Low-Dose Glucagon to Prevent and Treat Exercise-Associated Hypoglycemia in Individuals With Type 1 Diabetes: A Systematic Review and Meta-analysis

Abstract

OBJECTIVE

People with type 1 diabetes (T1D) struggle to manage exercise because of hypoglycemia risk.

RESEARCH DESIGN AND METHODS

This systematic review and meta-analysis evaluated low-dose glucagon's efficacy for preventing and treating exercise-induced hypoglycemia in T1D. Medline, Embase, and Cochrane CENTRAL were searched for randomized controlled trials and crossover studies until September 2024. The analysis included T1D adolescents and adults treated with low-dose glucagon versus nonglucagon treatments. Studies with glucagon-like peptides, noninsulin combinations, or uncontrolled exercise settings were excluded. Two authors extracted the data. The methodological quality was assessed with the Risk of Bias-2 tool and Grading of Recommendations Assessment, Development and Evaluations framework. Risk of Bias 2 informed a sensitivity analysis. The meta-analysis employed a random effects model to estimate the pooled treatment effects on hypoglycemia and time below range (TBR) (glucose <3.9 mmol/L), as well as secondary outcomes and adverse effects.

RESULTS

Of 6,792 records, 12 studies involving 248 individuals (mean age: 36 ± 10.5 years) met inclusion criteria. The meta-analysis showed significant reductions in hypoglycemia risk (risk ratio 0.54; 95% CI 0.35, 0.84) and TBR (−3.91 percentage points; 95% CI −6.27, 1.54) with low-dose glucagon. Sensitivity analysis yielded a slightly more confident effect size for hypoglycemia and TBR. However, overall adverse events increased with low-dose glucagon (risk ratio 2.75; 95% CI 1.07, 7.08). The included studies were few and heterogeneous, which may have influenced the overall outcomes.

CONCLUSIONS

Low-dose glucagon reduces exercise-induced hypoglycemia and TBR in T1D individuals. Future research should optimize glucagon dosage and timing for various exercise types and durations to confirm real-world effectiveness.

Graphical Abstract

Introduction

In people with type 1 diabetes (T1D), regular exercise offers several health benefits, including reduced risk of cardiovascular complications (1,2). However, managing blood glucose during exercise is challenging. Exogenously administered insulin levels in the blood cannot be lowered immediately, and glucagon secretion during hypoglycemia is diminished, making aerobic exercise prone to causing hypoglycemia (3–5). In recent years, advancements in treatment modalities have significantly improved diabetes management. Continuous glucose monitoring (CGM) systems have increased the time spent in euglycemia, reduced the incidence of severe hypoglycemia, and contributed to various other health benefits (6). Additionally, advancements in insulin pump technologies have played a crucial role in optimizing patient outcomes (7,8). However, incorporating these advancements into exercise routines requires careful planning and adaptation. Factors such as the type, duration, and intensity of physical activity, the circulating insulin levels, and individualized sensor glucose targets must be considered (6,9,10). Additionally, understanding and responding to trend arrows from CGM devices is essential for making timely adjustments, such as consuming carbohydrates. Incorporating strategies like bolus insulin adjustments into an overall exercise plan can further enhance glucose management during physical activity. Despite the well-documented benefits of regular exercise, barriers such as fear of hypoglycemia, difficulty with spontaneous activity, and uncertainty in adjusting insulin and nutrition remain significant challenges (11). These challenges have led to a growing focus on strategies to make exercise more accessible for people with T1D. The use of low-dose glucagon has been investigated as an alternative strategy for managing hypoglycemia during unplanned exercise. Exogenous glucagon serves as a substitute for the lack of endogenous glucagon action in people with T1D. An injectable dose of 100–300 μg of glucagon is effective by increasing blood glucose levels by ∼1.5–2 mmol/L within 15 min (12). It has also been shown to be noninferior to glucose tablets in terms of treatment success rates and achieving normoglycemia within 2 h postrescue (13,14). This rapid and predictable response makes it an attractive alternative to ingestible glucose for preventing and treating hypoglycemia. In addition to manual administration, researchers have explored incorporating glucagon into dual-hormone closed-loop systems, extending traditional insulin-only closed-loop systems. In these systems, glucagon is administered automatically, through either a dual-chamber insulin-glucagon infusion pump or a separate glucagon infusion pump. Algorithms for glucagon delivery vary (15–19): some microdose glucagon repeatedly to complement insulin delivery for optimal glucose control, while others use small doses as rescue to treat or prevent imminent hypoglycemia. Studies indicate that the primary benefit of the dual-hormone approach is during exercise (20). Glucagon is available as reconstituted powders (e.g., GlucaGen Hypo Kit; Novo Nordisk, and Glucagon Emergency Kit; Eli Lilly) or ready-to-use formulations (e.g., Balsamic; Eli Lilly, Gvoke; Xeris Pharmaceuticals, and Dasiglucagon; Zealand Pharma). Powder formulations require a multiple-step dissolvement process before use and face challenges like fibrillation and aggregation after dissolvement. In contrast, the ready-to-use glucagon formulations are quickly administered and resistant to aggregation but are only commercially available in high doses for severe hypoglycemia.

