web analytics

To mask or not to mask – my take on the recent Cochrane systematic review

Masking during the Covid pandemic has always attracted very polarised views. My take on it in March 2020 was to be sceptical:

You can see a lot of people wearing them, but the efficiency of the masks is actually minimal. 


The point I was making in March 2020 was that badly worn masks are actually counterproductive. I changed the view a bit later in June 2020:

Hence, any face covering that blocks the droplets will do the job.


but still wrote in July 2021:

Simple masks are effective, but mostly for protecting the public from the person who wears them, not the other way round.



 I do not like wearing a mask, although I have invested in a high-quality one by Airinum having worn similarly excellent homemade masks at the start of the epidemic.


In the 2018/19 flu pandemic, people were encouraged to contain the droplets associated with coughing and sneezing:

Actually, even before that, The Good Soldier Schweik would use every opportunity to spit into special vessels provided to control TB:

Cochrane review

So, how does the Cochrane review actually say? The authors are very cautious (as they should be). They conclude as follows:

Surgical masks

Wearing medical surgical masks probably makes little or no difference to the outcome of influenza-like illness (ILI)/COVID-19 like illness compared to not wearing masks.

The risk ratio (RR) is 0.95. What does this mean? It means that your risk of catching the disease (flu or Covid) if you wear a mask is slightly lower than the risk associated with not wearing.

But, the 95% confidence interval (CI) is 0.84 to 1.09, so we are 95% certain that the risk ratio is anything between either pretty good (0.84) or a bit rubbish (1.09).

Actually, we can also say that there is a 2.5% chance that masks reduce the risk by 16% or more. Conversely, there is a 2.5% chance that masks are more rubbish.

Fancy masks

This is a mask I still occasionally wear (more precisely, it is K95-certified which is similar but not identical to the US N95 or EU P2). The conclusion for N95/P2 masks is:

We are very uncertain on the effects of N95/P2 respirators compared with medical/surgical masks on the outcome of clinical respiratory illness (RR 0.70, 95% CI 0.45 to 1.10; 3 trials, 7779 participants; very low-certainty evidence).


To unpack this “mumbo-jumbo” of numbers, we are bit more certain that N95/P2 masks work (the risk ratio is 0.7, so the reduction could be 30% but as before, we can only say that we are 95% certain of RR being between 0.45 – a reduction by 55% – and 1.1 – an increase by 10%.

Some thoughts on this figures

As a statistician, I am always a bit wary of interpreting results of poorly designed experiments with many confounding factors and small replication. I am even more wary of meta-analysis studies like the Cochrane review, as there is a lot of arbitrariness in selecting and interpreting results.

It is absolutely OK if the results are clear-cut, but the large confidence intervals make me very sceptical. I suspect that in the studies included here, it was almost impossible to sufficiently control all factors associated with mask wearing, to draw any firm conclusions.


So, my TL&DR is that there is simply not enough evidence gathered by this type of experiments to draw any firm conclusions. If anything, there is a slight preference for risk reduction, but we cannot make this statements for sure.

The authors themselves start their conclusions that:

The high risk of bias in the trials, variation in outcome measurement, and relatively low adherence with the interventions during the studies hampers drawing firm conclusions.


Which masks, when, how and by whom were worn, matters. As do the type of disease, its mode of transmission, and overall human behaviour.

And that’s before we start taking into account possible drawbacks of the methodology and arbitrariness of assigning weights, as Tomas explains in his thread.

So, does it all matter?

Yes, because the next time we are hit with another pandemic, we will be without evidence that one of the simplest ways to control spread is missing because people will not take them up.

The Cochrane authors conclude that:

There is a need for large, well-designed RCTs [Random Controlled Trials] addressing the effectiveness of many of these interventions in multiple settings and populations, as well as the impact of adherence on effectiveness, especially in those most at risk of ARIs [Acute Respiratory Infections].


I finish with a quote from my July 2021 blog post:

But I am really puzzled by such a strong resistance to the idea of mask-wearing. Even in Western countries, we are taught from childhood that we should be doing things that protect us – and others – from risk, even if [they come at] a cost. “Do to others what you would have them do to you” is the Golden Rule widely accepted in our culture. How is it then that so many people in the West feel so strongly against mask-wearing?


In the last couple of days, there has been a lot of discussion on Tomas Pueyo thread cited above, pointing out problems with his comments. My article did not rely on his analysis and so I left it unedited.