Author + information
- Received July 26, 2004
- Revision received November 13, 2004
- Accepted November 16, 2004
- Published online March 15, 2005.
- William S. Weintraub, MD, FACC*,* (, )
- Elizabeth M. Mahoney, ScD†,
- Andre Lamy, MD‡,
- Steven Culler, PhD*,
- Yong Yuan, PhD§,
- Jaime Caro, MD∥,
- Sylvie Gabriel, MD¶,
- Salim Yusuf, MD, FACC‡,
- CURE Study Investigators
- ↵*Reprint requests and correspondence:
Dr. William S. Weintraub, Emory Center for Outcomes Research, 1256 Briarcliff Road, Suite 1N, Atlanta, Georgia 30306
Objectives We sought to evaluate the long-term cost-effectiveness of clopidogrel for up to one year after an acute coronary syndrome (ACS) without ST-segment elevation.
Background The efficacy of platelet inhibition with clopidogrel for up to one year after ACS was demonstrated in the Clopidogrel in Unstable angina to prevent Recurrent Events (CURE) trial, a randomized trial of 12,562 patients in 28 countries that was conducted between 1998 and 2000. Patients were given clopidogrel (300-mg load followed by 75 mg/day) versus placebo, both in addition to aspirin, for a mean of nine months.
Methods We used patient-level clinical outcomes and resource use from the CURE trial and estimates of life expectancy gains as a result of the prevention of the clinical events of death, stroke, and myocardial infarction on the basis of data from external sources.
Results Excluding clopidogrel costs, average costs of hospitalizations alone were $325 less for the clopidogrel arm (95% confidence interval −$722 to $45) using diagnosis-related group-based Medicare reimbursement rates. When including clopidogrel costs ($766 greater for the clopidogrel arm), average total costs were $442 higher for the clopidogrel arm (95% confidence interval $62 to $820). The incremental cost-effectiveness ratio (ICER) on the basis of the Framingham Heart Study was $6,318 per life-year gained (LYG) with clopidogrel, with 94% of bootstrap-derived ICER estimates <$50,000/LYG; based on Saskatchewan, the ICER was $6,475/LYG with 98% of estimates <$50,000.
Conclusions Platelet inhibition with clopidogrel in patients for up to one year after presentation with an acute coronary syndrome is both effective and cost-effective.
The Clopidogrel in Unstable angina to prevent Recurrent Events (CURE) trial evaluated the efficacy and safety of clopidogrel plus aspirin compared with aspirin alone when initiated early and continued long term (up to 12 months, average of 9 months), in patients presenting with unstable angina or a non–ST-segment elevation myocardial infarction (MI) (1). In CURE, clopidogrel reduced the risk of the primary outcome of cardiovascular death, MI, or stroke. Additionally, there was a reduction in refractory ischemia needing urgent intervention, other ischemic events, and pulmonary edema. A recent study evaluated the cost per event prevented with clopidogrel versus placebo on the basis of in-trial results from the CURE trial (2). To estimate the long-term cost-effectiveness of nine months of therapy with clopidogrel in terms of cost per life-year gained (LYG), it is necessary to evaluate the impact of clopidogrel on life expectancy. The purpose of the present study is to evaluate the long-term cost-effectiveness of clopidogrel by translating the reduction in primary fatal and nonfatal events observed with clopidogrel in the CURE trial into estimates of gains in life expectancy to arrive at an estimate of the incremental cost/LYG.
Design of the CURE trial
The CURE study, a large, multicenter randomized controlled trial, has been described previously in detail (1). Briefly, 12,562 patients were recruited from December 1998 to September 2000 at 482 centers from 28 countries. Patients were eligible if they were hospitalized within 24 h of onset of symptoms indicative of acute coronary syndrome (ACS) and did not have significant ST-segment elevation. Patients were assigned randomly to clopidogrel (loading dose of 300 mg followed by 75 mg per day) or a placebo for an average period of nine months. All patients received acetylsalicylic acid (75 to 325 mg daily). The primary clinical outcome was a composite of cardiovascular death, nonfatal MI, or stroke. All patients in both arms received conventional treatments, including thrombolytic agents, heparin, diuretics, antianginal therapy, antihypertensive medication, cholesterol-reducing agents, and glycoprotein IIb/IIIa antagonists during percutaneous coronary intervention (PCI). The CURE trial was approved by the institutional review board at all sites and analytic centers and complies with the Helsinki protocol for protection of human subjects.
