Everyone, please don’t miss reading Fosco’s closing point from his prior post, it’s dead on:
PS : regarding the early termination posts : no way it can happen whatever the effect size. This is an event driven study, you need to wait for 298 as per the protocol, even if early significance is reached with limited sample size, pharma need to stick to protocol as agreed with FDA. There is no such thing as 3 year measurement of efficacy or survival. It is all about having p<0.05 with sample of 298 events, eg logrank test on survival curves which is measuring the gap between two plots of data, if you stop earlier you are breaking the protocol, reducing the sample size, increasing the standard error which gives an estimate for the efficacy etc
The terminology in the declaration of the study includes "measured at year three" in the kaplan meier analysis - fyi. At year three on that chart, the spread goal is 10% improvement to OAS.
And I think it was yervoy stopped early by a couple or a few events. Same kind of study.
Can you make a kaplan meier style pseudo-analysis and fail to get p value .001 knowing what we think we know ?