato et al (2006) (24), note that "quality commercial masks are not always accessible." They designed and tested a mask made from heavyweight T-shirts, finding that it "offered substantial protection from the challenge aerosol and showed good fit with minimal leakage".Although cloth and surgical masks are primarily targeted towards droplet particles, some evidence suggests they may have a partial effect in reducing viral aerosol shedding (25).
When considering the relevance of these studies of ingress, it’s important to note that they are likely to substantially underestimate effectiveness of masks for source control. When someone is breathing, speaking, or coughing, only a tiny amount of what is coming out of their mouths is already in aerosol form. Nearly all of what is being emitted is droplets. Many of these droplets will then evaporate and turn into aerosolized particles that are 3 to 5-fold smaller. The point of wearing a mask as source control is largely to stop this process from occurring, since big droplets dehydrate to smaller aerosol particles that can float for longer in air (26).
Anfinrud et al (6) used laser light-scattering to sensitively detect droplet emission while speaking. Their analysis showed that virtually no droplets were "expelled" with a homemade mask consisting of a washcloth attached with two rubber bands around the head, while significant levels were expelled without a mask. The authors stated that "wearing any kind of cloth mouth cover in public by every person, as well as strict adherence to distancing and handwashing, could significantly decrease the transmission rate and thereby contain the pandemic until a vaccine becomes available."
An important focus of analysis for public mask wearing is droplet source control. This refers to the effectiveness of blocking droplets from an infectious person, particularly during speech, when droplets are expelled at a lower pressure and are not small enough to squeeze through the weave of a cotton mask. Many recommended cloth mask designs also include a layer of paper towel or coffee filter, which could increase filter effectiveness for PPE, but does not appear to be necessary for blocking droplet emission (6, 27, 28).
In summary, there is laboratory-based evidence that household masks have some filtration capacity in the relevant droplet size range, as well some efficacy in blocking droplets and particles from the wearer (26). That is, these masks help people keep their droplets to themselves.
4. Mask Efficacy Studies
Although no randomized controlled trials (RCT) on the use of masks as source control for SARS-CoV-2 has been published, a number of studies have attempted to indirectly estimate the efficacy of masks. Overall, an evidence review (29) finds "moderate certainty evidence shows that the use of handwashing plus masks probably reduces the spread of respiratory viruses."
The most relevant paper (30), with important implications for public mask wearing during the COVID-19 outbreak, is one that compares the efficacy of surgical masks for source control for seasonal coronavirus, influenza, and rhinovirus. With ten participants, the masks were effective at blocking coronavirus droplets of all sizes for every subject. However, masks were far less effective at blocking rhinovirus droplets of any size, or of blocking small influenza droplets. The results suggest that masks may have a significant role in source control for the current coronavirus outbreak. The study did not use COVID-19 patients, and it is not yet known whether seasonal coronavirus behaves the same as SARS-CoV-2; however, they are of the same genus, so similar behavior is likely.
Another relevant (but under-powered, with n=4) study (31) found that a cotton mask blocked 96% (reported as 1.5 log units or about a 36-fold decrease) of viral load on average, at eight inches away from a cough from a patient infected with COVID-19. If this is replicated in larger studies it would be an important result, because it has been shown (32) that "every 10-fold increase in viral load results in 26% more patient deaths" from "acute infections caused by highly pathogenic viruses".
A comparison of homemade and surgical masks for bacterial and viral aerosols (21) observed that "the median-fit factor of the homemade masks was one-half that of the surgical masks. Both masks significantly reduced the number of microorganisms expelled by volunteers, although the surgical mask was 3 times more effective in blocking transmission than the homemade mask." Research focused on aerosol exposure has found all types of masks are at least somewhat effective at protecting the wearer. Van der Sande et al (33) found that "all types of masks reduced aerosol exposure, relatively stable over time, unaffected by duration of wear or type of activity", and concluded that "any type of general mask use is likely to decrease viral exposure and infection risk on a population level, despite imperfect fit and imperfect adherence". Overall however, analysis of particle filtration is likely to underestimate the effectiveness of masks, since the fraction of particles that are emitted as aerosol (vs. droplet) is quite small (26). Analysis of seasonal coronavirus compared to rhinovirus (30) suggests that filtration of COVID-19 may be much more effective, especially for source control.
