Objective
The goal of this project is to develop a fully closed-loop system (FCL) for Type 1 Diabetes (T1D) able to: 1) track changes of IS in real time for each individual; and 2) adapt the control algorithm to take these changes into consideration. This system would allow to reduce the loss in performance that is generated by the transition from using hybrid closed-loop (HCL) automated insulin delivery systems, that leverage user interactions for pre-meal boluses and pre-exercise insulin reduction, to fully closed-loop systems that do not require such inputs.
A benefit of HCL systems lies in the fact that individuals provide helpful information each time they interact with the technology and can tune therapeutic actions based on their knowledge and prior experiences with meals and exercise. Given that this transfer of knowledge is not available in the fully closed loop design, the objective of this project is to automatically provide the FCL system with information about metabolic variability that would otherwise be provided by the individual's input, by adapting the controller insulin dosing with real-time tracking of IS. In this way, glycemic control will improve with increased protection against hypo- and hyperglycemia. In this regard, this project relies on the integration of the University of Virginia (UVA) FLC system with an algorithm developed by our group for the real-time tracking of IS from glucose sensor measurements, by first removing the necessity of having accurate meal information as an input to track IS, and then modifying the necessary FCL parameters to optimize insulin dosing modulation according to the obtained IS estimates. To validate this technology, a pilot clinical trial in 10 adolescents with T1D will be performed. Among all age groups, adolescents display marked metabolic variability and IS fluctuations. For them, the many stages of puberty and the influence of hormones contribute to changing IS as they grow. Furthermore, adolescents participate in physical exercise virtually every day, while also dealing with the stress of striving for greater independence, peer and sibling pressure, academic achievement concerns, and the burden of diabetes. Thus, testing the designed system in this patient population represents the strongest validation of the technology and bears large clinical relevance. Thus, this clinical trial will allow us to demonstrate (1) the safety and efficacy of using the SI-informed system to reduce glycemic variability resulting from circadian and inter-day variability in SI and (2) its ability to safely manage sudden fluctuations in SI such as those induced by physical activity.
Background Rationale
Type 1 diabetes (T1D) is a chronic condition where the body can't produce insulin due to the destruction of pancreatic β-cells. People with T1D require lifelong insulin replacement therapy to manage blood glucose levels within a safe target range and prevent hazardous diabetes complications. Despite advancements in glucose monitoring and faster insulin analogs, many individuals struggle to achieve optimal glycemic control as recommended by the American Diabetes Association. Intensifying insulin therapy can help, but it often leads to increased hypoglycemia, a major obstacle to optimal control. This creates a lifelong challenge for individuals with T1D: adjusting insulin doses to maintain target glucose levels while avoiding hypoglycemia.
Automated insulin delivery (AID) systems offer a potential solution to this optimization problem. These systems adjust insulin doses based on continuous glucose monitoring (CGM) readings and predictions of future glucose levels. Extensive research has shown that AID systems significantly improve glycemic control in people with T1D, reducing glycated hemoglobin levels and increasing time spent within the target range, while also mitigating hypoglycemia risks, across time and patient population. However, all currently available AID systems are hybrid closed-loop (HCL), relying on user inputs to optimize system performance. User interaction is needed for pre-meal boluses, and adjusting insulin before exercise, often requiring user-entered insulin therapy parameters (profiles of insulin-to-carbohydrate ratios, correction factors, and basal rates, that change according to the time of day) in an attempt to account for daily insulin sensitivity (IS) variations. Although user interaction is generally seen as burdensome, it provides valuable information and allows individuals to fine-tune therapeutic actions based on their knowledge and previous experiences with meals and exercise. However, as AID systems transition into fully closed-loop (FCL) design, user input is eliminated to minimize the burden. Unfortunately, this also means losing the benefit of user knowledge and experience, resulting in lower performance compared to HCL systems.
