Bioequivalence and Bioavailability Forum

Hi Ohlbe,

THX for the update!

Regrettably I couldn’t find the Nov 2018 guideline any more. It stated that “the variabilities associated with each product should be assessed”. I was always asking myself how that should be done.

• In Population and Individual Bioequivalence the \(\small{s_\textrm{wT}/s_\textrm{wR}}\) ratio was assessed and ‘similar’ variability was concluded for a ratio within 0.667 – 1.500. However, the power of comparing variabilities in a study designed to compare means is low.
  
• An alternative approach is given by the FDA in the warfarin guidance, where variabilities are considered ‘comparable’ if the upper confidence limit of \(\small{\sigma_\textrm{wT}/\sigma_\textrm{wR}}\) is ≤2.5.

This recommendation was dropped. Fine.

However, I don’t get why only a four-period, full replicate design (underlined in the original) will be accepted. Even if one compares – for fun – the variabilities of T and R, a 3-period 2-sequence full replicate design would do as well. Actually the estimated variabilities are more reliable in the latter, since for equally powered studies their sample size is higher. The confidence interval of the \(\small{CV}\) depends on its associated variance based on the \(\small{\chi^2}\)-distribution with \(\small{n-2}\) degrees of freedom, and, voilà, it will be narrower (see the end of the section there).

Or a bit provocative: Since a comparison of variabilities is no more required, why is a partial replicate not acceptable (as it is for C_max)? The estimated \(\small{CV_\textrm{wR}}\) is most accurate of all designs (I don’t like it but that’s another story).