An artificial betacell: assessment of the glucose analyser, infusion system and optimization of constants for the algorithms.

The glucose analyser and insulin infusion modules of the Biostator were tested. The infusion system was reliable since more than 99% of the computed volume of insulin solution was delivered by the infusion pump at infusion rates above 1/100 of maximum, and no insulin was adsorbed onto infusion bags or tubing. Blood glucose results from the Biostator were compared with routine laboratory methods during long-term feedback control. Both slope (0.73) and scatter (r=0.87) around the regression line were unsatisfactory when the recommended calibration procedure was used. Tests in fasting non-diabetic subjects showed a significant correlation between the variation in Biostator glucose read-out and the plasma protein concentration in the detector outflow. In diabetics the ratio between Biostator glucose read-out and laboratory glucose determinations declined significantly with time. These observations led to the introduction of a standardization procedure based on externally determined blood glucose concentrations. During long-term feedback experiments in diabetics this procedure resulted in a significant increase in slope (0.84) but no improvement in scatter around the regression line. Repeated OGTTs revealed a set of constants for the algorithms, which enabled normal glucose tolerance to be achieved with smaller amounts of insulin.