Montag, 6. April 2020

Masks, politics and the inexplicable superiority of South and East Asia

Introduction
Summary

Observational data
Clinical data

Consistent messaging and biological plausibility
References and Notes

Introduction
How is it possible that different health experts, organizations and governments provide such varying recommendations on mask use? For example, the World Health Organization (WHO) suggested in February of 2020 “that only those who are already sick with a respiratory illness should wear them” as did the American health authorities (reference at the end; and the US medical establishment  continues to be skeptical).

In contrast, Asian politicians urge the public to wear masks (Chan & Yuen 2020), the Austrian government plans to make mask wearing at super markets compulsory (reference) and the Austrian OEGIT (society for infectious diseases) states that wearing masks correctly can contribute to reduce rates of influenza infection (and by extension perhaps COVID19).

Before we get started, let me state my biases. I am not a fan of masks and I never considered it smart to wear them outside of crowded places. I was against wearing masks when there were less than a dozen documented cases due to my personal feelings about cost-benefits, although, now I am not sure if that was even a correct assumption. However, I am also getting sick of people downplaying their benefits, in part, perhaps due to some irrational bias against employing measures that were successful in Asia (test, trace, surveillance, masks, etc).

First, we have to distinguish two issues. Do masks work and do we have enough masks for health care workers (HCWs)? Even if masks work, some actors may see a benefit in downplaying their benefits or outright lying in order to stop a run on masks so as to protect HCWs. Lying is dumb, for obvious reasons, however.

The differences for the economy are tremendous:
  • If masks work for the community and we do not have enough, we need to produce considerably more. We need to start now.
  • If masks do not work for the community and we do not have enough, we only need to produce a little bit more for HCWs.
Summary
The below data is ordered from weakest to strongest:

The use of masks is biologically and physically plausible. Biology dictates that you need a certain number of viral particles to infect an organism. Masks reduce the number of emitted viral particles by filtering and through “dispersion” of the exhaled air. Countries with high mask usage have low rates of new SARS-CoV2 infections. In contrast to popular opinion China is one of these successful countries and to see this you just need to compare their peak case numbers to Europe’s which have not yet peaked.

Experimental studies (using “surrogate measures” not counting infections after mask use!) consistently suggest that any mask can reduce the number of viral particles sick individuals exhale. The efficacy of masks is as expected: cloth < surgical < respirator. Surgical masks (from here on mask refers to surgical masks unless stated otherwise) reduce the number of exhaled viral particles in large droplets, which is called source control. Whether masks are useful for self-protection and reduce the risk of inhaling viral particles is a bit more controversial. However, observational studies rather consistently support a benefit of masks for both self-protection and source control. Some of these studies focused on SARS and MERS, close but distinct relatives of SARS-CoV2, although their design could be better. However, even perfectly designed observational studies can be subject to biases, which is why we need large, properly designed randomized, controlled trials (RCTs).

Randomized trials do not show a consistent benefit of masks to protect from influenza. However, these studies are small and poorly designed. For example, people often do not wear their masks and are instructed to wear them too late. Importantly, if you combine studies that looked at masks and hand washing together, you can see a benefit but this is only evident when you exclude some particularly weak studies. This exclusion may introduce biases.
Furthermore, randomized trials suggest that respirators and masks have similar efficacy to protect health care workers from influenza, but few studies tested either of these against a “no mask” control. Hence in theory both could be ineffective. What is more, influenza data may not translate to SARS-CoV2. Healthcare settings may not translate to the community, because these are different exposure types (compare e.g. intubating a coughing patient and standing in the subway). Then again, just to note, if an infection in the public space were unlikely then social distancing in supermarkets and shops would be nonsensical as well - which is not inconceivable!

We can conclude that given a sufficiently dangerous and infectious pandemic, it makes sense to employ highly plausible and cheap measures even if the evidence base is not perfect. Among these unproven measures are masks and also cough hygiene, hand washing, surface cleaning and social distancing between strangers in the public*. This conclusion further suggests increasing the capacity to produce masks should be one key political goal.
*in contrast to that, social distancing that reduces the number of intense, long-term contacts is known to be effective, or at least supported by stronger evidence, e.g. school and workplace closures

Cross-country data sets the stage
Many of you may have seen the below image. Can we trust this data? The answer is not a binary yes or no, instead “it’s not that simple”. Generally cross-country comparisons are considered to provide very weak but nonetheless real evidence (See “Guns & Homicides” under Notes). Hence we consider them to be hypothesis-generating. They tell us what is worthy of further study. Furthermore, a proper cross-country comparison would adjust for confounding to strengthen the data. For example, many people in Asian countries wear masks, but they are also located in warmer climates that may reduce viral spread and their governments have implemented other policies. Eventually when we have data from dozens or hundreds of countries we will be able to adjust for these factors and study different levels of mask implementation and other public health measures. Meanwhile we have to do something because there is no time for studies right now.


Source: Twitter @jperla (feel free to claim credit if you invented this meme), may I add that China does use masks, and that’s alright, because if you plot the data on a per capita basis, then China stills forms part of the distinct South East Asian cluster marked by low peak viral transmission. This graph is somewhat misleading for very large and very small countries (see below).

So to illustrate this clustering better, I have prepared a graph showing a selection of countries, I choose small, medium and large countries in the West to kind of sort of match the size of the countries from the SEAsia dataset (Austria, Switzerland, Sweden, Norway, Netherlands, Italy, Spain, UK, France, Germany and US vs Taiwan, Singapore, Hongkong, Japan, Korea and China). The world average is also indicated in the graphs. All data is per capita. If you’ve been doing too much science and wonder about the significance, yes it is, yes it is (by the way, the appropriate test here is the conservative Mann-Whitney-U-Test that does not assume a Gaussian distribution).

For now let’s do a preliminary and trivial ‘adjustment for confounding’ for lack of a better word. Is there any consistent similarity between Asian countries and regions that have low infection rates (China, Taiwan, Hongkong, Japan, South Korea, Singapore) that could explain the data better than mask use? Let us look at these measures: border closures, contact tracing, testing and social distancing. We will use various sources that I’ve put under References and Notes, for example Wang et al. (2020) details the situation in Taiwan. I have also tried to reconstruct timelines of different measures based on (news) reports although the work isn’t finished, so all I can discern so far are rough overall patterns.

As of late March, out of these countries only one has implemented severe social distancing that we have seen in Europe, which is China with the famous Hubei lockdown. However, we do have to remember that people in many South East Asian countries practice moderate social distancing voluntarily and this may have been crucial when implemented before or around the time of the first diagnosed cases (also remember that endemic transmission will be going on at least 1-2 weeks prior to the first documented endemic cases because of the incubation time and delays in diagnosing cases.)


Source: Twitter/ RT: Eric Topol/ Financial Times. Pay attention to the absence of lockdowns in South East Asia and the impressive differences in case numbers on a logarithmic scale(!) The difference is so gigantic that you cannot plot Asia, Europe and the World on a linear scale in the same graph.

Out of these countries, as far as I am aware, Taiwan, Singapore, Japan and Hongkong imposed early restrictions on Chinese arrivals (at that time no full border closures). In contrast, South Korea was criticized for lax border controls. This is irrelevant as all these countries have endemic transmission now. Border closures would have only delayed the inevitable. Even given established endemic transmission the growth rates of new cases are lower in South East Asia than elsewhere.

