But the Provenge clinical data is understood to be really good. Didn't the drug improve survival in prostate cancer patients by 4½ months compared with men who got a placebo? I also understand the safety profile of Dendreon is exceptional. Yes, that's correct. The survival benefit and its low toxicity are Provenge's biggest selling points and could very well persuade experts on the March 29 panel to recommend approval. But the reliability of that survival benefit isn't crystal-clear. Huh? There's nothing clearer than survival in a cancer study. A patient is alive or dead, correct? True, but the Provenge phase III trial was designed to measure time-to-disease progression -- not survival -- as its primary endpoint. And on this score, the trial was a statistical failure. The measurement of survival was a secondary efficacy endpoint in the study, so while it was statistically significant, the validity of this finding is not a slam dunk. Even the authors of the Provenge clinical study, published last year in the Journal of Clinical Oncology, take a cautious approach: "While overall survival at 36 months was prespecified in the analysis plan, it was not specified as the primary efficacy end point, so the P value obtained for this analysis should be interpreted with caution," the authors wrote. A P value is a statistical measure of confidence that an outcome is the result of treatment, and not random chance. This is old news. Dendreon did its own analysis to confirm Provenge's survival benefit, correct? Indeed, and much of what Dendreon did is pretty convincing. For instance, skeptics had long thought that there were imbalances and other factors between the two arms of the Provenge study that favored the drug over placebo. This would explain why Provenge patients lived longer, skeptics believed. But when researchers crunched the numbers, the truth was actually reversed. There were few imbalances in the study's two arms, and in fact, there were several factors that actually worked against Provenge.