Many studies have explored the role of glucagon in preventing exercise-induced hypoglycemia in individuals with T1D, but with inconsistent findings. This systematic review aims to determine whether using low-dose glucagon in individuals with T1D reduces the risk of hypoglycemic events and/or the time spent below the target glucose range (<3.9 mmol/L). Additionally, we also explore other related glycemic outcomes as well as the incidence of adverse events with low-dose glucagon administration.

Research Design and Methods

This systematic review and meta-analysis adhere to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines and were recorded in the International Prospective Register of Systematic Reviews (https://www.crd.york.ac.uk/prospero, CRD42021262785).

Search Strategy and Selection Criteria

An electronic literature search for records published from inception to September 2024 was conducted in Medline (Ovid), Embase (Ovid), and Cochrane Controlled Register of Trials (CENTRAL via Cochrane Library). The search strategy used both index terms and free-text terms related to the key concepts of T1D and glucagon. The exercise was not included as a key concept in the search strategy, as preliminary test searches indicated that it would not yield additional relevant records. The Cochrane Highly Sensitive Search Strategy was used for Medline (Ovid) and Embase (Ovid) for identifying randomized controlled trials, and specific search filters were applied to include crossover studies and exclude animal studies. No language restrictions were applied. When a record indicating the registration of a potentially eligible trial was identified, additional searches were conducted to find any published articles reporting the trial’s results. The complete search strategy is available in Supplementary Table 1.

Backward and forward citation searches based on included studies were conducted using the online software citationchaser (21) to identify additional relevant records not identified in the database search.

All records were imported into EPPI-Reviewer 6 (ER6) (22), where duplicates were automatically removed using the system’s deduplication feature. Records retrieved through the citation search were also uploaded to ER6 and cross-checked for duplicates against the database search records.

Studies considered eligible for inclusion in the review were required to meet the following criteria: participants included children and adults with T1D exposed to both exercise and glucagon. Various forms of exogenously administered glucagon (native, soluble, nasal) were considered, along with different exercise modalities such as cycling, treadmill use, and jogging, performed within defined time intervals and controlled settings, clarifying how the session was monitored, encompassing both inpatient and outpatient settings. Comparator interventions included placebo, usual care, basal rate reduction, bolus reduction, carbohydrate intake, commercially available insulin pumps (continuous subcutaneous insulin infusion [CSII], sensor augmented pump [SAP], predictive low-glucose suspend [PLGS], and noncommercial insulin pump therapy [any pump systems and nonglucagon-driven closed-loop systems]).

Studies involving glucagon-like peptide analogs, glucagon in combination with noninsulin drugs, or exercise without defined time intervals and in uncontrolled settings (i.e., no information on exercise duration and monitoring conditions) were excluded.

The records were distributed among three independent reviewers (S.B.L., A.G.R., and K.N.), ensuring that every record was double-screened. Records that met the inclusion criteria were then retrieved for full-text screening. Discrepancies between reviewers were initially addressed through discussion, and, if unresolved, a third reviewer was consulted for arbitration. If primary data points were missing, we contacted the corresponding authors for clarification. The report, for reasons of excluding full-text assessed studies, is available in Supplementary Table 2.

