Abstract
Introduction
This year was an exciting time in the path forward toward commercialization and generalized availability of closed‐loop systems for clinical use. The number of companies looking to bring hybrid‐closed loop (HCL) systems to the market has multiplied. Results of the 6‐month randomized control trial of the Tandem's Control‐IQ hybrid closed‐loop system were presented at the 2019 American Diabetes Association (ADA) meeting and were recently published, showing time in target glucose range 70–180 mg/dL for participants using Control‐IQ was 71% compared to 59% for participants in the sensor‐augmented pump (SAP) group (P<0.0001). During the overnight period, time in range with Control‐IQ was 76% compared to 59% for SAP (P<0.0001) (1). First human study testing results of the iLet Gen3, a bionic pancreas platform connected to either the Dexcom G5 or the Senseonics Eversense implanted continuous glucose monitor (CGM), were also presented at the 2019 ADA meeting, showing 70.1% time in glucose range 70–180 mg/dL for those using the iLet vs. 61.5% for usual care (P=0.006) (2). Pilot data from the Diabeloop closed‐loop system was recently published (3), as well as a series of studies on the advancement of the Omnipod Personalized Model Predictive Control Algorithm for use in the Omnipod Horizon™ Automated Glucose Control System, showing that the Omnnipod system completely prevented overnight hypoglycemia after exercise and achieved 76% time in range (70–180 mg/dL) despite being challenged with missed meal bolus and overbolus scenarios (4,5). In addition, the community of people using do‐it‐yourself closed‐loop systems is constantly growing, and this year several studies and self‐reported data were published. Although there are debates regarding the use of these unregulated systems, future studies and industry collaborations such as the TidePool Loop project were recently established (6).
This year's article highlights new systems as well as advances coming to the next generation of closed‐loop systems—research platforms that will expand the ability to develop and evaluate these systems in the academic setting.
Advancement was also recorded in the field of decision support systems. Several studies were published, mainly focusing on people with type 1 or 2 diabetes using multiple daily injections (MDI). MDI users make up more than half of the type 1 population and most people with type 2 diabetes who are treated with insulin. This large population will probably not use closed‐loop systems but will still need advice and support to improve their metabolic control. Thus, decision support tools for MDI users have the potential to improve care for many people with diabetes. This year, the first large multicenter study was published demonstrating the safety and effectiveness of daily use of such a tool among people with type 2 diabetes. Titration guidelines have been transformed into decision support systems used by healthcare providers in clinics and medical centers, and are now being tested in small feasibility studies.
Key Articles Reviewed for the Article
Bergenstal RM, Johnson M, Passi R, Bhargava A, Young N, Kruger DF, Bashan E, Bisgaier SG, Isaman DJM, Hodish I
Madsen JOB, Casteels K, Fieuws S, Kristensen K, Vanbrabant K, Ramon‐Krauel M, Johannesen J; on behalf of the ABC consortium
Liu C, Avari P, Leal Y, Wos M, Sivasithamparam K, Georgiou P, Reddy M, J Fernández‐Real JM, Martin C, Fernández‐Balsells M, Oliver N, Herrero P
Mathioudakis N, Jeun R, Godwin G, Perschke A, Yalamanchi S, Everett E, Greene P, Knight A, Yuan C, Hill Golden S
Forlenza GP, Buckingham BA, Christiansen MP, Wadwa RP, Peyser TA, Lee JB, O'Connor J, Dassau E, Huyett LM, Layne JE, Ly TT
Buckingham BA, Christiansen MP, Forlenza GP, Wadwa RP, Peyser TA, Lee JB, O'Connor J, Dassau E, Huyett LM, Layne JE, Ly TT
Ekhlaspour L, Forlenza GP, Daniel Chernavvsky, Maahs DM, Wadwa RP, Deboer MD, Messer LH, Town M, Pinnata J, Kruse G, Kovatchev BP, Buckingham BA, Breton MD
Forlenza GP, Ekhlaspour L, Breton M, Maahs DM, Wadwa RP, DeBoer M, Messer LH, Town M, Pinnata J, Kruse G, Buckingham BA, Cherñavvsky D
Forlenza GP, Pinhas‐Hamiel O, Liljenquist DR, Shulma DI, Bailey TS, Bode BW, Wood MA, Buckingham BA, Kaiserman KB, Shin J, Huang S, Lee SW, Kaufman FR
Forlenza GP, Li Z, Buckingham BA, Pinsker JE, Cengiz E, Wadwa RP, Ekhlaspour L, Church M, Weinzimer SA, Jost E, Marcal T, Andre C, Carria L, Swanson V, Lum JW, Kollman C, Woodall W, Beck RW
Anderson SM, Buckingham BA, Breton MD, Robic JL, Barnett CL, Wakeman CA, Oliveri MC, Brown SA, Ly TT, Clinton PK, Hsu LJ, Kingman RS, Norlander LM, Loebner SE, Reuschel‐DiVirglio S, Kovatchev BP
Biester T, Nir J, Remus K, Farfel A, Ido Muller I, Biester S, Atlas E, Dovc K, Bratina N, Kordonouri O, Battelino T, Danne T, Nimri R
Boughton CK, Bally L, Martignoni F, Hartnell S, Herzig D, Vogt A, Wertli MM, Wilinska ME, Evans ML, Coll AP, Stettler C, Hovorka R
Petruzelkova L, Soupal J, Plasova V, Jiranova P, Neuman V, Plachy L, Pruhova S, Sumnik Z, Obermannova B
Deshpande S, Pinsker JE, Zavitsanou S, Shi D, Tompot R, Church MM, Andre C, Doyle III FJ, Dassau E
Chakrabarty A, Gregory JM, Moore LM, Williams PE, Farmer B, Cherrington AD, Lord P, Shelton B, Cohen D, Zisser HC, Doyle III FJ, Dassau E
Hajizadeh I, Rashid M, Turksoy K, Samadi S, Feng J, Sevil M, Hobbs N, Lazaro C, Maloney Z, Littlejohn E, Cinar A
Decision Support Systems
Automated insulin dosing guidance to optimise insulin management in patients with type 2 diabetes: a multicentre, randomised controlled trial
Bergenstal RM1, Johnson M1, Passi R1, Bhargava A2, Young N2, Kruger DF3, Bashan E4, Bisgaier SG4, Isaman DJM5, Hodish I4,6
1International Diabetes Center, Minneapolis, MN; 2Iowa Diabetes and Endocrinology Research Center, Des Moines, IA; 3Henry Ford Medical Center Endocrinology, Detroit, MI; 4Hygieia Inc, Livonia, MI; 5Department of Biostatistics, School of Public Health, University of Michigan, Ann Arbor, MI; 6Department of Internal Medicine, Division of Metabolism, Endocrinology and Diabetes, University of Michigan Medical Center, Ann Arbor, MI
Background
Insulin therapy is most effective if dosage titrations are done regularly and often; however, this is seldom practical for most clinicians, which can lead to an insulin titration gap. The d‐Nav Insulin Guidance System (Hygieia, Livonia, MI) is a handheld device that measures glucose, determines glucose patterns, and automatically determines the appropriate dose for the next insulin administration. The aim of this study was to investigate whether the combination of healthcare professional support plus the d‐Nav device is more effective than healthcare professional support alone.
Methods
Participants for this multicenter, randomized, controlled study, were recruited from three diabetes centers in Detroit, MI, Minneapolis, MN, and Des Moines, IA. Adult patients (21–70 years) who had been diagnosed with type 2 diabetes, had a hemoglobin A1c (HbA1c) level of 7.5% or higher (≥58 mmol/mol) and 11% or lower (≤97 mmol/mol), and had been on the same insulin regimen for the previous 3 months were eligible for inclusion. Exclusion criteria included body mass index of 45 kg/m2 or higher; severe cardiac, hepatic, or renal impairment; and three or more severe hypoglycemic events in the previous year. Once selected, participants at each site were randomly assigned (1:1) to either the intervention group (d‐Nav plus healthcare professional support) or the control group (healthcare professional support alone). Both groups were had three face‐to‐face and four phone visits during the follow‐up period (6 months). The primary objective was to compare average change in HbA1c from baseline to 6 months. Frequency of hypoglycemic events was used to evaluate safety. The primary objective and safety were assessed in the intention‐to‐treat population.
