The Right Drug the First Time

By Aaron Krol 

March 19, 2015 | For two and a half years in the mid-1990s, Alan Shuldiner had a peculiar morning routine. “I’d wake up around four, four-thirty in the morning,” he remembers. “I’d throw a chest of dry ice in my trunk, and I had a portable gas-driven generator and a small centrifuge in the hatchback of my Honda Accord.” Fully equipped, he would drive from his home in Baltimore to Lancaster, Pennsylvania, stopping briefly to pick up a female chaperon. Around six o’clock, they would reach the home of Sadie Beiler and begin the day’s work.

Shuldiner and Beiler were unlikely friends. He was an associate professor at Johns Hopkins University, a graduate of Harvard Medical School trying to start a career in genetic medicine. She was a mother of eleven from Lancaster’s large Amish community, whose religious beliefs forbade her from even riding in Shuldiner’s car without an escort. Yet she was a surprising advocate for genetic testing, serving as a scientific “liaison” to the Amish, whose huge, interrelated families and near-absolute ban on intermarriage with outsiders made them a perfect population for studying the inheritance patterns of disease.

Shuldiner’s research in this community began with diabetes testing, but as government and industry became fascinated with the genetics of common disease, his newly-founded Amish Research Clinic rapidly expanded. In the mid-2000s, the Clinic hit on one interaction that seemed destined to play a role in medicine. They were testing Amish patients’ responses to clopidogrel, a blood thinner that at the time was the second most-prescribed drug in the world. Armed with new tools for sequencing DNA and backed by related research at other labs, the group discovered that a single variant in a single gene, called CYP2C19*2, made their patients much less likely to respond to this drug.

Clopidogrel is the drug of choice for patients recovering from certain cardiovascular surgeries. After a coronary angioplasty, for instance, where a stent is inserted into the heart to open up a blocked blood vessel, patients are given clopidogrel to keep their blood flowing around the stent. That means a CYP2C19*2 carrier could be in real trouble after this kind of surgery: if clopidogrel can’t perform its anti-platelet functions, blood clots around the stent could quickly lead to a heart attack.

At the time, Shuldiner thought studies like his were building to a real breakthrough. A CYP2C19 test might become a routine measure in hospitals to stop physicians from giving clopidogrel to non-responders. “I maybe naively believed that these discoveries would make it very quickly into mainstream medicine,” he says.

Yet almost a decade after the first connections between CYP2C19 and clopidogrel were discovered, only a handful of medical centers have adopted these tests. Shuldiner, now a lead organizer of a group called the Translational Pharmacogenetics Project (TPP), is trying to figure out why.

Trial and Error 

Pharmacogenetics, the study of interactions between genes and drugs, has been around as a theoretical science since the 1950s, but as a tool for practicing medicine it’s much more recent. Scientists weren’t able to connect specific genetic variants to drug responses until the early 1990s, and for many years, there were few tools to turn these discoveries into clinical tests.

That’s changing fast, says John Logan Black, co-director of the Mayo Clinic’s Personalized Genomics Laboratory in Rochester, Minnesota. “With advanced genetic testing, where we’re able to see the whole exome and the whole genome, we now have the ability to do a very comprehensive job of looking at all of the genes that may impact a given individual for many, many medications,” he says.

A psychiatrist by training, Black has a special interest in the pharmacogenetics of antidepressants and antipsychotics, whose effects are notoriously hard to predict. Almost anyone living with mental illness can attest to the long, uncertain trial-and-error process of finding the right drug and dose to treat their symptoms. “If you’ve got a patient who’s really severely ill — they’re quite suicidal or they’re very psychotic — you don’t want to be messing around with medication trials,” says Black. “You want to get to the right drug the first time.”

That’s why the psychiatry department at Mayo has started genotyping genes related to patients’ tolerance of SSRIs and other psychiatric drugs. Programs like this are starting to spread to other parts of Mayo, too. Recently, the Clinic recruited over a thousand patients at high risk for heart disease to have a large panel of genes preemptively sequenced, and the results stored in the hospital’s electronic medical record (EMR). This genetic information can affect prescriptions of commonly used drugs like clopidogrel, codeine, and simvastatin.

As one of eight participants in the TPP, Mayo will also be tracking the rates at which physicians actually use these tests, and trying to identify trouble spots so other medical centers can learn from early stumbles.

The TPP grew out of the Clinical Pharmacogenetics Implementation Consortium (CPIC), a group that seeks out drug-gene pairs that could be used in clinical testing, and publishes guidelines for implementing them. According to Shuldiner, a member of CPIC, the organization originally thought that just getting this information in one place would spur hospitals to adopt more pharmacogenetic tests. Only slowly did they realize that most of the barriers were further downstream.

“We realized that we were writing these guidelines, but if we looked at our individual institutions, very few of them were actually implementing or adopting,” he says.

