Journal of Diabetes Science and Technology 2014, Vol. 8(1) 35 –42
© 2014 Diabetes Technology Society Reprints and permissions: sagepub.com/journalsPermissions.nav DOI: 10.1177/1932296813511730 dst.sagepub.com
Original Article
Developing a closed-loop insulin delivery system has the potential to transform diabetes management and improve the lives of those living with the disease. Since JDRF (formerly the Juvenile Diabetes Research Foundation) launched the Artificial Pancreas Project in 2006,1 various studies have reported the feasibility of different closed-loop algorithms to control blood glucose levels in patients with type 1 diabe-tes.2-6 Although much technical progress has been made, patients with type 1 diabetes carry a significant burden in the completion of day-to-day diabetes care tasks, and run the continual risk of hypoglycemia and hyperglycemia in their attempts to achieve optimal glucose control.7
Preliminary data from the Hypoglycemia-Hyperglycemia Minimizer (HHM) System, an automated insulin delivery device, are presented in this article. The control algorithm of the HHM System is designed to take action to reduce—if not prevent—glucose excursions outside of a prespecified target
zone by adjusting insulin delivery based on predictions of near-future glucose trends.
Following US Food and Drug Administration (FDA) guidance in preparation for human clinical trials, realistic aspects of the HHM System were investigated in silico, including protocol specifications and system characteristics.8 The “virtual cohort” investigated comprised 100 adult patients, and has been accepted by the FDA as suitable replacement for animal studies.9
1Animas Corporation, West Chester, PA, USA
2University of California, Santa Barbara, Santa Barbara, CA, USA 3Sansum Diabetes Research Institute, Santa Barbara, CA, USA 4University of Virginia, Charlottesville, VA, USA
Corresponding Author:
Daniel A. Finan, Animas Corporation, 965 Chesterbrook Blvd, Wayne, PA 19087, USA.
Email: [email protected]
Closed-Loop Control Performance of the
Hypoglycemia-Hyperglycemia Minimizer
(HHM) System in a Feasibility Study
Daniel A. Finan, PhD
1, Thomas W. McCann Jr, MBA
1,
Linda Mackowiak, MS
1, Eyal Dassau, PhD
2,3, Stephen D. Patek, PhD
4,
Boris P. Kovatchev, PhD
4, Francis J. Doyle III, PhD
2,3, Howard Zisser, MD
2,3,
Henry Anhalt, DO
1, and Ramakrishna Venugopalan, PhD
1Abstract
Background: This feasibility study investigated the insulin-delivery characteristics of the Hypoglycemia-Hyperglycemia Minimizer (HHM) System—an automated insulin delivery device—in participants with type 1 diabetes.
Methods: Thirteen adults with type 1 diabetes were enrolled in this nonrandomized, uncontrolled, clinical-research-center-based feasibility study. The HHM System comprised a continuous subcutaneous insulin infusion pump, a continuous glucose monitor (CGM), and a model predictive control algorithm with a safety module, run on a laptop platform. Closed-loop control lasted approximately 20 hours, including an overnight period and two meals.
Results: When attempting to minimize glucose excursions outside of a prespecified target zone, the predictive HHM System decreased insulin infusion rates below the participants’ preset basal rates in advance of below-zone excursions (CGM < 90 mg/dl), and delivered 80.4% less insulin than basal during those excursions. Similarly, the HHM System increased infusion rates above basal during above-zone excursions (CGM > 140 mg/dl), delivering 39.9% more insulin than basal during those excursions. Based on YSI, participants spent a mean ± standard deviation (SD) of 0.2 ± 0.5% of the closed-loop control time at glucose levels < 70 mg/dl, including 0.3 ± 0.9% for the overnight period. The mean ± SD glucose based on YSI for all participants was 164.5 ± 23.5 mg/dl. There were nine instances of algorithm-recommended supplemental carbohydrate administrations, and there was no severe hypoglycemia or diabetic ketoacidosis.
Conclusions: Results of this study indicate that the current HHM System is a feasible foundation for development of a closed-loop insulin delivery device.
