Role of glucagon in automated insulin delivery

Introduction

Type 1 diabetes (T1D) is caused by autoimmune destruction of the beta cells in pancreatic islets. Achieving near normal glucose levels in the treatment of T1D can help prevent the long term consequences of this disease.¹ Despite advances in medications and technologies for T1D treatment, attaining normal glucose levels remains very challenging and is often limited by hypoglycemia.^2,3 One contributing factor to hypoglycemia arises from the dysfunction of glucagon secretion in this disease. The absence of insulin alters the normal paracrine responses within the remaining cells of the islet leading to dysregulation of glucagon secreted from the alpha cells.⁴ Glucagon secretion can be inappropriately elevated in the setting of hyperglycemia or after a meal and can be blunted or absent during hypoglycemia.⁵ The impaired glucagon response contributes to ineffective glucose counter-regulation which greatly increases the risk of severe hypoglycemia, defined as hypoglycemia requiring the assistance of another person to treat.⁶ The treatment of T1D currently consists of multiple daily injections or continuous subcutaneous delivery of insulin through a pump. Newer pump technologies can now be controlled either directly by the patient or automatically through feedback from a continuous glucose monitor (CGM). ‘Automated insulin delivery’ or ‘insulin only/single hormone closed loop’ are two interchangeable terms for these newer pump systems that automatically dose insulin based on CGM values. Insulin analogs used in these treatments have significantly delayed pharmacodynamics compared to endogenous insulin, leading to periods of inappropriate hyperinsulinism which also contributes to hypoglycemia.⁷ Exercise and postexercise are especially high risk times for hypoglycemia due to changes in insulin sensitivity and increased insulin-independent glucose uptake into muscle.⁸ The glucagon response can be attenuated during exercise and the high catecholamine state of exercise can mask symptoms of hypoglycemia.⁹ Glucagon injected intramuscularly or subcutaneously in 1 mg doses is very effective to correct severe hypoglycemia. Over the last two decades, minidose glucagon, which is subcutaneous small doses of glucagon, was used in children, then adults to treat hypoglycemia when a person with diabetes is unable or unwilling to take oral carbohydrates.¹⁰ Within the last decade, research groups working on closed loop insulin and glucagon delivery systems are using automated microgram-sized doses of glucagon delivered through a subcutaneous pump systems with the goal of preventing or shortening periods of hypoglycemia and preemptively in the setting of exercise to decrease risk of hypoglycemia.¹¹

Brief overview of glucagon physiology and pharmacology

Glucagon is a 29 amino acid peptide made by alpha-cells of the pancreatic islet. In normal physiology, glucagon levels increase with hypoglycemia, fasting, exercise and protein-rich meals.¹² In normal physiology, hypoglycemia elicits an abrupt rise in glucagon, epinephrine and growth hormone which function to increase glucose.¹³ Glucagon’s main target is the liver where it binds to a G-protein coupled receptor and signals through cAMP to induce glycogenolysis and to a lesser degree gluconeogenesis to raise plasma glucose levels within 10–30minutes.^14,15 In T1D, glucagon responses are impaired or absent which further perpetuates episodes of hypoglycemia. Recurrent hypoglycemia results in reduced physiologic responses to hypoglycemia including impaired glucagon and epinephrine increases.^2,9

The dosing in standard hypoglycemia rescue kits is 1mg delivered intramuscularly, whereas the typical dosing used in dual hormone systems is microdoses of 10 to 100 micrograms delivered subcutaneously, typically amounting to less than 1mg/day. Importantly for use in dual hormone systems, subcutaneous injections of microgram doses of glucagon result in a dose dependent rise in plasma glucose levels at a wide range of starting glucose values. ^16,17 Unlike insulin, glucagon delivered subcutaneously demonstrates a more rapid onset and offset of action making it ideal for use in automated delivery systems.¹⁸

Glucagon in closed loop glucose control systems

Potential benefits of glucagon in closed loop systems

Closed loop systems are pump systems that deliver insulin and optionally glucagon dynamically with a controller algorithm based on values from CGM. Systems that deliver glucagon in addition to insulin are referred to as ‘dual hormone closed loop systems.’ A recent meta-analysis¹⁹ of available studies published in February 2018 showed that use of either single or dual hormone closed loop systems led to an increase in the percent time in the glucose target range (70–180 mg/dl, 3.9–10 mmol/L) by 140 minutes/day compared with study participants’ usual care. Time in hypoglycemia (<70 mg/dl or 3.9 mmol/L) also reduced by 20 minutes/day for participants using closed loop systems. In many clinical trials, insulin only automated delivery systems increased time in target range and reduced the time spent in hypoglycemia (<70mg/dL or 3.9 mmo/L) to 3% or less as compared with sensor augmented pump or pump use without CGM.²⁰ It is important to contextualize these findings; even if time in hypoglycemia is reduced to 3% on average, that is still 43 minutes per day. These numbers reflect the average experience of the clinical trial participants, who may have better diabetes control than a typical person with T1D. Even within these study participants, some people experienced significantly higher time in hypoglycemia than is represented by these mean results. The best insulin-only closed loop systems typically only achieve 70–80% time in target range and about 2–3% time in hypoglycemia. Therefore, it is important to consider additional strategies that can help people with T1D improve their glycemic control. Glucagon acts as the counterregulatory hormone to insulin in normal physiology and therefore is a natural consideration for inclusion in closed loop systems to reduce hypoglycemia. However, there will always be physiologic circumstances where glucagon cannot prevent hypoglycemia (see Potential limitations for use of glucagon in closed loop systems below).

