March 2008
Feature Story

Directors of molecular diagnostics laboratories are caught in a tug-of-war now over the new pharmacogenomics tests, such as those for determining warfarin sensitivity and predicting tamoxifen response. On one side are people who urge that these tests be adopted; among them are clinicians who believe the tests can provide greater safety or effectiveness, or both, for patients. They base their belief on biologic plausibility and evidence showing that genotype is associated statistically with better or worse outcomes. Another force for adoption is the general excitement in the media and, in consequence, among the lay public, largely generated by early adop­ters in the molecular diagnostics profession and by vendors, often under the ru­bric of personalized medicine. In addition, Food and Drug Administration label changes for warfarin and other drugs have spurred the push for adoption.

On the other end of the rope, exerting their strength to moderate the pace of adoption, are groups who find the clinical data inadequate to warrant general dissemination of pharmacogenomics, or PGx, testing, even for warfarin sensitivity, the most advanced application. (One laboratorian calls PGx testing for warfarin "the prototype or poster child for PGx.") These groups point out that no one has shown that using PGx test results actually brings about the predicted improvement in safety or response. Third-party payers are in this camp. But a recent American College of Medical Genetics report also takes this position (see "Statistical associations don't prove clinical utility"). To make matters more thorny, the public health implications are huge: Perhaps 1 million people initiate warfarin each year, and the consequences of misdosing can be severe.

Of course, molecular laboratorians themselves are not neutral observers. Many are excited by these powerful new molecular tools and want to be among the first to offer them as a clinical service. Over and over, they told CAP TODAY that they were offering PGx testing for warfarin, or would like to do so, even though there is not yet convincing evidence from randomized clinical trials to show significantly improved outcomes.

Ted E. Schutzbank, PhD, D(ABMM), technical director of CompuNet Clinical Laboratories in Dayton, Ohio, describes himself as "very excited and very frustrated at the same time" and says, of molecular testing in general, "No one wants to reimburse these tests." Referring specifically to PGx testing for warfarin, Dr. Schutzbank says, "There are two or three studies that definitely illustrate the utility of cytochrome P450 [CYP450] testing in warfarin sensitivity. Unfortunately, FDA did not go far enough when they relabeled warfarin. We were hoping they would recommend that PGx testing be implemented for warfarin therapy, but they only said that it could be useful. That was disappointing to many of us in the laboratory and in clinical medicine who understand that this testing could actually save lives." Asked about the evidence that PGx-guided warfarin therapy will save lives, Dr. Schutzbank says: "There is a large number of people who carry mutations in the CYP450 genes that affect warfarin metabolism. If you have the PGx data right up front, that will lead to faster appropriate dosing and fewer sequelae. The consensus of opinion falls in favor of using warfarin PGx to reach stable or optimum dosing much faster."

CompuNet Clinical Laboratories has not had many calls for PGx testing from clinicians, and thus hasn't brought it up yet. Dr. Schutzbank says: "A lot of physician education will be required-what the test can do for them and its limitations. But first we need to convince third-party payers that it is useful. That will not happen until large outcomes trials are completed." The insurance industry, he says, will not be convinced by the small studies that exist now.

ARUP Laboratories has offered PGx testing for warfarin for more than a year, says Elaine Lyon, PhD, medical director of molecular genetics and co-medical director of pharmacogenomics. The FDA's relabeling of warfarin was a "driving force" for ARUP. "I know ACMG has done an in-depth review and found that the clinical utility is still uncertain," Dr. Lyon says. "But we made a decision to go forward based on FDA relabeling and enough literature that we felt we could give an interpretation for the three variants that FDA recommends.

"An association has been shown between genotype and warfarin sensitivity," she continues. Less strong is that the time to get a stable international normalized ratio could be reduced-"that may not have been completely shown yet," she says. Dr. Lyon, who is associate professor of pathology at the University of Utah, says additional studies are needed.

Speaking of PGx testing in general, Barbara A. Zehnbauer, PhD, FACMG, says: "There are very few studies that stratify patients based on genotype, then verify whether they improved outcome or response based on those genotypes, which is how you want to use these tests clinically. Mostly there are retrospective studies classifying outcome or risk of adverse events stratified by genotype." Evidence for the value of PGx will not be complete until those prospective outcomes studies are done, says Dr. Zehnbauer, director of the molecular diagnostic laboratory and professor of pathology and immunology, Washington University School of Medicine. Even so, Dr. Zehnbauer says her laboratory has just started offering PGx testing for warfarin. "Partly it is a service," she explains. "One of our missions is supporting physicians in what they need to care for their patients. If we don't offer it, other labs will."

