NorthWest Biotherapeutics Inc (NWBO)

[mattpk]

What exwannabe is getting at, and he is entirely correct, that counting survival from any point other than randomization introduces a well-known statistical fallacy known as "the survivor's bias".

Randomizing all patients with a disease at the instant of diagnosis is fair game, and randomizing all patients at a later time point is also fair game, but randomizing at a later point in time and then including the time since diagnosis is what creates the bias. In any trial some patients are "sicker" than others and some percentage of these succumb to their illness early on, leaving those patients who suffer from a less advanced or milder form of the disease to be randomized later. Because these patients are inherently healthier (although still afflicted with the underlying disease) that creates a bias when comparing to historical data with different randomization points.

Good practice is to randomize all patients at a particular time and to initiate the next step of treatment immediately thereafter. It is OK to do it at step 1 or step 3 or step 5 or whenever it makes the most sense clinically, but survival time is counted from the day of randomization and no earlier.