Physical interventions to interrupt or reduce the spread of respiratory viruses. Part 1 - Face masks, eye protection and person distancing: systematic review and meta-analysis

Inclusion criteria

We included randomised controlled trials (RCT) and cluster-randomised controlled trials (C-RCT) including people of any age that tested the use of face masks (i.e. surgical or medical masks and N95 respirators), eye protection, or person distancing against standard practice, or a similar physical barrier, or compared any of these interventions. We only included studies that reported a measure of acute respiratory illness – such as influenza-like illness, influenza, or respiratory infections – and/or its consequences (e.g. days off work, complications, hospitalisation and death, if clearly reported as consequences of the respiratory illness). We also included relevant studies from the previous versions of this review.^7-9

Search strategy

We identified RCTs and C-RCTs studying effectiveness of eye protection (any purposed device excluding simple eyeglasses), masks (defined as any type of facial mask), and person distancing from our 2011 review. ⁸ These earlier studies were analysed using word frequency to create a new search string that was run in PubMed. ¹⁰ This search string was converted using the Polyglot Search Translator¹¹ and run in the following additional databases; the Cochrane Central Register of Controlled Trials, Embase and CINAHL. The search covered the dates October 2010 to 9 March 2020. Search strings for all databases are available in the appendices (Appendix 1). A backwards and forward citation analysis, using Scopus, was conducted on all new studies retrieved. Search and citation analysis results were screened using the RobotSearch tool to remove all obvious non-RCTs. ¹² Three authors (JC, MJ and ST) independently reviewed the titles and abstracts of the identified studies to assess eligibility for inclusion. Discrepancies were resolved by consensus.

Risk of bias assessment

Risk of bias was assessed by three author pairs independently (TJ, EB, LA, GB, MJ, EF) for the method of random sequence generation and allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), outcome reporting (attrition bias), and selective reporting (reporting bias). We used the Cochrane risk of bias tool. ¹³ For each item risk was either ‘high’, ‘low’ or ‘unclear’. Low risk of bias for the method of random sequence generation indicates that the method was well-described and is likely to produce balanced and truly random groups; for allocation concealment that the next treatment allocation was not known to participant/cluster or treating staff until after consent to join the study; for blinding of participants and personnel that the method is likely to maintain blinding throughout the study; for blinding of outcome assessors that all assessing outcomes were unaware of treatment allocation; for outcome reporting that participant attrition through the study is reported and reasons for loss are appropriately described; and for selective reporting that all likely planned and collected outcomes have been reported.

Data extraction and analysis

Six authors (TJ, EB, LA, GB, MJ, EF) independently extracted data in pairs. Discrepancies were resolved by consensus. Descriptions of the interventions were extracted using the Template for Intervention Description and Replication (TIDieR) template. ¹⁴ We entered outcome data in RevMan software and used a generalised inverse variance random effects method for pooling. The effect estimate was expressed as a risk ratio with 95% confidence interval. We calculated the I² statistic for each pooled estimate to assess statistical heterogeneity. ¹⁵ For studies that could not be pooled we report the effect estimates as reported by the study authors. We conducted a subgroup analysis for interventions aimed at protecting health care workers.

Differences between 2009 review and current review

The 2009 review included both randomised trials, cluster randomised trials, and observational studies. This update excluded the latter. This update was also split into two parts, according to intervention categories. This Part 1 is focussed on face masks, eye protection, and person distancing; Part 2 will address all other categories of physical interventions.

Results of the search

The updated search yielded 2468 references after removal of duplicates of which 2345 were excluded. A further 42 were excluded after review of the full text paper. Backwards and forward citation analysis identified a further 20 studies resulting in a total of 101 papers testing a range of barrier interventions aimed at interrupting the spread of respiratory viruses. For this part 1 of our review we included 15 RCT/c-RCTs, including 5 trials from the 2011 review^16-20 (see Figure 1).

