Where Do We Stand with Closed-Loop Systems and Their Challenges?

Abstract

Treatments for type 1 diabetes have advanced significantly over recent years. There are now multiple hybrid closed-loop systems commercially available and additional systems are in development. Challenges remain, however. This review outlines the recent advances in closed-loop systems and outlines the remaining challenges, including post-prandial hyperglycemia and exercise-related dysglycemia.

Introduction

Type 1 diabetes mellitus presents a significant lifestyle and financial burden to many individuals worldwide. The 2016–2018 T1D exchange registry data indicate that a minority of people living with type 1 diabetes are reaching the HbA1c goals recommended by the American Diabetes Association.¹ Systems that automate insulin delivery using continuous glucose monitoring (CGM) and insulin pump therapy, termed closed-loop systems, have the potential to both reduce the burden of diabetes management and improve glucose outcomes. The progress toward this goal in the past decade has been tremendous, but challenges still remain.

Advances in Delivery Systems

The MiniMed 670G system was the first commercialized hybrid closed-loop system. In this context, the term hybrid indicates that the system is designed to give boluses before meals. With the MiniMed 670G system, the bolus amount is determined by the user entering in the amount of carbohydrate to be consumed. The MiniMed 670G received Food and Drug Administration (FDA) approval in September 2016, with subsequent approval for use in children 7–14 years of age in June 2018. In a pivotal study of 30 adolescents and 94 adults, use of the MiniMed 670G system automated insulin delivery that targets a glucose of 120 mg/dL, termed Auto Mode, improved HbA1c and time in range (70–180 mg/dL), and reduced hypoglycemia (<70 mg/dL) as compared with a 2-week run-in period.²

See Table 1, which compares outcomes of the pivotal MiniMed 670G study in adults with other recently published closed-loop studies conducted over 3 months or longer. It is notable that this study was not randomized and was designed to primarily assess safety rather than efficacy. A safety study of 105 children ages 7–13 years demonstrated a similar improvement in HbA1c and time in range with the use of the MiniMed 670G system (Table 1).³ A 6-month multi-center, randomized trial assessing the efficacy of the MiniMed 670G was started in May 2017 (NCT02748018).

Table 1.

Summary of Recently Published Closed-Loop Clinical Trials of 3 Months' Duration and Longer

Closed-loop system	Population	Comparison	Primary outcome	Severe hypoglycemia/DKA
Control-IQ7	168 people with type 1 diabetes (ages 14–71 years) randomized, 112 to Control-IQ and 56 to sensor-augmented pump.	Randomized trial comparing Control-IQ with sensor-augmented pump therapy over 6 months.	Percentage of time spent with glucose 70–180 mg/dL mean (SD): 71% (12) with Control-IQ vs. 59% (14) with SAP (P < 0.0001).	No severe hypoglycemia events occurred. DKA occurred in 1 in the closed-loop group due to infusion set failure.
DBLG114	68 adults with type 1 diabetes randomized, 63 completed both periods.	Crossover trial comparing DBLG1 with sensor-augmented pump therapy, each period 12 weeks.	Percentage of time spent with glucose 70–180 mg/dL mean (SD): 69% (9) with DBLG1 vs. 59% (10) with SAP (P < 0.0001).	5 severe hypoglycemic events occurred in the DBLG1 group and 3 in the sensor-augmented pump group. No DKA events occurred.
FlorenceM10	86 people with type 1 diabetes (ages 6 years and older) randomized, 46 to FlorenceM and 40 to sensor-augmented pump therapy.	Randomized trial comparing FlorenceM with sensor-augmented pump therapy over 12 weeks.	Percentage of time spent with glucose 70–180 mg/dL mean (SD): 65% (8) with hybrid-closed loop vs. 54% (9) with SAP (P < 0.0001).	No severe hypoglycemia occurred. DKA occurred in 1 in the closed-loop group due to infusion set failure.
MiniMed 670G2	124 people with type 1 diabetes (ages 14–75 years).	2-week run-in phase with sensor-augmented pump therapy compared with 3-month use of 670G in Auto Mode.	Safety of 670G system use: 28 device-related adverse events occurred that resolved at home. There were 4 serious adverse events (appendicitis, bacterial arthritis, worsening rheumatoid arthritis, Clostridium difficile diarrhea).	No severe hypoglycemia or DKA events occurred.
MiniMed 670G3	105 children with type 1 diabetes (ages 7–13 years).	2-week run-in phase with sensor-augmented pump therapy compared with 3-month use of 670G in Auto Mode.	Change in HbA1c level: 7.9% (0.8) at baseline vs. 7.5% (0.6) after 3 months of use (P < 0.001).	No severe hypoglycemia or DKA events occurred.

