NorthWest Biotherapeutics Inc (NWBO)

[kabunushi]

Just knowing '100 still alive' isn't enough to predict results. We also need to know, as a very key number, how many of those alive are from the placebo group and how many treatment. Mathematically worst case, all 100 could be placebo. Practically speaking, even if 33 then the result looks like pure chance.

The split of placebo-to-treatment is key. Out of the 100, 80-20 treatment is probably a very clear great result. 75-25, I'm not sure. I don't know at all what kind of p value you'd get for that. But even a difference of 4-5 patients probably swings the p value a lot. Without the split number, just the 100 number is not all that meaningful, imo.

Note also that the median is the key statistic, that is the OS of the 100-101 longest surviving treated patients. That assumes zero lost to the study. Depending on how many patients were censored the 100 survivor data becomes more important. But we don't know that number and we don't know how many of the current survivors are from the placebo arm, and without that the blended number is not as interesting as you seem to think, imo.

[in edit] the longer the blended OS, the more likely it gets that the number of placebo survivors is very small -- we have a decent idea on the curve for SOC. My point though is that a small number diff in the split can move the p vale, and you didn't address the split statistic at all.