Summary of the Sponsor’s Pharmacoeconomic Submission
The sponsor submitted a cost-utility analysis comparing icosapent ethyl plus statin therapy to statin therapy alone for the reduction of CVEs (i.e., CVE-related death, non-fatal myocardial infarction, non-fatal stroke, coronary revascularization, and unstable angina) in patients 45 years of age or older with established CVD or 50 years of age or older with diabetes mellitus and other CVD risk factors, with elevated triglycerides (1.53 to 5.63 mmol/L or 135 to 499 mg/dL) and cholesterol (1.0 to 2.6 mmol/L or 40 to 100 mg/dL) on stable statin therapy.1,3 No other comparators were considered, based on the sponsor’s assertion that limited efficacy was observed with other therapies that aim to reduce triglyceride levels when administered in addition to statin therapy. The analysis was conducted from the Canadian public health care payer perspective with yearly cycles over a 20-year time horizon. A discount rate of 1.5% was applied to costs and QALYs.3 Baseline characteristics (e.g., gender, age, and distribution of statin intensity) upon model entry were based primarily on data from the REDUCE-IT trial.4 Statin regimens falling under each intensity category were defined by the 2013 American College of Cardiology/American Heart Association (ACC/AHA) Guideline on Treatment of Blood and Cholesterol to Reduce Atherosclerotic Cardiovascular Risk in Adults.8 Additionally, a proportion of patients were assumed to be receiving 10 mg ezetimibe based on data from the REDUCE-IT trial.
A Markov state-transition model was submitted to reflect the natural history of disease and the effects of treatment, consisting of the following five health states: CVE-free, death from fatal cardiovascular causes, non-fatal CVE, post–non-fatal CVE, and death from other (non-CVE) causes. Patients entered the model in the CVE-free state and were at risk of fatal or non-fatal CVE. Patients experiencing a fatal CVE transitioned to death from fatal cardiovascular causes, while patients experiencing a non-fatal CVE were further categorized to one of the following events: non-fatal myocardial infarction, non-fatal stroke, coronary revascularization, or hospitalization for unstable angina,3 and in the following model cycle moved into the post–non-fatal CVE state. While in the post–non-fatal CVE state, patients were at risk of subsequent fatal or non-fatal CVEs. In all alive health states, patients had a baseline risk of non-cardiovascular death (i.e., thereby entering the “death from other [non-CVE] causes” health state). Both death states were absorbing states in which patients remained upon entry ().
Natural history, in the form of transition probabilities, was based on parametric models for each of the individually included CVEs from the statin-only arm of the REDUCE-IT trial.4 Treatment efficacy for icosapent ethyl plus statin therapy was also informed by the REDUCE-IT trial. Transition probabilities for icosapent ethyl plus statin therapy were calculated by applying individual HRs for each of the primary CVEs to the transition probabilities expected for statin-only therapy. Treatment efficacy was applied only to the first five years of the model’s time horizon. To estimate the occurrence of secondary major adverse CVEs (MACE) for patients in the post-CVE state, a probability of any CVE event was applied, based on the data from the REDUCE-IT trial, with the type of event assigned based on the distribution of secondary MACE events reported in the REDUCE-IT trial (). Treatment-specific transition probabilities for secondary MACE events were applied for only the first five years in the model, after which the model assumed event rates were equal between groups and patients would not experience any subsequent CVE events.
All-cause mortality was obtained from Statistics Canada life tables,9 and, beyond year five of the model’s time horizon, patients with a history of CVE were assumed to be at an increased risk of death.10,11 Adverse event risks were derived from the REDUCE-IT trial.4
Health state utility values were obtained from the published literature. A baseline value of 0.762 was used to reflect the fact that patients had existing CVD, or diabetes and a high risk of CVD, at baseline. The utility value for acute or post-CVE was calculated by multiplying the utility values associated with the specific event by the baseline utility value noted previously. Disutilities for adverse events were also included based on values identified in the published literature.12–15 Drug costs for icosapent ethyl were obtained from the sponsor, while costs of statin therapy and ezetimibe were obtained from the Ontario Drug Benefit e-Formulary.16 No acquisition costs for icosapent ethyl were applied after the first five years in the model.4 Costs related to CVE and adverse events were obtained from a combination of the published literature and the Ontario Case Costing Initiative,17 while disease management resource use was obtained from a combination of clinical practice guidelines8 and clinical expert opinion, and costed according to the Ontario Schedule of Benefits.18
Sponsor’s Base Case
In the sponsor’s base case, icosapent ethyl plus statin therapy was $12,523 more expensive and produced more QALYs (0.29), resulting in an ICUR of $42,797 per QALY gained compared to statin therapy alone over 20 years (). The majority of the cost difference was driven by drug costs ($16,764), followed by the cost to manage the first CVE, whereas the QALY difference was driven primarily by the fewer number of patients experiencing coronary revascularization (Table 12). Icosapent ethyl plus statin therapy had a 70.4% probability of being cost-effective at a willingness-to-pay threshold of $50,000 per QALY.