DBLG1, Diabeloop Generation 1; DKA, diabetic ketoacidosis; SD, standard deviation.

Real-world data have identified challenges with use of the MiniMed 670G outside of clinical trials. This is likely due, in part, to issues with sensor calibrations and errors as well as forced exits from Auto Mode, which may be decreased in a more structured setting. In a retrospective cross-sectional analysis of 96 pediatric and young adult patients living with type 1 diabetes, the median time in Auto Mode was only 38.5% [interquartile range 0%–64%].⁴ The analysis did indicate that increased time in Auto Mode was associated with significantly lower HbA1c levels as well as an increased time in range.

Medtronic has developed a next-generation closed-loop system that includes automated correction boluses and a lower target glucose level. In a feasibility study of 12 adults using this prototype system for 4 weeks, time in range was significantly improved (85.3% [interquartile range 79.4%–88.4%] vs. 75.0% [66.6%–83.7%]) and mean sensor glucose was reduced (123.0 mg/dL [119.3–129.6] vs. 143.5 mg/L [135.8–154.5]), though with higher time in hypoglycemia (4.4% [3.3%–6.1%] vs. 3.0% [1.8%–3.8%]) as compared with a 1-week run-in phase using an open-loop system.⁵ Time in closed-loop during this study was very high at 99.98%. A study of 250 participants with type 1 diabetes is currently underway, testing an advanced hybrid closed-loop system funded by Medtronic (NCT03959423).

Control-IQ received FDA approval on December 13, 2019, making it the second commercially available hybrid closed-loop system in the United States. The Control-IQ system consists of a Dexcom G6 sensor and a t:slim X2 insulin pump running a closed-loop insulin delivery algorithm. This algorithm⁶ uses the programmed insulin pump settings but also automates correction boluses if sensor glucose is predicted to exceed 180 mg/dL and increases basal insulin delivery if sensor glucose is predicted to exceed 160 mg/dL. The system decreases basal insulin delivery if sensor glucose is predicted to fall below 112.5 mg/dL and stops basal insulin delivery if sensor glucose is predicted to fall below 70 mg/dL. In a randomized multi-center trial of 48 adolescents and 120 adults,⁷ use of Control-IQ increased time in range by 11% points (95% confidence interval [CI] 9–14). Use of Control-IQ also resulted in a reduced time in hypoglycemia (<70 mg/dL) as compared with sensor-augmented pump therapy by 0.88% points (95% CI −1.19 to −0.57). Use of Control-IQ also reduced HbA1c by 0.33% points (95% CI −0.53 to −0.13).

The pivotal Control-IQ study was, in part, funded by the National Institutes of Health (NIH), which has funded three other pivotal closed-loop trials.⁸ One of these pivotal trials is a multi-center randomized trial of two hybrid closed-loop systems, the FlorenceM, which is being studied in the United States, and CamAPS FX, which is being studied in the United Kingdom (NCT02925299). The FlorenceM system consists of a Medtronic 640G pump and Guardian 3 CGM, and the CamAPS FX consists of a DANA RS insulin pump and Dexcom G6 CGM. These two systems include a model predictive control algorithm that has been extensively tested⁹ and shown to improve time in range and reduce hypoglycemia over a 12-week period in children and adults with type 1 diabetes (Table 1).¹⁰ In 2020, the CamAPS FX app received CE mark in the European Union and United Kingdom. Another NIH-funded pivotal study is testing an advanced hybrid closed-loop system on the MiniMed 670G platform with a Fuzzy Logic algorithm (NCT03040414). This Fuzzy Logic algorithm has also been extensively tested and was shown to lower mean overnight glucose levels and reduce hypoglycemia as compared with sensor-augmented insulin pump therapy in children attending diabetes camp¹¹; more recently, it has been shown to improve time in range with home use over a weekend.¹²

