Genetic testing in patients with acute coronary syndrome undergoing percutaneous coronary intervention: a cost-effectiveness analysis

Authors

  • A. LALA,

    1. Department of Medicine, Division of Cardiovascular Medicine, New York University School of Medicine
    Search for more papers by this author
    • These authors contributed equally to this work.

  • J. S. BERGER,

    1. Department of Medicine, Division of Cardiovascular Medicine, New York University School of Medicine
    2. Department of Medicine, Division of Hematology, New York University School of Medicine
    Search for more papers by this author
    • These authors contributed equally to this work.

  • G. SHARMA,

    1. Department of Medicine, Division of Cardiovascular Medicine, New York University School of Medicine
    Search for more papers by this author
  • J. S. HOCHMAN,

    1. Department of Medicine, Division of Cardiovascular Medicine, New York University School of Medicine
    Search for more papers by this author
  • R. SCOTT BRAITHWAITE,

    1. Department of Population Health, Section on Value and Effectiveness, New York University School of Medicine, New York, NY, USA
    Search for more papers by this author
  • J. A. LADAPO

    1. Department of Population Health, Section on Value and Effectiveness, New York University School of Medicine, New York, NY, USA
    Search for more papers by this author

Jeffrey S. Berger, Director of Cardiovascular Thrombosis, NYU Langone Medical Center, 530 First Avenue, SK 9R, New York, NY 10016, USA
Tel.: +1 212 263 4004; fax: +1 212 263 3988.
E-mail: jeffrey.berger@nyumc.org

Abstract

Summary.  Background:  The CYP2C19 genotype is a predictor of adverse cardiovascular events in acute coronary syndrome (ACS) patients undergoing percutaneous coronary intervention (PCI) treated with clopidogrel. Objectives:  We aimed to evaluate the cost-effectiveness of a CYP2C19*2 genotype-guided strategy of antiplatelet therapy in ACS patients undergoing PCI, compared with two ‘no testing’ strategies (empiric clopidogrel or prasugrel). Methods:  We developed a Markov model to compare three strategies. The model captured adverse cardiovascular events and antiplatelet-related complications. Costs were expressed in 2010 US dollars and estimated using diagnosis-related group codes and Medicare reimbursement rates. The net wholesale price for prasugrel was estimated as $5.45 per day. A generic estimate for clopidogrel of $1.00 per day was used and genetic testing was assumed to cost $500. Results:  Base case analyses demonstrated little difference between treatment strategies. The genetic testing-guided strategy yielded the most QALYs and was the least costly. Over 15 months, total costs were $18 lower with a gain of 0.004 QALY in the genotype-guided strategy compared with empiric clopidogrel, and $899 lower with a gain of 0.0005 QALY compared with empiric prasugrel. The strongest predictor of the preferred strategy was the relative risk of thrombotic events in carriers compared with wild-type individuals treated with clopidogrel. Above a 47% increased risk, a genotype-guided strategy was the dominant strategy. Above a clopidogrel cost of $3.96 per day, genetic testing was no longer dominant but remained cost-effective. Conclusions:  Among ACS patients undergoing PCI, a genotype-guided strategy yields similar outcomes to empiric approaches to treatment, but is marginally less costly and more effective.

Introduction

Dual antiplatelet therapy with aspirin and a P2Y12 receptor antagonist is the mainstay of pharmacologic treatment in patients with ACS undergoing PCI [1–3]. The Trial to Assess Improvement in Therapeutic Outcomes by Optimizing Platelet Inhibition with Prasugrel–Thrombolysis in Myocardial Infarction (TRITON-TIMI 38) demonstrated the superior efficacy of prasugrel compared with clopidogrel in the reduction of major cardiovascular events, but patients treated with prasugrel also experienced more major bleeding [4].

Several large trials have identified loss-of-function alleles in the gene encoding cytochrome P4502C19 (CYP2C19) that predict a substantial reduction in the efficacy of clopidogrel [5–9]. The CYP2C19*2 allele (population prevalence of 20% to 65%) [10] causes a splicing defect and complete loss of enzyme activity; this presents phenotypically as poor response to clopidogrel. In the TRITON genetic sub-study among participants randomized to clopidogrel, carriers of the reduced function allele CYP2C19*2 had lower levels of the active metabolite of clopidogrel, diminished platelet inhibition and a significantly higher rate of major adverse cardiovascular events compared with wild-type individuals [8]. This finding was further supported by a meta-analysis of nine clinical trials (primarily of patients undergoing PCI), which concluded that carriers of the loss-of-function alleles were > 50% more likely to experience major cardiovascular events compared with wild-type individuals [11].

