Influence of arterial vs. venous sampling site on nicotine tolerance model selection and parameter estimation

In this modeling study we utilize previously published nicotine pharmacokinetic (PK) and pharmacodynamic (PD, heart rate) data to investigate the influence of PK sampling site (venous vs. arterial) on the selection of a specific PD tolerance model and estimation of its parameters. We describe a general model for tolerance which includes as special cases feedback (TF), and kinetic based tolerance (TK) models. A TK model has arterial plasma drug concentrations (Ca) driving (hypothetical) effect (Ce) and antagonist (Cm) site concentrations, which drive a non-feedback effect (Enf): tolerance depends on the relative rate of equilibration of Ce and Cm with Ca. The TF model adds feedback which makes tolerance depend on Enf, not just on drug kinetics for nicotine. The arterial-sampling-analysis (PKPDa) has Ca driving Ce and Cm. The venous-sampling-analysis (PKPDv) does the same but estimates Ca from venous data by means of deconvolution. A TF model (with Cm = Ce) was always selected in the PKPDa. According to this model tolerance developed rapidly with a median half-life of 6.6 min, and median decrease of effect due to tolerance of 31%. Different variants of the TF or TK models were selected in the PKPDv. Parameter estimates for PKPDv show higher variability, and, for the TF model, lower rate and extent of tolerance development and threefold increase in EC50. The study shows that (i) TF models are more appropriate than TK models to describe nicotine effect data, (ii) venous sampling may lead to incorrect model selection and inaccurate and imprecise parameter estimation in respect to arterial sampling, and (iii) arterial sampling should be preferred for accurate (non-steady-state) PD modeling.