Pandemic influenza A (H1N1) virus emerged in Mexico in early 2009 and rapidly spread worldwide. Severity of illness now appears to be more moderate than initially feared [Reference Fraser1, Reference Wilson and Baker2], although high population attack rates would be associated with significant numbers of severe infections, hospitalizations and deaths. While some governments, particularly in the developed world, have large antiviral stockpiles on hand and contracts for vaccines that are now in production, the primary interventions currently available in both developed and less-developed settings are non-pharmaceutical [3, 4]. At the population level, these can include border controls to delay cross-border transmission, and social distancing measures such as school or workplace closures. At the individual level, interventions to reduce transmission include improved hygiene and the use of face masks, respirators, and other physical barriers [Reference Jefferson5]. We conducted a systematic review [Reference Moher6] to investigate the evidence supporting the effectiveness of face masks in reducing influenza virus infection under controlled and natural conditions.
On 18 August 2009 we searched the following databases for articles published in English from January 1960 to August 2009: PubMed (1960–2009), Science Citation Index (Web of Science) (1970–2009), and the Cochrane Library (1988–2009). We searched for articles using the following search strategy:
#1: ‘facemask’ or ‘facemasks’ or ‘mask’ or ‘masks’ or ‘respirator’ or ‘respirators’ or ‘N100’ or ‘N99’ or ‘N95’ or ‘P2’ or ‘FFP2’
#2: ‘influenza’ or ‘flu’ or ‘respiratory virus’ or ‘respiratory infection’ or ‘respiratory tract infection’
#3: #1 and #2.
The search results were surveyed for methodological articles. Review articles were excluded, but the reference lists in all retrieved review papers were searched for additional related articles. In addition, a manual search was performed with the corresponding authors' reference database.
Two authors (B.J.C. and Y.Z.) independently evaluated the titles and abstracts of all studies for potential inclusion in this review. The same authors then reviewed full-length versions of selected articles to determine inclusion. When consensus was not reached, discussion and further study evaluation with other authors was used to resolve data extraction discrepancies. Articles were included in the review if they (1) described controlled volunteer studies of influenza virus filtration of face masks or respirators, (2) described observational or intervention studies of face masks or respirators to prevent influenza or influenza-like illness (ILI) in healthcare settings, (3) described observational or intervention studies of face masks or respirators to prevent influenza or ILI in community settings. Studies focused on specific non-influenza respiratory infections, such as SARS, were excluded. The initial search resulted in 279 citations. Fifty-six articles were accepted at the abstract stage and finally 12 articles were considered relevant for inclusion in this review (Fig. 1).
Experimental volunteer studies
We identified one study that examined the efficacy of face masks in filtering influenza virus in volunteer subjects. Johnson and colleagues tested the performance of surgical and N95 masks to filter virus in nine volunteers with confirmed influenza A or B virus infection [Reference Johnson7]. Participants coughed five times onto a Petri dish containing viral transport medium held 20 cm in front of their mouth. The experiment was repeated with subjects wearing a surgical mask, and wearing an N95 respirator. While influenza virus could be detected by RT–PCR in all nine volunteers without a mask, no influenza virus could be detected on the Petri dish specimens when participants wore either type of face mask. A limitation was that the study did not consider the role of leakage around the sides of the mask.
Studies in healthcare settings
We identified six studies of face mask use in healthcare settings (Table 1) [Reference Loeb8–Reference Hobday and Cason13]. Because the study designs, participants, interventions and reported outcome measures varied markedly, we focused on describing the studies, their results, their applicability and their limitations and on qualitative synthesis rather than meta-analysis.
RCT, Randomized controlled trial.
A randomized controlled trial in Canada found no significant differences in protection against laboratory-confirmed influenza infection associated with the use of surgical masks or N95 masks among nurses [absolute risk difference −0·73%, 95% confidence interval (CI) −8·8 to 7·3] with 24% of nurses in the surgical mask arm having laboratory-confirmed infection during an influenza season [Reference Loeb8].
A randomized controlled trial in Japan allocated 32 healthcare personnel to wearing surgical face masks or not, but was underpowered to detect significant differences between arms with one observed acute respiratory illness in each arm of the study during the follow-up period [Reference Jacobs9].
