Exercise-induced Glycemic Variations and Hybrid Closed-loop Systems

People with type 1 diabetes mellitus (T1DM) are continuously at risk of hypoglycemia, which is one of the main barriers to achieving optimal glycemic control.

Physical activity (PA) in T1DM is characterized by an imbalance between hepatic glucose production and glucose disposal into the muscle, increased insulin sensitivity and impaired counterregulatory hormonal response. Thus, PA could increase the risk of hypoglycemia in T1DM. Hypoglycemia can occur during exercise, as well as during recovery, and fear of hypoglycemia often results in either avoidance of exercise or overcompensatory treatment behaviors, which in turn result in worsened metabolic control and increased cardiometabolic risk.

Complexity of glucose homeostasis and an insufficient level of technology prevents tight blood glucose (BG) control regulation. Closed-loop artificial pancreas studies have shown reduction in the risk of hypoglycemia and increase in time in range (70-180 mg/dl) in T1DM patients. Although these systems work fairly well on overnight hypoglycemia, preventing low BG during and immediately after exercise remains a problem, due to the combination of a dramatic increase in insulin sensitivity with the delayed onset of subcutaneous insulin. Informing insulin dosing of PA could decrease this risk. Glycemic response to exercise varies based on type of exercise (aerobic or resistance), but also based on intensity and duration of exercise. Most existing hybrid closed-loop (HCL) systems use the Dexcom G6 glucose sensor, which has demonstrated good accuracy during aerobic, resistance and high intensity interval training (HIIT) exercise.

Current HCL systems offer the possibility to announce exercise to the system and this information is accounted for in the insulin dosing calculation. As a result, the system sets a higher glycemic target, increases the insulin sensitivity factor and/or avoids correction boluses during exercise. These systems perform well in preventing hypoglycemia, but compensatory hyperglycemia, due to higher carbohydrate intake and lower insulin delivery, often follows. One of the major issues with existing HCL systems is that exercise is considered a binomial entity, either present or absent, and factors like intensity, duration or type of exercise are not taken into consideration. However, these exercise variables can contribute to elicit completely different glycemic responses to exercise. Even more, these systems do not account for the delayed effect of exercise, i.e. the so-called activity-on-board (AOB), a quantitative representation of the previously performed PA that is still affecting the BG levels, and which is responsible for the post-exercise hypoglycemia, sometimes several hours after an exercise session. Current recommendations on how to manage BG levels in exercise have not yet been fully tested in the context of a closed-loop system.

The present study aims to develop a new and improved algorithm that uses information on exercise/physical activity variables to predict glycemic variations and modulate insulin therapy accordingly in order to avoid hypo- and hyper-glycemia and maintain glycemic levels in the desired range. This major aim is to be achieved in a multi-step manner. The first step will consist in data collection. Data relative to exercise/physical activity will be collected in two different settings. First, through the different exercise sessions, exercise-related data derived from heart rate monitor, strength training machines will be gathered. Second, data relative to spontaneous physical activity/sedentary behavior, outside the experimental setting will be gathered through movement trackers. Data relative to glycemic variations, derived from CGMs and data relative to insulin and carbohydrate intake from insulin pumps and food diaries. Through the exercise experimental sessions, investigators will be able to evaluate the effectiveness of hybrid closed-loop systems on maintaining glucose levels in range (70-180 mg/dl) and in preventing both hypo- and hyperglycemia, during and after exercise of different durations, intensities and types. In a second step, the relationships between exercise variables derived from both experimental sessions and PA monitoring outside of the experimental sessions, and glycemic, insulin and carbohydrate variations with physical activity will be investigated.

The third, and last step, will include further data analysis, testing, modelling and in-silico simulations in order to estimate the performances of a new decision support system for PA-informed insulin dosing against standard insulin dosing.

