Masks and respirators for prevention of respiratory infections: a state of the science review

The outcome in mask trials is an infectious disease which can transmit from person to person; risk of transmission is higher in hospitals and other closed settings. In RCTs of interventions against non-communicable diseases (e.g., hypertension), the intervention affects only the person receiving it. In trials involving a contagious infection, the intervention may affect people other than those receiving it by preventing onward transmission. This “herd protection effect” of N95 respirators has been described in health-care settings (138). It means prevention of infection in one individual can prevent a chain of subsequent transmission. To account for this, any clinical trial of respiratory protection in health-care and other closed settings should ideally be cluster randomized without appreciable contact between clusters. We acknowledge that this ideal design may not always be feasible, but findings from less optimal designs should be interpreted accordingly.

Respiratory virus activity (especially influenza) is highly variable by season and from year to year (139). In contrast, SARS-COV-2 is not clearly seasonal. A trial of masking that is conducted when the prevalence of the disease under investigation is extremely low (see, e.g., reference 140) may falsely conclude that masking is ineffective because few people in either arm develop the disease (hence, the study is likely to be underpowered). Ideally, any trial of prevention of influenza should be run over more than 1 year to overcome the risk of low seasonal influenza activity in a single year.

Mask trials may use clinical, laboratory-confirmed, or a combination of outcomes. Using “influenza-like illness” (ILI) (141) as a clinical outcome is problematic, particularly in studies that rely on self-reporting (with or without subsequent laboratory confirmation) because this relatively insensitive definition requires a fever of at least 38^oC in addition to one respiratory symptom such as cough or sore throat, and most adults do not have documented fever in the presence of a confirmed respiratory viral infection (142). Using ILI as the primary outcome measure may reduce the statistical power of a study. Some RCTs have used a definition of “clinical respiratory illness” (CRI), generally defined as two or more respiratory symptoms (cough, nasal congestion, runny nose, sore throat or sneezes) or one respiratory symptom and a systemic symptom (chill, lethargy, loss of appetite, abdominal pain, muscle or joint aches), which appears more sensitive (115).

Another limitation of these RCTs is including only influenza and/or COVID-19 as the primary outcome. Trial power may be reduced if tests are not conducted at the peak of viral shedding or are done incorrectly. Not accounting for influenza vaccination may also result in misclassification of people testing positive on serology as infected. In most mask trials, laboratory confirmation was through PCR, but some studies used serology as well. Influenza and SARS-CoV-2 have a higher risk of morbidity and mortality than other seasonal respiratory viruses measured in some RCTs. However, in reality, a wide range of pathogens cause respiratory illnesses and are transmitted in a similar way regardless of how severe the illness is. Although several RCTs report “other respiratory viruses” as an outcome, separate analysis is typically provided for influenza and, more recently, for COVID-19. A pooled analysis of two large RCTs reported significantly lower rates of laboratory-confirmed viral respiratory infections, even those assumed to be spread predominantly by large droplets, among participants who used a continuous N95 respirator (143).

Meta-analyses commonly and erroneously combine dissimilar interventions, such as a trial of advice to wear a mask when going out of the home combined with a trial of advice to wear a mask inside the home in the presence of a sick family member (144). Some community trials examined masks only in combination with another intervention such as hand hygiene. Such trials do not allow estimation of the efficacy of masks alone, though the rationale for such designs is pragmatic (hand hygiene should accompany mask use to prevent self-contamination) (145). Almost all meta-analyses of N95 use in health-care combine trials on intermittent and continuous use of the device. However, crucially, intermittent use of an N95 respirator by health-care staff when caring for infected or potentially infected patients or conducting an AGMP is a different intervention to wearing a respirator continuously during an entire work shift. In some RCTs, participants used masks and respirators only when they worked within 1-2m of patients with suspected or confirmed respiratory illness or when doing an AGMP, but not at other times (25, 146). This assumes health workers can accurately self-identify situations of exposure risk and fails to consider wider airborne exposure (including to asymptomatic or presymptomatically infected staff or patients) in health-care settings (48). The only trial to compare continuous and intermittent N95 use showed that continuous wearing of an N95 is protective, but intermittent use is not, and that intermittent N95 use and surgical masks were equally inefficient (147). This is consistent with trials that compared intermittent N95 use with surgical masks and found no difference (25, 146). Other examples of combining dissimilar interventions include combining masking to protect the uninfected wearer with masking of sick people (“source control”) to protect their contacts and combining studies of masking with studies of masking plus handwashing (9).