Results
Between February 2, 2015, and March 17, 2017, 236 patients were screened for eligibility, of whom 181 (77%) were enrolled and randomly assigned to the intervention (n=93) and control (n=88) groups. At baseline, the intervention group had mean HbA1c of 8.7±0.8% (72±8.8 mmol/mol) and the control group mean HbA1c was 8.5±0.8% (69±8.8 mmol/mol). The mean decrease in HbA1c from baseline to 6 months was 1.0±1.0% (11±11 mmol/mol) in the intervention group and 0.3±0.9% (3.3±9.9 mmol/mol) in the control group (P<0.0001). Both groups had similar frequency of hypoglycemic events: 0.29±0.48 events per month in the intervention group versus 0.29±1.12 in the control group (P=0.96)
Conclusions
When automated insulin titration guidance was combined with support from healthcare professionals, patients experienced better glycemic control than with professionals healthcare support alone. This combination approach facilitated safe and effective insulin titration in a large study group of individuals with type 2 diabetes. This combined approach should now be assessed across large healthcare systems to further validate the current findings and to examine cost‐effectiveness.
In this study, participants with type 2 diabetes using insulin used the d‐Nav insulin guidance system (Hygieia, Livonia, MI), a handheld device that contains a glucose meter and software that provided its users with dose by dose insulin recommendations, together with healthcare professional support. Both groups were contacted seven times (three face‐to‐face and four phone visits) during 6 months of follow‐up. Patients in both arms had a similar safety profile, but there was a significant improvement in HbA1c for users of the new system vs. those having healthcare professional support alone. Users of the system reported insulin dose adjustments on average 1.1 times per week and did more frequent fingerstick measurements than those in the control group. This study shows that automated titration guidance technology with healthcare professional support was superior to a healthcare professional support model alone for patients with type 2 diabetes on a number of different insulin injection regimens. The study presents a real‐time titration solution for a large population of people with type 2 diabetes using insulin in various ways. While this system has been studied previously in Europe (7), it will be important to see how well this and similar systems work without physician interventions, as it will be ideal if use of these clinical decision support systems could be scaled up for use in the larger population who may not have access to expert physician care.
No effect of an automated bolus calculator in pediatric patients with type 1 diabetes on multiple daily injections: the Expert Kids Study
Madsen JOB1, Casteels K2,3, Fieuws S4, Kristensen K5, Vanbrabant K4, Ramon‐Krauel M6, Johannesen J1,7; on behalf of the ABC consortium
1Department of Pediatrics, Herlev University Hospital, Herlev, Denmark; 2Department of Pediatrics, University Hospitals Leuven, Leuven, Belgium; 3Department of Development and Regeneration, University of Leuven, Leuven, Belgium; 4Interuniversity Institute for Biostatistics and Statistical Bioinformatics, KU Leuven–University of Leuven and Universiteit Hasselt, Leuven, Belgium; 5Department of Pediatrics, Skejby University Hospital, Aarhus, Denmark; 6Department of Endocrinology, Institut de Recerca Sant Joan de Deu, Hospital Sant Joan de Deu, Barcelona, Spain; 7Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark
Background
This multicenter crossover study investigated the potential beneficial effect of an automated bolus calculator (ABC) in children and adolescents with type 1 diabetes (T1D) treated with MDI.
Methods
Participants were randomized to either begin or end with a 5‐month intervention versus their regular treatment regimen (control), separated by a 2‐month washout period. During the intervention, participants were carefully instructed to use the ABC (Accu‐Check Aviva Expert, Roche) versus manual insulin calculations during the control period. Participants between 8 and 18 years of age with T1D were recruited from clinics in Denmark, Belgium, and Spain. Inclusion criteria included T1D for >1 year, a minimum of 3 months MDI treatment before inclusion, and HbA1c of 7.5%–11% (57–97 mmol/mol). Improvement in HbA1c was the main outcome, and improved quality of life (QoL) and glucose variability (time spent in target glucose) were secondary outcomes.
Results
A total of 65 patients with a mean age of 13.25 years and a mean HbA1c of 8.25% (66.7 mmol/mol) were included. Midway evaluation after 2 months of intervention showed no significant difference from the standard care (0.297 [95% CI −0.645 to 0.054]; P=0.10). The difference remained insignificant after the 5 months of intervention (−0.143 [95% CI −0.558 to 0.272]; P=0.51). Using the ABC did not change the time spent in target glucose range, nor did it change the QoL.
Conclusions
This study did not demonstrate beneficial additive effects of an ABC in children and adolescents with T1D treated with MDI in HbA1c, nor in any other endpoint investigated.
Automated bolus calculators are commonly used in insulin pumps, but not as often by MDI users. This study investigated whether using the Accu‐Check Aviva Expert in children and adolescents with T1D using MDI and carbohydrate counting would improve not only their metabolic control, but also their QoL and treatment satisfaction. The authors found no significant lowering of HbA1c or blood glucose variation, nor could they find significantly improved QoL or treatment satisfaction in the participants or their parents. High adherence (>80% use) also did not change the outcome of the results. The authors noted missed boluses could be expected to occur at just as high a frequency during the intervention period as in the control period, and an automated bolus calculator does not change this problem, and has limited effect if no blood glucose is measured or no carbohydrates are entered. As such, additional strategies beyond just adding an automated bolus calculator, for example addition of CGM, automated determination and adjustments of insulin dosing or other technologies, may be needed to help improve the efficacy of bolus calculators for children and adolescents with T1D treated with MDI.
A modular safety system for an insulin dose recommender: a feasibility study
Liu C1, Avari P2, Leal Y3, Wos M3, Sivasithamparam K2, Georgiou P1, Reddy M2, J Fernández‐Real JM3, Martin C4, Fernández‐Balsells M3, Oliver N2, Herrero P1
1Centre for Bio‐Inspired Technology, Department of Electrical and Electronic Engineering, Imperial College London, London, UK; 2Division of Diabetes, Endocrinology and Metabolism, Department of Medicine, Faculty of Medicine Imperial College, London, UK; 3Institut d'Investigació Biomèdica de Girona Dr Josep Trueta, Girona, Spain; 4Department of Computing and Communication Technologies, Oxford Brookes University, Oxford, UK
Background
Insulin dosing for people with T1D is a challenging and potentially risky task; therefore, there is a need to include safety measures as part of any insulin dosing or recommender system. This work presents and clinically evaluates a modular safety system that is part of an intelligent insulin dose recommender platform developed within the EU‐funded PEPPER project.
Methods
The proposed safety system is composed of four modules that use a novel glucose forecasting algorithm. These modules are predictive glucose alerts and alarms; a predictive low‐glucose basal insulin suspension module; an advanced rescue carbohydrate recommender for resolving hypoglycemia; and a personalized safety constraint applied to insulin recommendations. The proposed safety system was evaluated in a feasibility pilot study including eight adult subjects with T1D using multiple daily injections (MDI) over a duration of 6 weeks. Glycemic control and safety system functioning were compared between the 2‐week run‐in period and the end point at 8 weeks. A standard insulin bolus calculator was employed to recommend insulin doses.
Results
The percentage time in the clinically significant hypoglycemia range (<54 mg/dL, 3.0 mmol/l) significantly decreased from median (IQR) of 0.82% (0.05–4.79) at run‐in to 0.33% (0.00–0.93) at endpoint (P=0.02). This was associated with a significant increase in percentage time in target range (3.9–10.0 mmol/l) from 52.8% (38.3–61.5) to 61.3% (47.5–71.7) (P=0.03). There was also a reduction in number of carbohydrate recommendations for hypoglycemia treatment. Overall, glycemic control improved over the evaluated period.
Conclusion
A safety system for an insulin dose recommender has been proven to be a viable solution to reduce the number of adverse events associated to glucose control in T1D.
