Replies to post #2329 on Dendreon Corp fka DNDNQ

[iwfal]

Not only was it not prespecified (despite what mitch said), DNDN never gave the impression that the Cox analysis was done on something other than an ITT population. We now know that ~10% of the patients were thrown out, in a manner that was favorable to the provenge arm.

That isn't actually a surprise to me per se (i.e. I knew that they couldn't have prespecified all the knobs on Cox Regression and this is just one of the knobs - and human nature is you will tweak the knobs to get the right result, even if 'unintentionally'.)

What I find more disheartening is the misleading that went on in saying that the filtering methodology was prespecified. When clearly it was not.

Clark

[io_io]

<<<<<"We now know that ~10% of the patients were thrown out, in a manner that was favorable to the provenge arm.">>>>>


You know I read the Docs again last night, and you could say that my confidence moved up from 60% to 75%.


Your point PGS here has at first glance some validity, except that its a form of data-mining. There are so many variables that go into a trial and its measurement, and its the reviewers job to point out any suspect issues.

You seize (thats your nature, I accept) on the suspect issues, and you even say that the 10% of the patients were "thrown out".

"Thrown out" ?!? What are you implying ????


In any case, my larger point is that the reviewers were bound to find something to point to. Its too easy to say "maybe this, maybe that". If all the variables ran parallel, there would not need be panels, nor even review teams.