Barriers to the adoption of pharmacogenetic tests in clinical practice^3,4 include the fragmentation of health-care systems that preclude linking a “lifetime” genetic test result with future medical care, the low use of electronic medical records that are vital to linking test results with medication prescribing/dispensing/administration, health-care systems that do not reward the prevention of disease (or adverse drug effects), the lack of sufficient awareness about genomics on the part of many clinicians, and the fact that little of such testing is done preemptively and therefore the results are not available when the prescribing decision is made. Some of these barriers will persist for many years to come.

One barrier to clinical implementation of pharmacogenetics that is addressable⁵ is the lack of clear, curated, peer-reviewed guidelines that translate laboratory test results into actionable prescribing decisions for specific drugs. It is the goal of the CPIC (http://www.pharmgkb.org/views/project.jsp?pId=74) to provide such guidelines, the first of which is published in this same issue.⁶ The guidelines will center on genes (e.g., thiopurine methyltransferase (TPMT) and its implications for thiopurines) and drugs (e.g., warfarin and all the major genes that influence its action).

It is the goal of several laboratories to offer high-quality clinical pharmacogenetic tests at multiple loci for a low cost from a single DNA sample. When this occurs, it will allow for “preemptive” generation of pharmacogenetic test results; individuals would be able to have such pharmacogenetic results in their medical records, ready for use in the clinic if and when needed.⁴ This development could well have a substantial influence on several of the survey responses. For example, the cost of and reimbursement for the test results would likely be much less of a barrier. In addition, the preemptive availability of specific gene test results will shift the balance in favor of a prescribing model that requires gene-specific clinical guidelines: once the genetic test result is in the medical record, all the medications affected by that gene theoretically place that individual “at risk” if they are prescribed. Guidelines that state which medications would be most affected, and how their prescribing should be changed based on the gene test result, will be useful. Moreover, if the preemptive genetic testing encompasses the most important pharmacogenes, the clinician’s practice paradigm would probably change from “I want to prescribe drug x; I should order a test for gene z, check results, and then decide how to proceed with prescribing” to “I want to prescribe drug x; I will check the pharmacogene profile for this patient to see if there are genetic considerations regarding that medication in this patient.” High-risk genetic test results could be linked through decision-support tools to any attempts to prescribe, dispense, or administer high-risk affected drugs. One can see from this scenario that integration of medication use with laboratory testing via electronic links will be invaluable to the process of implementing pharmacogenetics in the clinic.²

It should be acknowledged that one limitation of such surveys is that there is a relatively small number of clinical experts who have experience in implementing pharmacogenetics; therefore, despite our attempt to direct our survey at individuals with a particular interest in this area, less than 20% of respondents described the use of pharmacogenetics at their organization as “routine” and only ~50% of respondents considered themselves practicing clinicians.

Each CPIC guideline will adhere to a standard format. Genes, drugs, and dosing recommendations will be categorized in each document. Specifically, each guideline will contain an introduction summarizing the drug dosing that is addressed as a result of specific genotyping tests, a focused literature review, gene-based information (genetic test interpretation for clinicians with population studies described if available, genetic test options, incidental findings such as “X diseases or conditions that have/have not been linked to variation in gene(s) Y, unrelated to medication use”), and drug-based information (background linking genetic variability to variability in drug-related phenotypes and levels of evidence and strength of recommendation for dosing recommendations, as discussed below). Table 1 provides an example of key data needed by clinicians: the assignment of likely phenotypes based on genotypes (Table 1). Each guideline also includes dosing recommendations for drug(s) based on genotype/phenotype, such as are included in the current package labeling for warfarin (Coumadin), along with a graded strength for each dosing recommendation, based on detailed levels of evidence graded as to its quality. The analyses of costeffectiveness are beyond the scope of these guidelines.

For clinicians, there will be a substantial need for gene-centric guidelines. Because the genetic test result has lifelong relevance, it would be optimal to have a mechanism for linking the information in the test result with all the potentially risk-related medications that may be prescribed to the patient over his or her lifetime, rather than only with the particular medication that may have prompted the ordering of the genetic test. Currently, pharmacogenetic tests are often ordered for the purpose of determining the dosage or whether to prescribe a particular medication linked to that gene. Indeed, some clinical laboratories offer the pharmacogenetic tests specifically paired with the agent of interest (e.g., the CYP2D6 test for tamoxifen therapy and the CYP2C9 test for warfarin therapy), and the laboratory provides test results in the context of implications for use of that particular agent. However, once the genetic test results are in the medical record, there are implications for all the agents whose effects are strongly linked to that particular gene, and these remain relevant for the lifetime of the patient. Moreover, the scientific data that link medications to variations in particular genes are constantly being updated. Therefore, a mechanism must be created for generating gene–drug interaction–related guidelines that can be updated to accommodate “new” drugs, and these updates must be freely available to clinicians. Of course, the alternative of having drug-centric guidelines will also have some value. Warfarin is a case in point:⁷ at least two genes (CYP2C9 and VKORC1) have a substantial impact on warfarin dosing, and therefore clinicians may become accustomed to ordering those two tests before starting the drug in a patient, and practice models may be developed that allow fast enough laboratory-test turnarounds to support the “test first, then prescribe” model.

A three-tier scheme is used to rate the quality of evidence linking drug-related phenotypes to specific genetic variations:

Likewise, a three-tier rating scheme is used for evaluating the strength of the recommendation: