Author + information
- ↵*Address for correspondence: Dr. Mark A. Hlatky, Division of Health Services Research, Stanford University School of Medicine, HRP Redwood Building, Room 150, Stanford, California 94305-5405
Cost-effectiveness analysis has been increasingly used in the field of cardiovascular disease, ever since its first application to medicine some 20 years ago (1). This technique provides a method to identify and weigh the benefits of an intervention against its costs. As discussed in recent overviews of cost-effectiveness (2–5), the effectiveness of an intervention is measured in health terms, typically as years of life saved (or as quality-adjusted life-years, to provide appropriate credit to interventions that relieve symptoms and improve quality of life). The costs of the intervention are measured in monetary terms. The health outcomes and costs of an intervention must be compared with an alternative, and the incremental cost-effectiveness ratio is then calculated as follows:where C1and C2are the costs of the two alternatives, and LY1and LY2are the health outcomes of the two alternatives. Cost-effectiveness ratios of $50,000/life-year or less are generally considered worthwhile in the United States because they compare favorably with the benchmarks provided by the cost-effectiveness ratios for accepted interventions, such as renal dialysis, which costs roughly $30,000/year (6).
Cost-effectiveness analysis has been applied most often to therapeutic interventions, such as thrombolysis for acute myocardial infarction (7), the treatment of elevated cholesterol levels (8)or the use of the implantable cardioverter-defibrillator for patients with a previous cardiac arrest (9). The framework of cost-effectiveness analysis can readily be extended to diagnostic or prognostic tests, but cost-effectiveness analyses of such tests pose conceptual and practical challenges. The result of a test is information about the patient’s diagnosis or prognosis. Although this information may provide some direct health benefits through reassurance of the patient (e.g., a normal coronary angiogram may reduce worry in a patient with chest pain), the major value of the test result is in allowing the physician to choose the therapy best suited for the patient. In turn, the chosen therapy can then provide specific health benefits to the patient. Thus, the health benefits of diagnostic or prognostic testing are largely indirect and are highly dependent on the efficacy of subsequent therapy.
The indirect effect of diagnostic or prognostic tests on health outcomes implies that the cost-effectiveness of these tests can be no better than the cost-effectiveness of the therapies prescribed based on their results. Imagine for a moment the perfect diagnostic test—100% sensitive, 100% specific, with no adverse effects and free of cost. If a positive result on this perfect test identifies a disease for which there is no effective therapy, then the test will have very little value. If the therapy provides only a slight benefit at great expense—say, $10 million dollars/year of life saved—the cost-effectiveness of even a perfect diagnostic test will be equally unattractive. In contrast, if therapy is highly cost-effective, then use of a perfect diagnostic test to look for patients with that disease will also be highly cost-effective. However, real-world diagnostic tests are imperfect, with <100% sensitivity and specificity, nontrivial monetary costs and, in many cases, some physical risk as well. False positive test results will lead to more tests or to unnecessary treatment, whereas false negative test results may provide inappropriate reassurance and a potential reduction in survival. The cost-effectiveness of diagnostic test strategies will therefore generally be less favorable than the cost-effectiveness of the associated therapy.
Testing after acute myocardial infarction
Patients with an acute myocardial infarction have an increased risk of cardiac death for the 6 to 12 months after hospital discharge (10). Many of the postinfarction deaths are sudden and unexpected, presumably due to cardiac arrhythmias. In principle, sudden deaths might be prevented by appropriate antiarrhythmic therapy, either drugs or devices such as the implantable cardioverter-defibrillator. With other therapies of proven benefit in the postinfarction patient (11)—including beta-blockers (12), aspirin (13), cholesterol-lowering drugs (14,15)and angiotensin-converting enzyme inhibitors (16)—the risk of death in the postinfarction year has been reduced considerably. In this setting, it is not clear that routine antiarrhythmic therapy can improve health outcomes further. It is logical to hypothesize that diagnostic testing might be able to identify a subgroup of patients in whom the risk of sudden death is sufficiently higher than average that intervention would be warranted and cost-effective.
