The Impact of Pharmacogenetics on Clinical Drug Development

THE IMPACT OF PHARMACOGENETICS ON CLINICAL DRUG DEVELOPMENT

Historically, fewer than 10% of new chemical entities (NCEs) entering preclinical development are approved for clinical use, often because of unaccept-able toxicity in animal studies or Phase I human trials or insufficient efficacy (Kleyn and Vesell, 1998). The cost of bringing a new drug to market is approximately $500–800 million; the costs of ADRS and treatment failures, discussed earlier, are staggering. The appli-cation of pharmacogenetic research and knowledge could result in streamlining and improving the clinical development process in several ways:

•   by initial toxicogenomic screening of compounds to detect selective metabolism, disposition or action related to known polymorphic enzymes, transporters or targets;

•   by providing an ‘insurance policy’ for drug development outcomes. If the results of pivotal trials do not show efficacy in the whole population, subsetting the population on the basis of genetics may enable identification of a group with positive results that were diluted in the entire population and

•   by enhancing the efficacy and safety profiles of medicines in targeted populations to enable better penetration in the marketplace.

Many pharmaceutical companies now routinely screen NCEs to see if they are metabolized selectively by known polymorphic enzymes, and development is discontinued or altered to include additional pharma-cokinetic studies for many of those that are because of the potentially increased risk of serious ADRs or lack of efficacy in subpopulations of patients (Zuhlsdorf, 1998).

Initially, much of the benefits of genomics was expected to ‘rescue’ NCEs after the drug had been ‘killed’. In reality, the value that late in the drug’s lifetime has not been the case. In contrast, collecting and using genomic information during drug develop-ment and at the time of launch is money and time well spent.

However, understanding why certain drugs that have been removed from the marketplace due to serious adverse events can help in developing follow-on compounds. For example, terfenadine (Seldane) caused ADRs in patients who had a specific CYP2D6 gene polymorphism and also were taking erythromycin. They were unable to metabolize terfe-nadine in this situation, which caused toxic accumu-lation of the drug in the body. The FDA worked with the pharmaceutical manufacturer to distribute appro-priate warnings about the possible risks of its use with concomitant medicines, but the company and FDA decided that the drug’s risk–benefit ratio did not justify its continued use. If a screening test to identify patients at risk for this problem had been available, it might have been possible to keep the drug on the market while protecting some of those most likely to experience toxicity from it (Bhandari et al., 1999).

Perhaps, the most striking example was the removal of Vioxx, a COX2 inhibitor from the marketplace. Identification of the cardiovascular risk took many years to identify with many millions of patients exposed. It is clear that new methods to study tens to hundreds of thousands of patients over multiple years will be needed. These methods will likely rely on technology such as electronic data capture that can collect data quickly and efficiently from the patients as well as the physicians. A low cost, efficient, patient-administered DNA collection approach will also help to identify patients at risk without compromising the access to drugs.

Discussion of the potential impact of pharmaco-genetics on clinical trial design is beyond the scope of this chapter, but it is clear that many phar-maceutical companies recognize its importance and are planning to initiate pharmacogenetic studies in the near future (Ball and Borman, 1997). Lichter and McNamara (1995) suggested one approach for incorporating pharmacogenetics into clinical trials:

•   Perform preclinical identification of metabolic pathways and population screening for common DNA sequence variants of the relevant enzymes, transporters, receptors and target genes (and their homologues), as discussed above.

•   Consider the ethnicity of study populations based on known differences in the frequency of specific polymorphisms.

•   During Phase I trials, type subjects for the genes known to control the drug’s metabolic pathway(s) to allow possible correlation of ADRs with geno-type and use this information as a basis for subject selection in Phase II and III studies.

•   During Phase II trials, type any identified rele-vant polymorphisms in the entire study group. Also type the gene product and related targets in all subjects, allowing assessment of allele frequencies in the population and in responders versus non-responders. Use these data as a basis for subject selection in Phase III trials.