Federico A. Monzon, MD, medical director of molecular diagnostics at The Methodist Hospital, Houston, is "a big proponent" of PGx testing, saying "it will help us better manage our patients." But he concedes there is still work to be done. "We are moving in the right direction with studies that have been published, especially with warfarin, where most of the demand for clinical use of pharmacogenomics will be in the next few years," he says. Adjusting warfarin dose by using an algorithm that incorporates PGx data plus clinical factors "gives people better dose estimates," Dr. Monzon says. Yet Meth­o­dist is not using PGx testing for warfarin at this point because its existing inpatient anticoagulation service is performing well. Pharmacy manages warfarin dosing for about half of inpatients and is doing an excellent job, Dr. Monzon says. "Using an algorithm that doesn't include pharmacogenomics, our pharmacists are getting results equivalent to other places that are using pharmacogenomics," he says, and adds: "We need to see if by adding pharmacogenomics, we can improve their performance before we try to implement it."

One of the first laboratories in the United States to offer PGx testing in the clinical arena was PGXL Laboratories in Louisville, Ky. "One of our main drivers has been getting word out to promote how pharmacogenomics is clinically useful and how this complex data can be translated into actionable information for the clinician," says Kristen Reyn­olds, PhD, vice president of laboratory operations for PGXL Laboratories and associate clinical professor of pathology and laboratory medicine at the University of Louisville School of Medicine. Her lab is "aggressive in getting that word out," Dr. Reyn­olds says, and has hired a sales force to accomplish this in the Louisville region.

Of the evidence for offering this test, Dr. Reynolds, too, mentions the FDA. "In updating the label for Coumadin, FDA felt there was significant evidence that not considering pharmacogenomics during warfarin therapy poses the greater risk from inappropriate dosing and a greater risk of unstable INRs," she says. "There is sufficient evidence to suggest that patients who have specific variant alleles in the CYP450 2C9 gene [CYP2C9] and/or the vitamin K epoxide reductase [VKORC1] gene may need lower doses." Though Dr. Reynolds acknowledges there may not be enough data from randomized clinical trials to say that initiating doses according to PGx results will reduce the number of bleeding events, she says: "We, and certainly Larry Lesko at FDA, refer to INR stability as a surrogate for clinical outcome because it is used by clinicians. There is a strong association between out-of-range INRs and adverse outcomes."

Brian F. Gage, MD, MSc, associate professor of medicine, Washington University, and medical director, Barnes-Jewish Hospital Anticoagulation Service, has for several years led studies of the association between variants in CYP2C9 and final stable warfarin dose. He has developed one of the most comprehensive algorithms for dosing warfarin that takes into account PGx results (www.warfarindosing.org). Dr. Gage believes that PGx testing will prove clinically beneficial but acknowledges that the evidence is not yet strong enough for these tests to be implemented widely. "There is certainly diversity of opinion among experts," he says. "In my own practice we are using pharmacogenomics-guided therapy as part of a clinical study. Outside of that study, the benefits are sufficient to justify pharmacogenomics testing in patients who would be at high risk of hemorrhage if they had a persistently elevated INR"-elderly patients taking anti­platelet therapy and patients with a history of hemorrhage.

Gregory J. Tsongalis, PhD, HCLD, FACB, director of molecular pathology, co-director of the pharmacogenomics program, and associate professor of pathology, Dartmouth-Hitchcock Medical Center, agrees more clinical trials data are needed. "We have a good handle on what the effectiveness will be of having pharmacogenomics testing available, both from medical and financial perspectives. We anticipate big savings from eliminating costs associated with adverse effects of overcoagulation and return visits to the ER or clinic. But there are very few data out there that actually show what that savings is," he says. Like Dr. Schutzbank, Dr. Tsongalis believes a lot of physician education is needed. "Part of all this is how quickly physicians are willing to use the assay," he says. "I am not sure this is as much of a slam dunk as people thought it would be."

