Rebound Hyperglycemia in Children and Adolescents with Type 1 Diabetes Using Automated Insulin Delivery Systems

Introduction

Automated insulin delivery (AID) systems have revolutionized the management of type 1 diabetes (T1D), improving glucose outcomes across different age groups. ¹ Despite these advances, hypoglycemia continues to be a major challenge to achieving optimal glycemic targets. ² During symptomatic and impending hypoglycemia, individuals often respond with excessive carbohydrate intake to alleviate or prevent symptoms rapidly, which may lead to rebound hyperglycemia (RH).^3–5

A recent study in adults with T1D demonstrated that RH follows approximately 35%–40% of all hypoglycemic episodes, occurring most frequently but with shorter duration in users of sensor-augmented and automated insulin pump-delivery systems. ⁴ Although AID algorithms are designed to attenuate glucose excursions through automated basal insulin infusion and correction boluses, the occurrence of and severity of RH may differ in children and adolescents using AID systems due to differences in alarm setting targets for hyper- and hypoglycemia, insulin sensitivity, hormonal counter-regulation, carbohydrate needs, and the degree of external supervision in diabetes care.^6,7 These physiological and behavioral differences may influence both the risk and treatment pattern of hypoglycemia. ⁸

Therefore, we aimed to quantify the frequency and severity of RH and to examine its associations with sensor glucose (SG) metrics in children and adolescents using AID systems in a real-world clinical setting.

Research Design and Methods

We conducted a cross-sectional real-world study including all children and adolescents with T1D followed at Steno Diabetes Center Copenhagen, Denmark. The study was approved by the Danish Data Protection Agency in the Capital Region of Denmark, R-22031406.

Eligible participants were <18 years old, using an AID system (Tandem Control-IQ [CIQ] or Medtronic MiniMed 780G [780G]), and had at least one continuous glucose monitoring (CGM) upload between September 2021 and August 2022 with ≥14 days of data and >70% CGM coverage. Individuals using open-source (“DIY-looping”) systems or with unclassified devices were excluded.

The details for data extraction and data analysis have been published elsewhere. ⁴ In short, personal characteristics—such as age, sex, insulin treatment regimen, diabetes duration, body mass index, and HbA1c—were obtained from electronic medical records (EPIC systems, WI, USA). CGM data were collected from Stenopool (CPH, DK), Diasend/Glooko (Glooko, CA, USA), or CareLink (Medtronic, CA, USA).

Hypoglycemia was defined as SG <3.9 mmol/L and hyperglycemia as SG >10.0 mmol/L without any duration requirement. Standard CGM metrics—time in range (TIR; 3.9–10.0 mmol/L), time below range (TBR; <3.9 mmol/L), time above range (TAR; >10.0 mmol/L), and coefficient of variation (CV)—were calculated. RH was defined as ≥1 SG value >10.0 mmol/L within 2 h of one preceding SG <3.9 mmol/L. Severity of RH was quantified by the area under the curve (AUC measured with the trapezoidal rule) of SG >10.0 mmol/L within the 2-h post-hypoglycemia period. Rebound hypoglycemia was defined as ≥1 value >10.0 mmol/L followed within ≤2 h by ≥1 value <3.9 mmol/L (Supplementary Fig. S1).

Primary outcome was RH severity. Secondary analyses examined associations of RH severity with CGM-derived TIR and CV, as well as differences in the frequency, duration, and severity of RH between AID systems.

Comparison between systems were statistically tested using the analysis of variance, adjusted for age, sex, and diabetes duration. If data remained skewed after log transformation, the Kruskal–Wallis test was used. Regression analysis was used to assess the association of RH severity with glucose metrics. Results are presented as mean ± SD, with significance at P < 0.05. Analyses were performed in SAS (version 9.2, SAS Institute Inc.; Cary, NC, USA).

Results

Among 3041 individuals with T1D using CGM, 390 were <18 years; 190 (49%) used AID (94 CIQ, 96 780G; Supplementary Fig. S2). Mean (±SD) age was 11.2 ± 4 years; 44% were female. Mean diabetes duration was 5.7 ± 5 years, HbA1c 53.9 ± 10.3 mmol/mol, TIR 69.5 ± 9.5%, TBR 2.6 ± 2.0%, TAR 27.9 ± 10%, and CV 37.8 ± 5%. Clinical characteristics differed slightly between systems, with 780G users showing shorter diabetes duration (7.0 ± 3.6 vs. 4.6 ± 3.3 vs. 7.0 ± 3.6, P < 0.001; Table 1).

OPEN IN VIEWER

Table 1.

Baseline Characteristics and Continuous Glucose Monitoring (CGM) Outcomes in Children and Adolescents with Type 1 Diabetes (T1D) Using Automated Insulin Delivery Systems (AID). Values Are Presented as Mean ± Standard Deviation (Std Dev) or as Number and Percentage. Participants Were Grouped by AID Systems: Dexcom CGM with Tandem Control-IQ (CIQ) and Guardian CGM with Medtronic MiniMed 780G (780G). The Table Compares Baseline Characteristics, HbA1c, Glycated Hemoglobin, Hypoglycemia Frequency, Rebound Hyperglycemia (RH) Frequency, Duration of Hyperglycemia following Hypoglycemia, Severity of RH Assessed by Area Under the Curve (AUC RH), and Hypoglycemia Preceding RH, Time Above Range (TAR, SG >10.0 Mmol/L); Time Below Range (TBR, SG <3.9 Mmol/L); Time in Range (TIR, SG 3.9–10.0 Mmol/L). P-Values Indicate Between-Group Comparisons.

