Use of the Omnipod 5 Automated Insulin Delivery System Activity Feature Reduces Insulin Delivery and Attenuates the Drop in Glycemia Associated With Exercise in a Randomized Controlled Trial

Abstract

OBJECTIVE

To compare the efficacy of enabling Activity feature 60 (AF-60) or 30 min (AF-30) before prolonged exercise versus the automated mode (Auto) in adults and adolescents with type 1 diabetes wearing the Omnipod 5 System.

RESEARCH DESIGN AND METHODS

In this three-way crossover study, 38 participants (age 30 ± 15 years; BMI 24.7 ± 4.1 kg/m²; HbA_1c 7.5% ± 0.9% [58 ± 11 mmol/mol]) from the extension phase of the pivotal trial of the Omnipod 5 System completed a 70-min treadmill session at 64–76% maximum heart rate in a postabsorptive state under each of the three conditions. Auto was resumed after exercise, and glycemia and insulin delivery metrics were examined in the 4-h postexercise period.

RESULTS

The percentage of participants who developed hypoglycemia during exercise did not differ significantly between Auto (42%) and AF-60 (29%; P = 0.34) or AF-30 (24%; P = 0.14). However, AF-60 and AF-30 reduced insulin delivery compared with Auto in the hour before (P < 0.001) and during exercise (P < 0.001). There was also a favorable attenuation in glucose drop during exercise when comparing Auto (−57 ± −35 mg/dL) with AF-60 (−44 ± −33 mg/dL; P = 0.02) and AF-30 (−36 ± −34 mg/dL; P = 0.01). In the postexercise period, glycemia and insulin delivery were comparable.

CONCLUSIONS

Enabling the Activity feature either 60 or 30 min before exercise reduced insulin delivery and attenuated glucose drops relative to Auto, but hypoglycemia incidence was not different across the three conditions. These findings support the use of the Omnipod 5 System for exercise but highlight the importance of using additional strategies, such as earlier use of Activity feature and/or carbohydrate intake to further reduce hypoglycemia risk.

Graphical Abstract

Introduction

Physical activity is a cornerstone of a healthy lifestyle and remains integral to the lives of individuals with type 1 diabetes. Despite the challenges of managing exercise-induced hypoglycemia, many individuals with type 1 diabetes engage in prolonged physical activity, lasting ≥45 min, such as hiking or walking, running, or cycling (1,2). Even those who are moderately active often participate in at least 30 min of endurance exercise weekly, although concerns about hypoglycemia often accompany these activities (3,4).

Automated insulin delivery (AID) systems improve the percentage of time glucose levels remain within a target range (time in range [TIR]; 70–180 mg/dL) and reduce hemoglobin A_1c (HbA_1c) levels compared with traditional methods such as multiple daily injections or open-loop (standard) pump therapy (5). Moreover, AID systems can improve glycemic management around exercise events by incorporating features such as a higher temporary target or exercise mode to help minimize hypoglycemia risk (6,7). Despite these advancements in technology, exercise continues to introduce unpredictable glucose variability, increasing the risk of hypoglycemia or hyperglycemia. The Type 1 Diabetes and Exercise Initiative (T1DEXI) study (8), which analyzed data from >500 participants with type 1 diabetes, revealed that exercise days improved TIR compared with sedentary days, particularly among AID users (80% vs. 72%). However, days with exercise sessions lasting ≥30 min were also associated with an increased percentage of time with a glucose level <70 mg/dL compared with sedentary days. Moreover, a higher percentage of days with a level 2 hypoglycemic event, defined as a glucose level <54 mg/dL for at least 15 consecutive min, occurred on active days (13% of days) compared with sedentary days (10% of days) (8). The risk of developing hypoglycemia during real-world exercise in adults and adolescents with type 1 diabetes varies markedly based on several factors, including baseline glucose level, glucose rate of change just before activity, insulin on board at exercise start time, activity type, and percentage of time with a glucose level <70 mg/dL in the last 24 h (9).