Data Extraction

Summary data were extracted from each eligible study, including study characteristics, population details, and outcome measures. Study characteristics encompassed title, author, publication year, journal, ClinicalTrials.gov identifier, country, study type, intervention and comparators, exercise type, duration and intensity, and the phase for the registered end points (the entire trial period). Population details included the number of participants, age group (adults or adolescents), the number of participants who completed the study, the treatment modality, such as multiple daily injections or insulin pump therapy, duration of T1D, body weight, BMI, HbA_1c levels, and total daily insulin dose. The included studies reported either mean and SD, median and interquartile range, or mean and 95% CIs for the end points concerning continuous outcomes. These data were harmonized to mean and SD (23). Data on nadir plasma glucose and coefficient of variation were collected as per protocol, but, because of limited reporting, these end points were excluded.

Outcome measures were categorized as primary outcomes, including level 1 hypoglycemic events (sensor glucose/plasma glucose [SG/PG] <3.9) and time below range (TBR SG/PG <3.9 mmol/L). Secondary outcomes comprised level 2 hypoglycemic events [SG/PG <3.0 or 3.3 mmol/L]), time spent in various glucose ranges (TBR2 SG/PG <3.0 or 3.3 mmol/L, time in range [TIR] SG/PG = 3.9–10.0 mmol/L, time above range [TAR] SG/PG >10.0 mmol/L, and TAR2 SG/PG >13.9 mmol/L), mean SG/PG, the number of carbohydrate treatments or the grams of carbohydrates consumed per participant per visit, and adverse events. The definitions of the outcomes were determined by the way they were reported in the studies.

Critical Appraisal

The included studies were independently evaluated for methodological quality by two reviewers (S.B.L. and A.G.R.) using the Cochrane Collaboration’s revised Risk of Bias 2 (ROB-2) tool (24). Bias was assessed across several domains: the randomization process, deviations from intended interventions, missing outcome data, outcome data points, selection of reported results, and overall risk of bias. To streamline the ROB-2 assessment, we categorized the end points into three groups: glycemic outcomes (including the number of hypoglycemic events and time spent in various glucose ranges), carbohydrate consumption, and adverse events. The ROB-2 assessment for all glycemic outcomes was combined, as the evaluation criteria for each domain were consistent across all glycemic outcomes. A sensitivity analysis was performed on the primary outcomes to evaluate the impact of studies with a high risk of bias on the overall results. Three studies were categorized as high-risk because of modifications in their primary end points during the trials and randomization process. In the sensitivity analysis, two studies were excluded because of changes in their primary end points reported in ClinicalTrials.gov.

The Grading of Recommendations Assessment, Development and Evaluations (GRADE) (25) approach was used to rate and assess the quality of the evidence for the primary outcomes. For the outcome “number of participants with a hypoglycemia event,” we did not downgrade the quality of evidence; however, for the TBR outcome, a downgrading occurred. A summary table of the findings can be seen in Supplementary Table 3.

Data Synthesis and Analysis

A descriptive summary and analysis of the included studies were conducted to outline key characteristics such as publication year, trial design, sample size, country of origin, participant age, types of interventions, intervention duration, and comparators used.

For the included studies, glucose levels were measured through either CGM, blood samples (obtaining plasma glucose) or blood glucose monitoring. Closed-loop systems were defined as insulin pump systems that were installed with a dosing algorithm for insulin alone (single hormone) or both insulin and glucagon (dual hormone) based on real-time CGM values and trends. In this systematic review and meta-analysis, these systems were categorized as closed-loop systems. Current or usual care was defined as any routine treatment provided to individuals with type 1 diabetes, including CSII, SAP, and PLGS pumps.

We structured the statistical analyses into the following comparisons: an intervention group (including all studies using low-dose glucagon in any form) versus a comparator group. The comparator group included single-hormone insulin pumps, SAP, PLGS pumps, CSII, glucose tablets, placebo, and basal insulin reduction. Additionally, we also divided the studies into subgroups regarding the intervention group (injectable glucagon vs. comparator). As per protocol, we did not perform any analyses on children.

The primary outcome, hypoglycemic events, was measured as the number of participants experiencing events and analyzed as a binary outcome. TBR was expressed as a percentage and analyzed as a continuous outcome. The end points were assessed during exercise, including the postexercise phase, and, in some cases, across the entire trial. A quantitative synthesis was performed to assess whether the included studies were sufficiently homogeneous for analysis.

Data were analyzed in R (version 4.3.0) using the package “meta.” The meta-analysis was performed using the random effect model. Risk ratio (RR) with corresponding 95% CI was estimated using the Mantel-Haenszel method for binary outcomes, including the primary outcome of hypoglycemic events. The mean difference with the corresponding 95% CI was estimated using the inverse variance–weighted method for continuous outcomes, including the primary outcome of TBR. Heterogeneity between studies was assessed by the I² statistic and Cochran’s Q test for heterogeneity. Publication bias was assessed visually using a funnel plot. All data generated or analyzed during this study are included in the published article (and the Supplementary Material).

