Rst b 20051693

Embed Size (px)

Citation preview

  • 8/12/2019 Rst b 20051693

    1/22

    Pharmacogenetics in drug regulation:promise, potential and pitfalls

    Rashmi R. Shah * ,,

    Pharmaceutical Consultant, 8 Birchdale, Gerrards Cross, Buckinghamshire SL9 7JA, UK

    Pharmacogenetic factors operate at pharmacokinetic as well as pharmacodynamic levelsthe twocomponents of the doseresponse curve of a drug. Polymorphisms in drug metabolizing enzymes,transporters and/or pharmacological targets of drugs may profoundly inuence the doseresponserelationship between individuals. For some drugs, although retrospective data from case studiessuggests that these polymorphisms are frequently associated with adverse drug reactions or failure of efcacy, the clinical utility of such data remains unproven. There is, therefore, an urgent need forprospective data to determine whether pre-treatment genotyping can improve therapy. Variousregulatory guidelines already recommend exploration of the role of genetic factors when investigatinga drug for its pharmacokinetics, pharmacodynamics, doseresponse relationship and drug

    interaction potential. Arising from the global heterogeneity in the frequency of variant alleles,regulatory guidelines also require the sponsors to provide additional information, usuallypharmacogenetic bridging data, to determine whether data from one ethnic population can beextrapolated to another. At present, sponsors explore pharmacogenetic inuences in early clinicalpharmacokinetic studies but rarely do they carry the ndings forward when designing doseresponsestudies or pivotal studies. When appropriate, regulatory authorities include genotype-specicrecommendations in the prescribing information. Sometimes, this may include the need to adjust adose in some genotypes under specic circumstances. Detailed references to pharmacogenetics inprescribing information and pharmacogenetically based prescribing in routine therapeutics willrequire robust prospective data from well-designed studies. With greater integration of pharmacogenetics in drug development, regulatory authorities expect to receive more detailedgenetic data. This is likely to complicate the drug evaluation process as well as result in complexprescribing information. Genotype-specic dosing regimens will have to be more precise and

    marketing strategies more prudent. However, not all variations in drug responses are related topharmacogenetic polymorphisms. Drug response can be modulated by a number of non-geneticfactors, especially co-medications and presence of concurrent diseases. Inappropriate prescribingfrequently compounds the complexity introduced by these two important non-genetic factors. Unlessprescribers adhere to the prescribing information, much of the benets of pharmacogenetics will besquandered.

    Discovering highly predictive genotypephenotype associations during drug development anddemonstrating their clinical validity and utility in well-designed prospective clinical trials will nodoubt better dene the role of pharmacogenetics in future clinical practice. In the meantime,prescribing should comply with the information provided while pharmacogenetic research isdeservedly supported by all concerned but without unrealistic expectations.

    Keywords: adverse drug reactions; doseresponse; drug interactions; ethnic differences;pharmacogenetics; regulatory guidelines

    1. INTRODUCTION

    Thispaperreviewsthe implicationsof genetic inuenceson doseresponse relationship of a drug and the currentapproaches to integrating pharmacogenetics in drugdevelopment. It also summarizes the regulatory frame-work that supports exploration of pharmacogenetic

    inuences during drug development and examinesfuture challenges in genotype-driven development andevaluation of new chemical entities (NCE) and theirclinical use.

    During the clinical use of a drug at present, aprescribing physician has no means of predicting theresponse of an individual patient to a given drug.Invariably, some patients fail to respond benecially asexpected whereas others experience adverse drugreactions (ADRs). Prescribing of drugs is therefore a

    relatively empirical trial and error process, frequentlyresulting in changes in the choice of a drug and/or itsdose. Since the majority of ADRs are type A (A foraugmentation and result from increased plasmaconcentrations) pharmacological reactions, they

    Phil. Trans. R. Soc. B (2005) 360 , 16171638doi:10.1098/rstb.2005.1693

    Published online 25 July 2005

    One contribution of 12 to a Discussion Meeting Issue Geneticvariation and human health.

    * ([email protected]). Previously Senior Clinical Assessor, Medicines and Healthcareproducts Regulatory Agency, London SW8 5NQ, UK. The views expressed in this paper are those of the author and do notnecessarily reect the views or opinions of the MHRA, otherregulatory authorities or any of their advisory bodies.

    1617 q 2005 The Royal Society

  • 8/12/2019 Rst b 20051693

    2/22

    ought to be predictable and preventable. This is ahighly desirable goal since a number of studies haveconsistently shown that ADRs and their managementconstitute a substantial burden on healthcareresources.

    These unexpected responses to drugs frequentlyresult from relatively rigid standard dose schedulesthat are usually recommended, ignoring the diverseinterindividual variability within the patient popu-lation. Hitherto, drugs have been developed, approvedand marketed on a one-size-ts-all basis frompopulation-based mean data on dose, efcacy andsafety. The broad assumption is that all patients are ahomogeneous group showing little or no interindivi-dual variability. This is evident from drug developmentprogrammes that have traditionally tended to minimizeor eliminate, rather than embrace, variability in thepopulations randomized into clinical trials and focusedon a very narrow dose range.

    Pharmacokinetics and pharmacodynamics are thetwo key components of the doseresponse relationshipof a drugthe response being a desired therapeuticeffect or an undesirable ADR. Frequently, there arewide interindividual variations in the pharmacokinetics(inuencing doseconcentration relationship) or thepharmacodynamics (inuencing concentration response relationship) of a drug. These variationsarise from non-genetic as well as genetic inuenceson the functional activity of drug metabolizing enzymesor pharmacological responsiveness of various drugtargets. In terms of genetic inuences, the presence of variant alleles often exerts inuences that far exceedthose due to other covariates that are usually investi-gated during drug development, for example age,gender, co-medications or the presence of concurrentdiseases such as renal or hepatic dysfunction. It istherefore not surprising that the genetic prole(genotype) of an individual may signicantly governthe safety and efcacy outcomes (phenotype) followingpharmacological interventions clinically. Indeed, gen-etic factors as a cause of variations in drug responseshave long been suspected.

    2. PHARMACOGENETIC INFLUENCES ONPHARMACOKINETICS

    Much of the attention in pharmacogenetics hashitherto focused on pharmacokinetics. This is notaltogether surprising since drug levels are easilymeasured and correlated to clinical response (thera-peutic drug monitoring). Furthermore, it was at thelevel of drug metabolizing enzymes that genetic factorswere rst found to inuence drug response. It is nowevident that most drug metabolizing enzymes areexpressed in genetically variant forms with alteredfunctional properties.

    One of the earliest examples of a geneticallydetermined variation in drug response is the prolongedapnoea that follows administration of suxamethonium

    chloride (a muscle relaxant) in some individuals. Thiswas found to be due to the inheritance of a variant formof plasma esterase, butyrylcholinesterase (designatedatypical cholinesterase). However, it was not until late1950s that studies on N -acetylation of isoniazid, a drug

    widely used for the treatment of tuberculosis, resultedin the rst systematic characterization of geneticpolymorphism in drug metabolizing enzymes ( Evanset al . 1960 ; Sunahara et al . 1961 ). A population couldbe divided into slow or rapid acetylators. Subsequently,variations in responses to a number of drugs metab-olized by acetylation were shown to be associated with N -acetylation status (slow or rapid acetylators) of individual patients ( Evans 1996 ; Furet et al . 2002 ;Hiratsuka et al . 2002 ).

    However, the number of drugs metabolized primar-ily by acetylation is very few. By far the vast majority of drugs are metabolized by enzymes that belong to asuperfamily known as cytochrome P450 or CYPs.Between 50 and 60% of drugs undergoing metabolicelimination are metabolized by CYP enzymes (alsoknown as isoforms). Although a large number of theseCYP isoforms are known to occur in nature, the onesthat metabolize a vast majority of drugs used clinicallyare CYP2D6, CYP3A4, CYP2C19 and CYP2C9(Daly 2004 ). Other isoforms involved relatively lessfrequently are CYP1A2, CYP2A6, CYP2B6, CYP2C8and CYP2E1. Increasingly, genetic polymorphisms arebeing uncovered not only in these CYP isoforms butalso in other drug metabolizing enzymes that arerelevant to the development and clinical use of medicines ( Daly 2003 ). These include various methyl-transferases ( Weinshilboum 1984 ) including thiopur-ine S -methyltransferase (TPMT) ( Schaeffeler et al .2004 ), UDP-glucuronosyltransferases (UGT)(Burchell 2003 ; Guillemette 2003 ), sulfotransferases(Coughtrie e t a l . 1999 ; Carlini e t a l . 2001 ) anddihydropyrimidine dehydrogenase ( Gardiner et al .

    2002 ).The potential clinical impact of pharmacogeneti-cally determined variability in the activity of drugmetabolizing enzymes, and therefore on a drugresponse, is best illustrated by genetic polymorphismof CYP2D6. It is not only the rst CYP polymorphismto be discovered and whose molecular basis wasdetermined but it is also the most widely studied andbest characterized. Studies in mid-1970s showed thatany given population may be divided into two CYP2D6drug metabolizing phenotypesextensive metabolizers(EMs) or poor metabolizers (PMs)depending ontheir ability to mediate CYP2D6-dependent hydroxy-

    lation of the (now obsolete) antihypertensive drugdebrisoquine ( Mahgoub et al . 1977 ; Eichelbaum et al .1979 ; Evans et al . 1980 ). This polymorphism resultsfrom autosomal recessive inheritance, in a simpleMendelian fashion, of alleles at a single locus onchromosome 22. Well over 70 CYP2D6 alleles havebeen identied to date and the details of these can beaccessed from a specially dedicated website at http://www.imm.ki.se/cypalleles/ .

