Projected impact of pharmacogenomic testing on medications beyond antiplatelet therapy in percutaneous coronary intervention patients

Abstract

Aim:

CYP2C19 genotyping is used to guide antiplatelet therapy after percutaneous coronary intervention (PCI). This study evaluated the potential impact of CYP2C19 and multigene pharmacogenomics (PGx) testing on medications beyond antiplatelet therapy in a real-world cohort of PCI patients that underwent CYP2C19 testing.

Methodology & results:

Multiple medications with actionable PGx recommendations, including proton pump inhibitors, antidepressants and opioids, were commonly prescribed. Approximately 50% received a CYP2C19 metabolized medication beyond clopidogrel and 7% met criteria for a CYP2C19 genotype-guided intervention. A simulation analysis projected that 17.5 PGx-guided medication interventions per 100 PCI patients could have been made if multigene PGx results were available.

Conclusion:

This suggests that CYP2C19 and multigene PGx results could be used to optimize medication prescribing beyond antiplatelet therapy in PCI patients.

Selecting treatment strategies that account for individual genetic differences offers enormous potential to improve health outcomes and lower healthcare costs. While more than 170 medications have germline genetic information in the drug label approved by the US FDA, pharmacogenomics (PGx) testing is not routinely used to guide prescribing in clinical settings [1,2]. PGx-guided prescribing offers the potential to optimize drug selection and dosing based on patient-specific genetic factors. However, in order for a gene–drug pair to be considered clinically actionable, there must be sufficient supporting evidence, an available genetic test and an alternative dose or drug for individuals with the ‘at-risk’ genotype. Moreover, a structured approach to implement PGx-guided prescribing must be in place [2].

The Clinical Pharmacogenomics Implementation Consortium (CPIC) publishes guidelines that systematically evaluate genotype and drug phenotype association data and translates this information into evidence-based recommendations for clinicians [3]. To date, CPIC has identified over 80 medications impacted by approximately 20 genes with level A or B evidence, whereby a prescribing action is recommended to avoid an adverse medication outcome and alternative therapies or dosing are highly likely to be effective and safe [2–4]. While approximately 7% of US FDA-approved medications have actionable PGx recommendations based on CPIC level A or B evidence, these medications have been estimated to represent 18% of the outpatient medications prescribed in the USA [2]. Furthermore, many of the at-risk genotypes that affect these medications are common, as 99% of individuals carry at least one actionable PGx genotype [5,6].

Currently, when PGx testing is performed, genotyping typically occurs for putatively functional genetic variants in a single gene and in a reactive manner after it has been determined that the patient will be prescribed a drug with actionable PGx recommendations. Notably, implementation of CYP2C19 genotype testing to guide the selection of antiplatelet therapy in patients with coronary artery disease (CAD) undergoing percutaneous coronary intervention (PCI) has become increasingly common [7,8]. It is now well established that patients with nonfunctional CYP2C19 variant alleles are at increased risk of adverse cardiovascular events after stent placement when treated with clopidogrel, which remains the most commonly used P2Y₁₂ inhibitor after PCI [9–12]. Consequently, use of alternative therapy (prasugrel or ticagrelor) is recommended in CYP2C19 nonfunctional allele carriers, which comprise approximately 30% of the US population [13].

Multigene preemptive PGx testing, in which a patient is simultaneously tested for multiple PGx actionable genes in advance of medication prescribing, likely offers advantages over single-gene testing due to decreasing genotype costs secondary to technological advancements and the potential benefits associated with PGx-guided prescribing [14]. However, the patient populations in which multigene preemptive PGx testing offers the greatest impact to avoid adverse drug outcomes have not been clearly defined, evaluated and validated. Due to the benefit of CYP2C19 genotype-guided antiplatelet therapy and high prevalence of polypharmacy among CAD patients due to advanced age and common comorbidities such as hyperlipidemia, hypertension, atrial fibrillation and depression, the cardiac catheterization laboratory offers potential to identify a high-risk population in which institutions can implement multigene PGx testing [15–18].

