Xuan Zhou, Fan Zhang, Peiming Ma
GlaxoSmithKline R&D, China
Introduction:
Issues often exist in parameter estimation of pharmacokinetic (PK) models for a parent drug and its metabolites. The model could be structurally non-identifiable even with perfect data, if the data for intravenous PK, or the metabolites alone PK, or the fractions of parent drug to metabolites are not available. With structurally identifiable model, real-life conditions may still cause the model parameters to be non-estimable.
To illustrate these points, we present a case of PK modeling for Wellbutrin XL (bupropion hydrochloride extended-release tablet). Bupropion XL is well absorbed and only small proportion of an orally administered dose reaches the systemic circulation intact due to high first-pass metabolism. Its three main active metabolites, hydroxybupropion, erythrohydrobupropion and threohydrobupropion, have been identified. Many existing PK studies for bupropion have been conducted over the past decades; however, their analyses have not yet provided enough insight into the conversion of bupropion to its 3 metabolites, and could not serve as a basis for better understanding of the underlying mechanism and for guiding the dose adjustment as a later goal.
Objectives:
To describe the PK of bupropion XL and its 3 active metabolites, and more importantly, to use this post hoc analysis as an example to discuss some common issues for modeling PK concentration data for parent compound and its metabolites when the drug was administered only orally.
Methods:
The data used for model building was from a GSK funded Phase I, open-label study (NCT02698553) conducted in healthy Chinese subjects. In this study, for safety consideration, dose titration strategy was used. 16 subjects received bupropion XL 150 mg once daily (QD) for 5 days, and then the dose was titrated to 300 mg QD from Day 6 to Day 14. Blood samples were collected pre-dose and Day 1, 5, 6, and 14-19. Three different model structures were investigated:
- first-pass effect not considered, the parent drug was either eliminated from the system or transformed into the 3 metabolites;
- first-pass effect considered, the dose enters simultaneously into the parent and metabolite compartments;
- first-pass effect considered, a dose apportionment was assigned independent of absorption rates with a fraction of dose leading to the parent, and fractions leading to the 3 metabolites as hypothetical metabolite absorption compartments prior to reaching the circulation.
The structural identifiability of these models was analyzed using Laplace transformation. Concentration data of bupropion and metabolites were simultaneously fitted using ADVAN5 and FOCE-I methods.
Results:
From the structural identifiability analysis in Model 1, the parameter ka, CLp/F, Q/F, V2/F and V3/F can be globally identified, while in Model 2, none of the parameters can be identified. In Model 3, only the absorption rate constants for the parent and all the metabolites (ka, ka4-ka6) can be identified. Model 3 was selected as final model based on mechanism and model performance. PK was adequately described for the parent drug by a two-compartment model with first-order absorption and linear elimination plus lag time, and for the metabolites by one-compartment models considering first-pass effects and systemic transformations. To have a structural identifiable model, the fractions of the parent converting to metabolites within circulation and the fractions of the parent and metabolites reaching circulation after absorption should be fixed. Those fractions were fixed to the proportions determined from the observed peak molar concentrations (Cmax) of the parent and metabolites (12.84%, 59.52%, 23.81%, and 3.83% for bupropion, hydroxybupropion, threohydrobupropion, and erythrohydrobupropion, respectively), which reflected the initial pre-systemic elimination, subsequent absorption and metabolism of bupropion.
Conclusions:
A population PK model was developed for bupropion and its three active metabolites from the study conducted in Chinese healthy volunteer. We believe a proper balance between complexity of mechanism and amount of data has to be reached for modeling parent-metabolite data.
Reference: PAGE 27 (2018) Abstr 8660 [www.page-meeting.org/?abstract=8660]
Poster: Methodology - Other topics