1. Unadjusted log-rank: HR 0.766 with p value 0.023.

2. Cox-adjusted: HR 0.775 with p value 0.032. 

3. Adjusted for Taxotere and timing: HR 0.763 with p value 0.036.

4. Cancer-specific survival:  HR 0.772 with p value 0.036.

5. Integrated data D9901+D9902a+IMPACT: HR  0.735 with p value 0.001. 

6. TTP again failed with p value 0.628. Not much of a surprise here.

7. 36-month survival for treatment arm consistent across all three trials at about 33%. 

8. 36-month survival for control arm  went up slightly to 23% from 21% in D9902a.

9. No serious safety issues. 

 

The survival data are remarkably consistent across all three datasets, D9901, D9902a and IMPACT. The Cox model, in fact, worsened the p value of IMPACT somewhat over the unadjusted log-rank data. But the Cox p value for the integrated data, D9901+D9902a+IMPACT, was a remarkable 0.001. That is, the chance of that dataset being a fluke is less than one in one thousand! This is a clear and unequivocal proof that Provenge works!