We hypothesize that PCI patients undergoing CYP2C19 genetic testing to guide antiplatelet therapy selection are also prescribed multiple medications, in addition to clopidogrel, that have actionable PGx recommendations for at-risk genotypes in CYP2C19 and other established pharmacogenes. However, it is not known how frequently this patient population is prescribed medications with actionable PGx recommendations and carry an at-risk genotype that increases risk for therapeutic failure or adverse events that could be avoided by preemptive PGx-based prescribing. Therefore, the objective of this study was to: describe the frequency of genetically actionable medication use beyond antiplatelet therapy in a real-world cohort of PCI patients that underwent CYP2C19 genetic testing; determine the proportion of PCI patients at risk for an adverse medication outcome based on their known CYP2C19 metabolizer phenotype; and evaluate the projected impact of multigene PGx testing on medication prescribing in CAD patients undergoing PCI.

Methods

Study design & population

This single-center, retrospective observational cohort study included 646 consecutive adult patients who underwent PCI with coronary artery stent placement at an academic medical center between 1 January 2015 and 31 December 2015. Patients were eligible for clinical CYP2C19 genetic testing at the interventional cardiologist’s discretion, which was clinically implemented in 2012 to guide antiplatelet therapy prescribing in high-risk patients [8,19]. Patients who died before discharge from their PCI hospitalization were excluded from this analysis. The study was approved by the Institutional Review Board. Due to the retrospective nature of the study, informed consent was not required.

Data abstraction

Demographic, clinical, medication and CYP2C19 genotype data were manually abstracted from encounters in the electronic health record (EHR). Physician-documented comorbid conditions were collected from the patient past medical history. CYP2C19 metabolizer phenotypes were assigned based on CPIC guidelines: ultrarapid metabolizer (UM; *17/*17), rapid metabolizer (RM; *1/*17), normal metabolizer (NM; *1/*1), intermediate metabolizer (IM; *1/*2, *1/*3, *2/*17 or *3/*17) or poor metabolizer (PM; *2/*2, *2/*3 or *3/*3) [20]. Medication use was reported for the total study population as well as within the stratum of the cohort that underwent CYP2C19 genotype testing.

Because antiplatelet therapy with aspirin and a P2Y₁₂ inhibitor (clopidogrel, prasugrel or ticagrelor) is indicated in all PCI patients and CYP2C19 genotype is already used clinically to guide antiplatelet therapy at our institution, the frequency of medication use beyond clopidogrel was the focus of the current analysis. Prescribed medications with CPIC level A or B evidence used to treat chronic medical conditions were collected at discharge from the index PCI hospitalization and at up to two follow-up encounters within 1 year of the index PCI (Supplementary Table 1) [3,4]. Outpatient cardiology and primary care provider (PCP) follow-up visits with a full medication history over 1 year were prioritized for data abstraction. Since PCI patients are treated by cardiologists, six additional medications used to treat cardiovascular diseases (CVD) with genetic information in the FDA label and CPIC level C–D evidence (propafenone, carvedilol, metoprolol, propranolol, rosuvastatin and hydralazine) were collected [1,4]. Medications used for the treatment of rare conditions and medications not commonly used in the USA were excluded due to the low likelihood of use within 1 year after PCI. Medications used transiently, such as antibiotics, antifungals and anesthesiology drugs and drugs with topical formulations, were also excluded because documentation of use in hospital discharge and outpatient encounters would likely underrepresent their actual use. Based on these criteria, five CPIC level A drugs and 31 level B drugs were excluded from data collection and 74 medications in total were included in data collection (Figure 1 & Supplementary Table 1).

Figure 1. . Overview of the number of pharmacogenomics actionable medications included in data collection.

The number of medications included in and excluded from data collection are described. The 74 medications included in data collection and the associated gene and CPIC level of evidence, are listed in Supplementary Table 1.

CPIC: Clinical Pharmacogenomics Implementation Consortium; CVD: Cardiovascular disease; PGx: Pharmacogenomics.

Data analysis

Descriptive statistics were used to assess demographics, clinical characteristics, CYP2C19 metabolizer phenotype and PGx actionable medication use in the full study population as well as the stratum that underwent CYP2C19 genetic testing. Data are reported as counts (%), mean ± standard deviation or median (interquartile range) unless otherwise indicated. Analyses were performed using SAS-JMP version 14.0 (SAS Institute, NC, USA).

