May 1, 2016 (Vol. 36, No. 9)
Eric T. Fung M.D., Ph.D. Vice President of R&D Clinical Applications Affymetrix
Standardized Assays Are Driving Preemptive Genotyping and Personalized Drug Therapies
For decades, genotyping has promised to serve as a practical means of relating genetic make-up and pharmacological efficiency—first at the level of patient groups and more recently at the level of individuals. Genotyping, however, still has a fairly limited role in determining which drug therapies, and which doses, should be used in specific circumstances.
If genotyping is to find widespread adoption, it will have to overcome several barriers, most notably variation in assays and delay in reporting, difficulty in translating genotype into specific actions, and a perceived lack of economic and/or clinical value. Technological advances coupled with changes in the availability of genetic information will dramatically change the landscape of pharmacogenetics.
The efficacy of any given drug therapy is dependent on a number of factors, most commonly described through the pharmacokinetic parameters of absorption, distribution, metabolism, and elimination (ADME). Together, these factors determine whether a patient will need increased or decreased dosages, or whether a given therapy will work at all in that patient. Additionally, these factors can determine drug-drug interactions for patients on polypharmacy.
Although a detailed description of specific genotype variants is beyond the scope of this article, a brief survey of the diversity of genotypes is helpful to provide a sense of the complexity that is inherent in genotyping, which has, in some ways, slowed the adoption of pharmacogenetics. As an example, human leukocyte antigen (HLA) genes are among the most highly polymorphic genes; more than 3,600 HLA class II alleles have been described.
More than 50 human cytochromes P450 (CYPs) have been identified, and most have at least several single nucleotide polymorphisms (SNPs), with CYP2D6 having over 100 identified SNPs. Specific combinations of polymorphisms are translated into star alleles, which are used to predict the impact on therapeutic response.
As might be expected, any individual enzyme can metabolize multiple drugs, and most drugs can be metabolized by multiple enzymes. Drugs can also inhibit metabolizing enzymes, while metabolizing enzymes can activate drugs by converting prodrugs into active metabolites. Generally, changes in functional activity of the enzyme are translated clinically by categorizing patients as poor, intermediate, extensive, or ultrarapid metabolizers.
The FDA has now included pharmacogenomics information in the labeling of 166 approved drugs, some of which include specific action to be taken based on biomarker information. Table 1 summarizes the biomarkers and indications for the pharmacogenomics labels. The FDA labels range from dosage and pharmacokinetics information to precautions and, in nine of the labels, boxed warnings to highlight potentially serious adverse reactions.
Most pharmacogenetics assays are currently offered as laboratory-developed tests; therefore, there is a wide range in the specific variants that are reported for any given target. As noted above, CYP2D6 has over 100 identified SNPs, and laboratories report various numbers of star alleles. Historically, this is because most genotyping assays involve methods based on the multiplex polymerase chain reaction (PCR). Accordingly, in these assays, the cost or effort to perform the genotyping is approximately proportional to the size of the panel.
Additionally, because some of the functional variants are copy number changes, multiple assays may be required (for example, quantitative PCR for copy number, plus PCR for genotyping). More recent advances in microarray technology make it possible to perform more complete genotyping and copy number analysis of known star alleles simultaneously across multiple genes, thus reducing the cost and increasing the efficiency of pharmacogenomics. For example, the Affymetrix DMET Axiom Assay can analyze over 4,000 genotypes across 900 genes along with copy number in a single assay.
From a regulatory perspective, it is likely that the disparate technologies laboratories use to generate their pharmacogenetics results will coalesce into a few, defined FDA-cleared devices. Because arrays can reproducibly provide comprehensive genotyping and copy number information at low cost, analytical and clinical validity can be readily demonstrated in a regulatory submission.
The translation of specific genotype combinations into actionable clinical utility is hampered by difficulties in interpretation. Part of this relates to the somewhat ambiguous notation of the impact of a given star allele; the designation “ultrametabolizer,” for example, does not obviously translate to a specific dose for a given individual.
Additionally, parameters such as ethnicity, age, body mass index, and gender can influence the pharmacokinetics in any specific individual. The establishment of guidelines can assist the practitioner in utilizing pharmacogenetics information to make therapeutic selections. At the forefront of establishing guidelines is the Clinical Pharmacogenetics Implementation Consortium (CPIC), which provides guidelines centered around specific genes as well as for specific drugs.
In most cases, physicians need to make treatment decisions immediately and cannot wait for genotype results. The obvious solution to this is preemptive genotyping, which is being deployed at five academic medical institutions (Mayo Clinic, Mount Sinai, St. Jude Children’s Research Hospital, University of Florida and Shands Hospital, and Vanderbilt University Medical Center) as part of the Translational Pharmacogenetics Program.
For preemptive genotyping to be widely deployed, the structure of electronic health records (EHRs) will need to evolve so that they enable the retrieval, storage, and reporting of complex genotyping data. Moreover, they will need to be able to provide the translation of star alleles with metabolizing status for specific drugs, dosing guidelines or suggestions for alternative drugs, and links to guidelines and other supporting information.
The most sophisticated embodiments of EHRs will also take into account other information that can influence dosing contained within the EHR, such as the patient’s ethnicity, weight, sex, and other medications. Most EHRs lack such capabilities, but two trends will substantially alter this landscape.
First, there is an increasing recognition of the role medical informatics plays in healthcare and an increased emphasis on this role at medical institutions, both academic and community-based. Second, the entry of high-tech giants such as Google and Apple into the medical informatics and large-scale genotyping/genetic analysis arena will accelerate the development of these tools.
Third-party payers have generally been reluctant to pay for most pharmacogenetics tests. The paucity of prospective randomized clinical studies showing either clinical or economic utility remains a fundamental hurdle for widespread adoption of pharmacogenetics. A likely path for the generation of clinical data will be through large, publicly funded genotyping initiatives in combination with investigator-initiated studies that rely primarily on mining EHRs for dosing, adverse reaction, and outcome information.
One such initiative is tied to the Million Veterans Program. It is mining data to explore the pharmacogenetics of metformin response in diabetics with renal disease.
Another push may come from consumers who choose to proactively obtain their pharmacogenetics information. Such activity will heavily depend on the appropriate EHR and bioinformatics infrastructure at primary care centers as well as harmonization of analytical test methods. These requirements suggest that consumer-driven work will lag the efforts at academic medical centers.
The pace at which pharmacogenetics is incorporated into healthcare will increase due to factors such as the decreasing cost of genotyping, the installation of a medical informatics infrastructure, and increased consumer demand for personal genotyping information. Moreover, these factors will reinforce each other and help preemptive genotyping become the norm rather than the exception.
As this trend gathers momentum, it will begin contributing to a virtuous cycle in which the increased availability of genotyping data associated with outcome information will permit the development of additional and more precise treatment algorithms. Technological advances in genotyping, most notably high-density genotyping at low cost with high reproducibility, and medical informatics will be key to making this a reality.