Karen Weck, MD, associate professor of pathology and laboratory medicine, director of the molecular genetics laboratory, and associate director of the UNC Institute for Pharmacogenomics and Individualized Therapy, University of North Carolina, Chapel Hill, says PGx testing "has the potential to save lives" and "huge potential for better understanding drug response for individuals." But as with any new development in medicine, she says, the desire to move forward quickly has to be balanced against the need to evaluate the testing's validity and utility. "That is where we are right now in this field," Dr. Weck says. "Pharmacogenomics has real clinical potential, but it has not been demonstrated yet. That is why it is important to do the clinical trials." While randomized trials are the ideal way to show clinical utility, Dr. Weck also points out that physician uptake drives clinical practice. "If clinicians are familiar with literature showing the effect of poor metabolism of warfarin, they may choose to use that information to dose people differently, even without a large trial that reaches significance." Some anticoagulation physicians would like to see the laboratory offer the assay even without a clinical trial, she says, particularly for patients who are difficult to dose.

Reimbursement is another reason for demonstrating clinical validity and utility, Dr. Weck adds, because Medicare and other third-party payers will be reluctant to approve payment without the data. She and her colleagues are embark­ing on a single-center study evaluating clinical utility and cost-efficacy of PGx-guided warfarin dosing among 500 patients. All patients will be dosed according to the Gage algorithm, with half of the patients having PGx data added in and the other half not, a design that will isolate any effect of genotype. Dr. Weck plans to have PGx data available by the second warfarin dose. Trial endpoints include number of clinic visits and INRs needed to get to target dose, as well as adverse events, physician uptake, and cost. "We have delayed offering pharmacogenomic testing for warfarin until the trial starts," Dr. Weck says. "Then all patients can get it as part of the trial." Conducting the trial will generate an infrastructure for implementing the test.

Criteria for the kind of data needed to prove clinical utility for PGx testing are simple but rigorous. Retrospective studies showing an association between genotype and INR demonstrate clinical validity, but they don't demonstrate clinical utility, says Margaret Piper, PhD, MPH, director of genomic resources at the Technical Evaluation Center, Blue Cross/Blue Shield Association. Clinical utility requires data from prospective randomized clinical trials showing that PGx-guided management improves outcomes, either hard endpoints or previously established surrogates. "For warfarin, there is evidence that links INR to bleeding, so you can use INR as a surrogate marker," Dr. Piper says. "But you can't use dose as a surrogate marker because dose is not an outcome. Genotype accounts for a significant proportion of final stable dose in multiple regression analysis," Dr. Piper continues, "but is that prediction enough to have a beneficial incremental impact on bleeding episodes?"

Incremental benefit is the other major issue in gathering data. "Genotyping for warfarin sensitivity is added to INR; it is not a replacement," Dr. Piper says. "So we need to know what it adds to the current situation." In a recent publication of a small randomized clinical trial, "the control arm was so well controlled by usual methods that genotyping added nothing significant," she says. (See "Statistical associations don't prove clinical utility").

Even while some investigators seek evidence for clinical benefit of variant alleles known to affect INRs, others are already looking to expand PGx testing for warfarin to include other P450 genes, such as CYP2C19, CYP3A4, CYP4F2, and CYP1A2. Ronald Przy­godzki, MD, associate director of the Genomic Medicine Office of Research & Development at the Department of Veterans Affairs, says current alleles "are a tremendous step ahead, but not nearly enough. They will catch a good number of people, but not everyone.

"Warfarin is a racemic mixture, with the S- and R-isomers," he continues. Though the R-isomer of warfarin is less potent, it should not be less considered in the metabolic pathways, he says. Of interest, the R-isomer undergoes metabolism through the aforementioned CYP pathways, "some of which may be secondarily affected by numerous drugs and extraneous habits like smoking."

At the last Association for Molecular Pathology annual meeting, Paul G. Rothberg, PhD, professor of pathology and director of the section of molecular diagnosis at the University of Rochester Medical Center, organized a workshop on PGx of warfarin therapy. "We asked the membership what they would be interested in hearing about, and the area of cardiac pharmacogen­omics came out on top," Dr. Rothberg told CAP TODAY. As a potential high-volume test, PGx for warfarin sensitivity could have a big impact, Dr. Rothberg notes. "There are few mass-market molecular genetic tests and this could be one of them," he says. "It could be right below cystic fibrosis screening of pregnant women."

