Exercise comes in a variety of forms and intensities and is important for health and fitness in diabetes . however, in people with diabetes, short term alterations in glucose homeostasis is often caused by increased exercise levels. The aimof this presentation is to highlight the general glycemic patters caused by the different types of exercise in people lving with diabetes who are taking insulin. The technical challenges to simulate the normal neuroendocrine responses to exercise in diabetes are also highlighted.
An overview of normal and abnormal physiology is examined
Prolonged aerobic exercise increases insulin stimulated and non-insulin stimulated glucose disposal and hypoglycemia develops. With intensive and competaive exercise events, particularly in the morning, stress hormones levels rise and hyperglycemia ensues.
Insulin delivery strategies for prolonged exercise and for short intense activities are needed to manage glycemia in patients living with diabetes
Exercise can be challenging for people with type 1 diabetes because exercise-induced hypoglycemia during exercise is common. The problem is compounded because different types of exercise can cause different responses. Anaerobic exercise can cause less of a drop in glucose than aerobic. And interval exercise can actually cause glucose to increase. People respond differently to exercise at different times of day with differing insulin on board.
We present data across multiple studies that include CGM, insulin, food, and physical activity metrics in people with T1D during different types of exercise including aerobic, anaerobic, and interval exercises and across various insulin infusion therapies including single and dual-hormone closed loop, sensor-augmented pump, and multiple-daily-injections. In addition to free-living data, we have data on the same subjects, using identical therapies, performing identical exercise at the same time of day to explore the limits of predictability under ideal, repeated circumstances. We present the design of various glucose prediction machine learning algorithms including random forests, multivariate adaptive regression splines (MARS), neural networks, and support vector regression algorithms.
Results indicate the importance of including heterogeneous training data across multiple exercise modalities, insulin therapies, and times of day when building predictive models. Prediction accuracy was not found to improve substantially when including data from prior exercise sessions. CGM features were the most relevant features in improving prediction accuracy.
Predicting glucose changes during exercise requires algorithms that have been trained on large heterogeneous datasets with observations from different exercise modalities, insulin therapies, and times of day.
Achieving the glycemic target for people with type 1 diabetes is challenging. Regular exercise improves glycemic control, fitness, body composition and metabolic profiles, with a recommended 150min of exercise weekly, spread over at least 3 days/week. However, glycemic control during and after exercise can be very challenging.
Artificial pancreas (AP) systems can lead to reduced exposure to hypoglycemia and increased time in range; however, preventing hypoglycemia during and immediately after exercise remains a challenge. Additional health monitoring data (e.g. heart rate and step counts) improved system's performance during exercise; nonetheless the risk for hypoglycemia was not eliminated. Furthermore, recent work has shown that an 80% reduction of basal infusion in the 90min leading to moderate exercise may allow for optimal glycemia.
We present the initial testing of such an exercise-informed AP system, capable of anticpating regular physical activity patterns.
Eighteen adults with T1D (>1year) were enrolled in an open-labelled randomized crossover study (15 completers). All subjects were experienced insulin pump users (>6 months), and had no history of severe hypoglycemia or diabetic ketoacidosis in the past 12 months; pregnancy and clinically significant cardiac conditions we also excluding.
Subjects completed a 4-weeks data collection period followed by two 32h supervised hotel admissions. During the data collection period, data from a continuous glucose monitor (Dexcom G6, Dexcom), activity tracker (Fitbit Charge 2, Fitbit) and subjects’ personal insulin pumps were collected. Subjects were instructed to exercise between 4pm and 7pm for at least 30 min/day and at least 4 times per week. Participants were then admitted to the hotel admissions, testing either a standard or an exercise-informed AP system. Admission started at noon on day 1, and included standardized meals at 1pm, 7pm, 8am, and 1pm; a 3x15min moderate exercise bout occurred at 5:30pm on day 1. During the admission, particpants used a prototype AP system consisting of a t: AP pump (Tandem Diabetes Care), a Dexcom G6 (Dexcom), and an activity tracker (Smartband 2, Sony), all connected to the DiAs platform (UVA) running the chosen algorithm.
Exercised-informed closed loop system significantly reduced the exposure to hypoglycemia during and immediately after exercise by consistently reducing insulin infusion in the hours leading to the bout, with no increased risk of hyperglycemia.
AP systems informed by commercially available activity trackers can learn from users exercise behaviors and approipriately anticipate such glycemic disturbances, potentially leading to reduced glycemic risk, and eventually a more automated system.