The impact of unmodelled interoccasion variability in bioavailability and absorption on parameter estimates in population pharmacokinetic analysis
S. Koehne-Voss (1), A. Gautier (1), G. Graham (1)
(1) Novartis Pharma AG Basel
Objectives: In their landmark paper Karlsson and Sheiner [1] showed that ignoring interoccasion variability in pharmacokinetic data can lead to biased population pharmacokinetic parameter estimates. They considered intravenous administration models in their simulations. We study the effect of unaccounted interoccasion variability in drug bioavailability and absorption on parameter estimates.
Methods: Simulations were performed in which data were generated from a one-compartment model with first-order absorption. The model included independent random effects for absorption (ka), apparent clearance and volume (CL/F, V/F), and bioavailability (F). Interoccasion variability between two occasions was present either in relative bioavailability (F) or absorption rate (ka). Two study designs were considered. In the first design single dose data were generated for occasion 1 and steady state data for occasion 2. In the second design steady state data were simulated on both occasions. Dense and sparse sampling strategies were considered. Simulated datasets were analysed with NONMEM V7.3 using the FOCEI method and an analysis model that matched the data generating model or an analysis model that matched the data generating model but with interoccasion variability not accounted for.
Results: Our simulations show that in the one-compartment model with first-order absorption not modelling interoccasion variability in ka leads to overestimation of ka and V/F. Estimated intersubject variability for ka is inflated. Estimates of CL/F seem to be less affected by unaccounted interoccasion variability in ka.
Unaccounted interoccasion variability in F can lead to biased estimates of apparent clearance and volume, although the bias was mostly small in our examples. Estimated intersubject variability can be seriously overestimated for CL/F, V/F, and F, with the bias for CL/F and V/F being more pronounced in designs that combine single dose and steady state dosing compared to designs with steady state dosing only. Intersubject variability for ka can be seriously underestimated.
Conclusions: When sampling pharmacokinetic data on several occasions interoccasion variability in absorption and bioavailability should be included in the model to avoid potential bias in population pharmacokinetic parameter estimates. In this simulation example bias was more pronounced in random effect parameters compared to fixed effect parameters.
References:
[1] Karlsson MO, Sheiner LB. The importance of modeling interoccasion variability in population pharmacokinetic analyses. J Pharmacokinet Biopharm (1993) 21(6):735-50.
