Replies to post #132695 on NorthWest Biotherapeutics Inc (NWBO)

[exwannabe]

More events were posted in this model as early as possible for each graph, in order to be as to be as conservative as possible.

You start by modeling the NWBO placebo arm per IMUC. But instead of a true model, you forced the events to create a "conservative" [shorter time to events] curve for that arm.

So now you have less censors on the placebo arm than a true model would have. When you subtract the numbers from the known overall data, this obviously increases the number still alive on the treatment arm.

In short, make a "bad" placebo arm estimate. Subtract from the blended data to get treatment arm. And the treatment arm will look good.

How about try this for run. Run your exact same methodology but force the censors on the placebo arm to be as old as possible (earliest recruits). That will look real ugly.