Current cardioplegic techniques do not consistently avoid myocardial ischemic damage in high-risk patients undergoing coronary artery bypass grafting. Alternatively, revascularization without cardiopulmonary bypass is not always technically feasible. We investigated whether an intermediary approach based on maintenance of a beating heart with cardiopulmonary bypass support but without aortic cross-clamping might be an acceptable trade-off. Thirty-seven consecutive patients underwent coronary artery bypass grafting (with an average of two grafts per patient) in a pump-supported, non-cross-clamped beating heart. Inclusion criteria were poor left ventricular function (18 patients; mean ejection fraction, 0.25), evolving myocardial ischemia or infarction (11 patients, 5 of whom were in cardiogenic shock), and advanced age (3 patients; mean age 79.5 years) with comorbidities. Results were assessed primarily on the basis of clinical outcome. In addition, measurements of plasma levels of markers of myocardial damage (troponin Ic) and systemic inflammation (interleukin-6, interleukin-10, elastase) were done in 9 patients before and after bypass. In 6 patients, right atrial biopsy specimens were taken before and after bypass and processed by Northern blotting for the expression of messenger ribonucleic acid coding for the cardioprotective heat-shock protein 70. These biologic data were compared with those from control patients who underwent warm cardioplegic arrest within the same time span. There was one cardiac-related death (2.7%), one Q-wave myocardial infarction, and no strokes. Four other deaths occurred from noncardiac causes, yielding an overall mortality rate of 13.5%. Limitation of myocardial injury was demonstrated by the minimal increase in postoperative troponin Ic levels (3.3 +/- 1.0 micrograms/L versus 6.6 +/- 1.5 micrograms/L in controls; p < 0.05) and the finding that heat-shock protein 70 messenger ribonucleic acid levels (expressed as a percentage of an internal standard) were significantly increased after bypass compared with pre-bypass values (279% +/- 80% versus 97% +/- 21%; p < 0.05). In the control group (cardioplegia), end-arrest values of heat-shock protein 70 messenger ribonucleic acid were not significantly changed from baseline (148% +/- 49% versus 91% +/- 29%), a finding suggesting a defective adaptive response to surgical stress. Conversely, peak levels of inflammatory mediators were not significantly different between the two groups. The eight grafts to the left anterior descending coronary artery that were assessed angiographically, by transthoracic Doppler echocardiography, or both methods were patent with satisfactory anastomoses. In select high-risk patients, on-pump, beating-heart coronary artery bypass grafting may be an acceptable trade-off between conventional cardioplegia and off-pump operations. It is still associated with the potentially detrimental effects of cardiopulmonary bypass but eliminates intraoperative global myocardial ischemia.