Ninety-nine adult mongrel dogs underwent acute ligation of the proximal left anterior descending coronary artery. Thirty minutes later, the occlusion was released to evaluate the effectiveness of five antiarrhythmic protocols in eliminating reperfusion ventricular fibrillation. The five protocols included: protocol 1 --i.v. lidocaine, preligation and prerelease (n = 19); protocol 2 -- i.v. lidocaine, prereperfusion only (n = 22); protocol 3 -- chronic, oral, daily amiodarone for 2 weeks preligation (n = 19); protocol 4 -- i.v. procainamide, preligation and prereperfusion (n = 21); and protocol 5 -- i.v. verapamil, prereperfusion (n = 18). Each regimen was evaluated with respect to the incidence of reperfusion ventricular fibrillation in dogs that survived to reperfusion, and the results were compared to 77 control dogs that underwent identical coronary artery occlusion and release procedures without drug therapy. The incidence of reperfusion ventricular fibrillation was as follows: protocol 1 -- seven of 15 dogs (47%); protocol 2 -- six of 18 (33%); protocol 3 -- 11 of 16 dogs (69%); protocol 4 -- eight of 17 dogs (47%); and protocol 5 -- 10 of 17 dogs (59%), compared with 36 of 60 (60%) in control dogs. Using chi-square analysis, protocol 2 was beneficial (p < 0.05). The dogs were then stratified into high- and low-risk subgroups based on the arrhythmic events of the antecedent coronary artery ligation periods, and predictive risk indexes for the occurrence of reperfusion ventricular fibrillation were developed. the Mantel-Haenszel method of statistical analysis revealed that none of these protocols resulted in a statistically significant reduction in the incidence of reperfusion ventricular fibrillation. Thus, use of these predictive indexes plus appropriate statistical methods has revealed, unexpectedly, limitations in the efficacy of a spectrum of antiarrhythmic agents in preventing reperfusion ventricular fibrillation.