A novel cost-effectiveness model of prescription eicosapentaenoic acid extrapolated to secondary prevention of cardiovascular diseases in the United States

Abstract

Objective: Given the substantial economic and health burden of cardiovascular disease and the residual cardiovascular risk that remains despite statin therapy, adjunctive therapies are needed. The purpose of this model was to estimate the cost-effectiveness of high-purity prescription eicosapentaenoic acid (EPA) omega-3 fatty acid intervention in secondary prevention of cardiovascular diseases in statin-treated patient populations extrapolated to the US.

Methods: The deterministic model utilized inputs for cardiovascular events, costs, and utilities from published sources. Expert opinion was used when assumptions were required. The model takes the perspective of a US commercial, third-party payer with costs presented in 2014 US dollars. The model extends to 5 years and applies a 3% discount rate to costs and benefits. Sensitivity analyses were conducted to explore the influence of various input parameters on costs and outcomes.

Results: Using base case parameters, EPA-plus-statin therapy compared with statin monotherapy resulted in cost savings (total 5-year costs $29,393 vs $30,587 per person, respectively) and improved utilities (average 3.627 vs 3.575, respectively). The results were not sensitive to multiple variations in model inputs and consistently identified EPA-plus-statin therapy to be the economically dominant strategy, with both lower costs and better patient utilities over the modeled 5-year period.

Limitations: The model is only an approximation of reality and does not capture all complexities of a real-world scenario without further inputs from ongoing trials. The model may under-estimate the cost-effectiveness of EPA-plus-statin therapy because it allows only a single event per patient.

Conclusion: This novel model suggests that combining EPA with statin therapy for secondary prevention of cardiovascular disease in the US may be a cost-saving and more compelling intervention than statin monotherapy.

Keywords:

• Cardiovascular disease
• Cost-benefit analysis
• Eicosapentaenoic acid
• Cost-effectiveness

Introduction

Cardiovascular disease (CVD) is the cause of one in every three deaths in the US, and results in direct and indirect costs exceeding $300 billion annually¹. The annual incidence of recurrent myocardial infarction (MI) and stroke are 210,000 and 185,000, respectively, in the US¹. Thus, despite best evidence-based treatment strategies, including high-dose statin therapy, there remains a high residual risk of cardiovascular (CV) events and significant health and economic burdens². Adjunctive therapies shown to reduce CV events when added to statin therapy include the omega-3 fatty acid eicosapentaenoic acid (EPA) and ezetimibe^3–6.

The Japan Eicosapentaenoic Acid Lipid Intervention Study (JELIS) was a long-term CV outcomes trial that explored the benefits of EPA in combination with statin therapy compared with statin monotherapy in adult patients with hypercholesterolemia³. The major finding of an analysis of the secondary prevention population from JELIS was a significant 23% relative risk reduction in the incidence of major coronary events (MCE) in the EPA-plus-statin group (8.7% vs 10.7%, adjusted hazard ratio [HR] = 0.77, 95% confidence interval [CI] = 0.63–0.96, p = 0.017, number needed to treat [NNT] = 49)⁵. Additionally, among 1050 patients with prior MI, the incidence of MCE in the EPA-plus-statin group was significantly reduced by 27% (adjusted HR = 0.73, 95% CI = 0.54–0.98, p = 0.033, NNT = 19)⁵. As no significant changes were observed in low-density lipoprotein cholesterol, total cholesterol, and high-density lipoprotein cholesterol levels in the secondary prevention population, the findings from JELIS suggest that treatment strategies, such as EPA, that provide benefits beyond lipid effects may also be important in reducing atherosclerotic CV risk^3–5.

It has been suggested that pleiotropic effects may be an important feature for interventional strategies to successfully reduce CV risk⁷. The broad range of potentially beneficial CV effects of EPA includes those on arrhythmia, inflammation, blood pressure, and lipids⁸. Furthermore, EPA is incorporated into membrane phospholipids and atherosclerotic plaques and has beneficial effects in multiple aspects of atherosclerosis including antioxidant properties; endothelial function; effects on macrophages, monocytes, and foam cells; inflammation; plaque progression, formation, and vulnerability; and thrombus formation⁹. Recent evidence suggests that EPA may also increase high-density lipoprotein anti-inflammatory properties and augment cholesterol efflux capacity from arterial wall macrophages¹⁰.

