NEW DRUG DEVELOPMENT, DRUG RESPONSE, AND PRECISION MEDICINES
In its broadest sense, pharmacogenomics can be defined as the investigation of variations of DNA and RNA characteristics as related to drug response. The last decade has seen a large increase in the amount of genomics data generated and, with it, increased the expectations of how improved understanding of disease will lead to the development of more effective therapies and personalized medicines. Despite the regular reports of novel genes being identified in a range of disorders, by 2012 the much-heralded promise of the human genome project has only started to materialize. However, it is important to realize that this does not represent a failure of the science to deliver as there are multiple clear examples of the predictability and clinical utility of pharmacogenetics but is a reflection of the length of time it takes to develop new drugs and implement changes in healthcare. The 1980s and 1990s saw a boom time for the pharmaceuticals industry producing many highly effective new classes of drugs from statins to proton- pump inhibitors and quinolone antibiotics. All were novel therapeutic approaches offering significant benefit to individuals and society. The science that drove many of these advances was based on the greater understanding of biochemistry and pharmacology that emerged during the 1970s and early 1980s. This 10- to 15-year time-lag from gaining scientific knowledge to developing therapies is typical for the pharmaceutical industry, and reflects the complexity of drug discovery and the time required for preclinical and clinical testing to ensure safety and efficacy. This chapter will introduce the major concepts of drug discovery and development and give a broad overview of how genetics and genomics is used across the whole drug discovery and development pipeline, from pre-target identification to post- marketing surveillance to help discover and develop improved medicines. It will describe some of the examples of how pharmacogenetics has impacted the lives of patients.
INTEGRATION OF GENOMICS INTO THE DRUG DISCOVERY AND DEVELOPMENT
Pharmaceutical companies have historically focused their drug discovery and development programs on finding therapies for broad use in large disease populations, the “blockbuster business model.” A blockbuster drug is usually defined as one with peak annual sales of greater than $1 billion and is generally developed for long-term use to treat common complex chronic disorders in the general population. The strategy to identify and develop blockbuster drugs has been the response to the high costs of drug discovery and development. A survey of the drug development costs of 68 new compounds from 10 pharmaceutical companies estimated that the cost to develop a new drug in 2000 was $802 million (DiMasi et al., 2003). The high costs of developing drugs can be attributed to two main factors: the large size and duration of the clinical trials required to provide the data to show safety and efficacy of the compound, and the high rate of attrition of compounds in clinical development; fewer than 10% of compounds entering phase I clinical development reach the market, the majority failing in clinical development due to lack of efficacy in phase II. The lack of recent research and development (R&D) success in finding blockbuster drugs, combined with financial pressure due to patent expiry and downward pressure on pricing, has led to a shift in strategy for many companies in the biopharmaceutical industry. Companies are shifting towards the discovery and development of stratified medicines. A stratified medicine is one that is targeted at a subgroup of a traditionally classified disease, such as Herceptin for the treatment of Her2-overexpressing breast cancer. Stratified medicines offer a significant opportunity: to the industry, as they have an increased probability of success and the potential of smaller programs; to the regulators, as the benefit–risk profiles of these medications are greater than with unselected medications; to the payers, as they are more cost effective; and most importantly, to patients, as they are more effective and safer therapies. Genomics has a large role to play in the development of stratified medicines, as many of the tools used to stratify the patient populations are genomic, such as selective epidermal growth factor receptor (EGFR) mutation status and Gefitinib, K ras mutation status and Erbitux and Vectibix, Alk4 mutation status and Crizotinib.
Pharmacogenomics—the investigation of variations of DNA and RNA characteristics (germline or tumor) as related to drug response in individual patients or groups of patients—is one of a number of methods employed by the pharmaceutical industry to stratify patient populations.
A major cause of the attrition of drugs for lack of efficacy is the heterogeneity of the diseases we currently classify as single entities. Most would be better referred to as syndromes rather than single diseases. The disease classification currently used is based on phenotypical consequences of disease processes rather than on the underlying pathological mechanisms. This has led to the clustering of heterogeneous disease syndromes based on symptoms rather than based on molecular pathology. Genomics will be an important tool in reclassifying diseases into a new molecular taxonomy of human disease. Oncology is one therapeutic area where this is most advanced, as the scientific evidence base for tumor etiology is more advanced than in other areas. The majority of drug development programs in oncology are now stratifying patient populations based on molecular changes in the tumor. During the period from 2005 to 2012, over 5 stratified medicines in oncology were approved (Table 8.1).
Table 8.1 STRATIFIED MEDICINES IN ONCOLOGY
Most of the current drug development programs in oncology are using a stratified medicine approach linking the target to the dysregulated disease pathways on the tumors and only being used when the right pathway is driving tumorigenesis. It is widely expected that this approach will expand across other therapeutic areas as our understanding of disease biology improves.
The Drug Discovery and Development Process
The generation of an idea that a particular protein might be a suitable therapeutic target for the treatment of a disease sets in motion what is often depicted as a linear process known as the “drug discovery and development pipeline,” in which new medicines follow a set route from early discovery and preclinical stages through a set of clinical development processes to the marketplace (Figure 8.1). In reality, the process is generally far from linear, but for the purposes of describing the component parts, we will consider it a sequential process.
Figure 8.1 The drug discovery and development pipeline.
The ultimate aim of the drug discovery process is to find a chemical (e.g., small molecule) or biological reagent, such as an antibody, that has the potential to be a drug that can be moved into preclinical and then clinical testing. In order to start the process of identifying a potential drug, a biological assay testing interactions with the drug target must be developed.
This assay is often based on a cloned and expressed form of the drug target and will be converted into a format that will allow high-throughput testing, as millions of chemicals may need to be screened in the assay. The need to screen millions of chemicals means that it is usually only feasible to screen one protein variant of the target in the high-throughput screen. It is therefore vital to screen the “right” variant. In the situation where there may be more than one form of the protein that can be included in the screen, it is important to know that the most biologically relevant and/or the most common variant is being screened, and it may be necessary to screen the chemical matter against more than one form of the protein. This is not always the most common form of the protein—Verumafenib, a novel drug for the treatment of malignant melanoma, was identified by specifically screening against the V600E mutated form of the BRAF protein to ensure that it only blocked signaling of the pathogenic form.
The high-throughput screens generally identify several potential “hits,” which need to be tested in more rigorous biological assays to determine the type of interaction and the effects and then refined using medicinal chemistry. Promising “leads” are then developed by a series of minor chemical changes to the original lead, and the final candidate is chosen based on the selectivity and potency criteria required for the drug candidate as well as the physicochemical properties of the molecule to ensure druglike properties. This candidate is then taken forward into preclinical testing.
The final testing phase is usually based on in vivo testing of the compound in animal models that have been demonstrated to have some translatability to the target human disease, or in a range of ex vivo models of human tissue that recapitulate components of the disease. The predictability and translatability of these models to humans varies with different diseases and is the focus of biomedical research in many therapeutic areas.
Once a drug candidate has been made, it goes into a preclinical toxicology testing that includes in vitro screening tests to identify potential pharmacological effects at other receptors that could lead to adverse events, and genetic toxicology testing, which evaluates mutagenicity and clastogenicity. Only if these are satisfactory does animal testing begin. The animal testing is done in two mammalian species and is staged to ensure that as few animals as possible are used and that major problems are picked up early. Toxicology studies to evaluate long-term exposure, reproductive toxicological effects, juvenile toxicity, and carcinogenicity are generally only performed once the data have been obtained from shorter-term human studies that support safety and efficacy. To date, toxicology induced by new chemicals are identified and classified by standard phenotypical and histological changes. While this picks up the majority of potentially toxic effects, it can be insensitive to subtle changes and can identify species- specific effects that can be difficult to interpret. A greater understanding of the molecular changes following drug administration could identify more subtle effects and species-specific effects. Similarly, the applicability of animal models of a disease could be assessed by evaluating molecular changes rather than phenotypical similarities that can be misleading. Greater emphasis is now being placed on molecular and biomarker changes that result in organ damage, where they are available (e.g. nephrotoxicity biomarkers).
Adverse events can be due to unexpected consequences of the primary pharmacology or to unexpected interactions with off-target proteins. Understanding the mechanism of the toxicological effects is important, as this allows a more quantitative evaluation of the risk of the event happening in humans. Genomics can be used to identify interactions with off-target proteins as transcription changes induced in the organ damaged by the compound can point to the mechanism of the toxicity. This is often referred to as toxicogenomics. Multiple consortiums (e.g., the Predictive Safety Testing Consortium [PSTC] and the Safety in Science in Medicines Education & Training [SAFESCIMET]) are currently working to identify genomic biomarkers that are more sensitive than current histopathological scores, allowing early detection of toxicology and the demonstration of species-specific toxic effects. Similarly, where specific organ toxicity is expected due to the mechanism of action of the compound or known off- target effects, then transcription changes can offer a more sensitive assay to detect early organ damage.
Once the initial in vitro testing and acute animal toxicology studies (which generally take 14 days) have been performed, then it is possible to start testing the candidate in humans. The human studies have traditionally been split into four phases (phases I–IV), each with specific aims (Box 8.1).
Box 8.1 HUMAN STUDIES HAVE TRADITIONALLY BEEN SPLIT INTO FOUR PHASES (PHASES I–IV)
|• Phase I—Pharmacokinetic and safety profiles in healthy volunteers
• Phase II—Safety and efficacy in patients, and the establishment of the dose response
• Phase III—Safety and efficacy at the chosen dosage
• Phase IV—Post-approval studies to answer specific safety or efficacy questions and to support commercial strategies
The first time a novel compound (or biological therapy) is tested in humans, a broad range of doses is tested, starting at very low exposures to minimize any risks to the clinical trial participants. Although these initial studies have generally been performed on healthy volunteers, there is an increasing trend towards incorporating patients as early as possible. The dose is escalated over several weeks, starting at a point between 10-fold and 100-fold below the expected pharmacological exposure levels, and rising to a maximum tolerable level, or several-fold beyond the expected maximum clinical dose (whichever is reached sooner). The aim is to identify common adverse events and their relationship with plasma exposure as well as to establish the basic pharmacokinetic (PK) parameters of the therapeutic agent. As drug development continues, more studies are performed to understand the effects of multiple dosing, specific drug–drug interactions, and food effects. The aim of these studies is to provide a more comprehensive understanding of the pharmacokinetics of the drug and any significant causes of variability in the pharmacokinetic profiles. Collections of DNA samples for pharmacogenomic analysis in phase I clinical protocols allow the assessment of the impact of known genetic variations on drug metabolism and transport.
There is a growing trend of performing some of these very early studies in patients, and these are often referred to as “phase Ib studies.” The primary intent of these studies is still to establish safety and pharmacokinetic action of the compound, but the use of patients allows early indicators of target engagement and biomarkers of efficacy to provide evidence that the compound is modulating the proposed mechanism. Where it is possible to biopsy disease tissue in these studies, transcription analysis can provide some evidence that the target pathway is being modulated. This is generally restricted to some tumor types and dermatological conditions like psoriasis, where it is possible to obtain high-quality tissue samples.
Phase II is traditionally divided into phase IIa, where the aim is to demonstrate the safety and PK parameters in patients, and IIb, where the aim is to establish efficacy and delineate the dose–response curve. However, most companies now endeavor to generate some data in the phase IIa studies to provide evidence of efficacy and confidence to progress into the more expensive and larger phase IIb dose-ranging study. This is a critical time in the development process, as up to 75% of all drug candidates will fail in phase II. If preclinical data or data from translational medicine studies have identified a patient population more likely to respond to the mechanism (e.g., BRAF activating mutation positive melanoma tumors for MEK inhibitors), then the studies can be restricted to this patient population to increase the likelihood of seeing an efficacy signal. Even when there is no strong a priori hypothesis, samples should be collected in phase II studies for pharmacogenomic analysis, as they are useful for testing less-validated hypotheses on the impact of genetic variation with respect to drug response. These studies are limited to detecting genetic variants with large effects, as these studies comprise relatively small numbers of patients (50–100). Samples for these pharmacogenomic studies may be collected with specific consent for genotyping of named genes within the protocol, which can be correlated with clinical data collected in the trial. With the ever-reducing cost of whole-exome and even whole-genome sequencing, there is a growing trend for collecting samples with broader consent to include sequencing studies.
Phase III trials form the basis of the regulatory approval and are often termed pivotal or registration studies. They are large studies evaluating the safety and efficacy of the candidate at the clinical dose and in the population where the drug will ultimately be used. The cost of this phase of development is significantly more than that of the others, so failure at this point has a major impact on the company. The larger numbers of patients included in these studies provide more power for pharmacogenomic analysis. In addition, these samples also provide a useful resource for more disease-focused phenotype– genotype correlations, and samples can be collected with broad consent for genotyping that allows the investigation of many candidate genes.
The patient population studies in the phase III program form the basis of the population approved to use the drug once it is launched. Therefore, if a genetically defined patient population is used in these studies, then the drug will only be approved for use in that group of patients. Even if the drug will only be used in a pharmacogenomically defined population, it is often necessary to include at least one study where all patient groups are included, to ensure that there is not an unexpected benefit in the nonselected population and also to provide a safety database for that group should they be prescribed the drug once it is approved. The inclusion of a prospectively stratified “all- comers” strategy also allows a more robust evaluation of the positive and negative predictive value of the test and, importantly, enables researchers to differentiate between a predictive pharmacogenomics test, where the test identifies subjects who differentially respond to the drug, from a prognostic test, where the test differentiated subjects with a more severe prognosis from the disease, regardless of treatment paradigm.
Drug testing does not stop with regulatory approval, and phase IV studies are run after the drug has been approved. Sometimes there are clinical studies required by regulatory authorities as a post-approval commitment. These generally test a specific question about safety and efficacy or are used to generate data to support commercial strategies. Studies conducted after the regulatory approval of the drug are an excellent resource for the implementation of a pharmacogenomics strategy because of the availability of larger sample sets. The potential to collect genomic samples from thousands of individuals recruited into large phase IV clinical studies presents the opportunity to link genomic data to good-quality clinical data, biomarker data, and, in many cases, long-term follow-up. An area where post-market pharmacogenomic surveillance can have a great impact is in addressing safety issues, thanks to the very large studies.
The availability of large numbers of patients on active treatments not only provides the material to look for pharmacogenomic effects but is also a valuable resource for understanding the molecular basis for disease, which in turn feeds back into idea-generation in the early-discovery section of the pipeline.
The studies performed within drug development programs are still classified according to this system, but, increasingly, companies are looking to generate potential signals of efficacy data in the early phase I and IIa studies (sometimes called the learn phase) to provide confidence that the compound will work, before investing in the more expensive phase IIb and III studies (sometimes called the confirm phase).
APPLYING GENOMICS TO DRUG DISCOVERY
Choosing the Best Drug Targets
One key area where genetics has impacted the drug discovery and development process is target selection. Between 50% and 75% of compounds fail in development due to lack of efficacy, and this is in large part because the target, and hence the mechanism of action of the drug, are not linked to the pathogenesis of the disease to which they are directed. Taking the view that the more you know about a drug target early in the discovery process, the less likely it is to fail in development due to lack of confidence in rationale (CIR), many companies are now investing up front in understanding the molecular genetics of the complex diseases we treat, and using genetics to identify novel targets and prioritize target selection from candidate gene lists for drug development programs. The advances in DNA sequencing, bioinformatics, and genetic analysis are providing great opportunities to use human genetics to identify novel targets.
Before 1990, pharmaceutical companies had worked on approximately 500 potential drug targets, with around 100 of these mechanisms having produced marketed drugs (Hopkins and Groom, 2002). Initial analysis of the final draft of the human genome project suggested that the total number of targets druggable with small chemicals might increase to 5000 (Drews, 2000). However, not all of these targets will be relevant to disease; therefore, current estimates are that there are 600–1500 drug targets in the human genome (Hopkins and Groom, 2002). This expansion of potential targets in concert with the rising costs of drug development means that the choice of targets is increasingly important. This number increases further when biological approaches are included.
