GPB
Based upon some research you found, I modified my position a bit.
1. I think you are correct that median PFS survival for the control group is not likely to increase very far beyond nine months - at best. The (at first) counterintuitive results from the pseudo-progression group were the primary reason for my change in position. In other words, even though selection of patients is done post chemo/radiation (instead of immediately after surgery), selecting out patients who demonstrate progression and/or pseudo progression at this time does not lengthen median progression free survival of the remaining patients by a great amount, because (as you already know) the pseudo progression group actually has better historical PFS (compared to the non-early progression population), thus washing out much of the potential benefit. So, I think you are correct in stating that even with delayed selection/filtering plus intent for complete resection, PFS in the control group is not likely to go very far beyond 9 months, and certainly 10.5 months would probably represent an outer median extreme result. So that makes the range somewhere between 6.9 -10.5 months -- (with 9 months being the best guess).
2. After reviewing many early interim success stories, I think you (or maybe it was somebody else) is correct that the DMC will not set as stringent a hurdle in the first interim analysis as I previously estimated. I now think their requirement will be somewhere between .0005 to .001. Hopefully the latter.
3. After "tediously" trying to experiment with different curves, I agree it is hard to get much better than .00001 results, but I still find that if the time to event is even longer than the phase 1 group, and we have enough data from the earlier trial patients; even if the control group has better results than before, the probability value can increase compared to phase 1 if the experimental curve flattens out more than the control curve -- (with an additional modification to my lame prior visualization) that being the control curve must drop off to near 0-5% around the same time as it did in phase 1, and the median PFS is closer to 9 months.
Interesting, by excluding pseudo progression patients, I suspect the tail may even be shorter for the control group in the phase 3 trial. I don't think NWBO knew this might happen when they started the phase (II)/III trial a long time ago. This could improve statistical significance by a small amount.
4. My optimism for successful first interim results is not diminished, and if instead the DMC simply continues the trial, I continue to believe (for reasons stated in prior posts) the 2nd interim results stand an even better chance of success. IMHO