Economic analysis plan and assessment of cost
The economic analysis plan was to compare the costs of the two treatment arms and, if the clopidogrel arm was more costly as well as more effective, to perform an incremental cost-effectiveness analysis (3,4). Costs included in the analysis are direct medical care costs for hospitalizations and the cost of clopidogrel. No data are available from the CURE trial to calculate indirect costs due to lost productivity. Cost after the first year and life expectancy were discounted 3% annually. The analysis uses U.S. dollars as the unit of analysis but uses resource use and clinical outcomes from all 12,562 patients.
Cardiovascular health care resource use associated with the index, and all follow-up hospitalizations were recorded prospectively, including diagnostic tests, therapeutic procedures, and medications. Ambulatory care, including outpatient diagnostic procedures and testing (other than coronary angiography), was not recorded. Because most resource-intensive procedures and tests are performed while patients are hospitalized, it is likely that most of the major components of health care resource use were collected. Possible exceptions would include same-day testing not requiring hospitalization, such as nuclear imaging studies or echocardiograms, and nursing home and rehabilitation stays after stroke. The use of medication other than study drug was not found to differ between the study arms. The use of open-label clopidogrel and adherence to double-blind treatment were recorded during hospitalizations and at periodic follow-up visits.
The initial and subsequent hospitalizations were assigned a diagnosis-related group (DRG), as used in the Medicare program in the U.S., by coders who were blinded to treatment group. Noncardiovascular follow-up hospitalizations were recorded but not included in this analysis because their frequency was rare and because they were distributed equally between treatment arms. Costs for each DRG were estimated using average Medicare reimbursement rates (5,6), which were obtained from the Medicare Part A data file (7), and average private payer reimbursement rates, which were obtained from the MEDSTAT database (5,8). A blended MEDSTAT-Medicare cost estimate was generated by applying MEDSTAT costs to the CURE trial patients younger than 65 years of age and Medicare costs to patients older than 65 years of age. The MEDSTAT estimates include professional costs; for Medicare, professional costs were calculated as a percentage of hospital costs by DRG, according to the method of Mitchell et al. (9). Costs beyond the trial period were estimated as the average per capita participant Medicare reimbursement of $4,370 in 2001 (10).
Life expectancy estimation
Life expectancy for patients with and without nonfatal events was estimated from two independent sources: the Framingham Heart Study (11,12) and the Saskatchewan Health database (13). For the latter, survival data on all 15,956 patients with acute MIs as defined by the International Classification of Diseases-9th Revision-code 410, from the years 1990 to 1995 with follow-up through the end of 2000, were analyzed using published methods (13). Briefly, mean survival beyond the end of the trial was estimated by integrating the survival curves, adjusted for various patient characteristics, including experience of specific subsequent ischemic events (13,14). Cox proportional hazards survival models for each time period were derived for patients with age, diabetes, previous MI, previous ischemic stroke, previous coronary artery bypass graft, and hypolipidemic use as covariates (13–18). The cumulative survival functions over the course of time were derived by applying the hazard functions in sufficiently brief intervals that the hazard can be assumed to be constant during the period.
For both the Framingham Heart Study and Saskatchewan Health database, gender and age-specific estimates of life-years lost due to events were obtained by subtracting life expectancy estimates for individuals with a given event pattern from life expectancy estimates for individuals with no events (14). These estimates were then applied to the CURE trial patient population. For patients who experienced multiple events of different types during the trial, lost life expectancy was estimated assuming a hierarchy of death, stroke, and MI. It was assumed that clopidogrel would be stopped at the end of the trial and that there would be no further reduction (or increase) in nonfatal events between the two arms. As a sensitivity analysis, cardiovascular death (as used in the clinical endpoint) rather than all-cause death (the primary analysis for the economic study) was used to estimate years of life lost. The difference between treatment groups in average life-years lost because of events (placebo − clopidogrel) yields an estimate of LYG with clopidogrel. Variability associated with these lost life expectancy estimates was not accounted for in the cost-effectiveness analyses.
Estimation of cost-effectiveness
The cost-effectiveness of clopidogrel was expressed as the incremental cost-effectiveness ratio (ICER), which is the added cost in the clopidogrel arm divided by LYG. Bootstrap methods (5,000 replicates) were used to estimate the 95% CIs for both cost and LYG with clopidogrel (19). Sensitivity analyses included reducing life expectancy gains by 50% and then by 80%, adding estimated costs associated with bleeding, adding costs due to added years of life, and quality adjusting survival. The impact of bleeding on cost could not be calculated directly because the costing was based on DRGs, which would not necessarily be affected by bleeding. Therefore, the impact of bleeding on hospital length of stay (LOS) was estimated using CURE trial data, and the concomitant cost increase was estimated assuming an average cost of $2,000 per day.