Many patients with type 1 diabetes struggle with hypoglycemia from exercise and may avoid exercise for this reason.^21,22 Exercise is a high risk time for hypoglycemia due to insulin independent disposal of glucose by contracting muscles, increased insulin sensitivity and increased insulin absorption from the subcutaneous space due to increased blood flow. ^23,24 The risk for hypoglycemia after exercise continues for many hours, with overnight being a particularly vulnerable time for hypoglycemia.²⁵ The current guidelines⁸ for glucose management with exercise focus on carbohydrate intake before exercise in addition to basal and bolus reductions; these approaches can lead to unwanted hyperglycemia. Additionally, requiring additional calorie intake for exercise is a major disadvantage for the over 50% of adults with type 1 diabetes who are overweight or obese and are exercising to help better manage their weight.²⁶ Subcutaneous injections of glucagon (150–200μg) delivered before or after exercise are effective to reduce or avoid hypoglycemia without need for preemptive carbohydrate intake.^11,27 Our research group and others have shown that dual hormone automated systems can help reduce exercise-induced hypoglycemia during aerobic exercise compared with single hormone automated systems.^28–30

Dual hormone systems may be indicated for certain patient populations. Patients with frequent hypoglycemia and hypoglycemia unawareness are often excluded from single hormone AID clinical trials, potentially skewing the results. These patients may experience more hypoglycemia than patients without hypoglycemia unawareness. Glucagon could serve as a useful line of defense against hypoglycemia for these patients. Additionally, dual hormone systems could prove useful to help reduce hypoglycemia awareness in these patients by reducing the occurrence of hypoglycemia.³¹ It is known that patients start to become more hypoglycemia aware after 3–4 weeks of no hypoglycemia events.³² More work in this patient population is anticipated in the future.

Closed loop systems typically use one of three strategies for mealtime insulin dosing; 1) a traditional premeal bolus based on carbohydrate counting is entered by the user, 2) a meal is announced to the system with limited information (i.e. small, medium, large) or no information on carbohydrate content, or 3) no meal announcement is made. Carbohydrate counting is burdensome for the person with T1D and tends to be fairly inaccurate.³³ A fully automated system that does not require a meal announcements would be ideal, however this approach is limited by postprandial hyperglycemia and late postprandial hypoglycemia.³⁴ Dual hormone closed loop systems may allow for fully automated mealtime insulin dosing as the glucagon theoretically could prevent late postprandial hypoglycemia. In practice, studies with dual hormone systems have yet to fully realize this potential.^35,36

Review of dual hormone closed loop studies

An early report of a dual hormone system used over 7 days in pancreatectomized dogs was published in 1982.³⁷ With improvements in CGM, algorithm and pump technology, advances in the field have been substantial over the last 10 years. Twenty dual hormone studies published since 2010 are included in this overview (Table 1). ^{12,28–31,35,36,48–51} The dual hormone systems in these studies are primarily compared to a control group. The control group in early studies typically involved participants using insulin pump alone (also known as CSII, continuous subcutaneous insulin infusion),while later studies used control groups wherein participants used sensor augmented pump (SAP) therapy. Most recently, the control group has been participants using predictive low glucose suspend systems (PLGS). Some studies compare single hormone to dual hormone systems. The sophistication of the dual hormone closed loop systems progressed from algorithms run on laptop computers with manual input of the infusion rates to algorithms run on a smartphone or infusion pump with automated infusion rate adjustment on a portable wireless system. The clinical studies have progressed from partial days under close supervision in an inpatient clinical research center to multiple days in a free-living outpatient setting. Some of the older studies report plasma glucose outcomes, while the newer studies report outcomes based on CGM data. There is now guidance on standardized reporting of closed loop study results.⁵² With all of this evolution, one cannot directly compare results of the older studies with the newer studies as they are fundamentally very different. However, this overview serves to show the trajectory of this field including advantages and disadvantages (see Table 1).

Table 1: Dual hormone studies published between 2010–2019.

Study characteristics and results for the twenty dual hormone studies reviewed.