The drug used in JELIS was EPA ethyl ester (Epadel^®, Mochida Pharmaceuticals Co., Ltd., Tokyo, Japan)^3,11. In the US, the high-purity EPA drug icosapent ethyl (Vascepa^®, Amarin Pharma Inc., Bedminster, NJ) is approved as an adjunct to diet and exercise to reduce triglyceride (TG) levels in adult patients with severe (≥500 mg/dL) hypertriglyceridemia¹². Vascepa is not approved by the US Food and Drug Administration to reduce the risk of coronary heart disease. The effect of Vascepa on the risk of CV mortality and morbidity has not been determined.

Icosapent ethyl is the ethyl ester of EPA¹² and is the same active moiety used in JELIS. Icosapent ethyl is the only prescription omega-3 fatty acid product approved in the US that does not contain docosahexaenoic acid (DHA). While EPA does not tend to raise low-density lipoprotein cholesterol (LDL-C), products containing DHA may complicate treatment strategies in patients with dyslipidemia, as DHA may raise LDL-C^13–18.

As JELIS was conducted in an exclusively Japanese population, we developed a model to explore the potential cost-effectiveness of high-purity prescription EPA-plus-statin therapy compared with statin therapy alone. We utilized event rates from the JELIS secondary prevention cohort for the statin + EPA arm to approximate reduced rates likely to be seen when extrapolated to a US population, and we utilized event rates for the statin-only arm, mainly from the Cholesterol Treatment Trialists’ (CTT) Collaboration¹⁹ analyses (and others as described in the Methods) as they may more closely represent event rates from the US. Sensitivity analyses were also conducted to determine scenarios in which EPA added to statin was not cost-effective. Of note, there is precedence for extrapolating clinical trial results from a purely Japanese population to a US population. The full panel report of the 2013 American College of Cardiology/American Heart Association (ACC/AHA) Guideline on the Treatment of Blood Cholesterol to Reduce Atherosclerotic Cardiovascular Risk in Adults considered the results from the Management of Elevated Cholesterol in the Primary Prevention Group of Adult Japanese (MEGA) trial to be generalizable to the US population. MEGA was one of four trials that formed the basis for the ACC/AHA recommendation for statin use in a primary prevention population²⁰. Additionally, data from two PROBE-designed trials, the Japanese Primary Prevention of Atherosclerosis With Aspirin for Diabetes trial and the Japanese Primary Prevention Project, were incorporated into a systematic evidence review for the US Preventive Services Task Force regarding aspirin for the primary prevention of cardiovascular events²¹.

Methods

Model design

The cost-effectiveness model was a deterministic decision analytic model designed to estimate total costs associated with the use of pharmacologic therapy and with major adverse CV and cerebrovascular events (MACCE) for statin-treated patients with and without prescription EPA use per quality-adjusted life year or per unit of utility. The model takes the perspective of a US commercial third-party payer and extends to 5 years.

Identification and definition of input parameters

The model includes inputs for CV events, costs, and utilities from multiple published sources. Many of the input parameters were obtained directly from peer-reviewed literature (Tables 1–3). Default values were provided as shown in the tables, but the user has the option to vary inputs to other pre-specified values or to add a value not included as a default.

Table 1. Base case input parameters: clinical events.

Parameter						Source(s)
Event, % experiencing over 5 years	EPA-plus-statin therapy		Statin monotherapy
Sudden cardiac death	0.7		2.0			JELIS^5; PROVE-IT-TIMI 22^23
Fatal MI	0.3		1.0			JELIS^5; Assumption*
Nonfatal MI	1.4		4.9			JELIS^5; TNT^22
Unstable angina	4.8		7.5			JELIS^5; Assumption†
CABG/PTCA	7.0		10.0			JELIS^5; CTT^19
Stroke	1.9		3.0			JELIS^5; CTT^19
Distribution of CABG vs PCTA, %
CABG	35					FREEDOM^24; SYNTAX^25
PCTA	65
Distributions of events over time, % per year
Year	1	2	3	4	5
CABG/PCTA	33.3	33.3	11.1	11.1	11.1	FREEDOM^24; SYNTAX^25; expert opinion, guided by literature
Other MACCE	20	20	20	20	20	Expert opinion, guided by literature

* Assumption based on expert opinion and values calculated from Dieleman et al.²⁶ for non-Japanese population.