This study describes use of only the safety module of an automated decision support system to recommend insulin doses for people with T1D who use MDI, which is one module of the PEPPER (Patient Empowerment through Predictive PERsonalised decision support) project. The dosing module was not tested in this study. In this feasibility study, preliminary data from adult participants on MDI in a nonrandomized open‐label study was presented. Eight subjects initially used the system for 2 weeks with the recommendations disabled, then used the full system for the next 6 weeks. Six participants completed the study, overall glycemic control improved with no change in hypoglycemia below 70 mg/dL but with a significant decrease in time spent in significant hypoglycemia range of below 54 mg/dL. Although early phase technical and user experience issues were observed in this initial version of the system, this preliminary study showed the potential of an automated decision support system for people with T1D using MDI therapy. A limitation to the study was the small number of patients and lack of control arm, making it difficult to determine if the improvement in control was due to the safety system or the use of the sensor. Future improvements to the user experience for this and similar systems will be very important to support long‐term use of these systems by the target population, as this has a direct impact on adherence and outcomes especially with prolonged use of a technology.
Development and implementation of a subcutaneous insulin clinical decision support tool for hospitalized patients
Mathioudakis N1, Jeun R1, Godwin G2, Perschke A3, Yalamanchi S1, Everett E1, Greene P4, Knight A5, Yuan C6, Hill Golden S1
1Division of Endocrinology, Diabetes & Metabolism, Department of Medicine, Johns Hopkins University School of Medicine, Baltimore, MD; 2Epic Information Technology Team, Johns Hopkins Health System, Baltimore, MD; 3Nursing Administration, ClinicalInformatics, Johns Hopkins Hospital, Baltimore, MD; 4Johns Hopkins Hospital, Baltimore, MD; 5Johns Hopkins Bayview Medical Center, Baltimore, MD; 6Armstrong Institute for Patient Safety and Quality, Johns Hopkins University School of Medicine, Baltimore, MD
Background
Insulin administration in hospitalized patients is common as one third of those hospitalized in the United Sttes have diabetes or hyperglycemia. Inappropriate dosing carries a risk of secondary complications. A clinical decision support tool (CDS) can assist the medical team in safe and effective insulin management.
Methods
CDS was developed based on existing clinical practice guidelines to determine subcutaneous insulin dosing or non‐critically ill hospitalized patients. The project was carried out at two academic medical centers that use the EpicCare electronic medical record.
Results
The developed CDS provide insulin dosing regimen by using personal data of the patient from the EpicCare including patient's body weight, diabetes type, home and hospital insulin requirements, and nutritional status. The recommended daily insulin dosing is divided into basal and nutritional doses with personalized correctional insulin scale. Emails, lectures, and videos were used to communicate this new tool to clinicians before implementation. Use of the tool was accompanied by a technical support team, and continuous modification of the tool was done according to users' feedback. Inclusion of programming in the electronic medical record (EMR) provider's community library, allowing easy dissemination of the tool to other institutions.
Conclusions
The authors of this study developed an EMR‐based tool for subcutaneous insulin dosing in non‐critically ill hospitalized patients. Further studies are needed to evaluate the adoption and effectiveness of the tool.
A clinical decision on insulin dosing is based on knowledge such as published recommendations and guidelines, data gathered from the patient, electronic medical records, and laboratory test results. The clinical team does not always have knowledge of all of the patient's information nor time to review all this valuable data. Therefore, an automated clinical decision support tool can help the clinical team to identify course of treatment, may minimize insulin dosing errors, and improve clinical outcomes for hospitalized patients. The study presents a way to automize clinical guidelines for insulin dosing in a personalized way integrated with the EMR of the individual patient. Although the study did not present clinical outcomes of using the tool, it showed a way to develop and implement such a tool within the clinical center by the local team. The main establishment is the ability to integrate patient medical data to decision making tool using the EpicCare system, which is mandatory for the prevention of medical errors, simplicity of use, and can be shared with other clinical centers. The safety and effectiveness of this tool remains to be proven in further clinical studies.
Closed‐Loop Systems
Commercial System Development
Performance of Omnipod personalized model predictive control algorithm with moderate intensity exercise in adults with type 1 diabetes
Forlenza GP1, Buckingham BA2, Christiansen MP3, Wadwa RP1, Peyser TA4, Lee JB5, O'Connor J5, Dassau E6, Huyett LM5, Layne JE5, Ly TT5
1Barbara Davis Center for Diabetes, University of Colorado School of Medicine, Aurora, CO; 2Division of Pediatric Endocrinology, Department of Pediatrics, Stanford University, Stanford, CA; 3Diablo Clinical Research, Walnut Creek, CA; 4ModeAGC LLC, Palo Alto, CA; 5Insulet Corporation, Acton, MA; 6Harvard John A. Paulson School of Engineering and Applied Sciences, Harvard University, Cambridge, MA
This manuscript is also discussed in the article on Advances in Exercise, Physical Activity, and Diabetes, page S‐109.
Background
The objective of this study was to assess the safety and performance of the Omnipod® personalized model predictive control (MPC) algorithm with variable glucose set points and moderate intensity exercise using an investigational device in adults with T1D.
Methods
A supervised 54‐hour HCL study was conducted in a hotel setting after a 7‐day outpatient standard treatment phase. Adults aged 18–65 years with T1D and HbA1c between 6.0% and 10.0% were eligible. Subjects completed two moderate‐intensity exercise sessions of >30 minutes on consecutive days: the first with the glucose set point increased from 130 to 150 mg/dL, and the second with a temporary basal rate of 50%, both started at 90‐minutes preexercise. Primary endpoints were percentage time in hypoglycemia <70 mg/dL and hyperglycemia ≥250 mg/dL.
Results
Twelve subjects participated in the study, with (mean±SD) age 36.5±14.4 years, diabetes duration 21.7±15.7 years, HbA1c 7.6±1.1%, and total daily dose 0.60±0.22 U/kg. Outcomes for the 54‐h HCL period were mean glucose, 136±14 mg/dL; percentage time <70 mg/dL, 1.4±1.3%; 70–180 mg/dL, 85.1±9.3%; and ≥250 mg/dL, 1.8±2.4%. In the 12‐hour period after exercise start, percentage time <70 mg/dL was 1.4±2.7% with the raised glucose set point and 1.6±3.0% with reduced basal rate. The percentage time <70 mg/dL overnight was 0±0% on both study nights.
Conclusions
The Omnipod personalized MPC algorithm performed well and was safe during day and night use in response to variable glucose set points and with temporarily raised glucose set point or reduced basal rate 90 min in advance of moderate intensity exercise in adults with T1D.
Performance of the Omnipod personalized model predictive control algorithm with meal bolus challenges in adults with type 1 diabetes
Buckingham BA1, Christiansen MP2, Forlenza GP3, Wadwa RP3, Peyser TA4, Lee JB5, O'Connor J5, Dassau E6, Huyett LM5, Layne JE5, Ly TT5
1Division of Pediatric Endocrinology, Department of Pediatrics, Stanford University, Stanford, CA; 2Diablo Clinical Research, Walnut Creek, CA; 3Barbara Davis Center for Diabetes, University of Colorado School of Medicine, Aurora, CO; 4ModeAGC LLC, Palo Alto, CA; 5Insulet Corporation, Billerica, MA; 6Harvard John A. Paulson School of Engineering and Applied Sciences, Harvard University, Cambridge, MA
Background
This study assessed the safety and performance of the Omnipod personalized MPC algorithm using an investigational device in adults with T1D in response to overestimated and missed meal boluses and extended boluses for high‐fat meals.
Methods
A supervised 54‐hour HCL study was conducted in a hotel setting after a 7‐day outpatient open‐loop run‐in phase. Adults aged 18–65 years with T1D and HbA1c 6.0%–10.0% were eligible. Primary endpoints were percentage time in hypoglycemia <70 mg/dL and hyperglycemia ≥250 mg/dL. Glycemic responses for 4 hours to a 130% overestimated bolus and a missed meal bolus were compared with a 100% bolus for identical meals, respectively. The 12‐hour postprandial responses to a high‐fat meal were compared using either a standard or extended bolus.
