A Randomized Crossover Pilot Study Evaluating Glucose Control During Exercise Initiated 1 or 2 h After a Meal in Adults with Type 1 Diabetes Treated with an Automated Insulin Delivery System

Abstract

Aims:

To assess the safety and efficacy of two exercise sessions performed 60- and 120-min postmeal with a combination of meal bolus reduction and increased glucose target to the automated insulin delivery (AID) system.

Methods:

A randomized crossover trial in 13 adult participants (6 females) living with type 1 diabetes using AID (A1c = 7.9% ± 0.6%, age = 53.5 ± 15.5 years, T1D duration = 29.0 ± 16.0 years) was conducted. Just before breakfast, at the time of meal bolus, the AID glucose target was increased from 6 to 9 mmol/L, and a meal bolus reduction of 33% was applied. Two 60-min exercise sessions (60% of VO₂ peak) were undertaken either 60 min (60EX) or 120 min (120EX) after a standardized breakfast, followed by a 90-min recovery period.

Results:

The mean reduction in plasma glucose (PG) levels from prebreakfast to postexercise (−0.8 ± 2.4 mmol/L vs. +0.3 ± 2.3 mmol/L, P = 0.082) were similar between 60EX and 120EX. From prebreakfast to postexercise, PG times in range (3.9–10.0 mmol/L; 63.4% ± 43.1% 60EX vs. 51.9% ± 29.7% 120EX, P = 0.219) and time above range (>10.0 mmol/L; 36.3% ± 43.3% 60EX vs. 48.1% ± 29.7% 120EX, P = 0.211) did not differ between interventions. The 60EX attenuated the glucose rise between premeal to pre-exercise (+1.8 ± 2.1 mmol/L 60EX vs. +3.9 ± 2.1 mmol/L 120EX, P = 0.001). No hypoglycemic events (<3.9 mmol/L) occurred during the study.

Conclusion:

Premeal announcement combining meal bolus reduction and increased glucose target was effective and safe during 60 min of moderate-intensity aerobic exercise, whether exercise onset was 60 or 120 min following a meal.

Clinical Trial Registration No.:

NCT04031599.

Introduction

Automated insulin delivery (AID) systems allowing dynamic insulin infusion from a subcutaneous insulin infusion pump have been shown to be safe and effective at enhancing glucose control in people living with type 1 diabetes (PWT1D).^1,2 Additionally, these new systems offer patients psychosocial benefits such as reduced anxiety, reassurance, and better sleep contributing to lowering diabetes self-management burdens.^2–4 Physical activity (PA) also provides a plethora of health benefits, including glycemic control but can be limited by fear and increased risk of hypoglycemia, even with AID systems.^5–7 Glucose regulation during exercise is contingent on several factors, including exercise timing and modality (type, intensity, duration, etc.) with prolonged aerobic exercise being associated with the highest hypoglycemic risk.⁸ A meta-analysis, including six randomized controlled trials recently concluded that closed-loop systems (CLS) significantly increase glucose time in range by 6.2% (TIR; 3.9–10.0 mM) during exercise compared with standard of care in PWT1D.⁹

Nevertheless, postprandial exercise remains challenging for adults living with type 1 diabetes, even with insulin pump therapy¹⁰ and AID.⁷ Indeed, in the postprandial state, rapid glucose changes (increase with carbohydrate absorption and decrease with muscle glucose utilization occurring with elevated plasma insulin from meal bolus) can reduce continuous glucose monitoring accuracy and increase lag time with interstitial glucose values. Altogether, these events can largely limit the ability of AID to mitigate the risk of hypoglycemia with basal rate modulation.

Previously, we reported that when a 60-min aerobic exercise bout performed 90 min after breakfast was not announced to the AID, participants with T1D still experienced hypoglycemia during and up to 1 h into the recovery period (% time <3.9 mM = 13% ± 19%).⁷ Announcing exercise 90 min before the onset significantly decreased the % time <3.9 mM to 7% ± 13%; (P = 0.005 vs. unannounced exercise) while adding a 33% meal bolus reduction to the exercise announcement lowered the time in hypoglycemia to 2% ± 6% (P < 0.0001 vs. unannounced exercise, P = 0.06 vs. exercise announcement alone). The combination of AID announcement and 33% meal bolus reduction was associated with a significantly smaller plasma glucose (PG) reduction during exercise than the other two strategies (−0.3 mmol/L vs. −2.6 and −2.4 mmol/L, both <0.0001).

