Pharmacogenomic Testing: Considerations for Clinical Diagnostic Laboratories

This month's expert is Michael Murphy. View the presentation and direct your questions to our online expert. AACC would like to thank Bayer HealthCare Diagnostics for making this program possible.

In regards to ethical issues, could you discuss this comment in more depth? "When do we test all patients receiving a particular drug?"
Washington, DC

Michael Murphy: This is an excellent question and one all of us in the industry struggle to make the decision, “When to test?”. In my opinion, when it is well established that there might be adverse drug reactions causing substantial harm to the patient, we should test. For example, we know after many years of research that Poor Metabolizers of the enzyme CYP2D6 are at risk of adverse events when they take tricyclic antidepressants. Given that Poor Metabolizers are present up to 6-10% of some ethnic groups, the number of affected patients can be significant. We should start introducing pharmacogenomic testing for many of the previous approved drugs that fit into a similar situation outlined above. The mandate to test must come from the FDA and needs to be reflected on the package insert (label) so that everyone knows when to test. Finally, there must be a cost-benefit analysis done for each of these cases. We should choose the drugs with the highest rates of prescribing that cause the greatest amount of morbidity and mortality in patients with altered metabolism (e.g., Poor and Ultrarapid Metabolizers).

I agree with the point that pharmacogenetic testing will help optimizing therapy. One problem not mentioned in the presentation, however, is: how much SNPs do we test? Taking CYP2D6 as an example, how many polymorphisms does one want to test to identify poor metabolizers? One SNP? Ten? And will this be the same SNPs for each ethnic population? The point is that this affects the value of a negative test (i.e. no polymorphism found, so no PM?!?). In reporting these tests, guidelines ask us to give an indication of the chance that a PM is missed (value of a negative test). This may be different when Caucasians are involved compared to, for instance, African Americans. How should we deal with this? Perhaps the FDA (or another regulatory body) may define that PMs should be excluded from 4 or 5 predefined major ethnicities with a confidence level of 90% or 95%? This would directly set a standard for how many polmorphisms should be tested. What is your opinion on this topic as an expert? Ron van Schaik Head Pharmacogenetics Core Lab AKC Dept Clinical Chemistry Erasmus MC Rotterdam The Netherlands
Rotterdam, The Netherlands

Michael Murphy: This is another excellent question that is both timely and relevant. For those whom might not understand the complexity of the question I would like to explain some background information. There are often numerous mutations found in a gene of interest that can alter the activity of the protein they encode. This activity change can range from little to no change all the way to a completely inactive protein (e.g., CYP2D6 enzyme in the liver). The mutations that abolish activity are usually referred to as “null” alleles. While there might be hundreds of mutations in a gene, we try to focus on those that are clinically relevant. Clinical relevance in this context refers the smaller set of mutations that can determine if a person is a Poor, Extensive, or Ultra-rapid metabolizer. In the case of CYP2D6, we know from years of research that it is fairly straight forward to identify Poor Metabolizers by selecting the most frequently occurring null alleles across all ethnic groups. Thus, a focused panel of alleles can be included in a diagnostic test providing easily 95-98% certainty that we have identified the patient correctly as a Poor Metabolizer. The range here (95%-98%) is a reflection of the differences in ethnic populations with regards to the prevalence of these targeted null alleles. Most important, this type of test can allow a medical practitioner to make a “yes or no” decision when it is well established that Poor Metabolizers are at risk of adverse events. The question of allele selection is less clear in the case of identifying the “Intermediate Metabolizers (IMs). The frequency of alleles that result in decreased enzyme activity (e.g., CYP2D6 *9, *10, *17, *29, *36, *41) depends to a large extent to the ethnic origin of the patient. What is very unclear at this time is when will this knowledge help to differentiate how we treat a patient who is Intermediate differently from one who is Extensive? This will probably have to be sorted out on a case by case or drug by drug basis.