Given the length of time it takes to get from an idea, to a compound, to the market, there are still only a few prospective examples of marketed compounds where genomics has provided a new drug target or supported its initial CIR; thus, there are insufficient data to show that having genetic or genomic CIR from complex traits has significantly increased candidate survival in the drug-development pipeline. Human genetics is a simple and effective way of beginning to assess the molecular evidence and provide the CIR for establishing a drug development program for a particular target. It is possible to retrospectively identify positive genetic associations between drug target and incidence or severity of disease for drugs that are currently widely prescribed; for example, angiotensin-converting enzyme inhibitors and hypertension (Zee et al., 1992; Province et al., 2003), β-agonists and asthma (Turki et al., 1995; Santillan et al., 2003), and serotonin reuptake inhibitors and depression (Ogilvie et al., 1996; Golimbet et al., 2004). Although this is not always the case, as the proton-pump inhibitors, used to treat gastroesophageal reflux disease (GERD), are one of the most frequently prescribed classes of drugs worldwide, but currently very little is known about the molecular genetics of GERD, and no reported association between the genes encoding the α and β subunits of the drug target hydrogen/potassium adenosine triphosphatase (ATPase) and the disease (Post et al., 2005). Knockout mouse data also provide evidence relevant to the function of target on the phenotype (Zambrowicz and Sands, 2003). The CIR for the statins, one of the most successful drug classes to be developed for the lowering of low-density lipoprotein (LDL) cholesterol, was derived from biochemistry. Interestingly the HMG-CoA reductase knockout mouse is lethal, and there are very few published genetic association studies on HMG- CoA reductase (Tong et al., 2004).
COMPLEX TRAIT GENETICS
The ability to carry out large-scale whole genome studies in well characterized populations extends the candidate gene approach, and has increased the potential to identify novel targets and new pathways that are relevant to disease. The challenge with these broad approaches is linking the findings back to our understanding of the disease process and using that knowledge to select a target. Linkage studies have had some success in identifying genetic variants associated with complex diseases; examples include phosphodiesterase 4D and stroke (Gretarsdottir et al., 2003), osmoprotectants taurine cyanate and nitrate (OTCN) cation transporter and DLG5 (discs large [Drosophila] homologue 5) genes with inflammatory bowel disease (Peltekova et al., 2004; Stoll et al., 2004), and 5-lipoxygenase- activating protein (FLAP) and myocardial infarction and stroke (Helgadottir et al., 2004). To date, these studies have provided some supporting evidence for the link between potential drug targets and disease, but only rarely are they the only evidence supporting this link. This is due to the fact that the reproducibility of early genetic association studies was poor, with many false positives reported; the identification of the causative variant is often challenging; hence making the prediction of whether the genetic variant is casing an increase or decrease in protein function can be a challenge. Three major advances have occurred in the last decade that have positively impacted the use of complex trait genetics. The first was the publication of the Wellcome Trust Case Control Consortium, which clearly demonstrated the need for larger sample sizes and rigorous quality control (QC) procedures (Frazer et al., 2004; John et al., 2004, The Wellcome Trust Case Control Consortium, 2007). The second advance has been the rapid development of DNA sequencing, which in 2013 was reaching a point where it is possible to sequence large cohorts of subjects, allowing the evaluation of rare variants as well as the common variants covered by the whole-genome association studies. The final advance is the development of bioinformatics and genetic analysis, which is allowing the combining of the genetic variations into pathway maps looking for dysregulated pathways rather than just individual SNPs. This is allowing the identification of optimal intervention points in pathways and the design of functional experiments that can confirm the direction of the dysregulation and hence whether an agonist or antagonist approach is required. There is therefore a renewed enthusiasm for the use of complex trait genetics to inform target choice, and the success of this will play out over the next five years.
SINGLE GENE DISORDERS AND TRAITS
Although the use of complex-trait genetics has yet to show real value, the use of rare genetic disorders has proven to be successful, albeit in a small number of cases. This approach of using the genetics of rare syndromes to identify drug targets with high confidence that pharmacological approaches will mimic the human phenotype has a growing precedence. The last five years have seen the first cohort of drugs to reach approval or late-stage clinical development where human genetics either identified the target or provided significant confidence in the approach. Examples of these drugs are included in Table 8.2, and include Maraviroc and chemokine receptor 5 (CCR5) (human immunodeficiency virus [HIV]), tofacitinib and the Janus kinases (JAK) (RA), romasozumab and sclerostin (postmenopausal osteoporosis), and vemurafenib and BRAF (melanoma).
Table 8.2 APPROVED DRUGS BASED ON GENOTYPE-PHENOTYPE CORRELATION
|Tofacitinib||JAK 3||Severe combined immunodeficiency|
|In development||Nav 1.7||Insensitivity to pain|
The identification of CCR5 as a potential therapeutic target for HIV infection came from the discovery that CCR5 was a coreceptor required for HIV infection, and from a genetic study of individuals who, despite multiple high-risk exposures, did not become infected with the virus. The genetic study demonstrated that individuals who were homozygous for this mutation (CCR5Δ32) and therefore had no functional CCR5 protein were apparently healthy and resistant to infection by HIV (Samson et al., 1996). Subsequent candidate gene studies have shown that heterozygosity for the CCR5Δ32 mutation is associated with slower progression to AIDS (Michael et al., 1997). Recent data have shown that a genetic polymorphism in the promoter of the CCR5 gene, resulting in increased CCR5 expression, is more common in individuals rapidly progressing to AIDS (Salkowitz et al., 2003). Thus, within seven years of the publication of genetic evidence that CCR5 would be a valid target in HIV therapy, clinical validation of this drug target was achieved with both Pfizer, Inc., and Schering-Plough publishing data showing significant viral load drops in patients with HIV infection treated with the potent CCR5 antagonists Maraviroc and Schering C, respectively (Feinberg, 2003).
The discovery of JAK and the identification of causative mutations in the JAK3 gene and severe combined immunodeficiency (SCID) highlighted the key role of this target in cytokine signaling and lymphocyte development and function, and provided CIR for the development of a selective JAK3 antagonist for the treatment of rejection in renal transplantation and rheumatoid arthritis. As with CCR5 above, the fact that individuals with the mutations only have the very specific effects of immunodeficiency and no other apparent deleterious phenotype means that these genetic data also provide confidence in safety (CIS) for the therapeutic approach (O’Shea et al., 2004).
Sclerosteosis is a rare genetic condition with only a small number of affected families in the world. A key aspect of the disease phenotype of sclerosteosis is bone overgrowth. This bone overgrowth is seen in the heterozygotes when they have generalized increase in bone density and mass, and the homozygotes when they have increased bone growth and density, which can lead to nerve-entrapment syndromes causing deafness and visual problems. The gene for sclerosteosis was identified in 2005, and the disease is caused by the absence of a protein called sclerostin. Sclerostin is a secreted protein that is highly amenable to a biologics approach, and reduction in circulating sclerostin will lead to increase in bone density. This led to a collaboration between UCB Celltech and Amgen to produce an antibody to sclerostin for the treatment of postmenopausal osteoporosis. This antibody has now been tested in phase IIb trials and has been shown to increase bone mineral density to a greater extent than do current therapies.
DRUGGABLE TARGETS APPROACH
An alternative strategy to the single-gene and whole-genome approaches is to carry out association studies in a subset of druggable target genes. Several companies have taken this approach to explore genetic associations with as many tractable targets as possible in a wide range of indications. Oxagen is a biopharmaceutical company specializing in understanding the genetic basis of common human diseases. One of the main areas of interest for the company is in G-protein coupled receptors (GPCRs); 20–30% of marketed drugs are targeted to the products of this class of genes. There are over 750 GPCR genes, thus Oxagen applied a filtering process to select the best targets for further analysis, based on expression profiling, known biology, whether they have a known drug targeted to them, or whether they are likely to be chemically tractable, before embarking on high-throughput genetic analysis (Allen and Carey, 2004). The Structural Genomics Consortium has focused on kinases (the Kinome). This consortium is funded by private and public sources and focuses on the identification of crystal structures of novel kinases and then the development of chemical tools. In concert with this, there has been considerable effort to identify kinases and their role in disease. Much of this has focused on the use of genetic mutations of kinases in cancer and genetic associations in conditions such as rheumatoid arthritis.