Results
Twelve subjects participated in the study, with (mean±SD): age 35.4±14.1 years, diabetes duration 16.5±9.3 years, HbA1c 7.7±0.9%, and total daily dose 0.58±0.19 U/kg. Outcomes for the 54‐hour HCL period were mean glucose 153±15 mg/dL, percentage time <70 mg/dL (median [interquartile range]): 0.0% (0.0±1.2%), 70–180 mg/dL: 76.1±8.0%, and ≥250 mg/dL: 4.5±3.6%. After both the 100% and 130% boluses, postprandial percentage time <70 mg/dL was 0.0% (0.0±0.0%) (P=0.5). After the 100% and missed boluses, postprandial percentage time ≥250 mg/dL was 0.2±0.6% and 10.3±16.5%, respectively (P=0.06). Postprandial percentages time ≥250 mg/dL and <70 mg/dL were similar with standard or extended boluses for a high‐fat meal.
Conclusions
The Omnipod personalized MPC algorithm performed well and was safe during day and night use in response to overestimated, missed, and extended meal boluses in adults with T1D.
These two studies, performed under medical staff supervision in an observed setting, build upon the initial safety study of the Omnipod personalized MPC algorithm (8) to test the system with additional challenges.
In the first study, the system proved effective in minimizing hypoglycemia after exercise and overnight with either an increased temp target of 150 mg/dL or a reduced basal rate of 50% started 90 minutes before exercise. Thus, both methods of exercise announcement were found equal in effectiveness. Still there is a need to find ways to overcome the risk of hypoglycemia triggered by spontaneous exercise and to determine how long before the planned exercise adjustments are needed to prevent hypoglycemia; this time duration is probably individual and depends on the nature of the physical activity.
In the second study, postprandial hypoglycemia <70 mg/dL was avoided for 83% (10/12) of subjects after the 130% overestimated bolus. In the case of a missed 50 gram of carbohydrates meal bolus, percentage of time glucose ≥250 mg/dL during the postprandial period was only 10.3%—lower than reported in prior studies with a missed meal bolus. Of note, there was no significant difference in overall postprandial algorithmic insulin delivery after the 100% and 130% meal boluses, suggesting the automated system improved the timing and distribution of insulin delivery after the meal. An interesting finding was related to the treatment of high fat meal. High fat meals may have delayed gastric emptying; therefore, subjects using pump therapy (open loop) use extended bolus to overcome the delayed raise in glucose levels. The authors showed in this study that extended or standard bolus for high fat meal yield similar glycemic outcomes during closed‐loop use.
The results of these two studies further support the development of the Omnipod personalized MPC algorithm for eventual outpatient studies.
Closed loop control in adolescents and children during winter sports: use of the Tandem Control‐IQ AP system
Ekhlaspour L1, Forlenza GP2, Daniel Chernavvsky3, Maahs DM1,4, Wadwa RP2, Deboer MD3, Messer LH2, Town M1, Pinnata J3, Kruse G5, Kovatchev BP3, Buckingham BA1,4, Breton MD3
1Department of Pediatrics, Stanford University, Palo Alto, CA; 2Barbara Davis Center for Childhood Diabetes, University of Colorado Denver, Aurora, CO; 3Center for Diabetes Technology, University of Virginia, Charlottesville, VA; 4Stanford Diabetes Research Center, Stanford, VA; 5Tandem Diabetes Care, San Diego, CA
Background
Closed‐loop systems have been shown to improve glycemic control throughout both day and night in a wide range of populations including adults, adolescents, and children. Testing these systems during intense and prolonged exercise in adolescents and children remains limited. The authors present the performance of the Tandem Control‐IQ artificial pancreas (AP) system in adolescents and children during a winter ski camp study, where high altitude, low temperature, prolonged intense activity, and stress challenged glycemic control.
Methods
This was a randomized controlled trial, included 24 adolescents (ages 13–18 years) and 24 school‐aged children (6–12 years) with T1D participated in a 48 hours ski camp (∼5 hours skiing/day) at three sites: Wintergreen, VA; Kirkwood, and Breckenridge, CO. Study participants were randomized 1:1 at each site. The control group used remote monitored sensor‐augmented pump, and the experimental group used the t:slim X2 with Control‐IQ Technology AP system. All subjects were remotely monitored 24 hours per day by study staff.
Results
The percentage of time within range (70–180 mg/dL) over the entire camp duration was significantly higher in the AP compared with the SAP group, 66.4±16.4 versus 53.9±24.8% (P = 0.01) in both children and adolescents. The Control‐IQ system was associated with a significantly lower average sensor glucose: 161±29.9 versus 176.8±36.5 mg/dL (P=0.023). There were no differences between groups for hypoglycemia exposure or carbohydrate interventions. There were no adverse events in both groups.
Conclusions
The use of the Control‐IQ AP was found to improve glycemic control and safely reduced exposure to hyperglycemia relative to remote monitored sensor‐augmented pump in pediatric patients with T1D during prolonged intensive winter sport activities.
Successful at‐home use of the Tandem Control‐IQ artificial pancreas system in young children during a randomized controlled trial
Forlenza GP1, Ekhlaspour L2, Breton M3, Maahs DM2, Wadwa RP1, DeBoer M3, Messer LH2, Town M2, Pinnata J3, Kruse G4, Buckingham BA2, Cherñavvsky D3
1Barbara Davis Center for Childhood Diabetes, University of Colorado Denver, Aurora, CO; 2Department of Pediatrics, Stanford Diabetes Research Center, Stanford, CA; 3Center for Diabetes Technology, University of Virginia, Charlottesville, VA; 4Tandem Diabetes Care, San Diego, CA
Background
Hybrid closed‐loop systems are now moving from research settings to widespread clinical use. In this study, the inControl algorithm developed by TypeZero Technologies was embedded to a commercial Tandem t:slim X2 insulin pump, now called Control‐IQ, paired with a Dexcom G6 continuous glucose monitor and tested for superiority against SAP therapy. Both groups were monitored by a physician throughout the clinical trial.
Methods
In a randomized controlled trial, 24 school‐aged children (6–12 years) with T1D participated in a 3‐day home‐use trial at two sites: Stanford University and the Barbara Davis Center (50% girls, 9.6±1.9 years of age, 4.5±1.9 years of T1D, baseline HbA1c 7.35%±0.68%). Study subjects were randomized 1:1 at each site to either HCL AP therapy with the Control‐IQ system or SAP therapy with remote monitoring.
Results
The primary outcome, time in target range 70–180 mg/dL, using Control‐IQ significantly improved (71.0%±6.6% vs. 52.8%±13.5%; P=0.001) as did mean sensor glucose (153.6±13.5 vs. 180.2±23.1 mg/dL; P=0.003), without increasing hypoglycemia time <70 mg/dL (1.7% [1.3% to 2.1%] vs. 0.9% [0.3% to 2.7%]; not significant). The HCL system was active for 94.4% of the study period. Subjects reported that use of the system was associated with less time thinking about diabetes, decreased worry about blood sugars, and decreased burden in managing diabetes.
Conclusions
The use of the Tandem t:slim X2 with Control‐IQ HCL AP system significantly improved time in range and mean glycemic control without increasing hypoglycemia in school‐aged children with T1D during remotely monitored home use.
These two studies build upon initial safety evaluation of the Tandem Control‐IQ HCL system (9), which uses a Dexcom G6 CGM connected to the Tandem X2 insulin pump. In these studies, the system was tested during winter sports, similar to a prior study evaluating the same algorithm that used a phone based HCL system in a ski camp (10). The results were very positive, with improved time in range and lower mean glucose with Control‐IQ compared to SAP. There was a small reduction in insulin use in the beginner skier group using closed‐loop control (CLC; 51.69 vs. 53.67 units) but a 32% reduction in the advanced skier group (30.1 vs. 44.03 units total insulin during the day). Significant amounts of additional carbohydrates were still consumed by both the SAP and Control‐IQ groups (∼50 g/day). The second study of Control‐IQ showed that in younger children using the system at home for 3 days, there was an increase time in range (71% vs. 52.8%) and lower mean sensor glucose (153.6 vs. 180.2 mg/dL) compared to SAP. Survey results showed use of the system was associated with less time thinking about diabetes, decreased worry, and decreased diabetes burden. In both studies, Dexcom Share was used by investigators to provide additional safety monitoring in these challenging conditions. Thus, team monitoring might have slightly improved outcomes for both groups. These studies set the stage for the longer outpatient trial of Control‐IQ that was recently published (1).