Summary of the clinical data
No differences existed between treatment groups in age, gender, or MI either at presentation or in previous history, diabetes, or hypertension (Table 1).The clinical outcome, the composite of all-cause death, MRI, or stroke by 12 months follow-up was significantly lower in the clopidogrel group, 9.3% versus 11.4%, p < 0.001, relative risk 0.80 (95% CI 0.72 to 0.90). There were trends toward a reduction in death from any cause as well as stroke. There was a significant reduction in MI. Bleeding was increased with clopidogrel. Hospitalizations by treatment group for each DRG assignment, and their associated unit costs are shown in Table 2.
Estimation of LYG
Using both the Framingham Heart Study and Saskatchewan Health database, there were trends toward life expectancy gains with clopidogrel as a result of the reduction in each of the individual events of death, MI, and stroke (Table 3).Overall, patients in the clopidogrel arm were estimated to have gained an average of 0.0699 life-years relative to patients in the control arm using Framingham data and 0.0682 life-years using Saskatchewan. Overall nine-month mortality for Saskatchewan patients with characteristics similar to the CURE placebo group was 7.7%, compared with 5.9% within CURE. Mortality at nine months in the CURE trial was 4.0% in patients who did not suffer and MI and 26% in patients who did suffer an MI between randomization and nine months (hazard ratio 5.74, p < 0.0001 adjusted for age, gender, previous MI, diabetes, smoking). This result compares with a nine-month mortality of 3.7% without an MI and 25% with an MI using Saskatchewan data. Similarly mortality at nine months in the CURE trial was 5.1% in patients who did not suffer a stroke and 23% in patients who did (hazard ratio 4.3, p < 0.0001 corrected for age, gender, previous MI, diabetes, smoking).
Exclusive of clopidogrel costs, initial hospitalization, rehospitalization, and total costs tended to be lower in the clopidogrel arm (Table 4).On the basis of Medicare costs, total costs were significantly higher in the clopidogrel group, whereas for MEDSTAT and the Medicare/MEDSTAT blend, the differences were not significant. Total costs based on Medicare for subgroups defined according to age, gender, diabetes, and previous MI are shown in Figure 1.Results for all subgroups resemble those for the total population, with trends toward slightly higher costs in the clopidogrel arm.
Using Framingham study-based estimates, not including costs beyond the trial period, the ICERs ranged from $4,833 with MEDSTAT to $6,318 with Medicare, with more than 90% of bootstrap estimates <$50,000/LYG (Table 5).Similarly, using Saskatchewan data-based estimates, ICERs ranged from $4,953 with MEDSTAT to $6,475 with Medicare, with more than 95% of bootstrap estimates <$50,000/LYG. The ICERs are systematically higher using the average Medicare cost per year to estimate costs beyond the trial period. A plot of the bootstrap-derived joint distribution of cost and effectiveness differences on the basis of Medicare costing and Framingham-based effectiveness estimates is shown in Figure 2,and cost-effectiveness acceptability curves on the basis of Medicare, MEDSTAT, and the blended costing approaches and Framingham life expectancy estimates are shown in Figure 3.The three costing approaches yielded similar results, with most of bootstrap-derived cost-effectiveness estimates <$15,000/LYG. Incremental cost-effectiveness ratios for subgroups, on the basis of Medicare costing and both Framingham and Saskatchewan life expectancy estimates, are shown in Figure 4.Results for most of the subgroups are similar to the overall estimate, with the exception of women, for whom the ICER was higher. Because of four more noncardiovascular deaths in the clopidogrel arm, ICERs for women are considerably lower when cardiovascular death was considered in the estimation of life-years lost (along with nonfatal stroke and MI), rather than all-cause death ($29,130 on the basis of Framingham and Medicare, $49,369 on the basis of Saskatchewan and Medicare). For all other subgroups there was little difference in ICERs on the basis of all-cause versus cardiovascular death.
These analyses required the use of external databases to project life expectancy beyond the end of the trial. The life expectancy gain with clopidogrel may be either smaller or larger than projected. If the estimated gain in life expectancy is only half of that projected, using the blended costing approach and Framingham life expectancy estimates, the ICER would be $9,820, with 91.9% of bootstraps samples <$50,000/LYG; on the basis of Saskatchewan data, it would be $10,065, with 96.7% <$50,000/LYG. If the life expectancy gain is just 20% of that projected, the ICER would be $24,549, with 77.7% <$50,000/LYG on the basis of Framingham, and $25,161, with 81.8% <$50,000/LYG based on Saskatchewan.
