College of American Pathologists
Printable Version

  Warfarin pharmacogenetics a waiting game


CAP Today




<em><strong>For warfarin pharmacogenetics testing</strong>, much rests on the results of the GIFT and COAG multicentered trials. Below, at Washington University, Dr. Charles Eby (left), medical director of the central labs for the trials, and Dr. Brian Gage, principal investigator for GIFT.</em>
For warfarin pharmacogenetics testing, much rests on the results of the GIFT and COAG multicentered trials. Above, at Washington University, Dr. Charles Eby (left), medical director of the central labs for the trials, and Dr. Brian Gage, principal investigator for GIFT.

May 2012
Feature Story

Karen Titus

Like Miller Lite in a Munich beer garden, pharmacogenetic testing to guide initial warfarin dosing is a tough sell. Who wants to persuade unwilling customers to pay for an unproven, unreimbursed test that even the experts pass up?

But if few are clamoring to use PGx for warfarin dosing, no one is placing it on the dustheap, either. The potential rewards are too enticing. That’s why even one of the more discouraged-sounding leaders in the field—say hello to Alan Wu, PhD—refuses to give up. “I always want to have the subject in the limelight, so we can have a discussion on it. If there’s no discussion, that means it’s dead. And I don’t think that’s the case,” says Dr. Wu, professor, laboratory medicine, University of California, San Francisco, and laboratory director, Clinical Chemistry Laboratory, San Francisco General Hospital.

Still, Dr. Wu can be forgiven for sounding blue. No professional groups have endorsed genotype testing for warfarin. Reimbursement discussions are virtually nonexistent. That, in turn, has squeezed vendors—in recent years, the number of companies that provide reagents and platforms for warfarin pharmacogenetic testing has dropped by half, to two.

Even at his own institution, Dr. Wu has been less than persuasive. Despite SFGH’s participation in promising studies, his clinical colleagues remain skeptical. “It’s been a disappointment for me,” he says. “I continue to write, continue to do studies, and hope that eventually we’ll see utilization. But right now, it’s not happening.

“We’ve tried everything,” he continues. “We’ve tried the educational route; we tried doing research studies with them; we tried to offer testing at reduced prices. And it just hasn’t come. It’s been frustrating.”

It’s also puzzling, perhaps, given that so much can go wrong with warfarin. This is not a drug for slackers.

Yet a certain languor persists. Charles Eby, MD, professor of pathology and immunology, medicine, Washington University School of Medicine, St. Louis, sees some foot-dragging among clinicians. Leaders in the field decry the lack of clinical evidence to support use of PGx testing, while general practitioners are still hesitant about genetic testing. “For many, many, many physicians, it’s a novelty,” says Dr. Eby, who chairs the CAP Coagulation Resource Committee. “It’s not part of their practice tradition.”

But using warfarin sans genotyping certainly is. Warfarin was essentially the sole anticoagulant from the mid-1950s until 2010, when the FDA approved the first direct thrombin inhibitor, dabigatran (Pradaxa). Some 50-plus years of institutional experience—overlaid by practitioners’ own medical memories—is a big tide to turn.

Dr. Eby draws an interesting comparison to gambling. “People who gamble remember their wins and forget their losses,” he says. “And I think people who prescribe warfarin maybe forget their bad outcomes: ‘Oh, that patient had a GI bleed. Well, that happens with warfarin, it hadn’t anything to do with dosing’—other than the fact that the INR was 10,” says Dr. Eby.

“I think there is kind of a built-in complacency with warfarin in medical practice,” he continues, even though physicians know it’s the second most common drug associated with drug-related complications or ED visits. “But we just kind of accept that. And if you buy into that, and you buy into your own selective recall of your experience dosing it, then you’re likely to say, ‘I don’t need this genetic information.’”

The fortunes of warfarin pharmacogenetic testing have ebbed and flow-ed, sometimes simultaneously. Case in point: the Centers for Medicare and Medicaid Services’ decision, in August 2009, not to reimburse testing as part of routine patient care. While some might be nonplussed by this (after all, the following year the FDA relabeled warfarin to recommend genotyping prior to initial dosing), Dr. Eby sees this as a positive step, since the agency will pay for warfarin genotyping if it’s done as part of a clinical trial.

Even advocates of warfarin pharmacogenetic testing agree that the clinical evidence needs to be stronger. Data to support its use have, like a Cubs World Series title, been scant.