Safety evaluation of the MiniMed 670G system in children 7–13 years of age with type 1 diabetes
Forlenza GP1, Pinhas‐Hamiel O2, Liljenquist DR3, Shulma DI4, Bailey TS5, Bode BW6, Wood MA7, Buckingham BA8, Kaiserman KB9, Shin J10, Huang S10, Lee SW10, Kaufman FR10
1Barbara Davis Center for Childhood Diabetes, Aurora, CO; 2Edmond and Lily Safra Children's Hospital, Sheba Medical Center, Tel Aviv, Israel; 3Rocky Mountain Diabetes and Osteoporosis Center, Idaho Falls, ID; 4USF Diabetes Center, Morsani College of Medicine, University of South Florida, Tampa, FL; 5AMCR Institute, Escondido, CA; 6Atlanta Diabetes Associates, Atlanta, GA; 7University of Michigan Medical School, Ann Arbor, MI; 8Department of Pediatric Endocrinology, Stanford University, Stanford, CA; 9SoCal Diabetes, Torrance, CA; 10Medtronic, Northridge, CA
This manuscript is also discussed in the article on Diabetes Technology and Therapy in the Pediatric Age Group, page S‐89.
Background
To evaluate the safety of in‐home use of the MiniMed 670G system with SmartGuard technology in children with T1D.
Methods
Participants (n=105, ages 7–13 years, mean age 10.8±1.8 years) were enrolled at nine centers (eight in the United States and one in Israel) and completed a 2‐week baseline run‐in phase in Manual Mode followed by a 3‐month study phase with Auto Mode enabled. Sensor glucose (SG), HbA1c, percentage of SG values across glucose ranges, and SG variability, during the run‐in and study phases were compared. Participants underwent frequent sample testing with i‐STAT venous reference measurement during a hotel period (6 days/5 nights) to evaluate the system's continuous glucose monitoring performance.
Results
Auto Mode was used a median of 81% of the time. From baseline to end of study, overall SG dropped by 6.9±17.2 mg/dL (P<0.001), HbA1c decreased from 7.9±0.8% to 7.5±0.6% (P<0.001), percentage of time in target glucose range (70–180 mg/dL) increased from 56.2%±11.4% to 65.0%±7.7% (P<0.001), and the SG coefficient of variation decreased from 39.6±5.4% to 38.5±3.8% (P=0.009). The percentage of SG values within target glucose range was 68.2±9.1% and that of i‐STAT reference values was 65.6±17.7%. The percentage of values within 20%/20 of the i‐STAT reference was 85.2%. There were no episodes of severe hypoglycemia or diabetic ketoacidosis during the study phase.
Conclusions
In‐home use of MiniMed 670G system Auto Mode for 3 months by children with T1D, similar to MiniMed 670G system use by adolescents and adults with T1D, was safe and associated with reduced HbA1c levels and increased time in target glucose range compared with baseline.
Most studies of closed‐loop systems included both adolescent and adult populations. As the system was found to be safe and effective at the former age groups, its use is expanding to younger age groups. This study, very similar in structure to the prior 3‐month trial of the 670G Automode in adults and adolescents (11), had children age 7–13 years of age perform a 2 week run‐in period followed by 3 months of Automode use. In this study, the Automode feature was used a median of 80.6% of the time, and the CGM was used a median of 90.9% of the time. Similar to the results of the prior study, this study showed time in target glucose range 70–180 mg/dL increased from 56% to 65%, sensor glucose variability decreased, and HbA1c decreased from 7.9% to 7.5%. Percent time <70 mg/dL also decreased from 4.7 to 3.0%. Mean ARD between SG and i‐STAT reference values was 11.9%. Although a single arm uncontrolled study, this study showed Automode in the 670G was safe and effective in a large group of participants aged 7–13 years old.
Predictive low‐glucose suspend reduces hypoglycemia in adults, adolescents, and children with type 1 diabetes in an at‐home randomized crossover study: results of the PROLOG trial
Forlenza GP1, Li Z2, Buckingham BA3, Pinsker JE4, Cengiz E5, Wadwa RP1, Ekhlaspour L3, Church M4, Weinzimer SA5, Jost E1, Marcal T3, Andre C4, Carria L5, Swanson V6, Lum JW2, Kollman C2, Woodall W2, Beck RW2
1Barbara Davis Center for Diabetes, University of Colorado Denver, Aurora, CO; 2Diabetes Study Group, Jaeb Center for Health Research, Tampa, FL; 3Division of Pediatric Endocrinology and Diabetes, Stanford University, Stanford, CA; 4Sansum Diabetes Research Institute, Santa Barbara, CA; 5Division of Pediatric Endocrinology and Diabetes, Yale University, New Haven, CT; 6Clinical Affairs, Tandem Diabetes Care, San Diego, CA
This manuscript is also discussed in the article on Diabetes Technology and Therapy in the Pediatric Age Group, page S‐89.
Background
This study evaluated a new insulin delivery system designed to reduce insulin delivery when trends in CGM glucose concentrations predict future hypoglycemia.
Methods
A 6‐week randomized crossover trial was conducted in patients with T1D (n=103, age 6–72 years, mean HbA1c 7.3%) to evaluate the safety and effectiveness of a Tandem Diabetes Care t:slim X2 pump with Basal‐IQ integrated with a Dexcom G5 sensor and a predictive low‐glucose suspend algorithm (PLGS) compared with SAP therapy. The primary outcome was CGM‐measured time <70 mg/dL.
Results
Ninety‐nine percent of participants completed the two study periods, and median CGM usage exceeded 90% in both arms. The median time spent below 70 mg/dL was significantly reduced from 3.6% at baseline to 2.6% during the 3‐week period in the PLGS arm versus 3.2% in the SAP arm (difference PLGS − SAP=0.8% [95% CI −1.1 to −0.5], P<0.001). The corresponding mean values were 4.4%, 3.1%, and 4.5%, respectively, representing a 31% reduction in the time <70 mg/dL with PLGS. Mean glucose concentration (159 vs. 159 mg/dL, P=0.40) and percentage of time spent >180 mg/dL (32% vs. 33%, P=0.12) did not differ between the two arms. One severe hypoglycemic event occurred in the SAP arm and none in the PLGS arm. Mean pump suspension time was 104 min/day.
Conclusions
The Tandem Diabetes Care Basal‐IQ PLGS system significantly reduced hypoglycemia without rebound hyperglycemia, indicating that the system can improve glycemic control on both adult and pediatric patients with T1D.
The Tandem Basal‐IQ PLGS System, first released commercially in August 2018, showed a 31% reduction in percent time below 70 mg/dL compared with SAP therapy in this randomized crossover outpatient trial. This occurred without increasing hyperglycemia (mean glucose was unchanged), and a slight increase in time in the target glucose range 70–180 mg/dL (65 vs. 63%, P<0.001). This may be explained in part by the PLGS algorithm's differences from other systems, where insulin resumption occurs on the first glucose reading past the nadir, whereas other systems do not resume insulin delivery until the sensor glucose has increased above a specified threshold and/or include a future predicted glucose value increasing above a predefined threshold. Also, the default settings for system alerts for automated suspension and resumption of insulin are set to off so as to not disturb the user. High system usability scores were given by the trial participants (12). Early real‐world outcomes have recently been reported that confirm the Basal‐IQ system's efficacy in a much larger cohort (13,14).
Subpopulations
Hybrid closed‐loop control is safe and effective for people with type 1 diabetes who are at moderate to high risk for hypoglycemia
Anderson SM1, Buckingham BA2, Breton MD1, Robic JL1, Barnett CL1, Wakeman CA1, Oliveri MC1, Brown SA1, Ly TT2, Clinton PK2, Hsu LJ2, Kingman RS2, Norlander LM2, Loebner SE2, Reuschel‐DiVirglio S2, Kovatchev BP1
1Center for Diabetes Technology, University of Virginia, Charlottesville, VA; 2Division of Pediatric Endocrinology and Diabetes, Department of Pediatrics, Stanford University School of Medicine, Stanford, CA
Background
Typically, CLC studies excluded patients with significant hypoglycemia. The authors evaluated the effectiveness of HCL versus SAP in reducing hypoglycemia in this high‐risk population.
