A run-to-run framework for prandial insulin dosing: handling real-life uncertainty

Individuals with type I diabetes require frequent adjustment of their insulin dose to maintain as near to normal glycaemia as possible. This process is not only burdensome but also, for many, difficult to achieve. As a result, control algorithms to facilitate insulin dosage have been proposed, but have not been completely successful in normalizing glycaemia. Here we present a novel run-to-run control algorithm to adjust the meal-related insulin dose using only post-prandial blood glucose measurements. For each meal independently, the insulin dose is adjusted based on the performance measure for the same meal the previous day. A robustness analysis is performed which considers the sources of uncertainty typically encountered in clinical use. This shows that the system remains stable even with large uncertainty. Copyright (C) 2006 John Wiley & Sons, Ltd.