Decontaminating N95/FFP2 masks for reuse during the COVID-19 epidemic: a systematic review

With the current COVID-19 pandemic, many healthcare facilities have been lacking a steady supply of filtering facepiece respirators. To better address this challenge, the decontamination and reuse of these respirators is a strategy that has been studied by an increasing number of institutions during the COVID-19 pandemic. We conducted a systematic literature review in PubMed, PubMed Central, Embase, and Google Scholar. Studies were eligible when (electronically or in print) up to 17 June 2020, and published in English, French, German, or Spanish. The primary outcome was reduction of test viruses or test bacteria by log3 for disinfection and log6 for sterilization. Secondary outcome was physical integrity (fit/filtration/degradation) of the respirators after reprocessing. Materials from the grey literature, including an unpublished study were added to the findings. Of 938 retrieved studies, 35 studies were included in the analysis with 70 individual tests conducted. 17 methods of decontamination were found, included the use of liquids (detergent, benzalkonium chloride, hypochlorite, or ethanol), gases (hydrogen peroxide, ozone, peracetic acid or ethylene oxide), heat (either moist with or without pressure or dry heat), or ultra violet radiation (UVA and UVGI); either alone or in combination. Ethylene oxide, gaseous hydrogen peroxide (with or without peracetic acid), peracetic acid dry fogging system, microwave-generated moist heat, and steam seem to be the most promising methods on decontamination efficacy, physical integrity and filtration capacity. A number of methods can be used for N95/FFP2 mask reprocessing in case of shortage, helping to keep healthcare workers and patients safe. However, the selection of disinfection or sterilization methods must take into account local availability and turnover capacity as well as the manufacturer; meaning that some methods work better on specific models from specific manufacturers. CRD42020193309.


Introduction
With the current COVID-19 pandemic, healthcare facilities have suffered from shortage of N95 or filtering facepiece 2 (FFP2) respirators due to concurrent increased demand and decreased production capacity and supply interruptions, or lack of resources. This emergency situation warrants the taking of extraordinary measures to maintain the security of health workers delivering care to COVID-19 patients. Decontamination and reuse of N95/FFP2 respirators is a promising solution, which has been envisaged by an increasing number of institutions all over the world during the first wave of the COVID-19 pandemic.
N95/FFP2 respirators are single use personal protective equipment and thus, their reprocessing was underexplored. Over the past decades, the idea of reprocessing N95/FFP2 respirators emerged during outbreaks due to SARS-CoV and influenza, but given that these events were geographically or temporally limited compared to the COVID-19 pandemic, there was no urge to proceed to legally challengeable protocols of reprocessing single use devices. The COVID-19 pandemic urged various stakeholders to consider N95/FFP2 respirator reprocessing as an alternative to protect health workers at the frontline. The main challenge today is to find reliable information on safe reprocessing within the vast quantity of publications on COVID-19 in the last months. Although in this review we focus on the efficacy of decontamination methods, it goes without saying that respirators should be reprocessed only if no better options are available. As most of the studies have so far been performed by industry, it is important for reviews on the topic to take the grey literature into account.
In order to have a clear understanding of the research that has already been conducted, we performed a systematic literature review for the question: What are the tested methods for decontaminating N95/FFP2 respirators, and what is the efficacy of those methods on viral contamination? The primary outcome was efficacy on reducing pathogens (viruses or bacteria) on pre-contaminated N95/FFP2 respirators. The secondary outcome looked at the physical integrity (fit/filtration/degradation) of the masks after the decontamination process.
The overall aim of this systematic review was to assess the microbiological efficacy of decontamination methods of N95/FFP2 respirators. We hope that this overview of the work performed thus far will help orient further research as well as help healthcare facilities make decisions regarding methods of respirator decontamination.

