Conventional ultrafiltration (UF) fails to reverse satisfactorily hemodilution and the rise in total body water (TBW) seen after cardiopulmonary bypass (CPB). We have modified the technique, timing, and placement of UF in the CPB circuit and in pilot studies observed controlled elevation of hematocrit and a significantly reduced rise in TBW. We have carried out a prospective randomized study in 50 children undergoing open-heart surgery, comparing modified UF (MUF) with nonfiltered controls. MUF was carried out for 10 minutes after completion of CPB to a hematocrit of 36-42. Fluid balance, TBW (by bioimpedance), and hemodynamics were recorded for 24 hours postoperatively. The results were analyzed using Mann-Whitney U test, comparing controls (n = 24) to ultrafiltered (n = 24). There was one death in each group. Blood loss (ml/kg/24 hr) was 19.5 median (range, 9-30) in the controls versus 12.5 (8-22) in MUF (p = 0.0002); blood transfused (ml/kg/24 hr) 15.5 (3-35) in controls versus 3 (0-11) in MUF (p = 0.0001); colloid transfused (ml/kg/24 hr) 12 (6-56) in controls versus 12 (0-28) in MUF (p = 0.18); percent rise in TBW 11.1 (4.3-16.8) in controls versus 4.0 (1.6-7.9) in MUF (p = 0.0001). There was rise in arterial blood pressure during MUF. Percent rise of systolic blood pressure was 1 (-4 to +9) in controls versus 49 (5-81) in MUF (p = 0.0001); percent rise in diastolic blood pressure 0 (-5 to +8) in controls versus 28 (3-47) in MUF (p = 0.0001). UF reduced the rise in TBW and donor blood requirement associated with CPB in children. The blood pressure rise observed during UF is as yet unexplained, but if proven safe the technique may permit donor blood-free cardiac surgery and prevent the accumulation of potentially dangerous excess tissue fluid.