Methods
Forty‐four subjects with T1D (25 women, 19 men), 37±2 years of age, with HbA1c 7.4±0.2% (57±1.5 mmol/mol), diabetes duration 19±2 years, on insulin pump therapy, were enrolled at the University of Virginia (n=33) and Stanford University (n=11). Inclusion criteria included increased risk of hypoglycemia confirmed by 1 week of blinded CGM, randomized to 4 weeks of home use of either HCL or SAP. Primary and secondary outcomes were risk for hypoglycemia measured by the low blood glucose index (LBGI)/CGM‐based time in ranges.
Results
Values are reported as mean±SD. From baseline to the final week of study, LBGI decreased more on HCL (2.51±1.17 to 1.28±0.5) than on SAP (2.1±1.05 to 1.79±0.98), P<0.001; percent time below 70 mg/dL (3.9 mmol/L) decreased on HCL (7.2%±5.3% to 2.0%±1.4%) but not on SAP (5.8%±4.7% to 4.8%±4.5%), P=0.001; percent time within the target range 70–180 mg/dL (3.9–10 mmol/L) increased on HCL (67.8%±13.5% to 78.2%±10%) but decreased on SAP (65.6%±12.9% to 59.6%±16.5%), P<0.001; percent time above 180 mg/dL (10 mmol/L) decreased on HCL (25.1%±15.3% to 19.8%±10.1%) but increased on SAP (28.6%±14.6% to 35.6%±17.6%), P=0.009. Mean glucose did not change significantly on HCL (144.9±27.9 to 143.8±14.4 mg/dL [8.1±1.6 to 8.0±0.8 mmol/L]) or SAP (152.5±24.3 to 162.4±28.2 [8.5±1.4 to 9.0±1.6]), P=NS.
Conclusions
Compared with SAP therapy, HCL reduced the risk and frequency of hypoglycemia, while improving time in target range and reducing hyperglycemia in people at moderate to high risk of hypoglycemia.
This study focused on the subpopulation of people with T1D that were at moderate to high risk of hypoglycemia. This vulnerable population may benefit from closed‐loop systems but have been previously excluded from these types of studies. Baseline time at <70 mg/dL (3.9 mmol/L) was higher than in most previously studied cohorts. The experimental and control groups were matched by predefined LBGI categories. The system used a phone based HCL system with a remote monitoring server. HCL significantly reduced hypoglycemia while improving time in target glucose range. Carbohydrate intake was not reported. These results support the safety and beneficial outcomes of HCL therapy for those with T1D at moderate to high risk of hypoglycemia. The next step is to work toward tracking and reducing the need for carbohydrate supplementation, test the system with minimal or no remote monitoring, and to evaluate the closed‐loop system compared to PLGS.
DREAM5: an open‐label, randomized, cross‐over study to evaluate the safety and efficacy of day and night closed‐loop control by comparing the MD‐Logic automated insulin delivery system to sensor augmented pump therapy in patients with type 1 diabetes at home
Biester T1, Nir J2, Remus K1, Farfel A2, Ido Muller I3, Biester S1, Atlas E3, Dovc K4, Bratina N4, Kordonouri O1, Battelino T4, Danne T1, Nimri R2
1Children's Hospital Auf der Bult, Diabetes‐Center for Children and Adolescents, Hannover, Germany; 2Jesse Z and Sara Lea Shafer Institute for Endocrinology and Diabetes, National Center for Childhood Diabetes, Schneider Children's Medical Center of Israel, Petah Tikvah, Israel; 3DreaMed Diabetes Ltd, Petah Tikvah, Israel; 4Department of Pediatric Endocrinology, Diabetes and Metabolic Diseases, University Children's Hospital, University Medical Centre Ljubljana, Ljubljana, Slovenia
Background
Previous diabetes risk evaluation and microalbuminuria (DREAM) studies evaluated the safety and efficacy of the MD‐Logic closed‐loop system (DreaMed GlucoSitter) for overnight glycemic control in various settings. This focused on evaluating the system for day and night use over 60 weekend hours at home compared with SAP therapy in participants with T1D.
Methods
A prospective, multicenter, crossover, controlled study was conducted. Participants were randomly assigned to either one weekend using SAP therapy (control) or one weekend on the MD‐Logic System (intervention). In the intervention arm, the amount of carbohydrate consumed was entered into the bolus calculator; a tablet computer was used to wirelessly automate the rest of insulin delivery. The primary endpoint was percentage of glucose values between 70–180 mg/dL.
Results
The intention‐to‐treat population consisted of 48 (19 males, 29 females) adolescents and adults experienced in sensor use. Median (IQR) age was 16.1 years (13.2–18.5); diabetes duration, 9.4 years (5.0–12.7); pump use, 5.4 years (3.1–9.4); and HbA1c, 7.6% (7.0–8.1). Percentage time within the target range (70–180 mg/dL) increased significantly from baseline (66.6% vs 59.9%, P=0.002) with the closed‐loop system versus control group, where it was unchanged (2.3% vs. 1.5%, P=0.369). A significant reduction in mean weekend glucose level per participant was also observed (153 [142–175] vs 164 [150–186] mg/dL, P=0.003). No safety signals were recorded.
Conclusions
Compared with SAP therapy, the MD‐Logic system was safe and associated with better glycemic control for day and night use. The absence of remote monitoring did not lead to safety signals in adapting basal rates or in administration of automated bolus corrections.
This study was the first evaluation of the MD‐Logic system at home during the daytime and overnight, rather than overnight only, for people with T1D. In addition to increased time in the target glucose range, there was a significant decrease in percent time above 180 mg/dL. A strength of the MD‐Logic system is its ability to detect the effect of meals and to respond accordingly by delivering insulin as a bolus, rather than responding by basal rate modulation alone, enabling the system to improve daytime glycemic control. In addition, after a pilot period of overnight remote monitoring for the first 10 subjects, the trial was done without remote monitoring, showing the system was safe to use without remote monitoring in this population of adolescents and young adults.
Fully closed‐loop insulin delivery in inpatients receiving nutritional support: a two‐centre, open‐label, randomised controlled trial
Boughton CK1,2, Bally L3, Martignoni F3, Hartnell S2, Herzig D3, Vogt A4, Wertli MM5, Wilinska ME1, Evans ML1,2, Coll AP1,2, Stettler C3, Hovorka R1
1Wellcome Trust–Medical Research Council Institute of Metabolic Science, University of Cambridge, Cambridge, UK; 2Wolfson Diabetes and Endocrine Clinic, Cambridge University Hospitals NHS Foundation Trust Cambridge, Cambridge, UK; 3Department of Diabetes, Endocrinology, Clinical Nutrition and Metabolism, Bern University Hospital, Bern, Switzerland; 4Department of Anaesthesiology and Pain Medicine, Inselspital, Bern University Hospital, Bern, Switzerland; 5Department of General Internal Medicine, Bern University Hospital, Bern, Switzerland
Background
Managing glucose levels is challenging in patients who require nutritional support in hospital. The focus of this study was to assess whether fully closed‐loop insulin delivery would improve glycemic control compared with conventional subcutaneous insulin therapy in inpatients receiving enteral or parenteral nutrition or both.
Methods
A two‐center (UK and Switzerland), open‐label, randomized controlled trial was carried out in adult inpatients who required subcutaneous insulin therapy and were receiving enteral nutrition, parenteral nutrition, or both. Participants were recruited from noncritical care surgical and medical wards. A computer‐generated minimization schedule categorized by type of nutritional support (parenteral nutrition on or off) and prestudy total daily insulin dose (<50 or ≥50 units) was used to randomly assign (1:1) patients to receive fully closed‐loop insulin delivery with faster‐acting insulin aspart (closed‐loop group) or conventional subcutaneous insulin therapy (control group) given in accordance with local clinical practice. Patients, ward staff, and investigators were blinded to CGM in the control group. Participants received a maximum follow‐up of 15 days or until hospital discharge. The primary endpoint was the proportion of time that sensor glucose concentration was in target range (5.6–10.0 mmol/L), assessed in the intention‐to‐treat population.