Many of these countriestest and trace”, but Japan does not. This is obvious if you look at the apparently high mortality rate in Japan (almost 3% despite being still early in the pandemic; as of late March) which suggests massive undertesting. However, even if you calculate a (naïve) mortality adjusted true case number then Japan and other Asian countries come out on top. Assuming a 1% true mortality rate Japan would have around 7000 cases, Italy about 1.5 Million (no this is not a typo) and Germany around 140 000 (5th of April). However, the legendary test and trace regime of South Korea may explain why they have already passed the peak of endemic transmission and Japan has not – although this remains speculation.

In contrast to popular opinion China is one of these successful countries and to see this you just need to compare their peak case numbers to Europe’s (which have yet to peak). Thus we can easily refute prevalent misconceptions like this one: “Sweeping mask recommendations—as many have proposed—will not reduce SARS-CoV-2 transmission, as evidenced by the widespread practice of wearing such masks in Hubei province, China, before and during its mass COVID-19 transmission experience earlier this year.”

Many people fail to understand the success of Mainland China because they compare the country’s response with countries that got prior notice. The PCR test for SARS-CoV2 was developed around the 13th of January. However, China reportedly had the first cases in November and endemic transmission by the end of December. Without a test and fighting an unknown disease, the country could have never followed a test and trace policy that is now widely implemented in both China and South East Asia. Given this disadvantage, the successful Chinese response is all the more surprising. 

Also consider these facts, not only was China the first country hit with absolutely no preparation, China’s GDP is about one third of Italy’s and Hubei is about the size of Italy with 50% higher population density. Please let’s not even get started on the population density of Wuhan vs Milan! Nevertheless, even Hubei never reached the same peak numbers as did Italy. If we’re making arguments against masks, at least we shouldn’t make disingenuous arguments that, if anything, support the opposite conclusion.


Perhaps a picture can better capture the level of population density (and mask compliance) in South East Asia. “Street in Hong Kong during the COVID-19 pandemic” (Source: Wikipedia)

I am not arguing that the data is correct or that it is strong. However, it is promising and it would be foolish to ignore it. The cross-country data consistently favors mask use.



Preclinical data and biological plausibility
Respiratory viruses can spread in three ways: respiratory aerosols and droplets or fomites (contaminated materials). Since aerosols are small enough to penetrate loose materials, theoretically surgical masks can help the most with the last two, by reducing the number of inhaled and exhaled droplets and also by reducing transfer of viral particles to your face through touching. Transmission through aerosols is considered as “airborne transmission” and for some pathogens there may be airborne transmission without aeorsols (MacIntyre and Chughtai 2015; Zhang et al. 2018).
We do not know yet how SARS-Cov2 spreads but at least a small element of airborne transmission is plausible, which would argue in favor of respirators.

Viruses are tiny so we would expect them to penetrate any surgical mask. However, droplets containing virus won’t. For those who have problems to imagine how that can work, I will provide an analogy. Imagine you stand in the bathroom behind your shower curtain or in your shower cabin. If you move the curtain or open the door you will immediately feel the temperature drop. Even though your shower curtain wasn’t forming a perfect seal it still took the heat, mostly stored in those invisible water droplets, some time to escape. In the same way there is nothing fundamentally impossible about a mask stopping viral droplets. Just enough to protect you.

More on physical probability
Let us look at exhalation first (source control). Some viral particles and even more droplets will be trapped in the mask due to simple adhesion, no matter how small they are. However, as some have argued the exhaled air will try to escape through the route of least resistance. With a mask that does not form a tight seal this should be around the side of your face/ears and above your nose/eye region. This by itself does NOT negate the benefits of the mask. By simple math we can see that we have split the airflow several-fold in different directions, reducing the number of viral particles that will hit others directly.

In addition one of these directions is never populated by people (region above your head), but theoretically exhaled droplets can and will diffuse to infect people around you. As is also obvious, the velocity of the exhaled droplets is greatly reduced because they take a detour and meet obstacles that cause turbulent airflow. This kind of deflection may be more relevant than filtration as suggested by ex vivo studies (Diaz and Smaldone 2010). No one quite understands the physics of this. There is a reason equations describing turbulent flow are part of the Millennium Prize Problems. Furthermore, when droplets exit the mask they will diffuse with slow velocity and in three dimensions, rather than being ejected from the mouth and nose, diffusing and travelling in the direction that you face in a cone shape (put another way, lateral dispersion through masks when coughing; Hui et al. 2012). In a situation where you speak with, or face, a stranger this can be a very relevant difference (think of shop clerks, subways, people asking for directions, etc.) Another complete wildcard is the milieu between the mask and skin, we’ll have air that is highly saturated with water so exhaled respiratory droplets should take up water and increase in size, which would change their physical properties. Generally, large droplets travel shorter which is beneficial.

The same basically applies in the opposite direction for inhalation. Just because droplets move around the mask does not mean they will remain equally infections. A droplet travelling around your eyes and nose (instead of through the mask) will be on a very turbulent path saturated with water and may well hit the mask or the inside of the mask, where it gets stuck.
  
The main point here is that the physics of mask usage are plausible, but sufficiently complex that we cannot use them to make a definite case for or against masks.

Laboratory studies
Here I will review a couple of experimental studies that try to determine the level of protection offered by masks. Furthermore, laboratory studies show that such an effect is highly plausible. It is evident that masks slow down the speed of exhaled air when coughing (Inouye et al. 2006). What about the particles that diffuse through and around the mask, how well are they filtered or stopped? The authors of this study (Milton et al. 2013) looked at people infected with influenza and measured the viral RNA in exhaled particles. They found that surgical masks reduced the RNA abundance in coarse particles >5um (micrometers) by 25x and by around x2.8 in small particles <5um. Combining the data they find in “..coarse and fine fractions, we detected viral RNA in 29 (78%) subjects when wearing facemasks and 35 (95%) when not wearing facemasks (McNemar's test p = 0.01). Surgical masks produced a 3.4 (95% CI 1.8 to 6.3) fold reduction in viral copies in exhaled breath.
Regarding methods, this study used a “specially designed aerosol sampler” that we will discuss below. Importantly, an earlier study using “deposition on petri dishes” to capture droplets also gave a encouraging results (Johnson et al. 2009).

Now on to the recently published paper (Leung et al. 2020). While yes, it is technically a randomized, controlled trial, it is not a clinical RCT. The study does not ask whether masks prevent viral infection. The study asks whether masks reduce the number of exhaled respiratory virus particles when you have a symptomatic infection. One strength of the study is its size and the fact that they looked at 3 different respiratory viruses. Exhaled breath was collected for 30minutes and people were allowed to cough naturally. As always viral particles were measured by PCR and the sampling is as above, <5um small droplet (here: aerosol) vs. >5um regular droplet (here: droplet). Cornoaviruses responsible for the cold were reduced in both exhaled droplets (p=.07) and aerosols (p=.02). Influenza viruses were only reduced in large droplets and there was no significant difference for rhinoviruses (also a common cold virus), although, eyeballing the data it looks promising. As you can imagine the statistics here are non-trivial because many patients never exhaled any viral particles during the study.

The machine used in the Leung and Milton study is called Gesundheit-II. My understanding is that the device captures most of the air you exhale (“collecting all exhaled breath”) hence it will also capture laterally dispersed droplets that would have never hit anyone. This, if anything, would make the above results more impressive! Image of the G-II and popular science explanation of the machine.

As we can see now, most of these studies generally concern “source control”. Exhalation of viral particles by symptomatically sick people. Participants that did not cough during the study also did not exhale measurable coronavirus particles (we’re talking here about the common coronavirus) and almost no influenza particles (Leung et al. 2020). However, there may be still some (mostly) asymptomatic or presymptomatic people that cough, not least because people cough for many reasons (e.g. allergies, airway irritation). This study still tells us very little about the efficacy of masks on the population level because “… exhaled breath collection was conducted for 30 min, this might imply that prolonged close contact [AND coughing] would be required for transmission” and we’re generally NOT wearing masks during work, school and when seeing friends.

