Was 9901 imbalanced for GS? and other questions.
The most commonly quoted reason for the analysts negative sentiments concerning the efficacy of Provenge is that the 9901 trial was only stat sig because the trial arms were not balanced. Specifically, they point out that the treatment arm had more GS<=6 and the control had more GS>=8. They beleive that since GS is inversely correlated with survival the trial was heavily biased and favored the treatment arm. My evaluation of this opinion is based on the following data.
% 3-year survival breakout by Gleason Score: 9901
GS<=6 Provenge arm: 41% or 9/22
GS<=6 placebo arm: 28.6% or 2/7
-----------------------------------------------
GS = 7 Provenge arm: 32.1% or 9/28
GS = 7 placebo arm: 11.1% or 2/18
-----------------------------------------------
GS = 8 Provenge arm: 40% or 4/10
GS = 8 placebo arm: 12.5% or 1/8
-----------------------------------------------
GS>=9 Provenge arm: 27.3% or 6/22
GS>=9 placebo arm: 0% or 0/12
Note: Could someone find the original source of these data? Not the original IV post. That's where I got it.
These data suggest that the treatment advantage increases with GS.
For GS<=7 18/50 alive with Provenge and 4/25 for control.
For GS>=8 10/32 alive with Provenge and 1/20 for control.
Additionally,
PC specific survival analysis was strongly in favor of the treament.
Sensitivity analyses checking for trial arm imbalances based on prognostic factors confirmed that the results were stat sig.
F_cking 9902A. What is the explanation for the poor stat results and lower survival times for 9902A? Without a reasonable explanation the 9901 results are unverified leading to much uncertainty. That's why the FDA didn't approve. Not some conspiracy theory.
With 500 patients the 9902B trial should be well balanced.
Questions:
1) Assuming a perfect balance between trial arms for all prognostic factors, does the Cox regression extend any additional statistical power? Or would it yield exactly the same p-value as for the log-rank test? Are we banking too much on the ability of the Cox model to deliver stat sig results at the interim?
2) The 9901 trial did not show a stat sig survival advantage until three years after the final patient was enrolled. At the interim look for 9902B, how many patients will have been enrolled for at least 3 years?
3) The patients in 9902B will be followed until death not for just 3 years as in 9901 and 9902A. This means that a patient, like Ed Garcia, who lives many years would be counted as living for all those extra months. Not merely the 3 years. Eight of the 9901 treatment arm survivor died after the completion of the 3 year followup. How long was their overall survival? What is the total survival of the 20 men who are still with us? And what about the cross-overs and placeboes? Are they still alive?
4) What is the PAP immune respsone index for these 9901 survivors after all these years?
5) What are the DNDN R&D scientists doing on a daily basis right now? They aren't involved in the trials or in the manufacture of Provenge for the trials. Is DNDN obligated to shareholders disclose specific R&D activity?
6) These guys with GS>=8 had some pretty aggressive tumors but still showed a enormous survival advantage over placebo. 9901 showed a stat sig effect on TTP only for GS<=7. The TTP curve for GS>=8 overlapped. Is this still further proof that the TTP endpoint is biased against Provenge?
7) Provenge isn't toxic to tumor cells and isn't supposed to be. Did the AC and do analysts really understand this point?
8) All but 1 out of 20 control patient with GS>=8 was alive after 3 years versus 32% (10/32) of Provenge treated patients. Why didn't DNDN point this out at the AC? Why doesn't someone familiarize these analysts and journalists with this GS survival breakdown?
All comments are appreciated.