Results
Ninety patients were assessed for eligibility, of whom 43 were enrolled and randomly assigned to the closed‐loop group (n=21) or the control group (n=22). Sensor glucose was within the desired target range 68.4±15.5% of the time in the closed‐loop group and 36.4±2.6% of the time in the control group (difference 32.0 percentage points [95% CI 18.5–45.5]; P<0.0001). One serious adverse event occurred in each group (one cardiac arrest in the control group and one episode of acute respiratory failure in the closed‐loop group), both of which were unrelated to the study. No adverse events related to study interventions were reported in either group. No episodes of severe hypoglycemia or hyperglycemia with ketonemia occurred in either study group.
Conclusions
Closed‐loop insulin delivery is an effective treatment option to improve glycemic control in patients receiving nutritional support in hospital.
This study assessed the safety and efficacy of CLC in an inpatient hospital setting receiving parenteral or enteral nutrition, where glycemic disturbances could have significant adverse effects on wound healing, hospital stay, morbidity, and mortality. As the glucose disturbance is defined and measured in this population, the use of fully automated closed‐loop systems is a prudent choice that can improve clinical outcomes. Subjects in the closed‐loop group had a higher percentage of time with glucose in the target range and less time above range than the control group. There were four hypoglycemic episodes in the closed‐loop group compared with nine in the control group. Notably, the time in range overnight was significantly higher for subjects in the closed‐loop group. This study shows that using closed‐loop insulin delivery for patients receiving parenteral or enteral nutrition is safe and effective and is similar to their prior study of CLC in the non‐critical care setting (15), shows CLC use can be very effective in the inpatient hospital setting. Thus, the use of a closed‐loop system in such patients may reduce the complexity of treatment. Nevertheless, the closed loop was compared with a control group that the treating medical team could not see, nor could they use the sensor glucose levels to modify insulin therapy (blinded CGM for the control group). Future studies are needed to confirm the utility of closed loop in such settings and to assess whether closed‐loop control can be translated into improved clinical outcomes.
Excellent glycemic control maintained by open‐source hybrid closed‐loop AndroidAPS during and after sustained physical activity
Petruzelkova L1, Soupal J2, Plasova V1, Jiranova P1, Neuman V1, Plachy L1, Pruhova S1, Sumnik Z1, Obermannova B1
1Department of Pediatrics, University Hospital Motol and Second Faculty of Medicine, Charles University in Prague, Prague, Czech Republic; 2Third Department of Internal Medicine, First Faculty of Medicine, Charles University in Prague, Prague, Czech Republic
This manuscript is also discussed in the article on Advances in Exercise, Physical Activity, and Diabetes, page S‐109.
Background
Officially licensed hybrid closed‐loop systems are not currently available worldwide; therefore, open‐source systems have become increasingly popular. The authors aimed to assess the safety, feasibility, and efficacy of an open‐source hybrid closed‐loop system (AndroidAPS) versus SmartGuard technology for day‐and‐night glucose control in children under extreme sports conditions.
Methods
Twenty‐two children (16 girls, 6–15 years of age, median HbA1c 56±9 mmol/mol) were enrolled in this pivotal winter sports camp study. The participants were divided into two groups using either the AndroidAPS or SmartGuard technology. Physical exertion was represented by all‐day alpine skiing. The primary endpoints were mean glucose level, time below the threshold of 3.9 mmol/L, and time within the target range of 3.9–10 mmol/L.
Results
The children using the AndroidAPS had significantly lower mean glycemia levels (7.2±2.7 vs. 7.7±2.8 mmol/L; 129.6±49 vs. 138.6±50 mg/dL, P<0.042) than the children using SmartGuard. The proportion of time below the target (median 5.0±2.5% vs. 3.0%±2.3%, P=0.6) and in the target zone (63±9.5% vs. 63±18%, P=0.5) did not significantly differ. The AndroidAPS group experienced more frequent malfunctions of the cannula set (median 0.8±0.4 vs. 0.2±0.4, P=0.02), which could have affected the results. No significant difference was found in the amount of carbohydrates consumed for the prevention and treatment of hypoglycemia [median 40±23 vs. 25±29 g/(patient·3 days)]. No episodes of severe hypoglycemia or other serious adverse events were noted.
Conclusions
This pilot study showed that the AndroidAPS system was a safe and feasible alternative to the SmartGuard Technology.
This study aimed to examine glycemic control using the AndroidAPS system (Android Smartphone with Android APS app, DANA Diabecare R pump, Dexcom G4™ Sensor, Xbridge which transmits the data from the sensor to the Smartphone) in 10 participants and using the commercially available SmartGuard Technology system with Predictive Low Glucose Management (PLGM) (Mini‐Med 640G pump, Enlite™ Sensor) in 12 participants under extreme sports conditions during a winter ski camp. Physical activity was represented by whole‐day downhill skiing (9 a.m. to 4 p.m.) and evening entertainment activities (disco dance, 8–11 p.m.). The experience level in downhill skiing was very good in both groups (all children were intermediate or advanced skiers). Overall, the children using AndroidAPS had significantly lower mean glucose levels than the children in the PLGM group, with similar levels of hypoglycemia and hypoglycemic events. Although there was constant supervision throughout the study and the study compared an AP system to PLGM, the results support the idea that do‐it‐yourself technology, in this case AndroidAPS, is a feasible tool for the optimization of T1D management during and after prolonged physical activity in children with T1D. As the number of people with T1D using these unapproved devices is increasing, physicians are increasingly confronted with T1D patients using them, and may need to be prepared to help them with setup and continued use, as well as troubleshooting connectivity and other common issues that may occur.
Next Generation Closed‐Loop
Design and clinical evaluation of the Interoperable Artificial Pancreas System (iAPS) smartphone app: interoperable components with modular design for progressive artificial pancreas research and development
Deshpande S1,2, Pinsker JE2, Zavitsanou S1,2, Shi D1,2, Tompot R3, Church MM2, Andre C1, Doyle III FJ1,2, Dassau E1,2,4
1Harvard John A. Paulson School of Engineering and Applied Sciences, Harvard University, Cambridge, MA; 2Sansum Diabetes Research Institute, Santa Barbara, CA; 3Labrasoft LLC, Leander, TX; 4Joslin Diabetes Center, Boston, MA
Background
There is an unmet need for a modular AP system for clinical trials within the existing regulatory framework to further AP research projects from both academia and industry. The authors designed and evaluated the interoperable artificial pancreas system (iAPS), a smartphone app that interfaces wirelessly with leading CGMs, insulin pump devices, and decision‐making algorithms while running on an unlocked smartphone.
Methods
After algorithm verification, hazard and mitigation analysis, and complete system verification of iAPS, 6 adults with type 1 diabetes completed 1 week of SAP use followed by 48 hours of AP use with the iAPS, a Dexcom G5 CGM, and either a Tandem or Insulet insulin pump in a U.S Food and Drug Administration (FDA)–approved investigational device exemption study. The AP system was challenged by participants performing extensive walking without exercise announcement to the controller, multiple large meals eaten out at restaurants, two overnight periods, and multiple intentional connectivity interruptions.
Results
Even with these intentional challenges, comparison of the SAP phase with the AP study showed a trend toward improved time in the target glucose range of 70–180 mg/dL (78.8% vs. 83.1%; P=0.31) and a statistically significant reduction in time below 70 mg/dL (6.1% vs. 2.2%; P=0.03). The iAPS system performed reliably and showed robust connectivity with the peripheral devices (99.8% time connected to CGM and 94.3% time in closed loop) while requiring limited user intervention.
Conclusions
The iAPS system was safe and effective in regulating glucose levels under challenging conditions and is suitable for use in unconstrained environments.
The various AP systems in development can be broadly grouped as those using dedicated embedded hardware or those relying on a dedicated locked‐down smartphone device (16), or some combination of the two. While the use of dedicated embedded hardware ameliorates issues related to connectivity, it is a less modular system for research purposes. A smartphone‐based solution provides modularity and access to software updates, which has enabled several outpatient studies and has moved the AP field forward. iAPS provides a path forward for testing and clinical evaluation of an interoperable AP with modern peripheral devices and state‐of‐the‐art algorithms. Use of an unlocked smartphone to control the pump, as used here, can result in better user acceptance and improved clinical outcomes, with the possibility of the AP system smartphone being the same device as the user's regular smartphone. This may improve QoL by allowing patients to carry only one phone devices (their own phone) to use for bolusing and other diabetes care activities.