			Participant Characteristics
First Author (ref)	Date of Online Publication	Location of Research Group	Number of participants	Mean age (years)	Mean DM Duration (years)	Mean A1c (%)	Duration	Study design	Treatment group(s)	Control group	Location	Standarized Exercise	Components	Monitoring		First Author (ref)	AID Initialization	AID Algorithm	Meals Announced?	Glucagon Delivery Strategy	Glucagon Delivered (SD)	Mean Glucose mg/dl (SD)		Time in Range 70–180mg/dl (SD)		Time in Hypoglycemia <70md/dl (SD)		Time in Hyperglycemia >180mg/dl (SD)		Hypoglycemia Treatments Per Day		Adverse Events
El-Khatib et al. (35)	4/16/2010	Boston, MA	11 adults	40	23	7.3	27 hours	Non-randomized	DH	No control group	Inpatient	No	Computer, Deltec Cozmo pumps, venous blood glucose samples, manual entry of rates	Onsite		El-Khatib et al. (35)	Weight only	Model preditive control	No meal annoucement allowed	Proportional derivative control to prevent <100mg/dl	1.70mcg/kg/day	164 (SD 17)		62%		<1%		38%		0		No nausea or vomitting with glucagon
Russell et al. (50)	8/28/2012	Boston, MA	6 adults	52	32	7.4	51 hours	Non-randomized	DH	No control group	Inpatient	No	OmniPod pumps, Navigator CGM	Onsite		Russell et al. (50)	Weight only	Model preditive control	Yes	Proportional derivative control to prevent <100mg/dl	3.6mcg/kg/day (5.1)	158 (SD 44)		68% (SD 12)		0.7% (0.8)		31% (12)		3.2g/kg/day (0.3)		Subject who received highest amount of glucagon
El-Khatib et al. (51)	2/4/2014	Boston, MA	12 adults/12 adolescents	45/15	27/7	7.3/7.9	48 hours	Randomized	meal prime, WO meal prime	No control group	Inpatient	Yes	OmniPod pumps, Navigator CGM	Onsite		El-Khatib et al. (51)	Weight only	Model preditive control	Yes	Proportional derivative control to prevent <100mg/dl	WMP 6.8mcg/kg/day (2.5); WOMP 6.6mcg/kg/day (3.5)	WMP 132 (9); WOMP 146 (9)	0.03	WMP 80%; WOMP 70%	0.04	WMP 5.1%; WOMP 3.6%	0.7	Not reported		0.92/day		Two subjects reported nausea.
Russell et al. (34)	6/17/2014	Boston, MA	20 adults/32 adolescents	40/16	24/9	7.1/8.2	5 days	Randomized, crossover	DH	Usual care with pump or SAP (45% own sensor use)	Adults: Hotel, Adolescents: Diabetes camp	No	Iphone 4S, G4 Dexcom, Tslim pumps M5	Onsite		Russell et al. (34)	Weight only	Model preditive control	Recommended but not required	Proportional derivative control to prevent <100mg/dl	0.82mg/day (0.41)	Adults DH 133 (13), control 159 (30.4)	<0.001	Adults DH 79.5% (8.3), control 58.8% (14.6)	<0.001	Adults DH 4.1% (3.5), control 7.3% (4.7)	0.01	Adults DH 16.5% (1.8), control 33.8% (16.4)	<0.001	Adults DH:2.2 (3.2), control 3.4 (3.1)	0.15	One patient reported nausea, two reported one episode of vomitting.
Russell et al. (40)	2/7/2016	Boston, MA	19 children	9.8	5	7.8	5 days	Randomized, crossover	DH	Usual care with pump or SAP	Diabetes camp	No	Iphone 4S, G4 Dexcom, Tslim pumps	Onsite		Russell et al. (40)	Weight only	Model preditive control	Recommended but not required	Proportional derivative control to prevent <100mg/dl	10.9 μg/kg/day(4.0)	DH 136.8 (10.8), control 167 (30.6)	<0.001	DH: 80.6% (7.4), control 57.6% (14)		(< 60) DH: 1.2% (1.1), control: 2.8% (1.2)	<0.001	DH 16% (6.4), control 36.3% (15.7)	<0.001	events/participant/study: DH 3, control 5	0.037	No difference in nausea rates DH vs. control.
El-Khatib et al. (41)	12/23/2016	Boston, MA; Worcester, MA; Palo Alto, CA; Chapel Hill, NC	43 adults	33.3	16.9	7.7	11 days	Randomized, crossover	DH	Usual care with pump or SAP (59% own sensor use)	Home	No	Iphone 4S, G4 Dexcom, Tslim pumps	Remote		El-Khatib et al. (41)	Weight only	Model preditive control	Recommended but not required	Proportional derivative control to prevent <100mg/dl	6.8 μg/kg/day	DH 140.4 (10.8), control 162 (28.8)	<0.001	DH 78.4% (6.0), Control 61.9% (14.4)	<0.001	(<60) DH 0.6% (0.6), control 1.9% (1.7)	<0.001	DH 19.8% (6.1), control 33.6% (16.4)	<0.001	DH 0.4 (0.3), control 0.9 (0.7)	<0.001	21 reported nausea with DH. 5 reported nuasea in control.
Haidar et al. (43)	1/30/2013	Montreal, Canada	15 adults	47.1	26.5	7.9	15 hours	Randomized, crossover	DH	SAP	Inpatient	Yes	Minimed veo pumps, Sofsensor Medtronic, manual entry of rates	Onsite		Haidar et al. (43)	Weight, TDD, ICR	Model predictive control	Yes	Logical rules based on glucose value and trend	0.076mg/visit	Dh 140.4, control 142.2	0.74	DH 70.7%, control 57.3%	0.003	DH 0%, control 10.2%	0.01	DH 29.3%, control 25.6%	0.74	At least one event per study: DH 7%, control 53%	0.02	None other than hypoglycemia