Considering how much early adopters like Mayo are investing in their pilot programs, offering a tried and tested path to implementation will be crucial to getting more centers on board.

“We’re early to this particular feature of clinical practice, and when you’re early you take risks,” says Black. “We think we have a model that will allow us to at least break even on this. We’re not looking at large margins in this arena, but it’s so important we feel that we have to do it.”

From Lab to Line Item 

One problem the TPP has encountered is that pharmacogenetic tests can be genuinely hard to perform. We’re used to laboratory tests giving fairly clear, yes-or-no answers, but the interactions between drugs and genes are complex and interconnected. “It takes time to figure out what it means when you’ve got variations in five or six genes that have an impact on one drug,” says Black. “I have difficulty doing it, and I’ve been doing this forever.”

Aniwaa Owusu Obeng is a pharmacist at Mount Sinai School of Medicine in New York, where results from a few pharmacogenetic tests are automatically reported to physicians through the EMR. She says that physicians’ demands for these alerts are stringent: “Show it to me as something I can read in less than 30 seconds and make a decision.”

That doesn’t leave much room for nuance when a center decides it’s time to implement a new test. However many variables have to be juggled behind the scenes, the results have to be a clear-cut recommendation: prescribe a drug, avoid it, or adjust the dose.

Timing is also a problem. Physicians would like to run their tests at the moment they’re making a medication decision, but a gene sequencing panel can take hours or even days to perform. That’s an unacceptable delay in an emergency situation, and a big barrier even in a routine patient visit.

For a large institution like Mount Sinai (which is not part of the TPP), one solution is to test patients in advance. Yet this approach is limited in its reach. “I think currently it is going to be kind of restricted to more progressive and larger academic medical centers,” says Stuart Scott of Mount Sinai’s Institute for Personalized Medicine. “It requires a lot of education and a lot of infrastructure development.”

It’s also expensive. Running a midsize gene panel can cost well over a thousand dollars, and because these tests are done preemptively rather than as a response to a specific medical problem, it’s hard to convince insurance companies to reimburse them. That creates conflict between payers and providers, because many medical centers would rather save money in the long term by combining many genes into one sequencing run.

“It could probably be cheaper to do a preemptive genotype on a panel, versus doing single tests,” says Owusu Obeng. Preemptive testing also means genetic information will be available as soon as a physician encounters a patient. “Clinicians can have it before they make prescribing decisions, and potentially reduce adverse events or trial-and-error periods,” she says. “All of that should be in the best interest of any insurance provider, so I think it’s about time that implementers get together and make a strong case for it.”

That case will be hard to make, however, without hard evidence that pharmacogenetic tests actually improve patient outcomes when implemented hospital-wide. 

Culture Clash 

Shuldiner, who moved on from Hopkins to become the Director of Personalized and Genomic Medicine at the University of Maryland, took a narrower approach in his own work with the TPP: testing patients for CYP2C19 variants before they go through coronary angioplasty, to make sure poor clopidogrel responders get a different blood thinner.

“I believe you’ve got to walk before you run,” he says. “While it’s our dream to be whole genome sequencing every patient, and embedding thousands of actionable variants in the electronic health record… if we learned how to do this for one gene, one SNP, one patient population, that would be a huge step forward.”

The University of Maryland Medical Center (UMMC) comes as close as possible to testing CYP2C19 at the moment a blood thinner needs to be prescribed. A blood draw taken during surgery is sent to a genomics laboratory, which tests the sample on the spot. When the results come back, the physician can use them to change a patient’s blood thinner prescription during recovery.

The program may seem modest, but it runs head-on into a culture clash at UMMC. “There are some cardiologists who get it, and they’re very excited to offer the new test and learn how to interpret it, and there are others who just don’t believe in this,” says Shuldiner. “Unless there’s a 20,000-patient, placebo-controlled, double blind randomized clinical trial, [many] cardiologists just aren’t interested.”

The issue of clinical evidence is a profound one for pharmacogenetics. When you’re trying to prevent an already-rare event like clotting around a stent, it really does take thousands of patients to prove you’re making a difference. Scientists like Shuldiner can show that CYP2C19 affects platelet counts after taking clopidogrel, and it makes intuitive sense that this could lead to clotting after an angioplasty. Scraping together the money for a clinical trial to settle the question, however, is another matter. Pharmaceutical companies, which usually pay for large trials, have little incentive to do so for tests that involve off-patent drugs like clopidogrel, especially when the results could restrict rather than expand the number of prescriptions.

That makes it hard to appeal to many physicians, who are naturally conservative about changing the way they treat their patients. “Any time you’re introducing something that is — I don’t want to be dramatic, but groundbreaking is a word that comes to mind — you’d better have some research to back it up, or people are just not going to listen to you,” says Black.

