Bioequivalence and Bioavailability Forum

Dear John!

❝ ❝ That would mean excluding the subject’s AUCs (AUC_t unreliable, AUC_∞ undefined). At least you could keep the subject’s C_max in the data set (which is more variable).

❝

❝ Actually that particular CRO (maybe there are others) kept the subject's Cmax, Tmax and AUCt but assigned AUCinf, kel and t1/2 as undefined or "N.C."

That doesn’t make sense from a (European) regulatory perspective, IMHO. For EMA the primary metric for extent of absorption is AUC_t, but you have to demonstrate that it’s a reliable estimate: AUC_t/AUC_∞ ≥0.8. For IR products this does not make sense, since absorption is expected to be already complete after 2–4×t_max. After this cut-off point estimates are stable, only variability increases (see papers by Endrényi, Midha, Tóthfalusi).
Interesting enough FDA (where both AUC_t and AUC_∞ are primary metrics) no clear statement is made about AUC_t. We find only:

[…] we recommend that 12 to 18 samples, including a predose sample, be collected per subject per dose. This sampling can continue for at least three or more terminal half lives of the drug. […] The sample collection can be spaced in such a way that the […] terminal elimination rate constant (λ_z) can be estimated accurately. At least three to four samples can be obtained during the terminal log-linear phase to obtain an accurate estimate of λ_z from linear regression.

It seems that AUC_t can be kept in the analysis.

❝ One more question, what if the last two points of the elimination phase is going in the positive direction and the points before them (lets say 3 points, yes after Cmax) are declining?

Good question, next question. For the FDA I would trigger reanalysis (pharmacokinetic repeat) – SOP in place, no cherry-picking (blinded for treatment), justification in the report which value was used in the final analysis.