Haidar et al. (30)	12/2/2014	Montreal, Canada	20 adults/10 adolescents	43/14	16/8	7.7/7.9	24 hours	Randomized, crossover	SH, DH	Usual care with pump	Inpatient	Yes	Minimed veo pumps, Sofsensor Medtronic, manual entry of rates	Onsite		Haidar et al. (30)	Weight, TDD, ICR	Model predictive control	Yes	Logical rules based on glucose value and trend	2.0 μg/kg/vist	DH 135 (34), SH 126 (27), control 144 (59.4)	DH:C 0.26	DH 63% (18), SH 62% (18), control 51% (19	DH:C <0.001	DH 1.5%, SH 3.1%, control 13.3%	DH:C <0.001	DH 20% (15), SH 20% (16), control 26% (22)	DH:C 0.83	At least one event per study: DH 21%, SH 17%, control 83%	DH:C <0.001	Not reported
Haidar et al. (44)	6/13/2015	Montreal, Canada	33 children	13.3	7.5	8.3	3 nights	Randomized, crossover	SH, DH	Usual care with pump	Diabetes camp	No	G4 dexcom, Accu-check combo pumps, tablet computer, manually entry of rates	Remote		Haidar et al. (44)	Weight, TDD, ICR	Model predictive control	N/A	Logical rules based on glucose value and trend	0.7μg/kg/night (1.0)	DH 138.6 (30.6), SH 145.8 (30.6), control 167 (25)	DH:C <0.001	DH 84%, SH 77%, control 54%	DH:C <0.001	DH 0%, SH 3.1%, control 3.4%	DH:C <0.001	DH 13%, SH 20%, control 38%	DH:C <0.001	events/night: DH 0, SH 0.04, C 0.15		Not reported
Haidar et al.(45)	11/3/2015	Montreal, Canada	21 adults/7 adolescents	39/15	21/8	7.4/7.7	2 nights	Randomized, crossover	SH, DH	Usual care with pump	Home	Yes	Paradigm veo pump, Enlite sensor, manual entry of rates	Onsite		Haidar et al.(45)	Weight, TDD, ICR	Model predictive control	Yes	Logical rules based on glucose value and trend	0.4μg/kg/night	DH 111.6 (28.8), SH 111.6 (30.6), control 120.6 (43)	DH:C 0.57	DH 93%, SH 91%, control 70%	DH:C <0.001	DH 1%, SH 5%, control 14%	DH:C <0.001	DH 0%, SH 0%, control 4%	DH:C 0.66	total events: DH3 SH 6, control 14,		Not reported
Gingras et al.(46)	6/15/2015	Montreal, Canada	12 adults	51.3	32.6	7.4	14 hours	Randomized, crossover	DH with ICR prandial dose, DH with qualitative prandial dose	Usual care with pump	Inpatient	No	Accu-check combo pump, Enlite Medtronic sensor, manual entry of rates	Onsite		Gingras et al.(46)	Weight, TDD, ICR	Model predictive control	Yes	Logical rules based on glucose value and trend	ICR 0.044mg/study, Qualitative 0.042mg/study	ICR 147.6 (37.8), Qualitative 151.2 (30.6), control 172.8 (36)	ICR:C 0.03	ICR 66.8%, Qualitative 64.2%, control 49.9%	ICR:C 0.10	ICR 0.1%, Qualitative 5.4%, control 5.6%	ICR:C 0.81	ICR 20.7%, Qualitative 29.3%, control 40.5%	ICR:C 0.03	Not reported		Not reported
Taleb et al. (12)	10/6/2016	Montreal, Canada	17 adults	37.2	23.1	8	90 minutes	Randomized, crossover	DH/SH with Continuous exercise, DH/SH with Interval exercise	No control group	Inpatient	Yes	Computer, Dexcom G4 sensor, MiniMed Paradigm Veo, manual entry of rates	Onsite		Taleb et al. (12)	Weight, TDD, ICR	Model predictive control	N/A	Logical rules based on glucose value and trend	DH Continuous 0.126mg/study (0.057), DH Interval	Not reported		DH Cont 100%, SH Cont 68.1%; DH Interval 100%, SH Interval 72.5%	Cont: 0.004 Interval: 0.11	DH Cont 0%, SH Cont 22.5%; DH Interval 0%, SH Interval 0%	Cont: 0.07 Interval: 0.04	Not reported		total events: DH Cont 2, SH Cont 4, DH Interval 1, SH Interval 6		Not reported
Haidar et al.(38)	1/18/2017	Montreal, Canada	23 adults	41	24	7.5	60 hours	Randomized, crossover	SH, DH	Usual care with SAP	Home	No	Dexcom G4, Sensewear, Accu-check combo pumps, manual entry of rates	Onsite		Haidar et al.(38)	Weight, TDD, ICR	Model predictive control	Yes	Logical rules based on glucose value and trend	7.9 μg/kg/study (4.1)	DH 142.2 (50.4), SH 142.2 (50.4), control 135 (54)	DH:C 0.16	DH 79%, SH 75%, control 64%	DH:C 0.31	DH 3.6%, SH 3.9%, control 7.9%	DH:C 0.002	DH 16%, SH 20%, control 15%	DH:C 0.13	total events: DH 6, SH 14, control 34		Not reported
