A pig model of cardiopulmonary bypass and cardioplegic arrest was used to compare leakage around manually inflatable and autoinflatable retrograde coronary sinus cardioplegia catheters. Warm and cold blood cardioplegia were alternately delivered through the catheter under study at both low (16 to 24 mm Hg) and high (30 to 40 mm Hg) perfusion pressures. During each experiment, the coronary sinus ostium was sealed around the shaft of the retrograde cannula so that all backflow occurring during cardioplegia delivery could be collected by a separate drainage catheter inserted directly into the coronary sinus approximately midway between the right atrium and the distal end of the balloon. Only two of the 52 manually inflatable cannulas leaked. The leakage in both cases was negligible (< 1% of the total retrograde cardioplegia flow). In contrast, leakage occurred with 57 of the 73 autoinflatable devices tested (p < 0.0001) and averaged 22% ± 3% (mean ± the standard error of the mean) of the total retrograde flow. The temperature of the cardioplegia solution had no effect on leakage. These results suggest that autoinflatable balloons inconsistently seal the coronary sinus during cardioplegia infusion, thereby decreasing the amount of flow that effectively reaches the myocardium. This should make surgeons cautious about using them with warm blood cardioplegia, which largely relies upon the delivery of sufficiently high retrograde nutritive flows for preventing cardioprotective aerobic arrest from becoming a life-threatening ischemic arrest. © 1994.