Day and Night Closed-LoopControl Using the IntegratedMedtronic Hybrid Closed-LoopSystem in Type 1 Diabetes atDiabetes CampDiabetes Care 2015;38:1205–1211 | DOI: 10.2337/dc14-3073

OBJECTIVE

To evaluate the feasibility and efficacy of a fully integrated hybrid closed-loop(HCL) system (Medtronic MiniMed Inc., Northridge, CA), in day and night closed-loop control in subjects with type 1 diabetes, both in an inpatient setting andduring 6 days at diabetes camp.

RESEARCH DESIGN AND METHODS

The Medtronic MiniMed HCL system consists of a fourth generation (4S) glucosesensor, a sensor transmitter, and an insulin pump using a modified proportional-integral-derivative (PID) insulin feedback algorithm with safety constraints. Eightsubjects were studied over 48 h in an inpatient setting. This was followed by astudy of 21 subjects for 6 days at diabetes camp, randomized to either the closed-loop control group using the HCL system or to the group using the MedtronicMiniMed 530G with threshold suspend (control group).

RESULTS

The overall mean sensor glucose percent time in range 70–180 mg/dL was similarbetween the groups (73.1% vs. 69.9%, control vs. HCL, respectively) (P = 0.580).Meter glucose values between 70 and 180 mg/dL were also similar between thegroups (73.6% vs. 63.2%, control vs. HCL, respectively) (P = 0.086). The meanabsolute relative difference of the 4S sensor was 10.8 6 10.2%, when comparedwith plasma glucose values in the inpatient setting, and 12.6 6 11.0% comparedwith capillary Bayer CONTOUR NEXT LINK glucose meter values during 6 daysat camp.

CONCLUSIONS

In the first clinical study of this fully integrated system using an investigational PIDalgorithm, the system did not demonstrate improved glucose control comparedwith sensor-augmented pump therapy alone. The system demonstrated goodconnectivity and improved sensor performance.

There have been a number of advances in developing automated insulin deliverysystems for optimizing glucose control in patients with type 1 diabetes with theultimate aim of reducing the burden of care for this condition (1–7). Early studies(8–10) demonstrated the feasibility of automated insulin modulation using

1Department of Pediatrics, Division of PediatricEndocrinology and Diabetes, Stanford UniversitySchool of Medicine, Stanford, CA2School of Paediatrics and Child Health, TheUniversity of Western Australia, Perth, WesternAustralia, Australia3Medtronic MiniMed Inc., Northridge, CA

Corresponding author: Bruce A. Buckingham,buckingham@stanford.edu.

Received 24 December 2014 and accepted 28March 2015.

Clinical trial reg. no. NCT02366767, clinicaltrials.gov.

This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc14-3073/-/DC1.

© 2015 by the American Diabetes Association.Readers may use this article as long as the workis properly cited, the use is educational and notfor profit, and the work is not altered.

Trang T. Ly,1,2 Anirban Roy,3

Benyamin Grosman,3 John Shin,3

Alex Campbell,3 Salman Monirabbasi,3

Bradley Liang,3 Rie von Eyben,1

Satya Shanmugham,1 Paula Clinton,1 and

Bruce A. Buckingham1

Diabetes Care Volume 38, July 2015 1205

DIABETES

CARESYMPOSIU

M

subcutaneous insulin pumps and subcu-taneous continuous glucose sensorswith the main focus being algorithm de-velopment. Further advances in bothsensor performance and algorithms in-crementally demonstrated improvedprotection against hypoglycemia, re-duced variability, and decreased meanglucose levels in controlled inpatient set-tings (11–13) as well as in outpatientstudies (2,4,6). The strategies for differ-ent control algorithms arewell describedin several recent reviews (14,15).The hybrid closed-loop (HCL) system

of Medtronic MiniMed Inc. (Northridge,CA) is the first fully integrated systemdesigned for continuous day and nightHCL control with the algorithm incor-porated within the insulin pump. TheHCL system requires meal announce-ment with an estimate of carbohydrateintake and a premeal insulin bolusto optimize postprandial glucose ex-cursions (9).The system consists of a Medtronic

fourth generation sensor (4S) anduses a proportional-integral-derivativewith insulin feedback (PID-IFB) algo-rithm with safety constraints to contin-uously modulate basal insulin deliverybased on sensor glucose values. The sys-tem configuration places the closed-loop controller on the pump andremoves the need for an intermediarydevice that contains the algorithm,such as a phone or portable computer.This reduces the need for connectivitybetween a controller and the insulinpump as well as between the controllerand sensor, two points of potential in-terrupted communication in a multi-device system (3–5).This report describes the clinical ex-

perience with using this HCL system in asupervised environment over multipledays for both day and night closed-loop control. The objective of this studywas to test the feasibility and efficacyof a preliminary algorithm in adoles-cents and adults with type 1 diabetesover 48 h in a research center followedby a 6-day period at a diabetes camp.

