Hmm...but doesn't this suggest then that cholesterol levels are merely a biomarker of something else that causes the mortality, and not a direct cause of it?
That is a good question. These results tell you that statins don't take your risk to zero. If you were totally protected from CAD as soon as you took a statin, then I agree that confounding by indication would not be a thing.
But, across all fields of medicine, we know that confounding by indication is a problem. Preventive medicines, by and large, are not sufficiently powerful to reduce risk to zero.
Right, but I suppose the model for what's occurring then is: High cholesterol either is a marker of some otherwise unrelated health issue, or high cholesterol causes damage to accumulate over time. Beginning to take a statin would prevent the accumulation of further damage, but presumably not reverse that damage that has already occurred. Since I assume there is some benefit of statin intake in terms of actual outcomes, it would seem the latter model fits the data more closely. However, you could imagine a world where statin intake is actually just correlated with "being concerned enough about your health to visit a doctor and follow their instructions", which may be correlated with a turn towards a healthier lifestyle, etc.