Even though exercise timing influences hypoglycemic risk, most studies investigating postprandial glucose management have used 90 min as exercise onset time.^7,11–13 In real-world conditions, the timing of postprandial exercise onset can vary widely based on each individual's schedule and preferences, and it is crucial to gain a better understanding of glucose regulation to allow safe postprandial exercise with AID. This pilot study aimed at evaluating the safety and efficacy of AID on glucose control using premeal exercise announcement (increased glucose target [6 to 9 mmol/L]) combined with 33% premeal bolus reduction) for a 1-h moderate-intensity aerobic exercise initiated either 60 or 120 min after breakfast. Using this glucose management approach, we hypothesized that comparable glucose control will be achieved between the two exercise timings.

Materials and Methods

Participants

This optional exercise substudy was offered to participants of a larger project investigating simplified meal bolus strategies in the context of AID (registration no. NCT04031599, https://clinicaltrials.gov). Participants of this substudy were recruited at the Montreal Clinical Research Institute from March 2020 until the sample size for the main study was reached in July 2021. To maximize the number of participants in the substudy, an amendment to the main study protocol was made in August 2021, which allowed the McGill research team to offer the substudy to all of their study participants actively enrolled. Recruitment at the McGill site took place between August 2021 and October 2021. The substudy was offered to all participants enrolled in the main study but was not obligatory for inclusion. This study was conducted according to the guidelines for good clinical practice (ICH E6 Guideline for Good Clinical Practice, May 1, 1996) and the Declaration of Helsinki, and was approved by our Institutional Ethics Committee at the Montreal Clinical Research Institute.

During the admission visit, each participant provided written informed consent and was offered the option of participating in the exercise substudy in addition to the main trial with a specific additional informed consent. The substudy interventions were conducted at the beginning of the day and in a fasting state. There was no active insulin on board from a preceding meal. Therefore, the two meal strategies that were tested in the main study had no potential carrying-over effect on the substudy outcomes. The substudy used a classical carb counting strategy for all interventions. All participants completed a run-in period of 12 days to optimize insulin doses and treatment before starting the main study and AID system. The period of AID use before completing the first substudy exercise intervention varied with a minimum of 3 days (median 14 days, range 3 to 21 days). The average time between the two substudy interventions was 19 days (median 19 days, range 6 to 28 days).

Sixteen of the 23 participants recruited to participate in the main study accepted to undertake the exercise substudy. Thirteen out of 16 participants completed the interventions at the Montreal Clinical Research Institute and were included in the analysis. Reasons for dropout included ergometer seat discomfort, study load too demanding, and a technical problem with the Dexcom. The participant flowchart is presented in Figure 1. Inclusion criteria for the main study were male or female ≥18 years of age, clinical diagnosis of T1D for at least 1 year, on insulin pump therapy for at least 3 months, and HbA1c <10%; and for the substudy, ability and willingness to perform two 60-min exercise sessions at moderate intensity. Exclusion criteria included clinically significant nephropathy, neuropathy, or retinopathy, recent (<6 months) acute macrovascular event, severe hypoglycemic episode within 2 weeks of screening or during the run-in period, and anticipation of a significant change in the exercise regimen between admission and end of the trial (i.e., starting or stopping an organized sport).

FIG. 1.

Participants flow chart.

Experimental design and randomization

We conducted a two-arm unblinded randomized crossover trial with an insulin–AID to compare two postprandial exercise timings (60 min vs. 120 min postmeal: 60EX and 120EX) in PWT1D. Before breakfast, the AID glucose target was increased from 6 to 9 mmol/L (up to the end of the exercise), and a meal bolus reduction of 33% was applied. One hour of moderate-intensity (60% VO₂ peak) exercise was performed followed by a 90-min recovery period. A crossover design was chosen to minimize the risk of confounding factors with awaited sample size. Block-balanced randomization was performed to determine the order of the interventions (60EX or 120EX). The study manager, who had no access to study participants, conducted the randomization and concealed 20 sequences into a randomization table. Following the admission visit, the research coordinator assigned the next sequence in the randomization table to each newly enrolled participant. The participants were blinded to the sequence allocation until the end of the admission visit.

