Bioequivalence and Bioavailability Forum

Hi Darborn,

❝ When calculating AUC_0–∞, I know some regulatory agencies use extrapolation while others use the actual value. For a long half-life pro­duct truncated at 72 hours, should I use AUC_0–t or AUC_0–72 for bioequivalence assessment? 

If it is an IR-product with a half life ≥ 24 h, AUC_0–72 (see this article about ICH M13A and the linked references).

If t_½ < 24 h, the primary metric is AUC_0–t. You have to estimate AUC_0–∞ as well but only to assess the extrapolated fraction (AUC_0–t should cover ≥ 80 % of AUC_0–∞ in ≥ 80 % of observations).

❝ I think these two value may differ when 72h has a LLOQ.

Of course. See this presentation (slides 19–24).

Hi Darborn,

❝ When calculating AUC_0–∞, I know some regulatory agencies use extrapolation while others use the actual value.

I am not sure I understood this sentence.

❝ For a long half-life pro­duct truncated at 72 hours, should I use AUC_0–t or AUC_0–72 for bioequivalence assessment?  I think these two value may differ when 72h has a LLOQ.

When you intend to evaluate BE by truncated AUC, like AUC_0-72 then you have some options, and I imagine this is the reason you ask.

1. For a profile with BLQs at the end you can do
1a. Let AUC_0-72=AUCt
1b. Extrapolation, which can be linear or log-based from the least measurable points in the elim phase.
I have seen both 1a and 1b accepted recently. Funnily enough if you do linear extrapolation, you may get negative concentrations on the extrapolated part depending on how ugly your data is. I have seen it accepted nonetheless.
2. For a profile where the last sample is not exactly 72hrs but measurable (ambulatory subjects are particularly unreliable when it comes to showing up on time), you can do:
2a. Let AUC_0-72=AUCt
2b. Inter- or extrapolation, which can be linear or log-based from the least measurable points in the elim phase.
This also seems to be accepted as far as I can tell.

At the end of the day, in BE we do not really care much if a profile can be said to be biased, rather we just want T/R estimates to be unbiased or at least have no arguments in favour of bias on the ratio.
So, as long as all profiles are treated in the same fashion, it is difficult for a regulator to argue that the ratio end up being biased. Therefore the various flavours and colors of AUC_0-72 are accepted even if something looks odd in someone's opinion from the perspective of a single profile.