Meta-analysis should combine only studies that were done in similar settings, but widely cited meta-analyses of mask trials combined dissimilar settings such as health-care and community settings (9). Even among community settings, there is heterogeneity. Households with an infected member are the highest-risk community setting, but a wide range of non-household community settings are commonly combined with household studies in meta-analyses (9). Moreover, in health-care settings, masks are used for occupational protection, and it is a responsibility of the employer to protect the health of employees (148). Masks may also provide source control to protect patients from asymptomatically infected health workers, but to our knowledge no study has investigated this. A meta-analysis of aged care outbreaks did show that universal masking of staff was protective and resulted in smaller outbreak attack rates (149).

When combining studies in a meta-analysis, the outcome being combined should be homogeneous. Mask trials may use clinical, laboratory-confirmed, or a combination of outcomes. Widely cited meta-analyses of mask trials combined dissimilar outcomes (9). The use of a definition of ILI (141) as a clinical outcome is problematic because most adults do not have fever (a prerequisite for the ILI definition) in the presence of a confirmed respiratory viral infection (142). Some RCTs have used a definition of CRI, which does not mandate fever and is more sensitive (115). A common mistake in meta-analyses is combining studies with dissimilar laboratory outcomes, such as RCTs whose outcomes were influenza confirmed by PCR plus serology with other RCTs that used only PCR (9). For influenza, serology may be positive following either infection or vaccination, and seropositivity is far more common than PCR positivity in such trials (25, 146). This flaw in the review process gives disproportionate weight to trials that used a less specific outcome (serology) than those that use PCR only. This can substantially distort findings because positive serology is numerically far greater than positive PCR results (9, 146). Furthermore, the interpretation of a single high titer compared to rising titers with paired sera may lead to misclassification. If participants are being vaccinated against the pathogen being tested, analysis should account for this (including the timing of vaccination relative to the positive test) to avoid misclassification bias.

To address our methodological concerns about previous studies (notably, the mask section of the 2023 Cochrane review [9]), we separated dissimilar settings, interventions, and outcome measures for a new meta-analysis of published RCTs. Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines were followed (PRISMA flowchart is available on request from CRM). Meta-analysis of single proportions was performed using R 4.1.3, using the ‘metafor’ and ‘meta’ packages. A random effects model was used throughout. Heterogeneity was assessed using I² and Q statistics. The restricted maximum-likelihood estimator method was used to estimate between-study variance (τ²). Statistical significance was defined as a two-sided P-value of <0.05. To examine the efficacy of masks and respirators in community and health-care settings, we searched PubMed and Embase for RCTs in these settings. We grouped community RCTs into “primary prevention” (masking to protect the wearer) (27, 144, 150–156) and source control (masking to protect others in the community) (157–159). Because health-care settings present different risks and issues, we examined the effectiveness of masks and respirators in community and health-care settings separately. For the meta-analyses reported here, we excluded RCTs of source control and analyzed only primary prevention RCTs. In community RCTs, we separately analyzed RCTs of mask and RCTs of masks plus hand hygiene. We analyzed outcomes of CRI, ILI, and laboratory (PCR)-confirmed influenza and other respiratory viruses (including influenza and COVID-19). Serological data were excluded to remove heterogeneity of outcome. We found six RCTs in health-care settings comparing various types of masks and respirators (24, 25, 115, 146, 147, 160). We analyzed data on targeted (intermittent) and continuous use of respirators separately.

For all trials, we extracted the following data: study design, year, type of intervention, primary and secondary outcomes, total number of participants, participants with each outcome, main results, and limitations as reported by authors. We tabulated study characteristics and performed meta-analyses using the subgroup analysis approach by outcome measures (161). We used the random effects model to estimate the respective pooled risk ratios (RRs) and 95% confidence intervals (CIs). We prepared forest plots to show pooled estimates and corresponding 95% CIs. Reanalysis of RCTs of Masks in Community Settings reports the findings of this reanalysis for community settings, and Reanalysis of RCTs of Masks and Respirators in Health-care Settings reports findings for health-care settings.

Mask wearers commonly experience local irritation of the skin and eyes, pressure effects, and a mechanical acne (“maskne”) resulting from contact dermatitis and disruption of skin microbiota (235). Risk of all these increases with duration of use (235) and appears to be high in health-care workers (230, 235). Headache is also common, particularly among those with a history of headache, and increases with duration of mask use (236). Thermal discomfort may occur, particularly in health-care workers wearing full PPE (237–239). Potential mitigations include local measures (topical treatment, cushioning tapes, and moving mask straps away from the ears) and workplace measures such as regular and timely “air breaks” (235, 240, 241). In the longer term, there is a need for improved designs that meet safety standards while ensuring that masks are comfortable to wear for extended periods (241), a topic that is discussed further in Toward Better Masks.

