Combining basal-bolus insulin infusion for tight postprandial glucose control: an in silico evaluation in adults, children, and adolescents

Abstract

Background: Achieving good postprandial glycemic control, without triggering hypoglycemia events, is a challenge of treatment strategies for type 1 diabetes subjects. Continuous subcutaneous insulin infusion, the gold standard of therapy, is based on heuristic adjustments of both basal and prandial insulin. Some tools, such as bolus calculators, are available to aid patients in selecting a meal-related insulin dose. However, they are still based on empiric parameters such as the insulin-to-carbohydrate ratio and on the physicians' and patients' ability to fit bolus mode to meal composition.

Methods: In this article, a nonheuristic method for assessment of prandial insulin administration is presented and evaluated. An algorithm based on set inversion via interval analysis is used to coordinate basal and bolus insulin infusions to deal with postprandial glucose excursions. The evaluation is carried out through an in silico study using the 30 virtual patients available in the educational version of the Food and Drug Administration-accepted University of Virginia simulator. Results obtained using the standard bolus strategy and different coordinated basal-bolus solutions provided by the algorithm are compared.

Results: Coordinated basal-bolus solutions improve postprandial glucose performance in most cases, mainly in terms of reducing hypoglycemia risk, but also increasing the percentage of time in normoglycemia. Moreover, glycemic variability is reduced considerably by using these innovative solutions.

Conclusions: The algorithm presented here is a robust nonheuristic alternative to deal with postprandial glycemic control. It is shown as a powerful tool that could be integrated in future smart insulin pumps.

Figure 1

Three-dimensional feasible set of insulin infusions and its corresponding 2D projection onto the basal–bolus dimensions.

Figure 2

Plot that illustrates all possible bolus administration modes. The figure is normalized with respect to the patient’s nominal basal and standard bolus from its I:C. Therapies with higher basal infusion than nominal correspond to the dual-wave and square bolus administration modes already implemented in insulin pumps. A decrement in the postprandial basal infusion results in the innovative temporal basal decrement mode. Therapies with nominal basal correspond to the standard bolus mode.

Figure 3

Plot that shows the evolution of the 2D basal–bolus projection feasible sets for a particular example and different carbohydrate content meals. For 40 g, any administration mode will lead to a good postprandial control according to the defined constraints. For 60 and 80 g, a square bolus is not feasible. For values greater than 100 g, only a temporal basal decrement bolus will lead to a good postprandial control. As the carbohydrate content of the meal increases, the projected feasible set shrinks, reducing the possible bolus administration modes. The vertical red line represents the standard bolus strategy, with basal equal to its baseline value.

Figure 4

The two different basal–bolus combination approaches compared with the standard solution. The green point represents the centroid basal–bolus combination, whereas the pink point represents the maximal-bolus solution. The grey point represents the standard bolus strategy with basal equal to its baseline value and the bolus given by the I:C. In this particular example, standard therapy is out of the set of feasible solutions.

Figure 5

Mean glucose response of the 10 adults in the UVa simulator. The blue line represents the response applying the standard bolus, whereas the green and the red line correspond to the centroid and maximal-bolus solution, respectively. These latter solutions produce a flatter profile than the one observed with the standard bolus, avoiding late hypoglycemia. In addition, the peak in the glucose profile remains similar or even lower.

Figure 6

Mean glucose response of the 10 adolescents in the UVa simulator. The blue line represents the response applying the standard bolus, whereas the green and the red line correspond to the centroid and maximal-bolus solution, respectively. These latter solutions produce a flatter glucose profile than the one observed with the standard bolus, avoiding late hypoglycemia. The peak in the glucose profile using any of the solutions is similar.

Figure 7

Mean glucose response of nine children analyzed from those available in the UVa simulator. The blue line represents the response applying the standard bolus, whereas the green and the red line correspond to the centroid and maximal-bolus solution, respectively. These latter solutions, although producing a slightly higher peak in the glucose profile than the standard bolus, achieve a flatter glucose profile, avoiding the severe late hypoglycemia.