•   If useful genetic markers of efficacy or ADRs are identified during Phase II, the Phase III group could be expanded to include a cohort prescreened to include likely responders and those at low risk of ADRs.

This approach is limited by its reliance on iden-tified candidate genes (genes selected on the basis of existing knowledge or an informed guess) and molecular pharmacology to identify drug–receptor interaction, and down-stream signalling pathways, and unexpected associations (either causal or resulting from LD) may not be recognized.

Another approach that is being used already by some pharmaceutical companies has been thought to hold even greater promise as technological advances increase the accuracy, feasibility and cost-effectiveness of high-throughput whole-genome scan-ning. The benefit of whole genome scanning has not yet been realized.

Regardless of the genomic approach, collection of a single blood sample for DNA analysis from all consenting participants in selected Phase II and clinical trials (after approval by the appropri-ate ethics review boards and provision of specific informed consent by subjects) enables pharmaceuti-cal companies to have the key samples needed in case a safety or efficacy question arises. This sample may be used to identify the occurrence of known polymorphisms affecting drug response, to evaluate candidate genes suspected of being involved in the disease or drug response and to assess patterns of SNP or haplotype occurrence related to efficacy or ADRs, allowing the creation of a SNP Print^SM to screen potential subjects or patients (post-approval) for their likely response to the drug or determine heterogene-ity of the disease in patients with similar phenotypes (Roses, 2000b).

Regulatory agencies might be concerned, appropri-ately, that the smaller numbers of patients in these streamlined clinical trials would be insufficient to detect rare ADRs (<1:1000) and that patients who did not receive or ‘pass’ the recommended MRT for the drug would nevertheless receive it and be at increased risk of harm. However, rare ADRs are not likely to be detected even in the relatively large clinical trials that are conducted now; it certainly is not feasible to enroll the approximately 65 000 patients that would be required to be 95% confident of detecting three or more cases of an ADR with an incidence of 1:10 000 (Lewis, 1981). The major, albeit rare, ADRs asso-ciated with dexfenfluramine, zomepirac, benoxapro-fen, troglitazone and terfenadine were not detected until after they reached the market. Extensive preapproval safety testing in even larger populations is a possible solution, although, as noted above, it will be impractical to identify very rare ADRs in clin-ical trial study populations, and the increased cost and delayed time to market is likely to create signif-icant financial barriers from the perspective of the pharmaceutical companies (and ultimately consumers and payers to whom the cost will be passed along) (Roses, 2000a).

One solution to this problem would be an exten-sive, regulated post-approval surveillance system that incorporates the collection of pharmacogenetic data. Roses (2000b) proposes that hundreds of thousands of patients receiving a medicine would have filter paper blood spots taken (perhaps from the original blood sample used for the MRT) and stored in a central location. As rare and/or serious ADRs are reported and characterized, DNA from affected patients could be compared with that of control patients, allowing ongoing refinement of the MRT. There is increas-ing pressure to improve the inconsistent and largely unregulated current system of post-marketing surveil-lance, and many authors agree on the need to incorporate pharmacogenetic data in some form into a revised system (Edwards and Aronson, 2000; Nelson, 2000).

Another approach is one that would put increas-ing control of medical data in the hands of those most directly affected by it – consumers. In this scenario, an individual could choose to have a one-time blood sample taken for DNA analysis and stored at a tightly secured central repository. As research into disease-related genes, genetic risk factors and genetic associations with medicine responses progressed, the consumer or a designated representative (such as a health care provider) could request that the sample be analyzed using relevant MRTs (including SNP Prints^SM and other markers. This ‘bank’ could serve as a central repository for the samples themselves and as a central database of information including well-established knowledge, current research and even opportunities for clinical trial subjects with specific conditions or genotypes. It could trigger genetic ‘alerts’ to consumers who chose to provide a medical and family history as new research results potentially relevant to them became available. A host of ethical, legal and social issues would need to be addressed as part of this venture, but it presents one option for an efficient, centralized and consumer-controlled bank of health-related genetic expertise and information.