Results

The database searches yielded a total of 8,154 records. After removing 1,362 duplicates, 6,792 records were screened for eligibility based on their titles and abstracts. Of these, 35 full-text articles were assessed, and 12 randomized controlled and/or crossover trial studies meeting the inclusion criteria were included (Table 1). No additional eligible studies were identified through citation search. The steps of the study selection are described as a flow diagram in Fig. 1.

Table 1.

Information on the included trials

Short trial title including year (reference)	Source of publication	ClinicalTrial.gov identifier	Study design	Participants (N)	Type of glucagon	Injection vs. infusion	Intervention	Comparator	Exercise duration (min)	Exercise type	Mean age (years)	Sex (male/ female)	T1D duration (years)	Weight (kg)	BMI (kg/m2)	HbA1c (mmol/mol)/ (%)	TDD (units/kg)
Van Bon (2012) (39)	J Diabetes Sci Technol	NA	Feasibility study, crossover	10	GlucaGen; Novo Nordisk	Infusion	DHCL	Open loop	30	Treadmill	55 [44–70]	8/2	35 [18–50]	NA	NA	8.0 [6.7–9.6]	NA
Haidar (2013) (28)	CMAJ	NCT01297946	RCT, crossover	15	Recombinant glucagon; Eli Lilly	Infusion	DHCL	CC	30	Bicycle	47 ± 12.3	6/9	27 ± 14.8	74 ± 16.9	26 ± 4.6	7.9 ± 0.7	0.6 ± 0.12
Haidar (2015) (29)	Lancet Diabetes Endocrinol	NCT01754337	RCT, crossover	30	GlucaGen; Novo Nordisk	Infusion	DHCL	SHCL, CC	60	Treadmill	33 ± 18.0	NA	16 ± 11.0	NA	24 ± 4.4	60 ± 11/7.7 ± 1.0	0.7 ± 0.16
Haidar (2016) (33)	JCEM	NCT01905020	RCT, crossover	28	NA	Infusion	DHCL	SHCL, CC	60	Treadmill/ bicycle	33 ± 17.0	13/15	18 ± 12.0	NA	25 ± 3.8	58 ± 11/7.5 ± 1.0	0.7 ± 0.24
Jacobs (2016) (40)	DOM	NCT02241889	RCT, crossover	21	GlucaGen; Novo Nordisk	Infusion	APX	CC	60	Treadmill	32 ± 7.0	9/14	15 ± 9.5	74 ± 15.8	25 ± 4.6	61 ± 11/7.5 ± 1.0	0.6 ± 0.21
Taleb (2016) (30)	Diabetologia	NCT01930110	RCT, crossover	17	Glucagon; Eli Lilly	Infusion	DHCL	SHCL, CC	60	Bicycle	37 ± 13.6	9/8	23 ± 11.7	NA	25 ± 4.0	64 ± 10/8.0 ± 1.0	46 ± 15.8*
Rickels (2018) (27)	Diabetes Care	NCT02660242	RCT, crossover	15	G-pen Mini; Xeris	Injection	150 MDG	CC, 50% BRR, 40% glucose tabs	45	Treadmill	30 C25-43)	8/6	22 (14–31)	NA	24 (23–27)	51 (48–60)/6.8 (6.5–7.6)	NA
Castle (2018) (17)	Diabetes Care	NCT02862730	RCT, crossover	20	GlucaGen; Novo Nordisk	Infusion	DHCL	SHCL, CC, PLGS	45	Treadmill	35 ± 4.7	9/16	20 ± 8.9	77 ± 16.0	NA	58 ± 10/7.0 ± 0.9	NA
Steineck (2019) (41)	Diabetologia	NCT02882737	RCT, crossover	14	GlucaGen; Novo Nordisk	Injection	200 MDG	PLC	45	Bicycle	48 [23–66]	8/6	25 [8–51]	79 [52–100]	26 [21–30]	54 [43–65]/7.1 [6.1–8.1]	34 [16–75]*
Wilson (2020) (26)	Diabetes Care	NCT03424044	RCT, crossover	23	Liquid stable; Xeris	Infusion	DHCL	SHCL, PLGS	45	Treadmill	32 ± 7.0	13/10	17 ± 8.1	80 ± 14.7	NA	7.1 ± 0.9	NA
Lindkvist (2023) (34)	Front Endocrinol	NCT04949867	RCT, crossover	11	GlucaGen; Novo Nordisk	Infusion	DHCL	SHCL	45	Bicycle	15 ± 1.47	9/2	6 ± 2.45	NA	21 ± 2.4	55 ± 9	0.9 ± 0.26
Aronson (2023) (42)	Diabetes Care	NCT03841526	RCT, parallel	44	Liquid stable; Xeris	Injection	150 MDG, 50% BRR	PLC 50% BRR, 150 MDG	30–75	Free living	36 ± 9 38 ± 12 40 ± 14	9/7 7/7 6/8	NA	80 ± 20 81 ± 16 80 ± 15	27 ± 5 27 ± 5 28 ± 5	NA	NA