Baseline testing

During the admission visit, eligibility criteria were assessed, and each participant provided their written informed consent. Medical data, HbA1c, anthropometric measurements, and records of insulin therapy (total daily dose, carbohydrate to insulin ratios, basal rates of 7 representative days) were collected. Cardiorespiratory fitness was assessed using a maximal graded exercise test adapted from Storer et al, on a cycle ergometer (ER 900; Ergoline, Germany) with power output increased by 10–20 W/min.¹⁴ Expired gas samples were analyzed through a mixing chamber using a Moxus cardiorespiratory test station (AEI Technologies). VO₂ peak corresponded to the highest 30 s mean value reached during the test. The results were used to estimate the exercise intensity (60% VO₂ peak) for the two subsequent exercise interventions. Participants were asked not to perform exercise (excluding light exercise or taking the stairs) or to consume alcohol 24 h before the next two intervention visits.

Procedures

On the morning of the two interventions, participants were admitted to the laboratory at 7:30 am after a ≥ 10 h overnight fast. A cannula was inserted into the antecubital vein of the arm for blood sampling purposes. At 8:00 am, a standardized breakfast (65 g carbohydrates, 21 g protein, 25 g fat, 577 kcal) was provided along with an insulin bolus reduced by 33% as compared with what was proposed by the AID algorithm. Increased glucose target was set in the CLS at 8:00 am for a total duration of 2 h (i.e., 1 h before exercise + 1 h of exercise for 60EX) or 3 h (i.e., 2 h before exercise + 1 h of exercise for 120EX). At 9:00 (60EX) or 10:00 am (120EX), participants started exercising and performed a 1-h exercise bout on the ergocycle at an intensity (watts) equivalent to 60% of their VO₂ peak. The two intervention visits were separated by a median of 19 days (range from 3 to 28 days).

The AID algorithm was based on a model predictive control algorithm, as previously described.^7,15,16 A minor change was made to the bolus calculator to account for recent basal insulin suspensions in the calculations of the negative insulin correction. The algorithm was programmed on a smart phone application (Oregon Health and Science University [OHSU]; iPancreas platform, OR),¹⁷ which was connected through Bluetooth to the insulin infusion pump (Tandem Diabetes Care, CA) and the real-time continuous glucose monitor (Dexcom, CA). The artificial pancreas system (algorithm) was initialized using records of participants' previous 3 days of insulin therapy (total daily dose, carbohydrate/insulin ratios and basal rates) and body weight obtained at the screening visit. Participants' usual fast-acting insulin analog was used (aspart or lispro).

Blood collection and analysis

During both interventions, venous blood samples were collected every 20 min before the onset of exercise, every 10 min during exercise, and every 15 min (up to 90 min) during the recovery period for a total of 2.5 (60EX) and 3.5 h (120EX). Each blood sample (4 or 6 mL) was collected using EDTA tubes. Blood samples were immediately processed to measure PG levels, using a YSI 2300STAT Plus analyzer (Yellow Springs, OH), and were stored for subsequent measurement of total insulin levels in duplicates using an enzyme-linked immunoassay (Millipore).

Outcomes and statistical analysis

The primary outcomes for our study were the mean change between prebreakfast PG levels and (1) glucose levels postexercise (Δ prebreakfast to postexercise), and (2) nadir in PG during exercise (Δ prebreakfast to nadir during exercise). Secondary outcomes included (1) the number of patients experiencing hypoglycemia requiring treatment (from prebreakfast to the end of the recovery period), (2) the mean decrease in PG levels during exercise, (3) the difference between glucose levels pre-exercise and the glucose nadir during exercise, (4) the percentage of PG time below range (TBR; <3.9 mmol/L), TIR (3.9–10.0 mmol/L), PG time above range (TAR; >10.0 mmol/L) for the prebreakfast to postexercise period, and the 60-min exercise period, and (5) the noninferiority of 60EX as compared with 120EX for PG TIR between prebreakfast and postexercise.