Methods
We performed a systematic review protocol according to the PRISMA checklist [1]. Considering that we expected many papers of interest not being available in PubMed, Cochrane or Embase, we broadened our search strategy towards PubMed Central and Google Scholar. The results were contextualized with further material from the grey literature, including unpublished data from the Geneva University Hospitals (HUG) laboratory and partner institutions.
All controlled original studies were eligible when applying a quantitative study design and measuring the effect of a decontamination strategy on a microbiological outcome such as respiratory viruses or bacteria, and published until the 17th of June, 2020. Studies with an English abstract were eligible when published in English, French, German, or Spanish.
We applied to following search strategy: (N95 OR FFP2 OR KN95) AND (decontamination OR disinfection OR sterilization) AND (reuse OR reprocessing OR reusing) AND (coronavirus OR "COVID 19" OR "SARS CoV-2" OR stearothermophilus OR influenza). The full search strategies are available in the Additional file 1. We included stearothermophilus in our search terms because it often used as an indicator pathogen decontamination testing, and influenza because we thought that it would possibly be a virus that would often be tested due to the 2009 pandemic. We included all studies that came up through the search and matched our search criteria, meaning that results from tests on bacteria and bacteriophages were also included in the table.
Results were stratified by the decontamination methods test organisms used, and whether the effect was disinfection (≥ log 3 reduction of the test microorganism) or sterilization (≥ log 6 reduction of the test microorganism). Any decontamination method was eligible if measuring its efficacy on a microorganism compared to a control. Primary outcome was disinfection or sterilization of the test microorganism. Therefore, all studies included were compatible with this outcome domain. Secondary outcomes included physical integrity, fit testing, and filtration capacity after reprocessing, and not all studies had data on this. We excluded articles other than original research.
The "study population" included N95, KN95, or FFP2 respirators or their equivalents. Outcomes were designated as "sterilization", "disinfection" and "failure". "Failure" was defined as having a lower than 3 log reduction of the test microorganism.
Titles and abstracts were screened independently by two reviewers. Duplicate articles were removed. After deduplication, 961 articles were screened. The full text of all potentially eligible studies was independently assessed by at least two authors. Disagreements were resolved by consensus or by consulting a third reviewer where necessary. Authors tabulatied the study intervention characteristics and compared them against the planned groups for synthesis. Inclusions and exclusions were recorded following the PRISMA guidelines, and reasons for exclusion were detailed. Data of each study was extracted by two review authors. The standardized extraction forms collected the following information: study title, author; year of publication; intervention(s), microorganism tested, microbiological outcome (log reduction), whether authors recommended their method of decontamination, whether disinfection or sterilization was achieved, data on physical integrity when available (including fit or filtration), and comments. These elements were included in the table (see Table 1), and were organized by type of intervention. These types of (liquid, gasses, heat, UV) were analyzed as individual subgroups.
All included studies were assessed for their inclusion by at least two review authors. All eligible studies were invitro studies; thus, only studies testing a method against a control were eligible. Studies that were finally not included were usually because the outcome didn't quantify a microbiological reduction. We expected to identify a large variety of procedures and methodologies. Thus, a descriptive analysis with narrative synthesis was planned. Ethical approval was not required for this review.

Results
Our search identified a total of 23 publications from PubMed, 18 from Embase, 0 from Cochrane, 229 from Pubmed Central, and 988 from Google Scholar. Originally, there were 1010 results using Google Scholar, but Google Scholar only allows access to the first 99 pages of results; and thus, we were not able to assess the last 22 articles from the Google Scholar results (Fig. 1). A total of 35 studies were included for final analysis . The following decontamination methods were identified (Table 1 and see complete table in the on-line Appendix): (saturated) steam (with pressure; autoclave), moist heat (without pressure; devices other than autoclaves), dry heat, UVGI (UVC and UVA), hydrogen peroxide vapor (HPV) and its associated forms (aerosolized or ionized) either alone or in combination with peracetic acid, peracetic acid (dry fogging system), ethylene oxide, ozone, ethanol, sodium hypochlorite, benzalkonium chloride, and detergent (non-antimicrobial) wipes. Individual studies were not assessed for bias or for risk of reporting bias. As a number of these small, in vitro studies were conducted my industry, it is possible that unfavorable results were not published as often.
