A review of artificial pancreas technologies with an emphasis on bi-hormonal therapy

Authors


Abstract

Since the discovery of insulin, great progress has been made to improve the accuracy and safety of automated insulin delivery systems to help patients with type 1 diabetes achieve their treatment goals without causing hypoglycaemia. In recent years, bioengineering technology has greatly advanced diabetes management, with the development of blood glucose meters, continuous glucose monitors, insulin pumps and control systems for automatic delivery of one or more hormones. New insulin analogues have improved subcutaneous absorption characteristics, but do not completely eliminate the risk of hypoglycaemia. Insulin effect is counteracted by glucagon in non-diabetic individuals, while glucagon secretion in those with type 1 diabetes is impaired. The use of glucagon in the artificial pancreas is therefore a logical and feasible option for preventing and treating hypoglycaemia. However, commercially available glucagon is not stable in aqueous solution for long periods, forming potentially cytotoxic fibrils that aggregate quickly. Therefore, a more stable formulation of glucagon is needed for long-term use and storage in a bi-hormonal pump. In addition, a model of glucagon action in type 1 diabetes is lacking, further limiting the inclusion of glucagon into systems employing model-assisted control. As a result, although several investigators have been working to help develop bi-hormonal systems for patients with type 1 diabetes, most continue to utilize single hormone systems employing only insulin. This article seeks to focus on the attributes of glucagon and its use in bi-hormonal systems.

Introduction

Type 1 diabetes mellitus (T1DM) is an autoimmune disorder characterized by hyperglycaemia resulting from absolute deficiency of insulin. Poorly controlled T1DM can lead to acute and chronic complications [1-4], and intensive insulin treatment has shown sustained benefit to minimize these long-term complications [5, 6]. However, intensive treatment with insulin is associated with increased risk of hypoglycaemia [7], due in part to dysfunctional glucagon secretion [8]. Severe hypoglycaemia can result in seizures, coma and if untreated, death [9], and for this reason, rapid-acting glucose (such as tablets, fruit juice or sweet snacks) or a glucagon emergency kit should always be available for those treated with insulin.

Conventional methods of self-monitoring of blood glucose (SMBG) and multiple daily injections (MDI) are very challenging for patients and families. Continuous glucose monitors (CGMs) can detect hypoglycaemia and glycaemic variability and can provide valuable information about patients' glycaemic trends. Randomized control trials and meta-analyses have shown that CGM use is associated with significant reduction in A1c [10]. A Cochrane meta-analysis reported significant decline in HbA1c using CGM with or without an insulin pump, although the reduction in HbA1c was greatest with sensor-augmented pump (SAP) therapy [11]. Novel analogues of insulin, such as aspart, lispro and glulisine possess improved pharmacokinetic properties [12] and are promising for closed-loop systems. The availability of CGMs and subcutaneous insulin pumps has somewhat eased this burden [13] but such devices are dependent on manual inputs and are prone to human error. Studies have shown that continuous subcutaneous insulin infusion (CSII) by pumps achieves better glycaemic control than MDI at higher initial HbA1c values [14-16]. The automatic delivery of insulin and glucagon has long been investigated as a possible alternative to self-administered insulin [17, 18]. Utilizing current pump and glucose sensor technology, closed-loop systems present a hopeful method of dealing with the issues of hyper and hypoglycaemia, while at the same time keeping the HbA1c at the goal recommended by expert panels. Within these systems, control algorithms interface with pumps and sensors to command hormone infusion rates. Several modern control strategies have already been successfully employed in various closed-loop designs. In essence, these three components—CGMs, control algorithms and pumps for hormones, comprise most currently-investigated closed-loop systems (figure 1).

Figure 1.

The general components of a closed-loop system for the management of type 1 diabetes: a continuous glucose monitor from which data is transmitted to a control algorithm, which in turn controls hormone delivery via a subcutaneous pump.

Research teams are working on developing artificial pancreas systems [19-26], including groups funded by the Juvenile Diabetes Research Foundation's (JDRF) Artificial Pancreas Consortium [27], the Diabetes wiREless Artificial pancreas consortiuM (DREAM) [24] and the Artificial Pancreas At Home [28]. These consortia encourage collaboration between research groups, reducing burden among individual scientists and promoting open exchange of information. An important accomplishment that has been very useful to these scientists has been the creation of a 300-person simulator designed for pre-clinical simulations of novel control algorithms to assess safety prior to human testing, reducing the need for animal testing. This simulator has been approved by the Food and Drug Administration for use in the USA [29]. Improvements in sensor accuracy [30] and in pump technology have also greatly improved the efficacy of artificial pancreas systems.