† Assumption based on values reported for PROVE-IT-TIMI 22²³ extrapolated to 5 years and corrected to reflect improved medical management.

CABG, coronary artery bypass grafting; CTT, Cholesterol Treatment Trialists’ Collaboration; EPA, eicosapentaenoic acid; FREEDOM, Future Revascularization Evaluation in Patients With Diabetes Mellitus: Optimal Management of Multivessel Disease; JELIS, Japan Eicosapentaenoic Acid Lipid Intervention Study; MACCE, major adverse cardiovascular and cerebrovascular events; MI, myocardial infarction; PROVE-IT-TIMI 22, Pravastatin or Atorvastatin Evaluation and Infection Therapy–Thrombolysis in Myocardial Infarction 22; PTCA, percutaneous transluminal coronary angioplasty; SYNTAX, Synergy Between Percutaneous Coronary Intervention With TAXUS and Cardiac Surgery; TNT, Treating to New Targets study.

Table 2. Base case input parameters: costs.

Parameter	Annual per-patient cost (USD)*	Source(s)
Sudden cardiac death	$34,501	Russell et al.^30, 1998
Fatal MI	$34,501	Russell et al.^30, 1998
Non-fatal MI	$47,139	Wang et al.^31, 2013; Azoulay et al.^32, 2003
Subsequent non-fatal MI	$3,446	Applies ratio from Cowie et al.^33, 2011
Unstable angina	$33,935	Kongnakorn et al.^34, 2009; Russell et al.^30, 1998
Subsequent unstable angina	$3,875	Bloom et al.^35, 1999
CABG	$48,050	Osnabrugge et al.^36, 2014; Brown et al.^37, 2008; Wang et al.^31, 2013; Magnuson et al.^24, 2013; Eisenberg et al.^38, 2005; Cohen et al.^25, 2014
PTCA	$27,363	Magnuson et al.^24, 2013; Cohen et al.^25, 2014
Subsequent revascularization	$5,183	Cohen et al.^25, 2014
Stroke	$67,383	Kongnakorn et al.^34, 2009
Subsequent stroke	$15,147	Applies ratio from Cowie et al.^33, 2011
US EPA (icosapent ethyl): 12•WAC	$2,504	Medispan – PriceRx 1/21/2015
Statins (annual)	$1,000	Assumption^†

* Costs were inflated to 2014 values and discounted 3%/year as described in the Methods.

† Statin costs were an approximation and blended from generic and brand costs³⁹ and can be changed in the model.

CABG, coronary artery bypass grafting; EPA, eicosapentaenoic acid; MI, myocardial infarction; PTCA, percutaneous transluminal coronary angioplasty; USD, United States dollars; WAC, wholesale acquisition cost.

Table 3. Base case input parameters: utilities.

Parameter	Value	Source(s)
Baseline utility	0.7800	Luengo-Fernandez et al.^40, 2013
Year with non-fatal MI or revascularization	0.6864	McConnachie et al.^41, 2014 (based on 24% decrement for half year)
Year(s) following non-fatal MI or revascularization	0.7332	Assumption: midpoint of baseline and event year*
Year with unstable angina	0.7059	McConnachie et al.^41, 2014 (based on 19% decrement for half year)
Year(s) following unstable angina	0.7332	Assumption: midpoint of baseline and event year*
Year with stroke	0.6353	McConnachie et al.^41, 2014 (based on 37.1% decrement for half year)
Year(s) following stroke	0.7077	Assumption: midpoint of baseline and event year*

MI, myocardial infarction.