    Since the wild type allele (CYP2D6 * 1) responsiblefor normal functional activity is dominant, only thoseindividuals carrying two CYP2D6 inactivating alleles(e.g. CYP2D6 * 3, CYP2D6 * 4, CYP2D6 * 5 or

    CYP2D6*

    6) are phenotypic PMs. However, amongthe EM phenotype, there are two subgroups of particular interest at either extreme of the EMpopulation distribution. One subgroup, termed theultrarapid metabolizers (UMs), is comprised of

    1618 R. R. Shah Pharmacogenetics and drug regulation

    Phil. Trans. R. Soc. B (2005)

    http://www.imm.ki.se/cypalleles/http://www.imm.ki.se/cypalleles/http://www.imm.ki.se/cypalleles/http://www.imm.ki.se/cypalleles/
  • 8/12/2019 Rst b 20051693

    3/22

    individuals with gene amplication and possessingmultiple copies of the gene for normal metaboliccapacity. UM phenotype may also be associated withinheritance of at least two unique alleles of CYP2D6(CYP2D6 * 35 and CYP2D6 * 41). The other group,termed the intermediate metabolizers (IMs), is com-prised of a heterozygous genotype (gene-dose effect)with a moderate impairment in drug metabolizingcapacity.

    The pharmacokinetic consequences of polymorph-ism in CYP2D6, summarized in table 1 , are thatrelative to EMs, the PMs experience far greaterexposure to the parent drug ( Idle & Smith 1984 ) anda markedly reduced exposure to the metabolitesgenerated by this enzyme. In contrast, UMs areexposed to high concentrations of rapidly accumulatingmetabolites and even at very high doses, attain only lowplasma levels of the parent drug. Retrospectivecandidate gene association studies on cases andcontrols have shown that the PM genotype is associatedwith an increased risk of a number of ADRs to drugsthat are primarily cleared by CYP2D6-mediatedmetabolism as well as being at risk of lack of efcacywhen the therapeutic effect of a drug is mediatedprincipally by its CYP2D6-generated metabolite. Forexample, since PMs cannot carry out the metabolicactivation of the pro-analgesic drug codeine tomorphine, they fail to derive adequate analgesicefcacy from codeine. In contrast, UM patients fail torespond to conventional doses of drugs metabolized byCYP2D6 when the therapeutic activity resides in theparent drug and often require megadoses of the drugconcerned e.g. the antidepressant nortriptyline or the

    antianginal drug perhexiline. UM patients are also atrisk of toxicity from rapidly accumulating metabolites.For example, rapid conversion of codeine to morphinein these individuals predisposes them to morphinetoxicity (e.g. epigastric pain). Table 2 summarizes otherimportant clinical outcomes that may be associatedwith the above pharmacokinetic consequences of CYP2D6-mediated metabolism.

    It is not only the ADRs and efcacy of a drug whenused clinically as a single agent that are inuenced byCYP2D6 genotype of the patient. Drugdrug inter-actions also show dramatic inter-genotypic differences.For example, CYP2D6 PMs (with alleles expressing no

    functional enzyme) do not show the drugdruginteractions predicted from in vitro studies. This ishardly surprising since there is no functional CYP2D6activity that can be inhibited. In contrast to some otherdrug metabolizing enzymes, CYP2D6 is not inducible(Eichelbaum et al . 1986 ; Wadelius et al . 1997 ; Dilgeret al . 1999 ; Branch et al . 2000 ). Likewise, UMs too mayfail to exhibit the expected drugdrug interaction. Incontrast to PMs and EMs, the UMs have a sufcientlylarge functional reserve of CYP2D6 activity that theywould most probably need much higher (and poten-tially toxic) doses of the inhibitor to elicit an interaction(Dalen et al . 1998 ; Dalen et al . 2000 ). Under normal

    conditions of use, the individuals most likely to displaya drug interaction are those who have an intermediateor otherwise compromised drug metabolizing capacity(IMs) or those who have inherited CYP2D6 alleleswith reduced or altered afnity for CYP2D6 substrates.

    At the level of CYP2D6, the anticipated dependence of drug interactions on the metabolic phenotype hasalready been conrmed for a number of CYP2D6substrates, for example encainide ( Turgeon e t al .1990 ), mexiletine ( Turgeon et al . 1991 ), desipramine(Brosen et al . 1993 ), propafenone ( Morike & Roden1994 ), codeine ( Caraco et al . 1999 ) and metoprolol(Hamelin et al . 2000 ). It should be obvious that inPMs, interactions at alternative pathways of metab-olism can nonetheless still occur. The point is well

    Table 1. Pharmacokinetic consequences of CYP2D6 poly-morphism.

    pharmacokinetic parameter consequences for the PMrelative to EM

    bioavailability 25 foldsystemic exposure

    C max 26 foldAUC 25 foldhalf life 26 foldmetabolic clearance 0.10.5 fold

    Table 2. Clinical consequences for PM and ultrarapid EMphenotypes of CYP2D6.

    clinical consequences for the PM

    increased risk of toxicitydebrisoquine postural hypotension and physical

    collapse

    sparteine oxytocic effectsperphenazine extrapyramidal symptomsecainide possibly ventricular tachyarrhythmiasperhexiline neuropathy and hepatotoxicityphenformin lactic acidosispropafenone CNS toxicity and bronchoconstrictionmetoprolol loss of cardioselectivitynortriptyline hypotension and confusionterikalant excessive prolongation in QT intervaldexfenuramine nausea, vomiting and headacheL -tryptophan eosinophilia-myalgia syndromeindoramin sedationthioridazine excessive prolongation in QT intervaltramadol hyper-anticoagulation from warfarin

    failure to respond

    codeine poor analgesic efcacytramadol poor analgesic efcacyopiates protection from oral opiate dependence

    clinical consequences for the ultrarapid EM

    increased risk of toxicityencainide possibly proarrhythmiascodeine morphine toxicity

    failure to respond

    nor triptyline poor antidepressant efcacy

    at normal dosespropafenone poor antiarrhythmic efcacyat normal doses

    tropisetron poor antiemetic efcacy atnormal doses

    ondansetron poor antiemetic efcacy atnormal doses

    Pharmacogenetics and drug regulation R. R. Shah 1619

    Phil. Trans. R. Soc. B (2005)

  • 8/12/2019 Rst b 20051693

    4/22

    illustrated by an interesting, but expected, genotype-dependent interaction between propafenone andrifampicinan enzyme inducer ( Dilger et al . 1999 ).CYP2D6 that primarily metabolizes propafenone (by5-hydroxylation) is non-inducible. However, coadmi-nistration of rifampicin decreased the oral bioavail-ability of propafenone from 30 to 10% in EMs and

    from 81 to 48% in PMs. Following oral propafenone,clearances through N -dealkylation (mediated byCYP3A4 and CYP1A2) and glucuronidation, but notCYP2D6-mediated 5-hydroxylation, increased regard-less of CYP2D6 phenotype indicating substantialenzyme induction of these enzymes. Thus, inductionof non-CYP2D6 pathways by rifampicin resulted in aclinically relevant metabolic drug interaction withpropafenone that was more pronounced in EMs thanin PMs with regard to percentage decrease inbioavailability of propafenone.

    The efcacy of proton pump inhibitors such asomeprazole and lansoprazole, both metabolized by

    CYP2C19, in reducing gastric acid secretion is due toparent drugs. Studies have shown that PMs of CYP2C19 have higher therapeutic response rateswhile EMs of CYP2C19 generally require higherdoses of these drugs ( Klotz et al . 2004 ). Arising fromlow gastric acidity, subjects of CYP2C19 PM genotypetend to have lower serumlevels of vitamin B 12 followinglong-term treatment with omeprazole ( Sagar et al .1999 ) and presumably other proton pump inhibitors.The antimalarial drug proguanil is activated byCYP2C19 to therapeutically more potent cyclopro-guanil, a strong dihydrofolate reductase inhibitor, andtherefore, PMs of CYP2C19 may be at risk of inadequate antimalarial efcacy ( Kaneko et al .1999 b). The dependence of drugdrug interactionson genotype has also been reported at CYP2C19 ( Choet al . 2002 ; Suzuki et al . 2003 ; Itagaki et al . 2004 ; Wanget al . 2004 ; Yasui-Furukori et al . 2004 a ,b).

    Apart from acetylation, polymorphisms of otherconjugation reactions such as glucuronidationmediated by UGTs are now also attracting increasingattention, especially in the eld of oncology. Glucur-onidation is by far the most important conjugationpathway in man. A multigene family encodes theUGTs and a relatively small number of human UGTenzymes catalyse the glucuronidation of a wide range of structurally diverse endogenous (bilirubin, steroidhormones and biliary acids) and exogenous chemicals.Genetic variations and single nucleotide polymorphisms(SNPs) within the UGT genes are remarkably common(Burchell 2003 ; Guillemette 2003 ). Two major isoformsof UGT, UGT1A1 and UGT1A9, show wide inter-individual variability in their activities and displaygenetic polymorphisms that have a signicant pharma-cological impact in terms of ADRs. Whereas irinotecanand avopiridol are metabolized by UGT1A, tranilastand atazanavir inhibit this isoform. Retrospectivestudies investigating the role of UGT1A isoforms in

    the safety of irinotecan ( Ando et al . 2000 ; Iyer et al .2002 ; Marcuello et al . 2004 ; Rouits et al . 2004 ),avopiridol ( Innocenti et al . 2000 ; Ramirez et al . 2002 ),tranilast ( Danoff et al . 2004 ) and atazanavir ( Shaw2002 ) have been most valuable in explaining the

    clinical concerns (myelosuppression, diarrhoea orhyperbilirubinaemia) associated with these drugs.

    Long after the removal of troglitazone (a novel andvaluable 2,4-thiazolidinedione oral hypoglycaemicagent with insulin-sensitizing activities) from themarket due to its frequent, severe and often fatalhepatotoxicity, Watanabe et al . (2003) undertook acasecontrol study of 25 cases of troglitazone-inducedhepatotoxicity and 85 controls that investigated 68polymorphic sites in 51 candidate genes related to avariety of biochemical functions. They reported astrong correlation between elevations of transaminasesand the combined glutathione- S -transferase GSTT1-GSTM1 null genotype (odds ratio 3.69, 95% CI of 1.35410.066, pZ 0.008). This retrospective obser-vation not only raises the possibility of a strong geneticsubstrate in troglitazone-induced hepatotoxicity butalso emphasizes the need for candidate gene associationstudies to investigate associations with a panel of manydiverse genes.

