More progress, fewer barriers for PGx testing

Karen Titus

January 2024—Sometimes even superb ideas can also turn out to be quite, well, bothersome. Zoom meetings. Bridal showers. Bike lanes. Parking apps. QR menu codes. And—if laboratories aren’t careful—the same can be true of pharmacogenomic testing.

Just ask Ann Moyer, MD, PhD, associate professor, laboratory medicine and pathology, Mayo Clinic. When it comes to pharmacogenomic testing, laboratory medicine brings significant expertise to the table. But in clinical settings, physicians who prescribe the medications need to be familiar with how to use the test results. They also need to work with the lab to decide which tests, for which genes or gene-drug pairs, will be most helpful for their patients, she says.

“Especially if you’re going to start incorporating clinical support alerts into the EHR,” adds Dr. Moyer, who was chair of (until Dec. 31; she is now advisor to) the CAP/ACMG Biochemical and Molecular Genetics Committee. “If the practice doesn’t actually want them, then you’re just going to end up annoying them.”

It’s a lesson Stuart Scott, PhD, learned the hard way. In his previous position at Icahn School of Medicine at Mount Sinai, New York, he and his laboratory colleagues developed a comprehensive pharmacogenomics panel “that we were very proud of,” recalls Dr. Scott, who is now at Stanford University, where he is professor, Department of Pathology, and director of the clinical genomics laboratory. “We put a lot of time and energy into making sure it was scientifically valid and sound and high-evidence content.”

So thorough was their work, in fact, that the report generated by the panel was a 30- to 40-page PDF file. Which, needless to say, didn’t work. “It was way too much information,” says Dr. Scott.

Happily, it can also be taken as a sign of progress that pharmacogenomic testing is now large enough to be an irritant. The field, young but not new, has seen earlier barriers to its use fall in recent years. What might have once been a lengthy steeplechase event has now settled into the more routine laps around the track. “We’re seeing more progress, rather than new barriers,” is how Dr. Moyer characterizes the field.

One of the issues that’s dropped away, Dr. Moyer says, is that in the past it was possible to test for only one gene at a time—hardly a cost-effective or efficient approach. Most of these genes are involved in drug metabolism, with some drugs metabolized by multiple enzymes encoded by multiple genes. “If you can only do one test for one gene at a time, you’re really limited in how a patient can actually use that information.”

Now that gene panels are available, one barrier has fallen.

Other barriers have dropped away as electronic health records and clinical decision support alerts have evolved to accommodate pharmacogenomic testing. Solid data alone isn’t enough—physicians need seamless ways to access it, both in the short and long term.

“Really, the dream is for this to be testing that you do up front,” Dr. Moyer says, with results available for use at different points in a patient’s life—acknowledging, of course, that the tests will change and improve over time.

But if the results are in a PDF, they can easily become buried in the EHR, “and no one can ever find them again,” she says.

That’s a lesson Dr. Scott and his Mount Sinai colleagues learned the hard

way.

“It just didn’t work,” Dr. Scott says. “It was too much information.” Even the best information is useless if no one can find it. Clinical colleagues worried about who was responsible for managing pharmacogenomic results and possibly missing crucial information. What would happen if one physician ordered pharmacogenomic testing, and then a year or so later, another physician prescribed a medication for that patient but missed relevant pharmacogenomic information?

A more immediate practical problem arose as well. When Dr. Scott and colleagues were translating their pharmacogenomic research into clinical implementation, back in 2009 and 2010, “We thought it was going to be the next big thing. But there was very little appetite from the providers. That was understandable in retrospect, because they just didn’t know what to do with the results.”

Now it’s possible to put pharmacogenomic, or PGx, data into discrete fields in the EHR. Much work has also gone into refining clinical decision support alerts to make PGx results more user-friendly. It’s now possible for the PGx testing to remain effectively invisible to the prescribing physician. “It’s just there in the background,” says Dr. Moyer. The alerts can then be set up “so that when the patient has an actionable genotype, it fires; if they don’t, it leaves them alone.”

Having a more integrated EHR-based reporting, coupled with clinical decision support, is crucial. “You remove the massive burden of giving all that information at once to any one provider,” Dr. Scott says. As it turns out, comedy isn’t the only field where timing is everything.