APX/APN, artificial pancreas with and without exercise announcement; BRR, basal rate reduction; CC, current care; CMAJ, Canadian Medical Association Journal; DHCL, dual hormone closed loop; DOM, Diabetes, Obesity and Metabolism; JCEM, Journal of Clinical Endocrinology and Metabolism; MDG, minidose glucagon (injectable); NA, not available; PLC, placebo; RCT, randomized controlled trials; SHCL, single hormone closed loop.

*Units only: Parentheses represent corresponding interquartile ranges; brackets represent corresponding ranges. All other values are in mean ± SD.

Figure 1.

Flow diagram of the study selection. The reports identified through the citation search are positioned at the bottom of the diagram to maintain a chronological flow. However, they underwent the same processes of deduplication, title and abstract screening, and full-text screening. Wrong study designs include studies that are not randomized or a crossover study. Wrong intervention involves uncontrolled exercise sessions or interventions other than glucagon, and the wrong population includes people with type 2 diabetes.

The 12 included studies are from 2013 to 2023, summarizing data from 248 individuals with a mean age of 36 ± 10.5 years. Eleven studies reported the sex distribution and reached 114 males and 110 females. Out of the 12 studies, one was designed as a parallel randomized study and one as a crossover feasibility study. Because of the inclusion of crossover designs, the number of data points for each outcome exceeds the overall total number of individuals included in this study. Risk of bias assessment is available in Supplementary Figs. 1–3.

For the hypoglycemic outcome, two studies (17,26) did not initially report the number of participants experiencing an event. Data were obtained after contact with the authors. Consequently, all 12 studies, comprising 546 data points, were included in the analysis (Fig. 2A). The pooled analysis showed a reduced risk of hypoglycemia with glucagon use (RR = 0.54; 95% CI 0.35, 0.84; P = 0.0062). Five studies (26–30) primarily drove this outcome. The I² statistic of 49% indicated moderate heterogeneity (Cochran’s Q test P = 0.028). The sensitivity analysis revealed a slightly weaker effect size for the hypoglycemic outcome (RR = 0.61; 95% CI 0.39, 0.96; P = 0.0319) and moderate heterogeneity (I² = 43.8%; P = 0.07).

Figure 2.

Forest plot for primary outcomes. A: Number of participants with a hypoglycemic event. Twelve studies were included in the analysis; in the intervention arm, 45 hypoglycemic events were compared with those in the comparator arm, with 147 events occurring; RR 0.54 (0.35, 0,84). B: TBR <3.9 mmol/L. Nine studies were included in the analysis, with a mean difference (MD) of −3.91 (−6.27, −1,54) percentage points between the two groups.

For the other primary end point, TBR, three studies were excluded: Steineck et al. (41) did not report this end point, and Aronson et al. (42) and Van Bon et al. (39) lacked the CIs required for meta-analysis. Therefore, nine studies, encompassing 521 data points, were included in the analysis (Fig. 2B). The pooled analysis showed a statistically significant reduction in TBR with glucagon use (mean estimated treatment difference [ETD] = −3.91 percentage points [pp]; 95% CI −6.27, −1.54; P = 0.0012), with five studies (17,27–30) contributing significantly to this result. The I² statistic (78.8%, P < 0.0001) indicated considerable heterogeneity between the studies. The sensitivity analysis changed the ETD to −2.90 pp (95% CI −4.94, −0.85), whereas heterogeneity remained considerable (I² = 76.1; P = 0.0008).

