[omginternets]

>At some threshold of disease viciousness, society (and the FDA) should relax substantially on the risk-aversion, because the patient is already facing the worst-case scenario.

This is a fair point, but the rest... well ... still manages to miss the above argument. Side-effects aren't observed in a vacuum, but efficacy is still observed in laboratory conditions, so we really can't comment on whether (or how) this molecule affects other systems.