New closed-loop insulin systems

Hybrid closed-loop insulin delivery systems are gradually transforming clinical management of type 1 diabetes. They comprise a subcutaneously worn continuous glucose monitor (CGM or glucose sensor) device, communicating with an algorithm that responds in real time to changes in sensor glucose levels, and adjusts the subcutaneous insulin infusion delivered by an insulin pump (Fig. 1).

Although the concept of glucose-responsive insulin delivery has been around for 50 years, earlier developments of closed-loop systems were hampered by the lack of accurate and reliable CGM systems, unavailability of wearable computational devices and secure wireless communication protocols, and challenged by limitations in insulin pump devices. With advances in CGM technology, the simplest form of automated insulin delivery was achieved with low-glucose suspend systems, where insulin delivery is suspended when the sensor glucose crosses a specified threshold, and predictive glucose management systems, where insulin delivery is suspended when an algorithm predicts that sensor glucose is likely to cross the low-glucose threshold [1, 2]. These systems reduce hypoglycaemia, although sometimes at the expense of increased hyperglycaemia [3]. These were conceptually landmark steps in the journey towards fully automated insulin delivery and a true ‘artificial pancreas’ (Fig. 2).

A closed-loop system is a more sophisticated system, with a control algorithm adjusting insulin delivery (up and down) in response to real-time sensor glucose levels and other inputs, such as meal intake (Fig. 3). The algorithm can accommodate variability of insulin requirements between and within individual users, and account for limitations of CGM accuracy and imprecisions of subcutaneous insulin delivery. Adaptation of the control algorithm to changes in physiological conditions with real-time adjustment of closed-loop control parameters is beneficial for optimal performance. Several different types of control algorithm have been developed, including model predictive control (MPC) algorithms, proportional integral derivative (PID) controllers and fuzzy logic control approaches [4]. MPC algorithms calculate insulin delivery by minimising the difference between model-predicted glucose concentrations and target glucose over a pre-specified prediction time horizon. PID controllers adjust insulin delivery by assessing glucose excursions from three perspectives: (1) deviation from target glucose (proportional component); (2) area under the curve between measured and target glucose (integral component); and (3) rate of change of measured glucose (derivative component). The fuzzy logic approach modulates insulin delivery based on approximate rules to express empirical knowledge of diabetes practitioners.

The first commercial closed-loop system, the MiniMed 670G (Medtronic, Northridge, CA, USA) was approved by the US Food and Drug Administration in September 2016 for use in people with type 1 diabetes aged 14 years and older [5]. This hybrid closed-loop system requires users to manually enter prandial insulin with automation of insulin delivery between meals and overnight. Several other hybrid closed-loop systems have since been commercialised and are increasingly being utilised in routine clinical care for people with type 1 diabetes.

Clinical studies evaluating the safety and efficacy of hybrid closed-loop systems have evolved from small, highly supervised studies, undertaken overnight or over 24 h in research facilities, to larger randomised controlled trials of unrestricted home living use, conducted over 6 months or longer. A meta-analysis, published in 2018, of 40 early outpatient studies, reports the efficacy and safety of hybrid closed-loop systems in people with type 1 diabetes [6]. Earlier closed-loop systems were associated with a 9.6 percentage point improvement in time in target glucose range (3.9–10.0 mmol/l) compared with comparator therapies (>2 additional h/day) and reduced time in hypoglycaemia (<3.9 mmol/l) by 1.5 percentage points (approximately 20 min/day) compared with control treatment. Hybrid closed-loop systems have a favourable effect on HbA_1c, with a reduction of 0.3–0.4% compared with control therapy in studies with a duration of more than 8 weeks per intervention. While this effect appears modest, this is despite the reduction in hypoglycaemia observed in several of these studies, as well as the low HbA_1c at recruitment, reflecting good baseline glycaemic control of study participants. Similar benefits have been reported in a meta-analysis of 25 studies in the paediatric population [7].

Individual randomised controlled trials demonstrate glycaemic benefits of hybrid closed-loop systems vs comparator therapies, but comparisons of efficacy between hybrid closed-loop systems across different studies is hampered by variation in baseline characteristics of participants, study duration and design. Studies including participants with variable experience in diabetes technology use (on multiple daily insulin injections and without previous sensor use) and from more diverse socioeconomic backgrounds are important to support generalisability of benefits.

