The Artificial Pancreas: How Closed-Loop Control Is Revolutionizing Diabetes

The artificial pancreas is a long-awaited goal for the management of type 1 diabetes, and its development was recently triggered by the development of continuous glucose sensors. Developing the artificial pancreas is a control engineering problem that is challenged by the large delays in insulin absorption, variability in system dynamics between patients and within the same patient, meals, exercise, and sensor errors. Model predictive controllers are at the forefront of current research due their ability to accommodate input constraints, insulin absorption delays, meals, and insulin boluses. Controller designs need to focus on the balance between system complexity, clinical benefits, and patient convenience. Randomized controlled clinical trials are the gold standard to assess artificial pancreas systems, but feasibility trials may be useful to assess the practicality of novel controllers and systems. Mathematical modeling and computer simulations may also accelerate the development of the artificial pancreas and allow optimization of controller designs prior to clinical testing. Topics of future research include artificial pancreas for type 2 diabetes, combining adjunctive therapies with novel artificial pancreas systems, and learning and adaptive algorithms.