A survey of 133 nurses in Hong Kong found that suboptimal adherence to wearing a face shield during high-risk procedures [adjusted odds ratio (OR) 3·56, 95% CI 1·18–10·69] was associated with higher risk of ILI, while suboptimal adherence to use of gloves and gowns were also associated with higher adjusted risk of ILI although not statistically significant [Reference Ng10]. Two other cross-sectional studies found no evidence for a protective effect of face masks against infection [Reference Al-Asmary11, Reference Davies12]. Finally, Hobday & Cason [Reference Hobday and Cason13] speculated that natural ventilation, hand hygiene and gauze face masks were associated with fewer observed deaths in open-air hospitals in Boston during the 1918–1919 influenza A (H1N1) ‘Spanish flu’ pandemic, although there were many potential confounders.
Studies in community settings
We identified four randomized controlled trials that examined the effectiveness of face masks to prevent respiratory virus transmission in community settings [Reference Cowling14–Reference MacIntyre16] (Table 2). In a household-based study in Hong Kong, index cases and household members were randomized to three arms, including control, hand hygiene and hand hygiene plus surgical masks (to be worn by the index case and household members) [Reference Cowling14]. In the primary intention-to-treat analysis there was no statistically significant difference in laboratory-confirmed influenza in household contacts across intervention groups. However when a pre-specified analysis restricted attention to 154 households in which the intervention was applied within 36 hours of symptom onset in the index case, statistically significant reductions in laboratory-confirmed influenza virus infections in household contacts were observed in the face mask and hand hygiene arm (adjusted OR 0·33, 95% CI 0·13–0·87). Adherence to the face mask intervention in index cases was moderate, but poorer in household contacts. The pilot study with a similar design was underpowered to identify significant differences between study arms [Reference Cowling15].
Another recent study randomized 145 symptomatic index cases aged 0–15 years from outpatient clinics and their household members to three arms: control, surgical masks (worn by household contacts only), or N95-type respirators (worn by household contacts only) without fit-testing [Reference MacIntyre16]. There were no differences in ILI in household contacts across intervention arms. A secondary per-protocol analysis found that adherent use of N95 or surgical masks significantly reduced the risk for ILI in household contacts (hazard ratio 0·26, 95% CI 0·09–0·77) compared to non-adherent mask use or allocation to the control arm.
Aiello and colleagues described a study in which 1437 university students were randomized by dormitory to three arms: control, surgical masks alone, and surgical masks plus hand hygiene [Reference Aiello17]. Students were followed for 6 weeks during the influenza season and assessed for clinically diagnosed or survey-reported ILI. Compared with the control group, significant reductions in ILI were observed during weeks 4–6 in the mask and hand hygiene group ranging from 35% (95% CI 9–53) to 51% (95% CI 13–73), after adjusting for vaccination and other covariates; similar reductions, although not statistically significant, were observed in the mask-only group compared to the control group. Neither mask use and hand hygiene nor mask use alone was associated with significant reduction in ILI rate cumulatively; continued subject recruitment (larger sample size) after study start, increased participation in the intervention later in the study, a late, mild influenza season, and/or interruption of the intervention for 1 week by spring break may explain this finding. The study was underpowered to determine the relative contribution of the protective effects of masks compared to hand hygiene.
Finally, Lo and colleagues [Reference Lo18] investigated respiratory virus isolations in specimens collected primarily from in-patients and compared virus isolations in Hong Kong in 2003 with the preceding years. Declines in the number and proportions of virus isolations were attributed to population increases in hygienic measures and widespread use of face masks, as well as social distancing during the SARS epidemic. However, the study could not distinguish the relative contributions of each intervention.