RESEARCH DESIGN AND METHODS

Medtronic HCL SystemThe Medtronic HCL system is a fully in-tegrated continuous glucose sensor andinsulin delivery system designed forcontinuous closed-loop control. The sys-tem consists of a 4S glucose sensor, a

sensor transmitter, and an insulinpump, as shown in Fig. 1. A PID-IFB al-gorithmwith safety constraints is incor-porated in the pump, allowing forcontinuous closed-loop control of basalinsulin delivery. The system also uses anext-generation Bayer CONTOUR NEXTLINK glucose meter, which allows thepatient to automatically send meterglucose values to the pump via radio-frequency connection.

Specifics of the control algorithmhave been previously described (16–18). Insulin delivery was modeled onthe multiphasic insulin response of theb-cell and consisted of three principalcomponents: proportional, integral,and derivative, with a modification toinclude feedback of a model-predictedinsulin profile (17,19).

The system operates in two modes:manual and closed loop. In the manualmode, the system uses patient-specificsettings, including basal rates, carbohy-drate/insulin ratio, insulin sensitivityfactor, and glucose targets. The pumpmay function as a stand-alone insulinpump. When used in conjunction withthe continuous glucose sensor, thereare also features such as automatic in-sulin suspension or predictive low glu-cose insulin suspension; however, thesefunctions were not activated during thisstudy.

With continuous glucose sensor in-put, closed loop may also be activated.The system uses the PID-IFB algorithmto continuously modulate basal insulindelivery. Instead of a preset basal rate,the algorithm determines insulin deliv-ery as sensor glucose inputs arrive anddelivers insulin as a small bolus or “mi-crobolus.” The system was set to a tar-get glucose of 120 mg/dL. Users mayalso inform the system of exercise, dur-ing which the target is increased to 160mg/dL.

The maximum insulin limit constrainsthe maximum insulin delivered at anytime by the algorithm. This value is spe-cific to each patient and is calculatedfrom fasting glucose values and totaldaily dose of insulin (TDD). The maxi-mum insulin limit adapts over timeand, for example, will increase in re-sponse to elevated fasting glucose val-ues. At the initialization of closed-loopcontrol, the TDD for each day of the pre-ceding week was manually entered intothe system.

A fixed carbohydrate/insulin ratiobased upon the TDD rule of 500/TDD,was used for premeal boluses. Correc-tion doses were initiated with meterglucose values .200 mg/dL, with a tar-get glucose of 180 mg/dL. The insulinsensitivity factor was algorithm derivedand also based upon the TDD.

Figure 1—Medtronic next-generation HCL system.

1206 Use of Medtronic Hybrid Closed-Loop System Diabetes Care Volume 38, July 2015

Study DesignParticipants were eligible to participateif they were between 14 and 40 years ofage, had a diagnosis of type 1 diabetesfor at least 1 year, and had been using aninsulin pump for at least 3 months. Exclu-sion criteria includeddiabetes ketoacidosisin the preceding 30 days, hypoglycemicseizure or loss of consciousness in thepreceding 3 months, and pregnancyor a medical condition considered tointerfere with the completion of theprotocol. There was no A1C exclusion cri-terion. The protocol was approved byStanford University Institutional ReviewBoard.

Inpatient PhaseAn initial inpatient study was conductedto evaluate the feasibility of system usein eight subjects, primarily between 14and 18 years of age, with type 1 diabe-tes. The HCL system was evaluated for48 h for each participant. A descriptionof the inpatient procedures is includedin the Supplementary Data.