	Total		Dexcom + tandem CIQ		Guardian + medtronic 780G		P value*
Variable	Mean/counts	Std dev/%	Mean/counts	Std dev/%	Mean/counts	Std dev/%	Overall
Baseline characteristics
Females, n (%)	84	44	47	50	37	38	0.11
Age, years	11.2	4	11.4	4.2	11	3.8	0.49
- Age group 0–5 years, n (%)	19	10	11	58#	8	42#	—
- Age group 6–12 years, n (%)	99	52	42	42#	57	58#	—
- Age group 13–17 years, n (%)	72	38	41	57#	31	43#	—
HbA1c, mmol/mol	53.9	10.3	53.2	8.1	54.7	12.1	0.29
T1D duration, years	5.7	5	7	3.6	4.6	3.3	<0.001
CGM outcomes
Sensor wear time, %	96.1	4.2	97.1	4.2	95.2	4.1	0.002
RH events per week, n	3.5	2	3.3	1.9	3.6	2.1	0.948
RH events per week (daytime), n	3.3	1.9	3.1	1.8	3.5	2	0.829
RH events per week (night), n	0.2	0.3	0.2	0.3	0.2	0.3	0.596
Duration of antecedent hypoglycemia per RH  event, min	26.3	19.3	23.8	25.2	28.8	10.2	<0.0001
Duration of hyperglycemia per RH event, min	96.6	39.1	109.4	41.2	83.9	32.2	<0.0001
AUC RH, mmol/L*min	1241.4	567.5	1424.8	606.8	1058.1	459.8	<0.0001
Hypoglycemia event week, n	9.4	5.5	9.3	5.7	9.5	5.3	0.707
Hypoglycemia leading to RH per week, n	3.7	2.0	3.8	2.0	3.6	1.9	0.156
Hypoglycemia leading to RH, %	40.7	16.9	40.9	17.5	40.6	16.5	0.492
Level 1 hypoglycemia (3.0–3.9 mmol/L)  leading to RH, %	40.6	18.6	40.5	18.6	40.8	18.8	0.686
Level 2 hypoglycemia (>3.0 mmol/L) leading	42.8	24.7	42.3	26.6	43.2	22.8	0.985
to RH, %
TIR, %	69.5	9.5	68.0	9.7	70.9	9.2	0.015
TBR, %	2.6	2.0	2.2	1.7	3.0	2.1	0.062
TAR, %	27.9	10.0	29.7	10.2	26.1	9.6	0.006
CV, %	37.8	5.1	37.5	4.9	38.0	5.3	0.935

All bold values are significant p-values.

^#This is the row percentage between AID devices.

AID, automated insulin delivery; AUC, area under the curve; CGM, continuous glucose monitoring; CIQ, Tandem Control-IQ; CV, time variation; RH, rebound hyperglycemia; T1D, type 1 diabetes; TH, rebound hyperglycemia; TAR, time above range; TBR, time below range; TIR, time in range.

Participants experienced a mean of 9.4 ± 5.5 hypoglycemia events per week, of which 3.5 ± 2.0 led to RH—equal to 41.0 ± 17.5% of all hypoglycemia events. The mean duration of hypoglycemia before the RH was 26 ± 12 min, while the RH lasted 96 ± 33 min. RH frequency did not differ between CIQ and 780G users (3.3 ± 1.9 vs. 3.6 ± 2.1, P = 0.57). However, 780G users had shorter and less severe RH episodes (AUC 1058 ± 460 vs. 1425 ± 607 mmol/L*min, P < 0.001) despite longer antecedent hypoglycemia duration (28.81 ± 10.2 vs. 23.81 ± 25.21 min, P < 0.001).

RH severity was inversely associated with TIR (r = –0.33, P < 0.001) and positively associated with CV (r = 0.12, P < 0.001), with no interaction by AID system (p for interaction = 0.67; Fig. 1).

OPEN IN VIEWER

FIG. 1.

Association between the severity of rebound hyperglycemia and glucose metrics by automated insulin delivery system. Scatter plots show the relationship between the area under the curve for rebound hyperglycemia (AUC RH, mmol/L × min) and time in range, defined as glucose 3.9–10.0 mmol/L (TIR, %), and glycemic variability, expressed as coefficient of variation (%). Data are shown for users of the Tandem Control-IQ system (CIQ; gray filled circles) and the MiniMed 780G system (780G; blue filled triangles). Symbols represent individual observations; lines indicate fitted linear regression trends, with shaded areas representing 95% confidence intervals.

