Application of Virtual Twin PBPK Models in Individuals with Obesity via CYP3A4 Phenotyping Using Endogenous Biomarker Data

Plasma 4β‐hydroxycholesterol (4β‐OHC) normalized for the levels of its parent cholesterol (4β‐OHC/C) is an endogenous biomarker for hepatic CYP3A4. This study evaluated CYP3A4 activity longitudinally in individuals while their obesity status was changing. 4β‐OHC/C was measured pre‐surgery ( n = 54) and 2 years after Roux‐en‐Y gastric bypass ( n = 30) and used as an input for creating virtual twins of each patient within physiologically‐based pharmacokinetic (VT‐PBPK) models, alongside other available data (e.g., demographics). Additionally, individual CYP3A4 abundance quantified from liver biopsies during surgery was available. Predicted CYP3A4 abundance using 4β‐OHC/C was within twofold of the measured abundance in 85% of the individuals ( n = 54). Pre‐surgery VT‐PBPK models informed by biomarker data predicted area under curve after intravenous midazolam (AUC inf,iv ) within twofold of observed values in 85% of individuals (geometric mean fold error (GMFE) = 1.58). VT‐PBPK models using CYP3A4 protein measurements showed similar performance (83% within twofold, GMFE = 1.66), verifying the biomarker approach. The post‐surgery VT‐PBPK models were revised to reflect surgery‐related alterations and changes in obesity status. Hepatic CYP3A4 activity at 2 years was predicted from 4β‐OHC/C measured in the same individuals. Biomarker‐informed models predicted midazolam AUC inf,iv within twofold for 87% of individuals (GMFE = 1.48; n = 30). ~60% of observed oral midazolam PK parameters were predicted within twofold, improving to 70% when the surgery model assumed full intestinal CYP3A4/5 recovery. This study highlights the utility of 4β‐OHC/C for continuous monitoring of CYP3A4 activity and possibility to use biomarker‐informed VT‐PBPK models for individual dose requirements of intravenously administered CYP3A substrates in individuals with obesity over the course of weight loss treatment.