Several ongoing studies should change matters, including the Genetics Informatics Trial (GIFT) of Warfarin to Prevent DVT, and the Clarification of Optimal Anticoagulation Through Genetics (COAG) trial. A third study, the EU-PACT (EUropean Pharmacogenetics of Anticoagulant Therapy—Warfarin), is also looking at patients with atrial fibrillation and DVT who are started on warfarin. All compare use of a dosing algorithm that uses warfarin genetic information in addition to conventional factors versus one that uses only clinical and demographic information.

The GIFT trial is NIH-funded and is looking at patients undergoing hip/knee replacement surgery who take warfarin to prophylax against DVT. Outcome in this study includes INR time within therapeutic range and bleeding and thrombotic complications (based on Doppler ultrasounds at four to six weeks postoperatively), according to Dr. Eby, whose colleague at Washington University, Brian Gage, MD, is the study’s principal investigator.

The COAG trial, also NIH-funded, is looking at patients with atrial fibrillation or venous thromboembolic events, as well as those who’ve had orthopedic joint replacement, who receive warfarin prophylactically. Outcome is INR time within therapeutic range after 30 days, says Dr. Eby, who is the medical director of the central laboratories for both trials. The trial is scheduled to complete enrollment at the end of 2012; the GIFT trial has three years of recruitment ahead and the aim is to recruit a fifth site.

In the meantime, several smaller developments may get momentum rolling.

Julie Johnson, PharmD, draws attention to a recent study by a group at the University of Utah (Anderson JL, et al. Circulation. 2012;125:1997–2005), which secondarily tested the relationship between genotype-guided dosing and standard care. Those in the first group had fewer patients out of range (both lower and higher) in terms of INR and significantly lower numbers of DVTs, bleeding events, and death. They also showed a higher percent of time in therapeutic range, says Dr. Johnson, distinguished professor of pharmacy and medicine, University of Florida, where she’s also director of the UF Center for Pharmacogenomics and the UF&Shands Personalized Medicine Program.

Dr. Johnson calls it a “very, very promising study, the strongest we have in terms of pointing to the potential clinical benefit.” It may also “push some clinicians to go ahead and implement pharmacogenetic testing,” even before results from the COAG and GIFT trials are in.

An earlier study by the International Warfarin Pharmacogenetics Consortium, published in the New England Journal of Medicine (2009;360:753–764), defined a dosing algorithm for a large multiethnic, multiracial population, using clinical and genetic information. The study suggested that using an algorithm with pharmacogenetic data allows clinicians to more precisely pick the warfarin dose patients will need long term.

Trials primarily focus on two genes, CYP2C9 (which affects warfarin metabolism) and VKORC1 (which determines warfarin sensitivity). Like the Founding Fathers, these two genes are well known, well-studied, and looked to for guidance. But they don’t tell the entire story.

While both are strongly predictive of warfarin dose requirements, the search for additional genes continues.

One study, led by the pharmacogenetics consortium, describes a novel variant that explains some additional variability in African-Americans, says Dr. Johnson, who was involved in the study. The researchers presented an abstract (No. 15518) at the American Heart Association meeting earlier this year and plan to submit a manuscript for publication this spring, says Dr. Johnson, who heads the consortium. The consortium is also in the early stages of looking at genetic data across racial-ethnic groups. The ongoing large trials might also help in the search for other polymorphisms, since researchers will archive DNA for use in future studies.

Of the promising genes being looked at, however, none explain anywhere near the variability that CYP2C9 and VKORC1 do. These are incremental additions.

Most studies have, until now, looked primarily at Caucasians, says Karen Weck, MD, professor of pathology and laboratory medicine and genetics, and director, molecular genetics, University of North Carolina at Chapel Hill. Currently known SNPs in VKORC1 and CYP2C9 are common in Caucasians, but less so in other groups. “And even in the IWPC algorithm for warfarin dosing, which included multiple ethnic groups, African-Americans were under-represented,” says Dr. Weck, who is also associate director, Institute for Pharmacogenetics and Individualized Therapy at UNC-Chapel Hill.

While the same VKORC1 and CYP-2C9 SNPs are important in African-Americans, other genetic factors have statistical significance in this group as well. “It’s an important point, because African-Americans tend to be more resistant to warfarin, and the common SNPs in VKORC1 and CYP2C9 cannot explain all of this resistance,” Dr. Weck says. “It’s important to include the genetic factors associated with extreme dose requirements.” There are a number of resistance variants in VKORC1 that have been described in the coding region, and some of them show up frequently in specific population groups—in Ashkenazi Jews, for example, the D36Y mutation is relatively frequent.

