Bioequivalence and Bioavailability Forum

Dear d_labes,

both your questions are quite closely related. I'll start with the futility bounds in sequential trials.

The question if "exceeding N_max" has any influence on type-I-error is solely a question of how that rule was specified when designing the trial. If it is interpreted as "test statistic at interim is too small" then you have used the statistical test to come to that conclusion and hence it will have an influence on your type-I-error. More precisely you have used a statistical test that gauges how likely it is that you will get a "positive" result in the end. Note that this is how futility bounds in sequential designs are usually derived and used.

If alternatively you have a situation were you decide to stop because of factors independent of the test statistic then it will not have an impact on the type-I-error as you have not performed a test. Examples of that include stopping a study because a competitor released a product that you don't think you can compete with, but also the senior manager that all of a sudden says: are you insane to think that we can do a trial with >N_max patients?

This is also the reason why doing a sample size review in a blinded fashion can be sensible. Note that I am saying "in a blinded fashion" meaning use techniques that do not use unblinded information not blind the statistician. If you use unblinded techniques most likely the test statistic comes in the mix in some form making it very tricky to argue that you have not also formally tested.

Hope that is helpful.