Continuous variables are presented as mean ± standard deviation or as frequencies for the number of patients experiencing hypoglycemia requiring treatment. Normality of data distribution was assessed using Shapiro–Wilk tests and boxplots within each experimental condition. Paired Student's T-tests were performed to compare all the outcomes except for changes in glucose (pre-exercise to nadir, pre-exercise to postexercise), as well as glucose TIR and TAR during exercise that were compared using Wilcoxon signed-rank tests. The comparisons were made between the 60- and 120-min exercise announcements. Two-tailed P-values <0.05 were considered statistically significant. A one-sided two-sample t-test with a margin of 10% was used to assess the noninferiority of 60EX as compared with 120EX (one-tailed P-values <0.05 was considered statistically significant) for PG TIR between prebreakfast and postexercise. All analyses were performed using SAS software, version 9.4 (SAS Institute). All data were included in the analysis and three data imputations (last and next observations averaged) were applied for three missed data points (problems with the catheter).

Given the preliminary nature of the substudy, we did not estimate a sample size but aimed to have at least 12 participants complete both strategies to explore noninferiority for glucose TIR between prebreakfast and postexercise.

Results

The characteristics of the participants who completed the study are presented in Table 1. Thirteen participants were included in each statistical analysis.

Table 1.

Participants' Baseline Characteristics

Characteristics	Values
Participants (M/F)	13 (7/6)
Age (years)	53 ± 15
Diabetes duration (years)	29 ± 16
Pre-AID glycated hemoglobin (%)	7.9 ± 0.6
Body mass index (kg/m2)	28.1 ± 4.0
Total daily insulin dose programmed in the AID (U/day)	52.9 ± 25.4
VO2 peak [mL/(kg·min)]	21.9 ± 3.8

Data are presented as mean ± standard deviation except for participants (M/F), which is presented as frequency.

AID, automated insulin delivery system.

Prebreakfast (fasting) PG levels were comparable in both conditions (P = 0.184) (Fig. 2a and Table 2). The rise in PG from prebreakfast to pre-exercise was less pronounced for 60EX as compared with 120EX (−2.1 mmol/L, confidence interval [95% CI]: −3.08 to −1.00, P = 0.001). Thus, participants started exercise with lower PG for 60EX (−1.4 mmol/L, 95% CI: −2.70 to −0.18, P = 0.028). Despite different PG levels at exercise onset, changes in PG from pre-exercise to postexercise, as well as to nadir during exercise were similar between conditions (both P ≥ 0.154). However, decrease in PG from pre-exercise to postrecovery was lower for 60EX as compared with 120EX (−1.6 mmol/L, 95% CI: −0.04 to 2.90, P = 0.045). For glucose TIR from prebreakfast to postexercise, 60EX was noninferior to 120EX within the preplanned noninferiority margin of 10% (P = 0.016). TBR, TIR, and TAR remained comparable between conditions from prebreakfast to postexercise (all P ≥ 0.211) and from exercise onset to postexercise (P ≥ 0.262). No hypoglycemic events occurred during 60EX interventions, nor during 120EX interventions (Table 2). Patients' insulin to carbohydrate ratios did not influence the reported results (data not shown).

FIG. 2.

Plasma (a) glucose and (b) insulin levels over the course of the two strategies. A standardized breakfast was consumed at −120 min (120EX) and −60 min (60EX). Meal bolus timing with exercise announcement is indicated by arrows. Areas between dotted lines correspond to the 60-min exercise sessions at 60% VO₂ peak on a cycle ergometer. Areas between solid lines correspond to the plasma glucose TIR (TIR; 3.9 to 10 mmol/L). Data are expressed as mean (standard deviation). 60EX, 60-min postbreakfast exercise condition; 120EX, 120-min postbreakfast exercise condition; TIR, time in range.

Table 2.