In this way, the system proposed in this project would bridge this performance gap by integrating the University of Virginia's FCL system with our group's algorithm for real-time insulin sensitivity tracking from glucose sensor measurements. This IS-informed FCL system will track individual insulin sensitivity changes in real time and adapt the control algorithm accordingly. By incorporating this adaptive component, the FCL AID system will optimize insulin dosing based on each patient's actual insulin needs and handle fluctuations in insulin sensitivity, improving glycemic control quality while minimizing user interaction. To validate this technology, a pilot clinical trial will be conducted in 10 adolescents with T1D. Adolescents experience significant metabolic variability and insulin sensitivity fluctuations due to puberty, hormonal changes, physical activity, stress, and the challenges of managing diabetes while striving for independence. This population represents a crucial test group as they face unique challenges and have diverse clinical needs. This will allow for robust validation of the technology and will provide valuable insights and contribute to improving diabetes management for this highly relevant population.
Description of Project
Insulin dosing in type 1 diabetes (T1D) is complicated by several physiological and psycho-behavioral aspects influencing insulin demand. Fluctuating insulin needs due to metabolic variability are mostly linked to changes of an individual’s insulin sensitivity (IS), driven by factors such as circadian rhythms, physical activity, psychological stress, menstrual cycle, etc. As a result, insulin requirements vary significantly over time within the same person, in addition to differences observed between individuals. Among all age groups, adolescents display marked metabolic variability and IS fluctuations. For them, the many stages of puberty and the influence of hormones contribute to changing IS as they grow. Furthermore, adolescents participate in physical exercise virtually every day, while also dealing with the stress of striving for greater independence, peer and sibling pressure, academic achievement concerns, and the burden of diabetes.
Automated insulin delivery (AID) has been shown to improve glycemic control in children, adolescents, and adults with T1D, and is now part of clinical practice. All AID systems currently on the market for the management of T1D are hybrid closed-loop (HCL) AID systems. These systems rely on user interaction for pre-meal boluses and pre-exercise insulin reduction, and often require user-entered insulin therapy parameters – i.e., profiles of insulin-to-carbohydrate ratios, correction factors, and basal rates, that change according to the time of day, trying to account for intra-day IS variability. These profiles require periodic adjustments, which is challenging and burdensome for both the individual and the healthcare provider and does not ensure optimal glycemic management given its trial-and-error nature. However, despite the burden of interacting with the system and informing therapy profile adaptation, hybrid approaches benefit from the fact that individuals learn how to adapt their therapy and provide helpful information each time they interact with the technology, based on their knowledge and prior experiences with meals and exercise.
As the next generation of AID technology is being engineered, most effort is being dedicated to the design of fully closed-loop (FCL) AID systems that do not require any user interaction. In this case, while user burden is minimized, the transfer of user’s knowledge and experience typical of HCL systems is lost – which is a factor contributing to the lower performance of FCL as compared to HCL. The goal of this project is to reduce this gap, proposing an FCL AID system that is capable of: tracking metabolic changes of an individual in real time; and adapting dynamically to them. The rationale behind the proposed FCL AID system is that by automatically providing information about metabolic variability that would otherwise be provided by the individual's input, adapting the controller insulin dosing with real-time tracking of IS will improve glycemic control with increased protection against hypo- and hyperglycemia. To this end, we plan to integrate our current IS tracking algorithm within the University of Virginia (UVA) FCL AID system. The project involves removing the necessity of having meal information as an input to track IS, evaluating the different modifications required into the system’s design to include the IS estimates, and optimizing the insulin dosing modulation with the estimated IS. The proposed system will be validated on simulated data, leveraging the UVA/Padova FDA-accepted T1D simulator, and then tested in a small clinical trial of 10 adolescents with T1D at the University of Virginia, to demonstrate safety and feasibility of the proposed IS-informed system.
If successful, this project will lead to a novel FCL system able to adapt and safely manage the natural variability of an individual’s insulin requirements, improving glucose control specially for adolescents, who arguably stand to benefit the most from this type of improvement in diabetes care.