To mitigate exercise-induced hypoglycemia, consensus guidelines recommend reducing basal insulin delivery, setting a higher glucose target for those using AID systems, and/or consuming extra carbohydrates before and/or during exercise (7,10–12). Although these strategies are effective in highly controlled clinical studies, real-world data suggest they are not routinely implemented (13,14). Using data from the T1DEXI study, Prévost et al. (15) analyzed nearly 4,000 exercise sessions and found that a higher glucose target was set 60 min before exercise in only 20% of the sessions. The suboptimal engagement with this strategy likely reflects the significant preplanning required, which is a known barrier to physical activity participation among individuals with type 1 diabetes (13,16). Surveys from 102 adults with type 1 diabetes who exercised regularly showed that only ∼12% made adjustments ≥90 min before activity (14). These findings further underscore the importance of evaluating insulin pump basal rate reduction strategies that can be implemented closer to the start of exercise, because they may be more practical and feasible in real-world settings. However, only a few studies have assessed the efficacy of setting lower basal rates within 1 h of exercise (17,18). In one of these studies, an 80% basal rate reduction initiated at 40 and 20 min before a 45-min cycle ergometer exercise session did not significantly affect TIR percentage or the percentage of time with a glucose level <72 mg/dL compared with a reduction in basal insulin delivery at exercise onset (17). However, initiating the basal rate reduction closer to the exercise session increased the likelihood of participants requiring carbohydrates to prevent exercise-induced hypoglycemia (17). Although emerging research has begun to explore the use of exercise-specific targets for AID systems (sometimes called Activity feature or Exercise mode), the optimal timing of such interventions remains unclear (19,20). Nonetheless, based on recent consensus guidelines for the use of AID around prolonged physical activity and/or exercise events, setting a higher temporary target is thought to be most effective if set 1–2 h before activity start time, and carbohydrates should be consumed as soon as glucose drops below 126 mg/dL (7).

The Omnipod 5 System (Insulet Corporation, Acton, MA) includes a mobile app and a tubeless insulin pump (pod) that includes Activity feature designed to reduce insulin delivery and temporarily raise the glucose target to 150 mg/dL for a user-defined duration, varying from 1 to 24 h. When using interoperable continuous glucose monitoring (CGM), the pump can work in automated mode (Auto) whereby a model predictive control algorithm delivers microboluses of insulin up to every 5 min to bring glucose levels toward user-defined glucose targets ranging from 110 to 150 mg/dL in 10-mg/dL increments (21,22). In real-world settings, the system demonstrated improvements in glycemic outcomes over standard pump therapy, with TIR increasing by 10.9% in preschool-aged children (2–5.9 years) (23) and by 15.6% and 9.3% in children (aged 6–13.9 years) and adolescents/adults (aged 14–70 years), respectively (22). Moreover, these glycemic benefits were sustained for up to 2 years, at least in preschool-aged children (24). Additionally, a recent retrospective analysis of 69,902 Omnipod 5 System users in the U.S. (≥2 years of age) who had at least 90 days of data showed <1.13% of time spent at a glucose level <70 mg/dL (25), which is well below clinical consensus guidelines (26). Although the abovementioned evidence supports the system’s overall efficacy in maintaining glycemia in the target range, it did not specifically examine the use of Activity feature. This feature aims to accommodate the reduced insulin requirements typically associated with physical activity.

The purpose of this study was to evaluate the efficacy of the Omnipod 5 System in reducing the risk of hypoglycemia during a 70-min endurance exercise session in adolescents and adults with type 1 diabetes. Insulin delivery and glycemic metrics were examined in the hour before exercise, during the exercise session, and 4 h after exercise. Outcomes were compared among three conditions: when Activity feature was enabled 60 (AF-60) or 30 min before exercise (AF-30) or when the system remained in Auto.

Research Design and Methods

Study Participants

This randomized controlled trial included a subset of 50 participants who were in a 12-month extension phase of a larger pivotal trial (ClinicalTrials.gov identifier NCT04196140). Written informed consent was obtained from all participants (or their guardians, where applicable). At the time of enrollment, eligible participants were between ages 12 and 70 years, had a diagnosis of type 1 diabetes for ≥6 months, had an HbA_1c <10% (86 mmol/mol), and demonstrated willingness and ability to complete exercise sessions. Exclusion criteria included a history of severe hypoglycemia (defined as a glucose event <54 mg/dL requiring third-party assistance), diabetic ketoacidosis, or an abnormal electrocardiogram within the past 6 months. At the time of the exercise sessions, all participants had been using the Omnipod 5 System in conjunction with a Dexcom G6 CGM (Dexcom, Inc., San Diego, CA) as part of the pivotal trial for ≥8 months.

