The clinical effects of fluid therapy designed to maintain different degrees of negative water balance during the first five days of life were determined prospectively in 88 very low-birth-weight infants. Infants with birth weights of 750 to 1,500 gm were matched for birth weight in 250 gm increments. RDS or no RDS, asphyxiated or not, and inborn or outborn. Each infant was randomized to either Group 1--fluids managed to allow 1 to 2% loss of BW per day to a maximum loss of 8 to 10%, or Group 2--fluids managed to allow 3 to 5% loss of BW per day to a maximum loss of 13 to 15%. The mean five-day cumulative fluid input in Group 2 was 220 ml/kg less than in Group 1, yet Group 2 lost only 41 gm/kg more than did Group 1 (8.8% of BW lost in Group 1 vs 12.9% of BW lost in Group 2, P less than 0.001). There were no statistically significant differences between the groups in incidence of clinically significant patent ductus arteriosus, intracranial hemorrhage, bronchopulmonary dysplasia, necrotizing enterocolitis, dehydration, acute renal failure, or metabolic disturbances. There was no difference in duration of respiratory support required, in time to regain BW, or in time to discharge. There was no difference in the neonatal mortality rate. Fluid input in VLBW infants can be flexible to allow the gradual loss of 5 to 15% of birth weight during the first week of life without adversely affecting outcome.