Well, when I use averages, and use the medians the investigators had hoped to achieve, it seems they work out to the PFS numbers being reached, based again, on averages, at the same time the company announced the number was reached.
One would hope that the investigators were basing those median PFS and OS numbers off of what they were seeing in an unblinded setting. So when one looks at them through the lens of averaging those months, versus looking for the middle number, they worked out to the same month.
That could have been a coincidence, of course. And there are likely other averages that could be applied to both the treatment and control arms to arrive at that 11.3 blended number. For instance, this:
82 x 9 PFS months = 738
165 x 12.5 PFS months = 2062.50
738 + 2062.50 = 2800.5 / 248 = 11.29
Or
82 x 10 PFS months = 1100
165 x 10.3 PFS months = 1700
= 2800 = 11.29
However, you can see how the treatment and the control numbers start to draw together - the control PFS has to come up, and the treatment PFS has to come down.
And I find it quite doubtful that the control patients in this trial could eek out an average of 9 or 10 months PFS from randomization, which is 12 or 13 months PFS from diagnosis. However perhaps a bear can come along and demonstrate how that is very possible.
I know you know this... but to be clear to others... these control patients would not have received DCVax-L prior to eventing. Of course, they only receive it after an event diagnosis. So there would be no confounding - except for, of course, if some treatment patients were initially showing a false progression that was initially ruled progression.
Anyhow, how long the adjudication process took is unknown. I can't remember exactly when the company indicated they were looking at scans, etc. (I'd have to look), but they could have still been in the process by February 2017. Or not. I don't know.
Still... this is the first time I've done those numbers with PFS. I've only done averages with OS numbers. And I found it heartening that the numbers the investigators chose that would make the trial stat sig, fit like a glove to the time that the PFS events were announced.
I could have missed something, or not thought of something. I'd imagine you, ex, Avii or a few others more mathematically inclined can tell me what that is. And these numbers aren't fail safe. They don't prove the trial will be stat sig. But they help point out that it very well might be. Which as you would know, is something that has concerned me with regards to PFS. And so I was pleasantly surprised that the numbers worked out the way they did.