Camp PhaseFor the main camp sessions, 21 sub-jects were recruited. Randomizationoccurred after the enrollment visitand was stratified by A1C. Subjectswere randomized to use either asensor-augmented pump with insulinsuspension using the commerciallyavailable Medtronic MiniMed 530G(Medtronic MiniMed, Inc.) (controlgroup) or closed-loop control with theHCL system.For the control group, an Enlite sen-

sor was inserted and linked to the 530Gpump. The 4S sensor was not compati-ble with the 530G pump. The suspendthreshold was set to 60 mg/dL. Inaddition, a low alert was set at 70 mg/dL,and a high alert was set at 250 mg/dL.Meter glucose testing was conductedusing a Bayer CONTOUR NEXT LINK glu-cose meter.For the HCL group, a 4S glucose sen-

sor was inserted. The HCL pump wasprogrammed with the subject’s currentpump settings and placed in manualmode. All calibrations were performedusing a capillary meter glucose valuemeasured by a Bayer glucose meter.The first calibration was entered be-tween 30 min and 2 h after sensor in-sertion. Sensors were calibrated at leastevery 12 h. A low alert was also set at70 mg/dL, and a high alert at 250 mg/dL.

After 5 h of manual operation, closedloop was started.

Both groups used premeal bolusesand were advised to enter the carbohy-drate amount into the bolus calculator.Research staff did not provide input forcarbohydrate counting. Subjects contin-ued in their group for the duration of the6 days and 6 nights of diabetes camp. Allsubjects were instructed either to washtheir hands with soap or use alcohol toclean fingers, discard the first drop ofblood, and use the second drop of bloodfor meter glucose testing. Capillary glu-cose testing was supervised in the HCLgroup. There were no restrictions forsubjects in either group, and all subjectsparticipated in the camp program.

Glucose Monitoring and Managementin the HCL GroupMeter glucose testing was routinely per-formed every 3–4 h, including beforemeals, before exercise, and beforebed. Overnight, subjects were tested at0000, 0300, and 0700 h.

The HCL system does not have thecapability for remote monitoring. Giventhat this was the first clinical experiencewith the system, a member of the re-search team accompanied each subjectto ensure that all system use was super-vised and that alarms were appropriatelymanaged. During the day, researchstaff was present in close proximity tothe subject. During the overnight pe-riod, subjects slept in their cabin withfellow campers without research staffpresent; however, in addition to theglucose testing at 0000 and 0300 h,the system was visually checked at0100 and 0500 h.

Subjects were advised to obtain a me-ter glucose value if the sensor glucosewas,70mg/dL and to treat with 15 g offast-acting carbohydrates. A repeat me-ter glucose was obtained after 15 min toensure that glucose values were .70mg/dL. Subjects were also advised tocheck for a meter glucose if the sensorwas .250 mg/dL .2 h after a meal. Ifthe meter glucose was confirmed to be.250 mg/dL, ketones were also ob-tained. If ketones were .0.6 mmol/Lor meter glucose was .400 mg/dL, asubcutaneous insulin correction dosewas given, the infusion set was changed,and closed loop was stopped for 2 h. Ifketones were ,0.6 mmol/L and themeter glucose was ,400 mg/dL, the

glucose value would be sent to the HCLpump for an automatic correction.

Closed-loop control would be stop-ped for the following conditions: meterglucose .400 mg/dL for .2 h, ketones.3 mmol/L, nausea or abdominal dis-comfort that did not resolve after 2 h inthe presence of ketones .0.6 mmol/L,any seizure or loss of consciousness, ormalfunction of the system that wasimposed upon subject safety.

At the end of the study, subjects com-pleted questionnaires to evaluate theirexperience.

Glucose Monitoring and Managementin the Control GroupIn the control group, subjects were un-der the care of the camp medical staff.This included premeal testing, and rou-tine testing at 0000 and 0700 h, and at0300 h if clinically indicated.

Subjects were also advised to obtain ameter glucose value if the sensor glu-cose was ,70 mg/dL and to treat with15 g of fast-acting carbohydrates.A repeat meter glucose was obtainedafter 15 min to ensure that glucose val-ues were .70 mg/dL. If the sensor glu-cose was .250 mg/dL, subjects wereadvised to check for meter glucose andketones and advise the camp medicalstaff prior to giving insulin.

Sample SizeThe inpatient study was designed to as-sess the feasibility of system use andwas not powered for significance. A co-hort of eight subjects was recruited togenerate 384 h of 4S sensor use andclosed-loop control.

