QSAR-based physiologically based pharmacokinetic (PBPK) modeling for 34 fentanyl analogs: model validation, human pharmacokinetic prediction and abuse risk insights

Fentanyl analogs, as emerging new psychoactive substances (NPS), pose a global public health threat due to widespread abuse, high toxicity, and frequent overdose fatalities. However, their structural diversity and scarce experimental pharmacokinetic (PK) data hinder hazard and abuse risk assessment. Conventional physiologically based pharmacokinetic (PBPK) models for these analogs are limited by reliance on time-consuming in vitro experiments or error-prone interspecies extrapolation for key parameters (e.g., tissue/blood partition coefficient, Kp).

To address this, we developed and validated a QSAR-integrated PBPK framework (QSAR: Quantitative Structure-Activity Relationship) for predicting human PK of fentanyl analogs. The workflow included: (1) Validating the framework via intravenous β-hydroxythiofentanyl in Sprague-Dawley rats (QSAR-predicted Kp via Lukacova method, GastroPlus^® modeling); (2) Comparing Kp accuracy (literature in vitro data, QSAR, interspecies extrapolation) in rat/human fentanyl PBPK models; (3) Predicting PK and tissue distribution (plasma +10 organs including brain/heart) of 34 human fentanyl analogs.

Key results: (1) For β-hydroxythiofentanyl, all predicted rat PK parameters (area under the plasma concentration-time curve from time zero to the last measurable time point [AUC_0-t], teady-state volume of distribution [V_ss], and elimination half-life [T_1/2]) of rats fell within a 2-fold range of the experimental values; (2) In human fentanyl models, QSAR-predicted Kp improved accuracy (V_ss error: >3-fold [extrapolation] vs. <1.5-fold [QSAR]) (3) Among 34 analogs, eight (e.g., p-fluorofentanyl); had brain/plasma ratio >1.2 (vs. fentanyl's 1.0), indicating higher CNS penetration and abuse risk.

This study demonstrates that the QSAR-PBPK framework enables rapid prediction of human pharmacokinetics (PK) for understudied fentanyl analogs without relying on scarce experimental data. For structurally similar, clinically characterized analogs (e.g., sufentanil, alfentanil), predictions of key PK parameters (e.g., T_1/2, V_ss) fall within 1.3-1.7-fold of clinical data, supporting the framework's utility for generating testable hypotheses about the PK of understudied analogs. It not only fills the data gap for fentanyl analog hazard assessment but also provides a scalable modeling strategy for PK evaluation of other NPS or illicit drugs.