Replies to post #46232 on Biotech Values

[iwfal]

One doc asked if his patients are going to respond as in 9901, or as in 9902A because for the latter he wouldn't adjust his treatment strategy for patients.

Well, that would be irresponsible. From a clinical standpoint the 9902a provided as much benefit as Taxotere. With less side effects.

Yes, I am being a little flip - but there is this strong, and irresponsible, tendency to view any trial that didn't hit p as complete failure, or even more strongly proof of lack of efficacy. When in fact it is fair to say, as the FDA stats reviewer did - that all of the endpoints were on the right side of HR=1.0. The p value should come off the totality of the data. And so should the HR. The odds that Provenge is a dud are pretty low - certainly lower than the 1 in 20 that you and David are bantering about. Again, when asked to speculate what the HR of 9902b would be in a known sicker population you speculated it would be 1.23 (if I remember correctly).

Clark

[gofishmarko]

>>> docs are reluctant to experiment on their patients by abandoning their effective treatment strategies for the latest small trial that showed a benefit. <<<

I have no doubt you're correct about that , but my sense is that cancer patients might come to a different conclusion than their docs , given all the information and the opportunity to weigh in.

For example , given a choice of the following two treatments , identical in all ways except presumed treatment effect and statistical robustness supporting that effect , I'm convinced that many docs and all statisticians would choose #1 , while most patients and other sane people would choose #2 :


#1 ) 2 months of increased survival with a 99.999% probability.

#2 ) 10 months of increased survival with an 80% probablility.


Investors , while not known for their sanity , would also choose # 2 , since it represents a "probable present value" of 8 months compared to only 2 months for #1.