From perspective of approval for ANY drug if FDA would insist to wait for Mean results then it could (would) take a decade or more on assumption that one single patient lasted that long. NOT PRACTICAL at all, right?
With stellar data a trial can be stopped and the next phase or NDA can now be applied. So you do not have to have final determination of long lived patients to move forward. At some point say hypothetically at 100% improvement the surviving patients could be added into the complete population with an asterisk. If the asterisk highlighted the CR's/ long term survivors and I was a doc or Patient it would be a welcome piece of information. These should be a red flag in the positive. If I were a patient facing certain death I would take a treatment with 5% chance of a CR. This information should be under consideration in the process at the FDA. You would assume they would have this information regardless of the end point in the trials.
The trial has a predetermined criteria for significance. When that point is reached, no further data is collected and they can determine if the difference between the MOS for the bavi arm and the MOS for the control arm was statistically significant. We don't know the criteria that's being used to determine that stopping point.