The ACMG issued a report on the potential impact [of PGx for warfarin dosing], Dr. Rothberg says, "and it did not come to the conclusion that it is proven effective, but that is a reasonable extrapolation of current data." Like Dr. Tsongalis, Dr. Rothberg is waiting for demand to build and figuring out what type of platform to bring in. Demand is not high in his area. "I have talked about it with some hematologists and they said to get ready-there will be demand." Clinicians are perhaps waiting for firm evidence. "I think they want to see a paper showing the clinical utility of the test and what fraction of complications can be avoided. Right now estimates are based on extrapolation, not on a randomized trial." Extrapolations can estimate the fraction of bleeding that arises from genetic causes. "Dr. Gage has made those kind of estimates," Dr. Rothberg says, "but we still need large clinical trials to demonstrate how frequently the use of genetic information prevents adverse outcomes."

Dr. Gage delivered the main talk at the AMP workshop. He noted that major bleeding is much more common with an INR greater than or equal to 4.0. S-warfarin, the active enantiomer, is metabolized by CYP2C9. In a retrospective cohort study, time to stable dose and time to first serious bleed were adversely affected in those who carried CYP2C9 variants *2 or *3 (Higashi MK, et al. JAMA. 2002; 287: 1690- 1698). Both variants reduce warfarin breakdown and render a person more sensitive, thus lowering the final stable maintenance dose (Gage BF, et al. Use of pharmacogenetics and clinical factors to predict the therapeutic dose of warfarin. Clin Pharmacol Ther. 2008. In press). In a recent study, presence of a *2 or *3 variant was associated with a two- to threefold increased risk of major bleeding, though the association before stabilization of warfarin dose was not significant (Limdi NA, et al. Clin Pharmacol Ther. 2008; 83: 312- 321). "Based on genotyping, we ought to be able to prescribe a dose of warfarin that would result in a therapeutic INR," Dr. Gage concluded.

Dr. Gage and colleagues have been working on a dosing algorithm for warfarin. They found that eight variables could account for 79 percent of the variation in therapeutic warfarin dose; three were gene­tic variables-CYP2C9*2 and *3 and VKORC1 haplo­type A (VKORC1 is the target for warfarin inhibition of coagulation). Others were smoking, first and second warfarin dose, and the INR on day three of treatment (Millican EA, et al. Blood. 2007; 110: 1511- 1515).

This is called the INR3 algorithm because it is intended to predict the revised dose that should be given after the day three INR has been obtained. For most warfarin PGx algorithms, genotype would be needed stat to affect initial dose (the exception is elective orthopedic or valve surgery). However, Dr. Gage says, "Genotyping does not need to be done stat. It's safe to begin warfarin with clinical dosing for the first day or two, then change to the pharmacogenomics-predicted dose once genotype is available." A 24-hour turnaround time for PGx results would be adequate for this purpose.

In his talk, Dr. Gage showed an Israeli randomized clinical trial evaluating PGx-guided warfarin dosing among 191 patients, using only variants of the CYP2C9 gene, that found a shorter time to therapeutic INR when genetic information was incorporated (Caraco Y, et al. Clin Phar­ma­col Ther. 2008; 83: 460- 470). A second, 200-patient U.S. randomized trial, which incorporated both CYP2C9 and VKORC1 genotypes, was "primarily a negative study," Dr. Gage noted. PGx-guided doses approximated stable doses more accurately. "However," the investigators reported, "percent out-of-range INRs, the primary end point, did not differ significantly between arms" (Anderson JL, et al. Circulation. 2007; 116: 2563- 2570). Dr. Gage has applied to the National Institutes of Health for funding for a multicenter randomized trial that would enroll about 2,000 patients initiating warfarin therapy. Its primary endpoint would be time in target INR range at 30 days. Secondary endpoints would be clinical adverse events at 90 days. Patients would be assigned to one of three arms: 1) PGx-guided dosing by the www. warfarin dosing. org algorithms for both initial dose and dose revision; 2) an algorithm that incorporates only clinical information (Lenzini PA, et al. Ann Pharmacother. 2007; 41: 1798-1 804); and 3) standard of care (to be determined).