Keywords
The design of the simulation study approximated the clin-ical study design as closely as possible, from both “system” and “protocol” perspectives.10 From a “system” perspective, the algorithm used in the simulation study was identical to that planned for the clinical study. From a “protocol” per-spective, two key parameters were intended to be adjustable throughout the clinical study to glean more targeted informa-tion: the ratios of prandial insulin boluses to the breakfast and lunch carbohydrate (CHO) quantities, relative to the nominal “matched” amounts based on the subject’s insulin-to-CHO ratios. The stated values of mismatch planned for the study were: (a) for breakfast, a matched bolus or 30% underbolus; and (b) for lunch, a matched bolus or 30% over-bolus. The rationale for this bolus mismatch stipulation was to ensure adequate glucose fluctuations and thereby investi-gate the algorithm’s insulin-dosing characteristics in response to potential hypoglycemic and hyperglycemic excursions. All stated values for these boluses were investigated in silico to demonstrate safety in the clinical study.
The results from the simulations did not indicate any safety concerns, regardless of the bolus mismatch parame-ters, and provided sufficient justification for moving forward with the clinical (feasibility) study.11,12
The feasibility study was executed to evaluate the perfor-mance of the HHM System in participants with type 1 diabe-tes in a clinical research center (CRC) setting. The primary objective of this study was to investigate the insulin delivery characteristics of the HHM System before and during glu-cose excursions below and above a prespecified target zone of 90-140 mg/dl. As the study was unpowered, targeted met-rics were designed and used to assess the potential of the HHM System to safely maintain glucose levels by adjusting insulin infusion amounts in quiescent times (including over-night) as well as in response to meal challenges.
Methods
Clinical Study Design
This was a nonrandomized, uncontrolled, CRC-based feasi-bility study. The study was conducted in accordance with the Declaration of Helsinki, International Conference on Harmonisation, Good Clinical Practice guidelines, and rele-vant local laws and regulations. The protocol and informed consent form were approved by Compass Independent Review Board, LLC (Mesa, AZ), and all participants pro-vided written informed consent.
The study objectives were to evaluate the performance of the HHM System in participants with type 1 diabetes in a CRC setting; investigate the insulin delivery characteristics of the HHM System in anticipation of, and during, glucose excursions below and above a prespecified target zone (90-140 mg/dl); and assess the ability of the system to safely maintain glucose levels by adjusting insulin infusion amounts in response to challenges of food and insulin boluses.
Subjects. The target cohort of this feasibility study was adults with type 1 diabetes using insulin pumps. Main inclu-sion and excluinclu-sion criteria are listed in Table 1. In addition, subjects with any other condition that, in the opinion of the investigator, would preclude successful participation in the clinical trial were excluded.
Investigational device. In this study the investigational device, the HHM System, comprised a continuous glucose monitor (CGM), insulin pump, automatic control algorithm, and central laptop (Figure 1).13
The algorithm of the HHM System had two modular com-ponents—the Zone Model Predictive Controller (Zone-MPC) and the Safety Supervision Module (SSM)—which worked to automatically and continuously dose insulin in response to changing CGM levels. The Zone-MPC algorithm used a mathematical approximation of insulin-glucose dynamics to predict near-future glucose trends from recent CGM measurements and insulin dosage amounts. CGM val-ues are automatically input to the algorithm every 5 minutes. The algorithm was designed to deliver insulin as needed to maintain glucose levels within the target zone (90-140 mg/ dl).14 The SSM used mathematical approximations of the insulin-glucose dynamics to continually assess and mitigate the risk of near-future hypoglycemia.15 It acted on the Zone-MPC’s recommended insulin infusion amount, and was designed to provide an additional safeguard against predicted near-future hypoglycemia. The action of the SSM was con-strained to approving or reducing (but not increasing) the insulin amount recommended by the Zone-MPC.