Single Hormone Systems

Interestingly, Kadish's closed-loop experiments in the 1960s used both intravenous insulin and glucagon. Later, research into closed-loop glycaemic control focused more on the delivery of insulin alone. In the late 1970s the Biostator was found to be useful as a research tool in the management of intravenous insulin delivery [31]. However, closed-loop control with subcutaneous insulin remained a challenge due to the significant absorption delay of older insulin preparations compared to newer analogues such as aspart, lispro and glulisine [12]. The use of glucagon in closed-loop systems fell out of vogue during this period of time, not resurfacing until the mid-2000s in subcutaneous delivery algorithms [32, 33]. With the introduction of insulin analogues, there was renewed interest in single hormone (insulin-only) closed-loop systems, with the development of more sophisticated algorithms [34-36]. Over the last two decades, closed-loop system technology has rapidly advanced and multiple small-scale clinical studies of single hormone systems have been completed.

Weinzmer and Steil tested a PID controller in fully automated (no human input) and hybrid (meal announcements) closed-loop systems, both in adults and paediatric populations [25, 37]. Although both settings were shown to effectively prevent glucose variations, hybrid control was more effective than fully closed-loop control in terms of minimizing post-meal glycaemic excursions.

Hovorka et al. initially implemented a Model Predictive Control (MPC)-based adaptive control algorithm to treat children and adolescents with T1DM and concluded that overnight closed-loop insulin delivery improved glycaemic control and reduced glucose excursion and nocturnal hypoglycaemia [20, 38]. Studies involving pregnant subjects with type 1 diabetes also showed that conventional insulin pump therapy and closed-loop control were comparable [39]. Multinational studies in type 1 diabetes subjects carried out in the USA and Italy, also using an MPC system developed by Cobelli et al., showed that MPC control was better than open-loop control, with reduced time spent outside the target range (70–140 mg/dl) and with reduced hypoglycaemic events overnight [40, 41]. Chassin et al. in 2004, and Kovatchev et al. in 2009, have completely designed and tested MPC algorithms using in silico methods [29, 42].

The MD-logic Artificial Pancreas system, developed in Israel by Atlas et al. [43], has been investigated in clinical trials [24] and initial results with a meal-detection system showed promise, with a reduction in post-meal and overnight hypoglycaemia. This closed-loop system utilizes a fuzzy logic control algorithm for delivery of insulin.

Intraperitoneal insulin delivery has also been explored in closed-loop settings by Renard et al. [44] as an alternative to subcutaneous delivery, and has showed quicker absorption profiles and reduced variability in plasma insulin levels. Clinical studies comparing intraperitoneal to subcutaneous delivery of insulin [45, 46] showed the improved kinetic characteristics of the former.

Various CGM studies show prolonged nocturnal hypoglycaemia in patients with T1DM [47, 48], providing the rationale for using SAP therapy or a low glucose suspend (LGS) feature when managing patients with T1DM. SAP has been shown to be effective in lowering HbA1c [49-51], though higher baseline HbA1c values predict better response. The LGS feature works by temporarily suspending basal insulin delivery for 2 h if there is no response to the hypoglycaemia alarm (Paradigm® Veo™, Medtronic, Minneapolis, MN, USA, available in the Europe). Early studies of LGS found a lower duration and severity of hypoglycaemia without leading to significant hyperglycaemia [52]. Moreover, transient overnight suspension of basal insulin delivery is not associated with diabetic ketoacidosis [53, 54]. Investigators have shown that LGS significantly reduced the frequency and duration of nocturnal and exercise-related hypoglycaemia especially in patients with hypoglycaemic unawareness [52, 55, 56].

Bi-Hormonal Systems

In 2010, our group carried out a crossover study in which we compared insulin-only closed loop control versus insulin plus glucagon control. We showed that the delivery of glucagon based predominantly on proportional and derivative error substantially reduced the rate of hypoglycaemia versus insulin alone [57]. At about the same time, a group from Boston employed a closed-loop bi-hormonal system using MPC for insulin and PID for glucagon delivery. After they accounted properly for the subject-to-subject variability in insulin kinetics, their system also achieved very good glycaemic control with minimal hypoglycaemia [22]. More recently, Russell et al. [58] found excellent glycaemic control with rare hypoglycaemia in the setting of high carbohydrate meals and exercise. Haidar et al. [59] employed a closed-loop bi-hormonal system in a clinical study against open-loop subcutaneous infusion. The closed-loop bi-hormonal system was superior to open-loop control in terms of reducing hypoglycaemia in 15 adults with type 1 diabetes. A total of 53% of subjects had a hypoglycaemic event less than 54 mg/dl during open-loop control treatment versus 7% of subjects during bi-hormonal closed-loop treatment [59]. The closed-loop subjects were given a mini-bolus of glucagon, on average, every 3.6 h and the change from falling glucose to rising glucose occurred on average after 20 min, though the specific nature of the glucagon algorithm was not disclosed [59].