Summary and Comparisons of Glycemic and Insulin Outcomes for the Two Strategies

		60EX (n = 13)	120EX (n = 13)	P	Effect size	95% confidence interval
Primary outcomes
Δ Prebreakfast to postexercise	Glucose	−0.8 ± 2.4	0.3 ± 2.3	0.082	−0.526	−2.64 to 0.18
Δ Prebreakfast to nadir	Glucose	−0.9 ± 2.5	0.3 ± 2.2	0.067	−0.559	−2.64 to 0.10
Secondary outcomes
AID breakfast bolus (U)	Insulin	5.8 ± 1.7	5.8 ± 1.8	0.680	−0.117	−0.19 to 0.13
AID hourly basal rate breakfast to postexercise (U/h)	Insulin	0.9 ± 0.6	1.0 ± 0.6	0.093	−0.507	−0.35 to 0.03
Prebreakfast (fasting)	Glucose	8.3 ± 1.9	7.6 ± 1.9	0.184	0.391	−0.33 to 1.53
	Insulin	154 ± 116	155 ± 117	0.271	−0.351	−4.44 to 1.39
Exercise onset	Glucose	10.1 ± 3.1	11.5 ± 2.5	0.028 a	−0.693	−2.70 to −0.18
	Insulin	247 ± 116	205 ± 121	0.125	0.448	−2.86 to 16.53
Postexercise	Glucose	7.3 ± 2.6	8.0 ± 1.7	0.397	−0.236	−2.18 to 0.93
	Insulin	238 ± 105	191 ± 101	0.040 a	0.671	0.30 to 10.90
Δ Prebreakfast to exercise onset	Glucose	1.8 ± 2.1	3.9 ± 2.1	0.001 a	−0.937	−3.08 to −1.00
Δ Exercise onset to postexercise	Glucose	−2.5 ± 1.5	−3.6 ± 2.3	0.155	0.421	−0.40 to 2.20
Δ Exercise onset to nadir	Glucose	−2.6 ± 1.4	−3.7 ± 2.2	0.154	0.422	−0.36 to 2.05
Δ Exercise onset to postrecovery	Glucose	−1.4 ± 2.4	−3.0 ± 3.4	0.045 a	0.621	−0.04 to 2.90
Time spent at plasma glucose % (prebreakfast to postexercise)
	TBR <3.9 mmol/L	0.2 ± 0.7	0.0 ± 0.0	0.337	0.277	−0.00 to 0.01
	TIR 3.9–10.0 mmol/L	63.4 ± 43.1	51.9 ± 29.7	0.219	0.360	−0.08 to 0.31
	TAR >10.0 mmol/L	36.3 ± 43.3	48.1 ± 29.7	0.211	−0.367	−0.31 to 0.08
Time spent at plasma glucose % (exercise onset to postexercise)
	TBR <3.9 mmol/L	0.4 ± 1.4	0.0 ± 0.0	0.336	0.276	−0.00 to 0.01
	TIR 3.9–10.0 mmol/L	60.5 ± 46.9	47.2 ± 30.7	0.276	0.317	−0.12 to 0.40
	TAR >10.0 mmol/L	39.1 ± 47.2	52.8 ± 30.7	0.262	−0.326	−0.40 to 0.12
Number of hypoglycemia events necessitating correction (n)
	During exercise	0 ± 0.0	0 ± 0.0	—	—	—
	During recovery	0 ± 0.0	0 ± 0.0	—	—	—

The P value is bolded whenever it is < 0.05, which is deemed as statistical significant.

^a Statistically different between 60EX and 120EX.

Glucose is presented in mmol/L. Plasma total insulin is presented in pmol/L.

60EX, 60-min postbreakfast exercise condition; 120EX, 120-min postbreakfast exercise condition; Δ, average change; TAR, time above range; TBR, time below range; TIR, time in range.

Plasma total insulin concentrations were comparable prebreakfast, as well as at exercise onset in both conditions (both P ≥ 0.125) (Fig. 2b). However, plasma insulin postexercise was significantly higher for 60EX (+47 pmol/L, 95% CI: 0.30–10.90, P = 0.040). No harms or untended effects were observed by study personnel or reported by participants during the study.

