Twenty-five of 49 patients who underwent a Fontan type operation had complex lesions other than tricuspid atresia with ventriculoarterial concordance. Three patients had significant subaortic stenosis. Thirty-four palliative operations, including nine Glenn shunts, were performed before the Fontan operation. Direct atriopulmonary anastomosis was performed in 21 patients. In four, valved conduits were used. Twelve patients had right atrioventricular valve patch closure (three had running and nine had interrupted suture technique). On the basis of the presence of increased or decreased pulmonary blood flow before any surgical intervention, patients were divided into Group 1 (previous pulmonary artery banding, N = 8) and Group II (pulmonic stenosis, N = 17). Postoperatively, in Group I, 87% had significant effusions, mean right atrial pressure was higher (20.6 +/- 6.5 torr), and hospital stay longer (31 days). In Group II, 40% had significant effusions, mean right atrial pressure was lower (16.5 +/- 4.3 torr), and hospital stay shorter (15 days). Significant atrioventricular valve patch disruption occurred in three patients (two had running suture technique), and conduit occlusion occurred in two. Four patients (three with subaortic stenosis and pulmonary artery banding) without an established Glenn shunt required Fontan takedown for persistent low cardiac output, two of whom died (2/25 or 8%). There were three late deaths (3/23 or 13%). Nineteen of 20 surviving patients observed from 2 months to 6 years are doing well. We believe that early Fontan takedown in patients with persistent low cardiac output, interrupted suture technique for atrioventricular valve closure, avoidance of valved conduits, and a preliminary Glenn shunt in patients with pulmonary artery banding and/or subaortic stenosis can further improve the results with the Fontan operation for complex lesions.