Castle et al.(39)	3/23/2010	Portland, OR	14 adults	36.7	14.1	7.6	28 hours	Randomized	Low gain, high gain glucagon delivery	Placebo delivery	Inpatient	No	Insulin via animas IR 1000 pump, glucagon via medfusion 2001 syringe pump, manual entry of rates	Onsite		Castle et al. (39)	TDD, BMI, Age	Fading memory proportional derivative	Yes	High gain, low gain	low gain: 0.746mg/day (134) vs. high gain 0.516mg/day	High gain: 138 (SD 17); Low gain: 157 (SD 24) placebo 135 (SD 16)		Numbers not reported	NS	DH 1.04%, control 2.77%	DH:C 0.04	Numbers not reported	NS	DH 1.1/study (0.5), control 3.9/study (1)	0.01	One subject reported nausea with during DH
Jacobs et al.(29)	6/24/2016	Portland, OR	21 adults	32	15.4	7.5	22 hours	Randomized, crossover	APX with exercise adjustment, APN with no exercise adjustment	Usual care with SAP	Inpatient	Yes	Google nexus smartphone, tslim pumps, dexcom G4	Onsite		Jacobs et al.(29)	TDD, BMI, Age	Proportional derivative	Yes	Reduced insulin and increased glucagon for exercise	APX 3.6μg/kg, APN 2.8μg/kg	APX 154.8, APN 145.8, control 154.8	APX:APN 0.032	APX 75%, APN 81%, control 72%	APX:APN 0.11	APX 0.3%, APN 3.1%, control 0.8%	APX:APN 0.001	APX 25%, APN 17%, control 27%	APX:APN 0.09	total events: APX 6, APN 9, control 7	APX:APN 0.08	Not reported
Castle et al.(28)	5/13/2018	Portland, OR	20 adults	34.5	20.2	7.5	4 days	Randomized, crossover	SH, DH, PLGS, CC	Usual care with pump or SAP (65% own sensor use)	Outpatient	Yes	Google nexus smartphone, tslim pumps, dexcom G5	Remote		Castle et al.(28)	TDD, BMI, Age	Fading memory proportional derivative	Yes	Reduced insulin and increased glucagon for exercise	0.510 mg/day (0.207)	DH 149 (38), SH 145 (31), PLGS 170 (49), control 164 (62)	DH:C 0.29	DH 72.0 (10.8), SH 74.3 (8.0), PLGS 65.2 (13.5), control 63.1 (17.3)	DH:C 0.010	DH 1.3 (1.0), SH 2.8 (1.7), PLGS 2.0 (1.5), control 3.1 (3.2)	DH:C 0.007	DH 26.7 (11.3), SH 22.9 (8.7), PLGS 32.8 (13.9), control 33.7(18.1)	DH:C 0.054	per day: DH 0.8 (0.7), SH 1.7 (1.4), PLGS 1.3 (1.3), control 1.5 (1.2)	DH:C 0.010	GI upset in 23% of DH, 0% SH, 13% PLGS, 5% control
Von Bon et al.(48)	10/16/2012	Amsterdam, The Netherlands	10 adults	55.4	34.6	8	8 hours	Non-randomized	DH	Usual care with pump	Inpatient	Yes	Computer, System gold medtronic minimed sensor, D-tron+ pumps, Polar HR monitor	Onsite		Von Bon et al.(48)	TDD	Proportional derivative control	No	Glucagon given if <117, bolus size based rate of fall of glucose	postbreakfast 0.04 mg, postexercise 0.12mg, postlunch 0.07mg	DH 156.6, control 162	DH:C 0.74	DH 62.3%, control 61.2%	DH:C 0.78	DH 5.3%, control 4.1%	DH:C 0.60	DH 32.4%, control 34.7%	DH:C 0.54	events: DH 4, control 2		Not reported
Von Bon et al.(36)	11/15/2013	Amsterdam, The Netherlands	16 adults	52.1	35.3	7.6	48 hours	Non-randomized	DH	Usual care with pump	Home	Yes	Computer, Sofsensor or Enlite medtronic minimed sensor, D-tron+ pumps	Onsite		Von Bon et al.(36)	TDD	Proportional derivative control	No	Glucagon given if <117, bolus size based rate of fall of glucose	Daytime 1.7–2.7mg, Nighttime 0.6mg	Day 2 median: DH 7.70 (2.29), control 8.84 (0.87)	DH:C 0.0273	Day 2 median: DH 76.5% (23.9%), control 66.0% (29.8%)	DH:C 0.1618	Day 2 median: DH 2.8% (9.8%), control 0.0% (11.1%)	DH:C 0.0172	Day 2 median: DH 18.3% (20.0%), control 31.0% (29.8%)	DH:C 0.0889	events: DH 6, control 6		Not reported
Blauw et al.(49)	3/22/2016	Amsterdam, The Netherlands	16 adults	41	18	7.7	4 days	Randomized, crossover	DH	Usual care with pump	Inpatient then Home	No	Inreda DiabeticBV(contains CGM, control algorithm, insulin pump glucagon pump), Enlite medtronic sensors	Onsite/Remote		Blauw et al.(49)	Weight only	Proportional derivative control	No	Triggered by glucose threshold	0.74mg/study	DH 133.2, control 145.8	DH:C 0.059	DH 84.7%, control 68.5%	DH:C 0.007	DH 1.3%, control 2.4%	DH:C 0.139	DH 11.9%, control 24.3%	DH:C 0.022	events: DH 12, control 21		One subject with nausea after 0.88mg glucagon, 6 glucagon infusion set occlusions
Abitbol et al.(31)	2/3/2018	Toronto, Canada	18 adults with hypo unaware, 17 adults hypo aware	45.6	26.9	7.7	11 hours overnight	Randomized, crossover	DH	SH	Inpatient	No	Enlite sensor, Minimed Veo pump, manual entry of rates	Onsite		Abitbol et al.(31)	TDD, ICR	Model predictive control	Yes	Logical rules based on glucose value and trend	Not reported	DH 122.4 (19.8), SH 142.2 (23.4)	DH:SH 0.01	DH 100%, SH 77%	DH:SH 0.04	Not reported		DH 0%, SH 17%	DH:SH 0.04	Hypo unaware: 0.38 events/night, Hypo aware 0.25 events/night		Not reported