    3. PHARMACOGENETIC INFLUENCESON PHARMACODYNAMICS

    It has been known for some time now that as withpolymorphisms of drug metabolizing enzymes, phar-macological targets of drugs also display geneticpolymorphisms and that these too inuence drugresponse. Variant alleles are known to occur not onlyat the genes expressing target enzymes, channels andreceptors but also at the genes responsible forintracellular signal transduction. Better characterizedamong these pharmacodynamic polymorphisms are thetargets involved in cardiac arrhythmias (mutations of sodium and potassium channels), asthma (mutationsof b2-adrenoceptors and of the core promoter of 5-lipoxygenase), cardiac failure (mutations of b2-adrenoceptors) and depression (mutations of thepromoter region of serotonin transporter). While mostindividuals may respond normally as expected,individuals with a genetic variant of a pharmacologicaltarget may exhibit a quantitatively or qualitativelydifferentexaggerated, inadequate or unexpected response even when the concentration of the drug iswithin the population-based normal therapeutic range.

    The QT interval of the surface electrocardiogram(ECG) reects the duration of ventricular actionpotential that is determined by a net balance betweeninward depolarizing and outward repolarizing currents,especially during phase 3 of the action potential. Themajor determinant of the outward repolarizing currentis known as IKr and is conducted by the rapidcomponent of the delayed rectier potassium channel.Reduction in this current results in QT intervalprolongation. Excessive prolongation of the QTinterval often leads to potentially fatal ventriculartachyarrhythmias, particularly a variety known astorsade de pointes (TdP). Over the last 10 years, many

    non-antiarrhythmic drugs have attracted considerableclinical and regulatory interest because of theirpotential to prolong the QT interval and induce TdP(Shah 2002 ). These drugs target and inhibit primarilythe IKr current.

    1620 R. R. Shah Pharmacogenetics and drug regulation

    Phil. Trans. R. Soc. B (2005)

  • 8/12/2019 Rst b 20051693

    5/22

    Following advances in molecular biology, geneticsand pharmacology of ion channels, it has becomeevident that there is a great diversity of genes thatcontrol the expression of these potassium channels(Escande 2000 ). Mutations of the genes that encodesubunits of these channels are common and give rise tocongenital long QT syndromes (LQTS). However, inview of the low penetrance of many of these mutations,the size of the population with dysfunctional potassiumchannels is substantially larger than that diagnosed byECG recording alone. Because of considerable overlap,measurement of the QT interval alone may not permitan accurate molecular diagnosis in families affected bythe congenital long QT syndrome. Not all allelecarriers have symptoms and only the DNA markersmake it possible to reach a genetic diagnosis in theseindividuals ( Vincent et al . 1992 ; Saarinen et al . 1998 ;Priori et al . 1999 ). Although the affected individualshave a normal ECG phenotype, they have diminishedrepolarization reserve nonetheless and are highlysusceptible to drug-induced QT interval prolongationand/or TdP, even at the recommended doses that arenormally safe. Individuals who develop drug-inducedprolongation of QT interval with or without TdP arenot usually genotyped but available evidence suggeststhat a substantial proportion of the cases of the drug-induced long QT syndrome might represent cases of forme fruste of the congenital long QT syndrome. Shah(2004) has recently reviewed the pharmacogenetics of drug-induced QT interval prolongation and TdP.

    Regarding polymorphisms of target receptors, indi-viduals who carry Arg16/Gly16 or Gly16/Gly16variants of b2-adrenoceptors have been shown to

    display a much less favourable immediate bronchodi-latory response to salbutamol, in contrast to those withwild type receptor characterized by Arg16/Arg16genotype. Polymorphisms of b2-adrenoceptors mayalso inuence airway responses to regular inhaledb -agonist treatment. Patients with Arg16/Arg16 geno-type who use salbutamol regularly show a small declinein morning peak expiratory ow (AM PEF). By the endof a 16-week study, Arg16/Arg16 subjects who hadused salbutamol regularly had an AM PEF 30.5 G12.1 l min K 1 lower ( pZ 0.012) than Arg16/Arg16patients who had used salbutamol only intermittentlyas needed. Subjects with Gly16/Gly16 genotype

    showed no such decline. Evening PEF also declinedin the Arg16/Arg16 regular users but not in those whoused it intermittently on as-needed basis ( Israel et al .2000 ). Similarly, asthmatic patients who carrymutations of the core promoter of 5-lipoxygenase(ALOX-5) respond poorly to ALOX-5 inhibitors suchas zileuton ( Drazen et al . 1999 ). Among patients withcardiac failure, those who are homozygous for theGln27 allele of the b2-adrenoceptor display a signi-cantly lower proportion of good responders to treat-ment with carvedilol than do the patients who arehomozygous or heterozygous for the Glu27 allele (26%versus 63%, pZ 0.003) ( Kaye et al . 2003 ). Cardiac

    failure patients who carry a b2-adrenoceptor allele witha mutation at codon 164 (Thr164/Ile164) have a 1-yearsurvival of 42% in contrast to 76% in those carrying thenormal wild type allele (Thr164/Thr164) ( Liggett et al .1998 ). Provided this observation can be replicated

    widely, it would argue for an earlier intervention(including cardiac transplantation) in the formergroup.

    Genetic polymorphism in the promoter region of theserotonin transporter (5-HTT) gene is reportedly adeterminant of response to uvoxamine, a selectiveserotonin re-uptake inhibitor (SSRI). The insertionvariant of this polymorphism (long allele) is associatedwith higher expression of brain 5-HTT compared tothe deletion variant (short allele) ( Weizman & Weizman2000 ). Patients who have one or two copies of the longvariant (homozygous l/l or heterozygous l/s) may showa better therapeutic response than patients who arehomozygous for the short variant (s/s). The safety andefcacy of a number of other SSRIs have also beenshown to correlate with these 5-HTT genotypes ( Kimet al . 2000 ; Pollock et al . 2000 ; Arias et al . 2003 ; Perliset al . 2003 ; Durham et al . 2004 ).

    There are now ever increasing numbers of reports of clinically highly relevant polymorphisms in otherpharmacological targets.

    4. PHARMACOGENETICS AND REGULATORYFRAMEWORK

    Given that genetic factors frequently determine inter-individual and inter-ethnic differences in pharmacoki-netics and pharmacodynamics of a drugwith all theassociated implications for failure of efcacy in somepatients and predisposition to ADRs and druginteractions in othersit is not surprising that regulat-ory authorities have long recognized the signicance of pharmacogenetics in drug development and are now

    increasingly directing their attention to addressingissues that may arise from genetic heterogeneity of thetarget patient population.

    A number of guidelines from the European UnionsCommittee for Proprietary Medicinal Products(CPMP), now known as Committee for MedicinalProducts for Human Use (CHMP), and the Inter-national Conference on Harmonization (ICH) alreadymake direct or indirect references to the need foraddressing genetic factors when developing an NCE(table 3 ).

    Indeed, the guideline on Pharmacokinetic studies inman, adopted by the CPMP as long ago as February

    1987, was probably the rst regulatory guideline toinclude direct references to genetic factors in determin-ing drug response ( CPMP 1998 ). This guidelinerecommends that metabolic studies should indicatewhether the metabolism of a drug may be substantiallymodied in a case of genetic enzyme deciency andwhether saturation of metabolism may occur, therebyresulting in nonlinear kinetics, within the dose levelsnormally used.

    The ICH guideline on Doseresponse informationto support drug registration describes how helpful theknowledge of the shape of individual doseresponsecurves is and it distinguishes individual curves from the

    population curve ( CPMP/ICH 1995 ). The guidelineclearly warns that Choice of a starting dose mightalso be affected by potential intersubject variabilityin pharmacodynamic response to a given bloodconcentration level, or by anticipated intersubject

    Pharmacogenetics and drug regulation R. R. Shah 1621

    Phil. Trans. R. Soc. B (2005)

  • 8/12/2019 Rst b 20051693

    6/22

    pharmacokinetic differences, such as could arise fromnonlinear kinetics, metabolic polymorphisms or a highpotential for pharmacokinetic drugdrug interactionsand recommends that in utilizing doseresponseinformation, the inuences of various demographicfeatures, individual characteristics (including meta-bolic differences) and concurrent drugs and diseasesshould be identied as far as possible.

    Since drug interactions are genotype-dependent asdiscussed earlier, the CPMP guideline on Investi-gation of drug interactions recommends that whenperforming mechanism-based in vivo studies, eitherwhen studying the effects of inhibition or induction onthe pharmacokinetics of an NCE, consideration shouldbe given to pharmacogenetic factors ( CPMP 1997 ).Subjects participating in metabolic in vivo interactionstudies should be appropriately genotyped and/orphenotyped (with respect to their drug metabolizingcapacity) at the beginning of the study if any of theenzymes mediating the metabolism of the interactingdrugs are polymorphically distributed in the popu-lation. As an extension to this, although not explicitlystated in the guideline, genotype of the donor liver usedfor in vitro microsomal studies should also beascertained. This CPMP guideline also recommendsinvestigation of drug interactions at sites other thanmetabolic route such as renal excretion and transportby efux pumps and P-glycoprotein. Pharmacogeneticfactors are also likely to be important when consideringand evaluating drug interactions at these transportersand/or P-glycoprotein.

    The CPMP guidance note on Investigation of bioavailability and bioequivalence also recommendsthat phenotyping and/or genotyping of subjects may beconsidered for safety or pharmacokinetic reasons(CPMP 2001 ).

    The US Food and Drug Administration (FDA)issued in April 1997 their guidance note Drugmetabolism/drug interaction studies in the drugdevelopment process: studies in vitro (FDA 1997 ).This states Identifying metabolic differences in patientgroups based on genetic polymorphisms, or on otherreadily identiable factors such as age, race, andgender, could help guide the design of dosimetrystudies for such populations groups. This kind of information also will provide improved dosing rec-

    ommendations in product labelling, facilitating the safeand effective use of a drug by allowing prescribers toanticipate necessary dose adjustments. Indeed, in somecases, understanding how to adjust dose to avoidtoxicity may allow the marketing of a drug that would

    have an unacceptable level of toxicity were its toxicityunpredictable and unpreventable. The JapaneseMinistry of Health, Labour and Welfares drugregulatory authority (Koseisho, now known as thePharmaceuticals and Medical Devices Agency,PMDA) has also issued guidelines in June 2001 thatrecommend genotyping in all drug developmentprogrammes for drugs that are metabolized by

    cytochrome P450s ( MHLW 2001 a ,b).Although the requirements to address these genetic

    factors are stated in different terms by differentregulatory bodies, the net effect of these requirementsis that new knowledge concerning pharmacogeneticvariations in drug disposition (pharmacokinetics) orresponsiveness of pharmacological targets (pharmaco-dynamics) will lead to additional requirements forpharmacogenetic documentation for NCEs.

