How precise is quantitative prediction of pharmacokinetic effects due to drug-drug interactions and genotype from in vitro data? A comprehensive analysis on the example CYP2D6 and CYP2C19 substrates

Abstract

Drug-drug interactions (DDI) and genomic variation (PG) can lead to dangerously high blood and tissue concentrations with some drugs but may be negligible with other drugs. Using a quantitative metaanalysis, we analyzed on the example of CYP2D6 and CYP2C19 substrates, how well the effects of DDI and PG can be predicted by in vitro methods. In addition, we analyzed the quantitative effect of prototypic inhibitors of the two enzymes in relation to their genetic deficiency. More than 600 published studies were screened which compared either human pharmacokinetics with and without comedication, or which compared human pharmacokinetics of deficient with extensive metabolizers, or which assessed metabolism by in vitro approaches. With human liver microsomes, the in vitro to in vivo agreement of fractional clearances was reasonably high if loss of substrate was quantified in the in vitro assays performed with and without enzyme specific inhibitors. Also a generally very high correlation between the clinical pharmacokinetic effects of inherited deficiency and inhibition by drug-drug interactions could be demonstrated. Most cases of poor correlation were explained by the lack of CYP2D6 versus CYP2C19 specificity of fluoxetine or by a poor knowledge of the quantitative contribution of the metabolic pathways if metabolite formation was quantified in the in vitro assays. The good correspondence of the in vitro data with clinical DDI and clinical PG studies may be a good basis for future application of these methods in drug development and drug therapy.

Introduction

Pharmacokinetic drug-drug interactions (DDI) may lead to serious adverse reactions and have caused numerous market withdrawals in the past (Mullins et al., 1998; Wienkers & Heath, 2005). Similarly, pharmacogenetic variation (PG) may compromise a safe and effective therapy with drugs. Therefore, consideration of DDI and PG in drug development and in daily medical care of patients is essential. But this is complicated by the fact, that the quantitative relevance of DDI or PG on systemic drug exposure widely differs between the different drugs. There are drugs for which DDI and PG effects are negligibly small, for other drugs DDI and PG effects are moderate and for some drugs they are very significant (Stingl, Brockmöller, & Viviani, 2013). Thus, a quantitative understanding and prediction of the magnitude of these effects is important, and this is particularly relevant to provide concise recommendations for an individually adjusted drug therapy (Kirchheiner, Fuhr, & Brockmoller, 2005).

Cytochrome P450 enzymes 2D6 and 2C19, together, are involved in biotransformation of about 25% of all small molecule drugs. About 8% and 3% of Europeans are so called poor metabolizers (PM) of CYP2D6 and CYP2C19 substrates, respectively (Ingelman-Sundberg, Oscarson, & McLellan, 1999). These PM are “natural knock-outs” providing unique opportunities to study the biological effects of the two enzymes in human beings. In the analysis presented here we utilize the unique opportunity of having a zero-activity comparison group (the PM) to assess the validity of frequently used in vivo methods for prediction of the fractional clearance via a specific enzyme.

In vitro methods, so-called reaction phenotyping, aim to identify the enzymes involved in the metabolism of a specific drug and their quantitative contribution to biotransformation best defined by fractional clearance (Houston & Galetin, 2003; Rodrigues, 1999; Venkatakrishnan et al., 2001a, Venkatakrishnan et al., 2001b; Wood, Houston, & Hallifax, 2017). Thus, those assays aim to predict the enzyme-specific contribution to in vivo pharmacokinetics in humans. Typically used reaction phenotyping methods (Table 1) are either performed as substrate depletion assay, quantifying loss of the substrate, or as metabolite formation assay quantifying formation of one or several metabolites. Both approaches have their advantages and disadvantages (Yang, Atkinson, & Di, 2016).

In vivo, the difference in total clearance between poor and extensive metabolizers (EM) corresponds to the partial clearance via the respective polymorphic enzyme (Kirchheiner et al., 2004; Kirchheiner, Meineke, Müller, Roots, & Brockmoller, 2002). In vitro data has been used to predict that difference in systemic exposure between extensive and poor metabolizers of CYP2D6 substrates (Gibbs, Hyland, & Youdim, 2006). Vice versa, data from in vitro approaches may then be validated using clinical pharmacokinetic data. Comedication with strong inhibitors can have effects similar to genetic deficiency (Rost, Brockmoller, Esdorn, & Roots, 1995). In vitro predictions of fractional clearance via a specific enzyme may therefore also be validated with in vivo data from drug-drug interaction studies (DDI).

Although the potential of in vitro methods to predict the in vivo pharmacokinetics has been shown for numerous single drugs, there are few comprehensive analyses on many drugs. And there are even less analyses comparing both, pharmacogenetics and drug-drug interactions, and how well both can be predicted by the in vitro methods. In vitro, there are several methods, but it is controversial which one is the best. Thus, here we want to extend research aiming to validate in vitro reaction phenotyping methods with in vivo data from PG studies and/or with in vivo data from DDI studies (Parmentier et al., 2019).

Thus, here we compared data from reaction phenotyping assays with in vivo PG and in vivo DDI data for substrates of CYP2D6 and CYP2C19. Our search strategy aimed to identify all available in vitro and in vivo data on CYP2D6 and CYP2C19 substrates suitable for our analysis. First, we compared reaction phenotyping results with clinical PG study data, second, we compared reaction phenotyping results with clinical drug-drug interaction data and, finally, we compared clinical PG data with drug-drug interaction data. First, this should allow to compare validity of different reaction phenotyping methods. Second, it should be possible to compare the in vitro assays' ability to predict PG as compared to DDI effects. Third, it will also be possible to compare in vivo PG effects with in vivo DDI effects. Finally, joint consideration of in vitro data, clinical PG and clinical DDI data may result in the best fractional clearance data.