Dr. Johnson suspects that because the influence of other genes appears to be small, physicians may have written them off. But multiple “small” additions have a way of adding up, she says.

It’s the common computational dilemma of personalized medicine: How personal can things actually get? A mutation that accounts for one to two percent looks fairly insignificant at the population level. “But if you’re in the group that carries that, it could well explain a much larger percent of that variation,” Dr. Johnson says. This is true even within the CYP2C9 and VKORC1 polymorphisms, she points out. As an individual, carrying the *3 mutation of the CYP2C9 gene is more important than carrying the VKORC1 SNP, since for every copy of the A allele, it explains about 25 to 30 percent dose reduction. But on a population level, the VKOR SNP is more revealing, simply because it’s more common than CYP2C9*3.

Dr. Weck has been involved in a study at UNC comparing patients whose dose is based on the so-called Gage algorithm (developed by Washington University’s Dr. Gage, and published on the Web site versus those whose dose is guided by a clinical dosing algorithm with no genetic information. She says one of the strengths of the UNC trial is that it includes a high percentage of African-American patients. But she worries that current pharmacogenetic algorithms may not have an effect significant enough to change practice. “It’s been difficult to show clinical utility in part because the extremes in response are rare,” Dr. Weck says. “There’s also a lot of noise in the system, due to other factors that can affect drug response.”

Research methods have also played a role in emphasizing common SNPs, Dr. Weck says. Most are found in genomewide association studies, in which common variants are more likely to rise to a level of significance. It’s the difference between finding a shell on a beach versus finding an oyster with a pearl.

That’s meant researchers have focused their efforts, understandably, on common variants associated with incremental differences in warfarin response. “I’m just afraid we may be throwing the baby out with the bathwater,” Dr. Weck says. What’s most important to know, she says, is who has extreme sensitivity or resistance to warfarin. “We’re concentrating on the common SNPs that have an incremental effect, and losing sight of the extreme dose requirements and those rare genetic factors.”

It’s not that Dr. Weck wants to cast aside the common man, or, rather, the common variants. “But they have a much less dramatic effect,” says Dr. Weck.

In a study she and her UNC colleagues did in collaboration with a Brazilian group (Orsi FA, et al. Thromb Res. 2010;2126:e206–e210), researchers found a coding region variant, V66M, in the VKORC1 gene, in two of seven highly resistant Brazilians of African descent.

But even though this and the other variants are known to be involved in extreme dose responses, they’re not typically included in the various dosing algorithms, she says. incorporates some rare variants, including CYP2C9*5, CYP2C9*6, and GGCX rs11676382; eventually, CYP2C9*8 and CALU rs339097 will be added as well. “However, as more variants are discovered, it takes time and money to quantify their effect on warfarin dose and then to update and validate all of the dosing algorithms at [the site],” says Dr. Gage, associate professor of medicine, Washington University School of Medicine. “We want to keep free, so we can’t include every rare variant. As is, we have to solicit donations and NIH funding to update the application.”

Lack of reimbursement hasn’t exactly inspired industry to put “develop warfarin PGx assays” at the top of their to-do lists. The UNC researchers have experienced firsthand the impact of the shrinking market. They launched their study using a test that has since been discontinued; the lab is now working to make an LDT to fill in the gap.

Warfarin pharmacogenetic tests will have a finite market, Dr. Johnson suggests. To her, single targeted tests eventually won’t make sense. “Large amounts of patient-specific genetic information will be available, so it won’t be necessary to order single tests,” she says. More and more, she’s hearing that insurance companies are agreeing to pay for a whole genome sequence for discovery of disease genes, such as in hypertrophic cardiomyopathy. “They realize that it’s cheaper in the end to pay for it all than to pay for it piecemeal,” she says. “I think that’s the future of pharmacogenetics as well.”

Both current and recent trials have highlighted the need for rapid genotyping.

Dr. Eby notes that one payer, Medco Insurance Co., sponsored an observational trial, the Medco-Mayo Warfarin Effectiveness Study, to look at whether patients would benefit from genetic information (Epstein RS, et al. J Am Coll Cardiol. 2010;55:2804–2812). “It was looked upon as an effectiveness study—here’s what happens when you use warfarin pharmacogenetic information in a more realistic study,” says Dr. Eby. “And it did show a clinically important difference in outcome.” The genotyped cohort had 31 percent fewer hospitalizations overall and 28 percent fewer hospitalizations for bleeding or thromboembolism during the six-month followup period.

