We evaluated the efficacy and safety of azimilide, a new class III antiarrhythmic agent that blocks both the slow and fast components of the cardiac-delayed rectifier potassium currents in 4 randomized, double-blind, placebo-controlled trials with similar protocols. The purpose of this study was to assess the relation between dose and effect. A total of 1,380 patients with a documented history of symptomatic atrial fibrillation (AF), atrial flutter, or both, were enrolled. After a 3-day loading period during which the assigned dose was given twice a day, subjects received placebo or azimilide (35, 50, 75, 100, or 125 mg once a day) for the duration of the study period. The primary end point of the studies was the time to symptomatic arrhythmia recurrence with a transtelephonic electrocardiogram typical of AF, atrial flutter, or paroxysmal supraventricular tachycardia. For each study, Kaplan-Meier estimates of the median time to recurrence were computed for placebo and for each azimilide dose. Cox proportional-hazards modeling was used to estimate hazard ratios for each active dose. Each of the 2 highest azimilide doses (100 and 125 mg/day) significantly prolonged the time to recurrence of arrhythmia. For the 100 mg/day dose, the hazard ratio was 1.34, 95% confidence interval 1.05 to 1.72; p = 0.02. For the 125 mg/day dose, the hazard ratio was 1.32, 95% confidence interval 1.07 to 1.62; p = 0.01. Patients with a history of either ischemic heart disease or congestive heart failure had a significantly greater treatment effect from azimilide than those without it. Torsades de Pointes occurred in 0.9% of patients receiving either of the 2 effective doses. Thus, doses of azimilide <100 mg/day are not effective for control of AF, whereas doses of 100 and 125 mg/day are effective with an acceptable risk of serious toxicity.