Let me expand the docetaxel comparison a little further
Docetaxel's p-value in the 772 patients clinical trial with way more events than IMPACT merely demonstrated a p-value of 0.01, and an HR of 0.76.
With that type of treatment effect, docetaxel may only demonstrate a p-value of 0.2 at 94 deaths (the events in trial 9901).
So another way to look at this is FDA in essence approved a drug with 20-fold (0.2 vs 0.01) higher chance of being a placebo with much more toxicity, had it been subjected to the same clinical trial condition of 9901.