Per request, 1 of two posts from a different mb.
Includes pictures of some docs, which don't seem to xfer, which is why I initially just posted the post #s.
"His [MT's] caution about an Interim halt was most certainly based on his knowledge of something but what that is we are all guessing. The powering of the interim, the OS needed at interim or at final to be seen or frankly another possibility would be that becasue the FDA permitted TD to go from P1 right to P3, isn't it possible that part of that agreement was to see HEAT through to final no matter interim numbers to establish full and unequivical safety since there was no P2 to establish safety signals, thus setting the bar so high it was virtually impossible but maybe TD came close hence the second look? All conjecture but interesting to say the least,"
I’d like to offer my perspective on this.
I doubt VERY much that they used standard OBF here at the interim. The FDA does not want PFS trials to stop at interims. Certainly not one stopped based on an alpha calculated at the “standard” 0.05 (2-sided) level - which would equate to 1-sided 0.003 hurdle at the 50% mark (slightly higher at 57% of information or whatever it worked out to – perhaps 0.00315).
I would not be at all surprised to learn that the stopping hurdle was in the neighborhood of 0.000003 (That was the START trial hurdle at 50% info fraction, they too had a SPA)
Why use such a strict hurdle for efficacy? Well, I think the main reason is that the FDA wants to see compelling (highly persuasive) statistical data, especially for a single registrational trial (especially one with PFS as a surrogate endpoint). MT basically said so much in his March comments about the “continue” explaining that the interim was essentially a surrogate of a surrogate and their expectation was “continue”.
Consider the recent approval of Inlyta: The FDA made it abundantly clear in their End of P2 meeting that they didn’t want to see the trial stopped for efficacy at the interim. However, ANY interim efficacy analysis requires SOME alpha adjustment (a cautionary note to DCTH investors – we will hear about this at their AC meeting, but I digress). How do you implement this? You assign a very stringent hurdle; still calculated using OBF methodology, but just not one based on a final alpha of 0.05.
------------------------------------------------------------------
------------------------------------------------------------------
And how did they decide to implement these requirements? They used asymmetrical stopping rules. When they say they used a “fraction” of the alpha for efficacy at the interim, you can be fairly certain it wasn’t the “standard” fraction (0.003 or so)
-----------------------------------------------------------------------
------------------------------------------------
So, what’s the relevance? If they indeed used a very strict interim hurdle, Kid’s theory about the interim “continue” indicating the control arm is achieving at least 19 months is based on an erroneous assumption. It is possible that the control arm is exhibiting less than that, and such an outcome would still be consistent with the interim “continue”. (and of course, this argument says nothing about the chance that control is doing better than 19, it just argues that the continue doesn’t necessarily limit the control to “at least” 19 months – 12 is possible as is 25)
