A simulation study to investigate the identifiability of parameters in a minimal PBPK model structure with target binding
Irene-Ariadne Kechagia, John M. Harrold, Juan Jose Perez-Ruixo, Aris Dokoumetzidis
School of Pharmacy, University of Athens, Greece; Amgen Inc, Thousand Oaks, CA USA
Objectives: To use simulations for investigating the ability to estimate the parameters of a minimal physiologically based pharmacokinetic (PBPK) model structure with target binding.
Methods: A minimal PBPK model of target engagement was constructed to represent localized inflammation. Most organs were lumped into either tight or leaky compartments, while a third compartment was constructed to understand target binding in a diseased tissue. Target was assumed to be primarily synthesized in the interstitial space (ISF) of the diseased tissue, but also, at a lesser extent in plasma and the ISF of other tissues, eliminated systemically from the plasma space, and distribute in the tissues through lymphatics, both uptake and recycling. Drug distributes into the compartments and binds to target in all compartments, while it is eliminated from plasma. The drug-target complex can distribute among the compartments and is eliminated from plasma. The parameters to be estimated were: Drug and complex plasma clearance; target plasma clearance and initial plasma concentration; and binding parameters. Two scenarios about the target half-life were considered, a short half-life (30 minutes, scenario A) and a longer half-life (6 hours, scenario B). The identifiability of the parameters was examined by two methods: (a) By simulating concentration time profiles with the model and attempting to estimate the desired parameters. The relative bias (RBIAS) and the standard errors (RSE) of the estimates were calculated in order to assess the accuracy and the precision of the estimation (method EST). (b) By evaluating the Fisher Information Matrix (FIM) for the true parameter values and calculating the RSEs of the parameters (method FIM). Two datasets were used: (a) with two outputs, i.e. total drug and total target in plasma, (b) with three outputs, i.e. same as in (a) plus the complex in plasma.
Results: The two methods EST and FIM produced similar conclusions, i.e. low RBIAS of EST corresponded to low RSEs in FIM and vice versa. For Scenario A, drug clearance was estimated with high precision (RSE<2%, RBIAS <1%), target parameters (RSE<20%, RBIAS <5%) and complex clearance (RSE<23%, RBIAS < 17%) were reasonably estimated but binding parameters were not (RSE>48%, RBIAS <130%). Three outputs gave better results than two, especially for target parameters and complex clearance. Similar results were drawn for scenario B (long target half-life).
Conclusions: In a minimal PBPK model structure including target binding, most parameters except the binding parameters can be estimated reasonably.
