Commentary IMPACT OF CYP2C9 GENOTYPE ON PHARMACOKINETICS: ARE ALL CYCLOOXYGENASE INHIBITORS THE SAME?

The market withdrawals of rofecoxib (Vioxx) and valdecoxib (Bextra) have focused considerable attention on the side effect profiles of cyclooxygenase (COX) inhibitors. As a result, attempts will be made to identify risk factors in the hope that physicians might be able to ensure patient safety. At first glance, CYP2C9 genotype might be considered a risk factor because many COX inhibitors are CYP2C9 substrates in vitro. This observation has led some to hypothesize that a reduction in clearance, in subjects expressing variant forms of the enzyme (e.g., CYP2C9*1/*3 or CYP2C9*3/*3 genotype), will lead to increased exposure and a greater risk of cardiovascular or gastrointestinal side effects. For any drug, however, one has to consider all clearance pathways. Therefore, a number of COX inhibitors were surveyed and it was determined that CYP2C9 plays a relatively minor role in the overall clearance (<20% of the dose) of sulindac, naproxen, ketoprofen, diclofenac, rofecoxib, and etoricoxib. CYP2C9 genotype would have no clinically meaningful impact on the pharmacokinetics of these drugs. In contrast, CYP2C9 genotype is expected to impact the clearance of ibuprofen, indomethacin, flurbiprofen, celecoxib, valdecoxib, lornoxicam, tenoxicam, meloxicam, and piroxicam. However, even when CYP2C9 is a major determinant of clearance, it is necessary to consider CYP2C8 genotype (e.g., ibuprofen) and, possibly, CYP3A4 activity (e.g., celecoxib, valdecoxib, and meloxicam) also. Events surrounding the market withdrawal of rofecoxib (Vioxx), a potent and selective COX-2 inhibitor, have raised concerns about the safety of other COX-2 selective inhibitors such as etoricoxib (Arcoxia), celecoxib (Celebrex), lumiracoxib (Prexige), and valdecoxib (Bextra) (Mukherjee et al., 2002; Bing, 2003; Couzin, 2004; Davies and Jamali, 2004; Fitzgerald, 2004; Kim and Reicin, 2004; Ray et al., 2004; Scheen, 2004; Bannwarth, 2005; Berenbaum, 2005). Such concerns finally resulted in the withdrawal of valdecoxib about seven months after the withdrawal of rofecoxib (Lenzer, 2005; Young, 2005). Nonselective COX inhibitors (NSAIDs) like naproxen, diclofenac, and ibuprofen have also come under scrutiny from regulators, physicians, and patient safety advocacy groups (McGettigan and Henry, 2000; Meagher, 2003). At the present time, it is thought that the CV and GI side effects of COX inhibitors are related to their mechanism of action. This involves the inhibition of COX, a hemeprotein that exists in two forms (COX-1 and COX-2). COX-1 is expressed constitutively in most tissues, whereas the expression of COX-2 can be induced by growth factors, cytokines, and vasoactive peptides such as endothelin. In response to cell damage, therefore, COX-2 is inducible by proinflammatory mediators and plays a role in the generation of prostaglandin E2, a major mediator of inflammatory response. On the other hand, the products of COX-1 are cytoprotective in GI epithelium, and selective inhibition of COX-2 is anticipated to reduce inflammation, and modulate pain, without the GI side effects characteristic of nonselective NSAIDs (e.g., peptic erosions, ulceration, and bleeding). As a result, it has become accepted that inhibition of COX-1 should be minimized, and the industry has focused on the design of potent and selective COX-2 inhibitors (Davies et al., 2000, 2003; Riendeau et al., 2001; Chavez Article, publication date, and citation information can be found at http://dmd.aspetjournals.org. doi:10.1124/dmd.105.006452. ABBREVIATIONS: COX, cyclooxygenase; P450, cytochrome P450; IC50(COX), concentration of inhibitor required to decrease COX activity by 50%; AUCpo(PM), area under the plasma concentration vs. time curve (oral dose) in subjects phenotyped as PM (expressing one or two variant alleles); AUCpo, area under the plasma concentration vs. time curve (oral dose); AUCpo(EM), area under the plasma concentration vs. time curve (oral dose) in subjects phenotyped as EM (expressing two wild-type alleles); [S], substrate concentration; PK, pharmacokinetics; PD, pharmacodynamics; PK-PD, pharmacokinetics-pharmacodynamics; Km, Michaelis constant; Vmax, maximal initial rate of metabolism; NSAID, nonsteroidal antiinflammatory drug; fh, fraction of dose eliminated in the liver; fm, fraction of total hepatic elimination via all cytochromes P450; fm,CYP, fraction of total cytochrome P450 metabolism catalyzed by an individual cytochrome P450 form; fm,CYP2C9(EM), fraction of total cytochrome P450 metabolism catalyzed by CYP2C9 in EM (wild type, CYP2C9*1/*1) subjects; GI, gastrointestinal; CV, cardiovascular; FMO, flavin-containing monooxygenase; AO, aldehyde oxidase; CLint f , intrinsic (metabolite formation) clearance; CLint,CYP2C9(EM) f , intrinsic (metabolite formation) clearance catalyzed by CYP2C9 (wild type) in EM subjects; CLint,CYP2C9(PM) f , intrinsic (metabolite formation) clearance catalyzed by CYP2C9 (variant forms) in PM subjects; EM, extensive metabolizer phenotype; PM, poor metabolizer phenotype; CLint , total intrinsic clearance (parent consumption); ADME, absorptiondistribution-metabolism-excretion; [E], enzyme concentration; kcat, first-order rate constant that relates Vmax to [E]; Cb, concentration of COX inhibitor in blood. 0090-9556/05/3311-1567–1575$20.00 DRUG METABOLISM AND DISPOSITION Vol. 33, No. 11 Copyright © 2005 by The American Society for Pharmacology and Experimental Therapeutics 6452/3061525 DMD 33:1567–1575, 2005 Printed in U.S.A. 1567 at A PE T Jornals on Jne 2, 2017 dm d.aspurnals.org D ow nladed from and DeKorte, 2003; Meagher, 2003; Justice and Carruthers, 2005). However, it should be recognized that although COX-2 is inducible, its products are not always proinflammatory. Constitutive COX-2 in the vasculature generates mainly prostacyclin, which is a vasodilator and inhibitor of platelet aggregation. Therefore, inhibition of COX-2 may alter the balance of prothrombotic (versus antithrombotic) eicosanoids and predispose susceptible individuals to CV side effects (Mukherjee et al., 2002; Meagher, 2003; Davies and Jamali, 2004; Justice and Carruthers, 2005). A complex picture is emerging and the clinical safety (both GI and CV) of COX inhibitors most likely depends on a fine balance of factors. These factors include COX-1 and COX-2 inhibitory potency, the IC50(COX) ratio (COX-2 versus COX-1), Cb/IC50(COX) ratios, PK, PD, PK-PD (dose response) for each COX form, tissue distribution of the inhibitor (relative to tissue distribution of each COX enzyme form), and therapeutic index (Riendeau et al., 2001; Meagher, 2003; Davies and Jamali, 2004; Lees et al., 2004; Justice and Carruthers, 2005). The situation is complicated further by the presence of allelic variant forms of COX-1 and COX-2, which may not only impact efficacy, but predispose individuals to different levels of risk (Halushka et al., 2003; Cipollone et al., 2004). Factors governing systemic clearance (PK) have received particular attention. This is because the majority of marketed COX inhibitors are well absorbed, metabolized extensively, subject to relatively minimal first-pass extraction, and exhibit linear PK ([S]/Km 0.1). Consequently, exposure and Cb/IC50(COX) ratios will depend on systemic clearance. In turn, systemic clearance will be governed by alterations in CLint tot in the liver (assuming that fh approaches unity). CYP2C9 (CYP2C9*1) is considered important, because it has been known for some time that the enzyme plays a role in the metabolism of many NSAIDs in vitro and is thus considered a major determinant of CLint tot (Zhao et al., 1992; Leemann et al., 1993; Miners and Birkett, 1998; Rodrigues and Rushmore, 2002). The catalytic efficiency (kcat/Km ratio) of the allelic variant forms of the enzyme (e.g., CYP2C9*2 and CYP2C9*3) is reduced because of a single amino acid substitution (Miners and Birkett, 1998; Takanashi et al., 2000; Tang et al., 2001; Rodrigues and Rushmore, 2002). Therefore, due to the high incidence of CYP2C9-related polymorphisms in some populations (e.g., the frequency of CYP2C9*1/*3 genotype is 12% in white subjects), one can hypothesize that the occurrence of side effects is increased in numerous subjects genotyped heterozygous, or homozygous, for the CYP2C9*2 or CYP2C9*3 alleles (Xie et al., 2002; Schwarz, 2003; Lee, 2004; Kirchheiner and Brockmoller, 2005; Rettie and Jones, 2005). However, the picture is not so simple and the data so far are not clear. For example, Wynne et al. (1998) hypothesized that PK might explain the risk of major GI hemorrhage with NSAIDs, with bleeders exhibiting a reduced clearance of NSAIDs compared with nonbleeders. A number of patients (n 50), hospitalized with GI bleeds while taking piroxicam, indomethacin, diclofenac, or naproxen, were evaluated. There were no significant differences in peak plasma concentration, time-to-peak plasma concentration, or AUCpo between bleeders and controls for any of the NSAIDs studied. The authors concluded that their results failed to support the hypothesis. In a separate study, Martin et al. (2001) evaluated the effect of CYP2C9 genotype on the incidence of gastric ulceration