NorthWest Biotherapeutics Inc (NWBO)

[kabunushi]

The statistical methods to try to account for crossover rely on a very big assumption which is that you can assign a standard value to how much benefit patients got out of the treatment. A big problem with this whole methodology is that it assumes that the benefit is the same for everybody and also that the benefit is uniform across time -basically that no matter how many months after surgery you start to get treatment the help you get from the treatment is exactly the same for every month you get the treatment and for every month thereafter. But in fact that might not be true at all: with this kind of treatment for this radical disease, there may be huge variation in how much patients benefit, possibly depending very much and varying a lot on how long after surgery they started getting dc-vax. I think it's very questionable if we can come up with a simple 'treatment effect factor', e.g. every month under treatment increases your OS by 2 weeks or whatever. It's a big assumption just to start with, that a simple treatment effect factor can be used to model actual patient responses.