HCP, there were a couple items in your post that I think warrant some clarification.
First,my recollection was that PPHM was indicating that it was the 1 mg/kg arm that was showing stat sig relative to the control arm. The post you were responding to suggested it was the Combined control arm that was stat sig, which is something that I don't recall PPHM has addressed in their communications.
Second, I don't believe it is correct to say that the 3 mg/kg interim MOS cannot improve as the status of censored patients changes. My post a little earlier today explains why.
Here is what you had written, which is a point that I had meant to address when you first raised it a few days ago, but was to busy at the time. My apologies for letting this float so long unaddressed.
"- Discrepancies occurrence is irrelevant for the 3mg arm because it was not affected.
- The 3mg arm MOS on 50% passing away can NEVER improve because its median comes out of an ascending ordered series (logic, your specialty!)
- The only thing that could make it look better would have to be that the median of the control arm shifts more to the left then the median of the 3mg arm.
- With the MOS of the control at 5.6, only 0.4 away from historical low of 5.2 and Bavi pollution of the control arm sample it is not impossible but if, then there is not much to gain.
- MOS in the control arm was reached long ago, since it is assumed PPHM tells us the good and hides the bad, they would for sure have given that info if it would have occurred in their Sept 7th PR.
So the answer is : NO, the 3mg MOS can't get better, at the best the control arm could get some worse. What troubles me is that by all posters it is someone that teaches logic that asks this question, so yes I may have been subconsciously been offering you the shoe to try out and you just did a quite convincing attempt to make it fit! "
My call out is that the MOS can indeed get better by showing longer survival for censored patients that event than survival indicated by an interim result-- if censored patients event over time where there are more patients that event "to the right" of the interim MOS value than to the left. Consider if there are 34 of 40 patients that evented and 6 censored such that the interim MOS is determined between patient number 17 and 18 in the order of patient survival posting. Let's assume that there are no censored patients in close timing to this patient number 17 and that the patients evented with the interim MOS result surrounding patient 17 are assigned letter designations E for interim evented patient 17 and sequencing earlier in the alphabet if evented earlier, later if evented later so that A represents patient 13 in the interim calculation and patient I (eye) represents patient 21 in the eventing sequence for the interim results.
The sequence for the interim result MOS is carved out for the "sweet spot" where changes in the MOS might be observed:
A, B, C, D, E, F, G, H, I
E is an interim MOS of 10 months and values shift by a Month for each letter shift
let's also assume that the eventing patients so labeled are each a month apart, with patient E surviving 10 months, patient D for 9 months etc. going to the left goes down alphabet sequence. Going to the right goes up alphabet sequence.
With the stage set, lets do a couple examples. Figure an example where the trial goes to completion and there are just 2 lost to contact or disqualified, making them permanently censored patients to the trial arm. Now the survival between patients 19 and 20 sets the final MOS, but which letter designation applies to that transition point within the sweet spot? The answer is, that it depends on where the previously censored patients were posted in the survival sequence. The MOS can remain with survival go patient E or move left or right, deep ding on when censored patients evented and when the permanently censored patients ended up.
Rather than making this post way too long by laying out all the combinations, I will describe one that can show how the MOS final can go longer than the interim determined value "E".
First, let's presume that the two permanently censored patients are spotted at timing before patient "A" for both the interim and final determination. That means that the early on censored patients don't affect the shift in MOS between interim and final since they are skipped over in. Oth analyses the same way.
Now, lets assume that all four of the patients that evented between when the interim and final MOS were determined shifted from being on the left of patient "A" to the right of patient "I". That means the addition of four evented patients to the pool would read the MOS final value two events to the right or for patient "G". Since patient "G" lived two months longer than patient E, who determined the interim result MOS, the final MOS value would post at two months longer. As my high school math teacher used to do when showing a proof for a theorem, we can write QED on the chalk board (that really dates me, LOL,).
My post earlier today about how the 3 mg/kg arm looks good relative to the control arm is because the display of evented and censored patients from Dr. Gerber's September 7 presentation shows the boundaries around which the MOS calculation can shift. Advantage stays with Bavi.
Best wishes and IMO.
KT
