Summer: I'm having a tough time getting my head around the proximity between the efficacy interim and the final event count being hit -- these two events appear to be only what, at most, 4 months apart (September for the interim, call it, and January for the final event).
There's nothing I can do to my model that would make sense of this if the efficacy interim were done at even as high as 80% of events. It just doesn't work.
Summer, if it's true that it took 3 months to do the analysis, then you may be right, but that was the third interim that looked at survival (two futilities before it, yes). It should not have taken 3 months.... I was thinking 6 weeks. Nov announcement of Sept event. Sept to Jan is 4 months.
The event rate isn't linear, of course, but near the end of the trial, I'm seeing about 15 events per month... something like that. Maybe 16.
Well, Summer did post something about how the deaths would occur in an asymptotic rather than linear fashion, which I think is consistent with dilivent's "rash of deaths" theory. It would probably be best to let Summer speak for herself though, especially since I don't really remember what an asymptote is.
But as long as we're asking her questions, I have one. If custirsen should fail to achieve statistical significance in SYNERGY, is the FDA likely to require a lower p-value than .05 in the AFFINITY trial? Since AFFINITY involves second-line patients and SYNERGY first-line patients, I'm not sure why the results in one should affect the requirements in another but maybe they would. OGXI has said the FDA has agreed that success in SYNERGY would suffice for approval (because the results in the P2 trial would count as the second trial it usually requires) but it's never said that about AFFINITY. Would it matter whether SYNERGY showed a trend in custirsen's favor?