The primary end point was the observed number of potential PGx-directed medication interventions that could have been made to optimize prescribing beyond antiplatelet therapy in the stratum of patients with CYP2C19 genotype results available (n = 380). A potential intervention was defined as medication with a clinically actionable recommendation prescribed to an individual that carried an at-risk CYP2C19 metabolizer phenotype (e.g., UM or PM). The total number of potential interventions, as well as the total number of unique patients with a potential intervention, were calculated to account for the possibility of multiple actionable CYP2C19 genotype-guided interventions within the same patient. The proportion of use by drug class and type of adverse outcome avoided (therapeutic failure or adverse event) were also calculated.

A secondary simulation analysis was conducted to evaluate the projected impact of a multigene PGx testing strategy on medication prescribing in the overall PCI patient population (n = 646). The simulation was designed to determine the number of PGx-guided interventions that could have been made based on genotype results in addition to CYP2C19. Actionable gene–drug pairs for which the drug was used in at least 2% of the study population were included. This resulted in inclusion of six genes in addition to CYP2C19 (CYP2C9, CYP2D6, G6PD, HLA-B, SLCO1B1 and VKORC1) and 19 medications in the analysis (Supplementary Table 2). At-risk genotype or phenotype frequencies, stratified by race, were abstracted from CPIC guideline supplements for CYP2C19 (UM, RM or PM), CYP2D6 (UM, IM or PM), SLCO1B1 (intermediate or low function), G6PD (deficient) and HLA-B (*58:01 carrier) [13,21–26]. For VKORC1/CYP2C9 (highly sensitive or sensitive warfarin responders), the ENGAGE AF-TIMI 48 trial was used in addition to the warfarin CPIC guideline to estimate phenotype prevalence [27,28]. For genes without metabolizer phenotypes, in which only allele frequencies were available, the Hardy–Weinberg equation was used to estimate the frequencies of the actionable at-risk genotypes. Because panel-based genotype results were not available among the patients in this cohort, the simulation analysis projected the population genotype/phenotype frequency estimates for each gene onto the PCI study population and assumed a population comprised of 80% Caucasian and 20% African–American patients. A potentially PGx actionable intervention was defined in the same manner as the primary outcome. The projected number of PGx-guided medication interventions that could have been made, if genotype results were available, was calculated by multiplying the observed frequency of medication use in the overall study population by the estimated at-risk genotype/phenotype frequency for each medication. The sum of the total number of potential PGx-directed interventions in the PCI study population (n = 646) and the projected number of interventions per 100 PCI patients were calculated for each gene and overall.

Results

Study population

The mean age of the PCI population was 63 ± 12 years, 70% were male and 20% were African–Americans (Table 1). Approximately half of the patients had an acute coronary syndrome indication for their current PCI, while 53% had a previous revascularization procedure. Common comorbid conditions included hypertension (88%), obesity (46%), diabetes (40%) and psychiatric disorders (18%), while active cancer and end-stage renal disease were prevalent in less than 5% of the cohort.

Table 1. . Study population baseline characteristics.

Characteristic	N (%)†
N	646
Age, years (mean ± standard deviation)	63.3 ± 11.9
Male	453 (70.1%)
Race/ethnicity
-Caucasian	461 (71.4%)
-African–American	128 (19.8%)
-Asian	3 (0.5%)
-Hispanic	22 (3.4%)
-Native American	19 (2.9%)
-Other/unknown	13 (2.0%)
Obese (body mass index >30 kg/m2)	300 (46.4%)
Current smoker	150 (23.2%)
Hypertension	568 (87.9%)
Diabetes	261 (40.4%)
Depression or other psychiatric disorder	117 (18.1%)
End-stage renal disease	20 (3.1%)
Active cancer	31 (4.8%)
History of myocardial infarction	195 (30.2%)
History of a prior revascularization procedure	345 (53.4%)
Acute coronary syndrome indication for PCI	315 (48.8%)
CYP2C19 genotype available	380 (58.8%)
-Previously determined genotype‡	92 (14.2%)
-Genotype during current admission§	288 (44.6%)

^†Data presented as N (%) unless otherwise stated.