Reducing hypoglycemia

Most studies show lower time in hypoglycemia with dual hormone than comparator groups (CSII, SAP, PLGS etc.), although not all reached statistical significance. A recent meta-analysis of seven dual hormone studies and sixteen single hormone studies with conventional pump therapy as comparator group was completed.²⁰ This meta-analysis found that the mean difference for time in hypoglycemia (<70mg/dl or <3.9 mmol/L) was reduced −3.78% (95% CI −5.58 to −1.97) for dual hormone systems compared to conventional pump. The reduction was less for single hormone systems compared with conventional pump, −1.88% (95% CI −3.40 to −0.36). In studies comparing a single hormone system directly with a dual hormone system, time <70mg/dl (3.9 mmol/L) is as low as 1% (representing about 15 minutes per day) for dual hormone systems. There is also a consistent trend towards reduced need for rescue carbohydrate intake with dual hormone systems. The effect of dual hormone control on overnight hypoglycemia in adults is less consistent. One study showed an insulin-only closed loop system and PLGS significantly reduced overnight hypoglycemia as compared to conventional pump therapy with little added benefit with the addition of glucagon.²⁸ A similar finding was seen in two other studies,^30,43 while a study in children aged 9–17 in a diabetes camp setting showed a statistically significant benefit of dual hormone over an insulin-only system and conventional pump therapy for reducing overnight hypoglycemia.⁴⁴ Longer term studies are needed to clarify whether dual hormone systems provide additional benefit over insulin-only systems for preventing overnight hypoglycemia.

Preventing exercise induced hypoglycemia

The preponderance of the evidence suggests that dual hormone closed loop systems reduce the occurrence of acute onset hypoglycemia caused by exercise. One study compared single hormone and dual hormone systems over the course of a four-hour interval during which participants were randomized to continuous aerobic or interval exercise for 60 minutes.¹² For continuous aerobic exercise, median percent time below 70mg/dl (3.9 mmol/L) was 22.5% for single hormone versus 0% for dual hormone (p=0.006). High intensity interval exercise showed median time <70mg/dl (<3.9 mmol/L) of 0% in both the single and dual hormone groups, demonstrating how various types of exercise impact glycemic control very differently. In another study, two dual hormone systems (one with exercise detection and dosing adjustment⁵³, one without) were compared to sensor augmented pump therapy.²⁹ The dual hormone system that adjusted dosing of insulin and glucagon following detection of exercise reduced time in hypoglycemia from the start of exercise until the next meal as compared to the dual hormone system with no exercise detection (0.3% vs 3.1% p= 0.001). In a follow-on study by the same group, dual hormone, single hormone, PLGS and current care (SAP or CSII alone) were compared in a four day outpatient study including 45 minutes of aerobic exercise on day 1 and 4.²⁸ The single and dual hormone systems included automated exercise detection and dosing adjustment. From the start of exercise until the next meal (approximately 4 hours), time <70mg/dl (<3.9 mmol/L) was lowest in the dual hormone group (DH 3.4%, SH 8.3%, PLGS 7.6%, current care 4.3%). Dual hormone showed lower time in hypoglycemia than SH (p=0.009) and PLGS (<0.0001), but not current care where subject driven pre-exercise insulin adjustment was allowed (p=0.49). Two other studies, showed no advantage of dual hormone over single hormone for hypoglycemia related to exercise.^30,45 Although in one of these studies the closed loop system was used by participants only during the overnight period starting a few hours after exercise.⁴⁵

Prandial glucose control

In theory, a dual hormone system could allow for more aggressive insulin dosing for carbohydrate intake to improve post-prandial hyperglycemia while glucagon prevents any postprandial hypoglycemia induced by more aggressive insulin dosing. To our knowledge this approach has not yet been tried in humans with a dual hormone system. Also, this approach does raise possible safety concerns due to possible hypoglycemia if glucagon delivery fails. Data from a related human study provides mixed results on whether this may be beneficial.⁵⁴ In this study, adolescents ages 12–18 on pump therapy were given a standardized meal on three separate occasions; they received their usual premeal insulin bolus, usual bolus plus pramlintide, or 60% increase of their usual insulin dose with rescue doses of minidose glucagon if glucoses dipped below 95mg/dl (5.3 mmol/L). The researchers saw that the 60% increased bolus did not result in statistically significant improvement in the area under the glucose curve as compared to the usual bolus, however there was a trend towards lower glucose values in the increased bolus group. Minidose glucagon injections (1–4 injections of 10mcg/ year up to 150mcg) were required around five hours after the meal and successfully raised glucose. Interestingly, the pramlintide plus insulin group showed a marked drop in the glucose in the early postprandial period, even resulting in early post-prandial hypoglycemia in some subjects. The results of the pramlintide arm of this study are further discussed further in “Other potential hormones for closed loop.” One potential limitation of using glucagon to prevent early postprandial hypoglycemia is that high circulating insulin concentrations are known to reduce the glucose raising impact of minidose glucagon.⁵⁵

Special considerations for glucagon in closed loop systems

Stable liquid glucagon formulations

Glucagon from standard hypoglycemia rescue kits is reconstituted in an aqueous solution. In this solution, glucagon forms fibrils and within days solidifies into a gel in additional to undergoing spontaneous degradation.⁴⁷ Several pharmaceutical companies are working towards stable liquid glucagon products that avoid these pitfalls.⁵⁶ There are two main approaches to improving the stability of glucagon, native human glucagon reconstituted in special carrier solutions and glucagon analogs with slight peptide alterations to promote stability and solubility. Until very recently, dual hormone studies used glucagon from a commercially available hypoglycemia rescue kit. Glucagon needed to be reconstituted and the pump reservoirs refilled every 24 hours during the studies. With several stable liquid glucagon products nearing FDA approval, this recent dual hormone study⁵⁷ used an investigational stable liquid glucagon product with results comparable with a prior study with the conventional glucagon preparation.²⁸

