Thread by @dvir_a

My thoughts on the @Pfizer+@Israel_MoH vaccine effectiveness study (you can call it open peer-review if you like). In the study, they show vaccine effectiveness (VE) of >90% for infection, symptomatic, hospitalization and severe disease by comparing to non-vaccinated. =>

1. Control – we know very well that those vaccinated (especially early) are from different socioeconomic groups. @MDCaspi showed it nicely. Those same groups also have higher infection rates. So, naively comparing to non-vaccinated is highly biased.
medrxiv.org/content/10.110…

In my analyses, I estimate this bias at 20-30%. Correcting for this is possible and will reduce a VE of 90% to ~87%. Not such a big deal. However, I really don’t understand why in the study they chose not to control for it.
medrxiv.org/content/10.110…

2. “Healthy bias” – they also do a comparison (results are not presented clearly) to a negative-control group, vaccinated group in days 0-6, where no VE should be seen). They note that they see **higher** infection rate in those days in the vaccinated group.

They use this result to negate the possibility of “healthy bias” – if you are sick before the vaccine you will not come to get your jab. But this is weird for at least three reasons:
A. There has to be “healthy bias”. It's an inherent bias, not something you can have or not.

B. The data MoH shared with the public is of days 0-13, so I can’t test it, but from all I heard from @RanBalicer and @KalksteinNir, who has access to HMO data (and cover ~80% of Israel), there is a very strong “healthy bias”.

C. See point 1. The vaccinated group has lower incidence rate. Even without healthy bias, the vaccinated group should have lower levels of infections.

I don’t know how to reconcile this, but maybe I am missing something.

3. Vaccine protects from infection – everyone is talking about this issue. I'll try adding my 2 cents. Using epidemiological questionings the data shows that ~1/3 of infected are symptomatic (side note – isn’t this much lower than expected?).

The logic they use to suggest that the vaccine protects from infection is that there is no difference in VE between all “positives” and only those “symptomatic”.

But that’s pretty weak. In theory that might be true, but the vaccinated people are exempt from getting tested in Israel (7+ after the 2nd dose, which is what they are looking at in this study).

Also looking at the numbers they actually see ~25% symptomatic in vaccinated compared to 34% in non-vaccinated. I would say that they are seeing more asymptomatic people in the vaccinated group, suggesting the protection from infection is ~2/3 of symptomatic disease.

In summary, I think that something is weird with the data. If the raw data used for this study would be shared with the public it would be much easier to get answers for the questions raised...