^‡Genotype test performed at a prior PCI admission.

^§Genotype test performed during the index PCI admission.

PCI: Percutaneous coronary intervention.

Frequency of PGx-actionable medication use in PCI patients

Overall, clopidogrel was prescribed in 68% of the cohort, with an alternative P2Y₁₂ inhibitor (prasugrel or ticagrelor) prescribed in 32% of patients. The most commonly used PGx actionable medications in the study population beyond antiplatelet therapy were omeprazole, oxycodone, pantoprazole, tramadol, glipizide, hydrocodone, esomeprazole, warfarin, citalopram, simvastatin, ondansetron, allopurinol, sertraline, paroxetine, escitalopram, venlafaxine, amitriptyline, mirtazapine and glimepiride (Figure 2A & Supplementary Table 2). In addition, there were 18 PGx actionable medications with observed use in less than 2% of the population that were collectively prescribed to 50 (7.7%) patients. When evaluated by therapeutic area (Figure 2B), the proportion of PGx actionable medications prescribed were most common for gastrointestinal indications (37%, proton pump inhibitors [PPI] and ondansetron), pain (26%, opioids) and mental health disorders (19%, selective serotonin reuptake inhibitors [SSRI] and other antidepressants). Additional disease states with PGx actionable medication use included CVD (9.4%, warfarin and simvastatin), diabetes (7.9%, sulfonylureas) and gout (3.6%, allopurinol). Of the CVD medications with FDA labeling information and CPIC level C–D evidence, metoprolol (59%) and carvedilol (19%) were the most commonly used, followed by rosuvastatin (7.4%), hydralazine (3.7%) and propranolol (1.2%) (Supplementary Figure 1).

Figure 2. . Frequency of multigene pharmacogenomics actionable medication use beyond antiplatelet therapy in percutaneous coronary intervention patients.

(A) Use frequency of PGx actionable medications in the overall study population (n = 646). Medication use reflects the number of patients with a prescription for the given medication at discharge from the index percutaneous coronary intervention hospitalization (dark) or initiation during a follow-up visit up within 1-year post discharge (light). Medications used in ≥2% of patients and the associated gene are presented. (B) The distribution of PGx actionable medication use by medication therapeutic area: GI, pain, mental health, cardiovascular (not including antiplatelet therapy and the CPIC Level C–D drugs described in Supplementary Figure 1), diabetes, gout and other (immunosuppressants, antiviral, cancer, cystic fibrosis and anticonvulsants).

^†Other medications is a composite of 18 individual medications with <2% use frequency in the study population (listed in Supplementary Table 2)

CPIC: Clinical Pharmacogenomics Implementation Consortium; GI: Gastrointestinal; PGx: Pharmacogenomics.

Observed number of potential CYP2C19 genotype-guided medication interventions

CYP2C19 genotype was available in 380 (59%) patients following the index PCI procedure, with 92 (24%) patients having undergone genotype testing at a prior PCI procedure. Among the 288 patients genotyped at the time of the index PCI, the median (interquartile range) genotype turnaround time was 1 (0–2) day and results were available for 199 (69%) patients by the day after the test was ordered. Overall, the distribution of CYP2C19 metabolizer phenotypes was 5% UM, 28% RM, 39% NM, 27% IM and 2% PM (Figure 3A).

Figure 3. . Observed frequency of CYP2C19 phenotypes and actionable medication use beyond antiplatelet therapy in genotyped percutaneous coronary intervention patients.

(A) CYP2C19 phenotype distribution in 380 genotyped patients: UM (n = 17; 4%), RM (n = 106; 28%), NM (n = 148; 39%), IM (n = 102; 27%) and PM (n = 7; 2%). The IM (*1/*2 = 73 [72%], *1/*3 = 0 [0%], *2/*17 = 29 [28%], *3/*17 = 0 [0%]) and PM (*2/*2 = 7 [100%], *2/*3 = 0 [0%], *3/*3 = 0 [0%]) phenotypes included multiple genotypes. (B) Use frequency of PGx actionable medications affected by CYP2C19 metabolism beyond antiplatelet therapy in CYP2C19 genotyped patients, sorted by medication class. Rabeprazole and clomipramine use was 0% and thus not included in the graph.