Continuous subcutaneous infusion pump design

All currently available CSII systems consist of a single reservoir or cartridge and a single canula for subcutaneous infusion. To date, dual hormone studies have required participants to wear two pumps and two infusion sites in addition to CGM. This is impractical for real-life use. People with T1D already struggle to find adequate locations for site and sensor placement while allowing previously used areas to heal. Some develop considerable lipohypertrophy or scar tissue from overuse of particular sites, which can impact insulin absorption kinetics if the site is used in the future.⁵⁸ Translation of dual hormone technologies to real world use will require development of pump technology with dual chambers and possibly dual cannulas, and such pumps are currently in development.⁵⁹ Even with this advancement, there will need to be careful design and training to fill and connect two chambers without inverting the medications. In practicality, this may prove to be too complex or burdensome for many individuals.

Potential limitations for use of glucagon in closed loop systems

Glycogen stores

Type 1 diabetes is associated with lower hepatic glycogen stores. Subjects with well controlled T1D had significantly lower glycogen stores than healthy volunteers after an overnight fast. ⁶⁰ Subjects with poorly controlled diabetes showed reduced glycogen synthesis and breakdown which improved but did not normalize with short term intensive diabetes control. ⁶¹ This raises the question of whether people with T1D can have sustained physiologic glucose raising responses to repeated doses of glucagon such as may be given during dual hormone control. In a study with participants with well controlled T1D, Magnetic Resonance Spectroscopy helped to answer this question. ⁶² Eight doses of glucagon (mean dose of 140mcg, total mean dose 1125mcg) dosed over 16 hours with repeated Magnetic Resonance Spectroscopy scans quantitated glycogen stores. There was a non-statistically significant trend towards a decline in glycogen stores in the fed state after glucagon dosing. In the fasting state, the baseline glycogen stores were lower, but there was a non-statistically significant increase in glycogen stores after the glucagon doses. Importantly, the impact of glucagon increasing glucose levels was maintained through the 8^th glucagon dosing. Longer term studies are needed to better understand if the glucagon response is maintained over time, but this study provides useful evidence that glycogen stores are not markedly reduced by repeated glucagon administration and the hyperglycemic response is also maintained.

Interactions between insulin and glucagon

Insulin analogs demonstrate slow absorption from the subcutaneous space and remain active in the body for 4–6 hours or more in most patients. In a person treated with insulin, hypoglycemia often occurs during a high insulin state. This reduces glucagon’s ability to counteract hypoglycemia as the ratio of insulin to glucagon within the portal venous system determines whether the liver can undergo glycogenolysis. A dual hormone study⁶³ analyzed circumstances under which glucagon dosing failed to prevent hypoglycemia. Many failures correlated with higher insulin-on-board conditions and also during sensor overestimation of glucose. This study⁵⁵ examined the effect of varying insulin levels on the endogenous glucose production rate produced in response to varying glucagon doses. Low and medium insulin infusion rates showed proportional endogenous glucose production response in relation to the glucagon dose. Whereas, under high insulin infusion, increased glucagon doses did not result in increases in endogenous glucose production.

Dietary and lifestyle considerations

The effect of glucagon on glucose is dependent on adequate hepatic glycogen stores. In normal subjects, low carbohydrate diets reduce hepatic glycogen stores.⁶⁴ In this study⁶⁵, adults eating a low carbohydrate diet (<50g/day for 1 week) versus a high carbohydrate diet (>250g/day for 1 week) showed that glucose levels increased less in response to subcutaneous doses of glucagon (100, 500mcg). Dual hormone systems may need to account for an anticipated blunted glucose response to glucagon in patients eating a lower carbohydrate diet.

Alcohol intake can lead to several factors which synergistically cause hypoglycemia in T1D. Ethanol inhibits hepatic gluconeogenesis⁶⁶, causes hypoglycemia unawareness⁶⁷ and impairs cognitive performance.⁶⁸ During ethanol intoxication, circulating glucose derives from hepatic glycogenolysis, therefore glycogen stores may be inadequate to allow for a hyperglycemic response to glucagon administration. This study⁶⁹ demonstrated a diminished hyperglycemic response to 100mcg glucagon following ethanol consumption (equivalent to 4–6 drinks) compared to placebo consumption. Encouragingly, a second subcutaneous glucagon dose of 100mcg two hours later showed a similar glucose raising profile, indicating that despite prior alcohol consumption, some degree of hyperglycemic response is still possible.

Safety of glucagon

The potential side effects of glucagon are listed in Table 2. New liquid stable glucagon products intended for long term use in subcutaneous infusion pump systems will need to go through thorough evaluation for safety and efficacy. Before these trials are done in humans, animal studies will be needed to show initial proof of safety. A recent animal study in rats and dogs with a glucagon analog showed promising safety results.⁸¹ Animals were dosed subcutaneously for 26 weeks (rats) or 39 weeks (dogs) with 10-fold varying doses of the glucagon product. At the equivalent dosing level of 1mg/day in humans, which is equivalent to current dual hormone studies, no unexpected adverse effects were seen in the animals. There was an increase in liver weight, glycogen vacuole formulation in the liver and liver function marker increases during the study, however this was believed to be due to the hyperinsulinism induced by the glucagon in these animals who did not have diabetes. We await results from long term animal and human safety studies for other stable liquid glucagon formulations in development.

