Dendreon Corp fka DNDNQ

[ocyanblue]

Slide 17 of the ECCO presentation listed the covariates used and their p-values. Three were not stat sig: disease location, LDH and weight. Disease location had p=.07 so it was arguably significant for the purpose of doing Cox regression (for example, the Tax trial used .10 for cut-off).

So, if DNDN had gone by the same procedure as done for 9901, they would have found just three significant covariates (plus treatment). But the inclusion of the two non-significant covariates in the model would unlikely have helped improving the p-value for treatment. So Monique's confirmation of all variations of the analysis should be believable.

For completeness, the only possibility of foul play is if the ECCO presentation did not show some other stat sig covariates beyond the five. Given that the reputation of Dr. Higano herself (and that of other co-authors) is on the line, this is extremely unlikely.