The artificial pancreas: making headway

The concept of ‘ closing the loop’ connecting glucose sensing to insulin delivery is not new. In 1964, Kadish was the first to use continuous real-time glucose monitoring to close the loop with an ‘ on-off system’ using an intravenous (IV) infusion of insulin and glucose. Ten years later, Albisser in Toronto and Pfeiffer in Ulm combined IV continuous glucose monitors with algorithms implemented on a microcomputer to automate IV insulin delivery thus facilitating the development of the Biostator (Miles Laboratory Inc, USA); the first commercial ‘ artificial pancreas’.1 These early achievements remained impractical for clinical use because they were dependent on glucose sensing in blood and IV insulin delivery. It took over 20 years for the first commercial subcutaneous continuous glucose monitor (CGM) to see daylight in 1999. The CGMS system (Medtronic MiniMed, USA) operated in the Holter-style to collect continuous glucose trace over three days. This and ever-improving insulin pump technology opened the door for other applications, such as real-time CGM, and rejuvenated interest in the artificial pancreas. In 2006, the Juvenile Diabetes Research Foundation (JDRF) embraced the concept of closed loop and allocated US $6 million in grants to investigate the benefits of technology controlling blood glucose level with the aim of accelerating the availability of the artificial pancreas.2 This gave the long-needed impetus required to move things forward. Supported by the JDRF, the University of Cambridge Artificial Pancreas Project was initiated at the end of 2006 with the aim of exploiting the existing technological base for the development of an artificial pancreas for overnight glucose control in children and adolescents with type 1 diabetes (T1D). The most urgent research question was whether the existing CGMs in combination with a control algorithm could facilitate safe and efficacious glucose control. The preliminary evidence is that improvements in glucose control can be achieved compared to the current best clinical practice,3, 4 but further research is required to establish clearly the benefits of the technology and pave the path for commercial exploitation.5 The (extracorporeal) artificial pancreas integrates three components, a subcutaneous CGM, a control algorithm, and an insulin pump, which could be linked by a wireless communication to enable data exchange. The CGM measures glucose concentration every 1–10min but the measurements are made in the subcutaneous (SC) tissues introducing a delay of about 10min due to the kinetics of glucose transport between the blood and interstitial fluid.6 CGMs are calibrated against finger-prick glucose measurements every 6–48h.7 Three CGMs are currently available on the market, the CGMS (Medtronic MiniMed, USA), the STS (DexCom, USA), and the GlucoDay (Menarini, Italy), whereas the FreeStyle Navigator (Abbott, USA) is anticipated to receive US Food and Drug Administration approval in 2007. The CGM measurements are utilised by a glucose controller, a computer program, which calculates an insulin infusion rate or an insulin microbolus every 1–10min. In prototypes, the control algorithm normally resides on a laptop or a handheld computer although commercial systems are likely to integrate it within an insulin pump or a handheld pump controller. The insulin pump delivers fast acting insulin into the SC tissue as a continuous infusion or as a bolus. Delays between insulin delivery and insulin-induced glucose lowering of about 140min are observed. The time-to-peak of plasma insulin concentration following SC delivery of fast acting insulin such as lispro or aspart is 40min.8 The additional 100min originates from the delay in insulin action (30min) and the propagation of insulin action to achieve the maximum insulin-induced blood glucose lowering (70min). This artificial pancreas based on a subcutaneous CGM and subcutaneous insulin delivery (SC–SC) setup is at the current research focus. Other closed loop arrangements are possible, particularly the fully implantable system with IV glucose sensing and intraportal insulin delivery,9 but these systems are hampered by the lack of commercially available intravenous CGMs and implantable pumps. Furthermore, only a little over 1000 insulin pumps have been implanted worldwide compared to over 300 000 SC insulin pump users in the US alone. The current experience suggests that an SC–SC system will not be fully autonomous but will require two types of end-user assistance: device handling and event disclosure. The former corresponds to inserting and calibrating the CGM, and replacing the insulin infusion set. The latter corresponds to the end-user communicating events, such as meals and possibly exercise, and manually triggering prandial insulin boluses. Currently, we would predict that the closed loop will be interrupted at meal times as safe and efficacious glucose control may be very difficult to achieve in the presence of system delays. At other times, the system will operate in a fully closed loop mode without user interaction. Over the next two years, the Artificial Pancreas Project at the University of Cambridge will test overnight closed loop glucose control. We will be studying five- to 18-year-old children and adolescents with T1D. The focus on overnight control reflects our expectation that the introduction of closed loop control in clinical practice will be staged, starting simply and progressing to more complex clinical management scenarios. Overnight glucose control is desirable for several reasons. It avoids postprandial conditions which are characterised by difficulties in predicting insulin requirements and thus the risk of postprandial hypoglycaemia. The effects of day-time exercise are restricted to delayed exercise-related hyperglycaemia. Overnight glucose control promises considerable benefits. The risk of nocturnal hypoglycaemia might be reduced, while morning normoglycaemia should allow a more consistent and reliable prandial breakfast insulin dosing. In children with T1D, nocturnal hypoglycaemia is frequent and contributes to the development of hypoglycaemia unawareness.10 It is also the most feared complication of T1D treatment and may be an impediment to intensified treatment. We will perform a number of small-scale studies at the Wellcome Trust Clinical Research Facility at Addenbrooke's Hospital, Cambridge. Each study will recruit approximately 12 subjects and will evaluate closed loop control against standard treatment by continuous SC insulin infusion. Our first two studies