Table 2.

Possible Side Effects of Glucagon

Site	Impact	Mechanism	Comments
Gastrointestinal	Nausea and vomiting.	Inhibition of gastric motility.	Common side effects to the 1 mg hypoglycemia rescue dose of glucagon, but much less common at the micro-dosing level used in dual-hormone systems.
Metabolism	Potential decrease in plasma triglyceride and cholesterol levels. Administration of high dose increases resting metabolic rate.	Promotes triglyceride lipolysis to produce free fatty acids and hepatic fatty acid oxidation for fuel substrates.70	In short-term human studies: administration of exogenous glucagon appears to have little effect on plasma lipid levels in healthy subjects or in subjects with T1D.71,72
			In a rat model: administration of glucagon over 21 days decreased plasma triglyceride and cholesterol levels with no change in liver fat.73
			In healthy subjects, glucagon infusion, along with a somatostatin infusion to inhibit insulin secretion, increased the resting metabolic rate by 15%.74
Cardiovascular	Administration of high dose glucagon can induce small increases in mean arterial pressure and heart rate without major effects of systemic vascular resistance.75	At doses of >1mg of glucagon, glucagon acts directly on cardiac tissues with chronotropic and inotropic effects via stimulation of catecholamines.	Older animal studies showed a detrimental effect of high dose glucagon on ischemic myocardium.76,77
			Human studies showed benefit or no effect of high dose glucagon in myocardial ischemia.78,79
Central nervous	Administration of high dose glucagon elicits feeling of satiety.	Glucagon crosses the blood brain barrier effecting the vagal system.	In one study, pre-prandial 1mg IM doses of glucagon over two weeks in healthy adults resulted in a reduction in intake of 440 calories per day and an average weight loss of 0.45 lbs compared to weight gain of 3.4 lbs in the placebo arm.80
Urinary	Increased natriuresis.	Possibly mediated by renal arterial vasodilation with an increase in renal blood flow.70
Respiratory	Pulmonary bronchodilation.	Smooth muscle relaxation.

Abbreviation: IM, intramuscular.

Data from Refs. 70–80

Other potential hormones for closed loop

A few other hormones are under consideration for adjunctive use with AID systems. Pramlintide is the synthetic form of amylin which is co-secreted from beta-cells in a ratio with insulin of 100:1 (insulin:amylin).⁸² Amylin secretion is lost along with insulin secretion in T1D. Pramlintide is effective in reducing post-prandial glucose spikes presumably by slowing gastric emptying. Over the course of a one year trial in T1D with administration before each meal, pramlintide reduced A1c by 0.2% to 0.6%.⁸³ Fixed premeal doses of 30mcg SQ pramlintide used in conjunction with an insulin-only closed loop system requiring meal announcement resulted in improved post-prandial glucose levels and reduced area under the curve for the glucose excursions.⁸⁴ Other studies have shown improved post-prandial control with adjunctive use of liraglutide (glucagon-like-peptide-1 receptor agonist) as a once daily SQ dose.^85,86 Most recently, a dual hormone system delivering a fixed ration of pramlintide to insulin (6mcg/unit) significantly improved time in target range (70–180mg/dl or 3.9–10 mmol/L) and glucose variability as compared to a single hormone closed loop system.⁸⁷

Conclusions

The inclusion of glucagon in dual hormone closed loop systems appears to be effective in reducing overall hypoglycemia and hypoglycemia related to exercise. Additional clinical trials are needed to assess if these benefits are persistent and the side effect profile remains favorable in longer term studies. There are technical and pharmaceutical advances that will be needed in glucagon formulations and pump technologies before this approach can be realistically feasible in free living conditions. Dual hormone closed loop control will likely not become standard of care for all people with type 1 diabetes. However, development and research should continue as it is likely to be an important option for those who suffer from recurrent hypoglycemia while using single hormone closed loop systems and in those who are physically active with recurrent hypoglycemia.

Key Points.

1. In type 1 diabetes, impaired glucagon responses contribute to ineffective counter-regulation which increases the risk of hypoglycemia.

2. Dual hormone closed-loop systems give microgram sized doses of glucagon under the control of a dosing algorithm.

3. As compared to insulin-only automated insulin delivery (AID) systems, dual hormone closed loop systems have been shown to reduce overall time in hypoglycemia and reduce time in hypoglycemia related to exercise.

4. Dual hormone closed loop systems will require a stable liquid glucagon product, several of which are under development by pharmaceutical companies.

5. Long term studies are needed to establish safety and tolerability of chronic use of low dose glucagon. Preliminary studies in humans and animals show favorable safety profiles.

Synopsis.

Treatment of type 1 diabetes with exogenous insulin oftentimes results in unpredictable daily glucose variability and hypoglycemia which can be dangerous. Automated insulin delivery systems can improve glucose control while reducing burden for people with diabetes. One approach to improve treatment outcomes is to incorporate the counterregulatory hormone glucagon into the automated delivery system to help prevent hypoglycemia that can be induced by the slow pharmacodynamics of insulin action. This review explores the advantages and disadvantages of incorporating glucagon into dual hormone automated hormone delivery systems.