Bioequivalence and Bioavailability Forum

Hi Simon!

❝ I'm not entirely sure why you'd estimate the area enclosed in the blue line - why is the line returning back to baseline at 9 hours?

❝ The AUC that the WNL NCA PD model calculates I've always considered analogous to AUCt because it draws that perpendicular line to the baseline at the last measured value.

I’m not sure whether AUEC is entirely analogous to AUC. Think about AUC (single dose) first. At t=0 the concentration should be ‘true’ zero (leaving homoeopathy and carry-over after washouts aside) and will return to zero at t≤∞. The theoretically best estimate of F would be the ratio of two AUCs extrapolated to t=∞. In relative (!) BA sooner or later absorption will be complete; sampling after this time point increases only the variance (concentrations approaching the LLOQ). In BE by convention (!) the ratio of AUC_ts is employed (if AUC_t ≥80% AUC_∞). Here the perpendicular drop makes sense because we have a true zero baseline – think about AUC as ∫C(t).
AUEC is different, IMHO. In Sandy’s example we start from a normalized baseline value (100%) and at the last measurement we may (!) still have a remaining effect (10% reduction). Unlike in PK in PD we often have to deal with highly variable measurements or observe a rebound (a negative effect or values above 100% in the example). In model 220 we assume a fixed baseline; fine. But then I would prefer to accept that the 9 h time point is not ‘the end of the story’ and extrapolate to the intersection of the regression line with the baseline (at 9.5 h). Only if I assume that the measured value at 9 h is pure random fluctuation I would use a perpendicular drop (or better: rise?) to the baseline. But of course this is pettifoggery.

❝ I think your second question is why doesn't WNL's NCA 220 model provide AUCbelow Baseline to inf. The short answer is I don't know…

❝ … but equally I'm not sure whether it could always be so easily calculated, what if the experiment had stopped at 5 or 6 hours and a 3 point regression would then continue downwards?

True, but does not speak against the method but against the design of the study. In PK extrapolation sometimes does not work as well (sampling stopped too early, long half life together with values close to the LLOQ).