Procedure. During the week before the CRC visit, individ-ual participants’ basal rates were assessed and optimized by
Table 1. Main Inclusion and Exclusion Criteria for Participation. Inclusion criteria
• Age 21-65 years
• Diagnosed with type 1 diabetes for ≥ 1 year
• Using an insulin infusion pump for ≥ 6 months with
commercially available rapid-acting insulin
• Hemoglobin A1c < 10%
Exclusion criteria
• Pregnant/nursing
• Skin conditions that would preclude wearing study devices • Diabetic ketoacidosis requiring an emergency room visit or
hospitalization in the previous 6 months
• Severe hypoglycemia resulting in seizure or loss of
consciousness in the previous 6 months
• History of a seizure disorder
• Use of oral prednisone or a beta-blocker • Intolerant of or unable to receive glucagon • Hematocrit level < 30% or > 55%
the investigative staff. The individual basal rates were used by the algorithm of the HHM System as the “default” insulin delivery rates when CGM values were predicted to be within the target zone; by extension, when CGM values were (or were predicted to be) outside the target zone, these basal rates served as the basis insulin delivery rate from which the algo-rithm might increase or decrease insulin delivery. CGM sensor insertion occurred 2-3 days prior to the CRC visit and was cal-ibrated by meter throughout the study per the manufacturer’s instructions. The CRC visit lasted approximately 24 hours for each participant, including approximately 20 hours of closed-loop control during which the HHM System was assessed. The sequence of events during the CRC visit was as follows:
•
• The participant arrived during the early evening of day 1.
•
• Closed-loop control was initiated at approximately 00:00 (midnight), and lasted until approximately 20:00.
•
• The breakfast meal on day 2 was administered at approximately 07:00, and included 1 g CHO per kg body weight, up to a maximum of 100 g CHO. – Coincident with the breakfast meal, an insulin
bo-lus was administered, which in some cases was an underbolus by 30% relative to their insulin-to-CHO ratio.
– The purpose for this deliberate underinsulinization was to ensure adequate algorithm activity in miti-gating the subsequent (potential) hyperglycemia. •
• The lunch meal on day 2 was administered at approxi-mately 13:00, and also included 1 g CHO per kg body weight, up to a maximum of 100 g CHO.
– Coincident with the lunch meal, an insulin bolus was administered, which in some cases was an overbolus by 30%.
– Similar to the rationale for the underbolus at the breakfast meal, the purpose for the deliberate overinsulinization at lunchtime was to ensure ad-equate algorithm activity in mitigating the subse-quent (potential) hypoglycemia.
•
• The participant was discharged in the evening of day 2. Glucose monitoring was performed using YSI 2300 STAT Plus (YSI Inc, Yellow Springs, OH) at the following time points: before meals and every 2 hours prior to enabling loop control mode; every 30 minutes during closed-loop control mode; every 20 minutes for 80 minutes follow-ing meals; every 15 minutes if hypoglycemic, until glucose rose > 80 mg/dl; and every 30 minutes if hyperglycemic until glucose fell < 300 mg/dl.
All diabetes care in the CRC was performed by study staff. A follow-up telephone call was conducted 24 hours after the CRC visit.
Data Analysis
The ability of the HHM System to anticipate and to preemp-tively respond to predicted below-zone and above-zone excursions based on CGM values was assessed, as was its ability to respond during below-zone and above-zone excur-sions. These metrics quantified differences in the HHM System–delivered insulin rates compared to the correspond-ing basal rates in the 15 minutes prior to an out-of-zone breach (defined as when the CGM tracing exited the glucose target zone), as well as any time during an out-of-zone excursion.
Glucose control metrics were calculated, but should be considered with the caveat that the primary objective of the study was not to obtain optimal glucose control, but rather to investigate the insulin-delivery characteristics of the HHM System including during periods of over- and underinsulin-ization. Glucose metrics based on YSI and CGM were calcu-lated for four time ranges: overall, overnight (approximately 00:00-07:00), post-breakfast (approximately 07:00-13:00),
Figure 1. The investigational device comprised a continuous subcutaneous insulin infusion pump (OneTouch® Ping® Glucose
Management System; Animas Corporation, West Chester, PA), a continuous glucose monitoring system (Dexcom® SEVEN® PLUS
Continuous Glucose Monitor; Dexcom, Inc, San Diego, CA), and the Hypoglycemia-Hyperglycemia Minimizer algorithm run on the University of California, Santa Barbara/Sansum Diabetes Research Institute Artificial Pancreas System (APS©) laptop platform.11 RF,
and post-lunch (approximately 13:00-20:00). Glucose met-rics were assessed as the percentage of time spent in an approximately normoglycemic range (70-180 mg/dl), a hyperglycemic range (> 180 mg/dl), and a hypoglycemic range (< 70 mg/dl) while in closed-loop control. Other calcu-lated metrics included the mean glucose for all participants.
