A portable insulin dosing device (Siemens) was used together with a programmable pocket calculator and a glucose analyzer for short-term adaptation of continuous intravenous insulin infusion to blood glucose alterations. A special algorithm was developed which utilizes a given blood glucose value and the glucose rate of change obtained from two to four consecutive samples as input variables. In contrast to current techniques of feedback-regulation, which require continuous glucose monitoring, intermittent blood sampling allows greater mobility of patients. With the semi-automatic feedback system, euglycaemic control was obtained for 12-h periods in ten Type 1 (insulin-dependent) diabetic patients (maximum value 9.50 mmol/l, minimum value 2.83 mmol/l). Severe hypoglycaemia occurred in no case and additional control by glucose infusion appeared to be unnecessary. Light exercise after termination of insulin dose for standard meals led to glycaemic excursions with a rapid decrease (mean 1.08 +/- 0.09 mmol/l), followed by a rebound (0.59 +/- 0.07 mmol/l) in each patient. The amplitude of these excursions decreased with increasing distance from the peak of the meal dose. Comparison of feedback-control alone with feedback by glucose plus preprogrammed dose (4 U/h) at the onset of the test meal revealed lower post-prandial glucose levels (post-prandial maximum +/- SEM: 6.49 +/- 0.18 versus 7.71 +/- 0.79 mmol/l) and a lower infusion rate of insulin for the combined regimen (mean post-prandial maximum +/- SEM: 8.4 +/- 1.2 versus 12.0 +/- 0 IU/h). The system is useful for programming of portable infusion devices and studies based on euglycaemic control in unrestrained patients.