will evaluate the control algorithm and the wireless communication technologies over a single night of closed loop control. The control algorithm will utilise the model predictive control (MPC) approach.11 We will implement the control algorithm on a laptop computer but the aim would be to port the system to a handheld computer. We plan to use the FreeStyle Navigator CGM12 or another commercially available CGM for glucose level measurements. Our subsequent studies will evaluate the effect of meals and exercise on the ability of the control algorithm to achieve safe and efficacious glucose control overnight. Prior to each study night, the subjects will consume meals of different caloric content and size. The effect of exercise at different levels of intensity will be studied. The final study plans to investigate glucose control over several consecutive nights in the home setting. This last study should provide a stepping stone for larger-scale, home-based studies. The primary outcomes measure of our series of studies will be the efficacy and safety of glucose control using the closed loop system compared to the current clinical practice. The studies provide considerable challenges from the regulatory viewpoint. We will be using a non-CE marked industry-developed CGM technology in combination with University-developed control algorithm(s) and an associated user interface. An extensive collaboration with the medical device industry is required to modify the CGMs and insulin pumps for our investigational purposes and to satisfy regulatory requirements. Our work is part of the JDRF's Artificial Pancreas Project2 which involves five additional research groups based in the US, who will investigate closed loop control in various settings and populations. Building on recent achievements, William Tamborlane and co-workers at Yale University will continue developing a closed loop system in the paediatric population with T1D using the ePID Medtronic MiniMed closed loop prototype, based on the proportional-integral-derivative (PID) control algorithm and the CGMS glucose monitor.3 Although initially conceived as a fully closed loop system, the recent work suggests that the ePID system achieves better performance with manual (open) as opposed to fully-automated prandial insulin dosing in line with theoretical expectations.13 In contrast, our work at the University of Cambridge is utilising the MPC paradigm to develop the glucose controller. The MPC approach is a modern control approach suitable for systems with long time delays and relies on a model of the pharmacodynamic effect of insulin. Whilst more complicated, the MPC approach promises a better performance than the PID control strategy although a head-to-head comparison has yet to be performed. Additionally, we will use an adaptive approach to deal with inter- and intra-individual variability in insulin requirements further improving the system performance. Other research groupings within the JDRF Artificial Pancreas Project will develop their own prototypes of the artificial pancreas and will test these in various populations, including adult subjects with T1D. Given the clinical, control engineering, and industrial expertise concentrated in the JDRF Artificial Pancreas Project and given the JDRF's concerted support and advocacy, the next two years provide an exciting and unique opportunity to bring the artificial pancreas closer to large-scale clinical trials, and to facilitate the professional and social acceptance of the technology— and, ultimately, to accelerate the availability of the first generation of commercial devices. The development and deployment of the artificial pancreas are likely to be staged with an objective of reducing progressively the risk of hypoglycaemia whilst maintaining or improving HbA1c. A ‘ leap’ to near-normal glucose levels is unlikely and should not be expected from the early generations of the artificial pancreas. However, it is a long-term goal. The treated subjects, the health care professionals, and the regulatory authorities have to become comfortable with the technology and an appropriate infrastructure has to be put in place. The staged rollout should not hinder the introduction of the artificial pancreas into clinical practice. Waiting for perfection could prevent the achievement of long awaited improvements in glucose control. The optimistic timescale of the first generation of a commercial artificial pancreas is around four to five years. The first small-scale clinical trials should finish within the next two years to be followed by larger-scale investigations to support regulatory submissions. Meanwhile, the status of the CGMs should progress from an ‘ adjunct’ to a ‘ replacement’ label to be commensurate with the way in which the artificial pancreas will use the CGMs. Although targeted to T1D in the first instance, the artificial pancreas should benefit other insulin-treated subjects with diabetes. This includes intensively-treated type 2 diabetes and insulin-treated gestational diabetes. In such populations, the delivery of exogenous insulin through the artificial pancreas is complemented by the residual endogenous insulin secretion reducing the system delays, increasing the responsiveness of the system, and possibly improving the overall performance compared to that shown in T1D. The cost of the artificial pancreas is likely to exceed the combined cost of the CGM and the insulin pump. Based on the figures valid in 2006, the annual cost of 24/7 usage of the DexCom CGM is £2200 (US $4380),14 and the annual cost of the insulin pump is in the region of £1500. Thus, should the artificial pancreas be available today, its annual cost would likely exceed £3700 to cover the developmental cost. Future savings could be achieved through the cost reduction of the components such as the use of insulin patch pumps and the anticipated reduction of the cost of the CGMs once the technology matures and competition increases. In conclusion, the JDRF Artificial Pancreas Project provides an exciting opportunity to capitalise on the recent developments in continuous glucose monitoring. The open research question is whether such monitors combined with a suitable glucose controller can be used to drive closed loop glucose control. This will be answered by building prototypes and rigorous clinical evaluations prior to moving to larger-scale home trials. Support by the JDRF, the European Foundation for the Study of Diabetes (EFSD), and the EU Clincip Project (Grant Number IST-2002-506965) is acknowledged.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.