I think the evidence that masks reduce the amount of - potentially infectious - particles exhaled by the source is convincing. While for viruses it is not clear if this translates into to a reduction in the infectiousness of the air and surroundings, there is at least one study for tuberculosis showing that during periods when the source wore a surgical mask the air in the medical ward was less infectious in a guinea pig model (Dharmadhikari et al. 2012). Tuberculosis is also a rather stringent model because these Mycobacteria (somewhere halfway in size between normal bacteria and viruses) are tiny and spread via aerosols.

It appears that laboratory and ex vivo studies are the origin of the myth that masks only work as source control. However, given the low viral shedding even by sick patients it would be technically challenging to determine if masks can protect you from someone else’s droplets. Absence of evidence is not proof of failure! Indeed most studies (and skeptics) incorrectly put prevention of aerosol transmission on some kind of pedestal. Given the above data I see no reason why masks wouldn’t mitigate droplet transmission during coughing. For example, van der Sande et al. (2008) found good inward protection from all types of masks whether cloth, surgical or respirator. Another study found that masks on the receiver actually increased the risk of aerosol inhalation vs just a mask on the source (Diaz et al. 2010). While a follow-up to this study showed that masks are most effective as source controls, a loose mask on the receiver is ineffective and a tight surgical mask has a small protective effect against aerosol inhalation (Mansour et al. 2013). This is a debate best left for another day, especially considering the highly artificial setups of these three studies.

Finally, while I do not agree with the conclusion of this article, it is only fair to consider some more arguments from the skeptics. The author points out that surgical masks and especially cloth masks show abysmal filter efficiencies (often 2-10%; but apparently a mean of 30-50% for surgical masks) for small particles and high flow rates. This usually means particle sizes <1um, which is considerably below 5um, an often used cutoff for the definition of droplets vs aerosols. This immediately raises an issue. Yes, it is possible that masks do not efficiently prevent aerosol transmission and this is consistent with the above studies. However, many viral diseases have a large droplet component!

Speculation about synergistic effects
As we will see below many studies show the combination of hand washing and mask wearing to be more effective than either alone. One way to explain this is through study power. Two effects that are real but not significant add up to a stronger (often marginally) significant effect. However, there could be also a biological reason. If you think about mask usage and generally all personal protective equipment (PPE), there is a lot of contamination going on when you take it off. When you take off gloves and a mask you’ll find some viral particles on your hands and perhaps your facial skin. If you touch your eyes or forget to wash your hands you’ll negate some of the benefit of the PPE. As absurd as it sounds, there may be other non-trivial effects, what if a mask increases the number of particles that stick to your face but reduces inhalation risk per se? A mask, contrast with no mask, gives droplets a lot of time to get stuck inside on your skin that is now extra wet. Phan et al. (2019), for example, find that respiratory viruses, at least in theory, can be found on hands and faces and below PPE of health care workers (HCWs).

Could masks benefit the whole society?
Assuming that masks work, even a small number of people wearing them would make a measurable contribution to reduced influenza burden (Tracht et al. 2010; Brienen et al. 2010; Yan et al. 2018). Others are similarly optimistic after measuring the filter properties of masks: “Any type of general mask use is likely to decrease viral exposure and infection risk on a population level, in spite of imperfect fit and imperfect adherence” (van der Sande 2008). In order to benefit society all masks need to do is reduce the R0 of a virus. They need not be perfect. If we push down the R0 then testing, quarantine, surveillance-contact-tracing and (mild) social distancing may be enough to stamp out the virus or limit its spread and save thousands of lives. In particular, masks could reduce the spread in crowded, public places where contact tracing would be insanely difficult. However, we should not expect too much from masks, because most viruses spread mainly through close contacts and people tend not to wear masks when they meet colleagues or friends.

Masks might also help to preserve our rights. South Korea has shown that such a regime works well while preserving civil liberties. It seems Europe has failed contact tracing and did little to reduce the R0 early on, so now we pay the price. Lock downs are a massive violation of civil liberties never seen in democracies before.

Do masks protect others or yourself?
Everyone repeats the claim that surgical masks are only good for source control (“Fremdschutz” in German), but it is difficult to find strong evidence for this assertion. Yes, surgical masks were designed to prevent the environment from the surgeon’s germs, but then again, by simple logic, what goes up must fall again and whatever leaves the lung also enters the lung by a similar route.

Indeed the German Robert Koch Institute (RKI) states that according to preclinical studies masks protect both the wearer and his environment. (The original is as follows: „Experimentelle Studien, z. B. an Kunstköpfen, lieferten…Hinweise, dass durch …[Masken]… sowohl ein besserer Schutz für Dritte…als auch für die tragende Person selbst erreicht werden kann…“)

Psychosocial factors and the adding up of unknowns
Pros
  • Less “face touching”
  • Direct protection (self)
  • Direct protection (others)
  • Anti-health halo effect (see below)

Cons
  • Health halo effect
  • Mask shortages for HCWs if everyone uses a mask
  • Improper use and handling
    • spread viral particles when taking off the mask
    • extended wear and reuse; mask will serve to spread pathogens
    • face touching if the mask feels uncomfortable

The health halo effect describes careless behavior that happens when the user feels “too safe” and offsets the benefit of an intervention e.g. someone may sneeze through a mask rather than into their elbow because they think the mask will stop viral particles. In contrast, I would argue that even a simple surgical mask is rather uncomfortable, so it discourages socializing and it also discourages speaking. That’d be an anti-health halo effect. Since people may fear that you are infected or unwilling to talk (because you wear a mask signaling “I am serious about disease prevention”), it might also serve to discourage social interaction with strangers through this way (see also Chan & Yuen 2020).

Similarly, a mask could promote face touching when it is too close to your eyes or when it is too loose and you have to adjust it or it can discourage touching your nose and mouth because it is a physical barrier. We can see that all these effects are unpredictable and hence we’d like to have studies that directly test whether masks work in the real world and not just in theory.

Observational data

Jefferson et al. (2009) combined six case control studies that showed, both alone and in combination, handwashing, masks, N95 masks, gloves and gowns to be significantly protective against SARS - a relative of SARS-CoV2. All or most of these studies were performed in health care workers (HCWs) and extrapolation to community acquisition is problematic. Their review of RCTs at that time only identified two studies and apparently most trials on mask use were published after 2008 when the pandemic flu struck (these will be reviewed below).

Bin-Reza et al. (2012) also finds the same when reviewing the available observational data. While they are supportive, unfortunately, these studies “were poorly designed, had many weaknesses and so were very difficult to interpret”. They go on to identify one of the better case-control studies (Lau et al. 2004) claiming the following: “The single case–control study that tried to address some of these limitations did not find that inconsistent use of masks or respirators was associated with SARS infection”. Interestingly the paper states that masks were basically always used “Almost 100% of the study respondents used either an N95 mask or surgical mask” and that “Having three or more personal protection equipment inconsistently used (including masks) was also a significant predictor of SARS infection for hospital workers in direct contact with SARS patients (OR = 7.84, p = 0.003); for those with direct contact with patients in general (OR = 10.83, p = 0.0007); and for those with no patient contact (OR = 3.4, p = 0.006) (Table1).”