Exercise Visits

Participants completed three treadmill exercise sessions separated by a minimum of 2 days in a randomized 1:1:1 crossover design (unblinded). The three sessions were as follows: 1) Activity feature enabled 60 min before and during exercise (AF-60), 2) Activity feature enabled 30 min before and during exercise (AF-30), and 3) standard Automated mode with no Activity feature enabled (Auto).

Before the first exercise session, each participant completed an abbreviatedtreadmill walking test to determine the treadmill speed and incline required to achieve a target heart rate range of 64–76% of their calculated maximum heart rate (220 minus age in years). Once the target heart rate range was attained, treadmill speeds and inclines were recorded and used in all exercise sessions.

In the 24 h before the exercise session, participants were asked to avoid participating in vigorous-intensity exercise, change their insulin delivery pod or CGM if either was set to expire on the visit day, and maintain a set target glucose on their AID system (the default and recommended set point was 120 mg/dL). On the day of each exercise session, participants were instructed not to alter their diabetes management and to eat lunch at least 3 h before their visit.

On arrival at the study site, staff confirmed that the participant’s last user-initiated bolus was at least 3 h before exercise and that the designated insulin delivery setting (AF-60, AF-30, or Auto) was activated as specified for the session.

Capillary glucose had to be between 90 and 200 mg/dL (inclusive) 5 min before the scheduled exercise start time. Capillary glucose was initially measured by the participant 15–20 min before exercise. At this time, if capillary glucose was <90 mg/dL, exercise was delayed by up to 30 min, and capillary glucose was rechecked every 15 min. If glucose was <90 mg/dL at these preexercise checkpoints, participants could consume up to 16 g (minimum 4 g) of fast-acting carbohydrates. If capillary glucose was between 200 and 300 mg/dL, a correction bolus was considered, and exercise could be delayed by up to 30 min. If capillary glucose was ≥300 mg/dL, blood ketones were checked with a ketone meter. If ketones were >1 mmol/L, the infusion site (pod) was changed, a correction bolus dose was administered, and the session was rescheduled. If ketones were ≤1 mmol/L, a manual correction bolus was considered. However, any individual who had a manual correction bolus within 3 h of the exercise start time was not included in the per-protocol analyses (Supplementary Table 1).

Exercise began at least 3 h after lunch or last meal (∼3:00 p.m.) and included 60 min of treadmill walking at the predefined heart rate range, with two 5-min breaks at 20 and 45 min. During each break and at the end of exercise, capillary glucose was measured (total time 70 min). If hypoglycemia (capillary glucose <70 mg/dL) occurred, participants stopped exercise, consumed 16 g carbohydrates, and rested for 15 min. Glucose was rechecked, and additional carbohydrates were provided as needed until glucose was ≥90 mg/dL, at which point exercise resumed. Participants with hyperglycemia (≥300 mg/dL) could continue unless ketones exceeded 1 mmol/L, in which case a correction bolus was administered, and the session was rescheduled.

At the end of exercise, when applicable, Activity feature was stopped, and Auto was resumed, unless deemed otherwise by an investigator for safety reasons (i.e., postexercise hypoglycemia). Participants were discharged from the trial facility after two sensor readings, measured at least 15 min apart, confirmed that glucose was between 70 and 300 mg/dL.

Outcomes

The primary outcome was the proportion of participants with a hypoglycemic event, defined as a capillary glucose reading <70 mg/dL during exercise. Time to the first hypoglycemic event was also evaluated for those who experienced an event.

Additional outcomes included carbohydrate consumption in the hour before and during exercise. CGM-derived outcomes assessed before exercise included percentage of time with a glucose level <70 mg/dL in the 24 h before exercise, change in glucose before exercise (change in CGM value from 5 min before to exercise onset), and sensor glucose at exercise onset. CGM outcomes assessed during exercise included mean glucose, nadir glucose, delta glucose (change in glucose at exercise onset to nadir), and glucose coefficient of variation. TIR percentage; time with a glucose level <70, <54, and >180 mg/dL; and number of participants with a CGM reading <54 and <70 mg/dL were assessed during and in the 4 h after exercise. Finally, insulin delivery in the hour before exercise, during exercise, and in recovery (for 4 h) was examined.

