Continuous arteriovenous hemofiltration (CAVH) is an increasingly popular technique in the care of critically ill children. The operating characteristics of the available circuits are largely unknown. Prior to introducing CAVH into our pediatric intensive care unit, we investigated the performance of three CAVH circuits: CAVH with postfilter dilution, CAVH with prefilter dilution (CAVHpre) and CAVH with dialysis counterflow. Using a neonatal lamb model, we measured filter blood flow (QB), ultrafiltrate rate (QU), arterial, venous and ultrafiltrate compartment pressures, oncotic pressure, plus urea levels in blood and ultrafiltrate fluid for the three CAVH circuit designs. Transmembrane pressure and urea clearance were calculated for various values of QB after varying a clamp on the arterial side of the circuit. The major finding, applicable to all circuits, was the wide variability of QB. Constant attention was required in order to obtain a consistent QB. Fluid clearance was effective with all three circuits. Urea clearance averaged 5-10 ml/min and was principally dependent on QU and independent of QB. The addition of dialysis counterflow did not increase urea clearance. The most convenient circuit we tested was CAVHpre, but the problem of unstable QB is common to all unpumped arteriovenous filtrate circuits. It is a major limiting factor in the practical application of this technology to critically ill children.