Replies to post #123896 on Biotech Values

[BTH]

higher than the 5% claimed by Martin, who essentially assumes the interim HR to be null. My question is how could he be so sure

I dont think this guy is the most ethical guy around.

[iwfal]

I will safely wager you the futility HR boundary will not be 1.0.

If they used 20% conditional power (a threshold suggested in one of my text books) I am pretty sure it would be an interim HR>1.0 (using OBF at 50% info fraction). I just don't know what threshold gets it under 1.0.

Then the non-termination for superiority then set the lower bound, somewhere at 0.71~0.72, per the OBF boundary. Without even knowing the exact the futility HR, it should be easy that we agree that if we let our simulator loose with an observed HR drawn from (0.71 to 1), the power of success at the second interim will be much higher than the 5% claimed by Martin, who essentially assumes the interim HR to be null. My question is how could he be so sure.

I'll acknowledge that Martin didn't put his story together in a completely coherent way - but I believe his thesis is that the drug is no better than placebo, and given this assumption then even if it passed the interim futility analysis then it is still has odds of less than 5% (when, prior to the interim, the odds of success for a worthless treatment were 2.5%). That said, upon further thought I will amend my previous statement to just say that 5% is roughly correct given those assumptions.

[iwfal]

ONTY -

I will safely wager you the futility HR boundary will not be 1.0.

FYI I did the math and if they were using 20% as the Futility Threshold then the HR boundary at the interim of this thread would be 1.03.

Obviously at this latest interim the same futility limit give a value around 0.92.

Finally - for those who follow ONTY if a Futility Threshold of 50% is used the HR at this last interim has to be about 0.885 or better. But obviously it is heavily dependent on the Futility Threshold picked (20-50% being a good range of the possibilities of what Merck might have picked).

PS Obviously I created a tool to make this calculation less onerous.

Background: Futility limits at interims are done in an inherently odd and confusing fashion. The question asked is: "Given the HR seen at this interim, and assuming that for all the remaining events the nominal HR is the HR originally assumed in calculating the power of the trial before the first patient was enrolled THEN if the power of the trial to be stat sig now is less than x% then halt for futility.". Typical values for x range from 20% to 50 or 60%. As far as I can tell there is little consensus about which x should be used.