Results
Participant Characteristics
The characteristics of the 13 participants who completed the study are summarized in Table 2.
Insulin Infusion Characteristics of the HHM System
The HHM System took significant preemptive, mitigating action prior to below-zone excursions (defined as the 15 min-utes prior to a below-zone CGM breach) by delivering, on average, 77.2% less insulin than basal during these times (Table 3). The HHM System did not significantly increase dosing relative to basal in the times prior to above-zone excur-sions. During below-zone excursions, an average decrease of 80.4% in HHM System insulin infusion rates was observed compared with basal rates. During above-zone excursions an average increase of 39.9% in HHM System insulin infusion rates was observed compared with basal rates.11The algorithm response did not result in a significant increase or decrease in overall total daily dose relative to the participants’ basal/bolus therapy regimens, but rather in a redistribution of insulin delivery based on CGM data. There was a 10.6% increase in the average HHM System insulin infusion rate compared with the average basal rate. During the overnight period, the breakfast period, and the post-lunch period the average HHM System infusion rate was 6.6%, 18.8%, and 3.8% greater than the average basal rate, respectively. The HHM System’s insulin infusion rates stayed at approximately basal-level rates for 9.7% of the total time spent in closed-loop control, were higher than basal rates 47.4% of the time, and were lower 42.9% of the time.
Overall, the average HHM System insulin infusion rate was 27% lower than that recommended by the Zone-MPC, reflecting the ability of the SSM to attenuate the insulin dose recommendations to mitigate the risk of hypoglycemia. The reductions in infusion rates by the SSM were 33.1% in the overnight period, 22.6% in the post-breakfast period, and 20.6% in the post-lunch period.
Glucose Metrics
Glucose metrics were calculated including the periods of deliberate underinsulinization and overinsulinization. Based on YSI measurements, participants spent on average 0.2 ± 0.5% of the closed-loop control time at glucose levels < 70 mg/dl and 69.6 ± 24.7% of the closed-loop control time at glucose levels 70-180 mg/dl (Table 4). During the overnight period, participants spent 0.3 ± 0.9% of the time < 70 mg/dl and 81.8 ± 35.6% of the time between 70 and 180 mg/dl. During the post-breakfast and post-lunch periods, partici-pants spent 0.3 ± 1.1% and 0% of the time < 70 mg/dl, respectively.11 Based on CGM measurements, participants spent on average 0.5 ± 1.2% of the closed-loop control time at glucose levels < 70 mg/dl and 62.2 ± 20.2% of the closed-loop control time at glucose levels 70-180 mg/dl (Table 5). During the overnight period, participants spent 0.1 ± 0.3% of the time < 70 mg/dl and 80.0 ± 27.5% of the time between 70 and 180 mg/dl. During the post-breakfast and post-lunch periods, participants spent 1.5 ± 3.8% and 0% of the time < 70 mg/dl, respectively.11
The mean ± SD glucose value based on YSI was 164.5 ± 23.5 mg/dl; the mean ± SD based on CGM was 175.1 ± 27.3 mg/dl.12 For CGM excursions < 70 mg/dl (n = 3), the average area between the CGM curve and the threshold of 70 mg/dl was 0.35 mg/dl•day overall and 1.69 mg/dl•day for the post-breakfast period.12 For excursions > 180 mg/dl, the average area between the CGM curve and the threshold of 180 mg/dl was 19.5 mg/dl•day overall, 12.5 mg/dl•day for the over -night period, 30.0 mg/dl•day for the post-breakfast period, and 25.1 mg/dl•day for the post-lunch period.
Following initiation of closed-loop control, and leading up to the breakfast meal, there was an evident narrowing of the ± 1 SD band in the CGM tracings (Figure 2), indicating the HHM System’s ability to control glucose to near-normo-glycemic values during the overnight period. Based on YSI, the mean ± SD meal peaks for the breakfast and lunch meals were 241 ± 47 mg/dl and 200 ± 50 mg/dl, respectively, and based on CGM they were 265 ± 63 mg/dl and 224 ± 51 mg/ dl, respectively.
