PBPK modeling of the antidepressant doxepin incorporating CYP2D6 genotype for precision pharmacotherapy

TL;DR

This paper develops a PBPK model for the antidepressant doxepin that accounts for genetic differences in CYP2D6 to improve personalized drug dosing.

Contribution

The study introduces a PBPK model for doxepin and its metabolite that incorporates CYP2D6 genotype to predict pharmacokinetic variability.

Findings

The PBPK model accurately predicted doxepin and N-desmethyldoxepin pharmacokinetics across different CYP2D6 genotypes.

Doxepin exposure increased with decreasing CYP2D6 metabolic capacity, consistent with clinical data.

The model captured genotype-dependent trends in metabolite formation and elimination despite limited genotype-specific data.

Abstract

Doxepin, a tricyclic antidepressant, exhibits substantial interindividual variability in pharmacokinetics, mainly attributable to genetic polymorphisms of cytochrome P450 (CYP) enzymes, particularly CYP2D6. Such variability may lead to clinically relevant differences in drug exposure, therapeutic response, and adverse effects when doxepin is used for antidepressant therapy. The present study aimed to develop and validate physiologically based pharmacokinetic (PBPK) models of doxepin and its active metabolite, N-desmethyldoxepin, incorporating CYP2D6 genetic polymorphisms to support precision pharmacotherapy. PBPK models were constructed using PK-Sim® (version 12.0) based on published clinical pharmacokinetic and pharmacogenomic data obtained after oral administration of doxepin. Model development was performed in a non-genotyped population. Subsequently, it extended to CYP2D6 ultra-rapid (UM), normal (NM), intermediate (IM), and poor metabolizer (PM) phenotypes by integrating genotype-specific metabolic activity. Model performance was evaluated by comparing predicted plasma concentration–time profiles, area under the concentration–time curve (AUC), and maximum plasma concentration (Cmax) with observed clinical data. The model's predictive capability was further assessed using goodness-of-fit analysis, visual predictive checks, and geometric mean fold error (GMFE). The PBPK models adequately reproduced the observed pharmacokinetics of doxepin and N-desmethyldoxepin, with most predicted AUC and Cmax values falling within a twofold error range across both non-genotyped and CYP2D6-genotyped populations. Simulations demonstrated an apparent genotype-dependent increase in doxepin exposure as CYP2D6 metabolic capacity decreased, consistent with clinical observations. Similarly, N-desmethyldoxepin exposure showed an inverse relationship with CYP2D6 activity, reflecting differences in metabolite formation and elimination. Despite the limited availability of genotype-specific concentration–time data for the metabolite, the model captured observed exposure trends, supporting its mechanistic plausibility. In conclusion, a PBPK framework integrating CYP2D6 genetic polymorphisms was successfully developed for doxepin and its active metabolite. The model reliably predicts genotype-dependent pharmacokinetic variability and provides a mechanistic basis for individualized dose optimization. This PBPK approach may serve as a valuable tool to support precision pharmacotherapy for doxepin, particularly in the context of antidepressant treatment.

Figures