I'll just add that if they meet the primary endpoint, they will likely do another analysis adding in the pseudoprogressors, according to the old leaked protocol, thereafter shared by one or more board members.
The number actually enrolled would still be higher. 348 + 32 = 380.
Let's take Chris's idea, and I think Senti's too, that they were very near or at 348 enrollment for the main group.
Take an extreme, if 100% PFS evented from the 8 patients in the placebo arm, you'd need another 8 patients to event from the treatment arm to reach a median for the pseudo progressive treatment arm. However, you may not get 16 events from the psPD by the time both groups are simultaneously unveiled. "MAIN ARM & PSEUDO-PROGRESSOR ARM (ABOVE) WILL BE UNBLINDED AT SAME TIME AT END OF STUDY." http://www.nwbio.com/AACR_poster_prolonged_survival.pdf
Notice that with the original group of 348, you'd need only 116 events from the treatment group to reach a median from their 232 enrollment, even if all 116 placebos evented.
So why choose 248 events from the main arm? Because, if there were theoretically no events from the pseudoprogressors, and you simultaneously unblinded them, see http://www.nwbio.com/AACR_poster_prolonged_survival.pdf, then you would not have enough events to get to a median for the the larger treatment group. However, if you were think ahead, you'd derisk that extreme possibility by simply adding 16 more events to the 232 events otherwise required by making the main trial primary endpoint analysis at 248.
Get it?
232 + 16 = 248.
If the extreme doesn't happen, statistically there is no problem, because the additional events from the psPD group will simply mean further powering for that separate endpoint. The trial was once again derisked, even for the post primary endpoint analysis.
Market Data 
Markets 
AI
