I wouldn't worry if overall survival is the primary or secondary endpoint. As long as it is a prospectively defined endpoint and the arms are well balanced through randomization, the results should be valid.
But I hate Dendreon's use of the word "similar" in the press release. Whereas they note that TTDP did not show a stat sig difference, they claim the survival compared to placebo is "similar" to the 9901 trial. But what the heck does that mean? A human and an elephant are similar if you consider that both are mammals. If I remember correctly, 9901 1) failed the ITT analysis at first but showed a subgroup survival benefit; but 2) over time, the ITT analysis become significant. So are we similar to 1 or 2?
Without clarification, it is hard not to conclude that the survival analysis failed to meet statistical significance despite their dancing around this point. And for a trial with 98 total patients, it is reasonable to accept that the trial was simply underpowered to show a statistically significant benefit. An honest PR would have been the best and most reasonable policy in this case.