Statistical Analysis

A sample size of 45 participants was required to provide 80% power to detect a difference in the proportion of participants with hypoglycemia in the Auto versus AF-60 group, assuming the proportions of participants with hypoglycemia would be 60% and 30%, respectively. To allow for 10% potential attrition, 50 participants were included in the exercise sessions. A per-protocol analysis was conducted, excluding participants who did not follow exercise protocols (Supplementary Table 1 lists protocol deviations). An intention-to-treat analysis is also reported in Supplementary Table 2. Statistical analyses were performed using SPSS version 29.0 (IBM Corporation, Armonk, NY) and GraphPad Prism version 10.11 (Boston, MA). P values, 95% CIs, and differences between the three sessions are reported. Statistical significance was defined as P < 0.05.

The Fisher exact test was used to assess for differences in categorical outcomes, including the proportion of participants experiencing hypoglycemia during exercise and the proportion of participants consuming carbohydrates before and during exercise across all three sessions. This test was also used to evaluate differences in the number of participants with CGM readings <54 and <70 mg/dL among the three sessions during and after exercise.

Kaplan-Meier survival curves were generated to visualize the time to first hypoglycemic event. A repeated-measures linear mixed-effects model adjusting for glucose at exercise onset and carbohydrate intake before and during exercise was used to determine differences in time to hypoglycemia and CGM metrics during exercise. A two-way repeated-measures ANOVA was used to determine differences in insulin delivery in the hour before and during exercise for each exercise session. For all other outcomes, a repeated-measures one-way ANOVA was used to determine differences in outcomes among the three exercise sessions. If residuals were skewed, a robust regression model was used. Pairwise comparisons were conducted, and multiple comparisons were corrected using the two-stage false-discovery rate procedure.

Results

A total of 38 participants (18 female) were included in the per-protocol analyses of this study. Participants were (mean ± SD) aged 30 ± 15 years (minimum–maximum 12–65 years) with a BMI of 24.7 ± 4.1 kg/m² (BMI z score for participants aged 12–20 years 0.45 ± 0.81 kg/m²), diabetes duration of 16 ± 11 years, and an HbA_1c of 7.5% ± 1% (58 ± 11 mmol/mol).

Medians and interquartile ranges (IQRs) for CGM readings from the hour before exercise to 4 h postexercise for each session are presented in Fig. 1. During exercise, the glucose coefficient of variation was lower by 3.64% with AF-60 (P = 0.03) and by 2.40% with AF-30 (P = 0.20) compared with Auto (Table 1). In the hour preceding exercise, the number of participants requiring carbohydrate supplementation was similar across all conditions (AF-60 n = 6 [16%] of 38; AF-30 n = 7 [18%] of 38; Auto n = 4 [11%] of 38; P = 0.72). Moreover, the average carbohydrate intake required to achieve a glucose level >90 mg/dL before exercise did not differ among study conditions. Across all conditions combined, when carbohydrates were consumed before exercise, the average intake was 21 ± 8 g (median [IQR] 16 [12–32] g).

Figure 1.

CGM measurements from 1 h before to 4 h after three exercise sessions, including AF-60 (purple), AF-30 (teal), and Auto (yellow). Solid lines are the medians, and shaded regions denote the IQRs. Target glucose range (70–180 mg/dL) is indicated in black dashed horizontal lines. Exercise period is denoted between dotted vertical lines from t 0 to 70 min (although exercise could have been prolonged if a hypoglycemic event <70 mg/dL occurred). Shaded vertical regions denote 5-min exercise breaks.

Table 1.