Discussion

The current study aimed to evaluate the safety and efficacy of a strategy combining a 33% meal bolus reduction with increased glucose target (6.0 to 9.0 mmol/L) for postprandial aerobic exercise performed 60 and 120 min after a standardized meal. Our results suggest comparable glucose control without hypoglycemia with both exercise timings.

Prolonged aerobic exercise tends to decrease glucose levels and thus increases hypoglycemic risks in PWT1D.^8,18 Furthermore, exercising in the postprandial state combines several challenges for AID: (1) high plasma insulin due to insulin on-board related to meal boluses, and (2) rapid PG changes (due to postprandial PG excursion and exercise) making input from continuous blood glucose monitors less accurate. Previously, our group tested the efficacy of announced and unannounced exercise to CLS during the postprandial state in PWT1D during a 1 h moderate-intensity exercise session performed 90 min after the meal.⁷ Hypoglycemia occurred particularly during the unannounced exercise strategy. However, we observed that combining a 33% meal bolus with increased glucose target was superior at reducing hypoglycemia than a 33% meal bolus reduction alone or no announcement. These results highlighted the independent and combined contribution of announcing exercise and bolus reduction to improve glucose control during postprandial exercise.

However, at different postprandial timings, carbohydrate absorption and plasma insulin on board will have variable and complex interplay,¹⁹ suggesting that data obtained with exercise onset 90 min after a meal might lead to different results with earlier or later postprandial exercise timings. Indeed, in real-world conditions, the timing of postprandial exercise onset can vary widely based on each individual's schedule and preferences. Thus, in the current study, we investigated two postprandial exercise timings with AID that have never been studied before.

International consensus suggests that AID announcements should be made at least 60 min before the start of prolonged aerobic exercise.²⁰ However, evidence evaluating the optimal timing and magnitude for meal bolus reduction with AID remains scarce and current recommendations are based on expert opinions. Our current findings suggest that exercise announcement combined with a 33% bolus reduction 60 or 120 min before exercise may provide similar glucose benefits in adult PWT1D, which supports the consensus statement. Yet, we should be cautious in drawing general conclusions as we observed numerically larger decreases in glucose levels in the 120EX arm compared with 60EX from the exercise onset to postrecovery period (−3.0 ± 3.4, 120EX vs. −1.4 ± 2.4, 60EX, P = 0.045, d = 0.621). This could be due to the higher PG at exercise onset observed for 120EX, which is often associated with more pronounced glucose changes during exercise.²¹ In our study, we noted a delay in the awaited decrease of plasma insulin during the first 30 min of exercise in both conditions.

As previously reported, increased blood flow and vasodilation during aerobic exercise can heighten insulin absorption from the subcutaneous tissue to the blood in PWT1D for both fast and intermediate-acting insulin.^22–24

Notably, the two strategies implemented in our study both led to the absence of time spent <3.9 mmol/L. While the lack of a control group (e.g., no exercise announcement) prevents us from concluding that these two treatments eliminate hypoglycemia, our results suggest that this approach is safe and effective for patients exercising 1 and 2 h following a meal. As compared with our previous study where exercise was announced 90 min before exercise onset,⁷ our participants had a lower VO₂ peak [21.9 mL/(kg·min) vs. 32.0 mL/(kg·min)], which could be attributable to their older age (53 years vs. 40 years), higher body mass index (28.1 vs. 25.3), longer diabetes duration (29 years vs. 23 years), and worse pre-AID therapy glucose control (A1c 7.9% vs. 7.3%).^22,23 This difference in cardiorespiratory fitness must have translated into a lower absolute exercise intensity at 60% VO₂ peak, which could potentially explain why no hypoglycemic events were observed. The proposed strategy should thus be explored in more fit PWT1D.

Furthermore, in the present study, patients undertook the first exercise session after a median of 14 days with AID, whereas in Tagougui et al, participants used AID starting only a few hours before exercise.⁷ Thus, our results probably better reflect real-world conditions.