In the analysis of carbohydrate consumption, a total of five studies with 361 data points were included, demonstrating a reduction in grams of carbohydrate consumption for those receiving glucagon (ETD = −0.76 g; 95% CI −0.06, −0.47; P < 0.0001) with no heterogeneity (I² = 0.0; P = 0.70).

In the analysis of TBR level 2 (SG/PG <3.0 or 3.3 mmol/L), five studies with 472 data points showed a nonsignificant ETD of 0.51 pp (95% CI −1.24, 0.23). Additionally, six studies with 268 data points did not demonstrate any difference in the number of level 2 hypoglycemia events (SG/PG <3.0 or 3.3 mmol/L, depending on the definition of level 2 hypoglycemia) in those receiving and not receiving glucagon (RR = 0.59; 95% CI 0.29, 1.20).

No difference was observed for TAR (SG/PG >10.0 mmol/L) analysis, including nine studies with 458 data points (ETD = −0.85 pp; 95% CI −2.65, 0.95). Moreover, three studies with 342 data points reported a TAR level 2 (SG/PG >13.9 mmol/L) of −0.16 pp (95% CI −2.19, 1.88).

For mean SG/PG levels, 10 studies with 531 data points found an ETD of 0.22 pp (95% CI −0.41, 0.86) for glucagon versus nonglucagon arms. Furthermore, TIR (SG/PG = 3.9–10.0 mmol/L) were analyzed from nine studies with 521 data points and demonstrated a nonsignificant ETD of 4.48 pp (95% CI −0.78, 9.73) in favor of glucagon.

Seven studies comprised a total of 303 data points reported on adverse events and demonstrated an increased risk for adverse events associated with glucagon administration (RR = 2.75; 95% CI 1.07, 7.08; P = 0.036). However, the analysis showed a substantial heterogeneity (I² = 66%; P = 0.007). Among specific adverse events, only nausea was pronounced, with an RR of 3.66 (95% CI 1.64, 8.16; P = 0.002) (heterogeneity I² = 0.0%; P = 0.70). Figure 3 represents an overview of both the primary and secondary outcomes.

Figure 3.

Summary of the results of all of the outcomes. Outcomes are shown as rows categorized as primary, secondary, and adverse outcomes and ordered by P value within each category. The x-axis displays the pooled treatment effect and 95% CI of the random effects model in original units for continuous outcomes (circles) and as RRs for binary outcomes (squares). AE, adverse event.

The subanalysis comparing subcutaneous injectable glucagon versus controls reveals no significant results favoring glucagon administration.

The forest plots for the sensitivity analysis, the secondary end points, and the subanalysis are available in the Supplementary Materials.

Conclusions

This systematic review revealed 12 randomized controlled and/or crossover studies meeting the inclusion criteria and representing data from 248 adults with an equal sex distribution. The meta-analysis done on these studies showed a reduced risk of hypoglycemia as well as less time spent in hypoglycemia (SG/PG <3.9 mmol/L) during exercise when receiving low-dose glucagon as either a subcutaneous injection or an infusion in people with T1D when compared with either single-hormone insulin pumps, SAP, PLGS systems, glucose tablets, placebo, or basal insulin reduction. The same was seen for carbohydrate intake, leading to a more minor consumption when glucagon was used. Not surprisingly, the risk of experiencing an adverse event increased when glucagon was administered, with nausea more pronounced. No severe adverse events were reported in these studies. For all other outcomes, such as level 2 TBR, level 2 hypoglycemia, TAR, mean plasma glucose, and TIR, no differences were seen in whether glucagon was given or not.

The subgroup analysis, which included two to four studies comparing injectable glucagon with a control group, did not show any significant differences, likely because of the limited number of included studies. The formulations used were either soluble or native glucagon, and no studies involving nasal glucagon were included.

A key strength of this review is the adherence to Cochrane guidelines throughout critical stages, including the development of a comprehensive search strategy, systematic screening and selection of studies, and the use of validated tools such as ROB-2 for risk of bias assessment and GRADE for evaluating the certainty of evidence. The risk of bias assessment identified some studies with high concerns, which were addressed through a sensitivity analysis. The GRADE assessment supports, to some extent, the strength of this review, as no downregulation was necessary for the outcome concerning the number of participants experiencing hypoglycemia. However, this was not the case for the TBR outcome, where a downregulation occurred because of inconsistency and risk of publication bias.