    5. PHARMACOGENETICS AND GLOBAL DRUGDEVELOPMENT

    In todays age, drug development programmes areundertaken at a global level. This is aimed at reducingthe costs, expediting the drug development process andaddressing the issues arising from global prescribing of drugs. However, the relative frequency of variousalleles of drug metabolizing enzymes varies in differentpopulations. Consequently, the frequencies of the PMphenotype as well of those with intermediate drugmetabolizing capacity also show a marked globalheterogeneity. For example, the frequency of CYP2D6 PM phenotype is much higher in populationgroups of western Caucasian origin (510%) than inFar East and Asian ethnic groups (02%) ( Bradford

    et al . 1998 ; Bradford 2002 ; Mizutani 2003 ; Shimizuet al . 2003 ; Ozawa et al . 2004 ). The frequency of PMsof CYP2C19 is lower in western Caucasians (24%)compared to the frequencies observed among Orientals(about 1525%), reaching as high as 6070% inVanuatu and other Pacic islands ( Kaneko et al .1999 c; Xie et al . 2001 ). Global heterogeneity andinter-ethnic differences have also been reported in thefrequency of variant alleles of the genes expressingmany other drug metabolizing enzymes ( Lin et al .1994 ; Collie-Duguid et al . 1999 ; Hon et al . 1999 ;Scordo et al . 2001 ; Lee et al . 2002 ; Xie et al . 2002 ) andof the gene expressing P-glycoprotein (MDR1, also

    known as ABCB1) ( Ieiri et al . 2004 ; Marzolini et al .2004 a ). For example, whereas the variant allele styledas MDR1 * 2 occurred in 62% of European Americans,it was reported in only 13% of African Americans ( Kimet a l . 2001 ), although the functional or clinical

    Table 3. Pharmacogenetics and CPMP and ICH guidelines.

    genetic factors in pharmacokinetics1 pharmacokinetic studies in man2 investigation of drug interactions3 ICHethnic factors in the acceptability of foreign clinical data4 investigation of bioavailability and bioequivalence5 ICHdoseresponse information to support drug registration . metabolic

    polymorphism . genetic factors in pharmacodynamics6 ICHdoseresponse information to support drug registration variability in

    pharmacodynamic response .

    1622 R. R. Shah Pharmacogenetics and drug regulation

    Phil. Trans. R. Soc. B (2005)

  • 8/12/2019 Rst b 20051693

    7/22

    signicance of most of the alleles of MDR1 iscontroversial and not yet adequately characterized(Chowbay et al . 2003 ; Eichelbaum et al . 2004 ; Soranzoet al . 2004 ).

    As with drug metabolizing enzymes, there also existinter-ethnic differences in the mutations of a number of pharmacological targets. For example, the frequenciesof Arg16/Gly16 mutation of b2-adrenoceptor (alsoresponsible for enhanced agonist-promoted down-regulation and more frequent in nocturnal asthma) inCaucasians, blacks and Asians are 0.61, 0.50 and 0.57,respectively ( Weir et al . 1998 ). In terms of haplotypes of mutations of b2-adrenoceptor, thirteen of the SNPs onb 2-adrenoceptor gene are organized into 12 haplotypesout of the theoretically possible 8192 combinations(Drysdale et al . 2000 ). Four of the observed haplotypeswere found in all the four populations sampled(Caucasian, African American, Asian, and HispanicLatino), although at markedly different frequencies,with one of them showing a more than 20 folddifference in its frequencies. Likewise, while thefrequency of long allele of 5-HTT may be as high as87% in some African American populations, it is as lowas 56% in some European Americans ( Lotrich et al .2003 )

    Regulatory guidelines also recognize the potentialinter-ethnic differences in pharmacokinetics and phar-macodynamics that might result from this globalgenetic heterogeneity. Information is therefore requiredon ethnic demography of clinical trial populations andpotential ethnic inuences on drug response(CPMP/ICH 1998 ). In terms of drug development,the signicance of these requirements lies in the facts

    that (i) there is an increasing globalization of drugdevelopment with clinical trials often conducted in ageographical population which may not be the ultimatetarget of the drug, (ii) more and more of the NCEs arefound to be substrates of polymorphic drug metaboliz-ing enzymes and/or targeted towards polymorphicreceptorsthis information is usually not known forolder drugs already on the market and (iii) that moderndrugs are more potent with narrow therapeutic indicesand therefore, relatively small differences in either thepharmacokinetics or pharmacodynamics may becomehighly relevant clinically. Inter-ethnic differences indrug response are well known ( Xie et al . 2001 ) and

    therefore, the sponsors of NCEs are now anxious toaddress the issues arising from global prescribing of their drugs. It is also evident that drugdrug inter-actions too may depend on ethnicity ( Caraco et al .1995 ).

    The ICH guideline on Ethnic factors in theacceptability of foreign clinical data recommends thesponsors and the regional regulatory authority in a newregion to assess an application for registration for theability to extrapolate to the new region those parts of the application based on studies from the trial region(CPMP/ICH 1998 ). To this end, it is recommendedthat the submission should include (i) adequate

    characterization of pharmacokinetics, pharmacody-namics, doseresponse, efcacy and safety in thepopulation of the trial region and (ii) characterizationof pharmacokinetics, pharmacodynamics and dose response in the new region. The guideline recognizes

    the role of genetic factors and when inter-ethnicdifferences are anticipated, bridging studies may berequired. Often, such studies are pharmacogenetic innature but at times, more extensive data may berequired.

    An argument is often advanced that interindividualvariability far exceeds inter-ethnic variability. This isnot in dispute but the argument overlooks the realitythat it is not ethnicity but the genotype of the trialpopulation that imposes the hurdle. The individuals of one genotype, for example the PMs of CYP2D6, formone distinct subgroup of regulatory interest regardlessof their ethnicity. Ethnicity becomes an important issueonly when the trial population is not characterized forits genetic prole, inter-genotype differences in phar-macokinetics or pharmacodynamics are not evaluatedand the frequency of the variant alleles is substantiallydifferent between the trial and the target populations.

    6. PHARMACOGENETICS AND DRUGPRESCRIBING INFORMATION

    When pharmacogenetic information from early phar-macokinetic studies is considered clinically relevant,regulatory authorities have always reacted with appro-priate labelling recommendations. Four drugs bestillustrate the current regulatory approach to incorpor-ating candidate gene-based pharmacogenetic data,gathered from prospective clinical pharmacokineticstudies, in the prescribing information when thisinformation is thought to be potentially relevant tosafe and effective prescribing.

    Thioridazine is metabolized by CYP2D6 and

    because of the risk of QT interval prolongation andTdP, it is contraindicated in patients with reducedactivity of CYP2D6. Sertindole, an atypical neurolepticagent, is primarily cleared by CYP2D6. In order toprotect the PMs who utilize an alternative pathwaymediated by CYP3A4, coadministration of sertindoleis contraindicated with ketoconazole and itraconazole,both powerful inhibitors of CYP3A4. The usual doseof the antidepressant escitalopram, the ( C )-( S )-enantiomer of racemic citalopram, is 10 mg oncedaily that may be increased to a maximum of 20 mgdaily. However, for patients who are known to be PMswith respect to CYP2C19, the recommendation dose is

    5 mg during the rst two weeks of treatment which maybe increased to 10 mg. Celecoxib, a COX-2 selectiveinhibitor, is predominantly metabolized by CYP2C9and, therefore, caution is recommended when cele-coxib is prescribed to patients known to be PMs of CYP2C9. Since uconazole inhibits CYP2C9, it is alsorecommended that celecoxib should be used at half thenormal doses in patients receiving uconazole. Arisingfrom the wide inter-ethnic differences in the pharma-cokinetics of this drug, an initial lower dose isrecommended in black patients.

    When additional data are or become available, anumber of sections of the prescribing information (e.g.

    dose schedules, contraindications, special warningsand precautions for use, drug interactions, ADRs)may have to be written in terms of pharmacogeneticprole of the patients. The most recently approveddrug that best illustrates this expected complexity of

    Pharmacogenetics and drug regulation R. R. Shah 1623

    Phil. Trans. R. Soc. B (2005)

  • 8/12/2019 Rst b 20051693

    8/22

    prescribing information is atomoxetine ( FDA 2002 ).This drug, approved by the US FDA in December2002, is indicated for attention decit hyperactivitydisorder and is metabolized primarily throughCYP2D6 and extensive genotype-related informationis included in a number of sections of the US labellingincluding drug interactions and ADRs.

    One risk that needs to be highlighted is that unlessthe positive and negative predictive values of genotype phenotype associations are high, dosing recommen-dations based exclusively on genotype may well lead to(sub-therapeutic) under-dosing in some individualssimply because of their genotypes and not necessarilybecause they are at risk of toxicity. For diseases that arelife threatening, this may have serious consequences forthe patients concerned. A collective analysis of the dataon irinotecan, indicated for metastatic colorectalcancer, well illustrates the point. UGT1A1 * 28 geno-type has a positive predictive value of only 50% and anegative predictive value of 9095% for toxicity and,therefore, the expert view from the US ClinicalPharmacology Subcommittee in November 2004 wasthat this test should not be used in isolation but coupledwith other information such as monitoring the patient,using a lower dose, pre-existing risk factors ( FDA2004 ).

    7. PHARMACOGENETICS IN DRUGDEVELOPMENT TO DATE

    In order to comply with various regulatory recommen-dations, sponsors of an NCE already conduct formalearly pharmacology studies in a genotyped panel of

    healthy volunteers to characterize pharmacogeneticinuences on pharmacokinetics of the NCE. Unfortu-nately, however, the ndings are rarely carried forwardto improving the designs and inclusion criteria forsubsequent dose nding and pivotal studies.