Time to hypoglycemia, glycemic metrics, and insulin delivery by session

Metric	AF-60	AF-30	Auto	Mean difference ± SE [P]
				Auto vs. AF-60	Auto vs. AF-30	AF-30 vs. AF-60
Time to hypoglycemia, min*	39 ± 4	37 ± 4	36 ± 3	−2.72 ± 4.93 [0.59]	−1.63 ± 4.87 [0.74]	−1.09 ± 3.23 [0.75]
Before exercise
Time <70 mg/dL 24 h before, %†	1.5 ± 1.9	1.7 ± 2.8	2.8 ± 4.7	1.22 ± 0.72 [0.10]	1.09 ± 0.57 [0.07]	0.13 ± 0.50 [0.80]
Change in glucose, mg/dL/min†	0.08 ± 1.10	−0.07 ± 1.34	0.26 ± 1.58	0.18 ± 0.30 [0.55]	0.33 ± 0.29 [0.26]	−0.15 ± 0.19 [0.43]
Glucose at exercise onset, mg/dL†	145 ± 29	133 ± 30	143 ± 32	−2.18 ± 6.69 [0.75)	8.95 ± 7.82 [0.26]	−11.14 ± 5.80 [0.06]
During exercise
Mean glucose, mg/dL*	125 ± 3	124 ± 4	121 ± 3	−4.22 ± 3.10 [0.18]	−3.69 ± 4.50 [0.42]	−0.52 ± 3.69 [0.89]
Δ glucose, mg/dL†‡	−44 ± 33	−36 ± 34	−57 ± 35	−13.16 ± 5.58 [0.02]	−20.16 ± 7.48 [0.01]	6.57 ± 6.02 [0.28]
Nadir glucose, mg/dL*	98 ± 4	96 ± 4	89 ± 4	−9.50 ± 3.83 [0.02]	−7.44 ± 4.71 [0.12]	2.07 ± 4.50 [0.65]
CV, %*	16.1 ± 1.2	17.4 ± 1.3	19.8 ± 1.5	3.64 ± 1.57 [0.03]	2.40 ± 1.84 [0.20]	1.24 ± 1.69 [0.47]
Time <54 mg/dL, %*	0.8 ± 0.3	0.7 ± 0.3	1.5 ± 0.8	0.76 ± 0.99 [0.45]	0.81 ± 0.72 [0.27]	−0.06 ± 0.49 [0.91]
Time <70 mg/dL, %*	4.5 ± 0.9	4.3 ± 1.2	6.1 ± 2.0	1.55 ± 1.81 [0.40]	1.79 ± 2.05 [0.39]	−0.23 ± 1.33 [0.86]
TIR 70–180 mg/dL, %*	92.3 ± 2.3	88.1 ± 2.7	87.8 ± 2.8	−4.49 ± 3.26 [0.18]	−0.30 ± 3.99 [0.94]	−4.19 ± 3.96 [0.30]
Time >180 mg/dL, %*	3.2 ± 2.0	7.5 ± 2.6	6.2 ± 1.9	2.96 ± 2.69 [0.28]	−1.33 ± 3.42 [0.70]	4.29 ± 3.76 [0.26]
N with CGM <54 mg/dL§	3 (8)	3 (8)	4 (11)	[>0.99]	[>0.99]	[>0.99]
N with CGM <70 mg/dL§	8 (21)	8 (21)	10 (26)	[0.79]	[0.79]	[>0.99]
4 h after exercise
Mean glucose, mg/dL†	160 ± 34	172 ± 48	154 ± 42	−6.28 ± 6.28 [0.32]	−16.69 ± 8.21 [0.05]	14.02 ± 8.21 [0.10]
Time <54 mg/dL, %†	0.7 ± 2.3	0.3 ± 1.0	0.5 ± 2.1	−0.22 ± 0.53 [0.68]	0.22 ± 0.29 [0.46]	−0.44 ± 0.36 [0.22]
Time <70 mg/dL, %†	1.9 ± 5.0	1.8 ± 4.1	3.0 ± 6.6	1.10 ± 1.15 [0.35]	1.33 ± 0.95 [0.17]	−0.21 ± 1.00 [0.84]
TIR 70–180 mg/dL, %†	65.3 ± 23.8	63.6 ± 29.0	67.2 ± 27.1	1.90 ± 4.72 [0.69]	2.72 ± 5.44 [0.62]	−2.85 ± 4.55 [0.53]
Time >180 mg/dL, %†	32.8 ± 24.6	34.6 ± 29.7	29.8 ± 28.9	−3.00 ± 4.91 [0.55]	−4.05 ± 5.53 [0.47]	3.06 ± 4.79 [0.53]
N with CGM <54 mg/dL§	4 (11)	3 (8)	3 (8)	[>0.99]	[>0.99]	[>0.99]
N with CGM <70 mg/dL§	6 (16)	9 (24)	10 (26)	[0.40]	[>0.99]	[0.57]
Insulin delivery
1 h before exercise, unitsǁ	0.29 ± 0.23	0.59 ± 0.48	1.27 ± 0.81	0.98 ± 0.13 [<0.0001]	0.68 ± 0.13 [<0.0001]	0.30 ± 0.08 [<0.001]
During exercise, unitsǁ	0.33 ± 0.31	0.31 ± 0.45	0.94 ± 0.67	0.61 ± 0.09 [<0.0001]	0.64 ± 0.10 [<0.0001]	−0.02 ± 0.07 [0.75]
4 h after exercise, unitsǁ	14.27 ± 8.26	14.43 ± 7.08	13.62 ± 8.02	−0.65 ± 1.18 [0.59]	−0.81 ± 1.11 [0.47]	0.16 ± 1.09 [0.89]

Data are given as mean ± SD or n (%) unless otherwise specified. Boldface type indicates significance at P < 0.05.

