Replies to post #45081 on Biotech Values

[iwfal]

It is bad science to analyze subset data from a trial, especially one with small patient enrollment numbers.

I agree. You can't look at xover survival vs placebo wo xover. Self selected subgroups are completely unreliable so saying that this data proves something is silly.

But by the same token Sher is essentially claiming that any trial with xover is unacceptable. And that is going against a LONG history of FDA precident. There is some truth to it - but the FDA long ago chose to say it was acceptable for ETHICAL reasons. Too bad. Get over it.

BTW - There is plenty of precident for survival being given its own alpha. And for single trials being acceptable. Is it mixed? Yes, more so than xover, but all of this sturm und drang is a little overdone. This is not ground breaking precident.

[mouton29]

Obviously there was no "subset analysis" for the basic overall survival analysis. That was statistically significant whether you examine 9901 or 9901 and 9902A combined. David's argument in rebuttal to Scherr's argument that the placebo itself mightt have a negative effect is a subset analysis, but given the design of 9902B -- the subject of an SPA -- I suspect we will never run a trial of Provenge v. inert placebo. If the FDA approves, obviously that won't happen, and if they require waiting till 9902B results are in.