The impact of using the well-stirred liver model in PBPK modelling for high-clearance compounds
Ruben de Kanter (1), Shirin Bruderer (2), Jochen Zisowsky (2), Jasper Dingemanse (2), Andreas Krause (2), Alexander Treiber (1) and Christopher Kohl (1)
(1) Drug Metabolism and Pharmacokinetics and (2) Clinical Pharmacology, Actelion Pharmaceuticals, Allschwil, Switzerland
Objectives: The well-stirred liver model is a widely used standard in Physiologically-based pharmacokinetic (PBPK) modelling. This analysis aimed at assessing the food effect observed for a high-clearance oral drug, assumed to be caused by a change in hepatic blood flow due to the presence of food.
Methods: PBPK modelling was performed using GastroPlus v.7 (Simulations Plus Inc.). Healthy human physiology was applied for all variables, including blood volume and flow values. In fed state (GastroPlus allows entering fed or fasted but no gradual switch between states), the liver blood flow was increased by 25%. Volume of distribution and tissue-to-blood partitioning were predicted. The clearance was fitted to the observed human plasma concentration-time profile after a single dose in either fasted or fed subjects.
Results: Clearance was estimated as 12.5 mL/min/kg (fasted) and 15.6 mL/min/kg (fed), volume of distribution was predicted to be 283 L. The observed food effect was a 15-fold increase in both Cmax and AUC. The observed concentration-time profile could not be described by the PBPK model for both fed and fasted state while keeping the intrinsic clearance (CLint) constant (which is assumed to be the case). Additional analyses using either the well-stirred and parallel tube liver models were performed to find a constant value for CLint in fasted and fed state that better described the observations. The parallel tube liver model, but not the well stirred model, was able to characterise the observed data substantially better, assuming the food effect is caused by a change in liver blood flow.
Conclusion: A limitation of the well-stirred liver model is that, for high-clearance compounds, the effect of a changed blood flow due to food is not correctly translated into a change in extraction ratio. This artefact, caused by an inappropriate model assumption, should be kept in mind when using commercial PBPK software packages that have the well-stirred model implemented as the only option.