*Adjusted means and SEs presented for each session. Adjusted mean differences, SEs, and P values were calculated from repeated-measures linear regression model adjusted for glucose at exercise onset and carbohydrate intake before and during exercise.

†Mean differences, SEs, and P values were calculated from robust regression model.

‡Change in glucose from exercise onset to nadir.

§P values determined from Fisher exact test. Counts differ slightly from total number of hypoglycemic events recorded during exercise, because these were based on capillary glucose measures, and a single participant could have sensor glucose readings that did not correlate with capillary glucose measure or fall into both hypoglycemia thresholds.

ǁMean differences, SEs, and P values determined from repeated-measures one-way ANOVA.

Table 1 reports time to hypoglycemia, CGM metrics, and insulin delivery before, during, and after exercise by session visit. When examining glycemic metrics before exercise, mean glucose at the start of exercise was similar between AF-60 (145 ± 29 mg/dL) and Auto (143 ± 32 mg/dL). However, mean glucose at the start of exercise during AF-30 tended to be lower than that with Auto, but not significantly (133 ± 30 mg/dL; P = 0.26). Glucose rate of change was similar among the three conditions at exercise onset.

Figure 2 illustrates the cumulative percentage of participants with a hypoglycemic event (defined as capillary glucose <70 mg/dL) during the 70-min exercise session for each of the three conditions. There was a tendency for AF-60 and AF-30 conditions to reduce the cumulative percentage of individuals with a hypoglycemic event by the end of exercise (AF-60 n = 11 [29%] of 38; AF-30 n = 9 [24%] of 38) compared with Auto (n = 16 [42%] of 38); however, these differences were not statistically significant (P = 0.34 and P = 0.14, respectively). Among those who developed hypoglycemia during exercise, the mean time to hypoglycemia onset was similar when Auto (36 ± 3 min) was compared with both the AF-60 (39 ± 4 min; P = 0.59) and AF-30 sessions (37 ± 4 min; P = 0.74). Despite a similar mean time to hypoglycemia onset, in those who developed hypoglycemia, a higher percentage of events occurred within the first 20 min of exercise for Auto (n = 4 [25%] of 16) compared with AF-60 (n = 2 [18%] of 11) and AF-30 (n = 0 [0%] of 9). For all conditions combined, the average carbohydrate intake required to return the glucose level to >90 mg/dL after hypoglycemia was 28 ± 11 g (median [IQR] 30 [16–32] g). When hypoglycemia occurred, the time it took to complete the exercise session was 90 ± 8 min (median [IQR] 90 [87–94] min).

Figure 2.

Percentage of participants with a hypoglycemic event (<70 mg/dL) over the exercise session time course in the three exercise sessions, including AF-60 (purple), AF-30 (teal), and Auto (yellow).

When CGM metrics were examined during exercise by session strategy, no significant differences were observed in TIR percentage metrics after adjusting for baseline glucose level at exercise onset and carbohydrate intake before and during exercise (Table 1). Nadir glucose during exercise was elevated with AF-60 and AF-30 compared with Auto, but significance was observed only between Auto and AF-60 (adjusted difference: AF-60 9.50 mg/dL; P = 0.02; AF-30 7.44 mg/dL; P = 0.12). When examining the change in glucose from preexercise to nadir glucose during exercise, changes with both AF-60 (−44 ± 33 mg/dL) and AF-30 (−36 ± 34 mg/dL) were significantly less than the change with Auto (−57 ± 35 mg/dL; both P < 0.05) (Table 1).

In the 4 h after exercise, there were no significant differences in TIR percentage metrics among sessions. However, as highlighted in Fig. 1 and Table 1, mean glucose was 6 mg/dL higher for AF-60 (P = 0.32) and 17 mg/dL higher for AF-30 (P = 0.05) compared with Auto (154 mg/dL).