A couple years later Offeddu et al. (2017) finds that both masks and respirators protect HCWs from SARS when analyzing observational studies. This is unsurprising because all the above studies are basically looking at the same SARS data. Some of the observational data suggested that respirators are superior to masks, but this is not obvious from the Offeddu data. Additionally, a recent review has brought to my attention the fact that there are actually observational studies suggesting that mask use is protective from SARS in the community setting, albeit there are only two such studies (Wu et al. 2004, Lau et al. 2004).

Similarly, a small study about MERS, a relative of SARS-CoV2, suggests that masks protect health care workers (Ki et al. 2019). In a retrospective study by Alraddadi et al. (2016) use of either a respirator or surgical mask was protective from MERS. Unpublished data on medarxiv paints basically the same picture for SARS-CoV2: “Of note, 10 of 213 medical staffs with no mask were infected by COVID-19 while 0 of 278 wearing N95 respirators was infected (6). Interestingly, a higher risk of infection was noticed in male professionals. This study called for the essential role of occupational protection (6).”

What do observational studies say about other diseases than SARS? This question is very important because SARS-CoV2 is unlike SARS which was “was less infectious than many other respiratory infections and was mostly nosocomial.” (MacIntyre and Chughtai 2015; TB: Schmidt et al. 2018) Respirators are used to protect HCWs from tuberculosis but, for obvious reasons, these studies generally look at combined measures. One study suggested reduced transmission of pertussis with face masks and a small study found no difference in RSV transmission (as per MacIntyre and Chughtai 2015; however, another review was a bit less pessimistic about RSV saying "eye protection appeared more effective than gowns and masks" French et al. 2016).

There is some indirect evidence from this influenza season in Hong Kong showing it to be unusually short, which could be due to a COVID-19 related surge in mask use and associated behaviour (Chan & Yuen 2020). The same may have happened in the past. One observational study in Hongkong (Lo et al. 2005) tried to quantify whether combined population level measures, implemented after/during a SARS outbreak, affected the prevalence of respiratory infection (fraction of positive tests for Influenza, RSV, Parainfluenza and Adenovirus). While flawed this is perhaps as good as it gets:
“Surveys conducted in April and May 2003 showed that most of the population wore a face mask (76%), washed their hands after contact with potentially contaminated objects (65%), used soap when washing hands (75%), covered their mouths when sneezing or coughing (78%), and used diluted bleach for household cleaning (>50%)”


Figure from: Lo et al. 2005. Hepatitis B is the negative control. It is not entirely clear, though, why the fraction of positive tests should decrease. The patients were presumably symptomatic when the test was ordered? What did they have? Did doctors order tests at a lower symptom threshold thus biasing the data? The authors note a surge in testing only during April, thus they think they can rule this out. Since the last SARS case was in early June they also think this can explain some level of "regression to the mean" seen in the second half of the year when people eased off the self-imposed measures.

There are also several studies that were published after the review by MacIntyre and Chughtai (2015) or were not mentioned by the authors. I present a couple with no guarantee of completeness. Ambrosch and Rockmann (2016) found that continuous use of surgical masks by staff halved the number of influenza-like illness in hospitalized elderly patients and Sokol et al. (2016) found a five-fold reduction of respiratory viral infections in a bone marrow transplant unit after mask use was implemented. Sung et al. (2016) also report a decrease in respiratory viral infections in their transplant unit. Finally, Thai Yeo et al. (2016) write in a short letter that universal mask policies also decreased the incidence of viral infections in the neonatal unit and that “taking our findings together with those of Sung et al [1], we believe that universal masking in high-risk closed settings…should be considered to reduce the risk of nosocomial RVI [hospital acquired viral infections]”

Furthermore, Uchida et al. 2016 found in analysis of Japanese elementary school students that mask wearing was equally effective to vaccination for the prevention influenza infection and handwashing had no impact on influenza risk. The study was relatively large as the authors could gather data from over 10 000 pupils and influenza was determined via rapid diagnosis kit. Presumably if the result was due to “healthy user” bias we would have expected a positive result for hand washing as well. Somewhat along the same lines,  study found that face masks might increase the risk for influenza in Korean school children while continuous use of face masks was protective (Kim et al. 2012). The school data contrasts with a systematic review of influenza prevention in long-term care facilities (Rainwater‐Lovett et al. 2013), the authors find that the use of “personal protective equipment” did not reduce the severity of outbreaks. However, neither did social distancing and the effects appear to be roughly positive, albeit non-significant.

Similarly, Hassan Emamian et al. (2013) found no effect of masks on respiratory tract infections in a nested case control study, if anything masks seemed detrimental. Some studies on influenza were negative (avian influenza A transmission, H7N7: te Beest 2010), while at least one study suggested that masks protect from hospital acquired swine flu (A/H1N1; Cheng et al. 2010). After a bit of digging I found that Offeddu et al. (2017) provides a summary of all observational studies in HCWs, while a mixed bag, they do generally favor the broad category of “any PPE” (i.e. masks or respirators) and the idea that they protect from influenza.



Figure from Rainwater‐Lovett et al. 2013; NPI = PPE = masks + addons. What can protect from influenza outbreaks in long-term care facilities?

Another question that may be of interest is whether masks can reduce infection rates during longer trips or during crowded events. A small case-control study by Zhang et al. (2013) suggest that masks protect strongly from acquiring influenza during a flight (H1N1pdm09). A study of military personnel also finds that combined measures (“eg, enhanced surveillance with isolation, segregation, personal protective equipment”) reduce rates of influenza infection in soldiers that are often confined in close space (Lee et al. 2010). Barasheed et al. (2016) performed a systematic review and found that face masks were protective against respiratory infections when used at “mass gatherings” (RR=0.89, p<0.01, n=13).

The issue of bias is quite significant for all of epidemiology. For example, Wada et al. (2012) note “…that wearing a face mask in public may be associated with other personal hygiene practices and health behaviors among Japanese adults. Rather than preventing influenza itself, face mask use might instead be a marker of additional, positive hygiene practices and other favorable health behaviors in the same individuals.” Similarly, we can imagine that case-control studies that ask health care workers to remember “PPE failure”, like not wearing a mask, could be subject to recall bias. If people assume that mask wearing is protective, they may be more likely to remember a failure if they got sick later. We cannot even rule out a simple post hoc ergo propter hoc conclusion! This is why we need controlled trials.

Clinical data
We will discuss two key papers here. One is a meta-analysis of mask use RCTs to prevent influenza infection and another one is the comparison between masks and respirators for influenza/other infections in HCWs.

Xiao et al. (2020) found that neither hand washing nor mask use led to a significant reduction of confirmed influenza infection. For example, with masks and handwashing combined “there was a nonsignificant RR reduction of 22% (RR 0.78, 95% CI 0.51–1.20; I2 = 30%, p = 0.25)” in lab confirmed viral infections while it was even lower for the mask only group (this is based on 10 studies). Below we will discuss this meta-analysis in more detail and why an earlier meta-analysis apparently arrived at the opposite conclusion -- finding that masks may be beneficial (for reasons that will become obvious I put the data in the chapter “A second look at the good studies and further meta-analyses”; Wong et al. 2014). You can also see MacIntyre and Chughtai (2015) who provide a narrative review of most of these studies.

Long et al. (2020), like most meta-analyses before, failed to find a difference between face masks and respirators when it comes to the protection of health care workers (HCWs). There was no difference in the prevention of laboratory-confirmed influenza or respiratory viral infections and influenza like illness (ILI).  The underlying and unspoken assumption of that paper is that respirators protect HCWs and that they have a small benefit over masks that is not visible in their data. Now from this we can conclude that face masks also help average Joe via two (admittedly very tricky) extrapolation steps. 1) Respirators protect HCWs, 2) The exposure types and pathogens are similar enough so that the HCW data translates to the community data.