* The model assumes that patients’ utility values increase halfway back to baseline (i.e. the midpoint value between the utility during the event year and the baseline utility).

Clinical events

Figure 1 depicts the CV event flow in the model. MACCE included in the model were sudden cardiac death, fatal MI, non-fatal MI, unstable angina, coronary artery bypass grafting (CABG), percutaneous transluminal coronary angioplasty (PTCA), and stroke. With the exception of stroke, the model used JELIS secondary prevention data for the EPA-plus-statin arm and values from CTT higher-intensity vs lower-intensity statin trials for the statin monotherapy arm (Table 1)^5,19,22–26. The model assumes that the event rates, except for stroke, will be similar to those from the JELIS secondary prevention study in the EPA-plus-statin arm, as the high plasma levels of EPA achieved by patients in the JELIS trial (170 μg/mL)²⁷ have also been achieved by a high-dose purified prescription EPA product in US patients^28,29. Stroke rates for the JELIS secondary prevention data are higher than US rates; therefore, stroke event rates from the secondary prevention studies in the CTT analysis¹⁹ were utilized in the model. The percentage reduction in relative risk for stroke between the EPA-plus-statin arm vs the statin monotherapy arm was modeled to be similar to the percentage reduction in relative risk for stroke between the EPA-plus-statin arm vs the statin monotherapy arm observed in the JELIS secondary prevention strata.

Figure 1. Decision analytic model flow.

Table 1 also presents other assumptions that were made in order to address two data gaps: in the statin monotherapy group, fatal MI events were assumed to be 1.0% and unstable angina was assumed to be 7.5%. These assumptions were informed by published literature^23,26. Of note, the JELIS data combined CABG and PTCA without presenting data on each end-point separately⁵. Because there are different costs for the two procedures, with CABG costs being substantially higher, attributing costs to these events required making an assumption about the distribution between the two. Using the base case, the model was set at 35% for CABG and 65% for PTCA (informed by the FREEDOM²⁴ and SYNTAX²⁵ studies). However, the user can select from pre-specified percentages, with CABG representing 100%, 65%, 35%, or 0% of interventions. The model also includes a user-defined field wherein the user can enter any number between 0% and 100% as the proportion of all CABG interventions. In this way, although a default value is identifiable in the model, the user can specify any value.

The JELIS secondary prevention report provided the number of events over the 4.6-year mean follow-up of the study, but it did not present data on the distribution of events during each year; therefore, assumptions about the distribution were required to populate the model. Based on a review of published studies among patient groups with risk factors^19,22–25, the per annum distribution of CABG/PTCA interventions were considered separately from other MACCE in the model. The default distribution for CABG/PTCA was assigned as 33% in each of the first 2 years, with the remaining 33% distributed evenly over the final 3 years. For other MACCE, the default distribution was equal over each of the 5 years. For both distributions, there are also built-in options to use alternative values built into the model (e.g. front-loaded with 25% in each of the first 2 years, with the remainder evenly distributed, or back-loaded with 15% in each of the first 2 years, with the remainder evenly distributed) as well as a functionality for users to vary the distribution over the time period as desired.

Costs

Published literature was assessed to identify the costs of treatment for the MACCE that were included as end-points in the JELIS secondary prevention strata: sudden cardiac death, fatal MI, non-fatal MI, unstable angina, CABG/PTCA, and stroke^{24,25,30–39}. These costs and sources are presented in Table 2.

In addition to costs for each event, the model applied costs for the year or years following each event. For stroke and non-fatal MI, the ratio of first-year costs vs following-year costs were applied based on published data, while for revascularization and unstable angina, estimates from specific studies were used (Table 2). Wholesale acquisition costs were used for prescription drug costs. Costs were inflated to 2014 values using price index information as reported by the Bureau of Labor Statistics (Series ID CUUR0000SAM, medical care) and were discounted at a rate of 3% per annum.