The sample size estimate for thecamp study was based upon data fromour previous camp studies (5). Poweranalysis was performed for a repeated-measuresANOVA.Thevariance-covariancestructure was assumed to follow a com-pound symmetry with a variance of27 mg/dL and covariance of 14 mg/dL.The mean percent time in range from70 to 180 mg/dL for the control groupwas estimated to be 62%, and the meanpercent time in range for the treatmentgroupwas estimated to be 75%. A sampleof 10 patients in each group, for a total of20 patients, would provide at least 80%power to observe a difference between62% and 75% in a two-sided test withan a level of 0.05. We aimed to recruit20 participants to account for subjectwithdrawal.

care.diabetesjournals.org Ly and Associates 1207

StatisticsFor the primary outcome of percenttime in range, sensor glucose valueswere compared between the groupsusing a two-way repeated-measuresANOVA, and results are shown as theleast squares mean 6 SE. Given themarked difference in sensor perfor-mance between the Enlite and 4S, onlysensors with a daily median absoluterelative difference (ARD) of ,15%were used in both groups for analysesof glucose outcomes.Depending on distribution, other data

are expressed as mean6 SD or as median(interquartile range [IQR]). Comparisonsfor meter glucose levels were madeusing a Mann-Whitney U test for non-parametric data and a Student t testfor data conforming to a normal distri-bution. Analyses were performed usingSigmaStat version 11.0. The sample sizecalculation was performed in Stataversion 13.1.

RESULTS

Inpatient StudiesData for the inpatient studies are in-cluded in the Supplementary Data.

Camp SessionDuring the camp session, 21 subjectswere enrolled and studied for a totalof 120 person-days. One subject fromthe HCL group had to leave camp after24 h and was replaced. The mean 6 SDage was 18.6 6 3.7 years (range 15.3–31.4), duration of diabetes was 9.1 64.7 years, A1C was 8.6 6 1.5% (range5.9–11.6) (70 6 16 mmol/mol, range41–103), and insulin dose was 0.8 6 0.2units/kg/day. Subjects in the HCL groupremained in closed loop for 93 6 3% ofthe scheduled time. There was continu-ous sensor glucose data available for98.9% of the time. Time off closed loopwas attributed to sensor change or tem-porarily suspending HCL after infusionset failure.There were no instances of diabetic

ketoacidosis or severe hypoglycemia re-sulting in seizure or coma that resultedin stopping closed loop for subjects inthe HCL group. There was an episodeof a hypoglycemic generalized tonic-clonic seizure in a control participant,which occurred after a dinner bolus of8.3 units was delivered with delayedfood consumption. The meter glucoseobtained at the time of the seizure was46 mg/dL. The subject made a full

recovery with the administration of in-tramuscular glucagon.

Sensor PerformanceThe overall mean ARD for the 4S sensorcompared with meter glucose valueswas 12.6 6 11.0%. The median ARDwas 9.7% (IQR 4.7, 17.6) (n = 742). ABland-Altman plot of 4S sensor glucosecompared with meter glucose is shownin Supplementary Fig. 1A. The meanARD values for the 4S sensor from day1 through day 6 were 15.9 6 12.4%,11.3 6 8.6%, 11.6 6 12.1%, 13.3 610.3%, 13.1 6 11.9%, and 11.3 69.9%, respectively.

The overall mean ARD value for theEnlite sensor, compared with meter glu-cose values, was 21.76 23.4%. The me-dian ARD was 15.9% (IQR 7.5, 28.2) (n =383). A Bland-Altman plot for Enlite sen-sor compared with meter glucose isshown in Supplementary Fig. 1B.

To assess the efficacy of the system,only sensor data with a daily medianARD of,15% were included in the anal-ysis for both the control and HCL groups.This left 32 person-days for the controlgroup and 51 person-days for the HCLgroup from a total possible of 60 ineach group.

Glucose ControlGlucose control, from sensor glucosevalues, for both groups using dayswith a daily median ARD of ,15% isshown in Table 1. In terms of the pri-mary outcome, the mean percent timein range 70–180 mg/dL during the dayand night was similar between thegroups (73.1% vs. 69.9%, control vs.HCL, respectively) (P = 0.580). The effectchanged over time (days of wear) (P =0.009), with the HCL group tending toimprove from 64.8% on day 1 to 77.7%on day 6, compared with the controlgroup, which deteriorated from 90.3%on day 1 to 56.2% on day 6. Data formedian percent time in range 70–180mg/dL per day are shown in Fig. 2A.Mean glucose values exhibited a similarpattern, with mean glucose improvingfrom 154 mg/dL on day 1 to 147 mg/dLon day 6 for the HCL group, whereas themean glucose in the control group in-creased from 130 mg/dL on day 1 to173 mg/dL on day 6.

During the overnight period, the over-all percent time in range 70–180 mg/dLwas 68.2% for the control group versus79.9% for the HCL group (P = 0.111).