Summary of Results of the Sponsor’s Base Case.
Summary of Sponsor’s Sensitivity Analyses
The sponsor evaluated several alternative assumptions using scenario analyses conducted probabilistically. The model results were most sensitive to its time horizon. Specifically, in adopting a lifetime time horizon, the incremental costs decreased and QALYs increased for icosapent ethyl with statins versus statin therapy alone, resulting in an ICUR of $32,925.
Additionally, several deterministic sensitivity analyses were conducted to test the sensitivity of individual parameter inputs on the overall economic findings. The model was found to be most sensitive to patients’ starting age, the HR for cardiovascular death, and the statistical parameters around cardiovascular death (mean and standard deviation).
Limitations of Sponsor’s Submission
Inappropriate time horizon: The sponsor considered a 20-year time horizon in its base case, citing previous CADTH Common Drug Review (CDR) reviews and recommendations for medications indicated for the treatment of high cholesterol. While CADTH recognizes prior reviews and the time horizons considered appropriate in these reviews, they were conducted before the latest edition of the CADTH Guidelines for the Economic Evaluation of Health Technologies.
5 The latest guidelines note that the time horizon should be long enough to capture all potential differences in costs and outcomes between the therapies under consideration. Thus, the appropriate time horizon for this chronic condition would capture the patient’s lifetime. The use of a shorter time horizon in the sponsor’s base case likely biased results against icosapent ethyl, as shown by the results of a scenario analysis conducted by the sponsor using a lifetime time horizon. A lifetime time horizon was used in the CADTH base case, and a 20-year time horizon was used in a scenario analysis.
Duration of treatment effects and costs of icosapent ethyl have been underestimated: Given a lack of comparative efficacy data beyond five years, the sponsor assumed treatment effects would apply only in the first five years of the model’s time horizon. In accordance with this, they also assumed that the cost of therapy would only be applied during the first five years of the model (i.e., no drug acquisition costs for icosapent ethyl beyond the first five years). In effect, the assumption made suggests all patients would discontinue icosapent ethyl after five years. Feedback from clinical experts consulted by CADTH indicated patients would be prescribed icosapent ethyl beyond the five-year time horizon that has been studied in the trial, as patients can be expected to continue to benefit from treatment. Further to this, there is no specific mention in their anticipated product monograph restricting treatment duration to five years. This assumption underestimated the costs and QALYs associated with icosapent ethyl. As drug costs are a key driver of cost differences in the model and the model is, to a greater degree, more sensitive to this parameter, the sponsor’s approach would bias results in favour of icosapent ethyl. CADTH addressed this limitation by extending the application of treatment efficacy (with respect to the primary CVE event) and costs for the entirety of the time horizon, beyond the initial five years. In light of limited long-term evidence, experts suggested it would be reasonable to assume that treatment benefits would remain constant with those observed during the trial period. As treatment-specific transition probabilities for secondary MACE events were available only for the first five years, CADTH continued to apply an increased mortality risk associated with CV events beyond the first five years in the model.
Incorrect post–non-fatal CVE utility values used: Upon review of the literature cited as the source of utility values for several non-fatal CVEs, several discrepancies were identified. The NICE report cited by the sponsor as the source for utility values related to revascularization did not include any values specific to revascularization, as this was not a specific clinical event in its model.
