Bioequivalence and Bioavailability Forum

Dear Jag!

❝ […] what is the proper way to determine Kel from the elimination phase, based on noncompartmental analysis, if the last measurable timepoint has a higher concentration than the adjacent timepoint?

Proper? Two pharmacokineticists = ≥3 methods.

❝ Should Kel be labelled as "Undefined"? Some CROs would say "Undefined" …

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).

❝ … but there are others who would determine Kel as long as the linear regression result based on the timepoints involved (at least 3) is acceptable.

Others = most, IMHO. As long as you are consistent, perform the estimation of λ_z blinded for treatment, and have an SOP in place it shouldn’t be problematic.^* PK software by default mark excluded datapoints both in the plots and tabulated output. A moderately higher last value is not that uncommon, especially if the last sampling interval is relatively short compared to the half-life. Just think about acceptable analytical (in)accuracy and (im)precision: 15% within the calibration range and 20% at the LLOQ.

Example: t_½ 8 h, last two sampling points (interval = t_½)
    theoret.  ±inacc/±imprec.  obs.
16    25         -2× 15%       17.5
24    12.5       +2× 20%       17.5
Even if the method is performing within specs, two values with deviations of opposite direction already lead to equal concentrations. A little bit more random noise and voilà!

---

• A friend told me that he once saw an SOP essentially stating: “Select datapoints in such a way that the residual AUC is <20% of AUC_∞.” Of course that’s not an outstanding clever idea.