Biotech Values

[DewDiligence]

GNVC: Those curves [#msg-38279815] look better than what I had imagined in my head. But I agree with you that it is not the best idea to jump in before the second look.

Presuming that relatively little alpha will be expended on the second interim look, it seems unlikely that the p-value threshold for the second interim will be met and hence the trial will likely continue to the final analysis. If the HR at the second interim is disclosed and is in the vicinity of 0.75, betting on the final outcome could be a decent proposition. The fly in the ointment is that there’s no assurance the HR at the second interim will be publicly disclosed.

This is somewhat tangential, but does anybody think they suffered from doing the uneven randomization in this trial?

I think they probably did for the reasons you mention (below).

It looks to me like the control arm is pretty jumpy in the 5-10 month range, and I wonder if that is partly due to the smaller numbers in this arm? If larger numbers on the control arm could "smooth out" the K-M profile, they could have had a nice clear separation starting from 3 months and lasting well into 20 months.

From a purely statistical standpoint, a 1:1 randomization is optimal; however, this must be balanced against the faster patient enrollment that is enabled by having a 2:1 (or even more unbalanced) randomization.