In response to these and other studies, the US Food and Drug Administration (FDA) announced a black box warning in March 2010 to alert physicians to the reduced effectiveness of clopidogrel in subjects who are poor clopidogrel metabolizers [12]. The FDA also noted the availability of genetic testing and advised physicians to consider using other antiplatelet agents or dosing strategies in patients identified as poor metabolizers.

Despite the FDA warning and the risk that clopidogrel non-responders face when on clopidogrel, it is still widely used. Furthermore, the clinical benefit of treating carriers of CYP2C19*2 with prasugrel instead of clopidogrel has not been adequately studied [8,11,13]. Prasugrel is also significantly more expensive than clopidogrel, and this difference will be magnified with the introduction of generic clopidogrel. Genetic testing also has unique associated costs, and the implications of its use in routine clinical practice are unclear. We therefore sought to evaluate the balance between potential benefits and healthcare expenditures of routine genetic testing-guided treatment in ACS patients undergoing PCI, compared with empirically treating all of these patients with either prasugrel or clopidogrel.

Methods

Decision model

We used a computer simulation program, TreeAge Pro (TreAge Pro Suite 2010, Williamstown, MA, USA), to develop a Markov model, analyze decision trees and perform sensitivity analyses. Our decision analysis compared three strategies: in addition to aspirin (i) treat all patients with clopidogrel and do not perform genetic testing; (ii) treat all patients with prasugrel and do not perform genetic testing; and (iii) perform genetic testing to determine which antiplatelet regimen to use (subjects with a loss of function allele were treated with prasugrel while wild-type subjects were treated with clopidogrel). Our base case was a hypothetical cohort of patients aged 60 years old (mean age in TRITON TIMI-38) with ACS undergoing PCI. We adopted a payer perspective and applied a 3% annual discount rate to costs and health utilities. The results of the analysis are expressed in terms of monetary costs (2010 US dollars), quality-adjusted life-years (QALYs) and incremental cost-effectiveness ratios (ICERs). We define the optimal or most cost-effective strategy as having the highest QALY benefit with an incremental cost-effectiveness ratio less than the US societal willingness-to-pay threshold ($100 000 per QALY gained) [14].

Accuracy of genetic testing  The pharmacogenetic-based strategy involved genotyping specifically for the CYP2C19*2 polymorphic variant, which accounts for 95% of loss of function mutations [8]. The sensitivity and specificity of the test were determined to be 99% and 99%, respectively, based on available data [15,16]. These values were varied in sensitivity analyses.

Clinical pathways  All post-ACS patients entered an event-free-survival state (Fig. 1). They could then remain in this state or transition to one of several other states, including (i) survival after one additional event, asymptomatic, (ii) survival after one additional event, symptomatic, (iii) survival after two or more events, (iv) cardiovascular death, (v) major bleed-related death, or (vi) other death. Mode of death, bleeding and thrombotic clinical events were adjudicated. If a non-fatal event occurred, it was further characterized based on severity as being symptomatic or asymptomatic (e.g. clinical MI vs. a biomarker-only MI). The severity of the non-fatal event determined whether a patient transitioned to a survival after one additional event, asymptomatic, state or a survival after one additional event, symptomatic, state, and each state had distinct implications for cost and quality of life. We assumed that once a patient had one non-fatal event and entered a survival after one additional event state (asymptomatic or symptomatic), they would be twice as likely to have a subsequent event (MI, cerebral event, major bleed, CV death or major bleed death). Subsequently it was assumed that once a patient entered the survival after two or more events state they would be four times more likely to have subsequent events [17–20] (Table 1).

Figure 1.

 Decision model analysis. CV = cardiovascular, MI = myocardial infarction, CVA = cerebral vascular accident, TIA = transient ischemic attack. *Survival branch corresponds to originating Markov state.

Table 1.   Baseline assumptions
VariableBase-case value (range)DistributionReference
  1. *Event probabilities and relative risks are aggregated across patients who are wild types and carriers, as reported in TRITON-TIMI 38 trial.

  2. Varies by age. Values provided for 60-year-old patients.

  3. Symptomatic MI was considered an MI that was a clinical MI (biomarkers and symptoms).

  4. §Range given for one-way sensitivity analyses. Does not correspond precisely to range for probabilistic sensitivity analysis.