Even the most organized and finely tuned system will sputter, however, if physicians lack basic PGx knowledge.

“The person prescribing the medication has to be at least somewhat familiar with pharmacogenomics, or be able to find the information they need to use it,” Dr. Moyer says. She’s seen improvement—another hurdle falling—in recent years, as education has expanded for clinicians and pharmacists. “That’s helped quite a bit.”

The biggest improvement she’s seen at Mayo came from its Right 10K study, involving more than 10,000 patients across numerous specialties and focused on using genomic data to individualize treatment. “That meant that many different physicians throughout the practice, and pharmacists, would be exposed to pharmacogenomic data on their patients,” Dr. Moyer says. “They could kind of get a taste of, ‘Well, what is this? Is it helpful?’” The effort included educational campaigns, Dr. Moyer says, “so people would know what to do when the data appeared, since they weren’t ordering the tests themselves.”

That turned out to have a leveling effect among all providers. In the past, some specialties have been quicker to adopt pharmacogenomics than others, in part for historical reasons, Dr. Moyer says. It depends in part, too, on the number of actionable drugs in a clinical specialty.

Another early barrier was outsized enthusiasm. Early adopters were understandably excited, perhaps “to the point that maybe as a field we weren’t all as realistic as we should have been,” Dr. Moyer concedes.

In practice, pharmacogenomics is focused primarily on drug metabolizing enzymes, and at present to a lesser extent on drug targets, Dr. Moyer says. “In general we can predict how the drug is going to be metabolized differently from one person to another. In those cases, we generally are able to predict toxicities, and maybe patients who needed an adjusted dose.

“But,” she continues, “we’re not at the point yet where we can say, ‘This is the drug you need to take.’ But early on, when the field was younger, I think people got the impression that it was going to be this magic bullet that would tell you exactly what drug to use. I think that’s jaded some clinical specialties to some extent.”

Some specialties, on the other hand, have shown great progress, with guidelines in place and high rates of adoption. Take, for example, the HIV drug abacavir. For patients who are positive for HLA-B*5701, it’s recommended that they not take the medication because they’re at higher risk for severe cutaneous reactions. In this area of medicine, Dr. Moyer says, it’s typical for PGx testing to be done up front prior to the medication being prescribed.

In other fields, knowing that a patient is a poor metabolizer of certain drugs may mean only a dose adjustment rather than a different drug. “It’s not always black and white,” she says.

Another earlier barrier, lack of reimbursement, is also dropping away, with a CPT code now available for PGx panel testing as payers recognize its benefits, says Dr. Moyer.

That’s been helped along by the strength and visibility of the Clinical Pharmacogenetics Implementation Consortium (https://cpicpgx.org), Dr. Scott says, with a growing literature acknowledging that medications that have CPIC guidelines have a level of evidence that warrants reimbursement. Though it’s limited to specific genes and medications, “just the fact they’ve come around to be more supportive is a huge change in the field.”

While individual institutions continue to make progress, the field has been pushed forward by larger entities such as CPIC that are doing some much-needed heavy lifting.

On a practical level, says Dr. Scott, CPIC answers the question uppermost in clinicians’ minds: If PGx testing is performed, how should they use the results clinically? CPIC has been publishing (in Clinical Pharmacology & Therapeutics) clinical practice guidelines for gene-drug pairs over the past dozen or so years.

All are structured the same way, Dr. Scott says: They include a section on the gene, one on the drug, and then a section containing the evidence supporting that association. They also have tables that link a particular genotype to the recommendation. CPIC does not advocate for or against testing. (“That’s a whole other debate,” Dr. Scott says.) But CPIC’s mantra, so to speak, is that because at some point all this data will become available—whether it’s driven by individual physicians ordering clinical tests, institutionwide endeavors, or even patients providing results from ancestry testing—providers will need help acting on results, Dr. Scott says.

These guidelines have become valuable for the whole PGx field, he says, because multiple professional societies have endorsed them and the authorship is international. “It’s a central place to see the evidence being curated systematically,” he says.

