1) The interim that they just failed had an info fraction of about 219/380=0.57. Since they are using Lan Demets that implies that the alpha assigned to this look was about 0.006. I don't have time to check enrollment ratio - but at 1:1 that would mean their liver PFS was worse than about 0.69.
2) Again I don't remember the details of the trial - but this was one of those YMB the-trial-is-running-very-long-so-it-must-be-a-success trials. And these results should put a damper on that thinking since 0.69 is a HR that would have been passed had the trial been running as long as contended (factor of >=2x was the contention) due to treatment being a miraculous cure.
3) Although the PR did clearly define their interim measuring methodology they were a mish mash mess on other topics. E.g.:
a) The most crucial question is unanswered - was there a futility test and what was it?
b) Are there other planned interims - and at what even count?
c) The PR implies that there is no statistical penalty for additional interims - which is absolutely incorrect. E.g. they just paid a 0.006 penalty and a future look will be assigned an even bigger alpha (i.e. bigger deduction from the final alpha)
d) They seem to be winging it on interims - how many they have and when. If done rigorously and with FDA prior approval this might be ok - but just the very casual attitude about it makes me suspicious that they are not doing it rigorously. And that will mean that even a great final HR and p value will still flunk the FDA.
Market Data 
Markets 
AI