"There is a sufficiently strong link between INR control and adverse events that a significant difference in INR control would justify wide adoption of these tests," Dr. Gage says. "Whether pharmacogenomics would be cost-effective depends on what genotyping would cost and the estimated clinical and economic benefit from reduction in adverse events. Those questions may end up unresolved after this trial." An estimate of cost-efficacy in the ACMG review was not optimistic (See "Statistical associations don't prove clinical utility").

Jeffrey L. Anderson, MD, FACC, FAHA, MACP, associate chief of cardiology at Intermountain Medical Center, was the principal investigator on the U.S. randomized trial with a negative outcome. Dr. Anderson, professor of medicine at the University of Utah School of Medicine, calls the trial "an interesting experience," one from which they learned a lot. One thing he learned: "Our inpatient anticoagulation management service does a very good job. When patients are being rapidly adjusted, that is hard to beat. I think that was one factor in not demonstrating benefit in our primary endpoint."

Dr. Anderson and colleagues are proceeding with additional studies. They intend to collaborate with Dr. Gage in the proposed trial and are "pushing ahead" with some QI initiatives within Intermountain. Since inpatient management was so effective, future studies will move to the outpatient setting. Also, they will power future studies to detect a 25 percent or even 20 percent reduction in out-of-range INRs, rather than the 50 percent they assumed in their first trial. "That will take a couple of thousand patients," Dr. Anderson says. It is important to show that PGx reduces out-of-range INRs, he says: "I think we will need to show that before we can sell this test widely."

In Dr. Anderson's view, warfarin dose adjustment needs to begin with the first dose. "It is important to get the test result back the first day and start out with the best dose so you can get outliers," he says. The rapid turnaround time of the platform they use, one being developed by Idaho Technology, is crucial to this goal. "All you need is a buccal swab," Dr. Anderson says. "You get a result back in an hour and plug it into the spreadsheet that spits out the dose." Their anticoagulation management service prefers that to having to track down outpatients with extreme INRs and asking them to come in to get adjusted.

Like Aristotle's theory of tragedy, laboratorians imple­menting PGx testing focus on the beginning-platform selection; the middle-interpretation; and the end-reporting.

Dr. Gage's group uses an in-house warfarin PGx test. He notes that the first FDA-cleared platform for this application was Nano­sphere's Verigene system, with Autogenomics' assay receiving clearance earlier this year. "We have experience with the Autogenomics and Third Wave platforms," Dr. Gage says. "Both are extremely accurate and fast enough to be used clinically." Dr. Tsongalis evaluated the Third Wave Invader 96-well plate assay and the Nanosphere Verigene assay. Turnaround time was 3.5 and 1.5 hours, respectively. He found the Invader assay sensitive and robust and the Verigene system "as turnkey as they get." In principle, Invader's batch format offers greater cost-efficacy, but that advantage would depend on how clinicians use the assay. If they want PGx data for the first warfarin dose, Dr. Tsongalis says, "I would be much more apt to use an assay like Nanosphere."

Dr. Zehnbauer will be using Invader. "It hasn't gone through FDA review yet," she acknowledges, "but we are using the Third Wave platform for another assay. It is very easy for technologists to operate." Dr. Lyon has been using Third Wave but is considering switching to Idaho Technology, which uses melt curve analysis and has a turnaround time of less than one hour.

After evaluating VKORC1 and CYP2C9 genotyping assays from Para­gon­Dx (formerly Gentris), Third Wave, and Auto­ge­nom­ics, Dr. Weck concluded that all three methods were good. In a comparison study of 20 patients on long-term stable warfarin therapy, all three assays had 100 percent accuracy for both genes. She adopted the Autogenomics platform, "because it is automated and easy to use," she says. It takes 10.6 hours for a six-sample run, which meets her goal of 24-hour TAT.

Dr. Tsongalis emphasizes the complexity of PGx interpretation and reporting. "I don't think we should go away from this conference thinking that it's all in place from the lab viewpoint," he said in his talk at the AMP workshop. "Warfarin is a good example because you have to take into account other factors than just genotyping," he says. "For pharmacogenomics tests to catch on, they have to be as simple for clinicians as possible." To accomplish this, Dr. Tsongalis has teamed up with clinical pharmacologists in his hospital; laboratory reports will include an interpretation and dose recommendation based on genotype. When clinicians get the reports, they will not have to do anything further. In addition, he says, "Asking MDs with a pharmacology background to put the result into clinical context gets around lab people not being able to make dose recommendations."