My opinion, and this is of course purely speculation, is that we can safely extrapolate the data. If anything, I would expect masks to be more effective and not les effective in the community setting. First of all, HCW studies only measure protection of the mask-wearer, whereas in the community masks are also worn by (a)symptomatic citizens to protect others (source control, German: Fremdschutz). In addition, community acquired infections may spread via moderate exposure, while HCWs will be overwhelmed regardless of protective equipment. (The opposite could be true as well: community acquisition is so rare that masks do not help but then neither would social distancing at the supermarket!) Although extrapolation is weak evidence, it cannot be ignored.

In fact, to be fair, the whole idea of “masks as source control” and “masks for the protection of HCWs” remains controversial. “Masks are also used to prevent surgical site infections in the [operating room] although most studies failed to show any efficacy against this indication…Only one clinical trial reported high infection rates after surgery if masks were not used by the surgeon in the [operating room]”. However, there is probably more to the story than meets the eye (as for example, Herron et al. (2019) found that only 18% of doctors were fully compliant with CDC recommendations on how to wear a mask).

In addition, some data “suggests that targeted masks and respirators are equally inefficacious (rather than equally efficacious)” for the prevention of influenza in health care workers (as reviewed by MacIntyre and Chughtai 2015). The closest we have come to an RCT with a real “no mask” control is a trial that compared regular masks with cloth masks and found that rates of ILI and lab confirmed viral infection were significantly higher with cloth masks (n~1400, MacIntyre et al. 2015). Finally, when I had almost finished preparation of this post, another study came to my attention. This suggests that my review is only scratching the surface! MacIntyre et al. (2017) performed a study in HCWs (or rather the re-analysis of two Beijing RCTs) that had a control arm and found basically a graded response for the prevention of influenza (control < surgical mask < respirator).

The extrapolation problem

Now let us look broader at the “extrapolation problem” as I will call it. Many people argue that we cannot extrapolate HCW data to normal people. If we grant this argument, we also have to pay attention to other issues of extrapolation. In fact, we cannot extrapolate from household to public mask wearing. A household contact is much closer to the HCW setting because you are continuously exposed to a very high virus titre. It may be that the viral load, in most cases, is several times higher than needed for an infection. In this case a reduction of inhaled viral particles by, let’s say, 50 to 80% would have no meaningful impact on infection rates. Contrast this to public mask wearing, where exposure is brief and less intense (e.g. in a restaurant, subway, supermarket, crowded street), hence viral titres may be closer to the minimum required (say, 100 particles are required for infection, than a 50% reduction from 1000 to 500 makes no sense but going from 150 to 75 may be useful).
If we accept this criticism we have to conclude that we have “no evidence of absence”; i.e. there are simply no properly designed RCTs studying mask use or hand hygiene for that matter. We could call it a day here, but let’s dig deeper to see if we can salvage this.

Study quality favors the null hypothesis: meta-analysis of Xiao et al.
Usually low quality RCTs tend to favor the interventional arm but below we will discuss why the opposite is the case for mask studies.

Quality of the data. To implement mask wearing within 2 days from symptom onset is just laughable. It would be a miracle if we saw a benefit in those studies. This is also not at all comparable to mask wearing in public which is done by asymptomatic individuals, pre-exposure, and hence always works from “day 0”. Since I can understand the rationale to perform these studies I would call them “heroic” rather than “junk science”. While they are definitely junk, there was still a nonzero chance of success and desperation requires heroic measures, doesn’t it?

We know that the incubation time of influenza is around 1 to 4 days (NB: some of these patients may not even had influenza if the inclusion criteria are ILI and not influenza Dx). As discussed above, the viral titres would be tremendous, but there is another issue that comes to mind. Viral transmission is the highest in the first couple of days and the mask RCTs generally recruit too late. Hence, all susceptible individuals may already have been infected at this stage and those who were able to mount an immune response, e.g. because they were infected with a similar strain before, would never have been infected with or without a mask. Even if you try to adjust for these issues, I don’t think it is possible. Let’s say you exclude the early infections in your analysis, then you may be only left with elite responders who had an immune response anyway and don’t need masks. Even if feasible, what does this do to your statistical power since most infections would have occurred early on?

Either way, among studies that adjust for this gigantic flaw of pre-exposure infection in their post-hoc analysis, most many find a preventative benefit of mask wearing (MacIntyre 2009; Suess 2012; Cowling 2009). Various fancy re-analyses can performed using advanced statistics but either way a “pattern of increasing efficacy with reducing delay was [seen]” (Lau et al. 2015 reanalyzing the Cowling data). Surprisingly no one went back to pool and re-analyzed the data for "early masking" intervention.

Compliance is another big issue that favors the null hypothesis in Xiao et al. (2020). In particular in household settings it is difficult to wear a mask at most times during the day (almost 16hrs / whenever in contact with the index case!) and virtually impossible to do this at night when sleeping in the same room as your sick spouse or child. In public it would be much easier to comply with mask use, because you only have to wear it when you are in a crowded place. (Of course, if you don’t wear it at school or work it won’t protect you or anyone else. However, this can be the difference between fully shutting down public life like in Germany and mere social distancing like in South Korea.)

Sample size is also a problem with these studies. Looking at the meta-analysis by Xiao et al. (2020) we have only 51 cases of PCR-confirmed influenza in the control group (in the mask only category) and 161 in the control group for mask+hand hygiene. We’re also ignoring the fact that these are cluster randomized RCTs so the number of independent datapoints is even lower (household data is correlated). Given these small sample sizes, it does not surprise that a single study may account for 30 to 40% of the data. Unfortunately, this happens to be one of the weaker studies with very delayed usage of masks and children who slept in the same room as their parents (Simmerman et al. 2011). I am rather surprised that even though data on influenza like-illness (ILI) and upper respiratory infections (URI) was provided by several studies, these outcomes were not analyzed by Xiao et al. This is unfortunate since these outcomes are considerably more common (caused by flu + cold viruses), although sometimes subject to bias when ILI is self-reported.

A second look at the good studies and further meta-analyses
Only 3 studies included by Xiao et al. (2020) implemented mask use before symptom onset which is consistent with current recommendations (Larson et al. 2010, Aiello et al. 2012, Aiello et al. 2010) and two of these studies report on the same cohort only during different years (Aiello et al. 2012, Aiello et al. 2010). Larson et al. found a significant reduction in secondary infections* with combined mask use + hand washing and their data is incorrectly reported in this meta-analysis as a null finding. Only the work by Aiello studied masks alone, masks + hygiene and controls. Masks + hygiene appeared to be superior to mask and control. In fact, we can see that efficacy, for laboratory confirmed influenza, is very roughly mask+hygiene > mask > control when you look at Table 2 in the meta-analysis and the primary data, but none of these findings were significant. Even so Xiao et al. attempted to pool these two studies, finding that “A pooled analysis of 2 studies in university residential halls reported a marginally significant protective effect of a combination of hand hygiene plus face masks worn by all residents (RR 0.48, 95% CI 0.21–1.08; I2 = 0%, p = 0.08)”. If you ask me, all three studies in fact are tiny, flawed and underpowered. The Larson study in particular, while it did have early mask intervention, suffers from other issues.

