[Gibbon1]

I saw an analysis from a epidemiologist who said the result is due to naive over adjustment for confounding variables. Specifically if you treat related confounding variables as if they are independent you end up with garbage.

They also said elsewhere that a lot of researchers from outside epidemiology are running really bad studies and publishing papers.

[himinlomax]

He's almost certainly wrong, the effect size on this one is astounding. See for example this: https://www.qeios.com/read/WPP19W.3

Even with conservative assumptions, smokers in this French study were 4 to 5 times less likely to be infected.

Keep in mind that drugs being tested are considered good when they reduce mortality by 10%. We're talking 75% reduction in infection rate here, and no increased mortality in those who are infected. It's really really hard to imagine what sort of confounding factor that could be at play here that would make the effect insignificant.

[sharken]

If the facts speak against what an epidemiologist says, then perhaps we should also use critical thinking there.