Comparison of common methodologies for accounting for IIV for oral bioavailability in the absence of intravenous data
Anne Brochot (1), Soha Freidy (1), Kai Bartlette (1), Alex MacDonald (1)
1 Allucent, LLC, Cary, NC
Objectives:
The oral bioavailability (F) of a drug is the product of the fraction absorbed (Fa), the fraction escaping the gut-wall metabolism (Fg), and the fraction escaping liver extraction (Fh). Every component of this product can be influenced by several factors such as the physicochemical properties of the drug or physiological issues (1). In turn, each of those factors contributes to potential inter-individual variability (IIV) in bioavailability. When modeling oral pharmacokinetic (PK) data using a nonlinear mixed effect method, two approaches are commonly used in the absence of intravenous (IV) data:
- Attribute the IIV linked to F to the apparent clearance (CL/F) and the apparent volume of distribution (V/F) parameters and estimate the potential correlation between the two;
- Fix the bioavailability to a relative value of 1 and estimate an IIV on this parameter.
While the modeling of oral or other extravascular pharmacokinetic data in the absence of corresponding intravenous data is common practice in drug development, there appears to be no clear consensus in the literature as to the most appropriate methodologies or a comparison of their relative merits. The objective of this analysis was to evaluate the suitability of these approaches using realistic PK data.
Methods:
To make the exercise representative, physiologically-based pharmacokinetic (PBPK) models of drugs from publicly available libraries in PK-Sim® (2) were used to simulate clinical study data with a phase 1 single ascending dose-like design with a dense sampling strategy. 6 drugs of different BCS class (verapamil and fluconazole, BCS class 1; montelukast and felodipine, BCS class 2; dapagliflozin and cimetidine, BCS class 3) were simulated, and the PK-Sim® bioavailability was output for each simulated subject. Each simulated dataset was modeled afterward using NONMEM V7.4. Bodyweight allometric scaling was introduced apriori on all clearance and volume terms. Model evaluation was based on goodness-of-fit plots, prediction-corrected visual predictive checks, and precision in parameter estimates. The best model without an IIV on F was determined for each drug and then IIV on F was included. The relationship between the post-hoc F ETA and the PK-Sim® bioavailability was evaluated.
Results:
All models included an IIV on CL/F and V/F as a minimum[AM1] , with the exception of fluconazole (IIV on absorption parameter) and montelukast (IIV on CL and peripheral distribution parameters). Partial or full correlation blocks were introduced where required. All models showed good predictive performance based on the [AM2] pcVPC, excluding fluconazole for which a dose dependency on absorption parameter was suspected as classical VPC looked good. With the exception of verapamil, all models showed a decrease in both objective function, AIC and residual unexplained variability (RUV) when an IIV on F was introduced. This improvement did not translate in significant enhancement of the pcVPCs, but correlation between the post-hoc ETA for F and the PK-Sim® derived bioavailability was observed in most cases.
Conclusions:
NONMEM was able to quantify the IIV on F in addition to apparent CL and V terms, and its presence improved the objective function and RUV in most case, and was mostly correlated to the “true” bioavailability. The choice to use or not an IIV on F, since not necessarily improving the predictive performances, could be made based on the objective of the modeling exercise, e.g. some bioavailability-specific covariates are suspected or if RUV is important for future simulations.
References:
[1] Rowland and Tozer’s Clinical Pharmacokinetics and Pharmacodynamics. Concept and applications. H. Derendorf, S. Schmidt. Fifth edition. Wolters Kluwer, 2020.
[2] https://www.open-systems-pharmacology.org/