IM: Intermediate metabolizer; NM: Normal metabolizer; PGx: Pharmacogenomics; PM: Poor metabolizer; RM: Rapid metabolizer; UM: Ultrarapid metabolizer.

Among the 380 PCI patients with a CYP2C19 genotype result, 194 (51%) were prescribed one or more of nine distinct PGx actionable medications, beyond antiplatelet therapy, affected by CYP2C19 alleles (Figure 3B). As a result of 35 (9.2%) patients taking more than one PGx actionable medication, there were 229 total instances of a CYP2C19 genotype-affected medication being prescribed to a patient with a genotype result available in the EHR. As summarized in Table 2, we observed 30 instances in which a CYP2C19 UM, RM or PM was prescribed CYP2C19-metabolized medication (7.9 per 100 patients). As a result of three patients taking more than one medication, an actionable PGx intervention based on the genotype result could have been made in 27 (7.1%) of the 380 genotyped patients (13.9% of the 194 genotyped patients taking a medication affected by CYP2C19 metabolizer status). The majority of the projected interventions would have served to avoid a potential therapeutic failure (87%) in a CYP2C19 UM or RM patient, with 13% preventing a potential adverse event in a PM (Table 2).

Table 2. . Summary of potential CYP2C19 genotype-guided medication interventions.

Drug class	Medication	Number of patients prescribed medication	At-risk CYP2C19 phenotype group	At-risk CYP2C19 phenotype frequency	Number of potential actionable PGx interventions†	Avoided clinical outcome
PPI	Esomeprazole, Omeprazole, Lansoprazole, Pantoprazole, Dexlansoprazole and Rabeprazole	165	UM PM	4.5% 1.8%	7 3	Therapeutic failure Adverse event
SSRI	Citalopram, Escitalopram and Sertraline	58	UM/RM PM	32.4% 1.8%	18 1	Therapeutic failure Adverse event
TCA	Amitriptyline	6	UM/RM PM	32.4% 1.8%	1 0	Therapeutic failure Adverse event
	Total	229			30

^†The number of potential actionable PGx interventions count the number of observed instances where a CYP2C19 genotyped patient (n = 380) was prescribed a CPIC level A or B evidence medication and carried an actionable CYP2C19 phenotype.

CPIC: Clinical Pharmacogenomics Implementation Consortium; PGx: Pharmacogenomics; PPI: Proton pump inhibitor; PM: Poor metabolizer; RM: Rapid metabolizer; SSRI: Selective serotonin reuptake inhibitor; TCA: Tricyclic antidepressant; UM: Ultrarapid metabolizer.

Projected number of potential PGx-guided interventions following multigene testing

A simulation analysis estimating the potential impact of multigene testing for CYP2C19, CYP2D6, G6PD, VKORC1/CYP2C9, SLCO1B1 and HLA-B in the overall study population (n = 646) projected that 113 total interventions (17.5 per 100 patients) beyond antiplatelet therapy could have been made based on actionable PGx results (Figure 4). A detailed summary, by gene and drug, is provided in Supplementary Tables 3–8.

Figure 4. . Projected number of actionable pharmacogenomics interventions beyond antiplatelet therapy by gene in percutaneous coronary intervention patients.

Based on a simulation analysis including the 19 most frequently prescribed medications and corresponding seven genes with actionable PGx recommendations, the projected number of potential PGx interventions were calculated as summarized in greater detail for each gene in Supplementary Tables 3–8. Data in the figure are presented, by gene, as the projected number of actionable interventions per 100 patients.

PGx: Pharmacogenomics.

Among the 396 instances of CYP2C19 metabolized medication use, we estimated 55 projected PGx interventions beyond antiplatelet therapy (8.5 per 100 patients; Supplementary Table 3). The most common projected interventions involved prevention of a potential SSRI (27 interventions) and PPI (13 interventions) therapeutic failure in UMs or RMs and avoidance of a potential PPI adverse effect in PMs (eight interventions). The number and distribution of projected interventions in the simulation analysis was consistent with our observed data in patients with available CYP2C19 genotype results.

