The Changing Paradigm of Drug Development and Delivery

THE CHANGING PARADIGM OF DRUG DEVELOPMENT AND DELIVERY

OTHER RELEVANT POLYMORPHISMS

Some sequence polymorphisms that are involved in disease pathogenesis also may be involved in determining drug response, directly or indirectly. One example is apolipoprotein E4 (Apo-E4), a risk factor for familial late-onset and sporadic Alzheimer’s Disease (AD) that has been reported to predict poor response to the cholinesterase inhibitor tacrine (Farlow et al., 1996). The presence of the E4 allele also may be a factor in the success of prophylactic oestrogen therapy for AD (Sadee, 2000). In addi-tion, Apo-E4 is associated with increased risk of coronary artery disease (CAD). Gerdes et al. (2000) reported that presence of this allele is associated with almost 2-fold increased risk of death in myocardial infarct survivors and that this increased mortality rate can be abolished by treatment with simvastatin (HMG-CoA/3-hydroxy-3methylglutaryl coenzyme A reductase inhibitor). Increased understanding of the underlying multigenic causes of AD, CAD and other diseases and neurodegenerative disorders may lead to the development of strategies for disease treatment and even prophylaxis for those at high risk of devel-oping a disease.

Another example of a disease-related polymor-phism that is predictive of drug response involves cholesteryl ester transfer protein (CETP) and pravas-tatin (HMG-CoA reductase inhibitor used to treat hypercholesterolemia). Kuivenhoven et al. (1998) reported a significant relationship between variation of the CETP gene and the progression of coro-nary atherosclerosis, independent of lipolytic plasma enzyme activity and plasma HDL cholesterol concen-tration. If these results are replicated, then the pres-ence of a homozygous polymorphism at this site could be used to predict whether treatment with pravastatin will be effective. Replication of the study has led to variable results, but Boekholdt

et al. (2005), conducted a meta analysis of seven large population-based studies (total patients >3500) and two randomized placebo control trials. They found that Taq1B polymorphism was associated with HDL-C and subsequent CAD, but not with Pravastatin therapy.

CURRENT CHALLENGES IN THE CLINICAL APPLICATION OF PHARMACOGENETIC KNOWLEDGE

It is clear from this cursory review of the current state of knowledge regarding the genetic basis of ADRs that many clinically significant genetic poly-morphisms affecting drug response in humans have been described already. The emphasis to date has been on identification of mutant alleles at a single gene locus (e.g. Phase I and II hepatic enzymes, TPMT and DPyDH), and this research has been fruitful. However, drug response depends on the drug’s inter-action with the many proteins involved in its absorp-tion, distribution, excretion and target site, each of which is coded for by genes that may be associated with common variants that may affect response. For example, one individual may exhibit polymorphisms of genes coding for two drug-related proteins: one that affects the degree of drug inactivation and one that determines the sensitivity of the drug receptor. The polymorphism affecting the drug’s metabolism would determine the plasma concentrations to which the individual is exposed, and the polymorphic recep-tor would determine the nature of the individual’s response at a given drug concentration. These poly-genic interactions are much more difficult to establish during the course of clinical drug trials than are the monogenic effects discussed above but may have an even more significant impact on drug response.

The effects of environmental factors may be modi-fied by wild-type or polymorphic genes, as well, intro-ducing more confounding variables. The majority of these gene variants are relatively uncommon in the general population, making it difficult to establish their role in drug response and demonstrate clini-cal relevance, especially for heterozygous individuals who are likely to exhibit more subtle effects than homozygotes.

Clinical research in disease genetics and pharma-cogenetics has, at times, produced discordant and contradictory results, creating confusion and result-ing in a lack of credibility in the minds of many health care providers. Inconsistent results may be due to several factors, including the lack of strict diagnostic criteria for study entry, the heterogeneous nature of the diseases being studied, the use of differ-ent end points and scales for assessing drug efficacy and ADRS, the presence of unknown or unidenti-fied environmental factors and the polygenic nature of many drug effects (Evans and Relling, 1999). It is crucial in clinical pharmacogenetic research that the study sample be of adequate size to demonstrate the necessary statistical power and that the results be rigorously confirmed in comparable populations by other researchers (Manasco, Rieser and Pericak-Vance, 2000). In addition, once a drug has been approved, ongoing, systematic centralized collection of meaningful, evaluable data regarding drug efficacy and ADRs does not occur routinely, and pharmacoge-netic data are rarely collected at all. As a result, oppor-tunities for increasing our knowledge of dramatic and subtle genetic effects on drug response both in large numbers of diverse patients and specific diagnostic subsets of patients are lost.

Doroshow, in his review of the gefitinib clinical trial and regulators at the FDA in their guidance documents on drug/device co-development highlight the need to prospectively collect biological samples linked to clinical data and consent during Phase III trials (www.fda.gov/genomics). Experience from the development of Herceptin, Iressa and now Tarceva highlight the need to have samples that accompany the clinical data to enable market-ready tests to be developed, reviewed and marketed at the time of drug release.

