NEW YORK – After launching at a single site five years ago, UCHealth in Colorado's clinical decision support (CDS) tool for prescribing a commonly used antiplatelet medication is live across the entire hospital system.
The team behind the CDS system developed and implemented it gradually, expanding it to new indications and sites, and is now using it as a foundation for PGx alerts for additional drug-gene pairs.
The CDS tool started by specifically warning physicians prescribing Bristol Myers Squibb/Sanofi-Aventis' P2Y12 inhibitor Plavix (clopidogrel) when a patient has a documented CYP2C19 gene variant in their medical record that suggests they would metabolize the drug poorly. Plavix is widely used as therapy after heart attack, stroke, and other heart problems, but for patients with certain genetic variants, the medication is less likely to be effective.
Since patients with these pharmacogenetic variants are left at risk for future adverse cardiac events, proactively identifying them and prescribing alternatives can lead to "high-impact outcomes," said David Kao, a cardiologist at UCHealth and medical director of the Colorado Center for Personalized Medicine at the University of Colorado Anschutz Medical Campus.
The CDS tool works by automatically sending an alert within UCHealth's Epic Systems electronic health record (EHR) system when a physician orders Plavix for patients deemed CYP2C19 intermediate or poor metabolizers based on information available in their medical chart. The alert is an interruptive warning — or a "pop-up" — and includes a brief description of the drug-gene interaction and risks, recommendation for alternative therapies, and external link to additional educational information.
Initially, in 2018, the team rolled out the CDS tool in a narrow indication, specifically focused on inpatients undergoing elective percutaneous coronary interventions after acute coronary syndrome at a single cardiac catheterization laboratory. The development team worked closely with the small group of cardiologists prescribing in that indication and at that site about what the alert should include and at what point in the care process it should occur.
Since then, the development team has expanded the PGx alert, first to include inpatients undergoing acute percutaneous coronary intervention procedures and then to fire for any Plavix order or refill, regardless of indication. Although the program started in the system's UCHealth Metro region, it's now used across the entire health system in both inpatient and outpatient settings, including 14 hospitals.
The CDS tool draws on data from a research biobank at the Colorado Center for Personalized Medicine, which was launched by CU Anschutz and UCHealth and began enrolling participants in 2015. UCHealth patients participating in the biobank provide a blood or saliva sample that's genotyped at a CAP-accredited and CLIA-certified lab at CU Anschutz, data from which is held in the biorepository and can be used for personalized medicine research.
Patients who participate in the biobank provide consent for research, recontact, and return of clinical genetic test results to the EHR. Having this data in the EHR then enables recommendations to be proactively displayed when a physician is ordering a prescription that could be impacted by a patient's genetics. Only participants in the biobank are able to participate in this PGx program.
Generally, it's not common for patients to be referred for genotyping before being prescribed Plavix, since patients might need the drug quickly in the midst of an acute issue, such as a heart attack, Kao said. That's why it can be challenging for a hospital system without such a biobank to set up this type of pharmacogenetic program.
Previous research has suggested that patients who received earlier CYP2C19 pharmacogenetic testing are more likely to be on genotype-informed treatment. In cases where PGx information isn't available, Plavix is often preferred by prescribers because patients on other drugs in the same class as Plavix have a higher risk of bleeding. And, with generic versions of Plavix on the market now, clopidogrel can be a less expensive option than newer P2Y12 inhibitors.
As of summer 2023, more than 227,000 patients at UCHealth had enrolled into the biobank, 41.5 percent of whom were normal metabolizers of Plavix based on CYP2C19 status, according to a paper published in the American Journal of Health-System Pharmacy earlier this year, on which Kao was the last author. One-quarter of participants were categorized as intermediate metabolizers, and 2.6 percent were poor metabolizers. In total, 55 patients have had a PGx alert pop up before a potential Plavix prescription.
This experience has provided the foundation for a broader PGx system at UCHealth, which now alerts for seven genes in which certain variants can affect response to 30 medications. For pharmacogenetic variants in CYP2C19, for example, the system flags potential drug-gene interactions for the seizure medication Onfi (clobazam), the antidepressants Celexa (citalopram) and Lexapro (escitalopram), and multiple proton pump inhibitors, in addition to Plavix. Other drugs included in the CDS system can be affected by variants in SLCO1B1, DPYD, CYP2C9, TPMT, NUDT15, and ABCG2. Drug-gene pairs determined to be lower risk are not interruptive alerts, and alerts are displayed in patient charts without a pop-up warning.
CYP2C19's impact on Plavix was the first drug-gene pair to be selected for the EHR-based clinical decision support "because the evidence is so strong about the relationship between genetics, the drug's pharmacology, and the potential adverse effects," said Christina Aquilante, a professor at the Skaggs School of Pharmacy and Pharmaceutical Sciences at the University of Colorado Anschutz Medical Campus and director of pharmacogenomics for the Colorado Center for Personalized Medicine, and first author of the American Journal of Health-System Pharmacy paper.
When selecting Plavix as the initial medication for PGx alerts, the team considered the strength of evidence for the clopidogrel-CYP2C19 association based on guidance from the Clinical Pharmacogenetics Implementation Consortium (CPIC) and the US Food and Drug Administration. CPIC, an internationally recognized guidelines body, recommends use of other P2Y12 inhibitors like Brilinta (ticagrelor) or Effient (prasugrel) for CYP2C19 intermediate or poor metabolizers, and the FDA has required a boxed warning on Plavix's label since 2010, explaining that the drug may not be as effective in CYP2C19 poor metabolizers and that doctors should consider switching such patients to another P2Y12 inhibitor. In addition, when selecting the first gene for the EHR-based clinical decision support tool, researchers considered "performance of the variants on the genotyping array, likelihood of detecting variant alleles in our population, large number of drugs affected by CYP2C19 variation, and corresponding prescribing frequencies of the affected drugs at our health system," they wrote in the paper.
There are also two clear alternative treatments for Plavix, which made setting up the clinical decision support straightforward, Kao added. Within the pop-up alert, the prescribing physician is presented with an option to remove Plavix and to start an order for an alternative medication. "In all of the pharmacogenetic applications, we really wanted to make choosing the recommended alternative as easy as possible, if not easier than not doing it," he said.
About 2,700 — or 1 percent — of biobank participants have received a PGx alert for any drug, study authors wrote in the paper. More than 3,300 PGx alerts have been fired in total, across all of those patients. Just over half of the PGx alerts were related to proton pump inhibitors, followed by statins (30.9 percent), escitalopram and citalopram (12.1 percent), and clopidogrel (1.6 percent).
The team continues to track the number of alerts and whether physicians change prescriptions based on them. Down the line, for the clopidogrel-CYP2C19 alert, Aquilante said they plan to evaluate whether the warnings affect the likelihood of a patient experiencing future major adverse cardiovascular events by comparing genotyped patients to a historical control population that wasn't tested.
"That's on our longer roadmap," she said.