Among the 314 instances of CYP2D6-metabolized medication use, we estimated 29 projected PGx interventions (4.5 per 100 patients; Supplementary Table 4). The projected interventions involved antidepressants (amitriptyline, paroxetine and venlafaxine) in CYP2D6 UMs, IMs and PMs (seven interventions), opioids (tramadol, oxycodone and hydrocodone) in CYP2D6 UMs and PMs (21 interventions) and ondansetron in CYP2D6 UMs (one intervention). Among the 44 and 41 patients taking warfarin and simvastatin, respectively, we projected 14 potential warfarin dose-optimization interventions in VKORC1/CYP2C9 sensitive or highly sensitive responders (2.2 per 100 patients; Supplementary Table 5) and ten potential interventions in patients with a SLCO1B1 intermediate or low function phenotype that could be prescribed a lower dose or alternative statin (1.5 per 100 patients; Supplementary Table 6). We also projected four potential interventions in patients with G6PD deficiency taking glipizide or glimepiride (0.6 per 100 patients; Supplementary Table 7) and one potential intervention for a patient taking allopurinol (0.2 per 100 patients; Supplementary Table 8), which is contraindicated in HLA-B*58:01 allele carriers [29].

Discussion

The projected impact of CYP2C19 and multigene PGx testing on medications beyond clopidogrel was evaluated in a real-world cohort of patients that underwent PCI and CYP2C19 testing. Our results demonstrate that medications with actionable PGx recommendations across multiple drug classes, most notably PPIs, antidepressants and opioids, are commonly prescribed in PCI patients. Among patients with an available CYP2C19 genotype result in the EHR, approximately 50% were prescribed a CYP2C19-metabolized medication beyond clopidogrel and 7% met criteria for a CYP2C19 genotype-guided medication intervention. Further, our simulation analysis projected that approximately 17.5 medication interventions per 100 PCI patients could have been recommended if multigene PGx test results were available. Taken together, these results suggest that multigene PGx testing could be used to optimize prescribing for multiple medications beyond CYP2C19 genotype-guided antiplatelet therapy in PCI patients and warrants further study.

Due to the benefit of CYP2C19 genotype-guided antiplatelet therapy after PCI, use of CYP2C19 testing in clinical practice to optimize antiplatelet therapy after PCI has become increasingly common [7,10,12]. Our results demonstrate that an opportunity exists to optimize the use of medications beyond clopidogrel based on the results of a CYP2C19 genotype test that has already been performed. Approximately 50% of PCI patients in our cohort were prescribed a CYP2C19-metabolized PPI or antidepressant at either discharge or an outpatient encounter within 1 year, with approximately 7% of patients qualifying for an evidence-based CYP2C19 genotype-guided medication intervention. In patients prescribed a PPI, interventions would include increasing the PPI dose in CYP2C19 UM patients to prevent therapeutic failure and selecting an alternative PPI in CYP2C19 PM patients to avoid an adverse event [30]. In patients prescribed a SSRI, interventions would include selecting an alternative agent in CYP2C19 UM or RM patients to avoid therapeutic failure and either a dose reduction or alternative agent in CYP2C19 PM patients to avoid a potential adverse event [23].

Because it has been estimated that approximately 99% of the population carries at least one at-risk PGx genotype, identifying patient populations that are commonly prescribed PGx actionable medications is an important prerequisite for efficient implementation of multigene PGx testing [5,6]. In addition to PPIs and antidepressants, opioids for chronic pain, warfarin for thromboembolic disease and simvastatin for hyperlipidemia were also commonly prescribed in PCI patients. Our simulation analysis projected that multigene testing in our study PCI population could have yielded 17.5 PGx-guided medication interventions per 100 patients, of which approximately half were derived from CYP2C19. The primary genes in addition to CYP2C19 yielding actionable results were CYP2D6 for antidepressants and opioids, VKORC1/CYP2C9 for warfarin and SLCO1B1 for simvastatin. These observations are consistent with two recent analyses that demonstrated multigene testing with CYP2C19, SLCO1B1, VKORC1 and CYP2C9 and multigene testing with CYP2C19 and CYP2D6 was more cost effective than CYP2C19 testing alone in patients undergoing PCI [31,32]. The projected number of PGx interventions in our study population to avoid toxicity in G6PD-deficient patients prescribed a sulfonylurea was low (0.6 interventions per 100 patients); however, most commercial multigene PGx panels do not include testing for G6PD deficiency. Taken together, our results suggest the cardiac catheterization laboratory is a logical ‘foot in the door’ setting for institutions to efficiently implement preemptive multigene PGx testing and can be used to inform the design of prospective studies that evaluate the feasibility and impact of this precision medicine strategy in PCI patients.