Safety Events
Under closed-loop control, there were a total of nine instances when the investigator administered algorithm-recommended CHO, which was 15 g of juice or glucose tablets (Table 6). Four of these instances occurred in one participant (partici-pant 12). There were no instances of investigator-initiated
Table 2. Patient Characteristics (N = 13).
Parameter Value
Age, years, mean ± SD 42.5 ± 12.7
Female, n (%) 11 (84.6)
Body mass index, kg/m2, mean ± SD 24.7 ± 5.1
Duration of diabetes, years, mean ± SD 27.2 ± 13.3 Duration of pump use, years, mean ± SD 9.6 ± 3.8 Hemoglobin A1c, %, mean ± SD 7.4 ± 0.8 Type of insulin used, %
Humalog 53.8
Novolog 38.5
Apidra 7.7
CHO administration (i.e., when the algorithm was not rec-ommending CHO administration). There was one instance of investigator-initiated treatment for hyperglycemia, per pro-tocol requirement (glucose level >300 mg/dl for 1 hour by YSI). At the time of this insulin dose, CGM glucose was decreasing, ketones were 0.1 mmol/l, and the participant was asymptomatic.
Protocol-defined glucose-related safety events (severe hypoglycemia or diabetic ketoacidosis) did not occur in any study participants. There was one adverse device effect related to bruising and bleeding at sensor insertion site (an expected event), which occurred prior to the CRC visit.
Continuous-Glucose-Monitoring-Related Issues
There were 128 total instances of missed CGM samples (on average, 9.8 samples per participant). These missed CGM
samples were due to radiofrequency communication issues or other communication issues related to the investigational setup, or due to display of “???” on the CGM receiver, indi-cating that the sensor is sending glucose readings but the receiver “does not understand.” The longest consecutive string of 10 missed CGM samples was due to device com-munication malfunction. For many participants, the longest string of missed samples was only 1 or 2 samples (corre-sponding to 5 or 10 minutes real time), indicating that the algorithm is able to cope with some missing data and carry on safely.
Discussion
This feasibility study evaluated the performance of the HHM System in participants with type 1 diabetes in a CRC setting. This study demonstrated that the HHM System was able to
Table 4. YSI Metrics While Under Closed-Loop Control: Percentage of Time Spent in Glucose Ranges of < 70 mg/dl, 70-180 mg/dl, and > 180 mg/dl.
Participant
% of time spent
Overall Overnight, 00:00-07:00 Postbreakfast, 07:00-13:00 Postlunch, 13:00-20:00 <70 70-180 >180 <70 70-180 >180 <70 70-180 >180 <70 70-180 >180 1a 0 95.3 4.7 0 100 0 0 85.7 14.3 0 100 0 2a 0 41.8 58.2 0 75.6 24.4 0 12.3 87.7 0 28.8 71.2 3a 0 72.6 27.4 0 100 0 0 63.4 36.6 0 46.5 53.5 4a 0 68.4 31.6 0 94.4 5.6 0 68.1 31.9 0 23.1 76.9 5 1.3 78.2 20.5 3.4 96.6 0 0 76.7 23.3 0 56.3 43.7 6 0 21.5 78.5 0 1.6 98.4 0 38 62 0 21.3 78.7 7 0 72.5 27.5 0 98.9 1.1 0 48.6 51.4 0 63.9 36.1 8 0 26.2 73.8 0 4.4 95.6 0 58.3 41.7 0 20 80 9 0 89.9 10.1 0 100 0 0 67.1 32.9 0 100 0 10 0 88.7 11.3 0 91.8 8.2 0 63.6 36.4 NA NA NA 11 0 73.5 26.5 0 100 0 0 12.5 87.5 0 100 0 12 1.3 81.5 17.2 0 100 0 4.1 39.7 56.2 0 100 0 13a 0 95.3 4.7 0 100 0 0 84.5 15.5 0 100 0 Average 0.2 69.6 30.2 0.3 81.8 17.9 0.3 55.3 44.4 0 63.3 36.7 Standard deviation 0.5 24.7 24.8 0.9 35.6 35.7 1.1 24.1 23.9 0 35.0 35.0
NA, not applicable.
aIndicates subjects who were deliberately under- and overbolused by 30% for breakfast and lunch, respectively.