With the increasing use of genetics to drive target-identification in well- defined patient populations comes the dilemma of knowing which of all the targets identified is the best to take forward. The application of whole- genome technologies to understanding common complex disease has also led to new potential targets if they could be drugged. This increase in the number and type of targets will provide unprecedented opportunity to fight disease if we can choose the right targets and the right therapeutic approaches.
Effect of Genetic Variation on Compound Screening
Regardless of the original source of the target, genetic analyses are important in understanding how to move forward in the drug discovery process. Undertaking a comprehensive analysis of the genetic variation that exists in putative drug targets will provide information that could have a powerful impact on drug-discovery processes downstream. In an internal study within Pfizer, Inc., comparing coding SNP (cSNP) frequency, a selection of 111 genes encoding potential druggable targets and 160 genes considered as “non-druggable” targets found that 15% (26/111) of the putative targets were not polymorphic at the amino acid level, while 40% (45/111) had one or two cSNPs. There are also well-documented differences in the frequencies of specific polymorphisms between ethnic groups. Prior knowledge of any polymorphisms in a target can be incorporated into target validation, lead optimization, and inform preclinical projects supporting the development of the compound. The effect of genetic variation can be assessed through in vitro assays that incorporate a comparison of polymorphic targets by using either cells or biological reagents obtained from donors of known genotypes (where available), or by site-directed mutagenesis. This will facilitate early assessment of the potential impact of genetic variation on the activity of compounds and offer the potential to choose candidates that are the least likely to be influenced by the target polymorphism (Penny and McHale, 2005).
Gaining an early understanding of the impact of genetic variation can increase confidence in chemistry (CIC). For example, CCR5 has been shown to be the second coreceptor required for primary HIV infection. As such, it was a very attractive drug target for the treatment of HIV, as blockade of CCR5 should reduce HIV entry into cells and hence lower viral turnover. There have been multiple polymorphisms reported in the CCR5 gene, and some of these have been associated with effects on HIV infection rates and/or progression from infection to AIDS. A key question that had to be asked was, What were the functional effects of these polymorphisms, and would they would impact the effectiveness of the therapy? Preclinically, it was possible to demonstrate that the predominant effect of the functional polymorphisms was to alter receptor expression rather than structure; hence, the variability could be managed by identifying a dose that could effectively inhibit viral entry across a wide range of receptor expression levels.
The pharmacogenomic studies included in the preclinical phase of drug discovery that provide CIR and CIC and support nomination of a candidate for development are not intended to replace any of the clinical studies required for exploratory drug development or predict response in patient populations. The preclinical strategy will produce data to inform the pharmacogenomic plan for compounds in exploratory and full development. The challenge facing pharmacogenomics specialists in the pharmaceutical industry is to use the available genomic data to improve the efficiency of clinical trials.
APPLYING GENOMICS TO DRUG DEVELOPMENT
There are several definitions of pharmacogenetics in the literature, but the term was originally used in 1959 by Vogel to describe the inter-individual differences in drug response due to variations in DNA (Vogel, 1959). Although this is the origin of the term, the concept of inherited differences in biochemical attributes dates back much further, with Garrod describing the inheritance of alcaptonuria and phenylketonuria in 1902, and Snyder in 1932 describing the inherited ability to taste (or not) phenylthiocarbamide (Garrod, 1902; Snyder, 1932). The article by Motulsky in 1957 was the first serious attempt to understand the basis of inherited inter-individual response to drug therapies, with descriptions of the effects of glucose-6-phosphate dehydrogenase (G6PD) deficiency and primaquine in African-American soldiers (Motulsky, 1957). During World War II, scientists from the University of Chicago observed that approximately 10% of black American soldiers and (rarely) some of the white soldiers developed hemolytic anemia of varying severity when given conventional doses of a then-new antimalarial drug, primaquine. Further investigation revealed that this was due to the lack of the G6PD enzyme in red cells, which was the same genetic defect that had been shown to be responsible for the development of hemolytic anemia in susceptible individuals following the ingestion of fava beans. This was one of the first descriptions of a Mendelian (X-linked) pharmacogenetic trait. Also, in 1957, Kalow and Genest described an autosomal recessive pharmacogenetic trait (Kalow and Genest, 1957). Approximately 1 in 2000 subjects undergoing anesthesia develop a prolonged pharmacodynamic effect of succinyl choline due to a deficiency in the enzyme pseudocholinesterase. This autosomal recessive trait has since been recognized in a wide variety of ethnic populations, and although the enzyme deficiency was identified in 1957, it was a further 30 years before the causative genetic mutations responsible for these reactions were identified (McGuire et al., 1989).
Pharmacogenetics remained a relatively small field until the 1990s, due to the fact that although it was well recognized that all drugs exhibited significant inter-individual variability in response, the genetic tools to examine this variability were not available. Apart from a few standard approaches (e.g., renal impairment studies and gender differences), there was limited investigation of this phenomenon during drug development. The approach of the drug companies and regulators alike was to ensure that all compounds had a sufficiently good therapeutic index that the average benefit significantly outweighed the potential risk. This has led to the withdrawal or termination of development of a number of compounds with good efficacy but an insufficient population-based safety profile, which can often be driven by a small number of potentially serious adverse events. These events can be categorized into those that are expected based on an understanding of the pharmacological action of the drug (type A), and those that correlate with plasma exposure levels or idiosyncratic (type B) (Rawlins and Thompson, 1991). The mechanisms of idiosyncratic reactions are generally unknown and do not have a clear dose–response relationship.
Inter-individual variation in drug metabolism is now a well-documented phenomenon, but it was not until Mahgoub et al, (1977 Lancet 2:854– 856) described the polymorphic metabolism of debrisoquin that significant interest grew in the genetic contribution. The cytochrome P450 (CYP) enzyme family protects the body from xenobiotic agents and is the major route of metabolism of many drugs (Danielson, 2002). Several of these enzymes (e.g., cytochrome P450 2D6, 2C9, and 2C19) are known to have functional genetic polymorphisms that result in significant reductions or increases in function (Lee et al., 2002; Shimizu et al., 2003). Genetic variation in cytochrome P450 2D6 (CYP2D6) is well characterized, and approximately 10% of Caucasians make no CYP2D6 enzyme. Experiments with the antihypertensive agent debrisoquin yielded the first proven examples of a pharmacogenetic effect. Debrisoquin is metabolized by the CYP2D6 enzyme. An individual who makes no CYP2D6 and takes a standard dose of debrisoquin will suffer a profound hypotensive event resulting from high plasma exposure levels due to an inability to metabolize the drug (Idle et al., 1978). Approximately 20% of all drugs are metabolized by CYP2D6, and subjects who are unable to make this enzyme are at increased risk of developing adverse events when taking one of these compounds (Cascorbi, 2003) (Figure 8.2).
Figure 8.2 Individual variation in drug metabolism.
The incorporation of genetic testing for CYP2D6 or related enzymes in clinical trials has the potential to identify, prospectively, subjects who are likely to have adverse events due to poor metabolism, or those who may have limited response through inadequate exposure because of ultra-rapid metabolism.
Many drug-metabolizing enzymes have genetic variants leading to reduced or increased function, with consequent impact on the PK variability. Despite this knowledge, there are few drugs for which pharmacogenetic tests are routinely applied, and only recently has it become accepted best practice to test for the presence of variation in the gene encoding the thiopurine methyltransferase (TPMT) enzyme before prescription of azathioprin and 6- mercaptopurine.4,5 Approximately 1 in 300 individuals is homozygous for mutations in the gene encoding the TPMT (Evans, 2004). If treated with a standard dose of azathioprin (6-mercaptopurine), these individuals have a substantially increased risk of developing the potentially fatal complication of red cell aplasia (Evans, 2004). Suitable dose reduction decreases this risk. The recent decision by the Clinical Pharmacology division of the FDA to recommend that subjects be tested for TPMT enzyme status (either phenotypically or genotypically) before dosing with 6-mercaptopurine is evidence of the increasing awareness of the value of understanding inter- individual variation in drug metabolism. Similarly, the recently approved drug Strattera from Eli Lilly provides safety data for poor and extensive metabolizers of CYP2D6, and the availability of a suitable test to distinguish these two groups is also included in the label, although there is currently no recommendation about using the test and adjusting the dose according to genotype.
