"If I remember correctly, the Provenge trials missed the primary endpoint which invalidated all further statistical analysis and OS was a safety endpoint for the Provenge trial. Doesn't that make this quite different as IPI passed the primary endoint?"
No and yes.
The FDA considers OS in a terminal indication to always be an implicit primary of some sort (though they had no idea what the alpha should be). Thus the OS analysis was not invalidated.
The problem was you still had a non "pure" single trial which is a big issue. If 9902A had confirmed the OS (and ignoring the CMC issues) they likely would have been approved in '07.
But I do agree the cases are not as similar as DD seams to imply.
Doesn't that make this quite different as Ipi passed the primary endpoint?
If BMY were seeking to market Ipi as an adjunct to gp100, there would be no problem insofar as the gp100±Ipi OS comparison was the primary analysis in the phase-3 trial. However, BMY is seeking to market Ipi monotherapy, and the statistical support for this comes from a secondary analysis in the trial.
Despite the above, I think the FDA will probably approve Ipi in the second-line setting, but (as stated at the bottom of #msg-51218113) I would not be unduly surprised if the FDA ends up asking BMY to run a new trial testing Ipi monotherapy vs some comparator.