Table 3. Ability of the Hypoglycemia-Hyperglycemia Minimizer (HHM) System Algorithm to Modulate Insulin Delivery, Relative to the Participants’ Corresponding Basal Rates, Both Prior to and During Out-of-Zone Glucose Excursions.
Below-zone excursions (n = 10) Above-zone excursionsa (n = 27)
Basal rate
(mean ± SD; U/h) (mean ± SD; U/h)HHM rate % change from basal (mean) (mean ± SD; U/h)Basal rate (mean ± SD; U/h)HHM rate % change from basal (mean)
Before excursions 0.77 ± 0.09 0.18 ± 0.33 −77.2 0.70 ± 0.21 0.71 ± 0.79 +1.6
During excursions 0.84 ± 0.19 0.16 ± 0.36 −80.4 0.90 ± 0.35 1.26 ± 1.14 +39.9
SD, standard deviation.
Table 6. Administrations of Algorithm-Recommended Supplemental Carbohydrates. Instance (sequentially) Participant YSI value at administration (mg/dl) CGM value at administration (mg/dl) 1 3 100.3 83 2 3 99.6 102 3 4 110.4 93 4 5 67.3 85 5 11 105.0 102 6 12 77.0 93 7 12 73.7 84 8 12 76.1 52 9 12 66.7 39 Mean ± SD 86.2 ± 17.3 81.4 ± 21.8
Each administration was 15 g of juice or glucose tablets. CGM, continuous glucose monitor; SD, standard deviation.
modulate insulin delivery based on CGM readings, and was effective in taking preemptive action to mitigate near-future
below-zone excursions by delivering less insulin than the corresponding preprogrammed basal rates.
The algorithm appeared to redistribute insulin delivery based on CGM data, without significantly altering the over-all total daily basal dose. To mitigate out-of-zone excursions, the HHM System decreased insulin infusion rates by an aver-age of 80.4% during below-zone excursions and increased infusion rates by an average of 39.9% during above-zone excursions.
Figure 2. Average Hypoglycemia-Hyperglycemia Minimizer (HHM) System results for all participants (n = 13). Upper plot: Glucose levels (mean ± standard deviation) based on continuous glucose monitoring (CGM). Also shown are the nominal meal times. The shaded area is the approximately normoglycemic range (70-180 mg/dl). Lower plot: HHM System–determined insulin delivered during closed-loop control, as a percentage difference from the participants’ respective basal rates (mean ± standard deviation).
Table 5. Continuous Glucose Monitor Metrics While Under Closed-Loop Control: Percentage of Time Spent in Glucose Ranges of < 70 mg/dl, 70-180 mg/dl, and > 180 mg/dl.
Participant
% of time spent
Overall Overnight, 00:00-07:00 Postbreakfast, 07:00-13:00 Postlunch, 13:00-20:00 <70 70-180 >180 <70 70-180 >180 <70 70-180 >180 <70 70-180 >180 1a 0 76.6 23.4 0 100 0 0 63.6 36.4 0 61.4 38.6 2a 0 30.8 69.2 0 56.7 43.3 0 8.2 91.8 0 20.5 79.5 3a 0 69.7 30.3 0 98.9 1.1 0 59.2 40.8 0 42.3 57.7 4a 2.3 56.7 40.9 0 75.6 24.4 6.9 59.7 33.3 0 19.2 80.8 5 0 74.8 25.2 0 100 0 0 64.4 35.6 0 53.5 46.5 6 0 53.1 46.9 0 82.3 17.7 0 70.4 29.6 0 12.0 88.0 7 0 51.9 48.1 0 77.3 22.7 0 36.1 63.9 0 36.1 63.9 8 0 20.2 79.8 0 0 100 0 51.4 48.6 0 12.9 87.1 9 0 85.7 14.3 0 100 0 0 75.3 24.7 0 77.8 22.2 10 1.0 81.4 17.5 1.2 78.8 20.0 0 100 0 NA NA NA 11 0 53.4 46.6 0 94.6 5.4 0 6.9 93.1 0 45.8 54.2 12 3.8 72.3 23.9 0 97.8 2.2 12.3 27.4 60.3 0 84.5 15.5 13a 0 82.0 18.0 0 78.0 22.0 0 71.8 28.2 0 97.1 2.9 Average 0.5 62.2 37.2 0.1 80.0 19.9 1.5 53.4 45.1 0 46.9 53.1 Standard deviation 1.2 20.2 20.4 0.3 27.5 27.5 3.8 27.0 26.4 0 28.8 28.8
NA, not applicable.