In 2004, Merck removed its COX2 inhibitor Vioxx from the marketplace after it was found to be associ-ated with increased risk of cardiovascular side effects. Considerable backlash directed towards the FDA and the pharmaceutical industry resulted. The increased scrutiny provided the impetus for enhanced pharma-covigilance efforts. Furthermore, there was a recog-nition that the AERS voluntary reporting system was not adequate to fully evaluate post-marketing safety.

Several of the guidance documents highlighted the opportunity the pharmacogenomics can play in iden-tifying populations at risk.

Much pharmacogenetic research to date has involved identifying and categorizing drug-related polymorphisms while relatively little has been done to determine clinical relevance in well-defined populations. Clinicians do not know which variants should be assessed, how and by whom that should be done, what drugs might be affected, what course of action would be appropriate based on the informa-tion obtained and who will cover the cost of the test. Should the dose be altered? By how much? Should the drug be avoided entirely? What about related drugs and polymorphisms? Must each be tested sepa-rately? What effects do the variants have on drug–drug and drug–environment interactions? What issues exist around professional liability and the ethical, legal and social aspects of such testing? Carefully designed, well-controlled clinical studies in appropriate popu-lations must be carried out to begin to answer these pressing questions, and the information then must be made available to clinicians and reflected in ethically grounded standards of clinical practice and compen-sation procedures.

In addition, standard pharmacotherapy references and treatment guidelines formulated by various health care organizations rarely contain relevant pharmaco-genetic information even when it is known, making it difficult for clinicians to gain access to existing data. Consumers, health care providers, payers and regu-latory agencies lack basic education with regard to pharmacogenetics and timely access to relevant new data as they emerge. Although the lay press occa-sionally spot-lights a tragedy that could have been averted through the application of existing pharmaco-genetic knowledge (such as ‘overdose’ deaths of slow drug metabolizers; Stipp, 2000), the need for increased professional and public awareness and education in this arena is equal to the need for continuing research.

Until recently, genotyping an individual was a labo-rious, time-consuming and costly proposition that was undertaken only if there was a high index of suspi-cion of an identified genetic disorder. The Human Genome Project and the multitude of high technology spin-offs from it are changing this situation. Auto-mated instrumentation, new bioinformatics systems and novel strategies derived from genomic research will enable researchers to evaluate and analyze the wealth of genetic information that will continue to emerge. High-throughput DNA sequencing, gene mapping and transcriptional analyses are becoming economically and scientifically feasible as a result of innovations such as DNA, cDNA (‘edited’ version of a gene, containing only the parts that will be expressed as proteins) and oligonucleotide microarrays and microfluidic analytical devices (Mancinelli, Cronin and Sadee, 2000).

Because genetic information does not change, it is conceivable that a consumer could have their genetic sample or genetic information in a central location that is available to healthcare providers in any location and at any time. The need for point of care genetic tests will ultimately be eliminated.

In contrast, tests that measure expression of a gene vary over time and thus the need for dynamic measurement for gene expression or protein expres-sion will be needed in some cases.

SINGLE NUCLEOTIDE POLYMORPHISMS, MEDICINE RESPONSE TESTS AND ‘GENETIC TESTS’

Many of the previously mentioned polymorphisms directly alter the metabolism, transport, action or excretion of medicines through identified (or identifiable) structural or functional effects; there is a causal relationship between the polymorphism and the phenotype. New genomic techniques such as those mentioned above are making it possible to detect associations (which may or may not be causal) between specific genetic markers and indi-vidual response to medicines. Single nucleotide poly-morphisms (SNPs, pronounced ‘snips’), single-base differences in DNA sequence, are the most common form of human polymorphisms. They occur with an average frequency of about 1 per 1300 base pairs, serving as easily identifiable virtual mileposts along the three billion base pair human genome (Interna-tional Human Genome Sequencing Consortium, 2001) (See Figure 51.1).

The SNP Consortium (a not-for-profit organization of pharmaceutical and bioinformational companies, academic centres and a charitable trust) produced an ordered high-density SNP map of the human genome, which is publicly available at http://snp.cshl.org. This map is being used to find disease genes and to correlate genetic information with individual responses to medicines. Although few of these SNPs are expected to be involved directly with disease or medicine response, they will be useful as analytical tools to track small segments of the genome. Indi-viduals who carry a particular gene variant (allele) are likely to carry variants of several SNP markers that are close to or within that allele because of the phenomenon of linkage disequilbrium (LD; when alle-les are in close physical proximity, they are likely to be inherited together).