In a field where not all guidance travels at the same speed, CPIC “has done a really good job of providing level of evidence guidelines,” agrees Karen Weck, MD, director of molecular genetics and pharmacogenomics, professor of pathology and laboratory medicine, and professor of genetics, University of North Carolina at Chapel Hill. CPIC assigns levels of evidence for genes-drugs (for example, 1A, 1B, 2A, 2B) based on the strength of the literature. “Many places, including ours, have strongly considered doing testing for all the level 1A CPIC recommendations, because CPIC does make specific dosing or therapeutic drug recommendations. That can be very helpful for clinicians when they order pharmacogenomic testing.”

CPIC is part of the Pharmacogenomics Knowledge Base, PharmGKB (https://www.pharmgkb.org/), an interactive tool funded by the National Institutes of Health that is also an excellent general resource for all things related to pharmacogenomics, says Dr. Weck, who is also director of the molecular genetics laboratory, McLendon Clinical Laboratories, UNC Hospitals. It provides links to the CPIC recommendations, gives prescribing information and clinical annotations, and lists all variants in a particular gene and their effects. “It’s well referenced, well annotated, and often updated. And it’s very user-friendly.”

It was the opportunity to work with PharmGKB and some of its developers, including Teri Klein, PhD, that drew Dr. Scott to Stanford. “It’s probably the single largest knowledge base for pharmacogenomics in the world,” he says.

He and Dr. Klein are also working to set up a comprehensive PGx sequencing panel in the clinical laboratory at Stanford, working hand in hand with IT. The goal is to deliver panel-based results directly into the patient’s chart and use EHR functionality to provide clinical decision support. “It really takes the burden off the physician,” he says.

Another catalyst is the Food and Drug Administration, which continues to include pharmacogenomic information on some drug labels.

Despite the criticisms the agency has historically received for not updating its labels, says Dr. Moyer, “I’ve definitely seen where they’ve gone back and added pharmacogenetics to a drug label when the evidence suggests they should.” She cites one example: the addition of NUDT15 to the thiopurine labels that previously included only TPMT.

Even more helpful, Dr. Moyer says, is the FDA Table of Pharmacogenetic Associations (https://bit.ly/FDA-PGxAssn), which summarizes the information contained in the drug labels.

The table is divided into three sections. The first shows the PGx associations for which data support therapeutic management recommendations. The second section shows associations for which data indicate a potential impact on safety or response, while the third section looks at potential impact on pharmacokinetic properties only.

Though the data in section one are clinically actionable, the other two sections are useful as well, says Dr. Scott. Section two is considered emerging evidence, while the information in section three can help temper unchecked enthusiasm. Explains Dr. Scott: “There’s a lot of literature that shows very robust association between genetic variants and drug phenotype. But if the drug phenotype is only the drug levels in your blood system, but it has no actual clinical outcome change, then that’s not really high enough evidence to suggest that people should do genetic testing in that context.”

The FDA has been crucial in promoting the use of pharmacogenomics, Dr. Scott says. The agency has essentially been doing its own scientific curation and evidence review, leading to those three tables. And for all of CPIC’s strengths, he says, the FDA carries with it a more formal authority.

Professional societies have also been using their influence to bring more clarity to the field.

The Association for Molecular Pathology pharmacogenomics working group, which Dr. Weck co-chairs (Drs. Moyer and Scott are members), has been working for five or six years to help push labs out of the starting blocks. Many labs want to set up PGx tests but don’t know what alleles to include, Dr. Weck says.

To date the group has published six papers in The Journal of Molecular Diagnostics, with a seventh—a recommendation for DYPD genotyping assays—set to be submitted for publication soon, says Dr. Weck. The group includes the CAP as well as CPIC, Pharm­GKB, the Dutch Pharmacogenetics Working Group (“which has been very active in the field,” Dr. Weck says), and the European Society of Pharmacogenomics and Personalized Therapy.

To develop their guidance, Dr. Scott says, “We went gene by gene, looking at the common PGx genes, including CYP2C19, CYP2C9, CYP2D6, TPMT, NUDT15, and others.” They looked to see whether each variant occurred at a high enough frequency in the general population, their functional impact, and whether reference materials are available to support validation—the group considers only alleles with such material to be tier 1 recommendations.