Utilities

Based on the literature, we estimated utilities for each event; the base case utility values are presented in Table 3^40,41. Utilities were decremented by a specified fraction for the year during which an event occurred. For example, the quality-of-life decrement associated with stroke was 37.1%. The model assumed that events would be evenly distributed throughout the year, so, on average, half the year would be affected; thus, the model applied this decrement for half of the year, while the other half was considered to remain at the baseline value (e.g. 0.78). If the patient died during the year, a utility value of 0 was assigned for half the year. In the years subsequent to an event, the model assumes that patients’ utility values increase halfway back to baseline (i.e. the midpoint value between the utility during the event year and the baseline utility). As in the case of cost inputs, values were discounted at a rate of 3% per annum.

Calculations

The cost-effectiveness model was developed in Microsoft Excel^® 2010 and all accompanying calculations were executed therein. For each model input type (i.e. pharmaceutical costs, event costs, utilities), the same process is implemented. Based on user inputs about events and their distribution over time, the outcomes for each year are calculated based on the proportion of patients who fall into each category. For example, the proportion of patients who have a non-fatal MI in year 1 of the model is used to apply a cost associated with MI to those patients and to decrement their utility values accordingly. These patients then remain in the model and accrue a lower cost (compared with the event cost) in each subsequent year as well as a higher utility (compared with the event year) for the remaining years in the model. A patient who suffers a fatal event during year 1 is assigned a cost for the event year as well as a lower utility value for that year. After the initial year, that patient is then removed from the analysis and receives no additional costs or utilities. The process is repeated for each subsequent year, with the pool of patients decreasing each year as fatal events occur. Patients who have additional non-fatal events are assigned adjusted costs and utilities according to their previous event (i.e. increased costs and decreased utilities; see Tables 2 and 3), but they are assigned the same cost for pharmaceutical agents annually as long as they remain alive. Incremental cost-effectiveness is calculated as the difference in total cost over 5 years (for EPA-plus-statin therapy minus statin monotherapy) divided by the difference in effectiveness over 5 years (expressed as total utilities for EPA-plus-statin therapy minus total utilities for statin monotherapy).

Sensitivity analyses

For each input parameter, sensitivity analyses were conducted assuming a range of plausible and extreme values to examine the influence of each parameter on the model and to test the robustness of the model. Rates, timing of events, patient utilities, costs of events, and overall assumptions were varied and tested. In addition to modifying individual model input parameters, two alternative scenarios were developed that involved different values for all clinical inputs. The first used JELIS event rates (‘JELIS scenario’) and the second assumed higher rates of effectiveness for EPA-plus-statin therapy (‘alternate scenario’). Cost and utility values from the base case were applied to these two alternative clinical scenarios. Results were interpreted based on the joint ACC/AHA statement on cost/value methodology⁴².

Results

The results for the base case of the model are presented in Table 4. Under the base case conditions, over the 5-year period, costs were lower ($29,377 vs $30,587) and total utilities accrued were higher (3.627 vs 3.575) for patients treated with EPA-plus-statin therapy compared with statin monotherapy, demonstrating that the EPA-plus-statin therapy dominated. In this base case scenario, EPA was cost saving. There were consistently higher annual costs for pharmacologic therapy in the EPA group, but, by the third year, there were savings in total costs due to the increasing costs for treatment of MACCE in the statin monotherapy group. The cost savings occurred despite the fact that, in the statin monotherapy group, CV mortality is assumed to be higher and, thus, more people drop out entirely and accrue no further costs.

Table 4. Model results: base case.

	EPA-plus-statin therapy	Statin monotherapy
First year costs* and outcomes
Pharmacologic agents	$3,497	$994
Costs of MACCE	$1,547	$2,666
Utilities	0.775	0.771
Second year costs* and outcomes
Pharmacologic agents	$3,386	$958
Costs of MACCE	$1,886	$4,020
Utilities	0.748	0.740
Third year costs* and outcomes
Pharmacologic agents	$3,277	$924
Costs of MACCE	$1,328	$5,422
Utilities	0.724	0.713
Fourth year costs* and outcomes
Pharmacologic agents	$3,173	$891
Costs of MACCE	$4,020	$6,452
Utilities	0.700	0.686
Fifth year costs* and outcomes
Pharmacologic agents	$3,071	$859
Costs of MACCE	$4,191	$7,401
Utilities	0.679	0.665
5 year totals
Total cost*	$29,377	$30,587
Total utilities	3.627	3.575
Incremental cost-effectiveness ratio	EPA-plus-statin dominates statin monotherapy

* All costs are per patient. EPA, eicosapentaenoic acid; MACCE, major adverse cardiovascular and cerebrovascular events.