There’s also some built-in skepticism toward pharmacogenetics in particular, because many drug-gene associations that scientists trumpeted after early experiments have not panned out in practice. “Overhyping some of these discoveries can lead to downstream problems,” says Scott. “For those clinicians who aren’t really believers or supporters of the field, if they [read about] the latest discovery in the New York Times, but it has no real evidence of validity or utility, that just provides them with more support.”

Even the most widely-adopted pharmacogenetic tests, like those connecting the genes CYP2C9 and VKORC1 with the blood thinner warfarin, have seen uneven results in the clinic. One large multicenter trial of warfarin testing found that treatment decisions based on simple variables like a patient’s age, race, and conflicting medications were just as successful as genetic tests at choosing the right prescriptions.

These results don’t mean that pharmacogenetics needs to go back to the drawing board. “What the warfarin trials have really shown is that there’s more research needed in different racial groups and different ethnicities,” says Scott. He points out that the clinical trial in question had a large African American population, but the genetic tests used were based on variants common in people of European descent. But studies like this do show that there are sincere reasons to be cautious when relying on this still-developing science, especially when decisions will affect patient care.

At the same time, the ultimate trump card would be positive clinical trials. Now that members of the TPP and other centers have started using CYP2C19 tests, the next step is to start a multicenter trial for clopidogrel that looks at clinical outcomes prospectively.

“The last thing we want to do in the field is practice snake oil genomic medicine,” says Shuldiner. “We need to be holding genomics up to the same evidence standards as any other field of medicine.”

Building Support 

Mount Sinai still uses warfarin tests, but it’s careful to cover all the bases. “For warfarin, there’s a whole list of other factors that need to be accounted for in the algorithm… which includes the age of the patient, the medications that they are on, their smoking status,” says Owusu Obeng.

“This is personalized medicine,” she adds. “It’s not necessarily just genomic medicine.”

That’s a key point about bringing DNA data into new fields. Genetics can supplement old ways of making treatment decisions, but not supplant them — and one lesson from the TPP’s experiences is that any medical center trying to change the way it treats patients has to be a little bit humble about the process.

“We’ve come to the realization that in many ways, these issues in pharmacogenetics are not much different from any other kind of biomarker test,” says Shuldiner. Genetics may attract some added skepticism, but problems like designing EMR alerts, or getting insurers to support new tests, are universal.

At the same time, where there’s a groundswell of enthusiasm for a new way of practicing medicine, clinics will eventually work out the wrinkles. One of the TPP’s biggest lessons has been that educating clinicians about pharmacogenetics is critical to the success of its members’ programs. That can take the form of training seminars, continuing medical education courses, or even guidance right at the point of care. An EMR alert that can direct physicians to the evidence for a drug-gene pair will be much more widely adopted than one that recommends a treatment without explanation.

Starting education early could also help clinicians accept pharmacogenetics as a normal part of medicine. Mount Sinai is starting to include pharmacogenetic tests in its medical school programs. “I think in the big picture it’s a fairly young field,” says Scott. “A lot of the clinician workforce didn’t have any formal education in pharmacogenetics at all, and still don’t.”

Owusu Obeng, a graduate of one of just two accredited residencies in the country for training pharmacists in pharmacogenetics, says that professionals like her who have specialized in the field need to be prepared to take on extra responsibilities.

“We didn’t think it was the role of the pharmacist to educate other providers initially, but I have to assume that role,” she says. At Mount Sinai, where clinical pharmacists make rounds with the medical teams, Owusu Obeng has a chance to add a genetic perspective when medical students and residents are getting their earliest exposure to the hospital environment. “By the time they get to be attendants or fellows and make these prescribing decisions, they would not only look at the clinical traits, but also think about the genomic aspects,” she says.

Pharmacists, she adds, “are best positioned to lead implementation in an institution. And you see that as you go around the country. With the institutions that are actually doing this, they have a heavy pharmacy presence as well.”

Meanwhile, at institutions that are embracing these tests, early adopters are confident that the evidence will eventually mount up to take pharmacogenetics nationwide. More recent warfarin trials, for instance, have started to show clearer clinical benefits of genotyping.

Black, whose laboratory often performs tests for warfarin response, is not surprised. When he sees a result that suggests a patient will be oversensitive to warfarin, Black’s procedure is to call the attending physician directly. “Nine times out of ten,” he says, “that physician is just trying to figure out what happened when he or she gave the patient warfarin, and their INR [a measure of blood clotting] went off the chart.”

And the physicians and laboratorians leading the way on pharmacogenetics are in no doubt what they would want for their own care. “If I was going to be prescribed clopidogrel tomorrow, I know that I’m a *2 carrier, and I know that means I have an increased risk for stent thrombosis,” says Scott. “I think logically that can only improve our clinical decision making.”

Correction 3/20/15: This article originally stated that Alan Shuldiner is currently working at the University of Maryland. At the time Shuldiner spoke to Bio-IT World for this article, he held joint positions at the University of Maryland and Regeneron Pharmaceuticals; however, he has since transitioned fully to Regeneron.