(Along those lines, the Centers for Disease Control and Prevention GeT-RM program, which characterizes reference materials, including for PGx, is another useful resource. Says Dr. Moyer: “This is helpful to laboratories when they are trying to validate their assay, so they can obtain samples that are known to have variants of interest, as well as to use as controls once their assays are live.”)

Though the recommendations of the AMP pharmacogenomics working group have been well received by clinical laboratories conducting PGx testing, Dr. Weck says, she acknowledges there often remains a disconnect between laboratory recommendations and offerings, and clinical recommendations and uptake.

“At least at my institution, that has historically limited the clinical utilization of pharmacogenomic testing,” she says. Nevertheless, the recommendations from the AMP and other groups, particularly CPIC, “have really moved the needle in terms of uptake of pharmacogenomic testing.”

And while CPIC might be seen by some as primarily aimed at end users, says Dr. Moyer, laboratories can incorporate that information into their electronic reports. CAP proficiency testing incorporates CPIC guidelines, she says. “We cite them regularly in the participant summary report discussions because we do want labs and lab directors to look at those.” Even if not every lab is providing medication recommendation information in its reports, laboratory directors and other lab professionals need to know about CPIC, she says, and be able to help users understand how to use PGx and direct users to the right resources.

Even as next-generation sequencing becomes more common, it’s not the only option in PGx testing, “which surprises people,” Dr. Moyer says.

“We still do a lot of targeted genotyping for pharmacogenetics, so it’s important how the lab designs its test.” Simply put, a specific gene variant won’t be detected unless it’s included.

That’s why the efforts to bring order to test design are so critical. “It used to be all over the map,” Dr. Moyer says, with different laboratories testing for different alleles. Published recommendations for what alleles to include in a clinical test, she adds, “are really helpful, because we need to be inclusive of the whole American population, which includes people of many different ancestral backgrounds.”

CPIC has also pushed the field to standardize nomenclature. “If different labs are using different nomenclature, that makes it hard for the end user to have any hope of understanding this information,” Dr. Moyer says.

As clinicians have become more familiar with PGx testing, Dr. Moyer says she and her colleagues are now receiving fewer calls from clinicians needing help in interpreting results. Mayo has its own PGx pharmacist specialists. But for institutions that don’t, she says, laboratories should know that even general pharmacists are being educated in pharmacogenomics. “They can handle a lot of questions.”

In a similar vein, she’d like to improve laboratory director education. Currently pharmacogenomics is a relatively niche field, Dr. Moyer says. But as the field expands, “We may need more labs that can handle it.”

Laboratories will also need to handle turnaround times for PGx testing. “We try to be quick about getting results back,” says Dr. Moyer. And ideally turnaround times will become less important in PGx, assuming preemptive testing becomes the norm. “People would get the testing before you really need it,” she says.

But new devices may also help speed matters along. At Mayo, “the cardiac cath lab has been really interested in rapid turnaround times for CYP2C19,” to help guide use of clopidogrel, Dr. Moyer says. The Cube CYP2C19 testing system (Genomadix), which offers rapid genotyping for antiplatelet therapy, might be an option, she says. “It’s a limited number of alleles, but for places where there’s a quick turnaround time indication,” such devices may prove useful.

Dr. Scott and his colleagues at Stanford are also looking at rapid testing; they set up the Cube system this past summer. “We’re hoping to go live in January,” he says. “Our neurology colleagues are interested in having rapid testing results—the same day—to help them with their platelet prescribing decisions in the context of minor stroke and TIA.”

Even as the barriers to pharmacogenomic testing tumble, progress isn’t always swift or straightforward.

Dr. Moyer is quick to weigh in. “For the last I-don’t-know-how-many years, I have been saying, ‘Oh, we’re going to switch to sequencing soon,’ instead of still using targeted genotyping. But at least at our institution, targeted genotyping is still faster and cheaper, and it makes it a lot more accessible to patients.”