Procedures of the methods were generally well described. Nine studies were performed before 2020, with the oldest publication from 2011. Many of these were performed in the context of past influenza pandemics (4 of the 9 studies used a strain of influenza as a test organism). In total, 70 individual tests were conducted, with 14 of the 35 studies evaluating more than one method. Sample sizes ranged from 3 to 115 tested respirators. Often, information regarding the sample size was unclear. The overwhelming majority of the experiments applied heat at different humidity levels, UVGI, or gaseous hydrogen peroxide (Table 1). Details regarding the studies such as the type of intervention, microorganisms tested, quantification methods, log reduction compared to control, data on physical integrity, fit, and filtration, and whether the method is recommended are provided in the table).
Numerous chemicals and processes can be used for disinfecting or sterilizing unsoiled N95/FFP2 respirators. None of these methods are new [37], but recognized processes for disinfecting or sterilizing medical equipment. Overall, 17 distinct processes (or distinct combinations of processes) were tested (see Table 2).

Prior knowledge and what this manuscript contributes
With the current COVID-19 pandemic, many healthcare facilities have been lacking a steady supply of filtering facepiece respirators (N95/FFP2). To better address this challenge, the decontamination and reuse of these respirators is a strategy that has been studied by an increasing number of institutions during the COVID-19 pandemic. Although the initial shortages in high-income countries have been resolved for the moment, this area of study is still crucial for low-resource areas, as we head into yet another wave of COVID-19. Learning how to navigate shortages of N95 masks safely, will no doubt have a positive impact on healthcare delivery in the future.
Prior to COVID-19, there were very few studies that looked into the reprocessing of respirators. One looked at S. aureus and mucin contamination [2], one looked at Bacillus subtilis spores, 3 three on MS2 bacteriophages [18,32,34], and four on influenza viruses [14,15,17,31]. A review on the same subject was published by Rodriguez-Martinez et al. as a preprint in July 2020 while we were still working on our review [38]. Still, our review is more comprehensive, as we took all of the grey literature into account, where the majority of the studies were published. Therefore, this review has over twice the number of studies included. This means that it provides far more resources for healthcare facilities, especially in low resource settings, that still do not have access to the supplies they need today.

Log reductions
The accepted log reductions for disinfection and sterilization and 4-5 log (depending on whether bacteria/yeasts/ or viruses) and 6-log, respectively. However, the US food and Drug Administration announced that for reprocessing masks during COVID-19, "bioburden reduction treatment should result in ≥ 3-log (1000-fold) reduction in microbial numbers and is consistent with a Tier 3 bioburden reduction system" [39]. Additionally, log reductions were often shown as a range or a minimum. Studies shows the reduction from 3-5logs, others 4-5 or as greater than 3. Numerous studies using masks contaminated by human samples (where the contamination was at 3log itself ) in an effort to replicate real life conditions were unable to show even a 3log reduction, or in some cases it was enough to show disinfection but not sterilization [4,12,21,23,28]. Sometimes papers didn't detail log reductions but used the terms disinfection and sterilization, when there was no bacterial growth on the medium [24,25]. Therefore, due to the danger of excluding methods that worked because the tests weren't designed to show as big of a log reduction as possible we decided that it would be better to count anything over a 3-log reduction as "probable disinfection", and use the generally accepted 6-log reduction for sterilization.
Referring to results in general, ethylene oxide, gaseous hydrogen peroxide, gaseous hydrogen peroxide with peracetic acid, peracetic acid dry fogging system, microwave-generated moist heat, and steam seem to be the most promising methods on decontamination efficacy, physical integrity and filtration capacity. However, successful decontamination or failure is not inextricably linked to the physicochemical process. UVGI, for instance, showed both very good results and failures. Thus, technical implementation and engineering may play an equally important role than the physicochemical properties of a method.
