I agree with what you say, but this the origins of a murky area of life. Torture is reprehensible and we should avoid it all all costs. Occasionally though, all cost situations arise, and men in the field choose to exercise what we in our armchairs might consider torture.
Randomized controlled studies should remain the standard. When they are possible, we exercise our judgement.
As Brignell has pointed out in his book, we don't know that smoking "causes" the cancer any more than fertilizer causes tomatoes to grow. It is pretty clear that if you want to avoid lung cancer, avoiding smoking is a good idea. Most smokers though roll the dice on this one and WIN. We can never forget that 11 out of 12 smokers WIN the dice roll on lung cancer. Start getting into RRs significantly less than 23 (and just about all are < 2), and you are in areas where people WIN 99% of the time.
Leaning on such studies is foolish. It may be necessary from time to time, but it should not become mainstream.
I would guess that the drug manufacturers are gambling that the worst case scenario of being sued for supplying a product later discovered to be less efficacious than claimed (or worse) will be more than balanced by the earnings prior to any claim and that perpetual inflation would discount the value of any penalty by the time settlement had to be made to the injured party's Estate.
If these are the standards by which supposedly full process trials are performed then I see no reason not to accept any epidemiological studies as equally valued the that value may be small.
Bear in mind humanity tends to believe in 'old wives' tales and the like for 'real' and 'natural' treatments and we might as well abandon the scientific method and go with consensus gut feel. I'm sure people would feel more comfortable with that.
All good preparation for the return to non-industrialised living.