The importance of using masks for health care workers has been observed (34) in three Chinese hospitals where, in each hospital, medical staff wearing masks (mainly in quarantine areas) had no COVID-19 infections, despite being around COVID-19 patients far more often, whilst other medical staff had 10 or more infections in each of the three hospitals.
Masks seem to be effective for source control in the controlled setting of an airplane. One case report (35) describes a man who flew from China to Toronto and then tested positive for COVID-19. He was wearing a mask during the flight. The 25 people closest to him on plane/flight attendants were tested and all were negative. Nobody has been reported from that flight as getting COVID-19. Another case study involving a masked influenza patient on an airplane (36) found that "wearing a face mask was associated with a decreased risk for influenza acquisition during this long-duration flight".
Guideline development for health worker personal protective equipment have focused on whether surgical masks or N95 respirators should be recommended. Most of the research in this area focuses on influenza. At this point, it is not known to what extent findings from influenza studies apply to COVID-19 filtration. Wilkes et al (37) found that "filtration performance of pleated hydrophobic membrane filters was demonstrated to be markedly greater than that of electrostatic filters." However, even substantial differences in materials and construction do not seem to impact the transmission of droplet-borne viruses in practice, such as a metaanalysis of N95 respirators compared to surgical masks (38) that found "the use of N95 respirators compared with surgical masks is not associated with a lower risk of laboratoryconfirmed influenza." Johnson et al (39) showed that "surgical and N95 masks were equally effective in preventing the spread of PCR-detectable influenza". Radonovich et al (40) found in an outpatient setting that "use of N95 respirators, compared with medical masks... resulted in no significant difference in the rates of laboratory-confirmed influenza."
One of the most frequently mentioned papers evaluating the benefits and harms of cloth masks have been by MacIntyre et al (41). Findings have been misinterpreted, and therefore justify detailed discussion here. The authors "caution against the use of cloth masks" for healthcare professionals compared to the use of surgical masks and regular procedures, based on an analysis of transmission in hospitals in Hanoi. We emphasize the setting of the study - health workers using masks to protect themselves against infection. The study compared a "surgical mask" group which received 2 new masks per day, to a "cloth mask" group that received 5 masks for the entire 4week period and were required to wear the masks all day, to a "control group" which used masks in compliance with existing hospital protocols, which the authors describe as a "very high level of mask use". It is important to note that the authors did not have a "no mask" control group because it was deemed "unethical to ask participants to not wear a mask." The study does not inform policy pertaining to public mask wearing as compared to the absence of masks in a community setting, since there is not a "no mask" group. The results of the study show that the group with a regular supply of new surgical masks each day had significantly lower infection of rhinovirus than the group that wore a limited supply of cloth masks. This paper lends support to the use of clean, surgical masks by medical staff in hospital settings to avoid rhinovirus infection by the wearer, and is consistent with other studies that show cloth masks provide poor filtration for rhinovirus (30). Its implementation does not inform the effect of using cloth masks versus not using masks in a community setting for source control of SARS-CoV-2, which is of the same genus as seasonal coronavirus, which has been found to be effectively filtered by cloth masks in a source control setting (30).
A. Studies of Impact on Community Transmission.
When evaluating the available evidence for the impact of masks on community transmission, it is critical to clarify the setting of the research study (health care facility or community), the respiratory illness being evaluated and what reference standard was used (no mask or surgical mask). There are no RCTs that have been done to evaluate the impact of masks on community transmission during a coronavirus pandemic. While there is some evidence from influenza outbreaks, the current global pandemic poses a unique challenge. A review (42) of 67 studies including randomized controlled trials and observational studies found that simple and lowcost interventions would be useful for reducing transmission of epidemic respiratory viruses. The review recommended that "the following effective interventions should be implemented, preferably in a combined fashion, to reduce transmission of viral respiratory disease: 1. frequent handwashing with or without adjunct antiseptics; 2. barrier measures such as gloves, gowns, and masks with filtration apparatus; and 3. suspicion diagnosis with the isolation of likely cases". However, it cautioned that routine longterm implementation of some measures assessed might be difficult without the threat of an epidemic."
