Introduction: Understanding disease-related changes in the pharmacokinetics of drugs in patients with nonalcoholic fatty liver disease (NAFLD) is of clinical importance to guide optimal dosing regimens for this patient population. Methods: A virtual NAFLD patient population, stratified into simple steatosis and the more advanced stage, nonalcoholic steatohepatitis (NASH) was developed using GastroPlus v.9.8.2 (Simulations Plus, Inc.) to support the implementation of a physiologically based pharmacokinetic (PBPK) modeling approach. The model accounts for (patho)physiological changes associated with the disease and alterations in protein levels of metabolizing enzymes and transporters pertinent to drug disposition . The virtual NAFLD population model was verified using previously published clinical pharmacokinetic data for pioglitazone, rosuvastatin and metformin as illustrative examples of drugs commonly used by patients with NAFLD. Results. PBPK model predictions of pioglitazone and rosuvastatin plasma concentrations, and hepatic radioactivity levels (represented by standardized uptake values [SUV]) of 11C-metformin were in good agreement with the clinically-observed data. Importantly, the PBPK simulations provided reliable estimates of the extent of changes in key pharmacokinetic parameters for the diseased population compared with the non-NAFLD control group, consistent with the reported changes in the corresponding clinical studies. Clinically-reported vs PBPK model predicted ratios (NASH patients divided by healthy individuals) for trough concentrations (Css,min) of pioglitazone were 1.46 vs 1.42, respectively. The corresponding ratios for the active metabolite of pioglitazone, hydroxy-pioglitazone, that is formed predominantly through a CYP2C8-mediated pathway were 0.68 and 0.66, respectively. This was in concordance with the proteomics-informed reduction in CYP2C8 abundance by ~34% that was implemented in the virtual NASH patient population. The clinically-reported vs PBPK model predicted ratios for the systemic exposure (AUC0-inf) of rosuvastatin in obese NASH patients compared to non-NAFLD lean individuals were 0.83 and 0.80, respectively, corresponding to a prediction fold-difference that was close to unity. The PBPK simulation highlighted that decreased OATP1B1 abundance, increased protein levels of BCRP, and differences in body weight and liver size may all contribute to the slight net reduction in systemic exposure of rosuvastatin observed in NASH patients compared to the non-NAFLD cohort. Meanwhile, simple steatosis and NASH appeared to have minimal impact on peak hepatic SUV of 11C-metformin, with observed and model predicted ratios to that of healthy individuals of 1.10 vs 0.95 for simple steatosis and 0.92 vs 0.77 for NASH, respectively. Conclusion: A virtual population model within the PBPK framework representing patients with NAFLD was successfully developed with a good predictive capability of estimating disease-related changes in drug pharmacokinetics. This supports the use of a PBPK modeling approach for prediction of the pharmacokinetics of new investigational or repurposed drugs for potential treatment of NAFLD, and may help inform dose adjustment of drugs commonly used to treat comorbidities in this patient population.
Funding: The financial support of Simulations Plus, Inc. is gratefully acknowledged