The hypoglycemia outcome exhibited moderate heterogeneity, indicating that the variation of the treatment effect across the studies is genuine and not due to random fluctuations. The results after the sensitivity analysis excluding two studies suggest that study-level factors might have driven the original findings. The revised RR with reduced heterogeneity provides a more robust, reliable, and consistent estimate of the effect. A similar observation was made regarding the TBR outcome, but the sensitivity analysis failed to reduce heterogeneity. One possible explanation is the exclusion of studies that did not report TBR data or the varying definitions of the end point. Although the modest reduction in carbohydrate intake of 0.76 g with glucagon treatment may not significantly alter the overall strategy for managing hypoglycemia, that is, consuming carbohydrates during exercise, this finding might also reflect variations in how carbohydrate consumption was reported.

It is well documented (31,32) that the risk of exercise-induced hypoglycemia increases when physical activity is performed in the afternoon and evening. Most of the studies included (17,26,28–30,33,34) in this meta-analysis focus on glucose levels during exercise in this high-risk period, thereby enhancing the relevance and reliability of the results.

The limitation of our systematic review is the small number of studies included in the meta-analysis for many of the end points and the heterogeneity reflected in the I² value, primarily due to the different study designs, that is, exercise type—duration, time of day for performing exercise, intervention—and comparator type, as well as the way data were measured and presented. Additionally, the different algorithms used in the closed-loop systems could have influenced the results because of their complexity in design and interpretation, with no option to adjust or compensate for these effects. Even though we tried to make the included studies as comparable and aligned as possible, some challenges were seen. The consequence was the exclusion of studies in the meta-analysis if data were not reported or too differently reported to be comparable, which weakened the statistical power of the results. Different glucagon formulations were used across the included studies. While the pharmacokinetics and pharmacodynamics of GlucaGen and Gvoke are similar, Gvoke has been associated with a higher incidence of reported adverse events (13,35). This was not possible to confirm in our data set since the reported adverse events came from studies using either type. However, only 7 studies out of the 12 included studies provided data on adverse events, which should be carefully considered in interpreting the results. Despite transient adverse events from these glucagon formulations, significant benefits are documented.

The use of low-dose glucagon can result in a smaller glucose peak compared with glucose tablets, which often cause a postprandial glucose spike (13,14,27). Additionally, the reduced need for carbohydrate consumption may decrease the risk of weight gain, an important consideration for many individuals with T1D. Since exercise is often part of a healthier lifestyle, minimizing the need to consume extra carbohydrates during or after exercise can support these goals. It is important to note that these types of drugs are currently approved only for the treatment of severe hypoglycemia, with approvals for nonsevere hypoglycemia still pending (36). More user-friendly versions of low-dose glucagon are therefore needed before glucagon truly can fulfill any role as a treatment strategy for mild hypoglycemia in people with T1D.

To our knowledge, this is the first systematic review examining the effect of low-dose glucagon in treating and/or preventing exercise-induced hypoglycemia in people with T1D. Although divergent results have been observed in the literature, there appears to be a tendency toward a positive effect of glucagon in this context, and the findings from this meta-analysis support this. Knowledge gaps remain regarding the optimal glucagon dosage, timing, and its effectiveness across different types and durations of exercise.

While minidose glucagon appears safe in the short term, long-term effects with regular administration in people with T1D do not exist. However, in people with congenital hyperinsulinism, a daily usage of glucagon has been reported to have a reversible necrolytic migratory erythema (37). Furthermore, glucagon may have a weight-reducing effect (38). Nevertheless, the findings in this systematic review have important implications for the development of future studies that could bring us closer to a new treatment or prevention strategy for exercise-induced hypoglycemia in people with T1D. It remains speculative whether certain subgroups of individuals with T1D benefit more than others. The predictable glucose-raising effect of glucagon may be particularly reassuring for those with a fear of hypoglycemia and/or impaired hypoglycemia awareness. Making exercise more accessible for all individuals with T1D will ultimately improve quality of life and help reduce diabetes-related complications.

This article contains supplementary material online at https://doi.org/10.2337/figshare.29437703.

Funding Statement

This project received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. All authors are employed at Steno Diabetes Center Copenhagen, a public hospital and research institution under the Capital Region of Denmark, partly funded by a grant from the Novo Nordisk Foundation.