DNDN - Any real statistician will tell you a trial shooting a p=0.01 with 125 people is more clinically persuasive than a trial shooting p=0.01 with 1250 patients.
Docs and amateur biostaticians hear "small trials pose the following problems..." and zone out the entire rest of the biostats lecture. With time and distance away from the lesson, this turns into "small trials are bad." Investors, who have an even shorter attention span than pre-med students, are even worse.
They miss the important part about describing strategies to ensure the hazards that come with small trials (mainly the magnification of small imbalances) do not harm the overall survival benefit.
When I see a 125-patient trial a company claims as a demonstration of a survival benefit, I do not automatically think, "Oh, that's too small." I think, "what have they done to correct for the probable imbalances inherent in small trials?"
Quibble if you want about the extraordnary steps Dendreon went through to prove the imbalances that were found did not explain the survival advantage they claim exists. But don't say "it was a small trial" and expect someone who knows better to listen.
Subset analysis of the control arm isn't bad science, its bad statistics when looked at in isolation. The fact guys who didn't crossover did worse than guys who did is not conclusive supportive evidence for efficacy -- but it doesn't hurt the case.
As Dr. Mule pointed out, we have no idea how any of these works. If we wait until we do, we won't approve one of these drugs any time in the next decade or so. If 9902b shoots p<0.05, would you withhold approval because the MOA isn't precisely known? If the answer is no, then the MOA argument is moot.
Unless otherwise indicated, this is the personal viewpoint of David Miller and not necessarily that of Biotech Stock Research, LLC
