Dendreon Corp fka DNDNQ

[iwfal]

Some 200 events were required for the Vital 1 interim; more than the total number of events in the combined 9901 and 9902a Ph3 trials.

Tax327 had a p value of 0.009 for 367 events. At only 200 events it would have had a p value nearer to 0.10 (above p=0.05). DSMB's typically stop for safety (i.e. not something the company gets to weigh in on) only when p value is a lot lower than that.

My limited understanding of the O'Brien Fleming method of allocating alphas lead me to believe that it also provides some sort of horizontal funnel plot that can be used to set some outer limits for futility and any stopping test for unexpected overwhelmingly positive results.

OF does not require futility testing per se. And many companies choose not to use futility testing - to be Dew-class cynical I would suggest a reason is that the companies need the trial as much as they need success. (and to be less cynical, many trials are done enrolling by the time the interim is performed so it doesn't cost too much to keep the trial going even if it is long odds.)


BTW - I believe a common futility test is if given the results actually seen plus assuming that the remainder of the trial goes the way you expected initially (i.e. using what HR you assumed before the first patient enrolled) then if you have less than 20% chance of stat sig the trial is halted for futility. I suspect that is a fairly high alpha for futility but I haven't done the math.