    It is most unusual to see dose nding studies thatinclude information on the genotype of the individualsrandomized. It is therefore uncertain that the entirerange of variability in doseresponse, and doserequirements likely to be encountered in the targetpopulation at large, has been studied. This deciencyhas serious implications for selecting the most appro-priate dose(s) for a pharmacogenetically heterogeneous

    population that will be randomized into the pivotalstudies and, ultimately, for the wider clinical use of thedrug. It may be noted that debrisoquine, the prototypesubstrate of CYP2D6, was found during its post-marketing period to be clinically effective in the doserange extending from 20 to 360 mg daily. In the contextof CYP2D6-mediated metabolism, the dose require-ments for nortriptyline and perhexiline also illustratethe point ( Meyer 2000 ; Sallustio et al . 2002 ). Fortherapeutic effect of nortriptyline, UMs require morethan 500 mg daily in contrast to PMs who need only2050 mg daily ( Meyer 2000 ). There is a range of doses in the groups between these two extremes.

    Perhexiline, an effective antianginal drug, was associ-ated with disabling neuropathy, hepatitis and weightloss. It was therefore withdrawn from clinical use in1985. When prescribed at the recommended dose of 100 mg three times a day, PMs are at a much greater

    risk of perhexiline-induced neuropathy and hepatitis(Shah et al . 1982 ; Morgan et al . 1984 ). It is now knownthat to maintain the plasma concentrations of perhexi-line within the therapeutic and non-toxic range, PMsrequire a dose of 1025 mg daily while EM andultrarapid EM require 100250 and 300500 mgdaily, respectively ( Sallustio et al . 2002 ). The perils of prescribing a standard dose of CYP2D6 substratedrugs to all patients, regardless of their CYP2D6metabolic capacity, are obvious. This one-size-ts-allapproach exposes some individuals to concentration-dependent ADRs. Kirchheiner et al . (2001) haveproposed a preliminary guidance for a number of drugs metabolized by CYP2D6 and CYP2C19 with aview to pioneering genotype/phenotype-specic doseschedules.

    Patients in pivotal clinical studies are seldom, if ever,genotyped. Even those patients that withdraw from thestudies because of failure of efcacy or development of a serious ADR do not attract any further attention. Andyet, these are the patients who are most likely torepresent or include outliers of pharmacogeneticinterest.

    8. LIMITATIONS OF CURRENT FOCUSON PHARMACOKINETICS

    Data from retrospective studies aimed at associatinggenetically determined variations in drug metabolizingactivity (genotype) with variations in drug response(phenotype) raised expectations in the 1980s and1990s that polymorphisms in drug metabolizingenzymes might substantially explain the lack of efcacy

    or induction of ADRs associated with their substratedrugs at a population level. If conrmed prospectively,pre-treatment genotyping of patients may offer excitingprospects of removing guesswork from prescribing andimproving therapeutic skills. However, despite thevariability in pharmacokinetics of a wide range of drugs metabolized by highly polymorphic enzymessuch as CYP2D6, there is insufcient evidence tosupport the notion that these polymorphisms areactually associated with altered outcomes and/or drugtoxicity in routine clinical practice ( Wedlund & de Leon2004 ). Meta-analysis of small-scale studies has showngenotyping to be less promising than had been

    anticipated. Predictive CYP2D6 genotyping is esti-mated to be benecial for treatment of about 3040%of CYP2D6 drug substrates, that is, for about 710%of all drugs used clinically, although prospective clinicalstudies are necessary to evaluate the exact benet of drug selection and dosage based on the CYP2D6genotype ( Ingelman-Sundberg 2005 ). Other drugmetabolizing polymorphisms also suffer from similarlack of prospective data on pre-treatment genotyping asa tool to guide prescribing.

    The disappointing and often conicting or ambig-uous reports on the clinical signicance of geneticallydetermined pharmacokinetic variability may simply

    indicate the limitations of single candidate geneapproach. But there are other more compelling reasonsfor guarded scepticism regarding the clinical signi-cance of pharmacogenetic variability in pharma-cokinetics. In the context of polymorphic drug

    1624 R. R. Shah Pharmacogenetics and drug regulation

    Phil. Trans. R. Soc. B (2005)

  • 8/12/2019 Rst b 20051693

    9/22

    metabolizing enzymes, some limitations in applyingpharmacogenetics to therapeutics are already self-evident. These are:

    (i) Pharmacokinetic variability will likely be rel-evant only when the pharmacology (activity andpotency) of the parent drug and its metabolite issignicantly different and the drug response hasa steep concentrationresponse curve. For thosedrugs where both the parent drug and themetabolite(s) are pharmacologically active, theconsequences of defective metabolism woulddepend on the overall contribution of the parentdrug and the metabolite(s) to the therapeutic (ortoxic) effects of the drug.

    (ii) Very few drugs are metabolized by a singleenzyme. Furthermore, PMs are often able toutilize alternative, but often less effective, path-ways of elimination.

    (iii) Many drug metabolizing enzymes are subject tovariant alleles which express enzymes withaltered substrate specicity or altered functionalactivity.

    (iv) Not all ADRs of a drug are geneticallydetermined even if a drug is metabolized by asingle pathway. Whereas only one or two of these may have genetic basis, others that areoften responsible for limiting the treatment haveno obvious genetic basis. Small systematicstudies suggest that often, not even the seriousADRs are associated with a specic genotype(Clark et al . 2004 ).

    (v) In addition to drug metabolizing enzymes,

    P-glycoprotein and associated organic iontransporters also inuence the disposition of many drugs. These play an important role in theabsorption of drugs and their transport into thecells and elimination into the bile or urine. Theactivities of these P-glycoprotein and transpor-ters are also polymorphically expressed andgenetically determined. Although their pharma-cokinetic effects may be moderate, they never-theless distort associations with othergenotypes.

    (vi) Not all toxic effects need have a pharmacoki-netic basis.

    Available data illustrate an important point from theregulatory perspective. Promises of potential clinicalbenets from integrating pharmacogenetics in clinicalmedicine are often based on inappropriate presump-tions on the role of polymorphic drug metabolizingenzymes or pharmacological targets. As discussedbelow, a number of drugs illustrate this point andoften, the results from pharmacogenetic studies of specic drug-induced toxic effects are inconsistent orcontradictory.

    Thioridazine has been shown in healthy volunteersto have a dose-related effect on ventricular repolariza-

    tion, primarily due to the parent drug with a possiblecontribution from the metabolites ( Hartigan-Go et al .1996 ). One recent study reported that CYP2D6 statusmight be an important determinant of the risk forthioridazine-induced QTc interval prolongation

    (Llerena et al . 2002 ) while another reported thatCYP2D6 genotype does not substantially affect therisk of thioridazine-induced QTc interval prolongation(Thanaccody et al . 2003 ). This discrepancy is notaltogether too surprising since the metabolite probablycontributes signicantly, but variably between individ-uals, to this toxic effect.

    The selective norepinephrine reuptake inhibitoratomoxetine is metabolized by CYP2D6. The peakplasma concentration ( C max ) of and the systemicexposure (area under plasma concentration versustime curve, AUC) to the pharmacologically activeparent drug are ve and 10 fold, respectively, higher inPMs compared to the EMs. However, when the safetyprole of this drug is scrutinized in terms of CYP2D6genotype (e.g. percentage of patients of each genotypediscontinuing therapy because of a side effect), it isquestionable if pre-treatment genotyping of patients iscost-effective or of any value clinically given the positiveand negative predictive values of the association.Atomoxetine is metabolized by dual pathways predominantly by CYP2D6 to pharmacologicallyequipotent 4-hydroxy-atomoxetine and to a lesserextent by CYP2C19 to almost inactive N -desmethyl-atomoxetine. Relative to atomoxetine, the plasma con-centrationsof 4-hydroxy-atomoxetine and N -desmethyl-atomoxetine are about 1 and 5%, respectively, in EMsand 0.1 and 45%, respectively, in PMs. In prospectiveclinical trials, many neuropsychiatric adverse eventswere generally only twice as frequent in PMs comparedwith the EMs and 5% of the EMs and 7% of the PMsdiscontinued treatment as a result. In terms of absolutenumbers, these translate into about seven EMs and one

    PM for every 100 un-genotyped patients discontinuingtreatment ( FDA 2002 ). To further add to thisdisappointment, the current safety concerns regardingatomoxetine are focussed on its hepatotoxic potential, asafety signal not evident during clinical trials.

    Two classes of drugs, antidepressants and neurolep-tics, have narrow therapeutic index and are generallymetabolized predominantly by CYP2D6. Althoughsmall retrospective studies appear to show a correlationbetween genotype and toxicity or failure to respond(Rau et al . 2004 ), overall analyses of studies correlatingCYP2D6 genotype with response to these drugs(referred to as phenotype) have been cautious in their

    conclusions ( Dahl 2002 ; Kirchheiner et al . 2003 a ,2004 b,c). The author of this paper analysed 17 studiespublished between 1995 and 2000, which had includedover 1350 patients receiving a range of neurolepticdrugs. These studies investigated associations betweenCYP2D6 genotype and drug levels, failure to respondbenecially, and frequency and severity of a number of ADRs such as neuroleptic malignant syndrome,extrapyramidal symptoms (EPS) and tardive dyskinesia(TD). Relationship with plasma concentrations wasshown for drugs with dominant CYP2D6-mediatedmetabolism but large intra-genotypic variability tendedto obscure its clinical value. However, there was no

    relationship evident between genotype and failure torespond benecially. There was only a general modesttrend observed towards a positive correlation betweenthe genotype, especially the presence of CYP2D6 * 10allele in the Japanese, and severity of TD and EPS.

    Pharmacogenetics and drug regulation R. R. Shah 1625

    Phil. Trans. R. Soc. B (2005)

  • 8/12/2019 Rst b 20051693

    10/22

    These disappointing ndings are hardly surprisingsince a number of these drugs are also metabolized bypathways other than those mediated by CYP2D6 andfrequently, these drugs have metabolites that arepharmacologically active in terms of efcacy andADRs.

