I will probably regret posting over here, so much noise (on both sides), but here are a few questions for vinmantoo:
1) You assert the second line trial was corrupted. No disagreement. But there are different types of corruption with different effects. Do you assert that there was anything corrupt about the trial other than a) switched doses between soc and 1mg arms, b) possible switched doses involving the 3mg arm (no evidence that this happened, but I'll grant it for now just for argument's sake) and c) the censored patients? If so, what sort of corruption and what is your evidence for this?
2) What effect would the dose switching (including in the 3mg arm if you want) have on the *efficacy* results? Would giving placebo in place of bavi to the experimental arm and bavi in place of placebo to the control arm tend to yield results that exaggerate or understate bavi's actual efficacy?
3) Do you think the FDA would approve a phase iii trial if there were greater than normal concerns about a drug's *safety*? Do you think that the various studies have made a strong case for bavi having a good safety profile?
Here is why I ask: IMO the sort of dose switching in the second line study would only serve to narrow whatever difference there might be between the actual soc efficacy and the actual bavi + soc efficacy. This sort of corruption, though, would make the trial less valuable for judging bavi's safety (if the control has received bavi, then you can't say "oh look, the bavi arm had the same AE's as the control so it is just as safe). Based on this sort of corruption, it is the safety element that is the bigger risk going into phase iii, but there is plenty of other data (other bavi trials, comparison of safety in corrupted control to that in historical controls) that mitigates it. In fact, in terms of efficacy, if the only corruption is the switched doses, then the originally reported *stat sig* efficacy is an understatement. Even with the trial's small n.