Dear Dr. Murphy, Congratulations for your excelent presentation. My questions are: 1) Are all the pharmacogenetic tests cited in pag 32 already approved by FDA? 2) Besides Gentris and Renassaince, there are other manufacturers of pharmacogenetic tests? 3) Which pharmaceutical industries are already using pharmacogenetic tests in their phases I to III clinical trials? 4) Could you send me the references of the informations on slides 30, 31 and 38? Thanks for your attention. Best regards, Prof. Rosario D.C. Hirata, Ph.D. School of Pharmaceutical Sciences Sao Paulo University Av. prof. Lineu Prestes, 580 05508-900 Sao Paulo, SP Brazil
Sao Paulo, SP, Brazil

Michael Murphy: The tests shown on slide 32 is a sampling of tests offered under Good Laboratory Practice (GLP) which is a regulation under the Food and Drug Administration (21CFR58). These rules are similar to CLIA in that all tests must be thoroughly validated and lab personnel have to be trained and proficiency tested. While I am not at liberty to share exact client names, I can tell you that most of the top 20 global pharmaceutical companies (ranking by revenues) are utilizing pharmacogenomics during clinical trials. This utilization starts at Phase I trials to ensure that patients with altered metabolism are included in the trial. Metabolism is examined in Phase II and III trials and when possible genetic markers related to efficacy such as transporters, target, and receptors are included. Gentris is one of a handful of clinical pharmacogenomic testing providers in the United States. As more turnkey solutions are launched we fully expect that the large reference labs that service clinical trials will also look to provide similar services. At this point there is a large regulatory burden in providing this kind of testing and it is still considered a high complexity, esoteric test environment. Our company and others are looking to change this so that all patients, not just those in clinical trials, will have access to tests that can help prevent drug metabolism-related adverse events. I will be glad to forward the references you requested.

Are you aware of any clinical recommendations that advocate or offer guidance for use of pharmacogenomic testing in a clinical setting?
Pasadena, CA

Michael Murphy: I am not aware of any guidance related to the use of pharmacogenomics in a clinical setting. Perhaps a helpful starting point would be the guidance published on pharmacogenomics by the FDA in March 2005. While this document is related to the importance of pharmacogenomic data during clinical trials it does outline what constitutes a “valid biomarker” such as many of the genes that encode for drug metabolism enzymes. We would expect such markers to eventually migrate into clinical practice assuming that if they are important in the development of drugs they might also be important prior to prescribing.

A follow-up to my previous question regarding guidelines. Are there any clinical situations where pretherapeutic pharmacogenomic testing for drug selection is even indicated or recommended?
Pasadena, CA

Michael Murphy: Yes, there are a few drugs that do have an indication for prospective pharmacogenomic testing. The Eli Lilly drug Straterra® (atomoxitine) has a “highly recommended” indication to genotype patients for CYP2D6 prior to prescribing. Poor Metabolizers of CYP2D6 have an increased rate of adverse events that might be avoided if it were known prior to treatment. More recently the CNS drug Abilify® also has recommendations for CYP2D6 genotyping. Other examples include a recommendation (or stronger) to genotype patients for TPMT (thiopurine methyltransferase) prior to treatment with azathioprines and mercaptopurines. One out of 300 patients have mutations resulting in no TPMT and these patients can suffer life-threatening adverse events. Last, I believe there is now a recommendation to genotype patients who are going to be treated with Irinotecan for the UGT1A1 to prevent some of the potential life threatening neutropenia adverse events. We certainly expect more drugs to be “re-labeled” as we gather the definitive evidence needed to show that these tests can prevent adverse events by prospective genotyping.

Is there a laboratory that can test for the CYP3A gene?
York, PA

Michael Murphy: I can not speak for others but I do know that Gentris offers CYP3A4 and CYP3A5 genotyping for patients in clinical trials. The limitation here is that we don’t fully understand why some drugs which appear to be “3A” substrates are affected by mutations while others are not. This is another example where we need to study the situation on a case by case basis.

Are some of the detoxification reactions linked to genes in pharmacogenetics dependent on a reducing power of the cells ? Would - in some instances - this be a limiting factor in the pharmacological treatment of a given compound ? Ole W Bøe/Clin Chem Consultant
Lillehammer, Norway

Michael Murphy: I am not completely sure I understand the question but I will try to answer. Most metabolism occurs in the liver and to some extent the GI tract and the kidneys. Genetic polymorphisms (mutations) in these genes does affect their activity and can be used to predict possible toxicity. Pharmacogenomics/genetics is however only one component to the final outcome. There is a large environmental component that must be taken into consideration including concomitant drug use, diet, smoking, and alcohol use to name a few. It is possible for someone who was genotyped as “Extensive” or “Normal” to phenotypically demonstrate toxicity simply because they were on one or more drugs that went through the same metabolic pathway (a phenomenon called “phenocopying”).