In real-life conditions, Franc et al reported low and similar daily TBR (<3.9 mM) (2.0% vs. 2.2%) for patients using CLS between days with and without unsupervised PA, respectively.²⁴ However, given the nature of this study, PA announcement (range 1 to 596 min before PA) and pre-PA meal timings varied greatly between PA days. Moreover, following PA announcement, the CLS recommended preventive CHO intake (41 g for days with PA vs. 22 g days without PA), which along with insulin adjustments, probably contributed to the prevention of PA-related hypoglycemic events. In a more controlled setting, Viñals et al used a multivariable CLS (insulin delivery and automatic CHO recommendation during and after exercise) to compare glycemic control during open-loop therapy to announced and unannounced exercises 4 h after a meal.²⁵ Using this approach in 10 PWT1D, the number of hypoglycemic events (i.e., PG <3.9 mM for ≥15 min) during and up to ∼2 h after exercise did not differ statistically between all three conditions (four unannounced, three announced, and eight open-loop, all P ≥ 0.218).

Exercise announcement to the CLS 20 min before exercise onset did not modify the blood glucose target by the algorithm. Interestingly in our study, we observed that announcing the exercise mode to the CLS did not significantly affect basal insulin reduction. Rather, the 33% reduction of the preceding meal bolus seemed to have had the dominant impact on reducing the risk of hypoglycemia. Based on these results, it is possible that a combination of feed-forward actions and meal bolus reduction may be effective at reducing the risk of exercise-related hypoglycemia in the postprandial state. However, while some studies with CLS combined with pre-PA CHO supplementation observed no or reduced PA-related hypoglycemic events in PWT1D,^26–28 chronic CHO compensation around exercise may be counterproductive to cardiovascular health and weight maintenance in PWT1D. In a previous report by our group, women living with type 1 diabetes reported insulin reduction as their preferred strategy to reduce hypoglycemic risk over CHO supplementation.²⁹ Without the use of CHO supplementation and exercise announcement 2 h before exercise, Morrison et al reported a mean TBR of 0.0% during and up to 2 h after moderate- and high-intensity exercise.³⁰ Contrary to our study, exercise sessions were shorter (i.e., 40 min at moderate intensity) and initiated at least 4 h following the participants' usual bolus and a standardized meal when most active insulin had already been cleared. Similar to our study, the lack of a control group in their study prevents broader interpretation.

Exercise announcement to AID aims at reducing hypoglycemia, yet it may also lead to hyperglycemia before and during exercise. In our study, 120EX had a significantly higher preexercise glucose level than 60EX (11.5 mmol/L vs. 10.1 mmol/L, P = 0.028) and numerically higher TAR (48.1% vs. 36.3%, P = 0.211). However, the pilot nature of this study precludes us from drawing definitive conclusions regarding glucose TAR and specific exercise timings. In the postprandial period with a preset exercise mode aiming for glucose target of 9.0 mmol/L, glucose TIR (3.9 mmol/L to 10.0 mmol/L) might not be the best measure to represent optimal glucose control. Instead, preventing exercise-induced hypoglycemia should remain the main priority for most patients.^5,6,31 To avoid hypoglycemia, several experts suggested that exercise should not be undertaken early after a meal at peak insulin action.^32,33 However, this is not always possible for PWT1D with time constraints or those performing active transportation immediately after a meal.

As in type 2 diabetes,³⁴ our data suggest that early exercise (i.e., <60 min postmeal) could be beneficial by limiting postprandial hyperglycemia even in the context of a reduced meal bolus. Overall, optimal timing of exercise announcement to AID will be dependent on several factors, including hyperglycemic risks, PWT1D time constraints, the type of exercise,¹² the ingestion of a snack during exercise,³⁵ and the possible incorporation of glucagon in AID,³⁶ which holds the potential to impact glucose outcomes. In this investigation, a typical mixed breakfast (577 kcal) moderate in CHO, protein, and fat (65 g = 45%, 21 g = 15%, and 25 g = 39%, respectively) was provided. The addition of fat (35 g) or protein (35 g) to a carbohydrate meal was previously shown to blunt glucose peak and prolong postprandial hyperglycemia in participants using CLS.^37,38 Thus, in the context of the current study, it is unlikely that meals higher in fat or protein would significantly influence the risk of hypoglycemia when exercise is performed 1 or 2 h after a meal.

