Bioequivalence and Bioavailability Forum

Helmut

Thank you very much for such a comprehensive reply. You will recall that my statistical knowledge amounts to less than a pimple on the backside of most users of this forum, so please excuse me if what I have written below seems garbled.

Firstly, it seems from your response that the method of Potvin, although having the drawback you mention, has the advantage of allowing you to calculate the size of the second group by using the data from the first group. This is logical. I had assumed that the regulators would be illogical and that, for a two-stage design, they would expect us to pre-specify the size of both first and second groups. Thus, we have planned a group sequential design with a first group of 24 to be followed by a second group of 24 if the interim analysis indicates the possibility of passing with 48. The idea would be to adjust the confidence intervals according the method of Gould 1995 (ie. as we plan two groups of equal size, we would calculate 93.9% confidence intervals at interim analysis).

I'll confess that I do not fully (or even slightly) understand the method then used to decide on stop or proceed; I was relying on my statistican to work this out from the Gould paper.

In your view, is Gould's approach unlikely to be acceptable to the regulators?

Kind regards

Andrew