GBM is a fast one at recurrence. Not a whole lot of time to move on to additional studies once you recur. This impact would likely be more muted as with crossover. My only question would be the mesnchymal thing and how that might help a crossover patient(but second surgeries were only done on 2, per Flip).
I am 100% sure that this works. If there weren't a crossover and we had the same JTM numbers there would be less question of crossover etc.
To say that the reducing of the control arm to 99 has 'little effect' is too sweeping an assertion. For maximum statistical power, you ideally have 1:1 treatment to control. 2:1 is bad enough; 70:30 weakens things further. You might end up with a p value of 0.04, when a slightly larger (or less lop-sided) ITT might have given you 0.05...
Alongside that you have the reduction from a planned 348 to 331 courtesy of the partial hold... More depowering.
Take these two power-reducing factors together (neither of which had been in NWBO's plan) and you effectively have a higher bar to negotiate.