    Non-steroidal anti-inammatory drugs (NSAIDs)are used widely and are responsible for treatmentlimiting gastro-intestinal side effects or hepatotoxicity.Indeed, hepatotoxicity is the most frequent cause of removal of NSAIDs from the market. Although mostNSAIDs are metabolized by CYP2C9, no associationbetween CYP2C9 genotype has been shown for eitherthe gastro-intestinal complications of NSAIDs gener-ally (Martin et al . 2001 ) or the hepatic complications of diclofenac ( Aithal et al . 2000 ). A more recent study hasalso found no evidence of impaired metabolism of oraldiclofenac in heterozygous and homozygous carriers of the CYP2C9 alleles * 2 and * 3 compared with thewild type allele and marked diclofenac-mediatedinhibition of COX-1 and COX-2 activity was detectedin all individuals independent of CYP2C9 genotype(Kirchheiner et al . 2003 c). The current consensus isthat CYP2C9 genotyping is unlikely to become routineclinical practice unless its value can be demonstratedin rigorous prospective studies ( Kirchheiner &Brockmoller 2005 ).

    Azathioprine and 6-mercaptopurine are metab-olized by the polymorphic TPMT. The activity of TPMT is inversely related to the risk of developingacute leucopenia. A number of studies have shown thatthe risk of azathioprine-induced acute leucopenia canbe greatly reduced by selecting the initial azathioprine

    dose based on TPMT genotype or phenotype(Colombel et al . 2000 ; Regueiro & Mardini 2002 ).However, an analysis of six clinical studies correlatingthe adverse effects of these drugs with TPMT genotyperevealed that an average of 78% of ADRs were notassociated with TPMT polymorphism. Pharmacoge-netic testing will thus not eliminate the need for carefulclinical monitoring of ADRs ( Schwab et al . 2002 ;Gearry et al . 2003 ; van Aken et al . 2003 ). Of course,this is not to suggest that it is not worth screeningpatients for genetic variants that may explain otherwell-established variations in drug responses but theabove analysis does illustrate the limitations of

    pharmacogenetics in routine clinical practice.Similarly, CYP2D6 polymorphism has proveddisappointing in predicting response to antihyperten-sive drugs ( Kirchheiner et al . 2004 a ; Schwartz &Turner 2004 ). Drug interactions too are similarlyrendered irrelevant if metabolites are pharmacologi-cally active. For example, venlafaxine (a dual mechan-ism-based antidepressant) is metabolized by CYP2D6but the labelling of this drug notes that in a clinicalstudy involving CYP2D6 PMs and EMs, the totalconcentration of active compounds (venlafaxine plusits major metabolite, O -desmethylvenlafaxine), wassimilar in the two genotypes. Therefore, no dosage

    adjustment is required when venlafaxine is coadminis-tered with a CYP2D6 inhibitor.Retrospective studies have also suggested that PMs

    of CYP2C9 are more susceptible to toxicity fromwarfarin or phenytoinboth drugs with a steep

    concentrationresponse curveand that there arewide inter-genotypic differences in dose requirements(Kidd et al . 2001 ; van der Weide et al . 2001 ; Higashiet al . 2002 ). Interestingly, however, although CYP2C9is intricately involved in the elimination of pharmaco-logically active ( S )-isomer of warfarin, its role in long-term safety of this widely used anticoagulant has yet tobe shown conclusively ( Takahashi et al . 2003 ; Kamaliet al . 2004 ; Takahashi et al . 2004 ). Available dataindicate that although CYP2C9 * 3/CYP2C9 * 3 geno-type is associated with dramatic over anticoagulationsoon after the introduction of this anticoagulant,overdose during the maintenance period is mostlyrelated to environmental factors ( Verstuyft et al . 2003 ;Peyvandi et al . 2004 ) which greatly inuence inter-individual variability in warfarin sensitivity. In onestudy, age and CYP2C9 genotype accounted for 12and 10% of the variation in warfarin dose require-ments, respectively ( Loebstein et al . 2001 ). Clearly,other factors such as variations in the activity of vitaminK epoxide reductase (VKOR) and diet also play animportant role. VKOR is the target of warfarin andthere are reports of familial occurrence of defects in aprotein of the VKOR-multienzyme-complex ( Olden-burg et al . 2000 ). Genetic control in the activity of VKOR has been described ( Rost et al . 2004 ) and Liet al . (2004) have recently identied the gene forVKOR. Polymorphisms of this VKOR gene may turnout to have a much greater effect on the response towarfarin than its CYP2C9-mediated metabolism.Similarly, despite the role of CYP2C19 in activatingthe antimalarial pro-drug proguanil to its therapeuti-cally potent metabolite (cycloproguanil), clinical obser-

    vations do not support the anticipated notion that thedrug will be ineffective in PMs of CYP2C19 ( Kanekoet al . 1999 a ).

    One of the uncertain aspects of pharmokineticvariability is the extent to which the disposition of many drugs is inuenced not only by the drugmetabolizing enzymes but also by P-glycoprotein.These are part of a larger family of efux transporters.They appear to have developed as a mechanism toprotect the body from harmful substances. P-glyco-protein is a 170 kDa membrane-bound protein thatfunctions as a membrane-localized drug transportmechanism with an ability to actively pump out a

    number of drugs. These transporters have beenidentied at a number of interfaces that the drugsmight cross, for example the intestinal wall, choroidplexus, gonads, placenta, renal tubules and biliarycanaliculi. Using ATP as an energy source, theytransport certain hydrophobic substances into thegut, bile or urine and out of the brain, gonads orother vital organs. Thus, they decrease oral bioavail-ability of drugs but once absorbed, decrease transfer of drugs across the bloodbrain barrier and reduce theirconcentrations in the central nervous system or acrossthe placenta to the foetus. In liver and kidney, thesetransporters are actively involved in secretion of drugs

    into the bile and urine, respectively. The expression of P-glycoprotein activity is under the control of MDR1(also known as ABCB1) gene ( Hoffmeyer et al . 2000 )and is an important factor in the disposition of manydrugs. The processes involved show considerable

    1626 R. R. Shah Pharmacogenetics and drug regulation

    Phil. Trans. R. Soc. B (2005)

  • 8/12/2019 Rst b 20051693

    11/22

    interindividual variability that is genetically deter-mined. Allelic variation in MDR1 gene is morecommon than had been previously recognized andinvolves multiple SNPs whose allelic frequencies varybetween populations, and some of these SNPs areassociated with altered P-glycoprotein function.Mutations at positions 2677 and 3435 are associatedwith alteration of P-glycoprotein expression and/orfunction. Two synonymous SNPs (C1236T in exon 12and C3435T in exon 26) and a non-synonymous SNP(G2677T, Ala893Ser in exon 21) are frequently linkedand the allele with this haplotype is styled MDR1 * 2.The AUC of fexofenadine was found to be almost 40%greater in individuals with * 1/ * 1 genotype compared tothose with * 2/ * 2 genotype. Those with the * 1/ * 2heterozygous genotype had an intermediate value.This suggests an enhanced in vivo P-glycoproteinactivity among subjects with the MDR1 * 2 allele ( Kimet al . 2001 ). This and related variant alleles also exertan inuence on the bioavailability and disposition of other drugs, although some data are contradictory(Hoffmeyer et al . 2000 ; Parker et al . 2003 ; Eichelbaumet al . 2004 ; Ieiri et al . 2004 ; Marzolini et al . 2004 a ).Chowbay et al . (2003) were able to show clinicallyrelevant substantial differences in the pharmacokineticsof cyclosporin in relation to MDR1 haplotypes (SNPsin exon 12, 21, and 26) in stable heart transplantpatients. From the overall evidence available to date, itappears that MDR1 polymorphisms have only amoderate impact on pharmacokinetics and pharmaco-dynamics of P-glycoprotein substrates ( Eichelbaumet al . 2004 ) and it remains unclear what polymorphismmay be responsible for which, if any, effects ( Soranzoet al . 2004 ).

    Organ-specic organic anion and cation transpor-ters are now recognized to play an important role in thetransport of some drugs into the cells and theirelimination into the bile or urine ( Kim 2004 ; Lee &Kim 2004 ; Marzolini et al . 2004 b; Mikkaichi et al .2004 ). Molecular studies have found evidence of genetic polymorphisms of these transporters in hep-atocytes ( Zhang et al . 1997 ; Tirona et al . 2001 ; Tirona& Kim 2002 ; Kim 2004 ). Mutations in the genescoding for these transporters may lead to dysfunctionalpolypeptides, which not only affect the pharmacoki-netics of the drugs concerned but may also intensify thepotential of some of these drugs to induce hepatotoxi-city ( Murata et al . 1998 ; Fouassier et al . 2002 ).

    9. RELATIVE IMPORTANCE OFPHARMACOKINETIC ANDPHARMACODYNAMIC POLYMORPHISMS

    It is now becoming increasingly evident that poly-morphisms of pharmacological targets (pharmacody-namic polymorphisms) may in fact be more importantand clinically relevant than polymorphisms of drugdisposition (pharmacokinetic polymorphisms). In apharmacogenetic study that compared paroxetine and

    mirtazapine in 246 elderly patients with majordepression, discontinuations due to paroxetine-induced side effects were strongly associated with the5-HTR 2A C/C, rather than CYP2D6, genotype. Therewas a signicant linear relationship between the

    number of C alleles and the probability of treatmentdiscontinuation. The severity of side effect in parox-etine-treated patients with the C/C genotype was alsogreater ( Murphy et al . 2003 ). Thus, although parox-etine is metabolized by CYP2D6, polymorphism of 5-HTR 2A appears to be a more important determinantof paroxetine-induced ADRs. In another study of 270cancer patients given anti-emetic therapy with5-HTR 3B receptor antagonists, approximately 30%suffered from nausea or vomiting despite these drugs.Ultrarapid metabolism of tropisetron (and to a lesserextent for ondansetron) was shown to predisposepatients to poor efcacy ( Kaiser et al . 2002 ). However,another study by the same group of investigatorsreported that patients homozygous for a deletionvariant of the promotor region of 5-HTR 3B experi-enced vomiting more frequently than did all the otherpatients. In terms of numbers needed to screen toidentify each case of vomiting, 5-HTR 3B polymorph-ism appeared to be complementary to CYP2D6polymorphism (30 for the combination versus 50 forCYP2D6 alone) ( Tremblay et al . 2003 ).

    The potentially greater importance of VKOR polymorphism relative to CYP2C9 polymorphism indetermining anticoagulant response to warfarin hasalready been referred to earlier.