Ensuring the safety of these decontamination methods is paramount. Reprocessing procedures using chemicals such EtOH as may leave a concentration that could be harmful for the user due to residual amounts of reprocessing substances or by-products produced during the process. Steps to ensure the safety of such reprocessing methods on porous masks must therefore be taken into consideration and implemented. If masks are reprocessed without determining the residual quantities of chemicals, this could prove harmful to the users.
We conducted a sub-analysis of the results by decontamination method.
For the decontamination methods using liquids One experiment tested benzalkonium chloride wipes on bacteria resulting in disinfection; fit and filtration testing failed [2]. Two experiments tested hypochlorite on bacteria (including spore-forming bacteria) resulting in disinfection [2,3]. Fit and filtration testing was performed in one of the studies and failed. One experiment tested non-antimicrobial detergent wipes on bacteria resulting in a failure to disinfect; fit and filtration testing failed as well [2]. Three experiments tested ethanol on viruses and spore-forming bacteria resulting in a mixture of disinfection and failure; fit and filtration testing failed [3][4][5].
For the decontamination methods using gases Though peracetic acid dry fogging system (PAF) is technically not a gas, dry fogging is a method that functions in a similar context as gas disinfection would, where the mask is enveloped in very small particles of the substance versus being doused in a liquid. Due to this, we have included PAF in this section. One experiment tested PAF on viruses resulting in a mixture of sterilization and disinfection; fit and filtration testing passed [21]. Two experiments tested ethylene oxide on viruses resulting in a mixture of sterilization and disinfection [8,21]. Fit and filtration testing was performed in both the studies and both passed [8,21]. Eleven experiments tested gaseous hydrogen peroxide on viruses, bacteria, spore-forming bacteria, and fungus, resulting in a mixture of sterilization, disinfection, and one failure [4,5,8,9,[21][22][23][24][25]. Fit and filtration testing was performed in nine of the studies, and all passed [4,5,8,[21][22][23][24]. Three experiments tested ozone on viruses and bacteria resulting in a mixture of sterilization, disinfection, and one failure [27][28][29]. Fit and filtration testing was performed in all of the studies; one passed [28], and the other two passed in the face piece but failed in the elastic band [27,29].
For the decontamination methods using UV light Fifteen experiments [3-6, 8, 9, 14, 15, 17, 30-35] tested UVGI on viruses, bacteria (including spore-forming bacteria, and fungus, resulting in a mixture of sterilization, disinfection, and failure. Fit and filtration testing was performed in seven studies [4,5,8,14,17,30,32]; one failed [4], one had mixed results [32], and five passed [5,8,14,17,30]. One experiment tested UVA on spore-forming bacteria resulting in in a failure to disinfect; fit and filtration testing was not performed [3]. For the decontamination using combinations of methods Two experiments tested gaseous hydrogen peroxide with peracetic acid on viruses and bacteria (including sporeforming bacteria) resulting in sterilization [6,26]; fit and filtration testing was performed for one study [26], which passed. Two experiments [8,36] tested UVGI with dry heat on viruses and bacteria resulting in a mixture of disinfection and failure; fit and filtration testing passed. One experiment tested UVGI with medium-humidity heat on viruses, resulting in a mixture of sterilization, disinfection, failure; fit and filtration testing passed [8].
The literature around reprocessing N95/FFP2 respirators has grown exponentially in the wake of the current COVID-19 pandemic. A large variety of methods were tested on different masks, and using different procedures. All methods were performed in controlled settings and without taking into account physical soiling by biological material from wearing the masks in daily routine. A number of reports did reflect on the potential risk of pathogens other than SARS-CoV-2 in biological fluids and how, as a consequence, reprocessing would need to be individualized. This means that the encouraging results from a number of these in vitro studies may not be applicable to healthcare settings. This should be taken into account when deciding whether to implement any FFP2 decontamination programs in a healthcare facility.
