Mathematical modelling of human pancreatic islet electrical activity for bio-inspired and fully automated insulin delivery system – MIMICbio

Type 1 diabetes (T1D) is a major disease linked to the destruction of pancreatic beta-cells, resulting in an absolute deficiency of insulin, the only hormone lowering blood glucose level. T1D accounts for 5-10% of the 537 million cases of diabetes worldwide in 2021, which is expected to increase to 642 million by 2040. In France, the number of T1D patients covered by the health insurance is 316 700. The DIALOG study showed that more than 40 000 patients have experienced severe hypoglycaemia, regardless of their mode of insulin therapy, 11% of whom requiring urgent medical care. It is thus widely accepted that T1D diabetes have a major impact on quality of life, notably linked to the mental workload due to T1D care, and can also affect the person's ability to carry out daily tasks. There is therefore a major societal challenge to be addressed.

In this context, development of automated insulin delivery (AID) technologies – also called Artificial Pancreas – dedicated to glycaemic regulation is a solution of great interest to provide a safe, reliable and comfortable care for T1D patients. The AID principle is the following: an electrochemical sensor is used for Continuous Glucose Monitoring (CGM), and this information is next processed by algorithms which drive an insulin pump to deliver the appropriate amount of insulin in real time. However, no definitive solution has yet stood out, since AID systems remain limited to partial automation, i.e. patients must announce meals, estimate the amount of carbohydrates taken during meals, and hypoglycaemia can still occur due to insulin over-administration. In previous studies, use of a static model of the hysteretic activation/inactivation of pancreatic islets (the micro-organs containing beta-cells) in AID therapy without meal announcement gave promising in silico results in the Food and Drug Administration (FDA)-approved UVA/Padova simulator. To increase the technology readiness level (TRL) maturity of islet-inspired AID system, insights of MIMICbio are threefold: A) Build new bio-inspired dynamic mathematical models of human islets based on experimental data (i) reproducing for the 1st time the multicellular activity of beta-cells and their specific kinetics during realistic meals, (ii) and considering the activity of alpha-cells, the 2nd main cell type within islets, and their physiological role as hypoglycaemia detector. B) Design a new model of meal intake regularity to learn from patient’s lifestyle and generate precursor signal for meal anticipation to raise the predictive level of AID systems for hyperglycaemia mitigation. C) Transpose these models into bio-inspired robust control algorithm setup to derive a fully automated health-aware control solution.

The strenght of MIMICbio's methodology lies in a therapeutic strategy that does not involve the use of new sensors or new insulin pumps, i.e. the bio-inspired control strategy is compatible with existing marketed technologies. Note that it could be also fully adapted to next developments in these fields (sensors, pumps, T1D care) to strengthen Europe's technological sovereignty in this area. So, MIMICbio's contributions can be quickly transposable in vivo to the clinic, after validation of proposed solutions, that are planned in the project, through in silico clinical trials in the FDA-approved simulator. This is a strategic choice made to accelerate the transfer of the solution into the clinic and to be ready in the global artificial pancreas market, which is expected to reach USD 624.64 million by 2030, according to Strategic Market Research.