Don -- what you posted isn't inconsistent with what I said.
I think the general point about survivorship bias is true. And the data evidences that. But I'll bet if you studied the individual cases, you would see a solid correlation between "cherry picking" from a big basket, and Phase III underperformance.
Again, if I do my preclinical work, have a solid Phase I to cover safety and dosage, do a single Phase II (ideally double blind and randomized, but practicality may dictate otherwise) in the indication that preclinical indicated would make sense, get great results in that single indication, and then set up a Phase III with similar (but not identical) parameters to the Phase II, one would expect much less "fading" in Phase III results than if it were a scatter-shot followed by a cherry picking.
The iwfal/DD point isn't a trial design point -- it's a statistical one. It's analogous to Type I error in a trial. The more "looks" you take, the greater the chance that you're cherry picking rather than finding real efficacy.
All I'm saying is: the facts in each situation matter. And perhaps no one is disagreeing with that here.
“The trick is in what one emphasizes. We either make ourselves miserable, or we make ourselves happy. The amount of work is the same.” Carlos Castaneda
