Bioequivalence and Bioavailability Forum

Dear KR!

❝ In EMEA guideline for two-stage design it is written that alpha is to be conserved.

... In EMEA draft guideline ...

❝ But nothing is clearly stated about CI (confidence interval). Will level of CI vary as per alpha

Yes, since the confidence interval is given by 1 – 2α.

❝ i.e. if we use Bonferroni's alpha spending function …

Some people say that Bonferroni is always the most conservative method (even in a stagewise procedure / no independent tests). No qualified opinion.

❝ … then do we need to estimate 95% CI for stage 1 and 90% for final analysis.

No. The overall α must be preserved at ≤0.05. For a two-stage sequential design Potvin et. al (2008) suggested α=0.0294 (according to Pocock’s method).

• In ‘Method C’ stage 1 α 0.05 or (!) 0.0294 – depending on intermediate power – and α 0.0294 in the pooled analysis.
  
• In ‘Method B’ α 0.0294 in both stage 1 and the pooled analysis.

For the reference and a flowchart see this post.
I would expect EMEA to give more details in the final guideline.