Glu4Pred: A computational tool for design and testing of insulin therapies for patients with type 1 diabetes based on interval simulation

Abstract

Diabetes management is a key factor in preventing clinical complications in type 1 diabetes (T1D) patients, where the use of manual or automatic insulin therapies are designed and implemented to counteract the effects of diabetes. However, a good glucose control may perform poorly if any therapy condition changes suddenly or over time, without the patient’s awareness. This phenomenon can seriously compromise the patient’s health and is usually attributed to uncertainty in the T1D models used to this end. Some simulation tools have been successfully designed to develop strategies for automated control in T1D patients; one of them even has the Food and Drug Administration’s (FDA) approval to be used for pre-clinical testing replacing animal experiments. However, none have explicitly included uncertainty as part of their simulation results. This work proposes a simulation tool named Glu4Pred that includes different sources of uncertainty as well as intra-patient variability to predict a more comprehensive dynamics of the glucose-insulin system. The Glu4Pred tool allows to define a complex scenario of several days or meals, where the patient’s variability and several sources of uncertainty are simulated obtaining as a result an envelope of glucose traces. This is achieved using different mathematical interval models of the glucose-insulin system. Furthermore, the tool includes a risk index outcome that quantifies the risk of experiencing different grades of hypo- or hyperglycemia in the postprandial state from interval simulation. As a result, a new computational tool integrating several sources of uncertainty through interval simulation, that can support the design of decision-aid systems for insulin therapy and glucose control in T1D, was developed.