Dendreon Corp fka DNDNQ

[iwfal]

For the truly unbalanced (between the arms) co-variates, these ones are LESS LIKELY to emerge as signifant prognosticators (by univariate analysis), and so would never even get into the Cox model.

Aha! If they were doing univariate analysis this would be significantly true. But I believe that they are doing at least a two variable Cox Regression analysis as the filtering process for precisely this reason. The language in the chart is a little ambiguous, but what I assumed certainly fits and makes a lot more sense:

"Significant of each individual prognostic variable considered in a model along with treatment arm."

To judge from comment c) at the endI take it you meant to write that you tend to believe the Cox Regression is "gamed" ?

? I was saying that I do not think they were intentionally gamed because I am not that cynical. But I can't rule out fooling themselves - but I wouldn't expect a factor of 14 or whatever it is.