Like most technologies in healthcare, I suspect the adoption of PGx testing will come down to issues of cost and efficacy. In you slide titled "Commercial Incentives" you note that an "obvious savings to healthcare" could drive acceptance of PGx applications, but studies indicate that you must do 12 genotypes to identify one Caucasian poor or ultra-rapid TCA metabolizer. Does this ratio confer obvious savings, and is it consistent with PGx testing for other therapeutic compounds? What is the cutoff (ratio) where these savings start accruing?
Vienna, Austria

Michael Murphy: 7) I agree, we must always consider that there is a cost benefit analysis that will drive adoption and justification for testing all patients for a particular drug or class of drugs. In order to make these types of analysis we must collect morbidity (and mortality) costs associated with continuing to do “trial and error” drug treatment. In the case of tricyclic antidepressants I am not aware of the study you cited but I do know that some of the severe adverse events for Poor Metabolizers are seizure, stroke, and death. Adding up the cost for such treatment when we are talking about 6-10% of the treated population should be significant. The next question is what is the cost of testing all patients in this situation. I will certainly be more difficult to arrive at a reasonable cost-benefit when the test cost upwards of $500 per patient. Dr. Peter Wedlund at the University of Kentucky published several papers in the past related to the cost benefit of treating patients with CNS drugs. By Dr. Wedlund’s estimates a test that cost $150 per patient can save $3000 in morbidity costs for patients being treated for depression. This data may have been updated as this was a paper I read several years ago. In the end the cost benefit will have to be studied for all new pharmacogenomic tests but I believe there are some obvious ones we can start with to get this into clinical practice.

Are you aware of any clinical recommendations that advocate or offer guidance for use of pharmacogenomic testing in a clinical setting?
Pasadena, CA

AACC Moderator: In response to Pasadena, the National Academy of Clinical Biochemistry is currently drafting “Guidelines and Recommendations for Laboratory Analysis and Application of Pharmacogenetics to Clinical Practice.” You can go to the NACB web site (http://www.nacb.org/lmpg/2006_lmpg_pgx.pdf) to see the progress of this project.

Are there specific or "rule of thumb" prescribing suggestions for PM who need to be on a specific drug? I know this also depends on other drugs used ("inhibitors"). This is what the doc will ask when given the PGX test results.
Helena, MT

Michael Murphy: This is another question we get often when we talk about pharmacogenomics to physicians and clinical diagnostic laboratories. There is no simple rule of thumb because it may depend on whether there are alternative routes of metabolism. When a drug is mostly metabolized by a single pathway we can provide some guidance on how to treat Poor Metabolizers (PMs). In general, the dose we would give to a PM would be 10-15 fold lower than that given to a normal or Extensive metabolizer. For most previously approved drugs given orally as a tablet this might be difficult. In my opinion, the doctors are really looking for a qualitative “Yes” or “No” answer. For drugs that cause severe adverse reactions in PMs I believe the physician will simply want to avoid prescribing (when possible) and try an alternative therapy. Last, there is a need for us to put together some kind of translation table so that the results of pharmacogenomic testing can be useful advice for all approved drugs. I believe that in the meantime we can at a minimum identify those at the extremes of metabolism (PMs and Ultra-rapid metabolizers) and either alter the starting dose or enhance our surveillance to try and mitigate known adverse reactions.

Is it enough to classify patients as poor, intermediate, extensive, or ultrarapid metabolisers, or would more quantitative descriptors (say, for example, a "75% metaboliser") be helpful in a clincial setting?
San Juan, P.R.

Michael Murphy: The problem here, whether we use a more quantitative reporting result (like 75% metabolizer), or the traditional “predicted phenotype” is that it does not usually translate into a treatment decision for the doctor. As an example, it is difficult to determine as a set rule whether there is really a difference in a way we would treat an Intermediate metabolizer versus an Extensive. There are a few examples in the literature where some have demonstrated a “gene dosage” dependent relationship but these are the exception, not the rule. I personally believe that we can accelerate the adoption of pharmacogenomic testing and utilization by first using it to help avoid all together dangerous drugs for PMs and UMs.

What is the cost of Phnotyping vs Genotyping?
Berkeley, CA

Michael Murphy: In general, the cost of phenotyping are much higher than genotyping. This is largely due to the fact that there are clinical costs and bioanalytical costs associated with phenotyping. The patient is usually given a non-efficacious dose of a drug metabolized principally by a single enzyme. Even Poor Metabolizers can have adverse events from simple drugs like dextromethorphan, the drug typically used for CYP2D6 phenotyping. This mandates keeping patients in the clinic for some period of observation. Usually urine or blood is collected 4-8 hours later and the sample sent off the bioanalytical lab. Genotyping only requires collecting only one EDTA (purple top) of blood. This can be down in a few minutes and the cost of the test is similar to bioanalytical testing.