    Hypolipidaemic responses to HMG-CoA reductaseinhibitors vary widely in the population. The systemicexposure to both enantiomers of uvastatin, an HMG-CoA reductase inhibitor, depends on CYP2C9 geno-type, with a three fold group mean difference in theactive enantiomer, and even greater difference in theinactive enantiomer, between the EMs and the PMs of

    CYP2C9. However, these differences in plasmaconcentrations of uvastatin were not reected in theeffect on cholesterol levels after 14 days of uvastatinintake in healthy volunteers ( Kirchheiner et al . 2003 b).Although several human cytochrome P450 enzymesmetabolize uvastatin, the pharmacokinetic differencesare not too surprising because CYP2C9 contributes5080%. Lack of a difference in pharmacodynamicresponses is, however, surprising. In contrast to theapparent clinical irrelevance of CYP2C9 polymorph-ism in uvastatin-induced changes in serum cholesterollevels, Chasman et al . (2004) have identied two tightlylinked SNPs in the gene coding HMG-CoA reductase

    that are associated with reduced efcacy of pravastatintherapy. It is not unreasonable to expect that thispolymorphism will also inuence uvastatin-inducedchanges in serum cholesterol levels. Compared withindividuals homozygous for the major allele of one of the SNPs, individuals with a single copy of the minorallele had 22 and 19% smaller reductions in totalcholesterol and LDL-cholesterol, respectively. In retro-spect, it is tempting to think that poor efcacy in a fewindividuals with anomalous target genotype may be thestimulus to driving upwards the recommended doses of HMG-CoA reductase inhibitors in the post-marketingperiod. For example, the recommended dose of

    cerivastatin was progressively increased from a maxi-mum of 0.3 mg daily at the time of its approval in 1997to a maximum of 0.8 mg daily at the time of itswithdrawal from the market in August 2001. However,it is uncertain whether the individuals with anomalous

    Pharmacogenetics and drug regulation R. R. Shah 1627

    Phil. Trans. R. Soc. B (2005)

  • 8/12/2019 Rst b 20051693

    12/22

    target genotype respond any better to higher doses.Undoubtedly, this must adversely affect the risk/benetof the drug in the wider population.

    Not surprisingly, the focus of current interest inpharmacogenetics of drug response has switched fromdrug metabolizing enzymes to the candidate genes of the pharmacological targets. Investigation of pharma-codynamic polymorphisms is now an active area of research.

    From a regulatory or clinical perspective, it is worthemphasizing that the consequences of polymorphismsin pharmacokinetics may be manageable by adjustmentof the dose to achieve the required therapeuticconcentrations (provided the toxicity from parentdrug or the metabolites does not supervene). However,the consequences of polymorphisms in pharmacologi-cal targets are far less likely to be easily managed.Indeed, these polymorphisms may have the effect of subdividing the target population or the disease intodiscrete subgroupseach requiring drugs that act atdifferent pharmacological targets. This has indeed beenthe experience with long QT syndrome which was atone time thought to be a single disease but now isrecognized to be a heterogeneous group of syndromeswith each syndrome having its own natural history andantiarrhythmic response to b-adrenoceptor blockingtherapy.

    10. PHARMACOGENETICS AND FUTURE DRUGDEVELOPMENT

    It is evident from the above discussion that forpharmacogenetics to be a truly valuable drug develop-

    ing and prescribing tool, pharmacogenetic approachesduring drug development will have to be moreholistic. These approaches will have to focus oninvestigating genetic inuences not only at pharmaco-kinetic (drug metabolizing enzymes and transporters)but also at pharmacodynamic (pharmacological tar-gets) levels to fully characterize the pharmacology of drugs. Clearly, a patients overall genotype thatdetermines a drug response (phenotype) must takeinto account the presence of normal wild type andmutant alleles in heterozygous and homozygous statesat both these key components of the doseresponserelationship. The situation becomes even more

    complex when one also considers the presence of multiple alleles at a single locus.There are considerable commercial, regulatory and

    clinical implications for the observation that a particu-lar genetic trait (single candidate gene or a SNP prole)confers susceptibility to toxicity or poor therapeuticresponse. For example, consider the fate of an NCEthat is less effective than an approved drug in the widerun-genotyped population but is found to have asuperior efcacy or safety prole in a genotypic subsetof the target population. It is therefore not surprisingthat, following the completion of the Human GenomeProject, immense efforts are under way at characteriz-

    ing normal nucleotide sequences as well as abbre-viated proles of nucleotide polymorphism(s)associated with diseases and with therapeutic responsesto drugs. The outcomes from these efforts areanticipated to provide a better and much greater

    understanding than has hitherto been possible of genetic factors underlying disease processes, develop-ment of new drug targets and biomarkers andresponses to drugs. Expectations have been raised,now higher than ever, that the goal of individuallytargeted therapy can be achieved.

    (a ) Establishing genotype/phenotype associationsTwo approaches have been usedcandidate geneassociation studies or genome-wide scans looking forSNP or haplotype proles associated with drugresponse. Despite the alleged limitations of candidategene approach, almost all the pharmacogenetic studiesto date have focused on this approach and mostsuccesses so far have resulted from it. Candidate geneassociation is mechanistic and relatively less resourceintensive but allows only a few genes to be studied. Incontrast, establishing a SNP prole from genome-scanis empirical, requires no knowledge of pharmacology of the drug and allows a much wider search. Therefore,

    although genome-wide association studies arecomplex, they may probably be the only way to bettercharacterize the genotype/phenotype relationships.However, it may be too optimistic to believe that allrelevant pharmacogenetic variations will be SNPs,especially as we already know examples of largedeletions, amplications and re-arrangements ( Idleet al . 2000 ). Genome-wide scans for SNPs may alsobe limited in their applications since it is known thateven at a single candidate gene locus, there isconsiderable allelic heterogeneity and there are wideinter-ethnic variations in the frequency of various SNPs(Cargill et al . 1999 ; Goddard et al . 2000 ; Stephens et al .2001 ; Ng et al . 2004 ). The prevailing SNP fever mayhave to be further tempered in the knowledge that thereare virtually no examples where a single DNA variantsite (genotype) can always be associated with aparticular trait (phenotype) in all subjects within allhuman populations ( Nebert 2000 ). Nebert et al .(2003) have summarized some of the major problems.

    Both within and outside the industry, the efforts atfullling the expectations of individually targetedtherapy are based on two approaches to clinical trials.

    ( b ) Classical clinical trials integrating

    pharmacogeneticsIt appears likely that it may soon become possible toroutinely and rapidly genotype individuals for a varietyof genetic traits at a very nominal cost. In order toexplore the role of pharmacogenetics in drug response,sponsors are now including a genetic extension to theusual clinical trial protocols, enabling them to collectand store blood samples for genetic analysis. Althoughthe patients would be required to give informedconsent for the main study protocol, the consent togenetic extension will be optional without prejudicingtheir chance of enrolment in the main study.

    The blood samples of those patients who consent

    would be analysed at a later date for an analysis of genotype/phenotype relationship to establish either acandidate gene or a SNP prole (from genome-widescan) associated with toxicity or failure to respond.This approach has the advantage of permitting analysis

    1628 R. R. Shah Pharmacogenetics and drug regulation

    Phil. Trans. R. Soc. B (2005)

  • 8/12/2019 Rst b 20051693

    13/22

    of the samples for specic genotypes related to drugdisposition as well as pharmacological targets.

    (c ) Enrichment design clinical trialsEnrichment design studies are another approachfavoured by some sponsors. This approach uses pre-enrolment genotyping to exclude (at present focusingon certain genotypes of drug metabolizing enzymes)from randomization those individuals who are unlikelyto benet or are likely to develop ADRs ( Murphy 2000 ;Murphy et al . 2000 ). Not unexpectedly, this design isadvocated in order to generate more robust evidence of

    efcacy but with trials of shorter duration and fewerpatients. It is also claimed to increase subject safety andeliminate the need for monitoring drug plasmaconcentrations. The advocates of this design clearlypresume (but seemingly without prospective evidence)a causal role for the genetic trait in drug response.

    The current efcacy-orientated approach to clinicaltrials already results in insufcient characterization of the clinical safety of an NCE. When a pharmacogenetictrait is found to improve efcacy, it is inevitable thatclinical trial populations will be highly select and evensmaller. The concern then will be further erosion incharacterization of the safety of the drug. Regulatory

    authorities are likely to approach with great caution anyclinical development programme heavily dominated byenrichment design studies. The advantages anddisadvantages of this design are summarized in table 4 .

    Clinical development programmes will then berequired to undertake specic additional safety andefcacy studies in genotypes excluded from pivotalstudies. Since it is very rare for a drug to be withdrawnfrom the market for failure of efcacy, there exists amore compelling case for conducting very carefullymonitored studies in genotypes suspected to be at risk.Thus, prospective genotyping should be used to ensureinclusion of important patient subgroups. The perils of

    excluding important subgroups from clinical trials arealready evident in the withdrawal of numerous drugsthat produce TdP or hepatotoxicitythe two seriousADRs that are most frequent in female gender ( Makkaret al . 1993 ; Shah 1999 )

    Probably the most important concern when inte-grating pharmacogenetics in drug development is thefact that rare or delayed ADRs (which are usually theserious ones and responsible for drug withdrawals)are unlikely to be observed during clinical trials.Perhexiline-induced neuropathy, for example, was notevident during clinical trials, and during its marketingpatients with neuropathy had taken the drug for a meanof 20 months. Therefore, in order to truly harness thepotential benets, pharmacogenetic studies will have tocontinue well beyond the approval of a drug into itspost-marketing surveillance period.

    Data protection, privacy and sample destruction willbe essential components of the consent for thesepharmacogenetic protocols ( CIOMS 2005 ). Inresponse to these developments, the CHMP hasrecently adopted a Position paper on terminology inpharmacogenetics ( CPMP 2002 ). This position paperdescribes ve categories of coding of the blood samples(identied, single-coded, double-coded, anonymizedand anonymous) for maintaining patient privacy with-out compromising the scientic and regulatory objec-tives of the study.