http://files.fast.ai/papers/masks_lit_review.pdf
Jeremy Howarda,c,1 , Austin Huangb , Zhiyuan Lik , Zeynep Tufekcim, Zdimal Vladimire , Helene-Mari van der Westhuizenf,g , Arne von Delfto,g , Amy Pricen , Lex Fridmand , Lei-Han Tangi,j , Viola Tangl , Gregory L. Watsonh , Christina E. Baxs , Reshama Shaikhq , Frederik Questierr , Danny Hernandezp , Larry F. Chun , Christina M. Ramirezh , and Anne W. Rimoint
a fast.ai, 101 Howard St, San Francisco, CA 94105, US; bWarren Alpert School of Medicine, Brown University, 222 Richmond St, Providence, RI 02903; cData Institute, University of San Francisco, 101 Howard St, San Francisco, CA 94105, US; dDepartment of Electrical Engineering & Computer Science, Massachusetts Institute of Technology, 77 Massachusetts Ave, Cambridge, MA 02139; e Institute of Chemical Process Fundamentals, Czech Academy of Sciences, Rozvojová 135, CZ-165 02 Praha 6, Czech Republic; fDepartment of Primary Health Care Sciences, Woodstock Road, University of Oxford, OX2 6GG, United Kingdom; gTB Proof, Cape Town, South Africa; hDepartment of Biostatistics, UCLA Fielding School of Public Health, 650 Charles E Young Drive, Los Angeles, CA 90095; iDepartment of Physics, Hong Kong Baptist University, Kowloon Tong, Hong Kong SAR, China; jComplex Systems Division, Beijing Computational Science Research Center, Haidian, Beijing 100193, China; kCenter for Quantitative Biology, Peking University, Haidian, Beijing 100871, China; lDepartment of Information Systems, Business Statistics and Operations Management, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong SAR, China; mUniversity of North Carolina at Chapel Hill; nSchool of Medicine Anesthesia Informatics and Media (AIM) Lab, Stanford University, 300 Pasteur Drive, Grant S268C, Stanford, CA 94305; oSchool of Public Health and Family Medicine, University of Cape Town, Anzio Road, Observatory, 7925, South Africa; pOpenAI, 3180 18th St, San Francisco, CA 94110; qData Umbrella, 345 West 145th St, New York, NY 10031; rVrije Universiteit Brussel, Pleinlaan 2, 1050 Brussels, Belgium; sUniversity of Pennsylvania, 3400 Civic Center Blvd, Philadelphia, PA 19104; tDepartment of Epidemiology, UCLA Fielding School of Public Health, 650 Charles E Young Drive, Los Angeles, CA 90095
This manuscript was compiled on April 10, 2020...
The science around the use of masks by the general public to impede COVID-19 transmission is advancing rapidly...
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...
"3. Filtering Capability of Masks
...Multiple studies show the filtration effects of cloth masks relative to surgical masks. Particle sizes for speech are on the order of 1 µm (20) while typical definitions of droplet size are 5 µm-10 µm (5). Generally available household materials had between a 49% and 86% filtration rate for 0.02 µm exhaled particles whereas surgical masks filtered 89% of those particles (21). In a laboratory setting, household materials had 3% to 60% filtration rate for particles in the relevant size range, finding them comparable to some surgical masks (22). In another laboratory setup, a tea cloth mask was found to filter 60% ofparticles between 0.02 µm to 1 µm, where surgical masks filtered 75% (23). Dato et al (2006) (24), note that "quality commercial masks are not always accessible." They designed and tested a mask made from heavyweight T-shirts, finding that it "offered substantial protection from the challenge aerosol and showed good fit with minimal leakage".Although cloth and surgical masks are primarily targeted towards droplet particles, some evidence suggests they may have a partial effect in reducing viral aerosol shedding (25).