Anticipated Outcome
If successful, this project will demonstrate the safety and feasibility of including real-time on demand estimates of the individual insulin sensitivity (IS) to improve glycemic control with a fully closed loop system, due to its ability to handle IS fluctuations originating from intra/inter-day IS variability and from acute disturbances such as physical activity (PA) vs. an IS-Naïve system. With the proposed IS-informed system, we expect to be able to track IS in the presence of varying disturbances in the CGM profile, obtaining reliable SI estimates using information available in a fully closed loop setting (without meal data). We intend to validate the IS-informed FCL system using data from different clinical trials and simulations in our FDA-approved T1D simulator. We also expect to validate the proposed IS-informed system in a small, randomized crossover clinical trial with 10 adolescents at UVA, in two 37- hour admissions where participants will use both systems in random order. With this pilot clinical trial, we expect to demonstrate the safety, feasibility, and efficacy of IS-FCL in (i) tackling intra-day and inter-day IS variability (SA2) and (ii) reducing the glycemic effects of sudden changes in IS generated by structured PA (SA3); as compared to our standard IS-naïve FCL system. For this, the expected outcomes are:
Primary outcome:
Increased time in target range (70-180 mg/dL, TIR) in the 37-hour admission period, using the IS-informed FCL AID system compared to the IS-naïve system.
Secondary outcomes:
• Reduced Time below 70 mg/dL (TBR) and low blood glucose index (LBGI), with non-superior time above 180 mg/dL (TAR) and high blood glucose index (HBGI), over the 16 hours after the exercise session, under the assumption that the IS-informed system can react to the increased IS due to the PA bout.
• Reduced TAR/HBGI with non-superior TBR/LBGI for the 5 hours after breakfast, under the assumption that the IS-informed system can react to the decreased IS due to the dawn phenomenon and morning resistance.
Exploratory outcomes:
• Similar glycemic outcomes in the 10 hours following dinner on day 1 compared to dinner on day 2 when the IS-informed FCL system is deployed, demonstrating the capability of this technology of handling inter-day IS variability.
Relevance to T1D
In adolescence, blood glucose control is very challenging in part because of an increase in insulin resistance during puberty, but also because of increasing independence and risk taking. Moreover, hormonal interplay, increased levels of physical activity and psychosocial stress during these stage in life, can increase metabolic-induced variations in IS. Given the suboptimal control of Type 1 Diabetes that is observed in adolescents even with HCL systems, adaptive algorithms able to manage the acute metabolic events and rapid insulin sensitivity (IS) changes typical of this age group, may be required to handle alterations in insulin needs and sustain optimal glycemic control over time. Evidence is emerging regarding safe use of AID systems during times where insulin action may be changing due to reduced or changed insulin clearance (e.g., chronic kidney disease in type 2 diabetes) or acute IS fluctuations during pregnancy. In the latter situation, AID systems were proven to adapt to the changes of IS during labor, delivery, and the immediate postpartum period. Similarly, in the case of T1D, an IS-informed bolus calculator based on real-time estimates of IS, was found to be safe and feasible in adults with T1D, appropriately reducing postprandial hypoglycemia following an exercise-induced IS increase. These findings highlight the importance of an adaptive algorithm that can adjust in response to individuals’ changing insulin requirements over time, using estimates of IS for dosing modulation and leveraging real-time estimation of this crucial metabolic parameter. Following this approach, the proposed IS-informed FCL AID system will allow to improve glycemic control for the adolescent population, who is exposed both to large physiologically-induced changes in IS due to the growing process and behaviorally-induced changes in IS due to the frequent physical activity typical of this age group.
In summary, this application proposes a novel, adaptive, and personalized FCL AID technology, that tailors dosing decision to the prevailing metabolic needs of the user. If successful, this project will lead to a new paradigm for FCL insulin dosing, that accounts for the actual insulin needs of the patient to make dosing adjustments.