The experiments used a vast array of viruses (both enveloped and non-enveloped), bacteria (including some spore forming bacteria), and fungi. The methods, even when using similar technology, cannot be compared directly because there were differences in protocols (tested pathogens; applied temperature, ppm, concentration, humidity) and study designs.
Sterilization was reached at least some of the time when using ethylene oxide, gaseous hydrogen peroxide, either alone or in combination with peracetic acid, moist heat, ozone, peracetic acid (dry fogging system), UVGI, either alone or in combination with humidity, heat, and steam. Only the two studies using gaseous hydrogen peroxide combined with peracetic acid were able to sterilize all of the tested microorganisms in all experiments. Moist heat and UVGI were often tested, but only rarely resulted in sterilization. A 3log reduction as an endpoint was reached by most of the tested methods except for non-antimicrobial wipes. There, however, a number of methods that failed fit and filtration testing.
Masks reprocessed by ethylene oxide, gaseous hydrogen peroxide, and peracetic acid dry fogging all passed fit testing and filtration capacity. Microwave generated moist heat also had little impact on mask integrity, fit testing and filtration capacity, most of the time. While ozone performed very well for the facepieces, there was a tendency to degrade elastic bands, which could pose a risk of failure during a risky procedure in daily practice. Saturated steam in autoclaves was promising in some studies but failed in others. Most likely, this is due to the overall quality of masks and manufacturing. Though we included all data on mask integrity, fitting and filtration capacity in a "pass/fail" manner, it is important to note that applied testing was very heterogeneous, ranging from simple visual inspection to very complex experiments.
Few of the studies looked at bioaccumulation of physical soil in the masks from breathing. When analyzing decontamination methods, most papers didn't take into account the bioaccumulation of protein from human breath, which could conceivably be significant, especially when performing numerous decontamination cycles. One study found the protein in exhaled breath condensate to be around 1.02 μg/ml [40]. One of the studies in this review looked at this data more closely, and found that the protein from human breath condensate accumulates at ~ 0.34 μg/minute breathing time [2]. Considering that physical soil affects the efficacy of decontamination, and that a soiled surface needs to be cleaned before it can be disinfected or sterilized, this area needs further study.
COVID-19 is a disproportionate threat to low income and low resource environments, where access to care, supplies or testing is challenging, and the living environment does not allow social distancing. Because many of the decontamination methods require costly equipment and specific infrastructure, special consideration should be given to studying simple, lower cost solutions. Some of the identified methodologies using a rice cooker, boiling pot of water, oven, or steam bag in a microwave, are potential procedures to be used in lower-resource settings or at home.
As single-use N95/FFP2 respirators do not tolerate washing with ordinary detergent, they may not be free of soiling. This interferes with the principle of disinfection and sterilization, that such procedures only can be effective if medical equipment has passed through a validated cleaning process. Thus, collecting, reprocessing and redistributing must be organized individually, which engenders a host of logistical challenges. Mask storage before reprocessing may also be a challenge, to avoid growing molds. Methods and logistical issues have a cost, which is much higher than the price per new mask. Furthermore, the capacity of reprocessing equipment such as plasma sterilizers or UVC-lamps is often limited and thus, not applicable in larger healthcare settings where thousands of N95/FFP2 respirators are needed every day.

Limitations
The majority of the identified studies were from the grey literature, were preprints, or were not yet peerreviewed. Many studies were conducted by industry, and may have also be subject to reporting bias. Thus, even if the quality of the experiments may have been acceptable for the most part, the generalization of the efficacy of tested methods is limited. Still, we deemed the extraordinary situation of the current COVID-19 pandemic as a reason to justify the inclusion of as many methods for mask reprocessing as possible, and to encourage further testing of the efficacy of some of the methods. Gaseous hydrogen peroxide [4, 5, 8, 9, 21-25] 11 UVGI [3-6, 8, 9, 30-35]