When considering the relevance of these studies of ingress, it’s important to note that they are likely to substantially underestimate effectiveness of masks for source control. When someone is breathing, speaking, or coughing, only a tiny amount of what is coming out of their mouths is already in aerosol form. Nearly all of what is being emitted is droplets. Many of these droplets will then evaporate and turn into aerosolized particles that are 3 to 5-fold smaller. The point of wearing a mask as source control is largely to stop this process from occurring, since big droplets dehydrate to smaller aerosol particles that can float for longer in air (26).
Anfinrud et al (6) used laser light-scattering to sensitively detect droplet emission while speaking. Their analysis showed that virtually no droplets were "expelled" with a homemade mask consisting of a washcloth attached with two rubber bands around the head, while significant levels were expelled without a mask. The authors stated that "wearing any kind of cloth mouth cover in public by every person, as well as strict adherence to distancing and handwashing, could significantly decrease the transmission rate and thereby contain the pandemic until a vaccine becomes available."
An important focus of analysis for public mask wearing is droplet source control. This refers to the effectiveness of blocking droplets from an infectious person, particularly during speech, when droplets are expelled at a lower pressure and are not small enough to squeeze through the weave of a cotton mask. Many recommended cloth mask designs also include a layer of paper towel or coffee filter, which could increase filter effectiveness for PPE, but does not appear to be necessary for blocking droplet emission (6, 27, 28).
In summary, there is laboratory-based evidence that household masks have some filtration capacity in the relevant droplet size range, as well some efficacy in blocking droplets and particles from the wearer (26). That is, these masks help people keep their droplets to themselves.
4. Mask Efficacy Studies
Although no randomized controlled trials (RCT) on the use of masks as source control for SARS-CoV-2 has been published, a number of studies have attempted to indirectly estimate the efficacy of masks. Overall, an evidence review (29) finds "moderate certainty evidence shows that the use of handwashing plus masks probably reduces the spread of respiratory viruses."
The most relevant paper (30), with important implications for public mask wearing during the COVID-19 outbreak, is one that compares the efficacy of surgical masks for source control for seasonal coronavirus, influenza, and rhinovirus. With ten participants, the masks were effective at blocking coronavirus droplets of all sizes for every subject. However, masks were far less effective at blocking rhinovirus droplets of any size, or of blocking small influenza droplets. The results suggest that masks may have a significant role in source control for the current coronavirus outbreak. The study did not use COVID-19 patients, and it is not yet known whether seasonal coronavirus behaves the same as SARS-CoV-2; however, they are of the same genus, so similar behavior is likely.
Another relevant (but under-powered, with n=4) study (31) found that a cotton mask blocked 96% (reported as 1.5 log units or about a 36-fold decrease) of viral load on average, at eight inches away from a cough from a patient infected with COVID-19. If this is replicated in larger studies it would be an important result, because it has been shown (32) that "every 10-fold increase in viral load results in 26% more patient deaths" from "acute infections caused by highly pathogenic viruses".
A comparison of homemade and surgical masks for bacterial and viral aerosols (21) observed that "the median-fit factor of the homemade masks was one-half that of the surgical masks. Both masks significantly reduced the number of microorganisms expelled by volunteers, although the surgical mask was 3 times more effective in blocking transmission than the homemade mask." Research focused on aerosol exposure has found all types of masks are at least somewhat effective at protecting the wearer. Van der Sande et al (33) found that "all types of masks reduced aerosol exposure, relatively stable over time, unaffected by duration of wear or type of activity", and concluded that "any type of general mask use is likely to decrease viral exposure and infection risk on a population level, despite imperfect fit and imperfect adherence". Overall however, analysis of particle filtration is likely to underestimate the effectiveness of masks, since the fraction of particles that are emitted as aerosol (vs. droplet) is quite small (26). Analysis of seasonal coronavirus compared to rhinovirus (30) suggests that filtration of COVID-19 may be much more effective, especially for source control.
