(2020-2023) Reference: PID2019-107722RB-C22. Funded by MICINN. 211.508,00 €. P.I. Josep Vehí. Other researchers: Ningsu Luo, Joaquim Armengol, Remei Calm. Coordinated project (UdG, UPV).
The TAILOR project is a multidisciplinary project in the areas of biomedical engineering, systems engineering and control, and endocrinology, addressing new technological challenges for automated glucose control in type 1 diabetes.
In the design of any system for glucose control in type 1 diabetes, maximization of perceived usefulness and ease of use must be pursued to cover the different psychosocial profiles of patients to whom the technology is addressed. This is still an unmet need in the artificial pancreas, where attrition of use has been recently reported during the post-market surveillance for the first commercial hybrid system.
Motivated by this fact, the project TAILOR addresses new technological challenges for automated glucose control in type 1 diabetes aiming at the improvement of glycemic control while matching the patients’ expected interaction with the system, yielding to patient-tailored solutions that maximize perceived usefulness and ease of use to foster the adoption of the technology. As ultimate goal, technological progress to widen the access of technological solutions to all patients that might benefit from them is pursued.
To achieve this goal, either in the artificial pancreas or new technological solutions for insulin pen users integrating a continuous glucose monitor (MDI+CGM), devices should be flexible enough to accommodate to different users, widening its target population. This implies that freedom should be given to the patient regarding management of meals and exercise (with/without announcement, carbs/glucagon) and use of extra wearables (always/sometimes/never) while improving glycemic control with respect to current therapy. To this end, learning tools for a better management of variability and system personalization will be key.
This subproject will address the following challenges:
- Improvement of glycemic control by means of better system personalization through the integration of learning tools into control algorithms to cope with patient’s variability. The behavior of the glucoregulatory system can vary very quickly, especially in hypoglycemic patients or in patients with high glycemic variability. System adaptation to personalize the controller’s tuning to the patient’s current metabolic condition is key for the performance of an AP system. Complexity of the cohorts at hand imposes advanced learning techniques for patient’s condition classification, detection and triggering of suitable adaptation laws.
- Alleviation of decision making burden and improvement of glycemic control in sensor-augmented MDI users. Technically, incorporation of CGM into MDI allows to “close the loop”, analogously to an artificial pancreas, although with distinctive features: (1) they are “human-inthe- loop” systems, that is, all actions require human intervention. Automatic features are then relegated to dosing computations, which can be based on CGM data, glucose predictions and estimations of insulin-on-board, affecting disturbances, insulin sensitivity, etc. in the same way an artificial pancreas does; (2) actions are not periodical (every 5 minutes), but event-driven (at meals) or self-driven (at any moment when a correction may be needed); (3) the above means that reliable long-term glucose predictors may be necessary; (4) modulation of long-acting basal insulin is difficult, limiting insulin actions to “above-basal actions”. In our opinion, conceptualization of MDI+CGM as an event-driven/self-driven closed-loop system can allow a rich transfer of technologies from artificial pancreas research, bringing significant improvements in glucose control for the big majority of patients using insulin pens.