The efficacy of the HHM System is further supported by the glucose data, which demonstrate that the system was able to maintain participants’ glucose in the approximately nor-moglycemic range (70-180 mg/dl) most of the time (an aver-age of 69.6% overall and 81.8% overnight, as measured by YSI) with very little time spent in the hypoglycemic range (< 1% of time with glucose levels < 70 mg/dl, as measured by both YSI and CGM).
The safety data also support the feasibility of the HHM System. During the study, there were no instances of severe hypoglycemia or diabetic ketoacidosis, no instances of investigator-initiated CHO administration, and only one instance of investigator-initiated treatment for hyperglycemia. The nine instances of investigator-administered algorithm-recommended CHO served as a preliminary evaluation of the algorithm’s warning capabilities, which will unquestionably be an important part of the overall HHM System.
Although there were 128 instances of missed CGM sam-ples (due to device communication or display issues, includ-ing related to investigational set-up), the algorithm was able to cope with the missing data and carry on safely.
Recent studies have also demonstrated feasibility of over-night closed-loop control systems in clinical settings.2,3,16,17 As in the current study, these studies demonstrate the feasi-bility of maintaining a majority of patients in (near) normo-glycemia while mitigating the risk of hyponormo-glycemia during defined periods of closed-loop control.
Limitations of this research are inherent in a feasibility study and include the small sample size, the relatively short term of observation, and the artificial, sedentary CRC-based environment. As the study design included adjustment of food and insulin variables, the glucose outcomes need to be interpreted with caution.
Conclusions
The results of this study indicate that the HHM System is a feasible foundation for development of a closed-loop control insulin delivery device; further studies are under way.
Abbreviations
CGM, continuous glucose monitor; CHO, carbohydrate; CRC, clin-ical research center; FDA, Food and Drug Administration; HHM, Hypoglycemia-Hyperglycemia Minimizer; SD, standard deviation; SSM, Safety Supervision Module; Zone-MPC, Zone Model Predictive Controller.
Authors’ Note
ClinicalTrials.gov identifier: NCT01401751.
Acknowledgments
The authors would like to thank JDRF and the staff of Sansum Diabetes Research Institute for their participation in this study. The authors received editorial support from Excerpta Medica for
preparation of this article. Parts of the study were presented as poster presentations at the 72nd Scientific Sessions of the American Diabetes Association (ADA), Philadelphia, PA (June 8-12, 2012) and at the 48th Annual Meeting of the European Association for the Study of Diabetes (EASD), Berlin, Germany (October 1-5, 2012).
Declaration of Conflicting Interests
The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Daniel A. Finan, Thomas W. McCann Jr, and Ramakrishna Venugopalan are employees of Animas Corporation. Linda Mackowiak and Henry Anhalt were employees of Animas Corporation at the time of execution of the study. Henry Anhalt is currently employed by Sanofi. Linda Mackowiak is a limited stock-holder of Medtronic, Abbott, and Edwards Lifesciences. Eyal Dassau has received product support from Insulet Corporation. Stephen D. Patek and Francis J. Doyle III have no conflict of inter-est. Boris P. Kovatchev has acted on the advisory board/consulted for Animas Corporation and Sanofi-Aventis; and has received research grant and product support from Animas Corporation, Insulet Corporation, Roche Diagnostics, Sanofi-Aventis, and Tandem Diabetes Care. Howard Zisser has received honoraria for scientific lectures and travel reimbursement from Animas Corporation, Cellnovo, Insulet Corporation, MannKind Corporation, and Roche; and research grant and product support from Animas Corporation, Abbott Laboratories, Dexcom Inc, Eli Lilly and Co, GluMetrics Inc, Insulet Corporation, LifeScan Inc, Medtronic Inc, Novo Nordisk, Roche, and Sanofi; and is a board member of Artificial Pancreas Technologies.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by JDRF and Animas Corporation.