The Omnipod Horizon system is currently in development by Insulet for commercialization (NCT04196140). The recently approved DASH system, which consists of a locked-down Android phone, took Insulet one step closer toward commercializing a hybrid closed-loop system. The Horizon system consists of the Omnipod patch pump, Dexcom G6 CGM, and a model predictive control algorithm. Multiple feasibility studies have shown that this system performed well in supervised free-living conditions with time in range (70–180 mg/dL) of above 70%.^13,14 A hybrid closed-loop system from Diabeloop received a CE mark in 2018. In a multi-center randomized crossover trial of 68 adults, the Diabeloop Generation 1 improved time in target range with a mean difference of 9.2% (95% CI 6.4–11.9), with five severe hypoglycemic episodes in the Diabeloop group and three episodes in the sensor-augmented pump group (Table 1).¹⁵ Multiple other companies are developing and testing novel hybrid closed-loop systems, including Lilly and BigFoot Biomedical.

Some people living with type 1 diabetes have not waited for companies to commercialize a system to fit their needs, instead opting for a do-it-yourself (DIY) system.¹⁶ These DIY systems, which are not currently FDA approved, use open-source phone applications such as Loop and Open Artificial Pancreas System (OpenAPS). As of September 2019, the OpenAPS web page cited 1569 users worldwide.¹⁷ Users may elect to use a DIY system over a commercial system, as these DIY systems typically have rapid innovation cycles and allow for more customization. Tidepool, a nonprofit group, is currently conducting an observational study of more than 1000 Loop users with the ultimate goal of delivering Loop as a supported, FDA-regulated mobile application (NCT03838900). Tidepool has partnered with Insulet, Medtronic, and Dexcom with the expectation of supporting future versions of Omnipod, MiniMed pumps, and Dexcom CGM.¹⁸

Drug Advances Within Closed-Loop Systems

Among the greatest obstacles that still exist with these insulin delivery systems are the delayed onset and offset of insulin delivered subcutaneously. In theory, a truly ultra-rapid-acting dose of insulin could reduce the burden of mealtime announcements and significantly reduce the risk of hypoglycemia due to increased activity. Fiasp insulin has been shown to have doubled the glucose-lowering activity in the first 30 min after delivery of a bolus via an insulin pump as compared with aspart insulin.¹⁹ Use of Fiasp in a closed-loop system, however, found no significant changes in glucose control as compared with aspart insulin.²⁰ It is not yet known whether making algorithm adjustments to account for the faster action of Fiasp will result in better control. There are also other ultra-rapid-acting insulin formulations in development, such as BioChaperone lispro, that have yet to be tested in a closed-loop system.²¹

The major target to reduce morbidity and mortality in type 1 diabetes is to restore insulin. There are additional pancreatic defects, however, as autoimmune beta cell destruction also affects amylin, and there is a variable dysfunction of glucagon secretion over time.²² These deficiencies result in abnormal physiological handling of meals, resulting in postprandial hyperglycemia. In normal physiology, amylin is co-secreted with insulin, assisting with satiety and slowing gastric emptying. Pramlintide injections before meals in the context of closed-loop system has been shown to reduce postprandial excursions.^23,24 Glucagon-like peptide-1 agonists such as liraglutide have a similar effect.²⁵ Co-infusion of pramlintide with insulin removes the burden of additional injections and may reduce the gastrointestinal side effects commonly seen with pramlintide injections.²⁶ In a closed-loop study, the addition of the automated infusion of pramlintide significantly improved time in range without increasing hypoglycemia.²⁷ A subsequent study is underway in investigating triple-hormone infusion with glucagon, pramlintide, and insulin (NCT03800875).