The HCL group performed more favor-ably over the course of the study andby day 6 had achieved 80.6% in rangecompared with 42.8% in the controlgroup. Data for median percent time inrange per night (70–180 mg/dL) areshown in Fig. 2B. There was also a trendtoward less hypoglycemia as well as atrend for less hyperglycemia, whichagain were more prominent by day 6.

Overall, the mean 6 SD percent timein range 70–180 mg/dL was 70.66 9.8%for the HCL group compared with 70.3614.9% for the control group. The median(IQR) for the same rangewas 70.0% (68.3,74.9) for the HCL group compared with74.4% (58.5, 82.0) for the control group.

Meter glucose values were also com-pared, and these data are shown in Ta-ble 1. There were more than double thenumber of meter glucose values ob-tained in the HCL group (n = 782) com-pared with the control group (n = 383).This occurred mainly due to the testingprocedures in the HCL group, withmeterglucose values obtained for all sensorglucose values outside of the 70–250mg/dL range as well as during the over-night period. For all meter glucose val-ues .250 mg/dL, a repeat test wasperformed after 1 h in the HCL group.There was an average of 0.4 events/person/day with meter glucose .250mg/dL in the control group versus 0.7events/person/day in the HCL group(P = 0.055). There were double the num-ber of tests performed 4 h after a meterglucose level of.250 mg/dL with a me-dian of 4 tests (IQR 3, 4) for the HCLgroup compared with 2 tests (IQR 1, 3)for the control group (P, 0.001), in partbecause of the less aggressive correctionboluses given in the HCL group com-pared with the control group.

There was only one meter glucosevalue,50 mg/dL in the HCL group com-pared with four values in the controlgroup over the week (P = 0.278). Therewere two meter glucose values .400mg/dL in the HCL group, comparedwith one in the control group.

Themedian sensor glucose values overthe course of 24 h for the control groupversus the HCL group are shown in Fig. 3.

Total Daily InsulinThe average daily insulin dose tended todecrease during the week of camp forboth the control and the interventiongroups. In the control group, the average

1208 Use of Medtronic Hybrid Closed-Loop System Diabetes Care Volume 38, July 2015

TDD was 63 6 17 units/day at baselinecompared with 576 23 units/day duringcamp (P = 0.078). In the HCL group, theaverage TDD was 58 6 17 units/day atbaseline versus 496 17 units/day duringcamp (P = 0.001).

Carbohydrate ConsumptionThe average daily amount of carbohy-drates consumed in the HCL group was179 6 27 g for females (n = 5, meanweight 73 kg) and 259 6 23 g formale subjects (n = 5, mean weight 77kg). This was similar to the amount ofcarbohydrates consumed in the controlgroup (158 6 56 g for females [n = 6,mean weight 63 kg] and 222 6 99 g formales [n = 4, mean weight 82 kg]).

CONCLUSIONS

Closed-loop control using a fully inte-grated sensor, algorithm, and pump sys-tem is an exciting innovation in thisrapidly advancing field, and this reportdescribes the first clinical experience us-ing theMedtronic HCL systemwith the 4Ssensor and an investigational algorithm.In a supervised camp setting, glucose con-trol was comparable to that achievedwith automated insulin suspension alone.

Over multiple days, however, there wasimproved performance of the system,achieving between 70% and 77% time inrange. The results are promising for thepotential impact of reducing the burdenof care of type 1 diabetes.

The overall glycemic control wasfavorable compared with other hybrid,single-hormone systems, such as pre-sented by the AP@home Consortium,where Leelarathna et al. (2) achieved amedian of 75% of time between 70 and180 mg/dL in an at-home, outpatientstudy over 7 days, compared with a me-dian of 70% for our cohort. Our results arealso similar to those of Kovatchev et al. (6),who reported a mean of 66% of time be-tween 70 and 180 mg/dL in a 40-h out-patient study of adults, compared withour mean of 70.6% for the same range.For a dual hormone system using bothinsulin and glucagon, Russell et al. (4)achieved an impressive 76–80% of timeinasimilar rangeover5days forbothadultsand adolescents, in a supervised studywithremotemonitoring. The systemsdescribed,however, are not fully integrated.

The HCL systemwas simple to navigateand use. Once the pump was operational

in manual mode with sensor glucose dataavailable, only the addition of TDD datawas required to start closed loop. In fu-ture versions, the system would be usedinmanualmode for 3–5 days, and no datawould be required to be entered prior toinitializing closed loop.