• Download figure
• Open in new tab

Figure 1: PRISMA Flow Diagram

Risk of bias

Reporting of sequence generation, allocation concealment (particularly in cluster-randomised trials) and blinding of outcome assessment was poor, leading to an unclear risk of bias judgement for 30% of studies (see Figure 2). The majority of studies were unblinded due to the nature of the masks, or insufficient information (70% of studies with high or unclear risk of bias). Only one study was blinded to staff. ²¹ For the remainder of the unblinded studies at low risk of bias, this was due to them having objective outcomes that were unlikely to be affected by unblinding. More than 80% of studies had no evidence of serious attrition and described reasons for losses to follow-up well. 70% of studies had no evidence of selective outcome reporting. One study had what appeared to be selective testing or reporting of viral tests, another had selective reporting of non-viral isolates and changes during the study that made planned outcomes unclear. The remainder of the studies had unclear risk of bias for this domain due to insufficient information reported.

• Download figure
• Open in new tab

Figure 2.

Risk of Bias in included studies - summary bar chart

Eye protection

We found no trials on the use of eye protection as a single intervention.

Masks

Nine trials compared masks with no masks. ^16-18,21-26 Two of these studies included health care workers^18,25 and 7 others included people living in the community. All trials were conducted in non-pandemic settings. A description of the interventions is presented in Table 1. Included trials are described in Table 2. Pooling of all nine trials did not show a statistically significant reduction of ILI cases (Risk Ratio 0.93, 95%CI 0.83 to 1.05) or laboratory-confirmed influenza cases (Risk Ratio 0.84, 95%CI 0.61-1.17) in the group wearing a mask compared to those not wearing a mask (see Figure 3a). Eight-seven percent of the weight of this analysis is carried by two studies from the same first author. ^16,22 A separate analysis of the two trials in healthcare workers also failed to show a statistically significant difference between the mask and no mask groups (Risk Ratio 0.37, 95%CI 0.05 to 2.50).

• Download figure
• Open in new tab

Figure 3a:

Meta-analysis of trials comparing masks vs no masks in healthcare workers and in community populations: effect on rates of Influenza-like Illness (ILI) and laboratory-confirmed influenza

Five trials compared surgical masks with N95/P2 respirators. ^25,27-30 All trials except MacIntyre 2009²⁰ included healthcare workers. Pooling of four trials showed no difference between surgical/medical face masks and N95 respirators for rates of clinical respiratory illness (Risk Ratio 0.70, 95%CI 0.45 to 1.10), for ILI (Risk Ratio 0.83, 95%CI 0.63 to 1.08), or for laboratory-confirmed influenza (Risk Ratio 1.02, 95%CI 0.73 to 1.43) (see Figure 3c). If only studies in healthcare workers are compared the Risk Ratio for ILI is 0.64, 95%CI 0.32 to 1.31. The outcomes ‘clinical respiratory illness’ and ILI were reported separately by the authors. Considering how these outcomes were defined it is highly likely that there is considerable overlap between the two and therefore these outcomes were not combined into a single clinical outcome. Harms were poorly reported, but generally discomfort wearing masks was mentioned²⁴ and Radonovich³⁰ mentioned that participants wearing the N95 respirator reported skin irritation and worsening of acne.

• Download figure
• Open in new tab

Figure 3b:

Meta-analysis of trials comparing masks vs no masks in healthcare workers: effect on rates of Influenza-like Illness (ILI)

• Download figure
• Open in new tab

Figure 3c:

Meta-analysis of trials comparing (surgical/medical) masks vs N95 respirators in healthcare workers and in community populations: effect on rates of clinical respiratory illness, Influenza-like Illness (ILI) and laboratory-confirmed influenza

MacIntyre 2015²⁵ also included a trial arm with cloth masks and found that the rate of ILI was higher in the cloth mask arm compared to medical/surgical masks (RR 13.25, 95%CI 1.74 to 100.97) and compared to no masks (RR 3.49, 95%CI 1.00 to 12.17).

Person distancing

One trial evaluated the effectiveness of quarantining workers of one of two sibling companies in Japan whose family members developed an influenza-like Illness (ILI) during the 2009-2010 H1N1 influenza pandemic. ³¹ Workers in the intervention group were asked to stay home on full pay until 5 days after the household member(s) showed resolution of symptoms or 2 days after alleviation of fever. Compliance was 100%. In the intervention group 2.75% of workers contracted influenza, compared with 3.18% in the control group (Cox Hazard Ratio 0.799, 95%CI 0.658 to 0.970, p=0.023), indicating a 20% reduction of infection in the intervention group. However, the risk of a worker being infected was 2.17-fold higher in the intervention group where workers stayed at home with their infected family members. The authors conclude that quarantining workers with infected household members could be a useful additional measure to control spread of respiratory viruses in an epidemic setting.