The loss of normal glucose counter-regulation in type 1 diabetes significantly increases a person's risk of hypoglycemia. Insulin pump systems with predictive low glucose suspend features significantly reduce, but do not eliminate, hypoglycemia.^28,29 Closed-loop systems have the additional advantage of frequent adjustment of insulin delivery rates to further reduce hypoglycemia. In a meta-analysis of outpatient closed-loop studies, the use of a closed-loop system reduced time in hypoglycemia by a mean of 1.49% (95% CI −1.86 to −1.11).³⁰ The addition of automated glucagon delivery to automated insulin delivery has been studied by multiple groups as a means to further reduce hypoglycemia.^31–34 Over 5 days of use, a dual-hormone closed-loop system significantly reduced mean glucose levels with no significant change in hypoglycemia as compared with insulin pump therapy.³² A second study with 11 days of use also showed a reduction in mean glucose levels and a reduction in hypoglycemia but with an increase in nausea.³⁵ This dual-hormone system, termed iLet and developed by Beta Bionics, is initialized only by inputting the user's weight. The iLet is also poised to be tested in an NIH-funded pivotal trial.

One of the barriers to a dual-hormone closed-loop system has been the need for a stable glucagon formulation. In 2019, the FDA approved Gvoke prefilled syringe and Gvoke HypoPen. These devices are filled with liquid stable glucagon. They are only approved for the treatment of severe hypoglycemia, but their approval may pave the way for future approvals, including for automated glucagon delivery. Other liquid stable glucagon formulations are under development, including dasiglucagon, a glucagon analog, and BioChaperone glucagon.³⁶ Long-term glucagon delivery has been shown to be safe in animals,³⁷ but more data are needed to support the safety of this approach in humans. Other potential issues with a dual-hormone approach include the additional cost of glucagon and the need for a second insertion site for glucagon delivery.

The Challenge of Exercise and Meals

Closed-loop systems have made the biggest gains in improving nocturnal glucose control. Daytime dysglycemia has been much harder to impact due to the effects of exercise and meals. Exercise stimulates a large influx of glucose into muscles and also increases insulin sensitivity. This dramatic change in insulin sensitivity, combined with the delayed offset of subcutaneous insulin, increases the risk of hypoglycemia even with the use of a closed-loop system. The MiniMed 670G allows the user to temporarily raise the glucose target to 150 mg/dL for exercise, but this often needs to be done 1–2 h in advance of exercise and possibly in combination with a snack to avoid hypoglycemia.

In a study by Breton et al., a closed-loop system significantly improved time in range in adolescents in a ski camp as compared with sensor-augmented pump therapy, but both groups still required multiple carbohydrate treatments with this prolonged exercise.³⁸ Forlenza et al. tested two closed-loop strategies to reduce exercise-related hypoglycemia.¹³ One strategy reduced the basal insulin rate by 50% starting 90 min before exercise. The other strategy raised the glucose target from 130 to 150 mg/dL starting 90 min before exercise. Time in hypoglycemia was low with both strategies, although some participants required carbohydrate either before exercise to prevent hypoglycemia or afterward to treat hypoglycemia.

Dual-hormone insulin and glucagon systems have also been used to reduce the need for carbohydrate supplementation to prevent exercise-related hypoglycemia. Our group demonstrated that a dual-hormone system significantly reduced exercise-related hypoglycemia as compared with a single-hormone system (3.4% vs. 8.3% time in hypoglycemia).³⁹ Both the dual-hormone and single-hormone systems automatically detected and responded to aerobic exercise. Exercise was detected by using accelerometry and heart rate data to estimate metabolic expenditure.⁴⁰ This approach is best designed for activities such as jogging where movement is easily detected in combination with an increase in heartrate as compared with activities such as cycling where the movement is more confined to leg movement. It is also well known that the glycemic response to anaerobic exercise is different than aerobic exercise, and there are likely multiple factors that impact glycemic response, including muscle mass, insulin-on-board, intensity, and duration of exercise.^41,42 A consensus statement⁴³ has been published with guidance on how to best manage glucose levels in exercise, but these recommendations have yet to be fully tested or implemented within the context of a closed-loop system. In addition, in the approach to glycemic control during exercise, these devices must account for whether exercise will be detected or manually entered by the user.