References
1. Kowalski AJ. Can we really close the loop and how soon? Accelerating the availability of an artificial pancreas: a
road-map to better diabetes outcomes. Diabetes Technol Ther.
2009;11(suppl 1):S113-S119.
2. Breton M, Farret A, Bruttomesso D, et al. Fully integrated artificial pancreas in type 1 diabetes: modular closed-loop
glucose control maintains near normoglycemia. Diabetes.
2012;61(9):2230-2237.
3. Dassau E, Zisser H, Harvey RA, et al. Clinical
evalua-tion of a personalized artificial pancreas. Diabetes Care.
2013;36(4):801-809.
4. Hovorka R, Allen JM, Elleri D, et al. Manual closed-loop insulin delivery in children and adolescents with type 1
diabetes: a phase 2 randomised crossover trial. Lancet.
2010;375(9716):743-751.
5. Russell SJ, El-Khatib FH, Nathan DM, Magyar KL, Jiang J, Damiano ER. Blood glucose control in type 1 diabetes with
a bihormonal bionic endocrine pancreas. Diabetes Care.
2012;35(11):2148-2155.
6. Steil GM, Rebrin K, Darwin C, Hariri F, Saad MF. Feasibility of automating insulin delivery for the treatment of type 1
7. Cobelli C, Renard E, Kovatchev B. Artificial pancreas: past,
present, future. Diabetes. 2011;60(11):2672-2682.
8. U.S. Department of Health of Human Services. Guidance for industry and Food and Drug Administration staff. The con-tent of investigational device exemption (IDE) and premarket approval (PMA) applications for artificial pancreas device sys-tems. November 9, 2012. Available at: http://www.fda.gov/ downloads/MedicalDevices/DeviceRegulationandGuidance/ GuidanceDocuments/UCM259305.pdf. Accessed May 22, 2013.
9. Kovatchev BP, Breton M, Man CD, Cobelli C. In silico pre-clinical trials: a proof of concept in closed-loop control of type
1 diabetes. J Diabetes Sci Technol. 2009;3(1):44-55.
10. Patek SD, Bequette BW, Breton M, et al. In silico preclini-cal trials: methodology and engineering guide to closed-loop
control in type 1 diabetes mellitus. J Diabetes Sci Technol.
2009;3(2):269-282.
11. Venugopalan R, Finan DA, McCann Jr TW, et al. Performance metrics of the Hypoglycemia-Hyperglycemia Minimizer
(HHM) System in a closed-loop feasibility study. Diabetes.
2012;61(suppl 1):A235.
12. Mackowiak L, Finan DA, McCann Jr TW, Venugopalan R, Zisser H, Anhalt H. Feasibility study assessing
Hypoglycemia-Hyperglycemia Minimizer (HHM) System in patients with type 1 diabetes (T1DM) in a clinical research center (CRC).
Diabetes. 2012;61(suppl 1):A233.
13. Dassau E, Zisser H, Palerm CC, Buckingham BA, Jovanovic L, Doyle FJ 3rd. Modular artificial beta-cell system: a prototype for
clinical research. J Diabetes Sci Technol. 2008;2(5):863-872.
14. Grosman B, Dassau E, Zisser HC, Jovanovic L, Doyle FJ 3rd. Zone model predictive control: a strategy to minimize
hyper- and hypoglycemic events. J Diabetes Sci Technol.
2010;4(4):961-975.
15. Hughes CS, Patek SD, Breton MD, Kovatchev BP. Hypoglycemia prevention via pump attenuation and red-yel-low-green “traffic” lights using continuous glucose monitoring
and insulin pump data. J Diabetes Sci Technol.
2010;4(5):1146-1155.
16. Nimri R, Atlas E, Ajzensztejn M, Miller S, Oron T, Phillip M. Feasibility study of automated overnight closed-loop glucose control under MD-logic artificial pancreas in patients with
type 1 diabetes: the DREAM Project. Diabetes Technol Ther.
2012;14(8):728-735.
17. Phillip M, Battelino T, Atlas E, et al. Nocturnal glucose control
with an artificial pancreas at a diabetes camp. N Engl J Med.