There is another meta-analysis that I ignored at first because of the title. The work by Wong et al. (2014) that focuses on hand hygiene but has an overlapping category with Xiao et al. (2020). In fact, the authors found that hand hygiene alone had no significant impact on influenza prevention, whereas hand hygiene + mask use did. As you may remember Xiao et al. also studied this latter category. So what is going on? Xiao includes 6 studies in their analysis (Aiello 2010, 2012, Cowling 2009, Larson 2010, Simmerman 2011, Suess 2012) and Wong includes 5 (Aiello 2010, 2012, Cowling 2009, Larson 2010, Suess 2012). The study that is missing was the largest and worst designed. Or rather than worst designed, it was simply hampered by the realities. Not only does Wong exclude this study, in contrast to Xiao, they also look at (sometimes self-reported) ILI on top of lab confirmed influenza. This gives more events and more statistical power. Either way, Wong et al. find the same using both outcomes.

Similarly, when we synthesize different types of (heterogenous evidence) we do see that: “If the randomized control trial and cohort study were pooled with the case–control studies, heterogeneity decreased and a significant protective effect was found (not illustrated: N = 1736; OR = 0.41 95% CI 0.18–0.92; I2 = 35%)” showing that facemasks reduce transmission of pandemic influenza (Saunders-Hastings et al. 2017).

*“Regarding URI, ILI, and influenza episodes, there was a significant decrease in secondary attack rates in the Hand Sanitizer and Face Mask group when compared with the Education group. [since in the Larson study households were only advised to wear masks after someone got infected, the raw rates of infections in each treatment group are meaningless and what matters are only the secondary infections. This contrasts with the Aiello work where students living in a dorm were advised to continuously wear a mask]”

Consistent messaging and biological plausibility
How can we recommend hand hygiene and respiratory etiquette purely based on biological plausibility if they are not backed by solid interventional studies (Xiao et al. 2020; Wong et al. 2014), but then fail to recommend mask wearing using the same criteria? Why do we recommend respirators for HCWs to protect themselves from SARS-CoV2 and other viral infections if there is no convincing evidence that they work better than masks? (Long et al. 2020) It’s all based on plausibility and weaker studies, yet somehow for mask wearing we want to use a higher standard of evidence. The most parsimonious explanation may be a combination of politics and second guessing what people want, because masks are difficult to wear (compliance issues), unpopular in normal times and in high demand right now. But please someone tell me ridiculous statements like this by the anti-mask crowd aren’t political theatre: “If masks had been the solution in Asia, shouldn't they have stopped the pandemic before it spread elsewhere?”. If we are going by that dumb logic, then, yes, masks are the solution because South East Asia is outperforming Europe – not only by a little – SEAsia is outperforming Europe by at least an order of magnitude when it comes to controlling the spread of COVID-19.

Another issue is if we provide polarizing recommendations, changing them to reflect the current situation will be seen as flip-flopping and erode public trusts. To no one’s surprise, the CDC now recommends masks (current as of early April), although, for availability reasons they recommend cloth masks.

The cost of masks, once the world market can produce them, is similarly low. If the argument is cost I don’t buy it. Either way it would behoove us to be honest with our messaging:
“Although, when looking at the most rigorous studies, we have no consistent data in favor or against the following ideas, nonetheless we think that people should wash their hands and practice good respiratory etiquette. We base this reasoning on very strong preliminary data. We also think that mask wearing may be similarly helpful but we fear that we could run out of masks for at risk populations and health care workers if everyone starts to go out wearing masks. Thus it would be wonderful if people simply stayed at home to reduce the burden.
Finally, perhaps there will be a reckoning and a time when we have to apologize for our incompetence and arrogance in the West. At this moment it appears that masks, swift responses with modest social distancing, and intense contact tracing may explain how South East Asia saved thousands of lives -- and how we failed.

Geographic note
Before I get tarred and feathered for my geography skills. The South East Asian elite cluster I defined is not geographically part of South East Asia, although it is located in East and South (East) Asia. I am thinking about different terminology for the future to make this less confusing.

References and Notes (selection, most missing references can be easily googled though!)
Surgeon General Urges the Public to Stop Buying Face Masks (29th February, 2020)

Main focus for preventing coronavirus spread should be hand hygiene, not face masks (February 27th, 2020)

China’s response:
(for reference new cases in Hubei peaked around the 4th of February)
Reconstructing China’s response is a project in and of itself for several reasons. China was struck early so they could not have taken the same measures as countries and regions that had time to prepare (Taiwan, Singapore, Hongkong, Korea, Japan). The country is huge with various policies. The data is also less reliable because China wages a propaganda war, particularly against the US, but the opposite is also true; the US is interested in making China look bad. We can expect the data for the other SEAsian countries to be more reliable. Relevant for our analysis: At first glance it seems that at least half a billion of people in China managed to stave off coronavirus without lockdowns. It is, however, not clear how many communities with endemic transmission are in this sample and what measures they implemented.

4th February: this article mentions tens of millions, especially in and around Hubei, as being quarantined or on lockdown
5th February: “Excluding municipalities in Hubei province, the center of the epidemic, 53 cities had imposed restrictions on movements of the public as of the late afternoon of Feb. 5, according to a tally by The Asahi Shimbun.” (**)
14th February: “Around 500 million people in China are currently affected by policies put in place restricting movement…not every city or province is facing Wuhan-like restrictions” (*)
Holiday extensions and postponement of school
Early April: China still has small numbers of endemic transmission. I think it is fair to look at this period to consider “how would a prepared China have dealt with COVID-19 if they had these tools 4 months ago”. It seems they are controlling endemic spread with the same measures as other SEAsian countries. There is also an element of targeted lockdowns (in contrast to Europe in the early stages, where we waited for the disease to get out of control) that makes for a nice game of viral whack-a-mole. Jia county in Henan province is on lockdown again.

This Worthing Herald article makes it sound like only few provinces were on lockdown and the Wikipedia article also only lists around 200-300 million, although both don’t seem quite reliable.


Hongkong’s response:
25th of January: first closures of large events (e.g. amusement parks)
28th of January: sizable and indiscriminate restrictions on travel from the whole of China, although, travel by bus and planes was still possible. Public employees suggested to work from home. (*)
29th January: closures of public venues like museums
3rd February: kindergarten and schools closed, expected to last until at least 20th of April. Churches and restaurants, reportedly, remain open. ***
March 25th: border closed to non incoming residents and strict quarantine measures implemented for everyone else**
March 27th: Life continues as normal, people do practice voluntary social distancing. People “went to restaurants, bars, Zara and Topshop, and caught public transport”. “Hong Kong kept an epidemic at bay by effectively identifying cases and isolating them, and with broad community support for social distancing, Cowling told the Guardian”
Early April: closures of bars, karaoke venues, etc no closure of other venues, this contrasts with more sweeping restrictions in Europe.


Japan’s response:
As per Wikipedia and the below sources:
February 1st, daily cases still at 0:
refuse entry to high risk individuals (from Hubei) (*)
February 25th, daily cases in this time range between 0 and 42:
large gatherings cancelled and schools closed, voluntary social distancing, support for telecommuting and people with cold symptoms urged to stay at home, ramp up production of masks and disinfectant, reduce community testing in favor of testing patients with pneumonia, "Maintain current entry restrictions and the recommendation for travel suspension to prevent a sudden influx of infected persons into Japan" (**)

** "Basic Policies for Novel Coronavirus Disease Control February 25, 2020" https://www.mhlw.go.jp/content/10200000/000603611.pdf
“Coronavirus in Japan: why is the infection rate relatively low?” (March 17, 2020 1.16pm GMT)
Note: the article mentions the low testing rate which is indeed abysmal, but generally comes off as typical Western apologism trying to predict doom for another South East Asian country (see similar articles on the “failures” of South Korea)

Taiwan’s response:
As per Wiki and the below sources:
January 5th: Monitoring of Wuhan arrivals (apparently no ban yet, see sources below) 
January 25th (daily cases still around 0): Hubei listed with level 3 travel alert; not clear if quarantine measures already in place? (**)
February 2nd: School season postponed, increase production of masks and disinfectant (*)
February 28th (peak daily cases of 5): Italy gets level 3 travel alert; quarantine measures implemented (***)

„Persons with low risk (no travel to level 3 alert areas)were sent a health declaration border pass via SMS (short message service) messaging to their phones for faster immigration clearance; those with higher risk (recent travel to level 3 alert areas) were quarantined at home and tracked through their mobile phone to ensure that they remained at home during the incubation period.” [right now I cannot figure out WHEN this came into effect]
Wang, C. Jason, Chun Y. Ng, and Robert H. Brook. "Response to COVID-19 in Taiwan: big data analytics, new technology, and proactive testing." JAMA (2020).


