Face masks to prevent transmission of respiratory infections: Systematic review and meta-analysis of randomized controlled trials on face mask use

Abstract

Objectives: To examine the use of face mask intervention in mitigating the risk of spreading respiratory infections and whether the effect of face mask intervention differs in different exposure settings and age groups.

Design: Systematic review and meta-analysis. We evaluated the risk of bias using the Cochrane Risk of Bias 2 tool (ROB2).

Data sources: We searched PubMed, Embase, Cochrane Central Register of Controlled Trials, and Web of Science were searched for randomized controlled trials investigating the effect of face masks on respiratory infections published between 1981 and February 9, 2022. We followed the PRISMA 2020 guidelines.

Eligibility criteria for selecting studies: We included randomized controlled trials investigating the use of face mask intervention in mitigating the risk of spreading respiratory infections across different exposure settings.

Results: We identified 2,400 articles for screening. 18 articles passed the inclusion criteria for both evidence synthesis and meta-analysis. There were N = 189,145 individuals in the face mask intervention arm and N = 173,536 in the control arm, and the follow-up times ranged from 4 days to 19 months. Our results showed between-study heterogeneity (p < 0.0001). While there was no statistically significant association over all studies when the covariate unadjusted intervention effect estimates were used (RR = 0.977 [0.858-1.113], p = 0.728), our subgroup analyses revealed that a face mask intervention reduced respiratory infections in the adult subgroup (RR = 0.8795 [0.7861-0.9839], p = 0.0249) and in a community setting (RR = 0.890 [0.812-0.975], p = 0.0125). Furthermore, our leave-one-out analysis found that one study biased the results towards a null effect. Consequently, when using covariate adjusted odds ratio estimates to have a more precise effect estimates of the intervention effect to account for differences at the baseline, the results showed that a face mask intervention did reduce respiratory infections when the biasing study was excluded from the analysis (OR = 0.8892 [0.8061-0.9810], p = 0.0192).

Conclusion: Our findings support the use of face masks particularly in a community setting and for adults. We also observed substantial between-study heterogeneity and varying adherence to protocol. Notably, many studies were subject to contamination bias thus affecting the efficacy of the intervention, that is when also some controls used masks or when the intervention group did not comply with mask use leading to a downward biased effect of treatment receipt and efficacy.

Trial registration: PROSPERO registration number CRD42020205523.

Fig 1. Prisma flow diagram of included articles.

Fig 2. Risk of bias.

Review authors’ judgement of each of the risk-of-bias item across the included studies evaluated using the Cochrane Risk of Bias 2 tool (ROB2).

Fig 3. Effect of face masks across 17 studies with covariate unadjusted risk ratio estimates and subgroup analysis by community setting.

Pooled effects across 17 studies of unadjusted risk ratio estimates for which the numbers of infected for each intervention arm were available using a random-effects meta-analysis model. TE = treatment effect and seTE = standard error of treatment effect, p-value = ξ² is the p-value for the test of homogeneity of study-specific effect sizes.

Fig 4. Effect of face masks across 18 studies using covariate adjusted odds ratios.

Pooled effects across all 18 studies using a random-effects meta-analysis model. Seven of the 18 studies provided data for adjusted odds ratio estimates. For those studies that did not report adjusted odds ratio estimates we calculated unadjusted odds ratio estimates using the numbers of infected reported in the paper. TE = treatment effect and seTE = standard error of treatment effect, p-value = ξ² is the p-value for the test of homogeneity of study-specific effect sizes.