For hospitalizations in which a major bleed occurred, LOS increased 5.94 days. In hospitalizations in which a life-threatening bleed occurred, LOS increased 4.57 days, and a major and life-threatening bleed 9.91 days. Combining the LOS data with the bleeding incidence from Table 1adds an incremental average of 0.05392 days to LOS. If an additional day in the hospital, independent of the DRG assignment, costs $2,000 per day, then incremental bleeding due to clopidogrel adds $108 to the average cost per patient. This would increase the ICER for Framingham, using Medicare costing, from $6,318 to $7,868.
Utility was not measured in CURE and, thus, we could not calculate quality-adjusted life-years (QALYs) directly using patient-level data. Because there were more nonfatal events in the placebo arm, it might be reasonable to expect utility to be higher in the clopidogrel arm, rendering results described above conservative. However, if utility is <1.0 but equal for both arms, the ICER in terms of cost per QALY gained would be higher than the cost per LYG estimate. If utility was 0.80 in both arms (20), then the ICER using Framingham life expectancy and Medicare costs would be $7,898/QALY gained. If costs in added years of life were also included, the ICER would be $11,430/QALY gained.
For patients with ACS, antiplatelet therapy with clopidogrel is cost-effective, with ICERs ranging from $4,910 to $6,473/LYG, and with >90% of bootstrap samples <$50,000/LYG for all models (21,22). These results were consistent for the three DRG-based approaches to costing hospitalizations and the two approaches to estimating the gain in life expectancy by the prevention of events. The results remained favorable when estimated direct medical care costs associated with increased life expectancy in the clopidogrel arm are included and when estimated quality-of-life adjustments were incorporated. The ICERs were favorable for nearly all subgroups considered. Only in women were estimated ICERs >$50,000 per LYG, due in large part to a trend toward more noncardiovascular deaths in women in the clopidogrel arm. Such results highlight the difficulty in drawing inferences with respect to both clinical efficacy and cost-effectiveness in underpowered clinical trial subgroups (23). Incremental cost-effectiveness ratios for women became more attractive when life-year estimates were made on the basis of cardiovascular death, which was a prespecified end point in the CURE trial.
Antiplatelet therapy with aspirin and adenosine diphosphate blockers has been shown to be beneficial in the treatment of vascular disease and to prevent thrombosis after PCI. In the CURE trial, clopidogrel therapy for up to one year was shown to prevent cardiovascular events in patients presenting with ACS (1). There was a slightly increased risk of non–life-threatening bleeding with clopidogrel, which largely was associated with procedures within the first 30 days after randomization, and managed successfully (24). The beneficial effects of clopidogrel also were found in the subset of patients who underwent PCI in the PCI-CURE study (25) and in Clopidogrel for the Reduction of Events During Observation (CREDO) trial (26).
The present analysis builds on an assessment of short-term cost-effectiveness conducted from the perspective of the United Kingdom, U.S., Sweden, France, and Canada on the basis of results from the CURE trial (2). In that study, cost per primary outcome event avoided with clopidogrel was £10,366 in United Kingdom, $22,484 in the U.S., SKr 127,951 in Sweden, €16,186 in France, and C$7,973 in Canada (2). In addition, a Swedish study suggested that clopidogrel is cost-effective in cost/LYG (27). One additional analysis, using decision analytic methodology, suggested that multiyear therapy with clopidogrel would only be cost-effective in patients who cannot tolerate aspirin (28).
A strength of the present analysis is that it was performed with patient-level data directly from the CURE trial. Moreover, as a randomized comparison, the assessment of both effectiveness and cost reflect the actual exposure to treatments independently of selection bias. In the CURE trial, concomitant medicine use reflected care consistent with American College of Cardiology/American Heart Association ACS guidelines; in particular, statins, angiotensin-converting enzyme inhibitors and beta-blockers. Thus, the CURE trial serves as a contemporary evaluation of the effectiveness of clopidogrel and aspirin versus aspirin and placebo.
One limitation of our study is that the CURE trial was a multinational trial and our approach of applying U.S. costs to trial-wide hospitalizations on the basis of DRGs does not fully account for possible differences in treatment practices and resource use between countries or health care systems. If a large proportion of patients come from countries for which the threshold for hospitalization differs considerably from that in the U.S., the difference in costs between treatments may be underestimated or overestimated. However, unless within a DRG the costs are higher in one treatment arm, this approach to costing hospitalizations should yield unbiased overall cost estimates and, in fact, should reduce unwanted variability in the evaluation of cost differences attributable to treatment.
