NorthWest Biotherapeutics Inc (NWBO)

[longfellow95]

45.9 - 94.5 is the 95% confidence interval but does not mean what many think. Even if all the data exists there will still be a wide CI because it is a measure of what the underlying distribution being sampled is. It is not an indicator of the error estimate on K/M vs the trial full dateset median.

Wrong!!
The CI just means there is a 95% chance of the eventual median being within the range.

How can you say:-

Even if all the data exists there will still be a wide CI because it is a measure of what the underlying distribution being sampled is.

And:-

What I mean is this. Say the trial comes in at 61 months. The K/M would converge on that as you get more data. And with all data it would be exact.

You are contradicting yourself.


What we can conclude is that the JTM CI (35.5 - 46.5) was way off. As the survival of the Top 100 membership defied the odds in the intervening year, the new CI by SNO was 45.9 - 94.5.
Which as good as tells us that there is a pronounced long tail of survivors and we don't have proportional hazards.

When the final membership of the Top 100 survivors is fixed, and 51 of those are dead, the median OS will be fixed.
I'm not sure the membership is yet completely fixed. We have to definitively know what the shortest survival time is to get into the Top 100.

Both CI's were no doubt based on proportional hazards and, as we know, they don't exist in this trial.
If the assumed long tail survival is as long and as fat as LL's statement suggests ('majority of the 3yr survivors haven't even progressed' that would strongly suggest that the median OS will be considerably higher than 58.4. Could easily be between 70 and 80.

And there almost certainly is a time bias. I thought that it was generally accepted that 2015 entrants are doing better than their predecessors.
Several possible reasons for that.


Mind you, none of the above tells us precisely what OS60 will be, nor (of course) what the delta will be.