Dr. Reynolds and her colleagues imagined clinicians asking, "What am I going to do with this information?" As opposed to reporting the following: "This patient has a CYP2C9 deficiency," they say: "Because this patient has a metabolic deficiency, they are more likely to take increased time to reach steady state, which affects interpretation of a stable INR." They provide the genotype but also a dose estimate example from their proprietary software, PerMIT (Personalized Medicine Interface Tool), a dynamic clinical decision support tool (Zhu Y, et al. Clin Chem. 2007; 53: 1199- 1205). "Perhaps more important," Dr. Reynolds says, "it gives the clinician information about when to trust that an INR is actually stable." To avoid giving dose recommendations, they say they are supplying an example of how information can be used in the specific case of that physician's patient.

In addition to reporting wheth­er a patient is homozygous or hetero­zygous for CYP2C9 and VKORC1, Dr. Zehnbauer's report will give a phenotype, such as "extensive metabolizer" or "poor metabolizer" and provide an interpretation that says what kind of impairment in warfarin clearance or response might result. "While we don't make dosing recommendations," Dr. Zehnbauer says, "we do have a state­ment saying the clinician may wish to consider altering dosing for this patient. And we will reference Web sites that have dosing algorithms that incorporate genotype." Clinicians are used to using algorithms to take into account BMI, gender, age, and concomitant medications, Dr. Zehnbauer says. "Genotype is just another variable with fractional incremental risk."

After PGx testing for determining sensitivity to warfarin, the next most important PGx application could be to predict response to tamoxifen. Accumulating evidence shows that CYP2D6 metabolizes tamoxifen, a first-line therapy for ER-positive breast cancer, to its active metabolite, endoxifen. "There is substantial literature indicating that women with poor-metabolizing CYP2D6 variant alleles have a poor response to tamoxifen," says Dr. Weck. "In postmenopausal women we can use aromatase inhibitors as alternative therapy," she says. "But in premenopausal women there is no really good alternative." UNC is initiating a clinical trial to investigate whether increasing tamoxifen dose from the typical 20 mg/day to 40 mg/day can overcome the intermediate-metabolizer phenotype (for example, in women who are heterozygous for a poor metabolizer variant). They will measure a surrogate outcome: endoxifen levels in plasma.

PGx testing for psychiatric drugs is further from clinical practicality. Relabeling of atomoxetine (Strattera), the only non-stimulant medication licensed for the treatment of attention deficit hyperactivity disorder, is based on altered metabolism by CYP2D variant alleles. "No study has demonstrated that giving an altered dose based on pharmacogenomics reduces adverse effects," says Dr. Reynolds, "but we believe that this is the way to go."

Use of PGx to identify variant alleles of CYP450 genes that metabolize selective serotonin reuptake inhibitors, or SSRIs, also lacks evidence, according to a review by the Evaluation of Genomic Applications in Practice and Prevention (EGAPP) Working Group (www. ­egappreviews.org). EGAPP "found insufficient evidence to support a recommendation for or against use of CYP450 testing in adults beginning [SSRIs]." The report notes that "CYP450 genotypes are not consistently associated with the patient outcomes of interest" and "discourages use of CYP450 testing for patients beginning SSRI treatment until further clinical trials are completed."

"Use of PGx testing [for SSRIs] is not now recommended, but that doesn't mean it's not something we should move forward," the VA's Dr. Przygodzki says. "There are many polymorphic markers in 2D6 and existing studies may only have looked at a handful of them," he says. "We need a good full-bore examination of all polymorphic markers. I'll admit-is this valid? I can't say because I haven't done the study."

Having PGx tests with short turnaround times to guide warfarin dosing will stimulate the development of other fast PGx tests, Dr. Tsongalis predicts. "It comes down to clinical need," he says. "Warfarin is the first therapy for which we need the genotype when the drug is prescribed. It's not a good example of cost savings based on a drug, but newer therapeutics are quite expensive." He can envision patients going from the laboratory to the pharmacy and, before they get to the pharmacy, a result is transmitted electronically and the pharmacist dispenses the right drug or right dosage. "It would not surprise me if eventually we would see pharmacogenomics be part of a pay-for-performance system," he says.

Related Links

• Statistical associations don’t prove clinical utility