Pharmacological options for rate control in atrial fibrillation are scarce. Ivabradine was postulated to reduce the ventricular rate in this setting. The objectives of this study were to evaluate the mechanism of inhibition of atrioventricular conduction produced by ivabradine and to determine its efficacy and safety in atrial fibrillation. The effects of ivabradine on atrioventricular node and ventricular cells were studied by invitro whole-cell patch-clamp experiments and mathematical simulation of human action potentials. In parallel, a multicenter, randomized, open-label, phase III clinical trial compared ivabradine with digoxin for uncontrolled permanent atrial fibrillation despite β-blocker or calcium channel blocker treatment. Ivabradine 1 μM inhibited "funny" current and rapidly activating delayed rectifier potassium channel current by 28.9% and 22.8%, respectively (P < .05). The sodium channel current and L-type calcium channel current were reduced only at 10 μM. Ivabradine slowed the firing frequency of a modeled human atrioventricular node action potential by 10.6% and induced a minimal prolongation of ventricular action potential. Thirty-five (51.5%) patients were randomized to ivabradine and 33 (49.5%) to digoxin. The mean daytime heart rate decreased by 11.6 beats/min (-11.5%) in the ivabradine arm (P = .02) vs 19.6 (-20.6%) in the digoxin arm (P < .001), although the noninferiority margin of efficacy was not met (Z = -1.95; P = .97). The primary safety end point occurred in 3 patients (8.6%) on ivabradine and in 8 (24.2%) on digoxin (P = .10). Ivabradine produced a moderate rate reduction in patients with permanent atrial fibrillation. The inhibition of funny current in the atrioventricular node seems to be the main mechanism responsible for this reduction. Compared with digoxin, ivabradine was less effective, was better tolerated, and had a similar rate of serious adverse events.