Monthly rate of events on prasugrel, %*
 Non-fatal MI
  First 30 days4.46 (4.10–5.10)Beta 
  After 30 days0.19 [4,12,21]
 Non-fatal cerebral event
  First 30 days0.79 (0.58–1.00)Beta 
  After 30 days0.03 [4,12,21]
 Non-fatal major bleed
  First 30 days1.18 (0.92–1.44)Beta 
  After 30 days0.05 [4,12,21]
 Cardiovascular death
  First 30 days1.15 (0.90–1.40)Beta 
  After 30 days0.07† [4,12,21,26]
 Major-bleed-related death
  First 30 days0.13 (0.04–0.21)Beta 
  After 30 days0.01† [4,12,21,26]
 Other death (non-CV, non-major-bleed-related death)0.05†NA[26]
Probability that event is symptomatic‡
 MI on prasugrel
  First 30 days0.18 (0.10–0.35)Beta 
  After 30 days0.90 (0.60–1.00) [4,12,21]
 MI on clopidogrel
  First 30 days0.23 (0.10–0.35)Beta 
  After 30 days0.90 (0.60–1.00) [4,12,21]
 Cerebral event on prasugrel0.71 (0.50–1.00)Beta[4,12,21]
 Cerebral event on clopidogrel0.67 (0.40–0.80)Beta[4,12,21]
 Life-threatening bleed on prasugrel0.51 (0.30–0.80)Beta[4,12,21]
 Life-threatening bleed on clopidogrel0.48 (0.30–0.66)Beta[4,12,21]
Quality-of-life estimates (utility)
 Event-free survival0.83† (0.80–0.85)Uniform[27]
 Survival after 1 additional event0.72 (0.65–0.80)Uniform[27,28]
 Survival after 1 additional event (symptomatic)0.55 (0.47–0.64)Uniform[27]
 Survival after 2 or more events0.50 (0.35–0.60)Uniform[27]
 Non-fatal MI0.70† (0.56–0.84)Uniform[27]
 Non-fatal cerebral event0.65† (0.52–0.78)Uniform[27]
 Major bleed0.80† (0.64–0.96)Uniform[28]
Relative risk of subsequent adverse events
 Increased probability of having subsequent event after having an initial event2 (1–4)NA[17–20]
 Increased probability of subsequent event after having 2 or more events (symptomatic or asymptomatic)4(2–6)NA[17–20]
Genetic test characteristics
 Sensitivity0.99 (0.85–1.00)NA[15,16]
 Specificity0.99 (0.85–1.00)NA[15,16]
 Population prevalence of carriers0.27 (0.10–0.40)NA[8]
Event costs
 MI
  Symptomatic$30 737 (24 881–36 592)Uniform 
  Asymptomatic$18 108 (15 000–21 000) [40,41]
 Cerebral event
  CVA$36 797 (18 046–63 352)Uniform 
  TIA$16 572 (13 000–19 000) [40,41]
 Major bleed
  Life-threatening$27 671 (20.729–34 612)Uniform 
  Non-life-threatening$14 735 (11 000–18 000) [40,41]
 CV death$23 246 (12 936–37 286)Uniform[40–42]
 Major-bleed-related death$11 761 (5388–17 244)Uniform[43]
 Other death (non-CV, non-major-bleed-related)$11 761 (5388–17 244)Uniform[43]
 Genetic test$500 ($60–$750)Gamma§[30]
 Clopidogrel (monthly)$30.00 ($30.00–$176.10)Gamma§[29]
 Prasugrel (monthly)$168.60 ($124.50–$207.30)Uniform[29]
 Event-free survival (monthly)$307.78† ($295.80–1613.80)Uniform[44]
 Survival after 1 additional event (monthly)$0.00 (0–1000) + cost of event free survival (monthly)Uniform[44]
 Survival after 1 additional event with symptoms (monthly)$776† (130–6000) + cost Event Free Survival (monthly)Uniform[28,44]
 Survival after 2 or more events (monthly)$776† (602–8822)+ cost Event Free Survival (monthly)Uniform[28,44]

The risks for the adverse events included in our model were primarily derived from event rates provided by the FDA [12,21]. Our model tracked QALYs, incidence of adverse events and net cost over 15 months and 10 years. The 15-month time horizon was used because this mirrored the period of data reported in the TRITON-TIMI 38 trial in which prasugrel was found to be superior to clopidogrel. For long-term analyses, the 10-year time horizon was used because this is a common time horizon for decision-making. All subjects remained on dual antiplatelet therapy for 120 months irrespective of whether they suffered bleeding events. We assumed a constant magnitude of benefit of antiplatelet therapy and constant rate of adverse events from antiplatelet therapy after the first 30 days because Kaplan–Meier curves suggested that events were comparatively rare after 30 days and their rate was relatively constant: 63% of all major events (including non-fatal MI, cerebral events, major bleeds, major bleed-related deaths and cardiovascular deaths) occurred in the first 30 days.

