They don't. But half the people got a placebo and half got the vaccine. 185 in the placebo group got covid, with 30 having severe covid, but 11 with covid and 0 severe in the vaccine group got it.

So do you really think that totally by chance, one random group of 15,000 people got a ton of exposure, and another group of 15,000 had little-to-no exposure? The probability of that is quite low.

Just to clarify, 185 got symptomatic covid. The 185 doesn't include the unknown number of asymptomatic people.

I am curious why they didn’t test everyone. Is it just a cost/test capacity thing? Wouldn’t testing to catch asymptomatic cases increase statistical power to allow the study to yield results sooner with smaller error bars? Also, isn’t preventing asymptomatic cases an important function of the vaccine (to get to herd immunity)?

Even without any rigorous stat modeling it is pretty clear it is tremendously effective:

https://twitter.com/tmprowell/status/1333442134553325569

I agree more testing could have been interesting, but this was a large trial with very clear results.

They don't fullly know. They rely on statistics.

I hope they at least ask subjects about exposure.

I'm in the AZD1222 (AstraZeneca) trial, and for the next two years, I need to fill out an online symptom questionnaire every single week. They also draw blood at the 6-month and 1-year mark, so presumably could test for antibodies.

Subject reports of exposure are notoriously unreliable. People forget. And even when anonymity is guaranteed they lie about shameful behavior.