Drug therapy tailored to the individual and rational discovery of better therapies are exciting prospects that have universal appeal. Science and technology have provided the framework and tools for genetic analysis of the human drug response, but whether these advances will improve patient care in a cost-effective manner remains to be established. It is the hope of personalized medicine that patient treatment guided by profiles individualized for specific drugs can provide an approach that will meet these objectives. It is reasonable to expect that risk profiles for drug susceptibility based on predictive genetic markers would provide the means to translate the molecular foundations of pharmacogenetics into these objectives.
The construction of risk profiles necessitates the collection of genomic data on a large scale. Pharmacogenetics is well positioned to construct such profiles suitable for medical practice from the extensive knowledge accumulated on monogenic pharmacogenetic traits. At present, the number of pharmacogenetic markers that have been characterized exceeds by a considerable margin the estimate of several dozen made almost two decades ago. Furthermore, recent molecular studies indicate that such traits are often associated with a limited number of functionally important variants, raising the prospect that the responsible genes may be cataloged in the near future.53,54 With the Human Genome Project completed, with genomics technologies for identifying and scoring polymorphisms, and with bioinformatics tools for handling large data sets well in hand, important parts of this task are already in place.
Despite the documented functional relevance of many drug-metabolizing genes, prospective gene-based drug prescribing has not yet been translated into clinical practice.13,34,55-59 Although scientifically sound, the pharmacogenetic evidence linking single-gene polymorphisms to functional outcome has often been incomplete or difficult to access, complicating its translation into clinical practice. The reasons for this are numerous and complex:
1. Pharmacogenetic studies often have not assessed clinically relevant endpoints. Many early investigations were performed to test proof of concept, i.e., to examine the relevance of genetic diversity to variations in human drug response. The pharmacogenetic studies performed during the 1960s and into the 1980s were not usually conceived or designed to assess clinical endpoints in a clinical setting. Only with ample evidence that genetic diversity was an important contributor to human drug response were the aims of research altered to assess clinically relevant endpoints.
2. Objective criteria appropriate for assessing clinically relevant outcomes were lacking. Most investigators in the earlier studies used one or more unfamiliar variables (genotypes, drug metabolite ratios, urinary drug/ metabolite patterns, drug plasma concentrations, etc.) to express the outcome, and the clinical implications of such variations may not have been clearly or sufficiently defined.
3. The statistical power of studies was low. Many of the early studies were limited to a small number of individuals or small populations, so their statistical power was low.
4. Many studies failed to evaluate the effects of modifying genes. The evidence suggests that most pharmacogenetic markers are monogenic in origin (i.e., in any particular family, only one responsible locus is thought to be defective). As these traits are examined in greater depth, we find that analyses at a single locus will reliably predict few phenotypic outcomes. Principal among the problems encountered in dissecting the molecular basis of even the simplest disorders attributed to mutations of a single gene, as Peltonen and McKusick60 observed recently, are the modifying effects of other genes. Existing information about monogenic disorders demonstrates that modifier genes can bring about substantial variations in the clinical phenotype for diseases such as cystic fibrosis and Hirsch-sprung's disease. Arguably, the metabolism of certain drugs is governed by a single rate-limiting step, and for them a close genotype-phenotype correlation might be expected. Admittedly, however, among all of the pharmacogenetic traits that have been characterized, none to our knowledge has tested the hypothesis that the modifying effects of other genes have no effect on the expression of the trait. Hence, establishing the relationship of genotype to phenotype for the human drug response represents a major challenge to the construction of genetic susceptibility profiles.
5. Drug susceptibility profiling implies a causal relationship between genotype and phenotype. Somehow, the belief that knowledge of the genotype is sufficient to determine the phenotype and predict the response of susceptible individuals has gained a credible audience. However, this notion ignores our ignorance of biology and defies expert opinion. First, the potential effects of modifying genes on the expression of a trait for a specific drug must be taken into account, as noted previously. Second, the functional consequences of genetic diversity must be precisely defined for each drug specified and for each of the separable phenotypes that characterize a given trait. This means that the drug susceptibility phenotype that is expressed in response to a specific drug must be causally related to the genotype (or haplotype) of a specific individual as precisely as data will allow.
6. Adequate methodology for high-throughput phenotyping is not available. As was noted in points 4 and 5 above, knowledge of genotype alone is usually insufficient to predict drug-specific phenotypes. Progress toward partial resolution of this problem has been made by technical means. Genotyping complemented by phenotyping with probe substrates has been used to determine the activity of specific drug-metabolizing enzymes.61-64
For example, Frye et al.61 showed that a cocktail of noninteracting probe drugs (caffeine, chloroxazone, dapsone, debrisoquin, and mephenytoin) can be administered simultaneously to estimate in vivo activities of several CYP450 and N-acetyltransferase enzymes. Subsequently, Dierks and co-workers62 developed a method for simultaneously evaluating the activities of seven major human drug-metabolizing P450s (CYP1A2, 2A6, 2C8, 2C9, 3A4, 2D6, and 2C19) using midazolam, bufuralol, diclofenac, ethoxyresorufin, S-mephenytoin, coumarin, and paclitaxel as probe substrates to monitor activity, and ketoconazole, quinidine, sulfaphenazole, tranylcypromine, quercetin, furafylline, and 8-methoxypsoralen as inhibitors to monitor inhibition in human liver microsomes. These techniques have been employed mainly in a research setting, but there is a pressing need for phenotyping methodology applicable to large-scale, high-throughput analyses of therapeutic drugs. The method of Frye et al. appears to be directly applicable to this end, and perhaps the other methods could be adapted to this purpose.
7. Knowledge of physiological or pathological factors that affect the regulation of genetic markers is lacking. Infections or inflammatory stimuli, or cytokines and interferons employed as therapeutic agents, as well as hormones and nutritional status can cause changes in the activities or expression of various forms of human drug-metabolizing enzymes.65-67 These factors have the potential to alter, adversely or beneficially, the therapeutic or toxic effects of drugs, and these effects are particularly crucial for drugs with a low therapeutic index. For certain drugs, the effects of these factors on the drug level have been known for many years (see Table 1 in Morgan65), but systematic investigation of their effects on the expression of Phase 1 and Phase 2 drug-metabolizing enzymes has only recently begun.
Was this article helpful?