I don't see why OS HR>0.8 is a big deal in this case. This is NSCLC against Tarceva not placebo. As long as stat significant, I really don't care if HR> or <0.8. Tarceva got approved with HR>0.8 against placebo.
I think a significant fraction of that ">0.8" is outright failure of the trial. And even if it gets, for instance, 0.83 it will be going against a MetMab label that says HR=0.5.