Can the risk of hypoglycemia before physical activity be predicted?

It cannot be emphasized enough: regular exercise is important, and even more so for people with type 1 diabetes (T1D), especially children and teenagers. Physical activity contributes to well-being and can help maintain a healthy weight. For people with T1D, it also enables the body to improve insulin sensitivity (i.e., respond better to insulin), which helps to reduce the doses required. Physical activity may also reduce the risk of cardiovascular complications (e.g., stroke).  

However, physical exercise—especially low- to medium-intensity, long-duration activities (e.g., leisurely walking or cycling)—can lower blood sugar levels and cause hypoglycemia. Hypoglycemia can occur not only while exercising (when muscles consume sugar), but also many hours after activity (when the body restores its sugar reserves). A recent Canadian study reported that 49% to 61% of active adults with T1D frequently experience hypoglycemic episodes during exercise, right after or overnight.

These episodes are often difficult to predict, and many children and adults living with T1D avoid playing sports because of their fear of hypoglycemia.   

And what if we could better predict the risk of hypoglycemia and use effective strategies for reducing the drop in blood sugar during and after a hike or a tennis game? Well, this challenge has been met in part by American and Canadian researchers. 

Leveraging artificial intelligence to predict the risk of hypoglycemia before physical activity 

Researchers collected data from 8,827 exercise sessions between 20 and 90 minutes carried out by 459 adult patients with T1D and fed this data to an artificial intelligence model. They were then able to successfully predict the risk of hypoglycemia during physical activity approximately three times out of four. The intelligent system’s predictions on the risk of hypoglycemia before the start of an exercise session were right in 77% of cases. This prediction model could eventually be integrated into closed-loop insulin delivery systems (or artificial pancreases) or mobile applications. 

How does it work? Most types of physical activity can be identified by algorithms (mathematical calculations) through continuous heart rate monitoring or specific sensors (e.g., built into a cell phone or smartwatch). Using these algorithms, the system is able to automatically apply or suggest strategies to anticipate and manage the risk of hypoglycemia, taking into account the user’s personalized parameters (e.g. insulin doses, time of last meal, estimated active insulin [amount of insulin that has been administered and has yet to complete its action], current blood sugar level) or type, duration or intensity of the activity.

Intelligent algorithms, when integrated into insulin pumps, can reduce or suspend insulin delivery, or recommend an amount of carbohydrates and when to consume them. Eventually, mini-doses of glucagon could also be administered (dual-hormone pumps currently in development). 

As for people who use insulin injections, the same kind of mobile application could suggest that they take a given amount of carbohydrates or modify the insulin dose according to the type of physical activity planned. 

Risk of hypoglycemia impacted by type of activity, blood sugar levels, and active insulin 

While analyzing the data from exercise sessions carried out by participants, researchers observed a correlation between the type of exercise and the risk of hypoglycemia.

• Free-living, unstructured aerobic exercise such as walking, hiking or physical labour is associated with a higher risk of hypoglycemia. 
• Structured exercise such as aerobic sessions (e.g., jogging, cycling, swimming), resistance exercise (e.g., weight training) or interval exercise (e.g., series of sprints) following a training video, is associated with a lower risk of hypoglycemia. However, participants might have been better prepared for structured physical activities (measuring blood sugar, eating carbohydrates, reducing insulin) and more aware of the risk of hypoglycemia than those who engaged in unstructured activity.

The researchers also found that certain blood sugar and active insulin characteristics increase the risk of hypoglycemia:

• Low or dropping blood sugar levels before starting physical activity.
• High amount of time spent in hypoglycemia (below 3.9 mmol/L) in the 24-hour period preceding exercise.
• High amount of active insulin in the body at the start of physical activity (an estimated value is displayed on insulin pumps).

Other considerations for better predicting and responding to hypoglycemia

To achieve greater accuracy in predicting hypoglycemia, the algorithms will have to take into account other factors associated with physical exercise, such as intensity, the evaluation of perceived exertion and competition stress. Further studies will be needed before prediction models can be integrated into insulin pumps or mobile applications.

References :

• Bergford, S. et al. (2023). The type 1 diabetes and Exercise Initiative: predicting hypoglycemia risk during exercise for participants with type 1 diabetes using repeated measures random forest. Diabetes technology and therapeutics, 25(9). DOI: 10.1089/dia.2023.0140. https://pubmed.ncbi.nlm.nih.gov/37294539/
• Paiement K. et al. (2022). Is better understanding of management strategies for adults with type 1 diabetes associated with a lower risk of developing hypoglycemia
• during and after physical activity? Can J Diabetes, 46(5): 526–534. https://www.canadianjournalofdiabetes.com/article/S1499-2671(22)00021-1/fulltext
• Brazeau A-S., et al. (2008). Barriers to physical activity among patients with type 1 diabetes. Diabetes Care, 31(11): 2108–2109. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2571055/