Repros Therapeutics Inc (RPRX)

[iwfal]

However, you will see plenty of trials that fall short of the assumed benefit margin, but the lower margin of advantage is still sufficient to hit statistical significance.

You appear to be implying that if you are stat sig, but fall short of the efficacy (x) assumed in the power calculation, that that is not part of the power calculation output. I.e. if the HR assumed in a power calc is 0.75, and the patient pop is such that the power calc says 90% power, then trials where you get p=0.04 and measured HR=0.82 are in addition to the 90%.

This is incorrect. The cases where the measured HR is worse than the assumed HR, but the trial is still stat sig, are part of the power percentage.


Biotechs as a group clearly overestimate the TRUE efficacy (the 0.75 in the above example) and thus overestimate the chance that they can hit stat sig (the power). And biotechs are normally off by a lot more than 10% on these power calculations. Hence all the efficacious drugs that miss stat sig.

Stated yet another way - on an endpoint where it would clearly be relatively cheap to up the power I would do so to account for enrollment criteria differences, SOC changes, ... .