in a relatively small number of subjects (n 23) receiving indomethacin, diclofenac, naproxen, ibuprofen, piroxicam, or sulindac. Although some of the subjects were genotyped CYP2C9*1/*2 (17%) and CYP2C9*1/*3 (13%), the incidence of ulceration was not associated with genotype. More recently, Martinez et al. (2004) were able to assess CYP2C9genotyped subjects receiving NSAIDs that underwent “extensive” CYP2C9-dependent metabolism (e.g., celecoxib, diclofenac, ibuprofen, indomethacin, lornoxicam, piroxicam, or naproxen) and other drugs that were not considered CYP2C9 substrates (e.g., salicylates and acetaminophen). The authors conclude that the association of variant CYP2C9 alleles and the risk of acute GI bleeding shows a gene-dose effect, and that it is higher in patients receiving drugs that are metabolized mainly by CYP2C9 (odds ratio of 2.6 when compared with nonbleeding subjects). It is concluded also that CYP2C9 genotyping may identify a subgroup of individuals who are at a potentially increased risk of acute GI bleeding. Interestingly, the observed risk was related largely to the CYP2C9*2 allele, which is unexpected because decreases in the kcat/Km ratio in vitro are more pronounced with recombinant CYP2C9*3 (Rodrigues and Rushmore, 2002). Therefore, the authors hypothesized that the association of CYP2C9*2 with NSAID-related GI bleeding risk may be related to a combined effect of mutations on CYP2C8 (CYP2C8*3 allele) and CYP2C9 (CYP2C9*2 allele), and the work of Yasar et al. (2002) was cited. This raises an interesting possibility that for substrates metabolized by both CYP2C8 and CYP2C9, an impaired clearance in vivo previously attributed to the CYP2C9*2 variant could in part be related to CYP2C8*3. But how many COX inhibitors are metabolized by CYP2C9 and CYP2C8 (Totah and Rettie, 2005)? The reports of Wynne et al. (1998), Martin et al. (2001), and Martinez et al. (2004) focused on the GI side effects associated with COX inhibitors. In all three cases, however, no effort was made to evaluate CYP2C9 genotype in relation to changes in PK and COX inhibition. More importantly, existing P450 reaction phenotype and clinical ADME data for the drugs in each study were not considered. Kinetic Considerations Before considering the role of CYP2C9, it is important to note that many pathways may contribute to the overall clearance of a drug. For example, an absorbed drug may be cleared unchanged via hepatic (biliary) and renal routes. As a result, not all of the dose is eliminated via hepatic metabolism (fh fm 1). Even if a drug is metabolized extensively, it is possible that multiple enzyme systems are involved and the overall clearance is governed by a combination of P450 and non-P450 (e.g., UDP-glucuronsyltransferase, FMO, or AO) pathways (fm 1). At the same time, even if drug elimination depends entirely on the P450 system (fm 1), it is possible that multiple forms of P450 contribute to the overall clearance (fm,CYP 1). Under different scenarios, therefore, the product fm fm,CYP does not equal unity (CLint tot CLint f ) (Rodrigues and Rushmore, 2002). For the sake of discussion, the theoretical relationship (eq. 1) between the product fm fm,CYP (specifically, fm fm,CYP2C9(EM)) and the AUCpo difference in PM (AUCpo(PM)) versus EM (AUCpo(EM)) subjects is shown below (Rodrigues and Rushmore, 2002). A similar relationship is commonly used to evaluate the effect of an inhibitor on the AUC of a substrate. In this instance, however, one is comparing the AUC ratio across subjects of different phenotypes, or genotypes, and it is assumed that the dose, the fraction of the dose absorbed, and the unbound fraction in blood is the same in both EM and PM subjects. In addition, it is assumed that gut first pass is negligible, that the drug is eliminated by the liver only (fh 1), that the elimination process is first order ([S]/Km ratio 0.1), and that no autoinduction occurs (not relevant following a single dose). One has to accept also that hepatic extraction is blood flow-limited (e.g., well stirred model). It is worth noting that recombinant CYP2C9*1 and CYP2C9*3 have been shown to exhibit nonhyperbolic (non-Michaelis-Menten) single Km kinetics with substrates such as naproxen (e.g., biphasic) and piroxicam (e.g., substrate inhibition) (Tracy et al., 2002). The impact of such nonhyperbolic kinetics in vivo is not known. However, if 1568 RODRIGUES at A PE T Jornals on Jne 2, 2017 dm d.aspurnals.org D ow nladed from product formation (parent elimination) is first order, then concerns about kinetic behavior at higher substrate concentrations ( Km) are minimized.