A 4 September headline in the Atlantic proclaimed “The Masks Were Working All Along.”
Many who were following the mask controversy were skeptical. Way back in January of 2020, before coronavirus hysteria enveloped the nation like a toxic fog, the CDC repeatedly recommended against face masks for the general public. But in March of that year, the CDC abruptly reversed its position. This was a surprise, coming as it did just two days after a review by the Minnesota Center for Infectious Disease and Policy concluded that masks don’t work.
Since the CDC reversed its position on masks, reviews by the Nordic Cochrane Center, the CDC itself, the New England Journal of Medicine, the World Health Organization, and the Norwegian Institute of Public Health all concluded that the masks don’t work. The only randomized controlled trial on the efficacy of masks in preventing the spread of COVID-19, published in March of 2021, found that – guess what – masks don’t work.
Moreover, there was some evidence the masks might actually be making things worse. A 2015 study found that cloth masks for nurses actually increased the rate of all respiratory infections, presumably by acting as a substrate for bacterial growth. (See my previous essay about the “re-analysis” of the nurse study.) Other studies have indicated that masks worn by operating room personnel may increase the rate of post-surgical infections.
So it came as a surprise to some when we were told the masks had been working all along. The basis for that conclusion was a huge randomized controlled trial of masks carried out in Bangladesh, then available only in preprint form, but since published in Science. The researchers were all eminently credentialed experts from such prestigious institutions as Yale and Stanford.
They randomly assigned six hundred entire villages to either the experimental group or the control group. The experimental intervention included free masks (either surgical or cloth), public information campaigns about the importance of mask-wearing, partnerships with local mosques to promote masking, and “mask promoters” whose assigned task was to stop individuals in public places not wearing masks and remind them to do so. The primary endpoint was symptomatic COVID-19 infection, conformed by PCR test.
The Atlantic article summed up matters thusly:
Masks work, period.
The randomly assigned pro-masking policy reduced the number of confirmed, symptomatic COVID-19 cases by nearly 10 percent, relative to the control group.
These conclusions have been repeated in such august fora as WebMD, Nature, and the Economist.
In a previous essay, I pointed out that the primary endpoint – symptomatic COVID-19 infection of any severity – is not a clinically relevant endpoint, and that the authors’ emphasis on relative reduction in risk – as opposed to absolute reduction in risk – obscures how tiny the promised benefits are.
Now, thanks to the intrepid Ben Recht, Associate Professor of Electrical Engineering at UC Berkeley, we know just how tiny these effects are. The authors of the original study – to their credit – have made the raw data available. The control group contained 161,861 individuals from 300 villages and the experimental group contained 178,322 individuals from another 300 individuals. There were 1,106 cases of symptomatic COVID-19 in the control arm as opposed to 1,086 in the treatment arm.
Now I’m going to get all mathematical on you for just a moment, but bear with me, okay?
Let nC = the number of individuals in the control arm,
nT = number of individuals in the treatment arm,
iC = the number of covid cases in the control arm, and
iT = the number of covid cases in the treatment arm.
The absolute reduction in risk is equal to:
iC/nC – iT/nT
That works out to a one in 1,538 (or 0.065%) reduction in the primary endpoint, covid infection of any severity.
But wait – it gets better (or worse, depending on your point of view). Remember that primary endpoint of symptomatic covid infection of any severity is not a clinically relevant endpoint anyway. The case-fatality rate for Bangladesh (number of deaths divided by the total number of cases) is equal to 1.57%. Multiply 1.57% by 0.065% and you get a figure of one in one hundred thousand.
We don’t need to calculate p values to figure out that such a puny promised reduction in risk is just not worth taking into account, but Professor Recht’s discussion (available here, here, here, and here) of the statistical gyrations the study authors had to go through to make this paltry figure “significant” makes salutary reading. As the saying goes, torture any dataset long enough and it will tell you what you want to hear.
Masks don’t work. It’s time to move on.
All this is unlikely to sway the pro-mask crowd, whose advocacy for muzzling the populace seems impervious to any actual data and has attained a level of fervor that can only be described as religious. But to anyone whose mind is not hermetically sealed, the lack of evidence for the efficacy of masks, combined with the innumerable instance of our rulers and their professional flatterers in the entertainment and media industries flouting the rules the rest of us are forced to live under, along with the vitriolic denunciation of doctors advocating for safe and effective early treatment, suggests that whatever is going on here is not about saving lives.
On the border between england and wales people are obliged to wear a mask on one side and can leave them off a step away on the other side. At first people who chose to use them were made to feel foolish by the media non-experts now people are making up their own minds and flouting rules in such numbers they couldn't police it any more. Seems the virus can hop around without a passport Thing is though when the rate is high in one area of high density we may reasonably think 1 in 1/2 thousand risk is quite high so worth weighing up whether to use the horrible things - or not.
How many physicists work for the CDC? That will tell you all you need to know about the CDC's competence to evaluate masking.