Model inputs

Probabilities of adverse outcomes  Impact of genotype on risk of events: Using published data [8], we derived risk ratios for adverse cardiovascular event rates in carriers of the CYP2C19*2 allele (risk ratio = 1.32) and wild-type subjects (risk ratio = 0.88) on clopidogrel vs. prasugrel, relative to the overall cohort (comprising 27.1% carriers and 72.9% wild-type individuals) (Table 2). This 50% increased risk (relative risk = 1.32/0.88 = 1.5) is consistent with other published estimates in the setting of PCI for ACS [11]. For example, compared with subjects receiving prasugrel, a wild-type patient treated with clopidogrel will have a 16% increased risk of non-fatal MI, whereas carriers will have a 74% increased risk of non-fatal MI (Table 2). Among subjects receiving prasugrel, there was no significant difference in event rates for wild-type patients or carriers between the no testing strategy and genetic testing strategy [22].

Table 2.   Relative risk of events for patients treated with clopidogrel vs. prasugrel based on TRITON-TIMI 38 [8]
 All comers*Wild-type†Carriers†
  1. Note: distributions for these variables were established as Log Normal.

  2. *Overall cohort and comprises 27.1% carriers and 72.9% wild-types.

  3. Confidence intervals for adjusted risks were derived by converting risk ratios for ‘All comers’ to their normal distributions, applying the adjustment, and exponentiating to recover the risk ratio.

  4. The upper bound of this relative risk was truncated.

Non-fatal MI1.32 (1.18–1.49)1.16 (1.091–1.237)1.74 (1.380–2.200)
Non-fatal stroke0.98 (0.69–1.41)0.86 (0.063–11.826)1.29 (0.014–12.00‡)
Non-fatal major bleed0.76 (0.60–0.97)0.67 (0.470–0.951)1.00 (1.000–1.006)
CV death1.12 (0.89–1.43)0.99 (0.956–1.016)1.48 (0.652–3.350)
Major-bleed-related death0.24 (0.09–0.63)0.21 (0.073–0.610)0.32 (0.145–0.694)

Non-fatal MI: The risk of non-fatal MI was derived from the TRITON-TIMI 38 trial and was adjusted for carriers of the loss-of-function allele and wild-type patients receiving clopidogrel.

For the purposes of our study, an asymptomatic MI was considered a biomarker-positive only MI, and a symptomatic MI was synonymous with a clinical MI. Because the relative risk for non-fatal MI for clopidogrel vs. prasugrel may be different for biomarker-positive only (asymptomatic) vs. clinical (symptomatic) MIs, and symptomatic MIs have greater impact on morbidity and mortality, we performed sensitivity analyses in which we varied the incidence of MI based on clinical severity.

Patients entered the survival after one additional event, asymptomatic, state after an asymptomatic MI and the survival after one additional event, symptomatic, state after a symptomatic MI (Fig. 1). The distribution of asymptomatic and symptomatic MIs was derived from FDA data [12]. The probability that an MI is symptomatic in patients on prasugrel or clopidogrel in the first 30 days was 18% or 23%, respectively [21], but was 90% for both after 30 days. The assumption that 10% of MIs were asymptomatic was based upon prior published studies and varied in the model to account for its effect (Table 1) [23,24].

Cerebrovascular event: A cerebrovascular event was defined as an ischemic or hemorrhagic cerebrovascular accident (CVA) or transient ischemic attack (TIA). For the purposes of our analysis, a TIA was considered an asymptomatic event (less severe) and an ischemic or hemorrhagic CVA was considered symptomatic. Patients transitioned to the survival after one additional event, asymptomatic, state after a TIA and the survival after one additional event, symptomatic, state after a CVA.

Major bleed events: The TIMI major bleeding definition was used for major bleeding events [25]. A life-threatening, non-fatal bleed was considered symptomatic and patients who experienced this event transitioned to the survival after one additional event, symptomatic, health state. The risk of experiencing a fatal major bleed in the first 30 days was 0.13% on prasugrel and fell to 0.01% every month thereafter. This proportional risk (relative proportion of death in TRITON accounted for by cardiovascular death and fatal bleeding was 73.5% and 11%, respectively) was applied to overall death rates from US Life Tables to account for the increased rate of death with advancing age [26].

Cardiovascular and non-cardiovascular death: The annual rate of death in patients with ACS undergoing PCI was 2.83% in TRITON-TIMI 38. Because 73.5% of these deaths were attributable to a cardiovascular cause, the rate of cardiovascular death was estimated to be 1.15% for the first 30 days and 0.07% per month thereafter (Table 1). This risk was also applied to overall death rates from US Life Tables to account for the increased rate of death with advancing age (Table 1).