As the clinical value of these tests becomes established and is translated into practice, so will the acceptability of requiring a metabolizing enzyme diagnostic before dispensing a drug. Clear demonstration of the advantages of prospectively using a diagnostic test versus clinical management of drug dosing will be vital if these tests are to be used in clinical practice. This will also allow the development of chemicals with narrow therapeutic windows and predominantly metabolized by a polymorphic enzyme. Many of these compounds have historically been terminated, as the risk of adverse events due to high plasma exposures outweighed the potential benefit. A clinically acceptable way of managing this risk would make the safe use of these compounds possible.
The importance of being able to predict drug response is highlighted by the fact that it has been estimated that approximately 30% of prescriptions written do not benefit the patient, and even in highly controlled environments, such as clinical trials, it is rare to get response rates significantly above 70% (Silber, 2000). If we assume that subjects take the medication in the prescribed manner, then lack of efficacy may result from inadequate exposure to the drug (PK variability), an inability to respond to the therapy due to genetic variation in the target and/or downstream effectors (pharmacodynamic [PD] variability), or because the pharmacological intervention does not alter the underlying pathophysiological process (disease heterogeneity). While some commentators have suggested that differences in disease genetics (disease heterogeneity) should be considered as separate from pharmacogenetics, at a practical level, understanding this genetic variation will result in the same outcome—for example, understanding the increased or decreased likelihood of response to therapy. Therefore, this group will be included in the PD variability subgroup.
There are now multiple examples of the use of pharmacogenetics to predict drug response. The majority of these are in oncology, where tumor mutations have been shown to drive pharmacodynamic response in multiple areas. The best known examples of this are Herceptin and Gleevec. In the case of Herceptin, amplification of the Her2 gene leads to up regulated Her2 protein expression in approximately 25% of all breast cancers. These tumors are responsive to Herceptin, whilst tumors with lower levels of expression of Her2 do not respond. Imatinib is a treatment for Philadelphia chromosome positive chronic myeloid leukemia specifically designed to target the BCR- ABL fusion protein generated from this chromosomal translocation. It also is active in tumors with mutated KIT genes (e.g., GIST). Table 8.3 contains a list of anti-tumor therapies aimed at genotypically defined tumors.
Table 8.3 ANTI-TUMOR THERAPIES FOR GENOTYPICALLY DEFINED TUMORS
|Gefitinib||Non–small cell lung cancer||EGFR|
|Erlotinib||Non–small cell lung cancer||EGFR|
|Crizotinib||Non–small cell lung cancer||Alk4|
Table 8.4 EXAMPLES OF DRUG RESPONSE MODIFICATION ASSOCIATED WITH GENETIC POLYMORPHISMS IN “DISEASE-MODIFYING” OR “TREATMENT- MODIFYING” GENES
Adapted from Evans WE, McLeod HL. Pharmacogenomics—drug disposition, drug targets, and side effects. N Engl J Med. 2003 Feb 6;348(6):538–549.
Vemurafenib is a very exciting example, as this compound was screened using the common V600E mutation of the BRAF gene. This mutation is present in approximately 60% of melanoma tumors. A counter-screen of non- mutated BRAF was also run, ensuring the identified compound was specific for the mutated allele. This drug is highly effective in V600E-positive tumors and has a very good safety profile, as it does not bind to the non-mutated protein, hence only working within the tumor cells.
Although the majority of examples are in oncology, there are exemplars in other therapeutic areas as well. One of the clearest examples is in the treatment of hepatitis C. Subjects who have the AA polymorphism in their interferon gene have a greater chance of responding to interferon therapy than do individuals who are AT or TT. Other examples exist, particularly in the rare disease field, where therapies are directed at specific genetic disorders, and in this case, it is disease genetics rather than pharmacogenetics.
Despite the success stories described over the last few years, most therapies tested to date do not appear to have a clear pharmacogenetic signature. It may be that the current approaches are unable to identify the correct genetic variation or (more likely) the combination of variants that can predict response, or it may be that genetic variation is not a major cause of the heterogeneity of drug response.
Predicting Type B Adverse Events
The last few years have demonstrated that pharmacogenetics can be used to predict some rare adverse events. Extreme pharmacodynamics adverse responses to drugs have been described in the past, such as malignant hyperthermia and inhaled anesthetics, succinyl choline deficiency, and prolonged paralysis. More recently, an immunogenetic explanation for rare hypersensitivity reactions was discovered. Abacavir was a key drug in highlighting the role of HLA variation and drug hypersensitivity. Two retrospective studies have identified the HLA-B*5701 allele of the major histocompatibility complex (MHC) class I B gene as a genetic determinant of hypersensitivity to abacavir (Ziagen) (Hetherington et al., 2002; Mallal et al., 2002). The availability of a relatively large patient population led to the identification of the HLA-B*5701-Hsp70-Hom variant haplotype in 94.4% of cases compared to only 0.4% of controls. Analysis in different ethnic groups, however, showed that HLA-B*5701 alone would not be sufficiently predictive of hypersensitivity in diverse patient populations, suggesting that other genetic determinants of hypersensitivity remain to be identified. Additional HLA associations with adverse drug reactions have been described. Chung et al. in 2004 described an association between HLA B1502 and Stevens Johnson syndrome in the Han Chinese population. Again, this association appears to be confined to the Han Chinese.
Predicting Type A Adverse Events
Adverse drug reactions (ADRs) are a major cause of morbidity, leading to approximately 5% of all hospital admissions, and severe adverse drug reactions are a leading cause of death in young adults. Despite initial optimism, pharmacogenetics has had limited impact in reducing this morbidity and mortality. There is, however, evidence that genetic variation can influence our risk of developing type 1 adverse events by either increasing our exposure to the active agent or altering the pharmacodynamics effects of the drug. Warfarin is one of the best understood examples of how genetic variation can influence the risk of adverse events. Bleeding events on warfarin are among the commonest adverse events resulting in significant morbidity. Underlying genetic variation accounts for at least 50% of the risk of developing a bleeding event. This risk is predominantly driven by two key genes; the drug-metabolizing enzyme cytochrome P450 2C19, and the gene encoding the vitamin K receptor. Studies by Pirmohamad et al. have shown that poor metabolizer status of cytochrome P450 2C19 have a Y-fold increase in plasma exposure of S-warfarin (the active moiety). The increase in exposure results in a Z-fold increase in bleeding risk due to pharmacokinetic variability. The vitamin K receptor is the target for warfarin and is required for the production of vitamin K–dependent clotting factors. A common variant in this receptor results in a decrease in vitamin K receptor function. Whilst this normally causes no significant sequelae, it does affect response to warfarin. Individuals who are homozygous for the rare allele have an increase in bleeding risk of Y when taking warfarin. By combining the results of these genotypes it is possible to refine an individual’s risk of developing a bleeding adverse event if they are given a standard dose of warfarin. Prospective trials are now ongoing to determine the utility of using genotype results to adjust the starting dose of warfarin.
Individualized Therapy—An Integrated Response
In real life, the response of an individual is based on both the plasma exposure and how that affects the various physiological processes in the target organs. Evans and Relling generated a hypothetical graph representing the PK and PD variation in concert (Evans and Relling, 1999).