    11. PHARMACOGENETICS AND FUTUREREGULATORY APPROACHES

    In future, substantial pharmacogenetic data will besubmitted with the dossiers of NCEs, raising issues thatwill be important for regulatory integration of thesedata in the overall drug evaluation and approvalprocesses ( CIOMS 2005 ). Whereas extensive pharma-cokinetic investigations during clinical developmentmay reveal the efcacy and safety implications of polymorphic pharmacokinetics of a drug, at least oneconcentration-controlled clinical trial may have to beconsidered in order to characterize variability inpharmacodynamics. Regulatory aspects most likely tobe inuenced are assessment of efcacy, dose sche-

    dules, ADRs and drug interactions in relation togenotype and communicating this assessment to theprescribing community. Sponsors will seek guidance onhow pharmacogenetic data ought to be presented andanalysed and may form part of the labelling. Regulators

    Table 4. Advantages and disadvantages of enrichment design pharmacogenetic studies.

    advantages1 reduction in number of dropouts from the study2 trials with smaller number of patients3 trials of shorter duration4 reduction in requirements for safety monitoring5 exploration of doses higher than otherwise possible

    6 elimination or reduction of inter-individual variabilitydisadvantages1 inadequate information regarding potential variability in the target population2 overestimation of the dose of a drug and its efcacy3 further erosion of short-term and long-term safety data4 distorted comparisons in active controlled trials5 arbitrary exclusion criteria since multiple enzymes frequently involved in drug metabolism6 disregard for the presence of multiple variant alleles at a given locus, which may have

    different substrate specicitywhich genotypes are candidates for exclusion?7 not possible to investigate safety and efcacy of even the lower doses in genotypes excluded8 as proposed and currently applied, this design overlooks the impor tance of pharmaco-

    dynamic polymorphisms and of haplotypes

    Pharmacogenetics and drug regulation R. R. Shah 1629

    Phil. Trans. R. Soc. B (2005)

  • 8/12/2019 Rst b 20051693

    14/22

    need to start addressing these issues and articulatespecic guidance.

    To this end, both the CHMP and the FDA havealready implemented measures for brieng meetingsor voluntary submission of genomic data, respectively(CPMP 2003 ; FDA 2003 ). The PMDA in Japanrecently issued a public consultation documentrequesting comments on their proposal for preparationof guideline for the use of pharmacogenomics inclinical trials. The outcome of this consultation hasresulted in March 2005 in a nal guidance note that isvery similar to the FDA guidance on voluntarysubmission of genomic data. Sponsors of NCEs whowish to explore pharmacogenetics during drugdevelopment are encouraged to collect these data andsubmit these exploratory and preliminary (probablynon-validated) data to expert regulatory groups fordiscussion without prejudice or any fears and concernsabout adverse impact from sharing these data on theirdevelopment programme (safe harbour). This con-cept of safe harbour implies that the exploratory datadisclosed to the regulatory authority will be beyond theregulatory reach for initiating any pre-emptiveregulatory action. These meetings are intended toprovide an informal forum for discussions between thesponsors and the regulators with a view to developingregulatory and scientic understanding and futurepolicies on safety and efcacy paradigms. If requested,the regulators would offer advice on improving ordesigning studies aimed at generating answers to thequestions that may emerge during the regulatoryassessment of the data submitted to support anapplication for a marketing authorization. Such bilat-

    eral discussions and examination of pharmacogeneticdata from a wide range of drug development pro-grammes should facilitate logical applications of pharmacogenetics to subsequent drug developmentand, ultimately, to clinical therapeutics.

    Regulatory agencies will want to see pharmacoge-netic data that are consistently reproducible andmore predictive of drug response on an individualdrug-by-drug basis before embracing pharmacoge-netics in evaluation, approval and labelling of drugs.Indeed, the entire dataset will need careful scrutinybefore pre-prescription genotyping can be advocatedfor any particular drug. In principle, evaluation of

    data in relation to genotype may appear to be arelatively straightforward procedure comprisingassessment of efcacy in clinical trial populationssub-grouped by genotype. In this context, genotypemay be regarded as just one more of the manyvariables that regulatory agencies frequently consider.Issues, however, will arise in terms of statisticalevaluation of efcacy. Obvious ones are (i) how willmultiple SNPs be weighted for inclusion in summarystatistics and (ii) variable degree of interactionsbetween various SNPs and between a SNP and anexternal factor.

    In order to maximize efcacy, sponsors of NCEs

    have traditionally frequently promoted higher thanoptimal doses ( Cohen 2001 ; Cross et al . 2002 ). Notsurprisingly, post-approval safety-motivated downwardchanges in dose schedules are frequent ( Cross et al .2002 ). One major risk is that unless the dose is carefully

    selected and matched with the genotype to achieveplasma concentrations within a carefully selectedtherapeutic window appropriate to each genotype, theadvantages from pharmacogenetic targeting may belost. In principle, recommending higher doses thanwarranted is equivalent to an otherwise optimal dose inthe presence of a metabolic inhibitor or geneticmutation. Therefore, arising from these interindividualdifferences in pharmacology, areas of regulatorysubmissions that are likely to attract close regulatoryscrutiny are the doseresponse studies and thestandard dose schedule that is usually recommended.Assessment of the posology of drugs with a continuousresponse variable depending on the genotype prolewill most likely pose a difcult challenge. When thevariability is a continuous parameter, specic doseschedules may be required for subsets of individuals.Therefore, when a genotype/phenotype association isshown for a drug metabolized by CYP2D6, as withperhexiline-induced neuropathy for example, poten-tially four dose schedules may be required for (i)ultrarapid, (ii) homozygous extensive, (iii) intermedi-ate extensive and (iv) poor metabolizer genotypes.

    Genotype-related information will need to becommunicated to prescribing physicians. This may berelatively easy for phrasing the indication, restrictingthe use of the drug to those genotype(s) showing themaximum benet and contraindicating its use in thosegenotypes excluded from the clinical trials or shown tobe susceptible to serious ADRs. It is also probable thatspecial monitoring requirements may have to berecommended for individuals of specic genotype.Drug interaction section would need to emphasize

    the risks in those with intermediate or poor metabolizergenotype. PMs have no functional enzyme to inhibit orinduce but they may require protection from unin-tended inhibition of alternative pathways of drugelimination. Pharmacogenetics, and its protagonists,promise to revolutionize therapeutics within the nextdecade. This promise is based on the ability to scan thegenome and the presumption of discovering anabbreviated prole of SNPs or haplotype associatedwith variations in drug responses. Therefore, stilllooking further in the future, the prescribing infor-mation may have to be phrased in terms of not only thedrug metabolizing enzymes or pharmacological targets

    but also in terms of SNPs or haplotypes.Equally important, however, is the fact thatregulatory evaluation of the drug concerned will needto run in parallel with the evaluation of the kit forgenotyping. These kits will require approval with theirown criteria for approval, especially specicity, sensi-tivity, positive predictive value and negative predictivevalue, and specic practical details to be included intheir product literature. Physicians will require readyaccess to rapid genotyping kits or facilities.

    In December 2004, the FDA cleared for marketingthe rst laboratory-based genotyping test system thatwill allow physicians to consider unique genetic

    information from patients in selecting medicationsand doses of medications. The new test is theAmpliChip CYP2D6 genotyping test. Approval forinclusion of CYP2C19 genotype testing followed in January 2005. The new test is the rst DNA

    1630 R. R. Shah Pharmacogenetics and drug regulation

    Phil. Trans. R. Soc. B (2005)

  • 8/12/2019 Rst b 20051693

    15/22

    microarray test to be cleared by a regulatoryauthority. It will screen a patient for 31 mutationsin the CYP2D6 gene and two mutations in theCYP2C19 gene and is expected initially to costbetween US $300 and 400. It is not intended to be astand-alone tool to determine optimum drug dosage,but along with clinical evaluation and other tools todetermine the best treatment options for patients.The AmpliChip CYP450 Test was launched inEurope in the fall of 2004. However, the physicianwill need guidance on genotype-related dose adjust-ments for a variety of drugs metabolized by theseisoforms as well as up to date information onpotential inhibitors and other substrates of CYP2D6and CYP2C19. An important issue in the use of thistest is who will provide this prescribing guidance andinformation. For the vast majority of drugs, there areneither the prospective data on the positive andnegative predictive values of genotypephenotypeassociations nor genotype-related optimal doses.

    12. PHARMACOGENETICS AND IMPACT ONFUTURE CLINICAL PRACTICE

    Whether the promise of pharmacogenetics is fullledremains to be seen ( Goldstein 2003 ; Nebert et al . 2003 ;Tucker 2004 ). In principle, genotype-based prescribingought to be more effective at improving response ratesand decreasing the socio-economic burdens of ADRs.However, there would have to be a considerable changein drug promotion and prescribing cultures.

    An important factor that is likely to limit thepotentially benecial application of pharmacogenetics

    is the interaction between the genotype and extrinsicfactors. Non-compliance by physicians with prescrib-ing information and patients alike is just one of thesefactors. The over-exalted benets of pharmacogeneticsusually ignore the contribution of these non-geneticfactors, which are the more frequent causes of variations in drug responses. A number of drugs havebeen withdrawn from the market not because of someabnormal pharmacogenetic trait in the patients butbecause of lack of attention to the prescribinginformation. For example, nine drugs have beenwithdrawn from the market over the last decadebecause of their potential to prolong the QT interval

    and/or induce TdP. For all these drugs, this unexpectedand undesirable response was related almost exclu-sively to non-genetic factors in the majority of thepatients who experienced this ADR. In only 38 of the341 cases of cisapride-induced ventricular tachyar-rhythmias was there an absence of any obvious riskfactor ( Wysowski e t al . 2001 ; Shah 2004 ). Evenphysiological states such as menstrual cycle canaugment the risk of drug-induced QT intervalprolongation.

    Drugdrug interactions are another major problemand have frequently resulted in withdrawal of drugsfrom the market, for example terfenadine, mibefradil,

    cerivastatin, cisapride and levacetylmethadol. Theinhibition of drug metabolizing enzymes by otherdrugs is an important point of intersection betweenpharmacogenetics and drug response. An individual of EM genotype can be readily converted into an

    individual of PM phenotype by concurrent adminis-tration of an inhibitor. For example, quinidine oruoxetine converts a CYP2D6 EM into a PM and anatural consequence of this iatrogenic phenocopyingis that many individuals pr