Health-related quality-of-life  Quality-of-life was estimated using the EuroQol-5D and other published literature [27,28] and expressed in health state utilities. Adverse events were associated with utility decrements and based on a multiplicative utility function for patients who experienced multiple adverse events. We used the relative prevalence of events in TRITON-TIMI 38 to estimate weighted utilities for non-fatal cardiovascular events that accounted for the relative prevalence of MI, stroke or non-fatal bleed.

Costs  We accounted for inpatient, outpatient and pharmaceutical costs using a payer perspective. We did not explicitly incorporate the cost of ACS treatment because this was incurred by all patients, irrespective of treatment strategy. The net wholesale price for prasugrel was estimated at $5.45 per day and a generic price of $1.00 per day was used for clopdiogrel [29]. The cost of genotyping was conservatively estimated to be high (base case, $500; range, $60–500), thereby biasing the analysis against a favorable cost-effectiveness ratio for genotypic testing [30]. This range was based on communication with commercial laboratories. Costs of adverse events, including non-fatal MI (asymptomatic and symptomatic), cerebral event (TIA and CVA), major bleeds (non-life-threatening and life-threatening), cardiovascular death and other death were estimated using mean national Medicare reimbursement rates for the corresponding diagnosis-related group code (Table 1). The cost of a non-fatal bleeding event was approximated using Medicare reimbursement for inpatient treatment of gastrointestinal hemorrhage [28]. All costs were varied in sensitivity analyses.

After an adverse event, monthly costs for the survival after one additional event, asymptomatic, state were considered to be equivalent to those for the event free survival state. The monthly cost for the survival after one additional-event, symptomatic, state was adjusted for the prevalence of different adverse events in patients who entered the state. Costs were adjusted to 2010 US dollars using the Bureau of Labor and Statistics Consumer Price Index Calculator [31].

Sensitivity analyses  We performed one-way sensitivity analyses of all variables included in the decision model over plausible ranges (Table 1). Ranges for adverse clinical events were derived from confidence intervals reported in TRITON TIMI-38 and the literature. In two-way sensitivity analyses, we calculated the cost-effectiveness ratios of genetic testing while varying the risk of non-fatal MI and CV death, test sensitivity and specificity, population prevalence of the loss-of-function allele, and cost of medications. We also performed two-way sensitivity analyses of the risk of non-fatal MI and major bleed events (fatal and non-fatal). We performed a probabilistic sensitivity analysis, randomly sampling (with replacement) a distribution of all variables 10 000 times and then simulating outcomes. We used β distributions for rates of adverse events and gamma distributions for cost of genetic testing and cost of clopidogrel.

Results

Base case analysis

In a population of 60-year-old patients presenting with ACS undergoing PCI, there was little difference between treatment strategies. The genetic testing strategy resulted in the greatest number of QALYs and was less costly than the no genetic testing strategies regardless of which empiric therapy was used (clopidogrel or prasugrel). These findings were present at 15 months and persisted at 10 years of follow-up. Over 15 months, genetic testing was $18 less costly and increased QALYs by 0.004 QALY, compared with treating all patients with clopidogrel. Similarly, genetic testing was $899 less costly and 0.0005 QALYs more beneficial, compared with treating all patients with prasugrel (Fig. 1). In the 10-year analysis, the direction of the findings was similar but the magnitude of differences increased, as genetic testing was $12 393 less costly and 0.117 QALYs (42.7 days) more effective when compared with a strategy of treating all patients with clopidogrel. Genetic testing was 0.013 QALYs (4.75 days) more effective and $28 855 less costly compared with a strategy of treating all patients with prasugrel (Fig. 2). Because the genetic testing strategy was dominant (more effective and cheaper) compared with empiric treatment strategies of clopidogrel or prasugrel, ICERs could not be calculated.

Figure 2.

 (A) Base case analysis at 15 months. Genetic testing strategy dominant compared with both no-testing empiric treatment strategies. Total costs were $18 lower with a gain of 0.004 QALY in the genetic testing strategy compared with the no-testing strategy with empiric clopidogrel, and $899 lower with a gain of 0.0005 QALY compared with the no-testing strategy with empiric prasugrel. (B) Base case analysis at 10 years. Genetic testing strategy dominant compared with both no-testing empiric treatment strategies. Total costs were $28 855 lower with a gain of 1.371QALY in the genetic testing strategy compared with the no-testing strategy with empiric clopidogrel and $12 393 lower with a gain of 0.152 QALY compared with the no-testing strategy with empiric prasugrel.