Main findings

Our results show that masks alone have no significant effect in interrupting spread of ILI RR 0.93 (95% CI 0.83 to 1.05) or influenza RR 0.84 (95% CI 0.61 to 1.17) in the all populations analysis. Our findings are similar for ILI in healthcare workers RR 0.37 (95% CIs 0.05 to 2.50) and for the comparisons between N95 respirators and surgical masks: RR 0.70 (95% CI 0.45 to 1.10) for clinical respiratory illness, RR 0.83 (95% CI 0.63 to 1.08) and influenza RR 1.02 (95% CI 0.73 to 1.43). Five of the trials contributing to the analysis were carried out by members of the same group. ^{20,25,26,28,29} On the basis of one trial²⁵ cloth surgical masks should not be used as they are associated with a higher risk of ILI and penetration of microorganisms. In general, harms were not or poorly reported, with general discomfort resulting in reduced compliance with wearing being the main issues.

One trial testing person distancing found a reduction in transmission to co-workers when those with infected household members stay home from work. However, staying home increased their risk of being infected two-fold. We were disappointed to find only one trial on person distancing which is currently the core of the global containment strategy. This points to the difficulty and lack of interest in carrying out such studies.

Even though this update of the review focussed only on randomised and cluster randomised trials, the available body of evidence is inconclusive. We found a body of relatively small trials conducted mostly in a non-epidemic context (low viral circulation), with the exception of the largest study which crossed during the active study period two of the highest reporting years for influenza in the United States, between 2010-2017. ³² The two largest studies with respect to event rates are^19,30 consistent regarding the direction of their findings of no differences between surgical or N95 masks.

Collectively, the evidence base was of variable quality. Inadequate reporting of sequence generation and allocation concealment was common. While allocation concealment of cluster-randomised trials is important this was rarely reported. Due to the nature of the intervention comparison, most trials were unblinded. However, blinding of outcome assessment is highly feasible and highly desirable, but was rarely done or reported. Outcomes were poorly defined with lack of clarity as to possible etiology of the agents (bacterial vs viral) in some studies. The cluster trials had insufficient attention paid to adjusting sample size calculations and analysis for clustering. As a consequence many trials were underpowered and had spuriously narrow confidence intervals around the effect size. The variable quality of the studies places some limits on the generalisability to the current COVID-19 epidemic, albeit it is a respiratory virus with a similar mode of transmission to those in the reviewed studies.

Comparison with other reviews

In a meta-analysis comparing surgical masks with N95 respirators Smith³³ pooled three trials^19,28,29 and found no significant difference (OR 0.89, 95% CI 0.64 to 1.24) for laboratory-confirmed respiratory infections or ILI (OR 0.51, 95% CI 0.19 to 1.41). A similar meta-analysis by Offeddu et al³⁴ concluded that based on two studies by MacIntyre et al^25,28 masks (either surgical masks or N95 respirators) were effective against clinical respiratory infections (RR 0.59, 95% CI 0.46 to 0.77) and ILI (RR 0.34, 95% CI 0.14 to 0.82). Pooling the same two studies they also found that N95 respirators were superior to surgical masks for Clinical respiratory infections (RR 0.47, 95% CI 0.36 to 0.62), but not for ILI (RR 0.59, 95%CI 0.27 to 1.28). ³⁴ The most recent meta-analysis by Long et al³⁵ included 5 studies comparing surgical masks with N95 respirators and found no difference (RR 1.09, 95% CI 0.92 to 1.28) against neither influenza nor respiratory viral infections (RR 0.89, 95% CI 0.70 to 1.11).35 By excluding the Loeb¹⁹ study (an open non-inferiority randomised, controlled trial carried out to compare the surgical mask with the N95 respirator in protecting healthcare workers against influenza) the authors found a significant effect against viral infections (RR 0.61, 95% CI 0.39 to 0.98). The authors do not report a rationale for the exclusion in the sensitivity analysis and do not report on exclusion of the studies with low weighting which arguably would be more relevant in a sensitivity analysis. The two studies which make up 96% of the weighting^19,30 clearly demonstrate no difference in the outcome events.