6 It appears that the sponsor selected the values for stable angina, with an assumption that there was no difference in utility values from the acute phase to the post-revascularization phase. Of note, the NICE report referenced by the sponsor as the source of this utility value cited another health technology assessment on the use of statins for prevention of coronary events.
7 In that study, revascularization was a possible clinical event in its economic evaluation, and the authors of that study assumed that post-revascularization patients would return to their baseline utility value.
7 Given the differences in revascularization rates between treatments, the assumption that the utility values for acute and post-revascularization are identical favours icosapent ethyl because it estimates fewer QALYs on statin-only therapy. This assumption overestimates the impact of revascularization events on utility, biasing results in favour of icosapent ethyl. A utility multiplier of one for post-revascularization was used in the CADTH base case. A lower value was tested in a scenario analysis, recognizing that some studies have suggested that post-revascularization does not achieve the same utility as return to baseline values, albeit not worse than acute revascularization.
19 Additionally, the NICE review cited by the sponsor as the source of utility values for post–non-fatal stroke, post–non-fatal myocardial infarction (MI), and post–unstable angina reported the utility values for these events as 0.628, 0.880, and 0.880, respectively, whereas the sponsor’s model used 0.683, 0.808, and 0.808, respectively.
6 The correct utility values from the NICE review were used in the CADTH base case.
Efficacy of icosapent ethyl may vary by risk stratum:
of the REDUCE-IT trial publication
4 shows HRs of the primary trial end point for various pre-specified subgroups. One such subgroup is a stratification by risk category, which included the following two risk strata: secondary prevention (established CVD) and primary prevention (diabetes with one other risk factor). For the composite primary clinical end point, Bhatt et al. identified an HR of 0.73 (95% confidence interval [CI], 0.65 to 0.81) in the secondary-prevention subgroup and an HR of 0.88 (95% CI, 0.70 to 1.10) in the primary-prevention subgroup, indicating that the efficacy of icosapent ethyl may vary for these subgroups.
4 The CADTH clinical review noted that, while there may be a difference of effect for the two subgroups, the REDUCE-IT trial was underpowered to detect whether such a difference was statistically significant. The clinical experts consulted by CADTH indicated that, if more patients had been included in the trials, it may have been reasonable to observe a clinically meaningful difference in effect between risk strata. If the clinical effectiveness of icosapent ethyl does differ by risk subgroups, the use of a combined HR reflecting the full population in the economic analysis would not allow comprehensive understanding of the differential cost-effectiveness across these distinct subgroups. CADTH was unable to conduct stratified analyses by subgroup due to a lack clinical data (HRs) for each of the individual cardiovascular outcomes stratified by risk strata. The potential cost-effectiveness of icosapent ethyl may differ between the primary- and secondary-prevention cohort, although the magnitude to which the ICUR may change is unknown.
Selective patient population: The comparative efficacy data incorporated in the model were based on the REDUCE-IT trial. As noted in the CADTH clinical review, the population studied was highly selective. A large number of patients were considered screening failures due to the study’s exclusion criteria (e.g., triglyceride level below 2.3 mmol/L [200 mg/dL] or above 5.6 mmol/L [500 mg/dL], congestive heart failure, active liver disease, or a planned coronary surgery or intervention). According to clinical experts consulted on this review, it is reasonable to expect that, in real clinical practice, many of the patients considered screening failures in the trial may be treated with icosapent ethyl. The clinical experts consulted on this review further noted that the available clinical studies provide limited evidence on the drug’s efficacy in patients younger than 50 years and older than 70 years. The generalizability of the economic results may therefore be restricted to the specific population enrolled in the REDUCE-IT study.
Not all comparators of interest were included: The sponsor did not include any regimen that included both statin therapy and an additional active treatment in the model, based on literature indicating no evidence of benefit with other agents currently prescribed for the same indication as icosapent ethyl. Icosapent ethyl is a highly purified version of EPA. Feedback from the clinical experts consulted by CADTH indicated that niacin, fibrates, and fish oils containing EPA (the active ingredient in icosapent ethyl), with or without docosahexaenoic acid (DHA), are currently used off-label in clinical practice in addition to the maximum tolerated dose of statin therapy. Although clinical experts consulted by CADTH noted that there is limited evidence of clinical benefit of these agents, which is in line with the sponsor’s rationale for their exclusion, icosapent ethyl would be expected to displace these alternative treatments. The cost-effectiveness of icosapent ethyl compared to these agents in statin-treated patients with elevated triglycerides and established CVD or those at high risk of CVD remains unknown.