Study design Participants Fifty men and women with T1DM will participate in the study. Preliminary testing Upon signing the informed consent, participants' baseline data will be collected. Then, preliminary testing will take place in the Exercise Physiology Lab at the University of Rome Foro Italico. Peak cardiopulmonary oxygen consumption (VO2peak) will be assessed using a graded cycle ergometer (Lode, Groningen, Netherlands) exercise test to volitional exhaustion. The maximum power output (Wmax) achieved during the test will be used to standardize exercise prescription.

The 1-RM test will be performed for the evaluation of the dynamic maximal muscle strength. The test will be performed before randomization, for lower body push movements and upper body push and pull movements on the leg press, chest press and low row Biostrength® line machines, respectively (Technogym S.p.A., Cesena, Italy). The 1-RM evaluation will be useful for the determination of the individuals' load during the exercise program. During the test, the velocity of the movement will be registered for the analysis of the muscle peak power. Before being assigned to any of the exercise trials, participants will be asked to wear a GENEActiv activity monitor at the wrist for seven consecutive days in free-living conditions. GENEActiv devices are wrist-worn instruments used for evaluation of free-living activities. These devices, which will be provided by the Exercise Physiology Lab, provide raw data for physical activity, sedentary behavior and sleep analysis, capturing movement, light and temperature data. Participants will be provided with a standardized diet for the whole duration of the study and will be asked to keep a food diary on the seven days they wear the accelerometer, 24 hours before the exercise trials and on the days of exercise trials.

Exercise trials Patients will undergo four different exercise trials, moderate intensity aerobic exercise (MIE), high-intensity interval exercise (HIIE), resistance exercise (RE) and combined aerobic and resistance exercise (COMB). The same exercise trials will be performed in two different conditions: Exercise mode set on "on" or exercise mode set on "off" on the insulin pump. Patients will also wear an accelerometer and a heart rate monitor on the trial days. The four exercise sessions will have the same duration of 50 min subdivided into 5 min of warmup, 40 min of work and 5 min of cool down. The 40 min of work will differ between trials as follows.

Moderate intensity exercise will consist of continuous exercise on a cycle ergometer at 40% Wmax.

High-intensity interval exercise will consist in three 10-min bouts of interval exercise at 1:1 work:rest ratio (5 x 1-min intervals at 90% Wmax with one minute recovery at 25% Wmax between intervals) interspersed by 3 min of recovery at 25% Wmax.

Resistance exercise will comprise a circuit of 8 groups of exercises repeated in 3 sets of 12 repetitions each at 70% of 1-RM. All exercises will be performed on the same machines used for strength evaluations during preliminary testing (Biostrength® line machines, Technogym S.p.A., Cesena, Italy), which will allow precise individualization of resistance exercise, recording of sessions and exact reproducibility on the next exercise trial for each patient.

Combined exercise will consist in 20 minutes of moderate-intensity exercise at 40% Wmax on a cycle ergometer followed by 20 minutes of resistance exercise consisting in a circuit training of 4 groups of exercises repeated in 3 sets of 12 repetitions each at 70% of 1-RM.

Experimental design Participants will be assigned the eight exercise trials (MIE-ON, HIIT-ON, RE-ON, COMB-ON, MIE-OFF, HIIT-OFF, RE-OFF, COMB-OFF) in random order. Randomization will be done by a computer-generated sequence. Each exercise session will be separated by at least 7 days.

Participants will be asked to abstain from strenuous exercise 48 hours before the exercise trial. The day before the exercise trials, participants will be asked to follow a standard diet and keep a food dairy.

On the morning of the trial, patients will present at the lab at 9.00 am, 2 hours after a standard breakfast, which was proceeded by an insulin bolus calculated by the participants bolus calculator. When allocated to one of the "ON" exercise trials, the patients will be asked to activate the exercise mode (temporary higher target) from 2 hours before the expected exercise trial until the end of the exercise trial. When allocated to one of the "OFF" exercise trials, the patients will start the exercise trial without announcing exercise to the pump. Heart rate, rate of perceived exertion (RPE), power output, angular velocity, number of repetitions and exercise load will all be recorded during the exercise sessions. CGM data, carbohydrate intake and insulin delivery data will be recorded for the following 24 hours.