Meals are another challenge presented by the delayed time to onset of subcutaneous insulin. Closed-loop systems have traditionally been hybrid closed-loop systems, where some insulin is given ahead of a meal. This has been shown to reduce postprandial hyperglycemia as compared with fully closed-loop insulin delivery.⁴⁴ Postprandial hyperglycemia can subsequently lead to late postprandial hypoglycemia, as the closed-loop algorithm ramps up insulin delivery in response to the hyperglycemia. One of the most common reasons for hyperglycemia is missed mealtime insulin, and the difficulty with managing mealtime insulin is a significant contributor to the burden of living with type 1 diabetes.⁴⁵ To reduce the burden of carbohydrate counting, the iLet, for example, has users give a general estimate of meal size rather than requiring an estimate of the number of grams of carbohydrate. One potential concern with using meal size estimates is that it may cause postprandial hyperglycemia and late postprandial hypoglycemia if significantly less insulin is given before the meal than would have been delivered if a user was entering in an estimate of carbohydrate content. This is of particular concern for fully closed-loop systems that do not include any type of meal boluses. Multiple models have been developed to detect missed mealtime insulin doses.^46,47 The use of such models may be leveraged in the future to better avoid postprandial hyperglycemia, particularly if these models can be used with ultra-rapid-acting insulin in the future.

Special Populations and Future Directions

Populations that will require additional considerations include those with chronic kidney disease, which affects 20% of people with diabetes in the most recent National Health and Nutrition Examination study,⁴⁸ and can significantly impact insulin clearance. Ongoing studies are examining whether or not closed-loop system will be able to better address the needs of this population. One such study has been published already, with a small inpatient trial showing improved time in range in people with type 2 diabetes and end-stage renal disease on hemodialysis with a fully closed-loop insulin delivery system.⁴⁹ Pregnant women are another group that could benefit from special attention and recommendations, with a study in 2016 demonstrating improved time in range with use of a closed-loop system in pregnant women with type 1 diabetes compared with sensor-augmented therapy.⁵⁰ Individuals with hypoglycemia unawareness are at particular risk of severe hypoglycemia and to date there have been a limited number of closed-loop trials in this high-risk population.⁵¹ Adolescents have been included in a number of closed-loop trials, but there are limited data in young children, who typically have a harder time announcing changes in activity or food intake to a hybrid system.⁵²

Closed-loop systems may in the future play a significant role in the management of diabetes in the inpatient setting. One study showed significant improvement in time in range with the use of a fully closed-loop system in inpatients with type 2 diabetes receiving noncritical care.⁵³ Patients who receive nutritional support in the hospital present an extra challenge. One study demonstrated a significant improvement in time in range in people with type 2 diabetes or stress-related hyperglycemia receiving nutritional support in the hospital with the use of a fully closed-loop system.⁵⁴ The ability to rapidly change the level of insulin delivery without subjecting a patient to treatment with intravenous insulin would be a tremendous improvement in inpatient diabetes care.

Closed-loop systems are ideally designed to improve glucose outcomes and reduce burden of diabetes care. As the field has evolved, there has become a better understanding of the importance of the user experience as well as the psychosocial issues with long-term closed-loop system use. As recently reviewed by Farrington,⁵⁵ there are both benefits and burdens with closed-loop system use and these psychosocial issues need to be better assessed in a more consistent manner and across a more diverse group of users.

Conclusions

Closed-loop systems are set to revolutionize the management of type 1 diabetes. Many clinical trials have shown that the majority of people using these systems in clinical trials can achieve a time in range of above 70% with a low time in hypoglycemia. With the designations of interoperable CGM and alternate controller enabled pumps, the FDA has laid the groundwork to allow for system interoperability, which ideally will enable users to choose which CGM system, pump system, and algorithm best fits their needs. Additional work is needed to reduce the burden of diabetes management, particularly for handling of meals and exercise. There also needs to be a continued movement toward making these systems more customizable to fit the needs of each individual person living with diabetes.

Author Disclosure Statement

M.J. has no relevant financial interests. J.R.C. has a financial interest in Pacific Diabetes Technologies, Inc., a company that may have a commercial interest in the results of this type of research and technology. J.R.C. has also consulted for Dexcom, Inc. and has been on advisory panels for Zealand Pharma and Novo Nordisk.

Funding Information

The time to prepare this article was supported by grant 1R01 DK120367 from NIH/NIDDK.