Peter G. Jacobs, United States of America
Oregon Health and Science University Biomedical EngineeringModerator of 1 Session
Presenter of 2 Presentations
PREDICTIVE FACTORS CONTRIBUTING TO GLUCOSE CHANGES DURING AEROBIC, RESISTANCE, AND HIGH INTENSITY INTERVAL TRAINING IN TYPE 1 DIABETES
- Peter G. Jacobs, United States of America
- Zoey Li, United States of America
- Gavin Young, United States of America
- Peter Calhoun, United States of America
- Robin L. Gal, United States of America
- Roy W. Beck, United States of America
- Jessica Castle, United States of America
- Mark A. Clements, United States of America
- Eyal Dassau, United States of America
- Francis Doyle III, United States of America
- Melanie Gillingham, United States of America
- Corby Martin, United States of America
- Michael R. Rickels, United States of America
- Susana Patton, United States of America
- Michael C. Riddell, Canada
Abstract
Background and Aims
People with type 1 diabetes (T1D) have difficulty with glucose control during exercise. Exercise type and other factors may have different impact on glycemic control. We used linear mixed effects modelling to identify physiologic features most predictive of changes in glucose during aerobic, resistance and high intensity interval training (HIIT).
Methods
Thirty seven potential predictive factors including glucose from continuous glucose monitoring, insulin, physical activity and food data collected during a 4-week free-living pilot study from 44 people with T1D (age 35±15 years, BMI 26.2±3.1 kg/m2, 19±13 years since diagnosis), were evaluated. Participants using multiple daily injections (n=9) or an insulin pump (n=35) were randomized to complete one of three 30-minute exercise videos twice each week (aerobic [n=19], resistance [n=14], or HIIT [n=11]). Completed exercise included 138 aerobic, 83 HIIT, and 82 resistance sessions. Change in glucose during exercise was calculated as the pre-exercise glucose value minus nadir glucose during exercise.
Results
Results showed that higher mean glucose 1 hour before (P<0.001) and lower mean glucose 24 hours before exercise (P=0.09) were associated with a greater drop in glucose during exercise. Higher insulin on board was also associated with a greater drop during exercise, but was not statistically significant after multiplicity adjustment (P=0.34).
Conclusions
Predictive features identified, including pre-exercise glucose level and insulin on board, may help inform new machine learning algorithms to better protect against hypoglycemia during physical activity.
Using machine learning to predict glucose changes during aerobic, anaerobic and mixed forms of exercise in patients with type 1 diabetes
Abstract
Background and Aims
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.
Methods
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
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.
Conclusions
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.