Our review highlights the limited evidence base supporting the efficacy or effectiveness of face masks to reduce influenza virus transmission. An important concern when determining which public health interventions could be useful in mitigating local influenza virus epidemics, and which infection control procedures are necessary to prevent nosocomial transmission, is the mode of influenza virus transmission between people and in the environment. Physical barriers would be most effective in limiting short-distance transmission by direct or indirect contact and large droplet spread, while more comprehensive precautions would be required to prevent infection at longer distances via airborne spread of small (nuclei) droplet particles [Reference Siegel19]. In healthcare settings, stringent precautions are recommended to protect against pathogens that are transmitted by the airborne route, including the use of N95-type respirators (which require fit testing), other personal protective equipment including gowns, gloves, head covers and face shields, and isolation of patients in negative-pressure rooms [Reference Siegel19]. There remains considerable controversy over the relative importance of the alternative modes of transmission for influenza virus. In a recent review, Brankston and colleagues concluded that natural influenza transmission in human beings occurs generally over short distance rather than over long distance [Reference Brankston20]. Based on the same evidence, Tellier had earlier concluded that aerosol transmission occurs at appreciable rates [Reference Tellier21], and cited further evidence in an updated review [Reference Tellier22]. Weber & Stilianakis [Reference Weber and Stilianakis23] found that contact, large droplet and small droplet (aerosol) transmission are all potentially important modes of transmission for influenza virus.
If airborne transmission were important, it would be less likely that surgical masks will lead to reductions in infectiousness or protection against infection, if worn by ill or uninfected people, respectively. The primary argument against airborne transmission is as much one of absence of evidence as evidence of absence. While there are documented examples of long-distance airborne transmission of other pathogens including varicella zoster virus and Mycobacterium tuberculosis, the literature contain few compelling examples of airborne transmission of influenza virus [Reference Brankston20], and several reports of scenarios where airborne transmission did not occur [Reference Blumenfeld24–Reference Han27]. Further indirect evidence such as the substantial benefit of hand hygiene to prevent influenza transmission [Reference Cowling14] is suggestive of direct or indirect contact as one of the most important modes of transmission for influenza virus in some settings. Further observational or intervention studies conducted in different latitudes during different times of the year could help to elucidate the role of temperature and humidity in mediating modes of transmission [Reference Lowen28].
We did not identify any experimental volunteer studies that investigated whether surgical masks or N95 respirators could protect against infection. We identified one experimental study of face mask performance which involved participants with confirmed influenza virus infection [Reference Johnson7], and the results suggested that surgical masks may be able to reduce infectiousness. In future similar studies it would be important to consider the potential for leakage around the sides of the mask in addition to direct penetration of infectious viral particles through the mask, if the results are to have practical implications for reduction of transmission in community and other settings [Reference Weiss29]. Further studies are needed to investigate how mask and respirator performance varies with temperature and humidity, or under working conditions when moisture in exhaled breath or sweat may build up in face masks and hinder filtration or fit [Reference Rengasamy, Zhuang and Berryann30].
Few studies have been conducted in healthcare settings, and there is limited evidence to support the effectiveness of either surgical masks or N95 respirators to protect healthcare personnel [Reference Loeb8–Reference Hobday and Cason13]. One recent large trial in nurses found no difference in effectiveness between surgical masks and N95 respirators, although the confidence intervals were wide enough to include moderate effect sizes [Reference Loeb8]. Further, larger studies are needed to confirm the non-inferiority of surgical masks. Guidance provided by the World Health Organization for protection of healthcare workers against pandemic influenza A (H1N1) virus infection recommends the use of standard and droplet precautions (including surgical masks or a face shield) during most patient interactions, while N95 or equivalent respirators are recommended for aerosol-generating procedures . One concern over the use of face masks or respirators in healthcare settings is the potential for negative psychosocial impacts on patients and children in particular, especially in regions outside Asia where masks are not routinely worn [Reference Beck32]. Long-term use of N95-type respirators is likely to lead to physical discomfort [Reference Li33], and has been associated with headaches [Reference Lim34]. Considerable resources might be required to make available N95 respirators and other protective equipment to large numbers of healthcare personnel through the course of influenza epidemics or pandemics. Finally, there are likely to be difficulties in ensuring compliance in healthcare workers [Reference Seale35]. Nevertheless personal protective equipment has led to major improvements in general infection control procedures in the hospital setting [Reference Lu36–Reference Rutala and Weber38] and should not be discounted due to the lack of available data examining influenza virus outcomes.