Table 5 presents results for sensitivity analyses that explore input parameters on rates and timing of events, patient utilities, costs of events, and overall assumptions (i.e. discount rate). In most scenarios, the results were not sensitive to multiple variations in model inputs and consistently identified EPA-plus-statin therapy to be the economically dominant strategy, with both lower costs and better patient utilities over the modeled 5-year period. Based on additional sensitivity analyses, EPA remained highly cost-effective (according to the framework of the ACC/AHA statement on cost/value methodology⁴²) when event rates for the EPA + statin group were ∼70% of the rates in the statin group, and met the upper end of the threshold for intermediate care value only when rates were ∼78% of the statin group’s event rates. This threshold analysis takes into account only differences in event rates; similar analyses could be conducted with costs or utilities, but, as there is less uncertainty around these input parameters and their impact has less influence on model results, similar analyses have not been conducted around them.

Table 5. Modeled sensitivity analyses.

Scenario and/or changes from base case	Cost/QALY
Base case	EPA plus statin dominates statin monotherapy
Vary event rates
JELIS scenario	$210,294/QALY; EPA is not cost-effective
Alternate scenario*	EPA plus statin dominates statin monotherapy
Sudden cardiac death event rate 0.7% for both EPA plus statin and statin only	EPA plus statin dominates statin monotherapy
Fatal MI event rate 0.4% for both EPA plus statin and statin only	EPA plus statin dominates statin monotherapy
For non-fatal MI, statin-only event rate 2%	$52,938/QALY; EPA is cost-effective with intermediate value†
For CABG/PTCA, statin-only event rate 8%	$12,070/QALY; EPA is cost-effective with high value†
Sudden cardiac death event rate 0.7% and fatal MI event rate 0.4% for both EPA plus statin and statin only	EPA plus statin dominates statin monotherapy
Sudden cardiac death event rate 0.7% and fatal MI event rate 0.4% for both EPA plus statin and statin only, event rate for non-fatal MI 3% for statin only	$100,357/QALY; EPA is cost-effective with intermediate value
Sudden cardiac death event rate 0.7%, fatal MI event rate 0.4%, and stroke rate 1.9% for both EPA plus statin and statin only, event rate for non-fatal MI 3% for statin only	$202,078/QALY; EPA is not cost-effective
CABG vs PTCA: all events CABG or all events PTCA	EPA plus statin dominates statin monotherapy
Vary timing of events
CABG: evenly distributed over 5 years	EPA plus statin dominates statin monotherapy
CABG: back-loaded (15% in each of first 2 years, evenly distributed afterward)	$798/QALY; EPA is cost-effective with high value†
MI: back-loaded (15% in each of first 2 years, evenly distributed afterward)	EPA plus statin dominates statin monotherapy
Vary costs
Discount rate: varies from 0% to 5%	EPA plus statin dominates statin monotherapy
Costs: varied by 25% above or below base case	EPA plus statin dominates statin monotherapy
Statin cost: decreased to $150 annually or increased to $2,000 annually	EPA plus statin dominates statin monotherapy
Vary inputs: utilities
Utilities: varied by 10% above or below base case	EPA plus statin dominates statin monotherapy

* Alternate scenario: statin-only group having ∼2–3.5× higher event rates than the EPA plus statin group; scenario influenced by well-controlled outcome trials (CTT¹⁹, TNT²², PROVE-IT–TIMI 22²³).

† Results were interpreted based on the joint ACC/AHA statement on cost/value methodology⁴².

ACC/AHA, American College of Cardiology/American Heart Association; CABG, coronary artery bypass grafting; EPA, eicosapentaenoic acid; JELIS, Japan Eicosapentaenoic Acid Lipid Intervention Study; MI, myocardial infarction; PTCA, percutaneous transluminal coronary angioplasty; QALY, quality-adjusted life year.