The importance of using masks for health care workers has been observed (34) in three Chinese hospitals where, in each hospital, medical staff wearing masks (mainly in quarantine areas) had no COVID-19 infections, despite being around COVID-19 patients far more often, whilst other medical staff had 10 or more infections in each of the three hospitals.
Masks seem to be effective for source control in the controlled setting of an airplane. One case report (35) describes a man who flew from China to Toronto and then tested positive for COVID-19. He was wearing a mask during the flight. The 25 people closest to him on plane/flight attendants were tested and all were negative. Nobody has been reported from that flight as getting COVID-19. Another case study involving a masked influenza patient on an airplane (36) found that "wearing a face mask was associated with a decreased risk for influenza acquisition during this long-duration flight".
Guideline development for health worker personal protective equipment have focused on whether surgical masks or N95 respirators should be recommended. Most of the research in this area focuses on influenza. At this point, it is not known to what extent findings from influenza studies apply to COVID-19 filtration. Wilkes et al (37) found that "filtration performance of pleated hydrophobic membrane filters was demonstrated to be markedly greater than that of electrostatic filters." However, even substantial differences in materials and construction do not seem to impact the transmission of droplet-borne viruses in practice, such as a metaanalysis of N95 respirators compared to surgical masks (38) that found "the use of N95 respirators compared with surgical masks is not associated with a lower risk of laboratoryconfirmed influenza." Johnson et al (39) showed that "surgical and N95 masks were equally effective in preventing the spread of PCR-detectable influenza". Radonovich et al (40) found in an outpatient setting that "use of N95 respirators, compared with medical masks... resulted in no significant difference in the rates of laboratory-confirmed influenza."
One of the most frequently mentioned papers evaluating the benefits and harms of cloth masks have been by MacIntyre et al (41). Findings have been misinterpreted, and therefore justify detailed discussion here. The authors "caution against the use of cloth masks" for healthcare professionals compared to the use of surgical masks and regular procedures, based on an analysis of transmission in hospitals in Hanoi. We emphasize the setting of the study - health workers using masks to protect themselves against infection. The study compared a "surgical mask" group which received 2 new masks per day, to a "cloth mask" group that received 5 masks for the entire 4week period and were required to wear the masks all day, to a "control group" which used masks in compliance with existing hospital protocols, which the authors describe as a "very high level of mask use". It is important to note that the authors did not have a "no mask" control group because it was deemed "unethical to ask participants to not wear a mask." The study does not inform policy pertaining to public mask wearing as compared to the absence of masks in a community setting, since there is not a "no mask" group. The results of the study show that the group with a regular supply of new surgical masks each day had significantly lower infection of rhinovirus than the group that wore a limited supply of cloth masks. This paper lends support to the use of clean, surgical masks by medical staff in hospital settings to avoid rhinovirus infection by the wearer, and is consistent with other studies that show cloth masks provide poor filtration for rhinovirus (30). Its implementation does not inform the effect of using cloth masks versus not using masks in a community setting for source control of SARS-CoV-2, which is of the same genus as seasonal coronavirus, which has been found to be effectively filtered by cloth masks in a source control setting (30).
A. Studies of Impact on Community Transmission.
When evaluating the available evidence for the impact of masks on community transmission, it is critical to clarify the setting of the research study (health care facility or community), the respiratory illness being evaluated and what reference standard was used (no mask or surgical mask). There are no RCTs that have been done to evaluate the impact of masks on community transmission during a coronavirus pandemic. While there is some evidence from influenza outbreaks, the current global pandemic poses a unique challenge. A review (42) of 67 studies including randomized controlled trials and observational studies found that simple and lowcost interventions would be useful for reducing transmission of epidemic respiratory viruses. The review recommended that "the following effective interventions should be implemented, preferably in a combined fashion, to reduce transmission of viral respiratory disease: 1. frequent handwashing with or without adjunct antiseptics; 2. barrier measures such as gloves, gowns, and masks with filtration apparatus; and 3. suspicion diagnosis with the isolation of likely cases". However, it cautioned that routine longterm implementation of some measures assessed might be difficult without the threat of an epidemic."
http://files.fast.ai/papers/masks_lit_review.pdf
