Replies to post #3625 on Repros Therapeutics Inc (RPRX)

[io_io]

<I think some are confusing this point with companies that clearly manipulate the press on trials, and the pushback on that.>

I agree. Most biotech CEOs have to mis-lead to the extent they can get away with.

Still, very few bemoan their bad luck due to a lack of trial design power - which they could easily do (blame it on the consultant, or on withdrawals, etc).

[iwfal]

There is no way to avoid it, it is just a fact. Well, you could avoid it by running "pre P3" trials to confirm, but that is obviously absurd.

Definitely OT. But...

Actually there is - and big pharma does it. But before going into that, a review of Dew's Program Survival Bias (which is hardly a new concept - since it is essentially pure Bayesian in origin).

If a trial gets a p value of x (lets go with 0.04), what is the chance that it is an efficacious drug?

Ans: less than the 2% a naive interpretation of statistics might have you guessing. Because there are soooo many more dud drug candidates than real ones.

More interestingly a similar effect should exist when you try to calculate the efficacy off that ph ii - you overestimate the efficacy when powering the ph iii. Because there are so many more ok drugs than great ones, so many more marginal ones than ok ones, ... .

Again, the point is that typically a data mined efficacy estimate will typically be significantly too high. So any powering assumptions driven off of it should be suspect unless you overpower - which I will suggest that big pharma does much more than small biotech. Alternatively you can power your trial by asking what is the minimum efficacy that will allow a market - but I have NEVER heard a company claim that they were 90% powering for an efficacy of x, because they figure x is the worst efficacy can be and still be sellable.