The effect of mask wearing on respiratory function has been extensively researched. Statistically significant but transient and clinically insignificant changes in gas exchange and pulmonary function have been detected in healthy individuals wearing masks or respirators during strenuous exercise (234, 242, 243) and during manual work (244). Many though not all primary studies found modest reductions in exercise performance, but participants often reported discomfort and subjective increase in breathing resistance. Accurate measurement of the physiological effects of mask wearing is challenging in several respects (245). Wearing spirometry apparatus over a face mask (242) can generate artifactual errors in estimates of breathing pressure and perceived respiratory effort that may exaggerate the impacts of masking on pulmonary function (246, 247). Overall, these findings suggest that individuals may choose to avoid high-intensity exercise while wearing a tightly fitting respirator. A concern raised early in the pandemic about the risk of cardiac dysrhythmias during exercise (248) was not confirmed in subsequent systematic reviews (240).

Although speculative claims have been made for widespread serious harms (especially to cardiorespiratory and metabolic health from masking), consistent findings across what is now a large body of research have shown a reassuring absence of clinically meaningful serious harms (230, 232, 240). Furthermore, no serious safety incidents have been documented despite periods of high community uptake of masking during pandemic waves (31, 249). Risk compensation behaviors (paying less attention to other protective measures when masked) have been proposed as an adverse effect of masking, but studies which looked for such behaviors did not find them, and indeed masking may lead to wearers adopting more protective behaviors (250). An exception is the work of Wadud et al., who found that community mobility increased after implementation of mask mandates (251) and characterized this change as risk compensation. However, as noted by Cooper et al. (252), trade-offs between (effective) mask mandates, on the one hand, and mobility and in-person work hours, on the other hand, mean that rather than being regarded as a negative consequence of population masking, mask effects could be framed as the potentiation of normal activities, be they economic (e.g., work), social (e.g., family visits), or educational (e.g., school attendance) without the worsening in population health status that would occur in their absence.

It is worth noting that the scientific literature contains peer-reviewed articles which selectively cite flawed empirical studies to support the argument that masks are universally harmful. In 2021, for example, a research letter in JAMA Pediatrics claimed that masks increased the carbon dioxide content of inhaled air in children aged 6–17 years (253). Shortly after publication, this letter was retracted by the journal following complaints by scientists, citing “fundamental concerns” with the study methodology and uncertainty regarding the validity of the findings and conclusions (254), but it has attracted over a million views. That evidently flawed study was later republished in a different journal (255). Similarly, a recent systematic review of harmful effects of masking on breathing (256) was quickly retracted by the journal because of concerns with scientific validity (257), but it continues to be circulated on social media.

“Children” range in age from 0 to 17 years. Infants are a special case: they have narrower airways, which may exacerbate the increased work of breathing during mask wearing, and they are less able to remove a mask if experiencing discomfort or obstruction. Masks are therefore not recommended for children under 2 years (258). Empirical studies are occasionally cited selectively in narrative reviews to support a strongly held view that masks are dangerous in older children. In some cases, the primary studies cited did not show the adverse effects claimed or did not include children. A 2021 narrative review examining the impacts of masking on respiratory infections in children aged 7–14 years did not identify any adverse effects, but only two randomized trials on children (124 participants) were included (259). Additional trials on school-aged children published since that review have shown no adverse impact of masking on cognitive performance (n = 133) (260).

While the absence of serious adverse effects in RCTs of masks in children is reassuring up to a point, the studies were small and had methodological limitations. However, despite widespread uptake of mask wearing in school-aged children around the world (e.g., masking of children was widespread in Asia in some settings and situations before the COVID-19 pandemic), there have, to our knowledge, been no formal reports of incidents of harm. An observational study of cloth masks in US primary schools (261) (n = 1,000 students in pre-K, kindergarten, first, or second grades) reported appropriate mask usage for an average of 77% of the school day over a 4-week period. No serious harms were reported, and there were fewer than 15 reports each of stress, ears hurting/headaches, communication difficulties, feeling hot, and difficulty breathing (the last resolved after a short break from the mask).