The results of our study demonstrates that approximately one of every six PCI patients met qualifications for a PGx-guided medication intervention beyond antiplatelet therapy may underestimate the value of preemptive genotyping since our analysis only considered prescribing during the 12 months after PCI. The number of interventions would be expected to increase over time as patients age and are prescribed additional medications. Furthermore, additional commonly prescribed medications with CPIC level C evidence, such as β-blockers, may be reclassified as clinically actionable as evidence supporting clinical use expands. Considering additional medications with PGx actionable recommendations by the Dutch Pharmacogenetics Working Group would also likely increase the number of potential PGx-guided interventions in PCI patients. Our results also demonstrate that most PGx-guided interventions would serve to prevent a potential therapeutic failure. While the clinical impact of such medication optimization interventions might not be immediate, the long-term effects could be significant to patients, providers and payers.

Notably, SSRIs are the first-line treatment for depression and anxiety and treatment failures can lead to significant health consequences for patients and increased healthcare costs that are especially problematic in the elderly [33]. Moreover, depression is more common in women than men and suboptimal management of depression is associated with adverse cardiovascular outcomes [34]. Given that over a third of the US population carries either a CYP2C19 or CYP2D6 UM or RM phenotype, PGx-guided strategies that enable more precise initial selection and dosing of antidepressants offer enormous potential to prevent or reduce the duration and impact of uncontrolled depressive symptoms [23], particularly in women, the elderly and other subsets of the population with a high frequency of antidepressant prescriptions. Furthermore, CYP2D6 genotype-guided opioid prescribing has been shown to improve pain control, with emerging data also showing benefit in CYP2D6 IMs and PMs [35]. While the effect of CYP2C19 genotype on PPI pharmacokinetics has been well documented, the clinical significance on clinical outcomes remains less clear [36]. It has been suggested that genotype-guided PPI prescribing has clinical benefit in the treatment of gastroesophageal reflux disease and Helicobacter pylori eradication, as well as the prevention of PPI-related adverse events such as infection and acute kidney injury [30]. Given the high prevalence of PPI use in PCI patients receiving antiplatelet therapy, studies evaluating the clinical effectiveness of CYP2C19 genotype-guided PPI prescribing are still needed.

As multigene PGx testing becomes increasingly common in clinical practice, a more comprehensive assessment of multiple medications (rather than a single medication) will be necessary to effectively use each patient’s genetic test results to optimize prescribing decisions and minimize healthcare costs. Notably, our results showed that use of noncardiac medications in PCI patients is common. While this represents an opportunity for medication optimization to occur across a patient’s entire medication list, many PGx actionable medications are used to treat psychiatric and pain conditions that are not typically managed by a cardiologist. Clinical pharmacists, clinical decision support tools and multidisciplinary collaboration are instrumental to improve the adoption and appropriate use of PGx-based medication optimization strategies and facilitate transitions of care in patients with a multitude of comorbidities and medications [37,38]. Innovative precision medicine services, such as a precision medicine pharmacist in cardiology or multiclinic settings, will become especially important as the frequency of PGx testing increases and new evidence linking genetic variation to suboptimal medication outcomes continues to emerge.

