site banner

Cochrane review is out and masks have weak evidence that they are not effective

vinayprasadmdmph.substack.com

This one is against rationalists because when Scott wrote his review that masks could be effective many of us trusted it.

I don't blame Scott for failing this one because doing review of hundreds of studies is hard and one person can hardly do it. But this clearly shows that rationalist way of thinking has no special formula, they can be easily mistaken and fall by accepting general consensus just like any other person.

I was impressed when Scott did his review about masks. I trusted it because there was no other clear evidence available. Cochrane hadn't done its review yet and NICE guidelines were silent on the issue. We vaguely knew from previous studies that masks are not effective, The WHO had said so. Suddenly everyone flipped and it was not because the evidence had changed. We simply wanted to believe that masks work and we mocked those who said “no evidence that masks help”.

Even with the belief that masks work, I never wanted mask mandates. I preferred recommendations only, so that no one was penalized or prohibited entry, travel etc if one doesn't want to wear mask. Scott unwillingly had been a catalyst for governments to introduce mask mandates and all this heavy handed approach has been for nothing.

Now we are back to square one, the evidence about masks is weak and it does not support their use even in hospital settings. We can all reflect now what happened in between during these 2 or 3 years. When I realized that Scott's review is clearly insufficient as evidence, I asked some doctors if they have any better evidence that masks work. Instead of getting answer I was told not to be silly, parachutes don't need RCTs and accused me of being covid denier for nor reason. Many so-called experts were making the same mistake as Scott by looking at the issue too emotionally. It is time to get back to reality and admit that it was a mistake and we should have judged the issue with more rational mind.

14
Jump in the discussion.

No email address required.

You cannot separate “telling people to wear masks work” from “wearing masks work” in the intervention. It is the real life we are talking about.

The argument that maybe the results would be better if we apply efforts to improve the compliance is a real one and was raised by the Cochrane group reviewers. Their answer was that no one has studied it, so we don't know and cannot claim that it would have helped.

I was just learning about different contraceptive methods. Their reported results of effectiveness are not some best case values but real life results from studies. https://en.wikipedia.org/wiki/Pearl_Index Even that is being criticised that in studies people get better counselling and training and may not represent the real life values. I find interesting that fertility rhythm method has very high theoretical effectiveness (slightly worse than condoms – https://en.wikipedia.org/wiki/Comparison_of_birth_control_methods) and yet it is heavily criticized by all experts in the field. It is always more easier to take a pill than measure temperature daily plus all other behavioural aspects.

Some argue that it still makes sense for their elderly relative to wear mask to protect themselves. Maybe, but I don't know your elderly relative. The statistical chances are that they are as much non-compliant as any other member of the population. Telling all hundred or thousand of them (how many readers do we have?) to wear a mask will statistically yield the same result as in those studies.

I find interesting that fertility rhythm method has very high theoretical effectiveness (slightly worse than condoms – https://en.wikipedia.org/wiki/Comparison_of_birth_control_methods) and yet it is heavily criticized by all experts in the field.

Interesting analogy -- 'withdrawal' might be a good parallel with masking, in that it actually works not bad if you can pull it off (out) but field results are poor due to, uh, implementation difficulties.

For this reason people are usually told "withdrawal is like 99% ineffective" -- which isn't really true, but serves the public health goal.

Leads one to speculate that the public health goal is quite different in the case of masking.

You cannot separate “telling people to wear masks work” from “wearing masks work” in the intervention. It is the real life we are talking about.

Then you just go back to the seat belt problem. Seat belts existed since the early 1960s but usage rates were abysmal for 25 years. Someone studying the effect of seat belts in 1985 would have found their efficacy lacking for the simple reason that few people used them. Seat belts had been required equipment in cars for some time but given that so few people were using them so long after their introduction, the Federal government could have been forgiven for scrapping the requirement altogether as ineffective. Instead, this is around the time state governments started requiring seat belt use among all front seat occupants. As the 1980s became the 1990s and more states started adopting such laws, auto fatalities, which had been more or less stagnant for decades, halved between the late 1980s and the present. Part of the problem could be that in a lot of places these so-called "mask mandates" were so inconsistently enforced and widely ignored that the actual effect was that of no mandate whatsoever, and that the real solution is stricter enforcement. This isn't necessarily a policy I would advocate for, but simply stating noncompliance is proof that the underlying implication is wrong is disingenuous, to say the least.

Most cars didn't have seatbelts then. Some of those cars are still running and it is legal to drive them without seatbelts.

RCT could be easily made by manufacturing a car with two models that are different only by presence of a seatbelt and randomly shipping to different dealers. If the car had a seatbelt, a dealer is obliged to explain a buyer how to use it properly. The car could have a mechanism installed that warns if the seatbelt is not in use and the dealer warns that defeating this measure will void the warranty. The compliance rate would be at least 50%. Then you just collect statistics from road accidents and related injuries. I am sure very soon this experiment would be stopped by an ethics committee because the seatbelt group would have huge difference that further studies would be unethical.

Maybe people who study road safety used a similar setup by comparing one model with a seatbelt to a different model without a seatbelt. It has some bias as assignment is not random and both groups can be different, for example, one model can be chosen by more careful drivers etc. It is very hard to control for all these factors afterwards. But even then they saw such a massive difference in injuries that could be explained only by seatbelt use, that it was made mandatory. It is easy to make mandatory rules in driving because most things in driving are mandatory, you have to stop at red light etc.

But the absence of RCTs and irrefutable evidence could be a minus because it was harder to explain people why seatbelts are protective. Many people said that seatbelts will protect you in minor accidents but in major crashes they would make you more likely to die. That's why we need a good evidence that seatbelts have a total protective effect from deaths although they won't protect in all possible cases. It would have improved adherence even without policing.

Someone studying the effect of seat belts in 1985 would have found their efficacy lacking for the simple reason that few people used them.

Nope. They'd find typical usage would (on average) result in people surviving the car crashes, and the non-usage in dying. We don't have anything like that justifying even an N95 mandate.

No, his claim is if you studied P(death | accident, car with seat belt) / P(death | accident) you wouldn't find an effect unless the sample size was massive because P(wears seat belt| car with seat belt) is very low, so P(death | accident, car with seat belt) = .1 * P(death | accident, wearing seat belt) + .9 * P(death | accident) which is .92 * P(death | accident), even if seat belts reduce crash risk by 80%!

We don't, but that's not what the study looked at.