To determine whether a continuous insulin infusion improves glucose tolerance in extremely low birth weight infants, we conducted a prospective, randomized trial in 24 neonates 4 to 14 days old (mean birth weight 772.9 +/- 128 gm; mean gestational age 26.3 +/- 1.6 weeks). Infants who had glucose intolerance were randomly assigned to receive either intravenous glucose and total parenteral nutrition with insulin through a microliter-sensitive pump or standard intravenous therapy alone. One infant assigned to receive insulin never required it. The groups were similar in birth weight, gestational age, race, gender, medical condition, and energy intake before the study. The mean duration of therapy was 14.6 days (range 7 to 21 days). During the study, the 11 insulin-treated infants tolerated higher glucose infusion rates (20.1 +/- 2.5 vs 13.2 +/- 3.2 mg/kg/min (1.1 +/- 0.1 vs 0.7 +/- 0.2 mmol/L); p less than 0.01), had greater nonprotein energy intake (124.7 +/- 18 vs 86.0 +/- 6 kcal/kg/day; p less than 0.01), and had better weight gain (20.1 +/- 12.1 vs 7.8 +/- 5.1 gm/kg/day; p less than 0.01) than the 12 control infants. The incidence of hypoglycemia, electrolyte imbalance, chronic lung disease, and death did not differ between groups. We conclude that a controlled insulin infusion improves and sustains glucose tolerance, facilitates provision of calories, and enhances weight gain in glucose-intolerant premature infants.