During closed-loop operation in thisstudy, both the carbohydrate/insulinratio as well as the insulin sensitivityfactor were algorithm derived. Therewas no ability to dictate predeterminedcarbohydrate/insulin ratios for specifictimes of the day. This did not reflectindividualized patient-derived settings,which may include, for example, moreaggressive carbohydrate/insulin ratiosfor breakfast. This contributed to thepostprandial hyperglycemia seen inthe HCL group. In future versions, therewill be the option of entering patient-specific carbohydrate/insulin ratios inspecific time blocks. With the limita-tions of current insulin pharmacokinetics,optimizing premeal boluses remainsthe challenge in an insulin-only deliv-ery system.

Although the system was safe, in thatthere was not a greater incidence of

Table 1—Glycemic outcomes

Treatment P value

Sensor-augmentedpump with thresholdsuspend at 60 mg/dL

(n = 10)HCL system(n = 10) Treatment Time

Interactionbetween timeand treatment

Sensor glucose for day and night (0700–0700 h)Mean glucose value (mg/dL) 147 6 8 157 6 6 0.274 0.071 0.006Percent time between 70 and 180 mg/dL 73.1 6 5.0 69.9 6 3.3 0.580 0.085 0.009Percent time between 70 and 150 mg/dL 58.0 6 6.2 51.8 6 4.1 0.375 0.211 0.064Percent time ,70 mg/dL 2.4 6 0.6 2.1 6 0.4 0.656 0.683 0.213Percent time ,60 mg/dL 0.7 6 0.4 0.5 6 0.3 0.699 0.274 0.164Percent time ,50 mg/dL 0.1 6 0.1 0.1 6 0.1 0.747 0.125 0.062Percent time .180 mg/dL 24.8 6 5.2 28.4 6 3.5 0.542 0.058 0.006Percent time .250 mg/dL 6.3 6 2.8 8.2 6 1.9 0.53 0.181 0.01

Sensor glucose at night (2300–0700 h)Mean glucose value (mg/dL) 149 6 10 146 6 6 0.75 0.005 0.013Percent time between 70 and 180 mg/dL 68.2 6 6.1 79.9 6 4.0 0.111 0.119 0.012Percent time between 70 and 150 mg/dL 51.7 6 8.8 59.8 6 5.8 0.421 0.132 0.058Percent time ,70 mg/dL 4.2 6 1.3 1.7 6 0.9 0.136 0.196 0.236Percent time ,60 mg/dL 1.7 6 1.0 0.5 6 0.7 0.298 0.076 0.188Percent time ,50 mg/dL 0.6 6 0.9 0.1 6 0.6 0.627 0.026 0.026Percent time .180 mg/dL 28.0 6 6.7 19.0 6 4.4 0.247 0.033 0.011Percent time .250 mg/dL 7.0 6 2.6 3.1 6 1.7 0.214 0.014 0.098

Meter glucose for day and night (0700–0700 h)Average meter glucose value 137 (121, 173) 152 (145, 179) 0.121Percent of meter values ,70 mg/dL 6.7 (5.4, 12.4) 4.7 (2.6, 8.3) 0.130Percent of meter values between 70 and 180 mg/dL 73.6 6 15.4 63.4 6 10.5 0.101Percent of meter values .180 mg/dL 15.5 (10.1, 43.4) 31.1 (20.3, 45.7) 0.076

Data for sensor glucose are reported as least squares mean 6 SE and include only sensor data where the daily median ARD,15%. Data for meterglucose are reported as mean 6 SD or median (IQR).

care.diabetesjournals.org Ly and Associates 1209

hypoglycemia, the automatic correc-tions did not correct glucose valuesinto target as rapidly as patients ex-pectedwith their own standard therapy.This was particularly problematic forsubjects who were accustomed to ac-tively correcting their glucose values toachieve optimal time in range. Thiscertainly contributed to the higher meanglucose values as well as the more pro-longed postprandial hyperglycemia seenin the HCL group. Future versions of thealgorithm may have a lower threshold toinitiate a correction dose as well as a

lower correction dose target to allow forincreased insulin dosing. In this investiga-tional algorithm, correction boluses weregiven automatically, and subjects werenot able to see the amount to accept, re-ject, or adjust the bolus. Subjects wereonly able to see the amount deliveredafter it was given. Although this involvedfewer buttons to push for the subject,many patients complained that they didnot feel comfortable not knowing howmuch insulin was given. These are impor-tant user interface considerations forfuture closed-loop development. The

degree of automation will affect patientbehaviors and the acceptability of thedevices. Automated insulin delivery sys-tems represent a shift in the current con-cepts of basal/bolus insulin delivery, andit is clear that patients will want somedegree of control of insulin delivery aspreliminary versions of these devicesbecome commercially available.