In real-life conditions, PWT1D previously reported consuming breakfast with a mean energy intake and composition of 531 kcal (CHO 58 g = 44%, Protein 23 g = 17%, and Fat 23 g = 39% of total kcal). Thus, the breakfast provided in the current study represents usual intake for these patients.

Our study has several strengths. First, the crossover design and head-to-head comparison importantly control potential confounding factors. Second, we use PG levels rather than CGM glucose to eliminate time delay and potential inaccuracy of sensors, especially under postmeal and exercise conditions. Third, it is worth noting that the fasting insulin level, a parameter that may impact glucose outcomes, was comparable in both groups. Several limitations also need to be addressed. First, this is a pilot study with a small sample size, which limits the power to detect the differences in all outcomes. Second, this study does not include a control group (e.g., without increased glucose target and/or full meal bolus group), preventing us from further interpreting our results. Since only one exercise modality (specific type, duration, and intensity) was assessed, it is possible that our findings do not apply to other exercise modalities. Finally, one might be cautious about applying these findings when exercise is performed at a different time of the day. Compared with afternoon exercise, morning exercise can be associated with higher glycemia³⁹ and a lower risk of postexercise hypoglycemic events in PWT1D.⁴⁰ Thus, it might be necessary for patients to adjust their approach (e.g., greater bolus reduction, additional pre-exercise carbohydrates) when exercising later in the day. Future large-scale studies are needed to further confirm our findings.

Conclusion

In conclusion, combining premeal exercise announcement (i.e., increased glucose target from 6 to 9 mmol/L) and a 33% meal bolus reduction may be an effective and safe strategy for glycemic control during postprandial exercise in PWT1D treated with AID exercising 60 or 120 min after a meal.

Authors' Contributions

R.R.-L. conceived and designed the study and acquired financial support. M.R., E.M.-C., J.M., and M.D. contributed to the data collection. R.R.-L., E.M.-C., J.M., Z.W., and S.T. contributed to the acquisition, analysis, and interpretation of data. E.M.-C., J.M., and Z.W. wrote the article (original draft). All authors reviewed the article and approved the final version of the article.

Author Disclosure Statement

R.R.-L. has received research grants from AstraZeneca, Eli Lilly, Merck, Novo Nordisk, and Sanofi-Aventis and has been a consultant or member on advisory panels for Abbott, Amgen, AstraZeneca, Boehringer, Carlina Technologies, Eli Lilly, Janssen, Medtronic, Merck, Neomed, Novo Nordisk, Roche, Sanofi-Aventis, and Takeda. R.R.-L. has also received honoraria for conferences by Abbott, AstraZeneca, Eli Lilly, Janssen, Medtronic, Merck, Novo Nordisk, and Sanofi-Aventis, and in-kind contributions related to closed-loop technology from Animas, Medtronic, and Roche. R.R.-L. benefits from unrestricted grants for clinical and educational activities from Eli Lilly, LifeScan, Medtronic, Merck, Novo Nordisk, and Sanofi and holds intellectual property in the field of type 2 diabetes risk biomarkers, catheter life, and the closed-loop system. R.R.-L. has received purchase fees from Eli Lilly in relation to closed-loop technology. Z.W. would like to thank Fonds de Recherche du Québec—Santé (FRQS) for the postdoctoral fellowship.

Funding Information

This study was supported by funding from the National Institute of Health (grant no. 1DP3DK106930-01). This study was carried out as part of the project of the International Joint Laboratory REGALE-1—Glycemic regulation during exercise in type 1 diabetes—in collaboration with URePSSS (Lille University) and IRCM (Montreal) and granted by I-SITE (Université Lille Nord Europe). S.T. was supported by grants from the Société Francophone du Diabète, Fondation pour la Recherche Médicale, and the Cardiometabolic Health, Diabetes, and Obesity Research Network. E.M.-C. was supported by the Fonds de Recherche en Santé du Québec (FRQS). R.R.-L. holds the J-A DeSève and Lamarre Gosselin diabetes research chairs.