36 mos survival analysis is better than median survival analysis.
Instead of median survival data, I used 36 mos survival data, and my initial worries have been dampened a little by doing that.
I used the following data:
In 9901-9902a, a total of 82 patients received Docetaxel
We know from Dr. Small's presentation that 37% of the (82) D9901-provenge patients received docetaxel, i.e. 0.37 x 82=30 provenge-arm patients received docetaxel. We also know from the same presentation that 49% of the placebo-arm patients received docetaxel. This translates into 22 placebo-arm patients receiving docetaxel. Thus, in total, 22+30=52 patients from D9901 received docetaxel. Subtracting these 52 patients from the total of 82 (D9901 and D9902a) docetaxel receiving patients, we get that D9902a contributed with 30 docetaxel treated patients. We have no data that enables us to split these 30 patients between the placebo and the provenge arm, so lets assume for the moment that the went 50-50 to the to arms. Thus, D9902a contributed with 15 Provenge-docetaxel treated patients to the analysis, making the total number of Provenge-docetaxel treated patients in the two arms equal 30+15=45 provenge-docetaxel treated patients. We know that this group had a median survival that slightly conservatively equals 36 mos. Thus, the provenge-docetaxel treated group contributes with 45/2=22 or 23 patients to the group of survivors after 36 mos.
Out of 147 provenge treated patients, 33% were alive after 36 mos, i.e. 48 or 49 patients were alive at that time. 22 or 23 of these 48 or 49 survivors belong to the provenge-docetaxel treated group of patients, and consequently, the remaining 26 patients belong to the provenge-NO-docetaxel treated group of patients. These 26 patients who survived without the help of Docetaxel should be related to the total group of provenge-treated patients who did NOT receive docetaxel. This latter number is calculated by subtracting the 15 D9902a-provenge-docetaxel treated patients and the 30 D9901-provenge-docetaxel treated patients from the total of 147 provenge treated patients. This makes the survival rate of provenge-NO-docetaxel treated patients equal 26/(147-15-30)= 25.5%
This number (25.5%) is not too bad compared to the 9901+9902 placebo-arm survival rate of 15%. AND, there is reason to believe that the assumption of splitting the 30 D9902a-docetaxel treatments evenly between the placebo-arm and the Provenge-arm is too conservative (more placebos got docetaxel).
Assuming that the split is 10-20 (provenge-placebo) increases the 25.5% above to 26.6%.
FURTHERMORE, 15% survival in the placebo-arm (combined 9901 and 9902 arms) benefits greatly from the cross-overs who got docetaxel, so looking only at the placebo-crossovers-NO-docetaxel is likely to be quite a bit lower.
CONCLUSION, we may not be in such a bad shape after all...
AI