The effectiveness of antiarrhythmic drugs in postinfarction patients has been controversial. Early studies of risk stratification after acute myocardial infarction identified reduced left ventricular function and ventricular ectopic beats as independent adverse prognostic factors (10). The association of ventricular ectopic beats with subsequent sudden death provided a strong rationale for use of specific antiarrhythmic therapy to reduce ectopic beats and the risk of cardiac death. The Cardiac Arrhythmia Suppression Trial (17)tested the postinfarction use of class I antiarrhythmic drugs and, surprisingly, documented a significant increase in cardiac death. Quantitative overviews have also suggested that class I antiarrhythmic agents actually increase postinfarction mortality (18). In contrast, early small randomized trials suggested that amiodarone might reduce the risk of death in postinfarction patients (19,20). The recently reported Canadian Amiodarone Myocardial Infarction Arrhythmia Trial (CAMIAT) (21)and European Myocardial Infarction Amiodarone Trial (EMIAT) (22)confirmed that amiodarone reduced sudden death, but they did not document a significant reduction in total mortality. These results have been interpreted as showing that amiodarone is ineffective after myocardial infarction (23), but neither CAMIAT (21)nor EMIAT (22)enrolled enough patients to exclude a clinically important effect of amiodarone in reducing total mortality. Recent meta-analyses (24,25)examined all available randomized trials and suggest that amiodarone does indeed reduce total mortality significantly by 10% to 20%. If one accepts that amiodarone is effective in reducing total mortality after myocardial infarction, the question becomes which patients, if any, benefit sufficiently from amiodarone to justify its risk and expense.
The study by Pedretti et al. (26)examined several strategies to select patients for postinfarction amiodarone treatment. They examined a noninvasive strategy and a two-step approach that combines a noninvasive test with the use of an electrophysiologic study in patients with a positive noninvasive result. The authors focused on heart rate variability as the prototype for a noninvasive test, based on encouraging early results in postinfarction patients (27). However, the specific noninvasive test chosen was not crucial to the analysis because Pedretti et al. show that other noninvasive tests might be used with cost-effectiveness similar to that of heart rate variability (26). The model used by Pedretti et al. did not consider use of amiodarone for all postinfarction patients. The authors’ implicit assumption was that routine use of amiodarone would not be realistic because the risk of sudden death in an unselected cohort of postinfarction patients is probably too low to justify use of amiodarone. Although the authors’ assumption is likely to be correct, it would have nevertheless been helpful to have included both the “treat all patients” and the “treat no one” strategies in any cost-effectiveness analysis of the use of diagnostic testing to select patients for treatment (28). It is only by comparing these two alternatives against the effect of testing followed by selective treatment that the effect of diagnostic testing can be identified separately from the overall effect of treatment in unselected patients.
The most important result of the report by Pedretti et al. was that the two-tier strategy of noninvasive/selective invasive testing was not cost-effective compared with the noninvasive strategy alone. This conclusion held regardless of whether the noninvasive test used was heart rate variability or alternative noninvasive tests alone or in combination. The invasive strategy would be economically attractive only if it had a substantially lower cost or if the sensitivity of invasive testing in the prediction of sudden death were twice as high as indicated in studies to date.
A second conclusion of the authors was that noninvasive testing may be a cost-effective method for identifying and treating patients at risk for sudden death after myocardial infarction. This conclusion is suggestive but must be considered tentative in light of the current uncertainty surrounding the optimal approach to postinfarction testing. Although Pedretti et al. have attempted to analyze the best available information on postinfarction test performance, their analysis can be no stronger than the data on which it rests. Most of the measures of test performance in the model were based on the data from a single study of 575 patients drawn from one hospital (27). We need more information on the performance of tests such as heart rate variability in larger, relatively unselected populations to be confident in the test’s ability to measure risk accurately. In particular, we need to identify tests or test sequences that identify patients at high risk of sudden death, yet low risk of nonsudden death. It is in these patients that the antiarrhythmic therapy has the greatest potential to improve outcome. Because the cost-effectiveness of diagnostic tests is closely tied to the efficacy and cost-effectiveness of therapy selected on the basis of test results, the goal of testing is to find postinfarction patients for whom antiarrhythmic therapy provides a clear health benefit at an acceptable cost.
☆ This work was supported by Grant HS-08362 from the Agency for Health Care Policy and Research, Rockville, Maryland.
Editorials published in Journal of the American College of Cardiologyreflect the views of the authors and do not necessarily represent the views of JACCor the American College of Cardiology.
- by the American College of Cardiology
- Goldman L,
- Garber A.M,
- Grover S,
- Hlatky M.A
- The Beta-Blocker Pooling Project Research Group
- Antiplatelet Trialists’ Collaboration
- Burkart F,
- Pfisterer M,
- Kiowski W,
- Follath F,
- Burckhardt D
- Ceremuzynski L,
- Kleczar E,
- Krzeminska-Pakula M,
- et al.
- Sim I,
- McDonald K.M,
- Lavori P.W,
- Norbutas C.M,
- Hlatky M.A
- Pedretti R.F.E,
- Migliori G.B,
- Mapelli V,
- Daniele G,
- Podrid P.J,
- Tramarin R
- Hartikainen J.E.K,
- Malik M,
- Staunton A,
- Poloniecki J,
- Camm A.J