This study was undertaken to evaluate the usefulness of a canine ischemic heart model achieved by coronary retrograde flow (RF) diversion, and to examine the effect of allopurinol on the myocardial infarction. The left anterior descending coronary artery (LAD) was occluded for 90 min followed by 4h reperfusion. Group 1 (n = 9) was a simple LAD occlusion group. In group 2 (n = 8), retrograde flowing blood from the distal of the occluded LAD was simultaneously diverted during LAD occlusion. In group 3 (n = 8), allopurinol was administered 60 min before ischemia with RF diversion followed by its continuous infusion. Infarcted myocardium was determined by triphenyl tetrazolium chloride staining and myocardium at risk by the dye double perfusion technique. RF diversion significantly reduced not only regional myocardial blood flow (RBF) (0.21 +/- 0.05 ml/min/g in group 1 vs 0.05 +/- 0.01 ml/min/g in group 2; p less than 0.05) but also its variance (p less than 0.01). Furthermore, the infarct size (infarct/risk ratio) in the allopurinol treated group was significantly reduced without any significant difference in rate pressure product, risk size or RBF (89.4 +/- 4.0% in group 2 vs 48.9 +/- 4.5% in group 3; p less than 0.01). We conclude that the canine RF diversion model is useful for myocardial infarct study because of the minimizing of the difference of collateral flow, and that canine myocardial xanthine oxidase may produce free radicals which take part in myocardial injury after reperfusion.