Several commentators have speculated that wearing a mask would increase hand-to-face contact among children. Empirical evidence has shown the opposite: either no difference or a decrease in hand-to-face contact among children wearing masks (262, 263).

Children’s own opinions on masks are highly relevant and rarely elicited. When they are, the responses tend to be positive. In a survey of 42,767 Canadian adolescents during the 2020/2021 school year, 81.9% supported wearing masks in indoor public spaces; 67.8% supported wearing them in school (23.1% were neutral and 9.1% opposed) (264). Coelho et al. conducted a survey and held focus groups to understand children’s experience of masking (265); acceptability appeared high, but some children reported barriers to communication, learning, and socialization. The authors recommended strategies to address these barriers, including the use of clear face masks. A recent systematic review of children’s views on masks covered 30 primary studies and concluded that acceptability appeared high, but comfort, fit, style (including age-appropriate designs), communication, and environmental issues were of high priority for children; the review also found that parents were sometimes more opposed to masks for their children than the children themselves were (266). These findings challenge statements by people speaking on behalf of children and claiming that masks are unacceptable to them (267, 268).

Table 5 summarizes empirical evidence on particular medical conditions, based mainly on previous reviews, along with suggested mitigations (230, 232, 233, 240). While primary studies were small, few in number and of variable scientific quality, common sense suggests a need for flexibility and precaution. Note that some of the conditions listed in Table 5 substantially increase vulnerability to respiratory infections, underscoring the point that the risks of masking should be balanced against the benefits.

During mask mandates, people who cannot wear a mask may opt to wear a distinctive lanyard in public or use a flashcard to communicate their exemption status. We found no research studies on the use or acceptability of such interventions.

These are recommendations for mask use by individuals for whom the consequences of infection are likely to be significantly worse than for others (including people who are elderly, pregnant, immunocompromised, or living with severe long-term conditions (327)) and those for whom avoiding infection is particularly important (e.g., those about to have surgery, travel, or participate in critical activities such as high-stakes sporting events). The scope of mask use would depend on individual risk assessment and management, but would typically be indoors and either intermittently or continuously, depending on the risk assessment. Requirements are likely to be increased in specific settings and during times of elevated infection risk; those at highest risk may be advised to wear respirator-grade protection. These policies may be developed or adapted by various interest groups and could be specified as requirements for some highly regulated activities such as elite competitive sports.

These are recommendations and requirements for mask use in workplaces (148) or public places such as health-care settings (328, 329) and long-term care facilities (149, 302) where risk of transmission is elevated or there are vulnerable people. Such policies should be developed and interpreted in the context of existing legal standards (e.g., those that cover protection in workplaces) and respiratory protection policies including formal fit-testing of respirators (110). Mask policies in specific settings are more likely to be effective and sustainable if all interested parties work collaboratively to develop them. In health care, for example, interested parties include IPC clinical teams, employers (responsible for occupational health and safety and maintaining workforce capacity, particularly during peaks of infection (148)), independent worker safety experts, staff unions (who advocate for workers’ conditions and safety), and patient advocacy groups (e.g., for the clinically vulnerable). The question of what kind of mask or respirator to provide in health-care settings and whether use should be continuous or intermittent is covered in Reanalysis of RCTs of Masks in Community Settings. Mask policies may also be needed in a range of non-health settings, particularly where staff and visitors are at higher risk of infection, for example, indoor, crowded environments with poor ventilation such as hospitality venues, and public transport (330).

An increase in seasonal respiratory infections such as influenza and RSV above a preset threshold locally or nationally could warrant setting-specific or more widespread mask use, along with other measures aimed at reducing infections and disease burden. Masking and physical distancing introduced to control COVID-19 infection were associated with a reduction in seasonal influenza (331), lending support for the benefits of such policies for preventing transmission of respiratory infections more broadly.

These policies follow the same principles as for seasonal respiratory infections but may be stricter, introduced earlier, and continued for longer, depending on the risk assessment (see next section). If there is early evidence that a pandemic will be particularly severe (e.g., fatality risk above a defined threshold), the pandemic response should attempt to eliminate, rather than merely reduce, the infectious hazard (332). The use of masks and respirators is particularly important in the early stages of a pandemic, when drugs and vaccines are unavailable and to protect health workers. A proactive elimination response, along with border controls, could exclude the emerging disease from a country for a prolonged period, as was achieved by a few countries in the first 1–2 years of the COVID-19 pandemic (333, 334). In such circumstances, mask requirements could be relaxed within a country while elimination is sustained but would still be required for border, health, and quarantine facility workers having contact with infected people entering the country and for local outbreaks. An effective pandemic mask policy can build on individual, setting-specific, and seasonal policies, all of which help to establish mask use as a familiar form of infection control. However, pandemic management needs additional preparedness, planning, and development during interpandemic periods (335). As this paper went to press, the U.S. Centers for Disease Control and Prevention issued interim guidance recommending respirators for worker protection against the emerging threat of Novel Influenza A Virus (“bird flu”), which is spreading among cattle and has infected humans (336).

