NorthWest Biotherapeutics Inc (NWBO)

[Pyrrhonian]

As long as the placebo is indistinguishable from tx and the treatment assignments remain blinded it's not easy at all. Most type one error comes from too much interaction and information sharing between the DMC and sponsor (even unintended things like body language).

Keeping management far away is def best. An appointed steering committee, which normally involves one or a few of the sponsor's staff, can act as an intermediary.

Any time data is unblinded there would be what is called alpha spent. Each time this occurs the p-value goes up, as bias is always potentially introduced when data is unblinded. Here is a good discussion on the topic: https://www.ma.utexas.edu/users/mks/statmistakes/errortypes.html