As system use extends, there needsto be flexibility and adaptability to in-dividuals with variable insulin sensi-tivities that may change with exercise,stress, illness, and hormone levels.Although there was a component ofadaptability with the maximum insulinlimit, this needs to be further investigatedwith studies testing system use over alonger time period. A system that is ableto quickly adapt, such as in the bihormo-nal approach used by Russell et al. (4), willundoubtedly be important for optimizingindividual control.

The Medtronic 4S sensor demon-strated vast improved accuracy over thecurrent commercially available Enlitesensor, with ameanARDof 12.5%. Sensoraccuracy is one of the critical determi-nants of a successful algorithm and af-fects how aggressively the algorithm canbe set in achieving glucose targets as wellas in affecting patient satisfaction and thepotential uptake of the technology. The4S sensor was well accepted by all ourparticipants, among both naive and expe-rienced sensor users.

We have consistently observed a de-crease in sensor accuracy when we movefrom inpatient to outpatient studies. In arecent study (5), the Dexcom G4P sensorerror increased fromameanARDof 10.4%during inpatient studies to 17.5% in thecamp setting. In both the inpatient andoutpatient settings, sensors were cali-brated against meter glucose values. Inthe inpatient setting, the reference glu-cose was measured by YSI, and in thecamp setting the reference was a meterglucose. The decreased accuracy of sensorglucose values may be attributed to thetechnique in obtaining the capillarysample. In the current study, subjectswere asked to wash their hands or usean alcohol swab, and to discard the firstdrop of blood and use the second dropfor testing. This resulted in only a min-imal decrease in 4S performance whenmoving from the inpatient to the campsetting.When glucose sensors are beingused to deliver insulin in a closed-loop

Figure 2—A: Percent time between 70 and 180 mg/dL over 6 days (0700–0700 h). Data arereported as median (IQR). B: Percent time between 70 and 180 mg/dL over 6 nights (2300–0700 h). Data are reported as median (IQR).

1210 Use of Medtronic Hybrid Closed-Loop System Diabetes Care Volume 38, July 2015

control system, we would advocate thatmeter glucose values be obtained usingthis technique.There were a number of limitations to

this study. The difference in sensor per-formance made a direct comparison ofglycemic outcomes prohibitive.We choseto use sensor data onlywhere themedianARD per day was,15%, to remove egre-gious errors in both groups. The degree ofsupervision was also different betweenthe groups. Our aimwas to have a controlgroup that reflected the standard care re-ceived at camp using a commerciallyavailable pump. This investigational HCLsystem did not have remote monitoringand required research staff be in closeproximity for the supervision of systemuse, which created a discrepancy in carebetween the two groups. We did not,however, influence clinical care byassisting in carbohydrate counting ineither group. An important objectivewas to see how the system would workwith minimal input from staff.As a first clinical study of the inte-

grated HCL system, it is reassuring tosee that the system performance wascomparable to automated insulin sus-pension alone in a supervised setting.The 4S sensor was a significant improve-ment over previous Enlite sensors andshould allow future versions of the HCLsystem to bemore aggressive with over-all glucose control.

Acknowledgments. The authors thank theparticipants and their families for taking part

in this study. The authors also thank themedicalstaff, directed by Dr. Kevin Kaiserman at CampConrad-Chinnock, and the camp staff, directedby Dr. Rocky Wilson at Camp Conrad-Chinnock,for making this study possible.Funding. This work is supported by a grantfromMedtronic Diabetes. T.T.L. is supported bythe University of Western Australia F.A. HadleyOverseas Medical Fellowship.Duality of Interest. T.T.L. has received hono-raria from Medtronic. A.R., B.G., J.S., A.C., S.M.,and B.L. are employees of Medtronic MiniMedand are Medtronic shareholders. B.A.B. is onmedical advisory boards for Sanofi; NovoNordisk; Becton, Dickinson and Company;Unomedical; and Medtronic. He has receivedresearch grant and/or material support fromMedtronic, Dexcom, LifeScan, Insulet, Bayer,Unomedical, and Tandem Diabetes Care. Noother potential conflicts of interest relevant tothis article were reported.Author Contributions. T.T.L. and B.A.B. de-signed the study, researched the data, andwrote the manuscript. A.R., B.G., J.S., A.C.,S.M., B.L., S.S., and P.C. reviewed the manu-script. R.v.E. provided statistical support. B.A.B.is the guarantor of this work and, as such, hadfull access to all the data in the study and takesresponsibility for the integrity of the data andthe accuracy of the data analysis.Prior Presentation. This study was presentedat the 75th Scientific Sessions of the AmericanDiabetesAssociation, Boston,MA, 5–9 June2015.

