Search

Diabetes mellitus designates a range of metabolic disorders characterised by hyperglycaemia due to deficient or absent insulin secretion, insulin action, or both. In particular, Type 1 diabetes is characterised by a total lack of endogenous insulin secretion which has to be replaced by exogenous insulin to control the plasma glucose concentration. An extracorporeal wearable artificial pancreas (AP) has been a research aim for over three decades. The research is motivated by the need to improve glucose control. Results of a major study, the Diabetes Control and Complications Trial (DCCT), have demonstrated that improvements in glucose control prevent or delay long term complications, which are the main causes of morbidity and mortality in subjects with Type 1 diabetes. Prior to a clinical evaluation, performance of new medical devices can be tested in silico. Such an approach has been adopted extensively by the pharmaceutical industry in the development of new drugs. In silica testing benefits from relatively low financial, human, and time costs by comparison with the resources required for a full clinical evaluation. The aims of the present thesis are to identify components of the AP, integrate them into a simulation environment, and design an in silico evaluation strategy for the development of closed-loop algorithms with the ultimate goal to assess safety and efficacy prior to clinical evaluation. In the present work, submodels of metabolic processes were linked to represent the characteristics of the glucoregulation in Type 1 diabetes. The submodels were associated with sets of parameters to account for variability in population and individual responses to meals and insulin therapy. The model of glucoregulation in Type 1 diabetes was extended by models of subcutaneous (sc) glucose sensing and sc insulin delivery to represent all aspects of the AP. A systematic approach was developed and employed to evaluate, in silica, the potential and limitations of an AP glucose controller. This was exemplified by evaluating a nonlinear model predictive controller. The robustness of the AP was explored by hypothesising various perturbations induced by different system components. A further objective included the establishment of a qualitative grading scheme of glucose control from the clinical viewpoint. This was followed by a comparison between results from simulations and a clinical trial of 24 hours, which gave the proof of concept of in silica testing. It was found that despite discrepancies due to initial conditions and meal differences, the simulations indicated well the outcome of the clinical trial. In conclusion, the thesis demonstrates the significant potential of in silica testing to make predictions about system behaviour aiding the assessment of safety and efficacy of control algorithms during the development of an AP.