Bioequivalence and Bioavailability Forum

Hi ElMaestro!

❝ ❝ […] OK, the framework showed that α is maintained. Of course this does not imply that skipping this step will not work at well. We just don’t have any evidence (i.e., a published reference). I don’t have the means to set up simulations but would be a rather interesting task. ElMaestro?

❝

❝ Yes, I can certainly set up some new simulations if necessary although this will be a spare time project and may take a lot of time. The author of the crackpot paper used at least what corresponds to two full months work time just to do the necessary programming, so I've heard.

I can offer you just a beer – and fame!

❝ Probably they just mean that classically, power is not associated with regulatory risk (patient's risk) and therefore it isn't what their concern is.

Yep.

❝ But for 2-stage designs, one may argue exceptions to this true general rule exist.

I would say so. They mixed up post hoc power with intermediate power.

❝ I am bewildered though, I am not sure what to simulate. If you guys can propose an Al Gore Rhythm (please make a graphic) then this would be a good starting point.

Somefink similar to what Detlew suggested last year (but α_1,2 0.0294 instead of α₁ 0.005 and α₂ 0.048).

❝ Is it possible to write one which does not involve intermediary power steps? I mean, at the end we do need to think about how many subjects to include at stage 2 and that means power doesn't it (and don't we)?

See also Detlew’s comments above.

---

Edit: See the simulations above.