Sensitivity analyses

Costs  Genetic testing remained marginally cost-saving (more effective and less expensive) compared with both no-testing strategies up to a clopidogrel cost of $3.92 per day. When increasing the cost to $4.67 per day, genetic testing remained dominant compared with treating all patients with clopidogrel and cost-effective compared with treating all patients with prasugrel, with an ICER well below $50 000/QALY (Fig. 3). Varying the cost of prasugrel from $3.00 to $6.00 per day or varying the cost of genetic testing from $60 to $500 did not substantially influence the results.

Figure 3.

 One-way sensitivity analysis of cost of clopidogrel ($1–6 per day). Genetic testing strategy was dominant compared with both no-testing empiric strategies up to clopidogrel cost of $3.92 per day. Up to a clopidogrel cost of $4.67 per day genetic testing remained dominant compared with empiric clopidogrel and cost-effective with ICER< $50 000/QALY compared with empiric prasugrel. At above $4.67 per day genetic testing remained dominant compared with empiric clopidogrel and was cost-effective with ICERS< $100 000/QALY compared with prasugrel up to a clopidogrel cost of $5.60 per day.

Prevalence of the CYP2C19*2 Allele  The TRITON TIMI-38 population primarily comprised Caucasians (92%), but the CYP2C19*2 mutation population prevalence is known to vary with race, with a higher prevalence among Asians. Therefore, it is possible that distinct preferred strategies may exist for different races, with important implications for patient-centered care. In sensitivity analyses, we found that genetic testing was no longer cost-saving compared witho empiric clopidogrel treatment if the mutation prevalence fell below 25%, though it remained the most effective strategy. When the mutation prevalence was < 10%, clopidogrel was the favored empiric strategy. When compared with empiric prasugrel, higher prevalence of the CYP2C19*2 allele made prasugrel a more effective strategy. When the mutation prevalence increased above 70%, genetic testing was no longer cost-saving compared with empiric prasugrel treatment, but it continued to remain the most effective strategy. Above a prevalence > 75%, prasugrel was the favored empiric strategy.

Relative risk of adverse events in CYP2C19*2 carriers vs. wild type  Previous trials have demonstrated an approximately 50% increased risk (RR = 1.5) of thrombotic events in carriers compared with wild-type subjects treated with clopidogrel [8,11]. Compared with empiric clopidogrel, genetic testing remained dominant (less expensive and more effective) across the range of relative risks (Fig. 4). Of note, we found that our results comparing genetic testing with empiric prasugrel were sensitive to variations in the magnitude of this risk. Above a 47% increased risk (RR = 1.47), genetic testing was the least expensive strategy and resulted in the highest number of QALYs compared with empiric treatment with prasugrel. Between an RR of 1.43–1.47, genetic testing was less expensive than empiric prasugrel but marginally less effective. However, empiric prasugrel was not cost-effective, with an ICER well exceeding $100 000/QALY. Below an RR of 1.42, treating all patients with empiric prasugrel was more cost-effective (Fig. 4).

Figure 4.

 One-way sensitivity analysis of relative risk (RR) of thrombotic events in wild-type (WT) individuals treated with clopidogrel compared with all individuals treated with clopidogrel (corollary of RR MACE in carriers vs. wild type in RED): cost-effectiveness. As the relative risk of thrombotic events in carriers vs. wild type increases, genetic testing is always dominant to empiric clopidogrel, and is cost-effective compared with empiric prasugrel between an RR of 1.43 and1.47. Genetic testing is dominant to empiric prasugrel above an RR of 1.47.

Non-fatal MI, CV death and bleeding  In one-way sensitivity analyses varying the baseline relative risk of MI from 1.1 to 1.29 among individuals treated with clopidogrel compared with prasugrel, the genetic testing strategy was cost-saving (more effective and less expensive) compared with empiric prasugrel and cost-effective compared with empiric clopidogrel (ICERs less than $10 000/QALY). At a 30–40% increased risk (RR 1.3–1.4) of non-fatal MI among patients treated with clopidogrel compared with prasugrel the genetic testing strategy was cost-saving compared with both no-testing strategies. Above a relative risk of 1.4, genetic testing continued to be cost-saving relative to empiric therapy with clopidogrel but was marginally less effective than empiric therapy with prasugrel. Varying the risk of having more severe or ‘symptomatic’ MIs did not significantly alter results.

TRITON TIMI-38 reported a non-significant 12% risk increase in the secondary endpoint of cardiovascular death in individuals treated with clopidogrel vs. prasugrel. We varied this point estimate across its 95% confidence interval and did not observe a significant impact on our results. Similarly, varying the risk of major bleeding or bleed-related deaths did not substantially impact our results.

