This study compares the clinical results of two basic variations of the modified Fontan/Kreutzer operation, direct atriopulmonary connection without an interposed valve versus atrioventricular connection using the native pulmonary valve and the potential pumping capability of the subpulmonary ventricular chamber. From January 1979 through June 1985, 84 patients with tricuspid atresia and ventriculoarterial concordance underwent the modified Fontan/Kreutzer operation at the Mayo Clinic. Sixty patients had atriopulmonary connection and 24 patients had atrioventricular connections. Preoperative characteristics of the two patient groups were similar, but there was a greater frequency of Waterston shunts in the atriopulmonary group (38% versus 17%) and greater frequency of Glenn shunts in the atrioventricular group (46% versus 15%). Mean pulmonary arteriolar resistance was 1.9 +/- 0.7 units in the atriopulmonary group and 1.1 +/- 0.8 units in the atrioventricular group (p less than 0.01). Early postoperatively, mean right atrial pressure was slightly higher in the atriopulmonary group than in the atrioventricular group (18 +/- 3 versus 16 +/- 3 mm Hg, p less than 0.01), but this difference was not reflected in the early or late results. Operative mortality was 5% for patients with atriopulmonary connections and 4% for patients with atrioventricular connections. At 3.5 years postoperatively, the overall survival rate was 89% +/- 4% for patients with atriopulmonary connection and 88% +/- 7% for patients with atrioventricular connections. We conclude that there is no important difference in the clinical outcome of patients undergoing modified Fontan/Kreutzer repair for tricuspid atresia with atrioventricular concordance with either of the two operative methods. The choice of the connection should be dictated by the anatomy, such as presence of pulmonary valve or arterial stenoses, size of outlet chamber, and the presence of anomalous coronary arteries.