Top-down approach for prediction of drug-drug interactions using physiologically based pharmacokinetic model
Motohiro KATO *
DMPK academy
[Purpose] Prediction of drug interactions using physiologically based pharmacokinetic (PBPK) models is widely used. Inhibition constants (Ki) are calculated using human liver microsomes, but discrepancies between in vitro Ki and in vivo Ki have been reported. Prediction using in vitro Ki values may give false negative predictions. Prediction methods using in vitro Ki are called bottom-up approaches, and prediction methods using in vivo Ki are called top-down approaches. The top-down approach is based on the results of clinical trials, and therefore has higher prediction accuracy than the bottom-up approach. Since only in vitro Ki values can be obtained at the drug discovery stage, a middle-out approach that combines bottom-up and top-down approaches is important in drug development. Since it is not easy to obtain the in vivo Ki value of an inhibitor from clinical trial results, a method to easily obtain the in vivo Ki value from the information provided in the drug package insert was developed. However, this method cannot be used for CYP3A4 substrates due to small intestinal metabolism. Therefore, a new method that incorporates a small intestinal inhibition model was developed.
[Methods] A simplified PBPK model with a small intestine metabolic inhibition model was constructed. The method for calculating the Ki value of the inhibitor and the fm of the substrate is briefly described below. Step 1: A 1- or 2-compartment model is selected from the four pharmacokinetic parameters (Cmax, AUC, Tmax, t1/2) of the substrate and inhibitor, and the parameters are estimated using the Bayesian method. Step 2: The PBPK model parameters are calculated using the compartment parameters and several other pieces of information.. Step 3: The Ki value of the inhibitor and the contribution ratio (fm) of the metabolic enzyme for substrate are calculated using the PBPK model by using a dichotomy method from the increase in AUC of the substrate in clinical trials.
[Results and Discussion] Six ketoconazole and midazolam interaction studies (single dose, once daily dose, twice daily dose) were analyzed. The geomean Ki values of ketoconazole was 1.71 ng/mL (0.89-2.61 ng/mL), which was lower than the in vitro Ki value (46.8 ng/mL) (3). In addition, three clinical trials of fluconazole and midazolam were analyzed. The geomean Ki value of fluconazole was 2824 ng/mL (2475-3286 ng/mL), which was similar to the in vitro Ki (1838 ng/mL).
[Conclusions] A novel top-down method for predicting drug-drug interactions due to CYP inhibition was developed. This method will make it easier to predict drug-drug interactions with high accuracy.
(1) Pharm Res. 2008; 25:1891-901
(2) M Kato, The 40th Annual Meeting of the Japanese Society for the Study of Xenobiotics 2025 (submitted)
(3) Xenobiotica. 2009; 39: 430-43