The environmental hazards posed by discarded masks need to be made clear, and information and resources need to be provided for more environmentally friendly disposal. Without public understanding and support, little change is likely to occur.

Many environmental reviews reproduce a widespread but probably flawed assumption that because masks may be contaminated with infectious organisms, all mask waste should be treated as a significant biohazard with substantial waste worker protection and special waste incineration (360, 368, 371, 382). While it is hypothetically possible that discarded single-use masks and respirators may act as fomites and contribute to virus spread (see What Are Masks and How Do They Work?), we believe that the biohazard of masks has been substantially overestimated in the past because of an assumed droplet mode of transmission. Scaling back the special biohazard measures in mask waste disposal would reduce their environmental impact. In addition, as Wang et al. recommend, dedicated mask disposal bins and systematic decontamination will maximize opportunities for recycling (360).

A reliable and scalable approach to recycling (or “upcycling”) such products has proved challenging for scientists. Difficulties include risk of contamination (e.g., with blood or non-mask waste), the multiple materials they contain (polymeric filters, aluminium nosepiece, and elastic loops), and adverse cost-benefit balance. However, studies have begun to demonstrate that shredded mask waste can be incorporated into roads and pavements, building materials (e.g., concrete), membranes and filtration materials, various kinds of fuel (through a high-temperature thermochemical process called pyrolysis), battery electrodes, adsorbents, and various specialist chemical products, in each case improving performance of the end material (360, 363, 364, 366, 369, 370). While few of these solutions are ready to be implemented at scale, they offer some hope for the future.

Reuse and extended use are partial solutions that could potentially be implemented promptly and in a highly cost-effective way. Surgical masks are not designed for reuse or even prolonged use, with a recommended maximum duration of 6 hours (383), but respirators are designed for prolonged use; a single respirator can be worn repeatedly and for long periods and still be effective (384, 385). A UK analysis showed that reusing respirators once reduced waste by 65%–80% when compared to single-use respirators but still generated more waste than single-use surgical masks, though this narrowed and then reversed with the number of reuses (386). Another study found similar results (387). Further research is needed, however, on when and how to decontaminate respirators prior to reuse. Although several decontamination methods like ultraviolet irradiation and hydrogen peroxide vapor have been explored (383), these can compromise fit and filtration performance, leading some manufacturers to advise against reuse (87, 388).

Elastomeric respirators provide a high level of protection as well as comfort and fit, are reusable, and have a lightweight solid frame and replaceable filters. They are a promising solution for both community and health-care facility use (389, 390). While their initial cost ($30–$150) is high compared to a single-use device, their long-term financial and environmental costs are lower. Depending on the model, they can be reused for 5 years or more, resulting in up to 96% less waste over single-use or decontaminated and reused options (386, 387).

Biodegradable polymers range from natural types like cellulose, chitin, and silk fibroin to semi-synthetic and synthetic varieties such as polylactic acid (PLA) and polyvinyl alcohol (PVA) (391). PLA and PVA, in particular, have shown significant potential as filter materials, with studies demonstrating their effectiveness in creating high-performance, environmentally friendly masks (391, 392). Wang et al. summarize additional studies on these and a range of additional materials including gluten, banana stems, and biodegradable nanofibers (see Toward Better Masks) (360). While cloth masks provide a lower level of protection than surgical masks (179), they can be reused. There is ongoing research into developing novel anti-microbial fabrics impregnated with graphene, copper, silver, or zinc nanoparticles, as well as design of fabric masks with improved fit that can perform as well as a surgical mask (92, 393, 394).

While 175 countries around the world have committed to ending plastic pollution, there are currently no international laws, regulations, or restrictions regarding mask disposal (360). Measures by national governments are needed, along with incentives (e.g., tax breaks) for those developing more environmentally friendly alternatives.

The above approaches should be pursued alongside other protections against airborne pathogens, notably attention to indoor air quality. If all indoor spaces were optimally ventilated, for example, protection would be less reliant on individuals’ ability and willingness to mask (395, 396).