Variation in drug-metabolizing enzymes can dramatically impact plasma exposure levels (see left-hand column in Figure 8.3). However, it is not until we integrate this with variation in genes affecting PD response (in the right- hand column) that we start to get a real understanding of the impact on response for the individual. It is important to realize that dose-related adverse events are observed in extensive metabolizers as well as poor metabolizers, but the incidence is dependent upon the frequency of variation in the genes affecting PD response. As the frequency of variation in genes affecting PD response approaches 0.5, the predictive power of a test solely looking at drug metabolism decreases. Similarly, the predictive power of a test evaluating variation in genes impacting PD response will vary depending upon PK variability. Most pharmacogenetic studies that have been published to date have concentrated on single genes or small numbers of candidate genes, which are likely to affect either PK or PD variability. It is unsurprising that these studies fail to demonstrate high positive or negative predictive information for drug response, as it is generally due to a combination of both of these factors. As we move forward, a more holistic approach to the examination of genetic factors impacting drug response should lead to the identification of sets of SNPs with higher predictive values, leading to improved prescribing (Table 8.4).
Figure 8.3 Drug response due to pharmacokinetic (PK) and pharmacodynamic (PD) interactions. The impact of genetic variation leading to altered plasma exposures depends on the variation in the genes leading to the effector mechanisms of the drug.
Improving Disease Classification: Stratified Medicines
The need to accurately and precisely characterize the disease under investigation has important implications in drug development. The current disease classification system has changed little in the last two hundred years and is based on the phenotypical clustering of symptoms. That is, diseases that present with similar symptoms have been classified as the same condition. These diseases are therefore more like syndromes and do not necessarily reflect a common underlying pathology. Similarly, there may be conditions with similar pathological mechanisms that are currently classified as different diseases, as the phenotypical features are not similar enough. A very clear example of this is in oncology, where many mechanisms are represented in subsets of organ-classified tumors: for instance, EGFR mutations are present in multiple tumor types. The knowledge from the outset of a drug discovery program that there are molecular subtypes of a disease means that appropriate preclinical experiments can be developed early to predict the likelihood of a pharmacogenomic effect, and this information can be used advantageously in the drug development program. Combining genotype data with other genomic data provides valuable information about the disease subtype. Integration of genotyping data with gene expression, for example, has identified subtypes of obesity phenotypes in a mouse model (Schadt et al., 2005). Using similar approaches and including microRNA, epigenetic, proteomic, and metabonomic analyses in well-defined patient cohorts will provide powerful tools to aid the dissection of the phenotype of disease in humans in order to drive the development of targeted therapies based on molecular sub-classification of diseases (disease stratification). This reclassification of disease has become the focus of several cross- academic/industry consortiums, and the next decade could see the development of new disease taxonomies reflecting the true molecular mechanisms of the pathologies, rather than their consequences.
One therapeutic area where using genetic and genomic technologies has undoubtedly had a major and measurable impact on understanding the molecular subtypes of disease is oncology. The advances in understanding the molecular mechanisms predisposing to cancer have seen the number of oncology compounds in clinical development rise from 10 to over 400 in a 10-year period. The majority of the new compounds now being tested are classed as “targeted biotech medicines.” Imatinib mesylate (Gleevec) and trastuzumab (Herceptin) were the first two such targeted compounds approved. Herceptin is a therapy targeting the HER2/neu receptor in breast cancer. The rationale for this therapy was based on a sound understanding of the underlying molecular pathology. It was known that only 20–30% of breast tumors overexpress this protein, and it was demonstrated in the drug development program that response to Herceptin was limited to subjects whose tumors overexpressed the target (Vogel et al., 2002). Similarly, Gleevec is a therapy targeting the fusion protein product resulting from the Philadelphia chromosomal translocation observed in most cases of chronic myeloid leukemia (CML) (Deininger et al., 1997). This therapy provided dramatic efficacy in cases of CML with the chromosomal translocation, and it was rapidly approved by the Food and Drug Administration (FDA).
Following the rapid approval and success of Gleevec and Herceptin, many other targeted cancer therapies have entered clinical trials, thus highlighting the absolute requirement to continue to investigate and understand the underlying molecular mechanisms that are associated with disease. Gefitinib (Iressa) was the first in class selective EGFR inhibitor to receive accelerated approval based on preliminary data from phase II studies in non–small cell lung carcinoma (NSCLC) patients. Activating mutations and overexpression of EGFR were known to occur in many cancers, providing CIR for development of an EGFR-inhibitor for cancer treatment. Inactivation of the EFGR gene in mice did not cause any major phenotypical effects, which fact in turn provided CIS with respect to pharmacological inhibition of this target (Wong, 2003). However, initial tumor response to treatment in the clinical trials of subjects with non–small cell lung cancer was only observed in 9– 19% of patients. Subsequent analysis to predict factors that would indicate good response to Iressa identified that female gender, nonsmoking status, and a specific histological subtype of tumor were associated with better response to therapy. Investigation of biological and markers of response failed to show an association with EGFR expression levels. However, somatic mutations in the ATP-binding site of the tyrosine kinase domain of EGFR were observed more frequently in the tumors of patients who responded to Iressa. The EGFR mutations are located close to the putative binding site for compounds like Iressa and lead to increased signaling in the growth factor pathway; therefore, tumors harboring these mutations are more susceptible to treatment with an EGFR inhibitor (Lynch et al., 2004). This highlights the importance of defining the molecular subtypes of disease and understanding the impact on response to therapy. Had the molecular profile of NSCLC been identified before testing in humans, it may have been possible to design preclinical cell-based assays to determine whether the genetic profile of the tumor would influence response to therapy and then inform clinical trial design.
The majority of oncology programs now in development are focusing on stratified populations based on genetic or genomic classifications of tumor type.
PHARMACOGENOMICS AND MARKETED DRUGS
In a recent study of adverse drug reactions (ADRs), 5% of hospital admissions in the United Kingdom were identified as being due to ADRs. Over 70% were considered avoidable, and while drug interactions accounted for the majority of the ADRs, and older drugs were implicated in the hospital admission, there is still a need to understand the underlying causes of all ADRs (Pirmohamed et al., 2004). It is difficult to detect rare adverse events in the confines of a clinical trial, due to the relatively small number of subjects in the study, and the current system for monitoring ADRs has been suggested to be “too disparate.” A move to a more comprehensive epidemiological approach to monitoring drug safety has been proposed. The inclusion of pharmacogenomic analyses within this approach would allow the systematic assessment of the contribution of genetic determinants to ADRs. Pharmacogenomic surveillance in large phase IV trials of approved compounds has the potential to have a great impact in addressing safety issues.
One therapeutic area where detailed pharmacosurveillance, including pharmacogenomic analyses, post-approval, is not new, is in the antiretroviral treatment of HIV infection. Viral resistance and drug toxicity are common and often lead to treatment failure. HIV genetic sequences are determined, and the viral load is constantly monitored to assess viral resistance to highly active antiretroviral therapy (HAART). Polymorphisms in drug transporters and drug-metabolizing enzymes have also been monitored in HIV therapy. Two retrospective studies have identified the HLA-B*5701 allele of the major histocompatibility complex (MHC) class I B gene as a genetic determinant of hypersensitivity to abacavir (Ziagen) (Hetherington et al., 2002; Mallal et al., 2002). The availability of a relatively large patient population led to the identification of the HLA-B*5701-Hsp70-Hom variant haplotype in 94.4% of cases compared to only 0.4% of controls. Analysis in different ethnic groups, however, showed that HLA-B*5701 alone would not be sufficiently predictive of hypersensitivity in diverse patient populations, suggesting that other genetic determinants of hypersensitivity remain to be identified. Implementation of pharmacogenetics post-approval will have a role in increasing the CIS of new products.
PHARMACOGENOMICS IN CLINICAL PRACTICE
There are two clear areas where pharmacogenomics has impacted clinical practice and is now being widely used. The first is oncology, where genetic profiling of non–small cell lung cancer, colorectal cancer, and breast cancer is routinely used to drive treatment choice. This has been a relatively rapid change in practice with the approval of multiple targeted drugs (Table 8.3) and is likely to grow further as our increasing understanding of tumor biology is matched with new targeted therapies (e.g., Crizotinib and Cetuximab). The second area is in infection, where we have seen routine testing become established for the HIV therapy abacavir and the hepatitis C therapy interferon C. Both of these areas are used to complex prescribing, which has enabled the more rapid integration of testing into the treatment paradigm.