Interestingly, our group and the Boston group found that in about 20–35% of cases in which glucagon was administered; it was not effective in preventing hypoglycaemia. For this reason, we undertook a more detailed analysis of the data in order to better understand why glucagon was sometimes unsuccessful. For example, we found that in the failures, estimated insulin-on-board was almost twice as high as in the successes. In addition, glucose at the time of glucagon delivery was significantly lower in failures than in successes, and sensors tended to overestimate glucose during failures. In summary, we found that glucagon may fail to prevent hypoglycaemia when insulin-on-board is high or when glucagon delivery is delayed due to overestimation of glucose by the sensor. We concluded that improvements in sensor accuracy and delivery of larger or earlier glucagon doses when insulin levels are high may further reduce the frequency of hypoglycaemia [60].

We and others believe that, for optimal efficacy of glucose control, a closed-loop algorithm should monitor insulin action. For this reason, we sought to determine if a model-based revision of the original algorithm could correctly detect, and reduce the hyperglycaemia from corticosteroid-induced insulin resistance in subjects with type 1 diabetes. This algorithm, termed the adaptive-proportional derivative (APD) system, correctly detected the reduced sensitivity, increased insulin delivery accordingly and minimized the resulting rise in glucose [61]. Currently, the APD system is being tested in a fully automated setup, with wireless communication from the sensors to the controller and then to the insulin and glucagon delivery systems (see components in figure 2), and plans are underway for out-patient studies in a hotel setting.

Figure 2.

The components of the Oregon Artificial Pancreas System: two Dexcom generation 4 continuous glucose monitors, controller (Viliv, YukYung Technologies Corp, Gyeonggi-do, South Korea), and two Insulet OmniPod® Personal Device Managers, one for insulin and one for glucagon.

Owing to rapid utilization of glucose by muscle and short-term improvement in the insulin sensitivity, the risk of hypoglycaemia increases during exercise. In addition, due to unavailability of effective method detecting physical activity, glycaemic control during exercise is challenging in the closed-loop setting. Using proportional-derivative controller Van Bon et al. [62] have shown that the efficacy of bi-hormonal closed-loop system was comparable to open-loop control.

Glucagon in Closed-Loop Systems

Glucagon use in the closed-loop setting is not new. Bi-hormonal delivery was first introduced by Kadish in 1964. Although the initial model was bulky and carried in a backpack, it was capable of continuous glucose monitoring, and insulin and glucagon delivery. In this model, glucagon was given intravenously to prevent severe hypoglycaemia. Subcutaneously delivered glucagon also has rapid onset of action and can be used to avoid the complications of intravenous administration. On the basis of the human pharmacokinetic data, El Khatib et al. [22] found that low dose glucagon is rapidly absorbed from the subcutaneous space with peak absorption at 23 ± 9 min, a result very similar to that of Graf et al. [63] who tested large SC glucagon doses in humans. In the Graf study, intramuscular glucagon had a faster onset and peak than SC glucagon. In terms of pharmacodynamic effects, the peak glycaemic effect after a large dose in the Graf study was approximately 30 min. In pigs from our laboratory, the peak glycaemic effect after small doses was also about 30 min [64].

Several investigators have used glucagon in closed-loop studies, though currently available preparations of glucagon are not FDA approved for long-term use in solution. In previous studies at Portland Oregon, glucagon was reconstituted every 8 h and delivered subcutaneously via an infusion pump. Although reconstitution of commercially available glucagon every 8 h appeared to be effective in a research setting, for outpatient closed-loop use of glucagon, more stable preparations must be developed which can indwell in the portable pump for at least 3 days. Commercially available glucagon forms amyloid fibrils almost immediately after reconstitution with water [64-66]. These fibrils can turn into a gel after prolonged ageing in solution, with the risk of pump occlusion. Glucagon fibrillation is affected by pH, osmolarity and concentration of glucagon solution [64]. Owing to the fact that the isoelectric point of glucagon is approximately 7, it is not soluble at neutral pH. Ward et al. [64] have shown significant decrease in glucagon fibrillation at pH 10 compared to pH 3. In addition, these fibrils have potential for cytotoxicity and cause cell death in cultured mammalian cells via apoptosis [64]. Onoue et al. [65] described kinetics of glucagon fibrillation and found that these fibres are similar to β-amyloid and prion protein, pathologically linked to diseases such as Alzheimer's and prion disease. Interestingly, some investigators reported that the amyloid β configuration is a normal storage form of mammalian hormones before exocytosis [67].