Diagnostic characteristics of genotype test  Assuming a willingness to pay of $100 000 per QALY, genetic testing remained cost-effective unless its sensitivity was < 79% or its specificity was < 84%, otherwise, a no-testing strategy with empiric prasugrel was preferred (Fig. 5).

Figure 5.

 Two-way sensitivity analysis of genotypic testing sensitivity vs. genotypic testing specificity. Shaded region identifies optimal/dominant management strategy. Genetic testing is superior when sensitivity> 84% and specificity> 79%.

Probabilistic sensitivity analyses

In the probabilistic sensitivity analyses, varying all variables simultaneously, a strategy of genetic testing to determine optimal antiplatelet strategy was cost-effective in 75% of the simulations assuming a willingness-to-pay threshold of $50 000 and in 78% of the simulations assuming a willingness-to-pay threshold of $100 000 per QALY. Even at willingness-to-pay thresholds of < $50 000, the genetic testing strategy had the highest probability of being cost-effective compared with both no-testing strategies (Fig. 6).

Figure 6.

 Probabilistic sensitivity analyses. Cost-effectiveness acceptability curves representing the probability that each treatment strategy is cost-effective for a given maximum willingness-to-pay threshold per QALY gained (over 15 months).

Discussion

Clopidogrel has been shown to reduce adverse cardiovascular events in the setting of acute coronary syndrome, but many patients continue to experience recurrent cardiovascular events. Compared with clopidogrel, prasugrel is superior in reducing the combined endpoint of non-fatal MI, non-fatal stroke and cardiovascular death (and increases the risk of significant bleeding events), possibly because genetic polymorphisms selectively reduce the antiplatelet effect of clopidogrel. Similar to Mahoney et al. [29], we replicated the finding that empiric prasugrel is cost-effective compared with clopidogrel, though our findings were much smaller in magnitude. In the current analysis, we found that a genetic testing strategy yielded clinically similar cost and health outcomes compared with strategies of no-testing and empiric clopidogrel or prasugrel therapy. Although the absolute differences are small, these findings were consistent, with genetic testing remaining the dominant strategy in numerous sensitivity analyses that varied event rates, cost and the impact of in-trial events on quality of life.

Our findings suggest that a genotype-guided strategy may be cost-effective if the risk of thrombotic events is at least 43% higher in carriers of the CYP2C19*2 allele compared with wild-type patients. However, the degree of increased risk in carriers compared with wild-type patients has been the subject of much debate. In a meta-analysis of nine studies including more than 9000 patients, 91% of whom underwent PCI (55% with a presenting ACS), the hazard ratio for major adverse cardiac events (MACE) in patients who were carriers of the loss-of-function CYP2C19 allele relative to wild-type patients was 1.57 (95% CI, 1.13–2.16; P = 0.006), with a notably wide confidence interval [11]. This relationship may also vary over time. For example, in a genetic sub-study from the PLATelet inhibition and patient Outcomes (PLATO) trial (in which only two-thirds of patients underwent PCI), the rate of CV death, MI or stroke among patients treated with clopidogrel was 5.7% in carriers of the CYP2C19 allele at 30 days compared with 3.8% in wild-type subjects (RR = 1.5; P = 0.028). However, at 12 months, this risk was somewhat attenuated, and the rate of adverse cardiovascular events in patients treated with clopidogrel was 11.2% in carriers compared with 10.0% in wild-type individuals [32].

Studies enrolling large numbers of conservatively managed ACS patients have failed to find significant evidence of an increased risk, probably because ACS patients undergoing PCI derive the greatest benefit from potent dual antiplatelet therapy [33]. For example, in the genetic sub-study from the Clopidogrel in Unstable Angina to Prevent Recurrent Events (CURE) trial, in which 15% underwent PCI, investigators did not find a significantly increased risk of events in carriers vs. wild-type patients [34]. Most recently, a published meta-analysis of 32 studies and over 42 000 patients failed to demonstrate a benefit of genetic testing for CYP2C19 polymorphisms [35]. The study included patients conservatively managed for ACS, with atrial fibrillation and with stable coronary disease. The benefit of clopidogrel in such a broad and varied population is modest; therefore, it was not surprising that testing for genetic polymorphisms affecting clopidogrel’s metabolism was not found to be efficacious. The population in which genetic testing is of particular interest comprises ACS patients undergoing PCI—a population at heightened risk for in-stent thrombosis. In the setting of PCI following ACS, the increased risk of cardiovascular events in mutation carriers was noted to be > 50% vs. wild-type subjects [8,11].

