what could go wrong?, page-53

These discussions appear to be confusing median PFS with time by which we have 90% of patients progressing.

Phase II had 2.4 months median PFS for control and 5.1 months PFS for HA - a 12 week advantage in median PFS for HA. Eye-balling the K-M curves, 90% of patients progressed by about 11 months in both control and intervention groups (no difference).

Phase III with a somewhat different population and a different regimen is powered to detect a 6 week difference in median PFS.

With 200+ per arm, would be VERY unlucky to get meaningful maldistribution of risk factors (an essential pre-condition of confounding). If there is a problem is more likely to be some unexpected characteristic in patient selection which gives longer PFS for both treatment or control (in which case we are stuffed), or failure to properly implement the study (e.g., choosing ineligible patients), in which case the very well paid research implementation agency could be sued (and it is insured for this).