Discussion

The decision analytic model presented here compared the cost-effectiveness of EPA-plus-statin therapy with statin monotherapy under a set of base-case assumptions as well as under more extreme assumptions using sensitivity analyses. The findings of the model indicated that addition of EPA to statin therapy for secondary prevention of cardiovascular diseases consistently resulted in both cost savings and better patient outcomes as measured by utilities. The base case as well as most sensitivity analyses resulted in cost-effectiveness ratios that indicate high (to intermediate) care value according to the framework of the ACC/AHA cost/value methodology⁴².

The framework and initial clinical data source for model input for the EPA-plus-statin arm of this model was JELIS. In the full panel report of the 2013 ACC/AHA Guideline on the Treatment of Blood Cholesterol to Reduce Atherosclerotic Cardiovascular Risk in Adults, JELIS was evaluated as the only omega-3 fatty acid trial in the expert panel’s systematic review²⁰. In JELIS, EPA was co-administered with a statin and compared with statin monotherapy in both primary and secondary prevention populations^3,5. EPA 1,800 mg was added to statin therapy in a Japanese population of men and post-menopausal women with baseline LDL-C levels ≥170 mg/dL, with and without coronary heart disease. Compared with statin monotherapy, co-administration of EPA and statins did not reduce LDL-C, and reduced triglycerides only modestly (estimated difference of 5%). However, combined EPA and statin therapy reduced coronary heart disease events by 19% (HR = 0.81, 95% CI = 0.69–0.95, p = 0.011) compared with statin monotherapy in the intent-to-treat patient population³. Similar magnitudes of risk reduction were observed in primary and secondary prevention populations^3,5. However, JELIS was not powered for sub-group analyses³. The ACC/AHA expert panel did not generalize the JELIS data to US populations due to the high dietary intake of omega-3 fatty acid, as fish intake is higher in the Japanese population compared with the US^3,20, and potentially due to the low statin doses used in the JELIS trial relative to current statin doses commonly used in US populations. With regard to the statin dose, the doses of simvastatin and pravastatin used in the JELIS trial were based on Japanese guidelines and were supported by the CV risk reduction benefits shown with the same pravastatin dose in the MEGA trial^3,43. The MEGA trial also included a Japanese population with a high dietary fish intake. However, it may be worth noting that the high plasma levels of EPA achieved by patients in the JELIS trial can also be achieved with a high-dose purified EPA in US patients²⁸, and blood EPA levels may be associated with clinical significance⁴⁴. An analysis of JELIS patients demonstrated that a higher on-treatment plasma level of EPA, but not DHA or other fatty acids, was significantly associated with a lower risk of MCE (HR = 0.71, 95% CI = 0.54–0.94, p = 0.018)²⁷.

This model utilized the framework of the JELIS secondary prevention trial to examine cost-effectiveness based on event rates in the EPA-plus-statin arm within JELIS compared with an event rate potentially more representative of the US population in the statin monotherapy arm based on event rates in the secondary prevention trials within the PROVE-IT-TIMI 22, TNT, and other CTT more vs less statin trials^19,22,23. However, the model was designed to be dynamic and allow users such as managed care organizations to input event rates from the patient lives they cover to determine a tailored cost-effectiveness estimate. An analysis from a claims database looking at the prevalence and the results on outcomes with an omega-3 fatty acid plus statin regimen among patients discharged with MI in Italy showed a synergism between the effects of statins and omega-3 fatty acids for most CV outcomes, including all-cause mortality compared with statin monotherapy (incidence rate difference = −2.5 per 100 patients/year; 95% CI = −2.2 to −2.9, p < 0.001)⁴⁵. As this model evaluated the cost-effectiveness of high-purity EPA-plus-statin therapy, it should be noted that the results would only be expected for a high-purity prescription EPA product and should not be translated to other prescription omega-3 fatty acid products that contain DHA or to omega-3 fatty acid dietary supplements such as fish oils^46–48. Costs may be different and DHA may raise LDL-C^13,14,17,18, which may potentially alter the risks and incidences of CV events included in this model. Furthermore, the EPA and/or DHA content of dietary supplements may vary from the amount stated on the package label and notably, dietary supplements are not required by the US Food and Drug Administration to demonstrate clinical safety or efficacy⁴⁹.

