Biomedical engineering research efforts have accomplished another level of a “technological solution” for diabetes: an artificial pancreas to be used by patients and supervised by healthcare professionals at any time and place. Reliability of continuous glucose monitoring, availability of real-time programmable insulin pumps, and validation of safe and efficient control algorithms are critical components for achieving that goal. Nevertheless, the development and integration of these new technologies within a telemedicine system can be the basis of a future artificial pancreas. This paper introduces the concept, design, and evaluation of the “intelligent control assistant for diabetes, INCA” system. INCA is a personal digital assistant (PDA)-based personal smart assistant to provide patients with closed-loop control strategies (personal and remote loop), based on a real-time continuous glucose sensor (Guardian RT, Medtronic), an insulin pump (D-TRON, Disetronic Medical Systems), and a mobile general packet radio service (GPRS)-based telemedicine communication system. Patient therapeutic decision making is supervised by doctors through a multiaccess telemedicine central server that provides to diabetics and doctors a Web-based access to continuous glucose monitoring and insulin infusion data. The INCA system has been technically and clinically evaluated in two randomized and crossover clinical trials showing an improvement on glycaemic control of diabetic patients.

For a diabetes mellitus patient, tight control of glucose level is essential. Results are reported of an investigation of the suitability of existing wearable continuous insulin infusors controlled and adjusted by a control algorithm using continuous glucose measurements as input to perform the functionality of an artificial pancreas. Special attention was given to the development of a continuous glucose monitor and to evaluate which quality of input data is necessary for the control algorithm. In clinical trials, it was found that for patients in a controlled environment an autonomously regulating control algorithm leads to an improved adjustment of patient glucose values and less overall insulin infusion as compared with the best fixed preprogrammed insulin infusion profiles of standard pump therapy. For the limited number of cases studied here, functionality of the control algorithm could tolerate some delay between the actual glucose values in the patient interstitial fluid and the algorithm input of up to 30 min. A quasicontinuous glucose measurement delivering actual glucose values every 5-10 min seems to be suited to control an artificial pancreas.

The objective of the project Advanced Insulin Infusion using a Control Loop (ADICOL) was to develop a treatment system that continuously measures and controls the glucose concentration in subjects with type 1 diabetes. The modular concept of the ADICOL's extracorporeal artificial pancreas consisted of a minimally invasive subcutaneous glucose system, a handheld PocketPC computer, and an insulin pump (D-Tron, Disetronic, Burgdorf, Switzerland) delivering subcutaneously insulin lispro. The present paper describes a subset of ADICOL activities focusing on the development of a glucose controller for semi-closed-loop control, an in silico testing environment, clinical testing, and system integration. An incremental approach was adopted to evaluate experimentally a model predictive glucose controller. A feasibility study was followed by efficacy studies of increasing complexity. The ADICOL project demonstrated feasibility of a semi-closed-loop glucose control during fasting and fed conditions with a wearable, modular extracorporeal artificial pancreas.

Within the ADICOL research project, we have developed a control-loop application for the automated delivery of insulin. The concept is unique in several aspects: a) Modularity: the ADICOL system consists of three elements, insulin pump, glucose monitor and expert-system, which all run as stand alone devices with the patient, but can be interconnected to form a body network. b) Subcutaneous Application: Both, glucose monitor and insulin delivery use the subcutaneous route to the body. c) Insulin Pump: The insulin infusion mimics a natural pancreas by employing a delivery interval of 3 minutes. d) Expert System: The expert system is running two different algorithms in a hand-held computer, taking advantage of the high-resolution insulin delivery. e) Algorithm: The Model-Predictive Algorithm is based on a feedback optimization of the insulin infusion rate and physiological parameters of the patient. f) Glucose Monitor: Our glucose monitor is based on the open-flow microperfusion sampling technique and employs amperometric glucose biosensors. g) Body Network: The interconnection of the 3 elements to form a control-loop system (or automated pancreas) is achieved by modern wireless technology. Various clinical tests have been undertaken to assess the performance of subsets, results will be shown. We present the overall set-up and its performance, and show latest results from various clinical human trials performed at the Universities of Graz, Austria and Perugia, Italy.

A specially designed field-effect transistor (FET) with a significantly enlarged gate area was applied in a classical urea enzyme FET (ENFET). The resulting high stability and sensitivity toward pH shifts make it predestinated for the measurement of H+ produced in the equilibrium of NAD(+)-dependent enzymatic reactions, especially when the equilibrium is shifted by a subsequent reaction. As a mode, the glucose dehydrogenase (GDH) reaction connected to a ion-sensitive field-effect transistor (ISFET) is demonstrated by which glucose could be determined in the range from 1 to 40 mM. A platinum electrode on the gate of the FET permits the measurement of reduction equivalents (NADH) by means of the recently reported chronopotentiometrical methods. Thus, in principle a way toward a redox-ENFET is shown.