However, this success has not been seen in all areas. Despite extensive knowledge of the genetics of CYP2D6 and related enzymes, and their involvement in the metabolism of many commonly used drugs, drug- metabolizing PG has had little impact in the clinic. Multiple case-control studies have implicated the role of genetic variants in these enzyme systems and the risk of adverse events (Brockmoller et al., 2002; Rau et al., 2004; Steimer et al., 2005). These studies have investigated both specific compounds and adverse event rates in drugs metabolized by polymorphic enzymes compared with non-polymorphic pathways. The failure to implement genetic testing for these variants in the clinic and appropriate adjustments in dosing is due to a number of factors. The lack of appropriately designed prospective trials demonstrating the clinical benefits of this approach, and inconsistency of results in some of the retrospective case- control studies, are often cited as reasons for the lack of clinical usage. Additional factors include the need for rapid, easy testing and increased genetic education for many healthcare groups: such as physicians, nurses, and pharmacists.
The degree to which pharmacogenetics is incorporated into mainstream clinical practice depends not only on the science but also on the regulatory and societal environment. To date, there has been little impact in the clinic, but the available tests have had limited predictive value, and there have been few good prospective studies performed. As the science progresses, the regulatory and societal factors will become more important. The nations’ regulatory authorities are responsible for ensuring that all drugs licensed for use have an appropriate risk–benefit ratio. However, this ratio is an average based on efficacy across the total treated population and on the adverse event rate across the same population. Approval can be, and has been, refused for drugs that offer significant benefit but have serious adverse effects in a few subjects. The ability to detect the subjects at increased risk of these adverse events would allow these drugs to be used safely. The number of drugs withdrawn in the last 5 to 10 years reflects the increasingly risk-averse regulatory environment. The potential of preventing these withdrawals in the future by identifying at-risk subjects has stimulated significant interest from the regulatory authorities. It is likely that, in the future, the identification or confirmation of these adverse events following a drug’s launch will stimulate research into the precise mechanism of the event and strategies to identify subjects at risk, rather than an immediate withdrawal of the drug. While studies to understand the mechanisms of ADRs have always been attempted during drug development, pharmacogenetics offers the potential not only to understand why a reaction has occurred but also to identify who is at risk of it, before administration of the drug. The regulatory authorities have been a key driver of the use of pharmacogenetics to improve the safety profiles of drugs.
An improved efficacy profile for a compound is important in the context of gaining drug approval, but it can be vital when drug reimbursement and use are considered. The use of drugs is primarily driven by the physicians who prescribe them and the healthcare infrastructures that reimburse their costs, such as the National Health Service (NHS) in the United Kingdom or health maintenance organizations (HMOs) in the United States. In order for the use of newer drugs to be justified, there needs to be significant benefit over existing therapies, which may be generic and have proven safety profiles. It is possible to use pharmacogenetics to improve efficacy profiles by identifying subjects who are likely to respond well (or those likely to get minimal benefit) and targeting the therapies accordingly. This use of pharmacogenetics is driven by the payers for the therapies, as the increasing pressure on healthcare budgets means that paying for more expensive branded therapies can only be justified for patients likely to gain significant benefit.
While the role of the regulators and healthcare payers in driving forward the use of pharmacogenetics is already emerging, it is unclear what the role of the patient will be, although it is clear that this could be significant. The risks of taking a medication must always be placed in context with the potential benefits and not treated in isolation. It may be perfectly acceptable to license and use a drug with significant risks if the potential benefits are substantial, as in cancer therapies. Meanwhile, in other situations, very little risk of adverse events can be tolerated (e.g., erectile dysfunction). The indication being treated and the current available therapies are the key determinants of the level of adverse events that would be tolerable for the efficacy observed. As the science becomes more sophisticated and the prediction gets better, it will then be possible to provide more refined risk–benefit ratios for each individual. It is unclear how this range of risk benefits will be managed. Traditionally the regulatory authorities have, in conjunction with independent experts, determined what is an acceptable population-based risk–benefit ratio. This average risk–benefit ratio may soon become a range of risks and benefits, and the acceptability of an individual risk–benefit ratio will become a question for the patient and his or her physician rather than the regulators. As the patients’ role in drug selection becomes more central to the prescribing process, so will their influence on drug licensing and the use of pharmacogenetics.
Pharmacogenomics offers great promise to all stakeholders in the healthcare community. To industry, it offers the potential of improving the efficiency of drug development by reducing the current high failure rate through better choice of targets and improved understanding of drug response early in development. To the healthcare providers, it offers the potential to reduce the burden of adverse events by identifying the subjects at increased risk and offering them alternative therapies, as well as targeting their resources to use newer, more expensive treatments on subjects who will derive most benefit. Finally, and most importantly, it offers to the patient the opportunity, with their physician, to identify from the range of available therapeutic options the one most suited to them. While pharmacogenetic testing is unlikely to be able to guarantee that the therapy will work and will not cause an adverse event, it will increase the probability that a drug will work and reduce uncertainty about adverse events, and provide a rational way of choosing between therapies.
As our understanding of genomics improves, so will our ability to determine key factors involved in variability of drug response. The quest for precision medicines will start at the beginning of the drug discovery process, with more comprehensive understanding of the molecular basis of the disease, patient stratification, and the role of the drug target in the pathological process. Significant variability in PKs will be explained by systematic evaluation of all the relevant metabolizing enzymes and transport proteins. The drug candidates will only be tested in patients with suitable variants of the drug target. Drugs will be approved with variable dosage levels dependent upon underlying genotypes affecting PKs and variation at the drug target. Finally, pharmacogenetics will not stop with a drug’s approval: post-marketing research will endeavor to identify the causes of rarer adverse events, leading to continuous refinement of how we use drugs throughout their lifecycle.
1. Allen MJ, Carey AH (2004). Target identification and validation through genetics. Drug Discov Today. 3(5):183–191.
2. Brockmoller JJ, Kirchheiner, et al. (2002). The impact of the CYP2D6 polymorphism on haloperidol pharmacokinetics and on the outcome of haloperidol treatment. Clin Pharmacol Ther. 72(4):438– 452.
3. Cascorbi I (2003). Pharmacogenetics of cytochrome p4502D6: genetic background and clinical implication. Eur J Clin Invest. 33(Suppl 2):17–22.
4. Chung WH, Hung SI, Hong HS, et al. Medical genetics: a marker for Stevens–Johnson syndrome Nature. 2004;428:486. (PubMed)
5. Danielson PB (2002). The cytochrome P450 superfamily: biochemistry, evolution, and drug metabolism in humans. Curr Drug Metab. 3:561–597.
6. Deininger MW, Goldman JM, et al. (1997). The tyrosine kinase inhibitor CGP57148B selectively inhibits the growth of BCR-ABL-positive cells. Blood. 90(9):3691–3698.
7. DiMasi JA, Hansen RW, et al. (2003). The price of innovation: new estimates of drug development costs. J Health Econ. 22(2):151–185.
8. Drews J (2000). Drug discovery: a historical perspective. Science. 287(5460):1960–1964.
9. Evans WE (2004). Pharmacogenetics of thiopurine S-methyltransferase and thiopurine therapy. Ther Drug Monit. 26(2):186–191.
10. Evans WE, Relling MV (1999). Pharmacogenomics: translating functional genomics into rational therapeutics. Science. 286(5439):487–491.
11. Feinberg J (2003). Meeting notes from the 43rd Interscience Conference on Antimicrobial Agents and Chemotherapy (ICAAC). New CCR5 antagonist shows antiretroviral effect. AIDS Clin Care. 15(11):94–95.
12. Frazer K, Seymour AB, et al. (2004). Identification of the genetic basis of individual differences in human serum high-density lipoprotein cholesterol (HDL-C) concentration. Proceedings-American Society of Human Genetics, Los Angeles, California, USA. Garrod A (1902). The incidence of alcaptonuria: a study in chemical individuality. Lancet. 2:1616– 1620.
13. Golimbet VE, Alfimova MV, et al. (2004). Serotonin transporter polymorphism and depressive-related symptoms in schizophrenia. Am J Med Genet. 126B(1):1–7.