The impact of closed-loop technology on quality-of-life measures has been explored in several studies [8]. Psychosocial benefits reported by users include reduced anxiety, improved sleep and confidence from improved overnight glucose control, less restrictive eating habits and ‘time off’ from the demands of diabetes management. Reported challenges include technical issues, alarm intrusiveness and equipment burden, in addition to initial difficulties trusting the system. Most participants in closed-loop studies report that they would continue using closed-loop therapy or would recommend it to others as the clinical benefits outweigh system shortcomings. Psychosocial studies have largely included participants involved in closed-loop trials who may not be representative of the wider population living with type 1 diabetes.

Details of the commercially available hybrid closed-loop systems can be found in Table 1 and an overview of key clinical studies are shown in Table 2.

The Diabeloop hybrid closed-loop system, comprising a handset containing the algorithm, Kaleido pump and Dexcom G6 sensor (DBLG1; Diabeloop, Grenoble, France), has been compared with sensor-augmented pump therapy in 68 adults with type 1 diabetes, in the home setting, with remote monitoring over 12 weeks [18]. Time with glucose in target range was 9 percentage points greater with closed-loop therapy than with control therapy (69% vs 59%) and time in hypoglycaemia was significantly lower with closed-loop use than during the control period. The DBLG1 hybrid closed-loop system has received the CE mark for use in adults with type 1 diabetes and is due to be commercialised.

The Insulet Omnipod Horizon hybrid closed-loop system, comprising the Omnipod pump and Dexcom G6 sensor (Insulet, Billerica, MA, USA), and Beta Bionics insulin-only iLet hybrid closed-loop system, comprising the iLet Bionic Pancreas System and Dexcom G6 sensor (Beta Bionics, Boston, MA, USA), are both currently being evaluated in pivotal trials with launches anticipated in the next 1–2 years.

The do-it-yourself (DIY) artificial pancreas system (DIYAPS) communities arose out of their frustration with the slow progress of medical device development cycles (the #wearenotwaiting movement). The communities develop and apply open-access closed-loop systems (e.g. the Open Artificial Pancreas System [OpenAPS], Loop and AndroidAPS), which do not undergo regulatory overview and approval. Without needing to await regulatory approval for new developments, these systems benefit from more rapid innovation cycles and can be more flexible in terms of customisation. In principle, access is open to anyone but users must be able to build and maintain their own system, albeit with support from the community itself. The role of healthcare providers in supporting the use of unregulated systems continues to be debated.

Several thousands of individuals use DIY systems globally. Observational before-and-after studies show improvements in time in target glucose range, HbA_1c and quality of life, but no longitudinal randomised controlled trials have evaluated the efficacy and safety of these systems [19]. A randomised clinical trial of a version of AndroidAPS is underway (ACTRN12620000034932p).

The physiological glucagon response to hypoglycaemia is often impaired in type 1 diabetes; therefore, addition of glucagon to a closed-loop system confers additional protection from hypoglycaemia and may allow more aggressive insulin delivery to achieve improved glucose control. Potential benefits are countered by increased system complexity, requirement for two separate infusion systems and the lack of approved room-temperature-stable glucagon for chronic subcutaneous delivery. There are currently no commercially available dual-hormone closed-loop systems, although several are in development.

The longest dual-hormone closed-loop home study, with remote monitoring, included 43 adults with type 1 diabetes with optional meal announcements over a period of 11 days. Time in target glucose range was increased (78% vs 62%) and hypoglycaemia (<3.3 mmol/l) was reduced (0.6% vs 1.9%) with dual-hormone closed-loop use compared with insulin pump therapy alone [20]. A shorter study, over 5 days, involving 32 adolescents with type 1 diabetes, demonstrated a 21 percentage point increase in time in target glucose range during the closed-loop period compared with the control period, but time in hypoglycaemia was similar between groups [21]. An outpatient study of over 60 h compared dual-hormone with single-hormone closed-loop systems in 23 adults with type 1 diabetes and showed no difference in time in target glucose range (79% vs 75%) or time in hypoglycaemia (<4.0 mmol/l; 3.6% vs 3.9%), but longer studies are required to fully investigate potential differences [22].