A new animal model of insulin‐glucose dynamics in the intraperitoneal space enhances closed‐loop control performance
Chakrabarty A1, Gregory JM2, Moore LM3, Williams PE4, Farmer B3, Cherrington AD3, Lord P5, Shelton B5, Cohen D5, Zisser HC6, Doyle III FJ1, Dassau E1
1Harvard John A. Paulson School of Engineering and Applied Sciences, Harvard University, Cambridge, MA; 2Vanderbilt University Medical Center, Nashville, TN; 3Department of Molecular Physiology and Biophysics, Vanderbilt University School of Medicine, Nashville, TN; 4Section of Surgical Sciences, Vanderbilt University School of Medicine, Nashville, TN; 5Physiologic Devices, Inc., Alpine, CA; 6Department of Chemical Engineering, University of California Santa Barbara, Santa Barbara, CA
Background
Current AP systems operate via subcutaneous (SC) glucose sensing and insulin delivery. Slow diffusion and transport dynamics across the interstitial space affects even the most sophisticated control algorithms in on‐body AP systems, which cannot react quickly enough to maintain tight glycemic control when ingesting meals or performing physical activity causes exogenous glucose disturbances. Recent progress in the development of an implantable AP includes the possibility of insulin infusion in the intraperitoneal (IP) space, where observed insulin–glucose kinetics are much faster than in the SC space.
Methods
A series of canine experiments were carried out to determine the dynamic association between IP insulin boluses and plasma glucose levels. Information from these experiments is being exploited to create a new mathematical model and to formulate a closed‐loop control strategy to be used in an implantable AP.
Results
These in silico experiments on an FDA‐accepted benchmark cohort demonstrate the possibilities of the proposed controller for implantable AP. The new controller design demonstrated a significant improvement over a previous design that used artificial data (time in clinically acceptable glucose range: 97.3±1.5% vs. 90.1±5.6%). Furthermore, this study investigated the ability of the recently proposed closed‐loop system to perform well despite delays and noise in the measurement signal (e.g., when glucose is sensed subcutaneously) and deleterious glycemic changes (e.g., sudden drop in glucose physical activity).
Conclusion
The proposed model based on experimental canine data leads to the generation of more effective control algorithms and is a promising step toward fully automated and implantable AP systems.
Prior studies using IP insulin delivery have shown promising results (17), particularly trying to improve upon the difficult challenge of preventing postprandial hyperglycemia since SC insulin delivery is often too slow to prevent elevated peak glucose levels after meals. In this study, the authors developed a new model of insulin delivery for the IP space based on canine data, and then evaluated the model in the UVA/ Padova FDA accepted metabolic simulator with unannounced meals and IP sensing and insulin delivery, as well as SC sensing and IP insulin delivery. The very high time in target range in this study shows that an AP system capable of leveraging the rapid sensing and actuation dynamics in the IP space could one day become a completely automated glucose management system for people with T1D.
Incorporating unannounced meals and exercise in adaptive learning of personalized models for multivariable artificial pancreas systems
Hajizadeh I1, Rashid M1, Turksoy K2, Samadi S1, Feng J1, Sevil M2, Hobbs N2, Lazaro C3, Maloney Z2, Littlejohn E4, Cinar A1,2
1Department of Chemical and Biological Engineering, Illinois Institute of Technology, Chicago, IL; 2Department of Biomedical Engineering, Illinois Institute of Technology, Chicago, IL; 3Department of Electrical and Computer Engineering, Illinois Institute of Technology, Chicago, IL; 4Department of Pediatrics and Medicine, Section of Endocrinology, Kovler Diabetes Center, University of Chicago, Chicago, IL
Background
Despite the recent advancements in the modeling of glycemic dynamics for T1D, many models rely on real‐time CGM feedback, which can decrease their ability to automatically adjust for unannounced meals and exercise without manual inputs from the patient/user. Generalized models may not accurately predict individual physiology or glucose excursions due to meals or activity.
Methods
This paper proposes an adaptive model identification process that incorporates exercise information and estimates of the effects of unannounced meals automatically without user input. The model estimates the effects of unknown consumed carbohydrates via an individualized unscented Kalman filtering algorithm employing an augmented glucose–insulin dynamic model, and exercise data is obtained from noninvasive physiological measurements. The proposed adaptive model identification algorithm integrates the additional meal and exercise data with personalized estimates of plasma insulin concentration.
Results
Clinical data from closed‐loop trials of the artificial pancreas system is employed to demonstrate the efficacy of the proposed personalized, adaptive modeling algorithm. The results of this experiment demonstrated accurate glycemic modeling with the average root‐mean‐square error (mean absolute error) of 25.50 mg/dL (18.18 mg/dL) for six‐step (30 minutes ahead) predictions.
Conclusions
The approach presented is capable of identifying dependable time‐varying individualized glucose–insulin models.
A major challenge to achieving a fully automated and reliable AP is the lack of an accurate model to represent the dynamic changes in physiology under various conditions. The authors' proposed adaptive and personalized modeling approach that considers the effects of unannounced meals and exercise on the transient glycemic dynamics was applied to 15 clinical data sets involving closed‐loop experiments. While collecting these data sets, subjects wore the BodyMedia SenseWear Pro3 (BodyMedia, Pittsburgh, PA) armband reporting physiological signals for over 60 hours. The results demonstrate that the use of the plasma insulin concentration as a filtered insulin input, the incorporation of estimates of the meal effect obtained through the nonlinear observer algorithm, and the inclusion of energy expenditure as an auxiliary input from the BodyMedia SenseWear armband improve the identified glucose–insulin dynamic models. The next step is finding the best way to integrate these additional signals into clinical use, as this is still a difficult challenge. With time in range approaching 70+ percent in many closed‐loop trials, it is still unclear what effect adding additional signals will have on time in range, and how the complexity of adding these additional signals will add or detract from the experience of wearing closed‐loop systems.
Summary
This year's studies confirm the continuous advances in the efficacy of both closed‐loop and clinical decision support systems on glycemic control. Published results from closed‐loop studies show clear improvements in glycemic control during the overnight period, as well as the daytime period in multiple different populations, including children, adolescents, those at high risk of hypoglycemia, and inpatients receiving nutritional support. Different systems were introduced with different combinations of pump and sensor, expanding the available closed‐loop options for people with diabetes.
Additionally, studies highlighting the effects of introducing decision support systems for patients on MDI therapy with type 1 and type 2 diabetes have now been published. Most of these systems are still in their early stages of development. A number of companies have announced their plans to develop and implement decision support systems in the form of smart pens (18,19); studies using these systems are still ongoing. These systems have great potential to empower people with chronic conditions to live better and healthier lives. Decision support systems could be the solution for frequent insulin titration between clinical visits and can be used in telemedicine, and therefore to bring expert knowledge to rural areas and avoid loss of work/school days. In addition, these systems have the potential to solve several critical problems in the healthcare system today, such as the shortage of endocrinologists, physician burnout, and more.
Advanced technology use will happen if it is seamless to employ, improves QoL, and has high patient satisfaction. Letting patients use a smartphone app on their own device to connect to or run the closed‐loop system, for example, is one way to achieve this. It is encouraging to see numerous publications this year addressing reduced burden of care and QoL (20 –22), as this will be a key issue moving forward.
Footnotes
Author Disclosure Statement
R.N. has received devices support for clinical studies from Medtronic, Dexcom, Abbott and Insulet Corporation. R.N. received honoraria for participating in the speaker's bureau of Novo Nordisk, Pfizer, Eli Lilly, and Sanofi. R.N. owns DreaMed stock and reports two patent applications.
J.E.P. has received grant support, consulting fees, and speakers bureau fees from Tandem Diabetes Care; grant support from Medtronic; grant support and consulting fees from Eli Lilly; grant support and product support, provided to his institution, from Insulet; and product support, provided to his institution, from Dexcom.
E.D. has received product support from Tandem Diabetes Care, Insulet Corporation, Dexcom, and Xeris; speaker's bureau fees from Roche; and consulting fees from Eli Lilly. He has several patents with royalties.
All other authors report no competing financial interest.