Quantitative Prediction of Drug Interactions Caused by CYP1A2 Inhibitors and Inducers

Abstract

Background

A simple method to predict drug–drug interactions mediated by cytochrome P450 enzymes (CYPs) on the basis of in vivo data has been previously applied for several CYP isoforms but not for CYP1A2. The objective of this study was to extend this method to drug interactions caused by CYP1A2 inhibitors and inducers.

Methods

First, initial estimates of the model parameters were obtained using data from the literature. Then, an external validation of these initial estimates was performed by comparing model-based predicted area under the concentration–time curve (AUC) ratios with observations not used in the initial estimation. Third, refined estimates of the model parameters were obtained by Bayesian orthogonal regression using Winbugs software, and predicted AUC ratios were compared with all available observations. Finally, predicted AUC ratios for all possible substrates–inhibitors and substrates–inducers were computed.

Results

A total of 100 AUC ratios were retrieved from the literature. Model parameters were estimated for 19 CYP1A2 substrate drugs, 26 inhibitors and seven inducers, including tobacco smoking. In the external validation, the mean prediction error of the AUC ratios was −0.22, while the mean absolute error was 0.97 (37 %). After the Bayesian estimation step, the mean prediction error was 0.11, while the mean absolute error was 0.43 (22 %). The AUC ratios for 625 possible interactions were computed.

Conclusion

This analysis provides insights into the interaction profiles of drugs poorly studied so far and can help to identify and manage significant interactions in clinical practice. Those results are now available to the community via a web tool (http://www.ddi-predictor.org).

Appendix

This “Appendix” illustrates the application of the general model for multiple CYP-mediated drug interactions to the case of interactions with ramelteon as the victim drug.

The model for an interaction mediated by three cytochromes (denoted 1, 2 and 3) and caused by an inhibitor and an inducer is described by Eqs. 8 and 9, respectively [88]:

$$ \frac{{{\text{AUC}}^{*} }}{\text{AUC}} = \frac{{{\text{CR}}_{1} + {\text{CR}}_{2} + {\text{CR}}_{3} + (1 - {\text{CR}}_{1} - {\text{CR}}_{2} - {\text{CR}}_{3} )}}{{{\text{CR}}_{1} \cdot (1 - {\text{IR}}_{1} ) + {\text{CR}}_{2} \cdot (1 - {\text{IR}}_{2} ) + {\text{CR}}_{3} \cdot (1 - {\text{IR}}_{3} ) + (1 - {\text{CR}}_{1} - {\text{CR}}_{2} - {\text{CR}}_{3} )}} $$

(8)

$$ \frac{{{\text{AUC}}^{*} }}{\text{AUC}} = \frac{{{\text{CR}}_{1} + {\text{CR}}_{2} + {\text{CR}}_{3} + (1 - {\text{CR}}_{1} - {\text{CR}}_{2} - {\text{CR}}_{3} )}}{{{\text{CR}}_{1} \cdot (1 + {\text{IC}}_{1} ) + {\text{CR}}_{2} \cdot (1 + {\text{IC}}_{2} ) + {\text{CR}}_{3} \cdot (1 + {\text{IC}}_{3} ) + (1 - {\text{CR}}_{1} - {\text{CR}}_{2} - {\text{CR}}_{3} )}} $$

(9)

The estimated IRs of fluvoxamine are as follows: IC_CYP1A2 = 1, IC_CYP2C19 = 0.98, IR_CYP3A4 = 0.3.

The estimated ICs of rifampicin are as follows: IC_CYP1A2 = 1.45, IC_CYP2C19 = 4.2, IC_CYP3A4 = 7.7.

Ramelteon is metabolized by CYP1A2, CYP2C19 and CYP3A4, with contribution ratios estimated from in vitro data as follows [93]: CR_CYP1A2 = 0.49, CR_CYP2C19 = 0.42, CR_CYP3A4 = 0.086 (CR set 1).

Because the CRs of ramelteon have not been estimated in vivo, one might also consider another set of CR values. For example, we may assume that the contribution of CYP3A4 might be lower or even negligible in vivo, with CR_CYP1A2 = 0.57, CR_CYP2C19 = 0.42, CR_CYP3A4 = 0.01 (CR set 2), or CR_CYP1A2 = 0.58, CR_CYP2C19 = 0.42, CR_CYP3A4 = 0 (CR set 3).

The predicted magnitude of the drug interactions is shown in Table 7.

Table 7 Predicted magnitudes of ramelteon drug interactions

As one can see, for the interaction caused by fluvoxamine, the predicted AUC ratio strongly depends on the estimated CRs for ramelteon, because fluvoxamine is a very potent inhibitor of CYP1A2 and CYP2C19, with IR values close to 1. Decreasing the CR of CYP3A4 and increasing the CR of CYP1A2/CYP2C19 result in a larger predicted AUC ratio, which better agrees with the observed ratio.