In this prospective multicentre cohort of unvaccinated HCW, FFP2 respirator use outside of AGP was marginally not associated with a decreased risk for SARS-CoV-2 infection compared to surgical masks. However, subgroup analysis suggested a protective effect for those with frequent COVID-19 patient exposure. Using FFP2 irrespective of the patient’s COVID-19 status did not provide additional protection for HCW involved in AGP. The large sample size, the dual approach for outcome assessment, and consideration of a variety of potential confounder variables (including personal risk factors, use of other PPE, and general risk perception) are among the strengths of this study.
This is, to our knowledge, the first prospective multicentre study comparing the effect of respirators and surgical masks regarding protection from SARS-CoV-2.
The overall association between FFP2 use and risk for SARS-CoV-2 infection was marginally not significant. This is probably a reflection of the heterogeneous study population, two thirds of which consisted of HCW with only sporadic (or even no known) COVID-19 exposure. However, for HCW with frequent exposure, we found a significant protective effect associated with FFP2 use. Several reports suggest that aerosol transmission is indeed a non-negligible mode of SARS-CoV-2 transmission and that respirators may provide additional protection compared to surgical masks [17,18,19, 21]. On the other hand, case reports have suggested that surgical masks are equivalent to respirators in protecting HCW from SARS-CoV-2 infection [22, 23]. These supposedly contradictory findings can be reconciled when considering a particular feature of SARS-CoV-2, namely its high overdispersion [24]. Overdispersion describes the highly variable transmissibility of infected individuals; in other words, only a minority of infected individuals actually transmit the virus to others, often within so-called superspreading events. As a consequence, the probability of being exposed but not infected is relatively high (irrespective of mask type). Supporting this hypothesis, a simulation study by Chen et al. describes the large variability in SARS-CoV-2 viral loads of infectious individuals and how this influences infection probability and the effectiveness of the different mask types [25].
Our effect size (aOR 0.7) was in the range of those reported by Lentz et al. (aOR of 0.4) and Martischang et al. (aOR of 0.7) [17, 19]. Since many HCW in our study did not consistently wear either FFP2 or surgical masks, we have to assume that the protective effect of FFP2 might be even higher in reality. However, the clinical significance of the protective effect mediated by FFP2 use can be questioned, given the dominating impact of extra-occupational exposures on the COVID-19 risk in HCW, as seen in other studies [26]. Moreover, the disadvantages of respirators (the discomfort over long periods of time, the possibility of a diminished protective effect without prior training and fit testing, and their cost) have to be considered when assessing the net benefit of FFP2 over surgical masks [27, 28].
Notably, we did not observe any protective effect of FFP2 for HCW performing AGP in the absence of COVID-19 suspicion in the patient. We extrapolate from these findings that universal FFP2 use in the hospital setting, where the average exposure risk is usually lower than during AGP, does not provide additional protection compared to surgical masks. We acknowledge however that in settings with a high proportion of undiagnosed, asymptomatic or presymptomatic patients, an additional benefit through universal FFP2 use cannot be excluded.
In sensitivity analysis, treating cantons and institutions as fixed effects alluded to a diminished association of the protective effect of FFP2 use (but only for the outcome of self-reported SARS-CoV-2 swab and not for seroconversion). This could be explained by differences in testing of HCW between institutions. However, the occurrence of regional differences and institutional factors contributing to the observed effect cannot be excluded. Similarly, excluding participants with positive household contact resulted in a non-significant association for the outcome of self-reported positive swabs. Yet again, the consistent result for seroconversion (which is a more objective outcome than self-reported swabs) strengthens the validity of our data.
To adjust for potential confounding, we included the use of gloves, gowns and goggles in our multivariable analysis. None of these measures were associated with any clear additional protective effect. Other associations with SARS-CoV-2 infection found in our study, such as the “protective” effect of active smoking or the increased risk associated with working as a nurse, have been discussed earlier [20].
Our study has limitations. First, residual confounding is possible. Yet, we have included multiple co-variables accounting for risk exposures and risk behaviours within and outside the hospital. Also, the fact that use of other PPE or universal respirator use among HCW performing AGP (representing HCW with particularly risk-averse behaviour) were not associated with reduced seroconversion rate, supports our argument of a valid multivariable model with low risk of residual confounding. Second, information about respirator use was collected in January 2021, when most SARS-CoV-2 positive participants had already had their infection. A positive test result could have led to a change in preferred mask type (in either direction). However, restricting the analysis to the time period close to the follow-up questionnaire showed similar or even stronger associations compared to the full model. Third, recall bias concerning use of mask type cannot be completely ruled out, though the ongoing debate about mask type in health care institutions, especially in the beginning of the study period, made their use a conscious choice and therefore likely to be remembered. Fourth, we did not specifically ask about type and duration of contact to individual COVID-19 patients, although type of profession, work percentage, or involvement in AGP can be regarded as proxy for this potentially important variable. Fifth, although we included multiple institutions, settings, and geographical regions in our study, the generalizability of the results can be questioned due to the fact that study participation was non-mandatory. However, distribution of key variables (e.g. age, sex, profession) were similar between the total HCW population (from the largest participating institution) and the cohort population [20]. Also, these results might not be valid for a vaccinated HCW population, as only data from non-vaccinated HCW were analysed. As more HCW become immunized the protective effect of FFP2 can be expected to diminish. Sixth, this study was performed before the emergence of the SARS-CoV-2 Delta and Omicron variants. The new variants show an increased transmission potential, which might lead to underestimation of the protective effect of FFP2 [29].