The results of comprehensive analysis of genetic variability in genes related to the action of the medicine in question or the disease process as well as whole-genome SNP scanning obtained during Phase II clinical trials of a medicine can be used to iden-tify specific SNP markers, patterns or haplotypes that correlate to patient responses (efficacy and ADRs). The medicine response-related data could form the basis for the selection of patients most likely to respond well in Phase III trials, possibly making these trials smaller, faster and more efficient if the goal of the study is to validate markers for efficacy. Studies designed to validate safety markers will require larger numbers of patients due to the rate of the adverse events.

The FDA has issued guidance related to co-development of a drug and predictive test. This guid-ance is focused on development of a predictive test if the markers are known at the time of Phase II trials. Unfortunately, this is not the usual case for drug development, but the emphasis on how samples are collected (with appropriate documented consent using processes that enable evaluation of chain of custody) and the need for a prospective approach to studying and replicating genetic findings will apply in other situations as well (www.fda.gov/genomics).

The FDA has also formed an interdisciplinary review group to review voluntary genomic data submissions (VGDS). The VGDS process enables sponsors to submit data to the agency and get feed-back without being required to submit the data to the NDA. If genomic data are to be included as part of the label or as part of the decision making process, then the VGDS process does not apply.

Functional enzyme analysis of TPMT in red blood cells before treatment with specific cancer chemother-apy drugs is one example of an MRT in current use in the United States. Another is HercepTest^®, which uses a polyclonal antibody to detect HER2 protein, reflecting HER2 expression in breast cancer cells; it is used to predict patient response to trastuzumab (Herceptin^®), a humanized monoclonal antibody against the HER2 receptor. Researchers already have developed other tools for assessing HER2 expression, including a monoclonal antibody test for the HER2 protein, a test for circulating HER2 protein (in the extracellular domain) and a test using fluorescence in situ hybridization (FISH) directly to determine the number of copies of the HER2 gene (not its protein product). Because of the strong corre-lation between overexpression of HER2 protein and response to Herceptin^®, the US FDA required that a test kit to assess HER2 protein expression be commer-cially available before drug approval – an example of a regulatory agency mandating the availability of a predictive test linked to use of a specific drug.

Transcriptional analyses, in which the expres-sion levels of DNA are measured, may provide another approach for predictive tests for medicine response (Kleyn and Vesell, 1998). RNA obtained from biopsied tissue and surgical specimens can be used for expression-based studies in some cancers, allowing detection of somatic changes associated with the development of some tumours and their response to chemotherapy. For example, the ampli-fication of the oncogene erb-B2 predicts a good response to treatment with a specific adjuvant ther-apy (cyclophosphamide-methotrexate-5-fluorouracil) for breast cancer (Muss et al., 1994). Alternately, the expression of genes predicting drug response can be assayed at the protein level using antibody-based tests of serum or other tissues (Kleyn and Vesell, 2000).

Ongoing genetic and genomic research undoubt-edly will result in the development of additional tools that can be incorporated into or used as the basis of medicine response tests. The rationale exists for conducting pharmacogenetic analyses to look for associations between drug responses (safety and effi-cacy) and genotype, and the technology exists for conducting these analyses. The missing piece is a pool of DNA samples with associated medical and medicine response data to facilitate the efficient conduct of pharmacogenetic research. Eventually, MRTs based on SNPs and other genetic polymor-phisms will enable health care providers to identify patients at high risk of developing a given disease, implement preventive therapy and lifestyle adjust-ments when appropriate and choose the medicines that are most likely to benefit the patient and least likely to result in serious ADRs (Mancinelli, Cronin and Sadee, 2000).

In the past, the term ‘genetic testing’ has been asso-ciated with the diagnosis of monogenic diseases such as cystic fibrosis and Huntingdon disease – conditions for which a causative, single, genetic mutation has been identified (Table 51.4). A newer area of genetic research and testing involves identification of genes related to the occurrence of common complex diseases such as asthma, heart disease and migraine. These diseases are likely to result from the interaction of multiple ‘increased risk’ or susceptibility genes with each other and possibly with environmental factors. These types of genetic research and testing (related to monogenic diseases and susceptibility genes) involve determining the likelihood of occurrence (prediction) or the presence (diagnosis) of disease in individuals. Although very useful and important, there are social and ethical risks related to the nature of the infor-mation revealed by these tests – what it means, who has access to it and how it can be used. There is general agreement on the need for genetic counselling to help patients and families understand and process the results of these disease- and risk-related genetic tests.

In contrast, the risks associated with tests to detect polymorphisms related to response to medicines, such as metabolic and drug receptor or target character istics and genomic profiles, are minimal: the data that are obtained are quite limited and specific to the drug(s) being considered. No information about disease, causal or susceptibility genes, is likely to be obtained. These tests, when validated, will be simi-lar to routine laboratory tests such as blood typing, drug concentration monitoring and liver enzyme anal-yses. Although health care providers would discuss the results with patients, there would be no need for genetic counselling and ongoing psychosocial support related to interpretation of the results, with rare exceptions.