South Korea’s response:
Lax border controls and freedom of travel were considered “mishandling” of the crisis (28th of February).

One of the most elaborate contact tracing regimes, reportedly, using CCTV cameras, cell phones and credits to track case movements and publishing their routes. Not clear if this is done routinely or only some of the time.
“The aggressive efforts by Hong Kong, Japan, Singapore and South Korea to investigate and isolate every possible infection is exactly what the World Health Organization has been calling for since January.”
“Contact tracing efforts were concentrated on the Shincheonji cluster”

Singapore’s response:
One of the most elaborate contact tracing regimes in the world:
“Its early success in keeping numbers low was due to a combination of medical surveillance, including extensive testing, comprehensive contact tracing and isolation, and risk-based community measures, according to Dr Vernon Lee, director of communicable diseases at the Singapore Ministry of Health.”

“As of March 5th, 2020, there have been 117 cases, of which 25 were imported. By devoting considerable resources including police investigation, 75 of the 92 cases of local transmission have been traced back to their presumed exposure, either to a known case or to a location linked to spread”
Ferretti, L., Wymant, C., Kendall, M., Zhao, L., Nurtay, A., Bonsall, D. G., & Fraser, C. (2020). Quantifying dynamics of SARS-CoV-2 transmission suggests that epidemic control and avoidance is feasible through instantaneous digital contact tracing. medRxiv.

“About 1 million people have downloaded TraceTogether app, but more need to do so for it to be effective: Lawrence Wong”

Late March: shut down of bars but “McDonald’s was crowded with schoolchildren studying and playing with their phones”
As of late March 22: tighter border controls, an increase in social distancing (events >250 people banned; increasing distance in super markets, etc), no lockdown.
“There is no full-scale lockdown. Schools and pre-schools are scheduled to reopen after a one week holiday with stricter measures in place to limit activities and keep students apart, including assigned seating at lunch.” (*)


Guns and Homicides:
Regarding the use of cross-country data in science, it is not unusual to compare countries to derive or validate hypotheses in the social sciences.
“We analyzed the relationship between homicide and gun availability using data from 26 developed countries from the early 1990s.  We found that across developed countries, where guns are more available, there are more homicides.  These results often hold even when the United States is excluded.”
Hemenway, David; Miller, Matthew.  Firearm availability and homicide rates across 26 high income countries.  Journal of Trauma.  2000; 49:985-88.

########################
“Why the U.S. is changing its mind on coronavirus face masks” (April 3, 2020 12:12 PM EST)

RKI recommendations
“Nach Sichtung der Literatur gibt es insgesamt wenige aussagekräftige Studien aus dem Krankenhausbereich zur Effektivität des Tragens von MNS bzw. Atemschutzmasken zur Verhinderung einer Influenzainfektion. Die vorliegenden Studien zeigen die Effektivität des Tragens von Masken im Allgemeinen im Vergleich zum Nichttragen von Masken. Es gibt (wenige) Hinweise für eine Überlegenheit von FFP2 gegenüber dem MNS.“

Xiao, Jingyi, et al. "Nonpharmaceutical Measures for Pandemic Influenza in Nonhealthcare Settings-Personal Protective and Environmental Measures." Emerging infectious diseases 26.5 (2020).

Bin-Reza, F.; Lopez Chavarrias, V.; Nicoll, A.; Chamberland, M.E. The use of masks and respirators to prevent transmission of influenza: a systematic review of the scientific evidence. Influenza Other Respir. Viruses, 2012, 6(4), 257-67.

Long, Y., Hu, T., Liu, L., Chen, R., Guo, Q., Yang, L., ... & Du, L. (2020). Effectiveness of N95 respirators versus surgical masks against influenza: A systematic review and meta‐analysis. Journal of Evidence‐Based Medicine.

Lo, J. Y., Tsang, T. H., Leung, Y. H., Yeung, E. Y., Wu, T., & Lim, W. W. (2005). Respiratory infections during SARS outbreak, Hong Kong, 2003. Emerging infectious diseases, 11(11), 1738.
Physical interventions to interrupt or reduce the spread of respiratory viruses: systematic review.

Jefferson T, Del Mar C, Dooley L, Ferroni E, Al-Ansary LA, Bawazeer GA, van Driel ML, Foxlee R, Rivetti A.
BMJ. 2009 Sep 21;339:b3675. doi: 10.1136/bmj.b3675. Review.

Lau, Joseph TF, et al. "SARS transmission among hospital workers in Hong Kong." Emerging infectious diseases 10.2 (2 van der Sande, M., Teunis, P., & Sabel, R. (2008). Professional and home-made face masks reduce exposure to respiratory infections among the general population. PLoS One, 3(7). 004): 280.

Tracht, Samantha M., Sara Y. Del Valle, and James M. Hyman. "Mathematical modeling of the effectiveness of facemasks in reducing the spread of novel influenza A (H1N1)." PloS one 5.2 (2010).

Brienen, N. C., Timen, A., Wallinga, J., Van Steenbergen, J. E., & Teunis, P. F. (2010). The effect of mask use on the spread of influenza during a pandemic. Risk Analysis: An International Journal, 30(8), 1210-1218.

Zhang, Yun-Hui, et al. "Role of viral bioaerosols in nosocomial infections and measures for prevention and control." Journal of Aerosol Science 117 (2018): 200-211.
[general interest review]

Zhang, L., Peng, Z., Ou, J., Zeng, G., Fontaine, R. E., Liu, M., ... & Chuang, S. K. (2013). Protection by face masks against influenza A (H1N1) pdm09 virus on trans-Pacific passenger aircraft, 2009. Emerging infectious diseases, 19(9), 1403.

Sokol, K. A., De la Vega‐Diaz, I., Edmondson‐Martin, K., Kim, S., Tindle, S., Wallach, F., & Steinberg, A. (2016). Masks for prevention of respiratory viruses on the BMT unit: results of a quality initiative. Transplant Infectious Disease, 18(6), 965-967.

Wong, Valerie WY, Benjamin J. Cowling, and Alison E. Aiello. "Hand hygiene and risk of influenza virus infections in the community: a systematic review and meta-analysis." Epidemiology & Infection 142.5 (2014): 922-932.

Inferring influenza dynamics and control in households. Lau MS, Cowling BJ, Cook AR, Riley S. Proc Natl Acad Sci U S A. 2015 Jul 21;112(29):9094-9. doi: 10.1073/pnas.1423339112. Epub 2015 Jul 6.

Lau J, Tsui H, Lau M, Yang X. SARS transmission, risk factors and prevention in Hong Kong. Emerg Infect Dis 2004;10:587–92

Wu J, Xu F, Zhou W et al.  Risk factors for SARS among persons without known contact with SARS patients, Beijing, China. Emerg Infect Dis 2004;10:210–16.