Three controlled studies of face mask effectiveness in the community setting used case-ascertained designs, where ill index cases were recruited from outpatient clinics and households were followed up for 7–10 days to observe secondary transmission [Reference Cowling14–Reference MacIntyre16]. The Hong Kong study applied surgical face masks to index cases and their household contacts [Reference Cowling14, Reference Cowling15], while the Australian study applied surgical masks or N95-type respirators to household contacts only [Reference MacIntyre16]. Neither study provides conclusive evidence that face masks are effective in primary intention-to-treat analyses, although statistical power was limited. Adherence was moderate in both studies, and a per-protocol analysis of the Australian study suggests that masks could be effective in reducing risk of infection [Reference MacIntyre16]. In the Hong Kong study, index cases not allocated to the face mask intervention reported use of face masks, indicating some degree of contamination of the intervention, while adherence was lower in household contacts and the results may primarily support the use of masks in ill members to reduce infectiousness [Reference Cowling14, Reference Cowling15].
The effectiveness of face masks is probably impacted by compliance issues in both the healthcare and community setting [Reference Cowling14, Reference Cowling15, Reference Seale35]. Various studies show a lower level of compliance with face masks [Reference Cowling14, Reference Cowling15] or find lower reported acceptability of face masks [Reference Stebbins, Downs and Vukotich39] compared to hand hygiene behaviours and other non-pharmaceutical interventions. However, these studies do not seek to explain the reduced compliance, nor do they measure levels of compliance in the midst of an outbreak of pandemic influenza. Future research endeavours should investigate the influence of cultural and sociobehavioural factors (e.g. fear, stigma, altruism) on levels of compliance during a pandemic. Use of face masks in the community was very common during the SARS epidemic in Hong Kong, but not in Singapore [Reference Leung40], and cultural differences could also affect compliance.
Pandemic guidance provided by the World Health Organization for community settings advises that masks may be worn although effectiveness is uncertain particularly in open spaces . Other health agencies, such as the US Centers for Disease Control and Prevention, are not recommending masks in the community setting, with the exception of high-risk individuals who care for the sick or spend time in large crowds in areas affected by the pandemic . Wearing masks incorrectly may increase the risk of transmission . Further studies of face mask use are now underway, including some with prospective designs that follow cohorts of initially uninfected people. These studies will be particularly important in addressing compliance to and effectiveness associated with sustained use of face masks beyond the acute scenarios of existing studies [Reference Cowling14–Reference MacIntyre16]. While fewer resources are required to conduct studies with outcomes based on self-reported signs and symptoms of acute respiratory infection, future studies could include acute and convalescent serology or repeated collection of clinical specimens to provide results specific to influenza virus infection.
In conclusion there remains a substantial gap in the scientific literature on the effectiveness of face masks to reduce transmission of influenza virus infection. While there is some experimental evidence that masks should be able to reduce infectiousness under controlled conditions [Reference Johnson7], there is less evidence on whether this translates to effectiveness in natural settings. There is little evidence to support the effectiveness of face masks to reduce the risk of infection. Current research has several limitations including underpowered samples, limited generalizability, narrow intervention targeting and inconsistent testing protocols, different laboratory methods, and case definitions. Further in-vivo studies of face masks in infectious individuals are warranted to determine the proportion of exhaled virus that is trapped by the mask. More detailed volunteer challenge and volunteer transmission studies could be designed to include both infectious and susceptible participants, to evaluate the efficacy of face masks both in reducing infectiousness and reducing susceptibility. However, such studies would require substantial resources, and contrived experiments may have limited generalizability to the natural setting. Large intervention studies in healthcare and community settings are likely to provide the best evidence of the effectiveness of face masks in reducing transmission in pandemic and inter-pandemic periods and are an urgent priority to guide pandemic preparedness for second and subsequent waves of pandemic influenza A (H1N1) and future pandemics.
We thank Lincoln Lau for technical assistance. This work was supported by the Area of Excellence Scheme of the Hong Kong University Grants Committee (grant no. AoE/M-12/06). The funding body was not involved in the collection, analysis and interpretation of data, the writing of the manuscript, or the decision to submit for publication.
DECLARATION OF INTEREST