>>13175571The relative risk is the property of the treatment. It just compares the number of infections in the control group vs the vaccine group.
For example you have 100 people in each group and there were 4 infections in the control group and only 1 infction in the vaccine group.
So the relative risk reduction is 1 - 1/4 = 75%. But the chance to get sick was 4% and %1 in each group, so the ARR is 3%.
ARR is always much lower since the baseline risk is very low to begin with. RRR is used when the baseline risk varies from group to group or it is uknown.
It can also vary in pandemics obviously because of seasonal fluctuations or masks/social distancing etc.
Basically your RRR means "whatever risk you have has been reduced by this much". And then if you know your baseline risk you can apply RRR to get the ARR.
For example for Pfizer:
Vaccine: (N=18,198 not infected, 8 infected) 0.00043960874
Placebo: (N=18,325 not infected, 162 infected) 0.00890207715 ~ 0.8% NNT = 125
The first observation is you don't need to know N to get RRR. You just 1 - 8/162 and you get 95%
The baseline risk for that *particular* control group was 0.8%. And it has been pretty much reduced to zero.
BTW by infection they mean "symptomatic infection". So a chance to get sick in that particular group was 0.8% which is fairly low.
And they reduced it to 0.04%. So the ARR is ~0.8%.
But if for example the trial is done somewhere in India at the peak of infections with no masks or social distancing, the baseline risk will be much higher.
So the ARR will be higher and NNT will be lower. But we know RRR from the same vaccine is 95% so we multiply it by the new baseline risk to get the new ARR.
>What is the scale of the ARR?It is just the difference between two percentages.
>risk of severe disease? risk of death?Apparently for Pfizer it is a risk of symptomatic infection.
