NorthWest Biotherapeutics Inc (NWBO)

[longfellow95]

Well, it was supposed to be a third. But at some point during the long trial, the ratio became a bit skewed, and ultimately it ended up being 29.9% instead of 33.3%.
This only came to light when they published the JTM interim report on 29 May 2018.

https://translational-medicine.biomedcentral.com/articles/10.1186/s12967-018-1507-6

And since then, this randomisation skew has been much discussed, as to how and why it came about. And there is no definitive answer.

It has been suggested that after the partial hold on recruitment was imposed in Aug 2015, when they had already randomised between 300 and 310 patients, that they were allowed to continue to randomise those already in the screening process, but only to treatment.
Thus they managed to get to 331, with 232 being in treatment.
The original target was 348. And two thirds of 348 is 232...
So they fully recruited to the intended treatment arm number.
But the control arm should have got to 116, and only got to 99.
But the skew could have been there before the partial hold.
We just don't know.

No end of reasons have been put forward as to why this might have came about. But nobody actually knows.

On the crossover issue. Patients can only cross over after having been deemed to have disease progression. Disease progression is the only trigger to cross over. (Though some of those progression decisions might have been pseudo-progression mistaken for actual disease progression.)

And patients got randomised after concurrent chemorad, which takes 3 months from surgery. If they progressed (or pseudo-progressed) during that 3 months, they were not randomised into the main trial, but were entered into the information arm, so they still got DCVax treatment.

And you are completely right that we don't have a 'clean' comparison between control and treatment on overall survival, seeing as two thirds of the control arm went on to receive DCVax after crossing over after progression. And the 30-33 who didn't crossover can't really be see as a representative control.
Because individuals in this group were more likely to be 'iller'.
Perhaps going straight to hospice after disease progression, and not wanting to cross to DCVax treatment. Or they were more likely to be the ones who 'jumped ship' from the DCVax trial, perhaps joining a different trial for patients with recurrent GBM, or just going missing completely ('Lost to follow-up').
And 30-33 is way too small a comparator anyway.

And that is why it was essential to carry on the trial to an unusual degree of maturity. To demonstrate that even if the arm by arm comparison has some confoundment (due to crossover), the survival of the treatment arm is wholly exceptional.
Then you get into the vexxed question about whether comparison with 'historical controls' is valid and / or acceptable to a regulator. No straightforward answer to that one. Other than if survival is demonstrated to be 'out of the ballpark' good, then it becomes harder for a regulator to say no.
The 'historical control' could be a similar treatment population in a similar trial (Optune is the obvious example), or simply the wider GBM population outside of a clinical trail. But that has its own issues because the ordinary GBM population is older on average and very much includes those who progress rapidly, or are immuno-suppressed.
Basically, it will be a very helpful thing if DCVax-L treatment results comfortably surpass Optune treatment results, especially at all the milestones beyond 2yrs. So that means percentage survival at 2yrs, 3yrs, 4yrs, and 5yrs.