No outpatient treatment, rehabilitation, or nursing home resource use was collected in the CURE trial, and associated costs are therefore excluded from this analysis, as well as indirect costs associated with lost productivity. However, those costs might be expected to be higher in the placebo arm because of the higher in-trial nonfatal event rate, which would result in our estimates being conservative.
The costs of bleeding were not directly measured in the CURE trial. However, when analyzed using LOS as a proxy for cost, the effect on the ICER was modest. This sensitivity analysis may double count some of the cost of bleeding because some of this cost may be accounted for by DRG assignment, thus increasing the ICER.
External databases were used to estimate life expectancy. Multiple improvements in medical care have extended life of patients with vascular disease such that a database like Framingham may not adequately reflect these improvements. However, the Saskatchewan patient sample was based on index hospitalizations occurring between 1990 and 1995, yielding similar life expectancy results to Framingham. That the two external databases yielded similar results offers reassurance that the estimates used in this analysis are reasonable. More importantly, although nine-month mortality was actually slightly higher in the Saskatchewan patient sample than in the CURE trial, mortality rates with and without an MI were similar between the two sources. Although the Saskatchewan data still may not perfectly reflect current prognosis during a longer term, a greater hazard of death without contemporary therapy would mean the projected life expectancy would be shorter, rendering these results conservative. Similarly, the data from Saskatchewan was all from MIs, which may have a higher hazard long term than the ACS patients in the CURE trial, again rendering the present analysis conservative. In addition, there is little doubt that events do predict future increased mortality (29,30). If the prognostic importance of events in the CURE trial was just 20% of that in Framingham and Saskatchewan, the ICERs would still be reasonable. The impact of an MI occurring in the CURE trial between randomization and nine months follow-up was large, with a mortality of 26.3% versus 4.0% in patients suffering versus not suffering an MI; this suggests that these events are worth preventing.
From a U.S. societal perspective, clopidogrel for up to one year in the setting of ACS is cost-effective according to commonly used benchmarks.
The authors thank Patrick Moneuse for statistical assistance.
Supported by grants from Sanofi-Aventis and Bristol-Myers Squibb to Drs. Weintraub, Caro, and Yusuf. Dr. Yuan is an employee of Bristol-Myers Squibb. Dr. Gabriel is an employee of Sanofi-Aventis.
- Abbreviations and acronyms
- acute coronary syndrome
- Clopidogrel in Unstable angina to prevent Recurrent Events trial
- diagnosis-related group
- incremental cost-effectiveness ratio
- myocardial infarction
- length of stay
- life-year gained
- percutaneous coronary intervention
- quality-adjusted life-years
- Received July 26, 2004.
- Revision received November 13, 2004.
- Accepted November 16, 2004.
- American College of Cardiology Foundation
- Lamy A.,
- Chrolavicius S.,
- Gafni A.,
- et al.
- Gold M.R.,
- Siegel J.E.,
- Russel L.B.,
- Weinstein M.C.
- Drummond M.F.,
- O'Brien B.,
- Stoddart G.L.,
- Torrance G.W.
- Pedersen T.R.,
- Kjekshus J.,
- Berg K.,
- et al.
- ↵Centers for Medicare and Medicaid Services. Medicare Part A. Available at: www.cms.gov/statistics/medpar/shtstay01drg.pdf. Accessed January 10, 2005.
- The Medstat Group
- ↵Mitchell J, Burge R, Lee A, et al. Per case prospective payment for episodes of hospital care. Springfield, VA: U.S. Department of Commerce National Technical Information Service, 1995.
- Fried V.M.,
- Prager K.,
- MacKay A.P.,
- Zia H.
- Peeters A.,
- Mamun A.A.,
- Willekens F.,
- Bonneux L.
- Caro J.,
- Migliaccio-Walle K.,
- Ishak K.J.,
- Weintraub W.S.
- Nieto F.J.,
- Coresh J.
- Lee J.,
- Yoshizawa C.,
- Wilkens L.,
- Lee H.P.
- Efron B.,
- Tibshirani R.
- ↵Bauer C, Chapman D, Lin P, et al. The CEA registry: standardizing the methods and practices of cost-effectiveness analysis. Harvard Center for Risk Analysis. Available at: http://www.hsph.harvard.edu/cearegistry/. Accessed January 10, 2005.
- Goldman L.,
- Gordon D.J.,
- Rifkind B.M.,
- et al.
- Yusuf S.,
- Mehta S.R.,
- Zhao F.,
- et al.
- Lindgren P.,
- Jonsson B.
- Kleiman N.S.,
- Lakkis N.,
- Cannon C.P.,
- et al.