Database analyses for the prediction of in vivo drug-drug interactions from in vitro data

Abstract

Aims: In theory, the magnitude of an in vivo drug-drug interaction arising from the inhibition of metabolic clearance can be predicted using the ratio of inhibitor concentration ([I]) to inhibition constant (K(i)). The aim of this study was to construct a database for the prediction of drug-drug interactions from in vitro data and to evaluate the use of the various estimates for the inhibitor concentrations in the term [I]/K(i).

Methods: One hundred and ninety-three in vivo drug-drug interaction studies involving inhibition of CYP3A4, CYP2D6 or CYP2C9 were collated from the literature together with in vitro K(i) values and pharmacokinetic parameters for inhibitors, to allow calculation of average/maximum systemic plasma concentration during the dosing interval and maximum hepatic input plasma concentration (both total and unbound concentration). The observed increase in AUC (decreased clearance) was plotted against the estimated [I]/K(i) ratio for qualitative zoning of the predictions.

Results: The incidence of false negative predictions (AUC ratio > 2, [I]/K(i) < 1) was largest using the average unbound plasma concentration and smallest using the hepatic input total plasma concentration of inhibitor for each of the CYP enzymes. Excluding mechanism-based inhibition, the use of total hepatic input concentration resulted in essentially no false negative predictions, though several false positive predictions (AUC ratio < 2, [I]/K(i) > 1) were found. The incidence of true positive predictions (AUC ratio > 2, [I]/K(i) > 1) was also highest using the total hepatic input concentration.

Conclusions: The use of the total hepatic input concentration of inhibitor together with in vitro K(i) values was the most successful method for the categorization of putative CYP inhibitors and for identifying negative drug-drug interactions. However this approach should be considered as an initial discriminating screen, as it is empirical and requires subsequent mechanistic studies to provide a comprehensive evaluation of a positive result.

Figure 1

Qualitative zoning for the prediction of drug–drug interactions involving CYP inhibition. The curve represents the theoretical curve based on equation 1. F-: false negative, T-: true negative, F+: false positive, T+: true positive

Figure 2

Relationship between the observed AUC ratio and the various [I]/K_i ratios for drug-drug interactions involving CYP3A4 (•), CYP2D6 (▴) or CYP2C9 (▪). The curves represent the theoretical curves based on equation 1 using average systemic total drug plasma concentration ([I]_av– panel A), average systemic unbound drug plasma concentration ([I]_av,u– panel B), maximum systemic plasma concentration ([I]_max– panel C), and maximum hepatic input concentration ([I]_in– panel D)

Figure 3

Identifying drug-drug interaction studies involving either reversible (closed symbols) or mechanism-based inhibition (open symbols). CYP3A4 (•), CYP2D6 (▴), CYP2C9

Figure 4

Data from studies involving CYP3A inhibitors. Fluconazole (n = 7) (○), itraconazole (n = 17) (▪), ketoconazole (n = 11) (▴), HIV protease inhibitors (n = 2) (▵), others (n = 35)

Figure 5

Data from studies involving CYP2D6 inhibitors. Quinidine (n = 7) (▴), citalopram (n = 3) (□), fluoxetine (n = 8) (▪), fluvoxamine (n = 2) (○), sertraline (n = 6) (▵), others (n = 25) (•)

Figure 6

Comparison of the use of three in vivo metrics for interactions involving CYP2D6 inhibition. The ordinates show the observed ratio of AUC (A), plasma concentration at a single time point (B), or metabolic ratio (C) for the substrate in the presence and absence of inhibitor. Open symbols represent studies involving paroxetine.

Figure 7

Simulation of the effects of changing the rate of absorption of the inhibitor (A) or liver blood flow (B). The AUC ratio was calculated as 1 + [I]_in/K_i and plotted against 1/K_i. The following parameter values for inhibition by quinidine were used, D = 268 nmol, τ = 24 h, & CL/F = 412 ml min⁻¹ in equations 4 and 6 A— : ka Fa = 0.1 min⁻¹, —: ka Fa = 0.01 min⁻¹, …: ka Fa = 0.001 min⁻¹B— : Qh = 1610 ml min⁻¹, —: Qh = 805 ml min⁻¹, …: Qh = 3220 ml min⁻¹