An overview of interpretability of two models of unbound fraction that are used in combination with the well-stirred model for predicting hepatic clearance of drugs

Abstract

Hypothetical and experimental models of unbound fraction have been proposed to facilitate predicting hepatic clearance (CLH) of drugs from values of intrinsic clearance for the unbound drug (CLint-in vitro-unbound) in the well-stirred model (WSM). The hypothetical model (fu-adjusted) is adjusting the unbound fractions determined in plasma in vitro to estimate the maximum unbound fractions at the hepatocytes if each drug-protein complex in plasma becomes fully dissociated at the membrane by any albumin (ALB)-facilitated hepatic uptake mechanism. The model of fu-adjusted is also adjusting the unbound fraction for a pH gradient effect across the membrane. Whereas the new experimental model (fu-dynamic) measures the unbound fractions resulting to the dynamic dissociation kinetics from proteins in a presence of plasma and liver enzymes in an in vitro assay. The objective of this study was to conduct an in-depth analysis of previous CLH predictions made with these unbound fractions in a companion manuscript. Furthermore, a new dataset on seven statins was also included in this study. Finally, the physiological basis of fu-adjusted has been redefined to extend its applicability with more drugs. In this case, there are lower concentrations of binding proteins in liver versus plasma that could also explain the higher unbound fractions for that organ. The outcomes associated to additional analyses pointed out that fu-adjusted, again, generally provided the most accurate predictions of CLH because fu-dynamic has generated superior biases of underpredictions or overpredictions. For slowly metabolized drugs bound to ALB, fu-dynamic was definitively less accurate than fu-adjusted. For other drug properties, fu-dynamic fared better but it was still not generally more accurate than fu-adjusted. Furthermore, experimental values of fu-dynamic were sometimes incoherent. For example, drugs bound to alpha-acid glycoprotein (AGP) did not follow the assumption of fu-dynamic (i.e., values of fu-dynamic did not correlate with values of CLint-in vitro-unbound) by contrast to those drugs bound to ALB. Therefore, the current experimental setting for fu-dynamic might be unsuitable in some circumstances. Overall, this study confirmed that calculated values of fu-adjusted were as accurate as experimental values of fu-dynamic and can even be more accurate. A guidance on which unbound fraction to use in WSM is also provided.