Kim, C. O., Nam, C. M., Lee, D. C., Chang, J., & Lee, J. W. (2012). Is abdominal obesity associated with the 2009 influenza A (H1N1) pandemic in Korean school‐aged children?. Influenza and other respiratory viruses, 6(5), 313-317.

Johnson DF, Druce JD, Birch C, Grayson ML. A quantitative assessment of the efficacy of surgical and N95 masks to filter influenza virus in patients with acute influenza infection. Clin Infect Dis 2009;49:275-7.

van der Sande M, Teunis P, Sabel R: Professional and home-made face masks reduce exposure to respiratory infections among the general population. PLoS ONE 2008;3(7):e2618

More on Asian measures:

Pop sic articles:

Masks as Source control (Masken als Fremdschutz)

MacIntyre CR, Chughtai AA. Facemasks for the prevention of infection in healthcare and community settings. BMJ 2015;350:h694.

Cheng, V. C., Tai, J. W., Wong, L. M. W., Chan, J. F., Li, I. W., To, K. K. W., ... & Yuen, K. Y. (2010). Prevention of nosocomial transmission of swine-origin pandemic influenza virus A/H1N1 by infection control bundle. Journal of Hospital Infection, 74(3), 271-277.

PLoS One. 2010 Nov 17;5(11):e13998. doi: 10.1371/journal.pone.0013998.
Surgical mask to prevent influenza transmission in households: a cluster randomized trial.
Canini L1, Andréoletti L, Ferrari P, D'Angelo R, Blanchon T, Lemaitre M, Filleul L, Ferry JP, Desmaizieres M, Smadja S, Valleron AJ, Carrat F.

MacIntyre, C. Raina, et al. "A cluster randomised trial of cloth masks compared with medical masks in healthcare workers." BMJ open 5.4 (2015): e006577.

Emerg Infect Dis. 2020 May 17;26(5). doi: 10.3201/eid2605.190994. [Epub ahead of print]Nonpharmaceutical Measures for Pandemic Influenza in Nonhealthcare Settings-Personal Protective and Environmental Measures. Xiao et al. 2020

Ki, Hyun Kyun, et al. "Risk of transmission via medical employees and importance of routine infection-prevention policy in a nosocomial outbreak of Middle East respiratory syndrome (MERS): a descriptive analysis from a tertiary care hospital in South Korea." BMC pulmonary medicine 19.1 (2019): 190.

Influenza Other Respir Viruses. 2020 Apr 4. doi: 10.1111/irv.12745. [Epub ahead of print] Medical Masks vs N95 Respirators for Preventing COVID-19 in Health Care Workers A Systematic Review and Meta-Analysis of Randomized Trials. Bartoszko et al. 2020

X. Wang, Z. Pan, Z. Cheng, Association between 2019-nCoV transmission and N95 respirator use. medRxiv, 2020.2002.2018.20021881 (2020)10.1101/2020.02.18.20021881).

Hui, D. S., Chow, B. K., Chu, L., Ng, S. S., Lee, N., Gin, T., & Chan, M. T. (2012). Exhaled air dispersion during coughing with and without wearing a surgical or N95 mask. PloS one, 7(12).

Inouye, Sakae, Yasuaki Matsudaira, and Yoshibumi Sugihara. "Masks for influenza patients: Measurement of airflow from the mouth." Japanese journal of infectious diseases 59.3 (2006): 179.

Leung, N. H., Chu, D. K., Shiu, E. Y., Chan, K. H., McDevitt, J. J., Hau, B. J., ... & Seto, W. H. Respiratory Virus Shedding in Exhaled Breath and Efficacy of Face Masks. 2020.

Milton DK, Fabian MP, Cowling BJ, et al. Influenza virus aerosols in human exhaled breath: particle size, culturability, and effect of surgical masks. PLoS Pathog 2013;9:e1003205.

Chan, Ka Hung, and Kwok-Yung Yuen. "COVID-19 epidemic: disentangling the re-emerging controversy about medical facemasks from an epidemiological perspective." International Journal of Epidemiology (2020).

Sung, A. D., Sung, J. A., Thomas, S., Hyslop, T., Gasparetto, C., Long, G., ... & Chao, N. J. (2016). Universal mask usage for reduction of respiratory viral infections after stem cell transplant: a prospective trial. Clinical Infectious Diseases, 63(8), 999-1006.

Dharmadhikari, A. S., Mphahlele, M., Stoltz, A., Venter, K., Mathebula, R., Masotla, T., ... & van der Walt, M. (2012). Surgical face masks worn by patients with multidrug-resistant tuberculosis: impact on infectivity of air on a hospital ward. American journal of respiratory and critical care medicine, 185(10), 1104-1109.

Yan, J., Guha, S., Hariharan, P., & Myers, M. (2019). Modeling the Effectiveness of Respiratory Protective Devices in Reducing Influenza Outbreak. Risk Analysis, 39(3), 647-661.

#################
Summary of meta-analyses and systematic reviews (selection):
Bartoszko et al. 2020 point out that in 4 RCTs masks and respirators were similarly effective but only one even “evaluated coronaviruses separately and found no difference between the two groups (p=0.49)”

Jefferson et al. 2011 
“Overall masks were the best performing intervention across populations, settings and threats.”

Cochrane update 2014: masks and surgical wound infections
“Three trials were included, involving a total of 2113 participants. There was no statistically significant difference in infection rates between the masked and unmasked group in any of the trials.”
Cochrane 2002:
“Two randomised controlled trials were included involving a total of 1453 patients. In a small trial there was a trend towards masks being associated with fewer infections, whereas in a large trial there was no difference in infection rates between the masked and unmasked group. Neither trial accounted for cluster randomisation in the analysis.”
Bahli 2009: “No significance difference in the incidence of postoperative wound infection was observed between masks group and groups operated with no masks (1.34, 95% CI, 0.58-3.07). There was no increase in infection rate in 1980 when masks were discarded. In fact there was significant decrease in infection rate (p < 0.05).”

Offeddu et al. 2017: HCW data
“Meta-analysis of randomized controlled trials (RCTs) indicated a protective effect of masks and respirators against clinical respiratory illness (CRI) (risk ratio [RR] = 0.59; 95% confidence interval [CI]:0.46-0.77) and influenza-like illness (ILI) (RR = 0.34; 95% CI:0.14-0.82). Compared to masks, N95 respirators conferred superior protection against CRI (RR = 0.47; 95% CI: 0.36-0.62) and laboratory-confirmed bacterial (RR = 0.46; 95% CI: 0.34-0.62), but not viral infections or ILI. Meta-analysis of observational studies provided evidence of a protective effect of masks (OR = 0.13; 95% CI: 0.03-0.62) and respirators (OR = 0.12; 95% CI: 0.06-0.26) against severe acute respiratory syndrome (SARS).”

Cowling et al. 2010: 3RCTs included, masks show promise, inconsistent data

Saunders-Hastings et al. 2017:
“Meta-analyses suggest that regular hand hygiene provided a significant protective effect (OR=0.62; 95% CI 0.52-0.73; I2=0%), and facemask use provided a non-significant protective effect (OR=0.53; 95% CI 0.16-1.71; I2=48%) against 2009 pandemic influenza infection… If the randomized control trial and cohort study were pooled with the case–control studies, heterogeneity decreased and a significant protective effect was found (not illustrated: N = 1736; OR = 0.41 95% CI 0.18–0.92; I2 = 35%).”

Keine Kommentare:

Kommentar posten