Insulin delivery was assessed the hour before, during, and 4 h after exercise (Table 1 and Fig. 3). In the hour before exercise, compared with Auto, insulin delivery was significantly reduced when Activity feature was enabled (mean difference: AF-60 −0.98 units; P < 0.001; AF-30 −0.68 units; P < 0.001). Similarly, both Activity feature sessions significantly reduced insulin delivery during exercise (mean difference: AF-60 −0.61 units; P < 0.001; AF-30 −0.64 units; P < 0.001) compared with Auto.

Figure 3.

Box plots demonstrating the relationship between session and insulin delivery before (hashed lines) and during (solid color) exercise. Three exercise sessions include AF-60 (purple), AF-30 (teal), and Auto (yellow). Box plots extend from the 25th to 75th percentile; whiskers extend from the 10th to 90th percentile; horizontal lines within the boxes depict the medians, with the plus signs indicating the means. *P < 0.05 by two-way repeated-measures ANOVA.

Conclusions

This study evaluated the efficacy and safety of using the Omnipod 5 System Activity feature 60 or 30 min before the onset of endurance exercise in limiting hypoglycemia during exercise in individuals with type 1 diabetes. Although the use of Activity feature at both time points lowered insulin delivery and attenuated the drop in glucose during exercise, neither strategy significantly reduced hypoglycemia risk relative to Auto when used at these time points. In Activity feature conditions, hypoglycemia occurred in approximately one-quarter of study participants, suggesting that although there may be some protective effect of Activity feature on the drop in glucose during prolonged endurance exercise in the postabsorptive state, it should either be set earlier before exercise or be used as only one component of a broader exercise strategy for individuals living with type 1 diabetes (7,27).

Indeed, use of a multipronged approach to exercise management when using AID systems was recently endorsed in a position statement from the European Association for the Study of Diabetes and the International Society for Pediatric and Adolescent Diabetes, which recommends initiating higher temporary targets 1–2 h preexercise and consuming carbohydrates in small amounts (10–20 g) just before exercise if sensor glucose is <126 mg/dL (7). Although likely beneficial, use of higher temporary targets 1–2 h before physical activity occurs in <20% of all exercise sessions in real-world settings (13). Individuals are more likely to engage in strategies that require little to no preplanning, such as setting a higher temporary target at exercise onset or simply consuming carbohydrates as needed (13,14). The findings of this study suggest that setting Activity feature 30 or 60 min before exercise onset reduces insulin delivery and offers some protection against a drop in glycemia during exercise compared with Auto, but the risk of hypoglycemia remains.

Of note, we did not observe statistically significant differences in TIR, nadir glucose, or incidence of hypoglycemia between the two Activity feature conditions, and use of this feature provided only modest benefit over Auto when used within 1 h of exercise. These findings align with previous studies using standard pump therapy that explored basal rate reductions both closer to and further from exercise onset in a postabsorptive state (17,18). Specifically, Roy-Fleming et al. (17) found no significant differences in TIR during a 45-min cycling session, which occurred after an 80% basal reduction initiated 40 or 20 min before exercise or at exercise onset (t −40 63% ± 37%; t −20 66% ± 25%; t 0 65% ± 31%). Furthermore, an 80% basal rate reduction initiated 90 or 40 min before exercise (60 min of cycling) showed time to hypoglycemia was delayed in the t −90 min condition; however, the incidence of hypoglycemia during activity was not different between conditions (18).

Unlike a static basal rate reduction as highlighted in the abovementioned studies, the AID system in this study uses a model predictive control algorithm to tailor insulin delivery based on current glucose levels and trends, aiming to maintain a target glucose level of 150 mg/dL when Activity feature is enabled (21,22). Although we were unable to analyze minute-by-minute insulin delivery rates, Tagougui et al. (28) demonstrated that even with a full meal bolus dose, indicating exercise to an AID system 90 min before prolonged exercise tended to result in slightly lower circulating plasma insulin compared with no indication, highlighting the ability of an AID system to reduce insulin delivery in anticipation of and during exercise. Although we observed significant reductions in the amount of insulin delivered in the hour before and during exercise under both Activity feature conditions compared with Auto, this did not translate into meaningful differences in the glucose level at exercise onset.

In this study, a higher percentage of hypoglycemia events occurred with Auto within the first 20 min of exercise compared with both Activity feature conditions. Therefore, enabling Activity feature 30–60 min before activity may offer some protection against hypoglycemia for shorter-duration events, but setting it earlier for longer-duration events should be reinforced (7).

