> DNDN -- Shouldn't they be able to look at any submission and come up with an all-encompassing, single numerical estimate of the probability of Type 1 error for efficacy? If they can, what is that number for Provenge, so far...<
I can't answer your question, but it's my opinion that Provenge would have been approved if 9902a had achieved a log-rank p-value of 0.10 and maybe even if it had achieved 0.15.
Although 9902a was barely mentioned at the advisory panel, I think it was the centerpiece of the FDA's rationale for withholding approval. A p-value of 0.33 was simply too big a miss to be considered supportive, and I think many DNDN longs were intellectually dishonest is refusing to acknowldge this.
Market Data 
Markets 
AI