CADTH Common Drug Review Reanalyses
CADTH undertook the following reanalyses to address the limitations of the model:
applying a lifetime time horizon (i.e., 46 years)
assuming icosapent ethyl would be used for the entire time horizon, meaning that treatment effects and drug acquisition costs were applied for the model’s full time horizon
adjusting utility values to reflect appropriate values from the cited literature:
6
post–non-fatal stroke: 0.628 (standard error [SE] 0.040)
post–unstable angina: 0.880 (SE 0.018)
post–non-fatal myocardial infarction: 0.880 (SE 0.018)
post-acute coronary revascularization multiplier of 1.0 (SE 0).
7
Results of the reanalyses are presented in . Compared with the sponsor’s base case, the CADTH base case resulted in an increase in total costs and total QALYs for both icosapent ethyl plus statins (costs $117,105; QALYs 12.03) and statin therapy alone (costs $67,713; QALYs 11.56), due primarily to the increased time horizon and the assumption that patients would remain on treatment over their lifetime. Incremental costs ($49,392) increased to a greater degree than incremental QALYs (0.47), resulting in an ICUR of $105,053 per additional QALY gained for icosapent ethyl plus statins compared to statin therapy alone.
Summary of Results of the CADTH Base Case.
Additional scenario analyses were undertaken to consider alternative scenarios from those in the CADTH base case:
20-year time horizon: While the latest CADTH economic evaluation guidelines indicate that a lifetime time horizon is most appropriate, a 20-year time horizon was previously used in other CDR pharmacoeconomic submissions for similar indications. Hence, a scenario analysis applying a 20-year time horizon was considered for comparison purposes.
Alternative post-revascularization utility value: In the CADTH base case, it was assumed that patients would return to their baseline utility values following a revascularization procedure. To determine the impact of this assumption on the CADTH base case, an alternative post-revascularization utility value was identified from a systematic review.
20 Specifically, this systematic review identified literature from a study of patients at multiple time points following percutaneous coronary interventions.
19 The difference between the two time periods (0.868 − 0.776 = 0.092) was applied as the sponsor’s post-revascularization utility value.
Full results of CADTH scenario analyses are presented in Table 15. In the 20-year time horizon scenario, incremental costs decreased to $43,522 from $49,392, as reported in CADTH’s base case, while QALYs decreased, resulting in an ICUR of $141,118 per QALY gained. The reduction in costs was primarily due to lower icosapent ethyl acquisition costs, while reduction in QALYs was due to a shorter horizon, over which clinical gain continued. The scenario applying an alternative post-revascularization utility value indicated that the model is sensitive to this parameter and that the estimate of icosapent ethyl’s cost-effectiveness varies depending on the utility input used. This scenario analysis resulted in an increase in the incremental QALYs and a consequent decrease in the ICUR to $93,657 per QALY gained.
CADTH undertook price-reduction analyses, shown in . In the CADTH base case, icosapent ethyl would be the optimal therapy at a willingness-to-pay threshold of $50,000 per QALY gained at a price reduction of 43%.
CADTH Reanalysis Price-Reduction Scenarios.
Patient Input
No patient input was received for this review.
Conclusions
In statin-treated adult patients with elevated triglycerides and other risk factors or at high risk of CVD, the CADTH base-case reanalysis estimated that icosapent ethyl plus statin therapy resulted in increased costs and greater QALYs than statin therapy alone, resulting in an ICUR of $105,053 per QALY gained. Results were primarily driven by drug acquisition costs, with a price reduction of 43% required for icosapent ethyl plus statin therapy to be cost-effective at a willingness-to-pay threshold of $50,000 per QALY.
The cost-effectiveness of icosapent ethyl compared to relevant comparators currently used in clinical practice, in addition to statin therapy, or in a broader clinical population beyond what has been studied in the REDUCE-IT trial, is unknown. Cost-effectiveness may further differ among patients classified by different risk strata (i.e., primary prevention or secondary prevention) within the indication.