Current AID and exercise guidelines recommend setting a higher temporary target 1–2 h before activity and consuming 10–20 g carbohydrates just before exercise if preexercise glucose is <126 mg/dL (7). In the current study, participants only consumed carbohydrates when glucose was <90 mg/dL before exercise (16–24 g) or if hypoglycemia (<70 mg/dL) occurred during activity. Future work should more closely follow current recommendations to set Activity feature earlier and initiate carbohydrate intake at a higher glucose level to minimize hypoglycemia risk (7,29). Although our study was underpowered to conduct formal predictive analyses, exploratory comparisons (Supplementary Table 3) revealed that sessions with hypoglycemia were characterized by more time spent with a glucose level <70 mg/dL in the 24 h before exercise, slight preexercise glucose declines compared with relative stability, and lower preexercise glucose levels. These findings are in line with other work suggesting that other factors, such as recent glycemic trends, seem to influence hypoglycemia risk during prolonged activity (30).

Study strengths include highly controlled in-clinic exercise sessions under standardized conditions in a crossover design, per-protocol analyses, and a diverse participant cohort spanning various ages and diabetes durations. We also adjusted for known covariates that influence hypoglycemia risk during exercise (9,30). However, this study had several limitations. First, the study was underpowered to detect significant differences in hypoglycemia risk, as 12 participants were excluded from analyses because of protocol deviations. Second, the highly controlled nature of this study limits its generalizability to different exercise types, intensities, durations, and times of day. Third, although postexercise hypoglycemia incidence was reported, we did not standardize or control for participant behavior after discharge. Moreover, our strategies were tested during a postabsorptive state, where bolus insulin on board was low, which tends to attenuate hypoglycemia risk during exercise when using AID systems (7). Finally, Activity feature set at exercise onset was not evaluated, although this has been reported as a strategy initiated by ∼30% of individuals with type 1 diabetes who participate in endurance-based activities (14). As such, although our data suggest little difference between AF-30 and AF-60 in insulin delivery or glycemic outcomes, the benefit of setting Activity feature at exercise onset remains unclear.

In conclusion, our study demonstrates that in preparation for endurance exercise completed in a postabsorptive state, enabling Activity feature 30 or 60 min before activity reduces insulin delivery and minimizes glucose declines relative to Auto. However, the use of Activity feature at these time points does not eliminate hypoglycemia risk and is only marginally better than simply remaining in Auto during exercise. Therefore, continued optimization of algorithm targets and activation timing, along with carbohydrate consumption when needed and vigilance in glucose monitoring, as recent guidelines suggest (7), remains crucial for minimizing hypoglycemia risk during exercise.

This article contains supplementary material online at https://doi.org/10.2337/figshare.29425637.

Appendix

Omnipod 5 Exercise Research Group. Members include Sue A. Brown, Mary Voelmle, Emma Emory, Viral N. Shah, Halis K. Akturk, Nicole Schneider, Hal Joseph, Prakriti Joshee, Christie Beatson, Bruce W. Bode, Brooke Narron, Tricia Lopez, Mei Mei Church, Molly Piper, Jimena Perez, Suzan Bzdick, David W. Hansen, Sheri L. Stone, Ruth S. Weinstock, Irl B. Hirsch, Jesica Baran, Subbulaxmi Trikudanathan, Nancy Sanborn, Dori Khakpour, Jennifer L. Sherr, Michelle Van Name, Michelle Brei, Melinda Zgorski, Amy Steffen, Lori Carria, Sanjeev N. Mehta, Lori M. Laffel, Lindsay Roethke, Margaret Fisher, Rebecca Ortiz La Banca, Lisa Volkening, Louise Ambler-Osborn, Christine Turcotte, Emily F. Freiner, Bruce A. Buckingham, Dessi Zaharieva, Ryan Kingman, Kaisa Kivilaid, Krista Kleve, Trang T. Ly, Bonnie Dumais, Todd Vienneau, and Lauren M. Huyett.

Funding Statement

This study was funded by Insulet Corporation. D.P.Z. receives research support from the Leona M. and Harry B. Helmsley Charitable Trust and the International Society for Pediatric and Adolescent Diabetes. D.P.Z. is a board member of Breakthrough T1D.