Long term, however, she hopes that as pharmacogenomic testing as a whole turns to more sequencing, laboratories will be able to layer pharmacogenomics onto other testing. For example, might it be possible to add DPYD onto tumor sequencing, given its importance for metabolizing fluoropyrimidines? The only caveat, Dr. Moyer says, would be that when sequencing tumor tissue for the genes that are specific to the chemotherapy metabolism, labs need to be aware that the tumor could have additional mutations not present in the germline. “You’d have to confirm your findings if you found that someone had reduced metabolism.”

Dr. Weck and her UNC colleagues have performed PGx testing for a number of clinical trials looking at the clinical utility of genotype-guided dosing, including for warfarin, tamoxifen, and irinotecan. But those efforts haven’t been adopted in clinical practice, primarily because the results of the clinical trials have not demonstrated improved outcomes.

In contrast, CYP2C19 genotyping has had the greatest uptake by UNC clinicians. “Cardiologists who are involved in placing coronary artery stents have ordered CYP2C19 genotyping in every patient to identify poor and intermediate metabolizers who should receive an alternative platelet inhibitor. Therefore, we perform CYP2C19 genotyping three times a week to identify patients who are at risk of thrombosis due to lack of clopidogrel response in the setting of coronary artery stents or atrial fibrillation.”

The UNC molecular genetics lab has also worked with Epic and its genomics module to provide clinical decision support tools for any patient who is prescribed clopidogrel. If clopidogrel is ordered in a patient known to be a CYP2C19 poor or intermediate metabolizer, a best-practice alert pops up to suggest the patient receive a different drug or alternative dose. They use Epic’s genomics module to include a genomic indicator specifying a patient’s genotype information up front, for various genetic information, including PGx. “So if the information is available, it can be part of the electronic medical record as a discrete result, and ideally pharmacogenomic information would be available preemptively” if the patient ever needs a drug. Dr. Weck says routine, preemptive PGx genotyping as part of annual medical visits is “the dream of pharmacogenomics, but that’s still not happening.”

They also have clinical decision support in place for TPMT and NUDT15 genotyping, which is recommended for patients who are prescribed thiopurine drugs.

Dr. Weck is a champion of clinical decision support. “CDS does help drive the practice” of pharmacogenomics, she says, and is likely to keep gaining traction in medicine in general. CDS can also be helpful in another sense, she adds, by blurring the distinction between primary care physicians and specialists: “It doesn’t matter who orders the drug if an alert pops up in the electronic medical record.” CDS tools at her institution have to be approved by a clinical decision support committee, which includes clinicians and, if relevant, pharmacists.

On a related note, she says the FDA package insert for lecanemab, a recently approved drug for treating Alzheimer’s disease, recommends APOE genotyping. “We’re bringing up this testing at UNC—the pharmacy requires APOE genotyping before the release of the drug and considers it to be contraindicated in APOE4 homozygotes” who are at high risk of amyloid-related imaging abnormalities, or ARIA. This is a fairly new application of PGx testing associated with drug toxicity, she says.

As pharmacogenomics spreadsits wings, what else should individual laboratories be doing?

“You have to partner,” says Dr. Moyer. A good place to start is to let clinicians know testing is available, “and that you’re willing to partner with them—if they’re interested.”

The more labs can do to make PGx testing seamless for their colleagues, the better, she continues. Noting that short and shrinking appointment times hinder most physicians, “If those of us in lab medicine can make it as easy as possible for our colleagues, that’s a good role for us to be playing.”

Dr. Scott agrees. “The lab can build the best test we want, but if providers are uncomfortable with it, then it’s not going to take off anywhere.”

While he and his lab colleagues are used to working with numerous specialties with an interest in PGx testing—cardiovascular medicine, behavioral health, pain management, and oncology—“We’ve also enjoyed working with primary care physicians. They’re quite open to this type of information, with the caveat that they’re very busy. If we can deliver the results efficiently, without adding extra work for them, we’ve found this group to be fairly open to leveraging that information for their clinical decisions.”

But, he cautions, “It’s already hard enough for primary care physicians and specialists to manage what they have now. To add another layer of burden on top of it usually makes them shy away.” Ideally, labs will understand the operational barriers well in advance, as opposed to finding them out after a test launches. “That does a disservice to the whole thing if it’s found out too late.” 

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