Blood glucose levels were compared in eight type I diabetic subjects who were given closed-loop infusions of insulin by intraperitoneal (i.p.) and intravenous (i.v.) routes, in a cross-over randomized study. After a test meal, plasma glucose peaks were significantly higher with i.p. than with i.v. infusion (174 +/- 22 versus 129 +/- 29 mg/dl) and marked hypoglycemia occurred after 180 min in five of eight subjects. These observations appear to be the consequence of a 60-min lag in insulin rise with i.p. administration. Because of this difference in plasma glucose rise, twice as much insulin was administered i.p. than with i.v. Plasma insulin rose to similar values in both cases. Therefore, with present closed-loop systems, i.p. insulin infusion does not lead to better control of glucose levels than i.v. infusion and does not prevent hyperinsulinism. Adjustments of the artificial B-cell algorithms and the injection of a bolus dose must be tested so that the potential advantages of the i.p. route may be achieved.