Limitations

This model should be interpreted in the context of several important limitations. First, our model is only an approximation of a potential reality and is unable to capture all complexities of a real-world scenario. Many of the model inputs were derived from the literature and not from a single outcomes trial with a priori pharmacoeconomic end-points. The model allows for only a single event per year per patient and does not fully account for the possibility that patients could experience multiple events in a year or over time; there were no data to estimate such event rates or whether there were increases in costs or decreases in utilities associated with multiple events. Therefore, findings may under-estimate the cost-effectiveness of the EPA-plus-statin combination due to potential reductions in more than one event per patient. A recent IMPROVE-IT trial analysis showed a slightly greater relative risk reduction in total CV events (first and recurrent) compared with the first event only⁵⁰. The model reported herein was developed from the perspective of the US healthcare system, even though the EPA-plus-statin arm event rates were derived from the JELIS secondary prevention strata, a purely Japanese population. For the EPA-plus-statin arm, the model assumes that the event rates will be similar to the JELIS secondary prevention strata and this may not be the case in the US population. The JELIS secondary prevention strata were not representative of the US population with established CVD, as the rates of CVD in a secondary prevention population are higher in the US than in Japan^51,52. However, because of this difference, and because the EPA plasma concentrations in JELIS (170 μg/mL) can be achieved with a high-purity EPA prescription product in the US^28,29, one may expect the difference in event rates between statin monotherapy vs EPA-plus-statin therapy to be larger in the US^3,27. The cost/quality-adjusted life year for the exact JELIS scenario in this report was not cost-effective for EPA. This may be a result of applying lower coronary event rates from a Japanese patient population, without adjustment, to a Western population inclusive of drug and intervention costs. The model is not unique in extrapolating Japanese event data to the US^20,21. This model also excluded indirect costs and costs of adverse events and assumed sufficient adherence to medication; all of these approaches are consistent with previous cost-effectiveness analyses of lipid-modifying treatments⁵³. Additionally, the model is not meant to present a societal perspective and does not include non-medical or patient out-of-pocket costs. Based on our deterministic model design, an acceptability curve could not be derived.

Conclusions

This novel model suggests that combining EPA with statin therapy for secondary prevention of CVD in the US may be cost saving and appears to be a more compelling intervention than statin monotherapy. The robustness of this decision analytic model was confirmed by performing a range of sensitivity analyses by varying key inputs and assumptions. However, the treatment effect of icosapent ethyl (high-dose purified EPA) on CV outcomes remains to be proven; the results of the ongoing REDUCE-IT study (clinicaltrials.gov NCT01492361) utilizing icosapent ethyl with statin therapy will allow for a more refined cost-effectiveness microanalysis based on outcomes data in patients with persistent hypertriglyceridemia, including those from the US and other Western countries⁵⁴.

Transparency

Declaration of funding

This study was designed and sponsored by Amarin Pharma Inc., Bedminster, NJ, USA.

Declaration of financial/other relationships

JKS and CKH-L are employed by Exponent, which received grant/research support from Amarin Pharma Inc. JN is on the speakers’ bureaus of Amarin Pharma Inc., Arbor, Amgen, Kowa Pharmaceuticals America, Inc., Merck, Regeneron, and Sanofi; is a consultant to Amarin Pharma Inc.; owns stock in Amarin Pharma Inc., Amgen, Pfizer, and Regeneron; and provides advisory services to Amarin Pharma Inc. SP and PBE are employees and stock shareholders of Amarin Pharma Inc. SC is a consultant to Amarin Pharma Inc. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Acknowledgments

Editorial assistance was provided by Peloton Advantage, LLC, Parsippany, NJ, USA, and funded by Amarin Pharma Inc., Bedminster, NJ, USA.