Cost-effectiveness of implantable cardioverter defibrillators relative to amiodarone for prevention of sudden cardiac death

Background: Implantable cardioverter defibrillators (ICDs) are remarkably effective in terminating ventricular arrhythmias, but they are expensive and the extent to which they extend life is unknown. The marginal cost-effectiveness of ICDs relative to amiodarone has not been clearly established. Objective: To compare the cost-effectiveness of a third-generation implantable ICD with that of empirical amiodarone treatment for preventing sudden cardiac death in patients at high or intermediate risk. Design: A Markov model was used to evaluate health and economic outcomes of patients who received an ICD, amiodarone, or a sequential regimen that reserved ICD for patients who had an arrhythmia during amiodarone treatment. Measurements: Life-years gained, quality-adjusted life-years gained, costs, and marginal cost-effectiveness. Results: For the base-case analysis, it was assumed that treatment with an ICD would reduce the total mortality rate by 20% to 40% at 1 year compared with amiodarone and that the ICD generator would be replaced every 4 years. In high-risk patients, if an ICD reduces total mortality by 20%, patients who receive an ICD live for 4.18 quality-adjusted life-years and have a lifetime expenditure of $88 400. Patients receiving amiodarone live for 3.68 quality-adjusted life-years and have a lifetime expenditure of $51 000. Marginal cost-effectiveness of an ICD relative to amiodarone is $74 400 per quality-adjusted life-year saved. If an ICD reduces mortality by 40%, the cost-effectiveness of ICD use is $37 300 per quality-adjusted life-year saved. Both choice of therapy (an ICD or amiodarone) and the cost-effectiveness ratio are sensitive to assumptions about quality of life. Conclusions: Use of an ICD will cost more than $50 000 per quality-adjusted life-year gained unless it reduces all-cause mortality by 30% or more relative to amiodarone. Current evidence does not definitively support or exclude a benefit of this magnitude, but ongoing randomized trials have sufficient statistical power to do so.