>> I think three respected posters had mostly agreed with the possibility of such a result in their respective simulations - what you need to keep in mind is that their simulations were assuming that the control arm would perform up to its "best" historical norms {I think they all assumed a 10 month MOS}
Why assuming 10 months, not 10.4 months? Why not even higher? The 10.4 months from SNY/REGN trial included both squamous and non-squamous NSCLC where squamous typically has lower mOS than non-squamous on docetaxel. SUNRISE enrolled only non-squamous NSCLC. Since SUNRISE enrolled patients from Korea and Taiwan, why not use even higher numbers by incorporating those ethnic groups? That type of simplified methodology doesn't work in handicapping clinical trials. If you still can't see it, oh well.
and I think three respected posters had mostly agreed with the possibility of such a result in their respective simulations
Actually, the sims all hinted to the possibility of exactly what happened.
I never saw golfho's "work", but the other 2 had decent numbers showing how the control arm should be progressing. But the result of that left the Bavi arm with very few events, far fewer than what one might rationally expect.
So at that point we knew the control arm was doing exceptionally well. It still could have played out OK (with Bavi besting an outperforming control). But certainly what happened was in play.