A needle-type glucose sensor has been developed using a platinum electrode covered with immobilized glucose oxidase. Experiments with albumin-saline solution in vitro showed that at 5.5 mmol/l glucose concentration the output current generated was 1.2 +/- 0.4 nA (mean +/- SD). The current increased as a linear function of glucose concentration over the range (0-27.7 mmol/l). The response time to reach 90% of the final plateau value was 16.2 +/- 6.2s. The signal-to-noise ratio of the sensor was 15.8 +/- 2.6 decibels. The temperature coefficient in output was 2.3 +/- 1.0%/degrees C. The current output was not affected significantly by changes in oxygen tension of the solution in the range 25-150 mmHg. In vivo, the output current of sensors inserted into the subcutaneous tissues of dogs was directly related to blood glucose concentrations after oral glucose or meals. Daily checking of the sensors maintained in subcutaneous tissues in five dogs showed that the sensitivity decreased gradually to 87.2 +/- 7.6% at 72 h, and dropped significantly to 57.4 +/- 7.0% of the initial output by 96 h. A wearable artificial endocrine pancreas (18.0 x 17.7 x 7.9 cm, 700 g) was developed, consisting of a needle-type glucose sensor, a microcomputer system and a pump driving mechanism. Three pancreatectomized dogs were fitted with the system by inserting the sensor into subcutaneous tissue. By renewing the sensor every fourth day, the device could maintain the daily glucose variations in diabetic dogs within the range 5-9.5 mmol/l for 7 days.

Long-term glycemic normalization in diabetic dogs has been achieved using a portable insulin delivery device. Six pancreatectomized male dogs carried the device for periods of 7 ± 2 (range –12) mo. Pulses of insulin were delivered every 90 s into the inferior vena cava through an exteriorized indwelling catheter In the postabsorptive period, insulin was infused at the mean rate of 0.44 ± 0.03 mU/min/kg and glycemia was 93 ± 7 mg/dl, while in the postprandial period insulin was infused for 7 h at the mean rate of 3.10 ± 0.22 mU/min/kg. A standard daily diet was provided, and the resulting glycemic profiles were similar to an age-, sex-, and weight-matched group of six healthy controls. Peripheral insulin levels in the infused diabetic dogs (15 ± 1 μU/ml) were not significantly different from nondiabetic controls in the fasting state and were unresponsive to caloric provocation. However, in the postprandial period of enhanced delivery, insulin levels in the diabetic dogs were two to four times higher than in the controls. Unaccountable episodes of hypo- or hyperglycemia did not occur. Techniques for long-term vascular access were developed such that catheter-related complications, including infection and thrombosis, were not encountered. This study demonstrates the feasibility of long-term glycemic control in unrestrained diabetic animals using a preprogrammed waveform of insulin delivery. Hyperinsulinism accompanies the glycemic normalization when the hormone is delivered peripherally.

The dynamic property of glucose-induced insulin secretion was simulated with the aid of a control theory, and the relationship between the stepwise input of glucose concentration and the biphasic response of insulin as an output was expressed in the transfer function of proportional plus derivative action to glucose concentration. By utilizing this model, the following computer algorithm for the artificial beta cell was made; rate of insulin infusion = Kp BG + Kd ΔBG+ Kc, where BG was the abbreviation for glucose concentration, ΔBG was the rate of change in BG, Kp and Kd were the coefficient for the proportional and derivative action, respectively, and Kc the constant for basal insulin secretion. Microcomputer, insulin pump and digital printer were packed in small case, which was connected with glucose analyzer. Experiments were carried out in the depancreatized dogs to validate the importance of infused insulin based on derivative action to glucose concentration in glucose regulation. Kd/Kp ratio was changed when insulin was infused following iv glucose bolus injection. The results revealed that when the derivative action was added to the proportional action properly in insulin infusion regulatory system, insulin requirement was the smallest and glucose regulation was the best. The characteristics of the system recognized in clinical application were 1) rate of insulin infusion was small enough to keep the plasma concentration of insulin at a physiological level, then insulin requirements were reduced to around a half of those given subcutaneously, 2) glucose or glucagon to restore hypoglycaemia was not essential.

Long term glucose control in pancreatectomised dogs has been obtained with portal insulin therapy. When compared to a previous similar study using peripheral infusions, 20% less exogenous insulin was required and peripheral fasting insulin levels were 30% lower. Animals (n = 5) were unrestrained, conscious and carried a programmable insulin pump for 163-224 days. In the post-absorptive state blood glucose was normal (87 _+ 5 mg/dl) as was plasma insulin (10 + 1 mU/1) with porcine insulin infused at a basal rate of 0.36 _+ 0.01 mU/kg/min. Following ingestion of a standard mixed meal the infusion rate was increased to 2.47 + 0.09 mU/kg/ rain for 71/2 h resulting in post-prandial normalisation of blood glucose. Peripheral plasma insulin levels were twice normal during the post-prandial infusion, but only half those previously reported with peripheral infusions. Insulin clearance rates were 37 ml/kg/min in the basal state and rose significantly post-prandially. In the absence of extra mealtime insulin the clearance rate was unaffected by the resulting post-prandial hyperglycaemia and similar to values observed with insulin infused peripherally at 0.45 + 0.03 mU/kg/min. No significant increase in the post-prandial rate of insulin clearance relative to the fasting rate was observed with peripherally administered insulin. It was thus concluded that portal insulin replacement in pancreatectomised dogs could normalise both blood glucose and insulin in the fasting state, but post-prandial peripheral insulin levels remained elevated.

The metabolic and hormonal consequences of long term intravenous insulin replacement were studied in 11 pancreatectomised dogs. Insulin was delivered into the portal circulation of six animals for 164-224 days and into the peripheral circulation of the remainder for 123-365 days. Infusion rates were initially adjusted to achieve normoglycaemia in the fasting (0.37 +/- 0.01 mU Kg-1 min-1 portal; 0.45 +/- 0.03 mU kg-1 min-1 peripheral) and post-prandial states (2.57 +/- 0.07 mU kg-1 min-1 for 7 1/2 h portal; 3.16 +/- 0.18 mU kg-1 min-1 for 7 h peripheral). Animals were fed their usual mixed diet and blood samples were drawn from indwelling catheters at regular intervals for 24 h. A matched group of six normal dogs was similarly studied. Significantly less insulin was needed for glycaemic normalisation with portal (1.05 +/- 0.03 U kg-1 day-1) compared with peripheral (1.27 +/- 0.08 U kg-1 day-1) infusions, but post-prandial insulin levels were not normalised. Glucagon levels were normal and unaffected by the route of insulin infusion. Lactate and pyruvate responses were exaggerated post-prandially in the diabetic compared with the normal dogs. Fasting non-esterified fatty acid levels were suppressed with peripheral but normal with portal insulin infusion. There were only minor differences in the branched chain, essential and other non-essential amino acids except for alanine which was significantly above normal in the diabetic animals. Fasting levels of insulin, lactate, pyruvate and non-esterified fatty acids were normalised only with portal infusion while glucose, glucagon, 3-hydroxybutyrate and most amino acids were normalised regardless of the route of infusion. We conclude that the metabolic regulation achieved with portal insulin replacement is closer to normal than that achieved with peripheral infusion.