Pramlintide is an analogue of amylin, which is co-secreted with insulin from beta cells and reduces postprandial glucose excursions by slowing gastric emptying [23]. A novel dual-hormone closed-loop system delivering a fixed ratio of pramlintide:insulin was evaluated during a 24 h inpatient study in adults with type 1 diabetes. The dual-hormone system improved time in target range compared with an insulin-alone system (84% vs 74%), an effect attributable to improved daytime glucose control [24]. Gastrointestinal symptoms were reported more frequently during use of closed-loop systems with pramlintide as compared with insulin only. Pramlintide co-delivery may support the development of fully closed-loop systems, obviating the need for manually initiated prandial insulin delivery.

High quality user and healthcare professional training is essential for ensuring that the clinical benefits of hybrid closed-loop systems are realised in the real-world setting. This is an important consideration for health economic analyses, to support adoption, implementation and reimbursement. Establishing realistic expectations of hybrid closed-loop therapy and reiterating the importance of core diabetes skills and tasks is important to promote long-term use and optimal clinical outcomes. Training programmes have been developed, using online and face-to-face approaches, to support users to maximise glycaemic and quality-of-life benefits of closed-loop therapy [25].

Early real-world use of the first commercially approved hybrid closed-loop system (Medtronic 670G) exposed issues around usability. The system required significant user input to remain in Auto Mode and, in one prospective observational study, one-third of users discontinued use of Auto Mode during the first year after initiation [26]. Factors influencing discontinuation include CGM issues (calibrations), number of alarms and efforts to limit Auto Mode exits [27]. Usability issues can prevent realisation of the benefits of closed-loop systems as increased time in Auto Mode is associated with improved glycaemic outcomes [28].

Some first-generation hybrid closed-loop systems use a relatively high glucose target (6.7 mmol/l) and lack flexibility to adjust the target to suit the needs of the user. This makes the system unsuitable for those aiming for tighter glycaemic control, including pregnant women.

Postprandial glucose excursions remain a challenge for closed-loop systems due to inherent delays in subcutaneous insulin absorption. User interaction with accurate carbohydrate counting and pre-meal bolusing is required for optimal glycaemic control. Attempts to reduce user burden with simplified meal boluses or fully closed-loop systems have resulted in compromised glycaemic control [29, 30].

Managing physical activity can be challenging primarily due to increased hypoglycaemia risk and altered insulin sensitivity. Even with closed-loop glucose-responsive insulin delivery, users usually need to plan for exercise, announcing exercise to the algorithm in advance, and may still require carbohydrate intake to prevent hypoglycaemia [31, 32]. Carbohydrate loading before exercise can be problematic with glucose-responsive insulin delivery, often resulting in hypoglycaemia during exercise.

Important ethical considerations include ensuring equitable access, training and support for closed-loop technology, and protecting user confidentiality and safety from security breaches [33].

Future closed-loop systems will benefit from improved individual components; smaller, more accurate CGM devices with longer wear-time and smaller insulin pumps with the user interface transferred to a smartphone/watch will improve usability and minimise device burden. Interoperable devices and data management platforms offer flexibility for users to create their own personalised closed-loop ecosystem.

The introduction of new faster-acting insulin analogues (Fiasp and Ultra-Rapid Lispro) provides an opportunity to potentially improve performance of closed-loop systems with faster onset and offset of insulin action following subcutaneous delivery. Short studies comparing faster-acting insulin with standard insulin using the Medtronic 670G system have not shown significant benefits, but longer studies are required to fully evaluate closed-loop systems with faster-acting insulin [34, 35]. Faster-acting insulins are not yet approved for use in the t:slim X2 pump and, hence, the Control-IQ closed-loop system.

Integration of additional signals to algorithms, such as heart rate or accelerometery, to more quickly detect physical activity than with CGM alone, may reduce exercise-related hypoglycaemia. If efficacious, this would be particularly beneficial in young children in whom activity is usually spontaneous and unpredictable and hypoglycaemia is a major concern.

The last 5 years has seen the successful transition of closed-loop systems from research into routine clinical practice for management of type 1 diabetes. There is still scope for further improvements to optimise postprandial glucose control, exercise management and usability before this technology can be said to truly ameliorate the burden of diabetes. Widespread adoption and reimbursement of closed-loop systems will be critical to ensuring equitable access to this technology.