But the study has been criticized—appropriately so, says Dr. Eby—because of the delay in providing genetic information to the treating physician. It took, on average, 20 days, during which time most of the variation in warfarin dose would have been worked out through trial and error. There’s also some suspicion, he says, that patients who agreed to be in the trial (and who thus received warfarin genotyping), for whatever reason, are healthier or more compliant than patients who declined to participate.

Most patients in this country are started on 5 mg a day, says Dr. Johnson, then adjusted up and down guided by INR testing. It usually takes one to three months to stabilize patients, but during that period, the rate of serious adverse events—including bleeding and thromboembolic events—is as high as it is every year thereafter, she explains.

Ideally, says Dr. Eby, clinicians would have pharmacogenetic information before the second dose is given. But variations in rates of metabolism as well as sensitivity do not necessarily become evident within the first two doses, he says. Dr. Gage’s site includes algorithms that allow incorporation of genetic information up to 10 days within starting warfarin. Beyond that, it’s unlikely to be useful, because trial and error will accurately determine a patient’s dose.

But while it seems to make sense to start off with a bang, like a Brahms symphony, that’s not always the case. Dr. Gage says that for orthopedic or valve replacement populations, sending the test to a reference lab may not be a hindrance, since these patients typically know they’ll be taking warfarin weeks in advance. “Even for other indications, the first couple of doses do not need to be tailored to genotype,” he says. “In fact, in current clinical trials”—including GIFT—“the protocols ignore the CYP2C9*2 and CYP2C9*3 variants when recommending the initial dose(s), to prevent a delay until the INR is therapeutic in these slow metabolizers.”

If warfarin pharmacogenetics testing hasn’t gotten off to a white-hot start, it shouldn’t be considered a dying ember, either. Observers suggest that it was overhyped at the beginning, as is so often the case with genetic testing. From there, anything less than immediate, unerring success would make it seem a failure.

“Genotype alone is helpful, but to dose warfarin well, one also needs to consider clinical variables, INR response, and pharmacokinetic principles,” says Dr. Gage. “Our expectations were too high.”

There are, to be sure, scattered pieces of evidence that clinical interest in warfarin PGx persists. Dr. Gage reports that use of the Web application has grown each year; currently, he says, the site averages 2,400 visitors a week. Some are using it for randomized studies, others for educational purposes, he says. Most are using the site clinically—without entering genotype.

Dr. Weck and her lab began offering warfarin genotype testing clinically in early April. Within a week came a request to do testing for an African-American patient with extreme resistance—her daily dose was 15 mg, yet her INR refused to budge from 1.2. “It’s not surprising that the current pharmacogenetic algorithms didn’t account for her warfarin resistance. We may want to sequence the VKORC1 gene and see if she has a resistance variant.”

Dr. Johnson and her colleagues at Florida plan to launch a clinical program for personalized medicine in late May or early June. They will start with clopidogrel and intend to add warfarin within a year. The plan is to do testing on a panel of 256 polymorphisms, some of which will be used clinically, some of which may eventually see clinical use, and some of which will be held for research.

The Florida program will, eventually, make warfarin pharmacogenetic testing part of the standard order set. (Clopidogrel, for example, will be part of the standard order set for patients going to the cath lab.) Genotype data will be uploaded to the medical record automatically, as would any test results. Dr. Johnson and her colleagues are also creating clinical decision support alerts, so clinicians know how to act on the information. If the lab breaks down the barriers to warfarin PGx testing, she predicts, clinician resistance will melt away.

A grant will pay for the testing early on, but Dr. Johnson anticipates that a broad panel will be no more expensive than a single pharmacogenetic test. “And we can get pharmacogenetic information that would be relevant over a person’s lifetime”—about drug metabolizing enzymes and drug transporters, for example, and other areas that have strong support in the literature. If a pharmacogenetics panel has been run for a patient who then needs warfarin in, say, two years, the genetic information will already be available to guide dosing, with no waiting and no additional payment needed.

“It’s preemptive testing,” says Dr. Johnson.

Not every laboratory will be able to replicate such a setup, but that doesn’t give them a pass, Dr. Johnson warns. Labs will in time have to figure out a way to get pharmacogenetic information into their clinicians’ hands. “It’s not obvious right now how to do that. So labs need to at least be ready when clinicians ask.” And eventually they, or their patients, will ask—of that, she has little doubt.

Karen Titus is CAP TODAY contributing editor and co-managing editor.