Generic clopidogrel is now available in the United States and its cost is likely to decline with time. However, consideration of the impact of a reduced price for clopidogrel (range from $1.00 to 4.00 per day from current price of $4.62 per day) did not substantially alter our results (Fig. 3). The cost of genetic testing is also likely to decline with the advent of new technology. In addition, the time between performance of the test and availability of test results is also likely to shorten, making point-of-care genetic testing feasible [36].

Limitations

Although this study used detailed data provided in FDA documents on prasugrel and clopidogrel in the setting of ACS, there are several limitations. The TRITON TIMI-38 trial was a multicenter and multinational study and healthcare delivery costs were not available at the individual patient level. Also, we accounted for the impact of genetic testing only for the CYP2C19*2 allele. Though this mutation represents 95% of loss-of-function mutations [8], we cannot comment on the cost-effectiveness of genetic testing for other mutations, such as *3, and gain-of-function mutations such as *17, where individuals are ‘ultra’ metabolizers of clopidogrel and may be at increased risk of bleeding [37]. Furthermore, across the United States, platelet reactivity testing is being used in cardiac catheterization laboratories to determine optimal antiplatelet therapy. The interplay between genetic testing and phenotyping testing for platelet reactivity is complex and still not completely understood. The current analysis does not address the economics of platelet reactivity testing.

Though prasugrel is a preferred therapy in a majority of subgroups, post hoc subgroup analyses from TRITON-TIMI 38 revealed a trend toward worse clinical outcomes in patients with a prior history of TIA or stroke, body weight < 60 kg or age over 75. Therefore prasugrel may not be considered optimal therapy in these patients. In the model all subjects remained on dual antiplatelet therapy for 120 months, assuming a constant magnitude of benefit and constant rate of adverse events because Kaplan–Meier curves suggested that the rate of events beyond 30 days was relatively constant. This is not real-life practice and may have altered the results. An additional limitation of our analysis was that we assumed that the relative risk of death in our population, compared with the general US population, remained fixed over time.

Though no studies to date have found a difference in bleeding events in carriers of the CYP2C19*2 polymorphism compared with wild-type individuals, these studies have been underpowered. We attempted to address this limitation by varying the relative risk of bleeding in carriers compared with wild-type individuals treated with clopidogrel from RR 0.7 to 1.3; our results were unaffected. However, it is possible that the relative risk falls outside of this range. Further studies are therefore needed to better elucidate the relationship between genotype and bleeding risk in patients treated with clopidogrel.

Though a willingness-to-pay threshold of $100 000/QALY was used according to established literature [30], similar results were seen with a willingness-to-pay threshold of $50 000/QALY. In fact in base case analyses the genetic testing strategy was dominant (cheaper and more effective), and continued to remain so across most sensitivity analyses.

Finally, although genetic testing was shown to be cost-saving compared with treating all ACS patients undergoing PCI empirically with prasugrel or clopidogrel, the absolute health and cost differences were small. Clinicians must continue to use judgment in evaluating patients who are at high risk of stent thrombosis and bleeding to determine optimal antiplatelet therapy. Two ongoing trials DANTE (Dual Antiplatelet Therapy Tailored on the Extent of Platelet Inhibition) and ARCTIC (Double Randomization of a Monitoring Adjusted Antiplatelet Treatment Versus a Common Antiplatelet Treatment for DES Implantation, and Interruption Versus Continuation of Double Antiplatelet Therapy, One Year After Stenting) will provide additional insights into the efficacy of alternative treatment strategies.

Recently, ticagrelor, an oral reversible antiplatelet agent, was approved by the FDA for use in the United States in patients with ACS [38]. In a cost-effectiveness analysis, empiric ticagrelor strategy was found to be cost-effective in ACS patients compared with a genetic testing strategy [39]. Further analyses evaluating genetic testing and different antiplatelet agents are needed to place results of the current analysis into context.

Conclusions

Genetic testing allows clinicians to identify patients with ACS undergoing PCI who are more likely to benefit from clopidogrel or prasugrel based on the absence or presence of the CYP2C19*2 polymorphism. Adopting this strategy yields clinically similar outcomes to empiric approaches to treatment, though it may modestly improve quality of life, reduce healthcare costs, and provide clinicians with additional guidance regarding the optimal choice of an antiplatelet regimen. Further studies are needed to determine the optimal approach to antiplatelet therapy in ACS patients.

Funding/Support

J. S. Berger was partially funded by an American Heart Association Fellow to Faculty Award (0775074N) and a Doris Duke Clinical Scientist Development Award (2010055).

Disclosure of Conflict of Interests

The authors state that they have no conflict of interests.

Ancillary