Bioequivalence and Bioavailability Forum

Hi John,

❝ I was reading a protocol last week and see this "subject data to be excluded from BE stats if auc/aucinf > 20%". Can one do that at all??

<nitpicking>

You mean the extrapolated part = 100×(AUC_∞–AUC_t)/AUC_∞ >20% or
the covered part = 100×AUC_t/AUC_∞ <80%.

</nitpicking>

Ha-ha! Of course some “clever” CROs love to exclude subjects → decreased power → study fails → perform a new study → earn more money…

Is this an IR formulation? If yes, absorption is essentially completed at 2–4 times t_max. For the FDA you can exclude AUC_∞ and go ahead with AUC_t (though I don’t think FDA requires that; the ≤20% residual area is EMA-style and not mentioned in any of FDA’s guidances). I would never exclude all data of a subject. At least I would keep the subject’s C_max – which is due to its higher variability the win-loose metric anyhow.
If this is an extended release formulation, AUC_∞ is more important than AUC_t since we may run into flip-flop PK and want to compare absorption. However, if you opt for exclusion of AUC, still keep C_max.

Just for completeness, EMA’s GL:

[…] AUC_(0-t) should cover at least 80% of AUC_(0-∞). Subjects should not be excluded from the statistical analysis if AUC_(0-t) covers less than 80% of AUC_(0-∞), but if the percentage is less than 80% in more than 20% of the observations then the validity of the study may need to be discussed.