There are several limitations to this study that should be acknowledged. The simulation analysis was limited by the use of projected at-risk genotype frequencies and a probability-based estimation of potential PGx interventions in the overall population, rather than patient-level data based on observed genotype results and medication use. However, the simulated CYP2C19 analysis yielded very similar results to the observed analysis in patients with an available CYP2C19 genotype, which increases confidence in our results. While the frequency of patients at risk for a PGx-associated adverse medication outcome were calculated, direct evaluation of therapeutic failures and adverse effects for each medication in individual patients were not completed in the current study, which is a limitation. The impact of PGx-guided prescribing on adverse medication outcomes will require prospective investigation in subsequent studies. Due to the retrospective design, the observed medication use frequency may not reflect more contemporary prescribing patterns. Although medication use was evaluated at multiple encounters over time, misclassification bias cannot be excluded. Further, medications used transiently, such as antibiotics, antifungals and anesthesiology drugs, were not included since our data abstraction strategy would likely underestimate their actual use. Notably, the antifungal voriconazole has actionable prescribing recommendations based on CYP2C19 genotype results [39]. Thus, exclusion of these medications from our analysis likely underestimated the projected impact of PGx testing in this population. Additionally, some patients likely receive some medications for noncardiac disease states at other centers that may not be documented in the EHR, further underestimating the number of actionable PGx recommendations. At last, although the study population was derived from an academic, tertiary referral center with a population that is approximately 20% African–American and representative of surrounding areas, medication use frequency at this single-center may not be generalizable to other PCI populations. Notably, our simulation analysis did not project the impact of multigene PGx testing in Asian populations, which may have further underestimated the projected impact of PGx testing due to the higher prevalence of CYP2C19 IMs and PMs compared with Caucasian and African–American populations. Furthermore, we did not evaluate the impact of gender on the frequency of actionable PGx recommendations, which is an important future direction due to the higher prevalence of antidepressant use in women compared with men. Prospective studies in larger and more diverse multicenter populations and over longer time horizons after PCI are warranted to assess the projected impact of PGx testing on medication prescribing overall and within key demographic subsets such as race/ethnicity, gender and age groups. In addition, future studies evaluating the projected impact of multigene PGx testing on medication optimization beyond PCI patients and across multiple clinical settings are needed.

In summary, the current study demonstrated that medications across multiple drug classes with actionable PGx recommendations are commonly prescribed in PCI patients receiving antiplatelet therapy. The projected impact of PGx testing on optimal medication prescribing beyond clopidogrel in this real-world patient population suggests that multigene PGx testing could be beneficial among PCI patients who are already indicated for CYP2C19 genotyping and the cardiac catheterization laboratory could serve as a logical setting for institutions to efficiently implement multigene PGx testing into routine clinical practice. Future prospective studies evaluating the feasibility and effectiveness of multigene PGx testing in PCI patients are needed.

Summary points.

• The patient populations in which pharmacogenomics (PGx) testing offers the greatest potential impact on medication prescribing remain unclear.

• We hypothesized that percutaneous coronary intervention (PCI) patients are prescribed multiple medications, in addition to antiplatelet therapy, that have actionable PGx recommendations for at-risk genotypes in CYP2C19 and other pharmacogenes.

• This study evaluated the potential impact of CYP2C19 and multigene PGx testing on medications beyond antiplatelet therapy in a real-world cohort of PCI patients that underwent CYP2C19 testing.

• Our results demonstrate that medications with actionable PGx recommendations across multiple drug classes, most notably proton pump inhibitors, antidepressants and opioid analgesics, are also commonly prescribed in PCI patients.

• Among PCI patients with a CYP2C19 genotype result available, 51% were prescribed one or more CYP2C19-metabolized medications beyond clopidogrel and 7% met criteria for a CYP2C19 genotype-guided medication intervention.

• A simulation analysis projected that 113 interventions (17.5 per 100 PCI patients) could have been recommended in the study population if multigene PGx results were available. Approximately half of the interventions would be guided by genes other than CYP2C19.

• These results suggest that CYP2C19 and multigene PGx testing could be used to optimize prescribing for multiple medications beyond CYP2C19 genotype-guided antiplatelet therapy in PCI patients and warrants further study.

• As new evidence emerges and PGx testing becomes increasingly common in practice, clinical pharmacists, clinical decision support tools and inter-disciplinary collaboration will be instrumental in the adoption and appropriate use of PGx-based prescribing.