14. Gretarsdottir S, Thorleifsson G, et al. (2003). The gene encoding phosphodiesterase 4D confers risk of ischemic stroke. Nat Genet. 35(2):131–138.
15. Helgadottir A, Manolescu A, et al. (2004). The gene encoding 5-lipoxygenase activating protein confers risk of myocardial infarction and stroke. Nat Genet. 36(3):233–239.
16. Hetherington S, Hughes AR, et al. (2002). Genetic variations in HLA-B region and hypersensitivity reactions to abacavir. Lancet. 359(9312):1121–1122.
17. Hopkins AL, Groom CR (2002). The druggable genome. Nat Rev Drug Discov. 1(9):727–730.
18. Idle JR, Mahgoub A, et al. (1978). Hypotensive response to debrisoquine and hydroxylation phenotype. Life Sci. 22(11): 979–983.
19. John S, Shephard N, et al. (2004). Whole-genome scan, in a complex disease, using 11,245 single- nucleotide polymorphisms: comparison with microsatellites. Am J Hum Genet. 75(1):54–64.
20. Kalow, W, Genest K (1957). A method for the detection of atypical forms of human serum cholinesterase; determination of dibucaine numbers. Can J Biochem Physiol. 35(6):339–346.
21. Lee CR, Goldstein JA, et al. (2002). Cytochrome P450 2C9 polymorphisms: a comprehensive review of the in-vitro and human data. Pharmacogenetics. 12(3):251–263.
22. Lynch TJ, Bell DW, et al. (2004). Activating mutations in the epidermal growth factor receptor underlying responsiveness of non–small cell lung cancer to gefitinib. N Engl J Med. 350(21):2129–2139.
23. Mallal S, Nolan D, et al. (2002). Association between presence of HLA-B*5701, HLA-DR7, and HLA- DQ3 and hypersensitivity to HIV-1 reverse-transcriptase inhibitor abacavir. Lancet. 359(9308):727–732.
24. McGuire MC, Nogueira CP, et al. (1989). Identification of the structural mutation responsible for the dibucaine-resistant (atypical) variant form of human serum cholinesterase. Proc Natl Acad Sci U S A. 86(3):953–957.
25. Michael NL, Louie LG, et al. (1997). The role of CCR5 and CCR2 polymorphisms in HIV-1 transmission and disease progression. Nat Med. 3(10):1160–1162.
26. Motulsky AG (1957). Drug reactions, enzymes, and biochemical genetics. J Am Med Assoc. 165(7):835–837.
27. Ogilvie AD, Battersby S, et al. (1996). Polymorphism in serotonin transporter gene associated with susceptibility to major depression. Lancet. 347(9003):731–733.
28. O’Shea JJ, Husa M, et al. (2004). Jak3 and the pathogenesis of severe combined immunodeficiency. Mol Immunol. 41(6–7):727–737.
29. Peltekova VD, Wintle RF, et al. (2004). Functional variants of OCTN cation transporter genes are associated with Crohn disease. Nat Genet. 36(5):471–475.
30. Penny MA, McHale D (2005). Pharmacogenomics and the drug discovery pipeline: when should it be implemented? Am J Pharmacogenomics. 5(1):53–62.
31. Pirmohamed M, James S, et al. (2004). Adverse drug reactions as cause of admission to hospital: prospective analysis of 18,820 patients. BMJ. 329(7456):15–19.
32. Post JC, Ze F, et al. (2005). Genetics of pediatric gastroesophageal reflux. Curr Opin Allergy Clin Immunol. 5(1):5–9.
33. Province MA, Kardia SL, et al. (2003). A meta-analysis of genome-wide linkage scans for hypertension: the National Heart, Lung and Blood Institute Family Blood Pressure Program. Am J Hypertens. 16(2):144–147.
34. Rau T, Wohlleben G, et al. (2004). CYP2D6 genotype: impact on adverse effects and nonresponse during treatment with antidepressants—a pilot study. Clin Pharmacol Ther. 75(5):386–393.
35. Rawlins, M, Thompson J (1991). Mechanisms of adverse drug reactions. In: D Davies, ed. Textbook of Adverse Drug Reactions. Oxford, UK: Oxford University Press. :18–45
36. Roses AD, Burns DK, et al. (2005). Disease-specific target selection; a critical first step down the right road. Drug Discov Today. 10(3):177–191.
37. Salkowitz JR, Bruse SE, et al. (2003). CCR5 promoter polymorphism determines macrophage CCR5 density and magnitude of HIV-1 propagation in vitro. Clin Immunol. 108(3):234–240.
38. Samson M, Libert F, et al. (1996). Resistance to HIV-1 infection in Caucasian individuals bearing mutant alleles of the CCR-5 chemokine receptor gene. Nature. 382(6593):722–725.
39. Santillan AA, Camargo CA Jr, et al. (2003). Association between beta2-adrenoceptor polymorphisms and asthma diagnosis among Mexican adults. J Allergy Clin Immunol. 112(6):1095–1100.
40. Schadt EE, Lamb J, et al. (2005). An integrative genomics approach to infer causal associations between gene expression and disease. Nat Genet. 37(7):710–717.
41. Shimizu T, Ochiai H, et al. (2003). Bioinformatics research on inter-racial difference in drug metabolism. I. Analysis on frequencies of mutant alleles and poor metabolizers on CYP2D6 and CYP2C19. Drug Metab Pharmacokinet. 18(1):48–70.
42. Silber BM (ed.) (2000). Pharmacogenomics, biomarkers and the promise of personalised medicine. Pharmacogenetics-Pharmacogenomics. Snyder L (1932). Studies in human inheritance. IX. The inheritance of taste deficiency in man. Ohio J Sci. 32:436–468.
43. Steimer W, Zopf K, et al. (2005). Amitriptyline or not, that is the question: pharmacogenetic testing of CYP2D6 and CYP2C19 identifies patients with low or high risk for side effects in amitriptyline therapy. Clin Chem. 51(2):376–385.
44. Stoll M, Corneliussen B, et al. (2004). Genetic variation in DLG5 is associated with inflammatory bowel disease. Nat Genet. 36(5): 476–480.
45. Thompson JF, Durham LK, et al. (2005). CETP polymorphisms associated with HDL cholesterol may differ from those associated with cardiovascular disease. Atherosclerosis. 181(1):45–53.
46. Thompson JF, Lira ME, et al. (2003). Polymorphisms in the CETP gene and association with CETP mass and HDL levels. Atherosclerosis. 167(2):195–204.
47. Tong Y, Zhang S, et al. (2004). 8302A/C and (TTA)n polymorphisms in the HMG-CoA reductase gene may be associated with some plasma lipid metabolic phenotypes in patients with coronary heart disease. Lipids. 39(3):239–241.
48. Turki J, Pak J, et al. (1995). Genetic polymorphisms of the beta 2-adrenergic receptor in nocturnal and non-nocturnal asthma. Evidence that Gly16 correlates with the nocturnal phenotype. J Clin Invest. 95(4):1635–1641.
49. Vogel CL, Cobleigh MA, et al. (2002). Efficacy and safety of trastuzumab as a single agent in first-line treatment of HER2-overexpressing metastatic breast cancer. J Clin Oncol. 20(3):719–726.
50. Vogel F (1959). Moderne probleme der humangenetik. Ergeb Inn Med Kinderheilkd. 12:52–125.
51. Wong RW (2003). Transgenic and knock-out mice for deciphering the roles of EGFR ligands. Cell Mol Life Sci. 60(1):113–118.
52. Zambrowicz BP, Sands AT (2003). Knockouts model the 100 best-selling drugs—will they model the next 100? Nat Rev Drug Discov. 2(1):38–51.
53. Zee RY, Lou YK, et al. (1992). Association of a polymorphism of the angiotensin I-converting enzyme gene with essential hypertension. Biochem Biophys Res Commun. 184(1):9–15.
54. Wellcome Trust Case Control Consortium. Genome-wide association study of 14,000 cases of seven common diseases and 3,000 shared controls. Nature. 7 June 2007;447:661–678.