I am beginning to wonder if ARQL sucks so incredibly badly because of this selected trial with MetMab. Isnt the Roche trial more highly selected than that of the ARQ trial? Granted, ARQ has a little head start on Roche but a targeted, more defined trial is going to have a chance of a higher degree of success, and better comp results.
I would like to see a statistics about this. However, a decent size trial powered 90% to show a 25-35% improvement at final should be powered at least 20% at interim assuming the interim is at 1/2 # of events and uses standard OBF under the exact assumptions used during trial design. So I say the probability of early unblinding at interim should be higher than this at least if you think original trial design assumptions are correct (or if you think the trial will be a success at the end).