Owing to its cytotoxicity and potential to occlude the pump catheter, glucagon fibrillation is one of the major hurdles in the success of bi-hormonal closed-loop systems. Matilainen et al. [68] have reported some improvement in glucagon stability with cyclodextrin but this additive did not completely block fibril formation. Although the potential importance of inhibiting glucagon fibrillation has been recognized for several years, and several methods have been used to block fibrillation with some success, relatively few attempts have been made at preventing the degradation (i.e. deamidation, oxidation and isomerization) of glucagon.

Future Directions

Many investigators have been working on testing closed-loop systems in different settings. Inpatient and short-term outpatient studies have shown promising results. However, the ultimate goal is to move to chronic use in the outpatient setting. Cobelli et al. [69] reported a successful feasibility and safety study of a wearable closed-loop system in an outpatient (hotel) setting, using a smartphone, and the DREAM group recently published data from a multinational camp study in children showing reduced overnight hypoglycaemia with lower overnight median glucose levels [70]. Such outpatient studies are necessarily needed at this juncture in order to continue working envelope towards a workable, ambulatory device.

Investigators have explored accelerating absorption of rapid-acting analogue insulins through site-warming techniques [71] and co-administration of hyaluronidase [72, 73] but there remains a need for more rapid insulin absorption. Weinzimer et al. [74] investigated the effect of pre-meal pramlintide in the closed-loop setting, showing some benefits in preventing post-meal excursions. Sensor technology has advanced significantly over the last 10 years but occasional inaccuracy continues to be a problem. The use of redundant sensor technology with either similar [75] or differing sensor technologies may play an important role in optimizing sensor accuracy, though further reductions in sensor error may make redundancy less useful.

Further integration of devices and optimization of hardware reliability are also needed if closed-loop systems are to become widely accepted. Currently, researchers use multiple handheld devices to receive sensor signals, compute hormone delivery rates and control hormone delivery pumps. The complexity of such systems in their current forms are problematic and cause sensor signal dropouts, failure of pumps to receive hormone delivery commands, and overall system failure requiring professional intervention, which is not acceptable in the outpatient setting.

Small, automated doses of glucagon successfully minimize the incidence and duration of hypoglycaemia in bi-hormonal closed-loop systems [22, 57]; however, instability of glucagon is a large hurdle in this regard. Commercialization of a bi-hormonal pump will require stable glucagon preparations that can remain in a wearable pump for at least 3–7 days.

To date, studies using closed-loop systems have been limited to treating patients for less than 1 week. The experience of testing such systems outside of the hospital has been even more limited. Commercialization of closed-loop systems requires much longer-term outpatient studies showing safety and should include assessments of metabolic control and quality of life [24, 26]. To be used independently by the patient, the closed-loop device should be safe, effective, portable and user-friendly. From the patient's perspective, automated closed-loop treatment will be an important achievement that will improve glycaemic control while reducing the risk of hypoglycaemia.

Conclusion

Even with great improvement in the accuracy and reliability of glucose sensors and insulin pump technology, patients with T1DM still struggle with hypoglycaemia and achievement of HbA1c goals. Although single-hormone closed-loop systems are clearly superior to standard open-loop insulin therapy, the risk of hypoglycaemia remains a challenge, especially in situations such as post-exercise and the late post-prandial period. Bi-hormonal closed-loop systems utilizing glucagon have been shown to further reduce the incidence and duration of hypoglycaemia. However, glucagon stability remains a hurdle to the successful use of this hormone in closed-loop systems and there is a need for stable preparations of glucagon to improve chronic closed-loop treatment. After promising results from inpatient studies, investigators are now testing these systems in the outpatient setting. Closed-loop systems show promise in improving the safety and effectiveness of T1DM treatment.

Acknowledgements

We appreciate the financial support from NIH K23 grant to Dr J. El Youssef (5K23DK090133-03) and NIH grant to Oregon Clinical and Translational Research Institute (UL1 RR024140); and grants from Juvenile Diabetes Research Foundation, Legacy Good Samaritan Foundation and the HEDCO Foundation.

Conflict of Interest

W. K. W. has filed patents covering a dual hormone closed loop algorithm and a method for biochemically stabilizing glucagon.

Ancillary