Discussion

In this real-world study of children and adolescents with T1D using AID systems, RH was a frequent consequence of hypoglycemia, occurring after approximately 41% of all hypoglycemic events. Importantly, RH severity was associated with lower time in range and higher glycemic variability, underscoring RH as a clinically relevant parameter for achieving glucose targets for AID users.

The observed RH frequency was similar to that reported in adults with T1D using sensor-augmented insulin pump or AID systems, with RH occurring after about 30%–40% of hypoglycemic episodes. ⁴ We can only speculate on whether children and adolescents may be more vulnerable to experiencing RH due to heightened fear of hypoglycemia, higher SG alarm thresholds, variable symptom perception, parental involvement, and challenges in accurately estimating carbohydrate amounts during acute events.^5,9,10

As for adults, children, and adolescents using AID systems may require lower carbohydrate amounts to treat hypoglycemia due to reduced insulin-on-board resulting from the automated insulin suspension with decreasing glucose values. This aligns with updated international recommendations suggesting 5–10 g of carbohydrates for hypoglycemia treatment in AID users, rather than the previously recommended 15 g.^8,11 As the time of data collection was in the early phase of AID treatment, these recommendations were not yet widely implemented in clinical practice.

In this study, we found an inverse association between RH severity and TIR and a positive association between RH severity and glycemic variability, highlighting that RH may contribute to the overall glucose instability.

While RH frequency did not differ between CIQ and 780G users, RH episodes were shorter in duration and less severe among 780G users, despite a longer duration of antecedent hypoglycemia. These differences may partly be explained by differences in user characteristics between device groups, including the shorter diabetes duration among 780G users. However, without C-peptide measurements, the clinical relevance of this difference remains uncertain, as diabetes duration alone may not reliably indicate residual β-cell function. ¹² In addition, AID system–specific algorithm features may also influence post-hypoglycemia glucose trajectories. CIQ is generally considered a treat-to-range system, targeting glucose levels approximately between 6.25 and 8.9 mmol/L, whereas the 780G system is a treat-to-target system with selectable glucose targets of 5.5, 6.1, or 6.7 mmol/L. Furthermore, active insulin time is fixed at 5 h in CIQ but adjustable from 2 to 8 h in 780G. Automated correction boluses also differ between systems, with corrections targeted to approximately 6.1 mmol/L in CIQ and 5.5, 6.1, or 6.7 mmol/L in 780G. These algorithmic differences may add to the explanation of shorter and less severe RH episodes observed among 780G users. Data on individual pump settings, active insulin time, insulin-on-board, and the amount and frequency of autocorrections during the post-hypoglycemia period were not available in our dataset. We were therefore unable to evaluate RH frequency or duration in relation to these parameters, which should be addressed in future studies. However, as RH frequency remained high regardless of AID system, further optimization of both AID algorithms and hypoglycemia treatment strategies is warranted. Educational interventions tailored to AID users may help reduce RH, including a lower hypoglycemia alarm target, guidance on lower carbohydrate requirements (5 vs. 15–20 g of carbohydrates) during AID therapy, and education regarding expected post-hypoglycemia glucose trends—especially for children and adolescents who may have greater difficulty limiting carbohydrate intake during hypoglycemia treatment.

Strengths of this study include the relatively large pediatric cohort, the use of real-world CGM data, and the detailed assessment of RH severity using glucose AUC to capture duration and magnitude of the RH events to determine the associations of RH with the overall glycemic outcomes.

Several limitations should be acknowledged. First, the cross-sectional design precludes causal inference. Second, potential selection bias may have influenced comparisons between CIQ and 780G users. Third, information on carbohydrate intake, insulin dosing, physical activity, and other factors surrounding hypoglycemic events was not available. Fourth, only AID systems available during the study period were included, which may limit generalizability to newer technologies. Similarly, the relatively favorable glycemic outcomes observed among study participants may limit the generalizability of our findings to populations with higher HbA1c levels. Fifth, differences in glucose outcomes may partly reflect differences between the CGM sensors used with each AID system (Dexcom for CIQ and Guardian for 780G), as previous studies have shown varying results regarding the extent to which sensor differences influence reported glycemic outcomes.^13–15 Therefore, it remains difficult to determine whether, and to what extent, sensor-related differences may have influenced our findings. Finally, hypoglycemia and RH were defined based on single CGM values without minimum duration thresholds, which may have led to overreporting of events. However, defining hypoglycemia as at least one CGM value below 3.9 mmol/L (70 mg/dL) allowed us to capture all events in which users were likely to respond to perceived low glucose levels, supporting the clinical relevance of the definition.

Conclusion

RH is common among children and adolescents with T1D using AID systems and is associated with poorer sensor-derived glucose outcomes, such as TIR and glycemic variability. Although AID technology may attenuate RH severity, it does not eliminate the occurrence of RH. Optimizing hypoglycemia treatment strategies remains a key target for improving glycemic outcomes in pediatric AID users, irrespective of the AID system.

Authors’ Contributions

A.G.R., K.G.T., J.S., and K.N. conceived the study. AGR analyzed data and drafted the article. K.G.T. contributed to data collection. K.G.T., J.S., and K.N. critically reviewed the article. All authors approved the final version.

Supplemental Material

Supplemental Material