The findings from several systematic reviews and meta-analyses over the last decade have not demonstrated any significant difference in the clinical effectiveness of N95 respirators or equivalent compared to the use of surgical masks when used by healthcare workers in multiple health care settings for the prevention of respiratory virus infections, including influenza.

Our 2011 review⁸ showed a clear protective effect of wearing surgical masks and hygienic measures compared to not wearing masks in the SARS 2003 outbreak (RR 0.32, 95% CIs 0.26 to 0.39). The evidence was based on case-control studies carried out during the outbreak.

Relevance of the Findings in the Clinical Setting

Our findings are highly relevant in the setting of epidemic and pandemic respiratory infections and the current global pandemic of COVID-19 underscores the point. The evidence supports that SARS-CoV-2 is spread through respiratory droplets and/or contact routes which places it in the route of transmission for which health care workers would be required to wear masks.^36,37 The WHO China Joint Mission on COVID-19 of 75,465 cases supports person-to-person droplet and fomite transmission, with the majority of transmission occurring within families in close contact with each other. ³⁶ A recent report in a clinical setting of intubation and non-invasive ventilation in which 41 health care workers were exposed over a prolonged period a within close proximity to a COVID-19 + patient revealed no transmission events to the Health care workers based on repeated testing during which majority (85%) of the health care workers were wearing a surgical mask and other appropriate PPE while the remainder wore an N95 respirator. This latter finding supports the results of the studies which were reviewed and brings into focus the importance of the use of masks as a component of personal protective equipment in the current COVID-19 pandemic. The current COVID-19 pandemic has elicited conflicting recommendations with several institutions and countries suggesting that only N95 or equivalent masks should be used as a component of the personal protective equipment for health care workers and not a surgical mask. The WHO recommendations emphasize that in the setting of epidemic and pandemic respiratory virus infections transmitted predominantly by the droplet route, one of the most important elements is strict adherence in the use of personal protective equipment of which the facial mask is only one component³⁸ and suggest surgical masks for routine care and reserve the N95 mask for aerosol generating medical procedures. Despite the methodological issues outlined, our review of the available literature did not find any differences in the clinical effectiveness of either type of mask in the setting of respiratory viral infection transmission to health care workers. Our review also identified a dearth of reported findings related to the harms of N95 respirators and they need to be considered in any RCTs or C-RTs, especially in the setting of a global pandemic with the potential of frequent and prolonged use. Many such harms were identified in the setting of the SARS epidemic in 2003 and in the ensuing years and included respiratory fatigue, increased work of breathing, poor work capability, increased nasal resistance, fatigue with minimal workloads, elevated levels of carbon dioxide, facial dermatitis, acne and potential self-contamination events.^39-44

Limitations

Though the trials in this review provide a reasonable body of evidence, there are several important limitations. First, there is considerable clinical heterogeneity between the designs, and substantial statistical heterogeneity for some analyses. The latter is not readily explained by differences in the study questions. Second, the range of viral infections studied is limited, with a particular focus on influenza; no studies include SARS-CoV-2. None of the studies in health care worker included undertaking aerosol-generating procedures for which WHO currently recommends the N95 or equivalent mask. Finally, the studies provide sparse and unsystematic data on any harms, such as the discomfort, dehydration, facial dermatitis, distress, headaches, exhaustion or other problems caused by masks such as the N95. Some studies measured adherence which was generally high despite the mask discomfort.

Conclusion

Despite the lack of evidence, we would still recommend using facial barriers in the setting of epidemic and pandemic viral respiratory infections, but there does not appear to be a difference between surgical and full respirator wear. Despite the methodological concerns, our review of the available studies demonstrates consistency in the finding of no difference between surgical and N95 or equivalent masks as a physical intervention to interrupt or reduce the spread of respiratory viruses, mainly influenza. The consistency of the finding across multiple studies of variable quality adds epidemiologic strength of association.

The fact that all included trials were conducted in relatively low transmission periods limits generalisability to an epidemic of the global size of COVID-19. We excluded in this part of the review trials testing the combination of hygienic and barrier methods. These have shown to be effective in observational studies carried out during the SARS 1 epidemic.⁸

Based on the evidence of the previous SARS epidemic large trials comparing full facial protection with surgical masks need to be carried out to settle the matter, given the difference in wearability, harms and costs. Funding for such trials and research once the epidemic has passed, is critical to inform future preparedness for global epidemics.