References1. Hovorka R, Elleri D, Thabit H, et al. Overnightclosed-loop insulin delivery in young peoplewith type 1 diabetes: a free-living, randomizedclinical trial. Diabetes Care 2014;37:1204–12112. Leelarathna L, Dellweg S, Mader JK, et al.;AP@home Consortium. Day and night homeclosed-loop insulin delivery in adults with type1 diabetes: three-center randomized crossoverstudy. Diabetes Care 2014;37:1931–19373. Thabit H, Lubina-Solomon A, Stadler M, et al.Home use of closed-loop insulin delivery for

overnight glucose control in adults with type 1diabetes: a 4-week, multicentre, randomisedcrossover study. Lancet Diabetes Endocrinol2014;2:701–7094. Russell SJ, El-Khatib FH, Sinha M, et al.Outpatient glycemic controlwith a bionic pancreasin type1diabetes. NEngl JMed2014;371:313–3255. Ly TT, Breton MD, Keith-Hynes P, et al. Over-night glucose control with an automated, uni-fied safety system in children and adolescentswith type 1 diabetes at diabetes camp. DiabetesCare 2014;37:2310–23166. Kovatchev BP, Renard E, Cobelli C, et al.Safety of outpatient closed-loop control: firstrandomized crossover trials of a wearable artifi-cial pancreas. Diabetes Care 2014;37:1789–17967. Nimri R, Muller I, Atlas E, et al. MD-Logicovernight control for 6 weeks of home use inpatients with type 1 diabetes: randomized cross-over trial. Diabetes Care 2014;37:3025–30328. Hovorka R, Allen JM, Elleri D, et al. Manualclosed-loop insulin delivery in children and ado-lescents with type 1 diabetes: a phase 2 rando-mised crossover trial. Lancet 2010;375:743–7519. Weinzimer SA, Steil GM, Swan KL, Dziura J,Kurtz N, Tamborlane WV. Fully automatedclosed-loop insulin delivery versus semiauto-mated hybrid control in pediatric patients withtype 1 diabetes using an artificial pancreas. Di-abetes Care 2008;31:934–93910. Kovatchev B, Cobelli C, Renard E, et al. Multi-national study of subcutaneous model-predictiveclosed-loop control in type 1 diabetes mellitus:summary of the results. J Diabetes Sci Technol2010;4:1374–138111. Hovorka R, Kumareswaran K, Harris J, et al.Overnight closed loop insulin delivery (artificialpancreas) in adults with type 1 diabetes: cross-over randomised controlled studies. BMJ 2011;342:d1855.12. O’GradyMJ, Retterath AR, Keenan DB, et al.The use of an automated, portable glucose con-trol system for overnight glucose control in ado-lescents and young adults with type 1 diabetes.Diabetes Care 2012;35:2182–218713. Kovatchev BP, Renard E, Cobelli C, et al.Feasibility of outpatient fully integrated closed-loop control: first studies of wearable artificialpancreas. Diabetes Care 2013;36:1851–185814. Peyser T, Dassau E, Breton M, Skyler JS. Theartificial pancreas: current status and futureprospects in the management of diabetes. AnnN Y Acad Sci 2014;1311:102–12315. Thabit H, Hovorka R. Bringing closed-loophome: recent advances in closed-loop insulindelivery. Curr Opin Endocrinol Diabetes Obes2014;21:95–10116. Steil GM, Rebrin K, Janowski R, Darwin C,SaadMF.Modeling beta-cell insulin secretiondimplications for closed-loop glucose homeostasis.Diabetes Technol Ther 2003;5:953–96417. Steil GM, Palerm CC, Kurtz N, et al. The effectof insulin feedback on closed loop glucose con-trol. J Clin Endocrinol Metab 2011;96:1402–140818. Steil GM, Panteleon AE, Rebrin K. Closed-loop insulin delivery-the path to physiologicalglucose control. Adv Drug Deliv Rev 2004;56:125–14419. Steil GM, Rebrin K, Darwin C, Hariri F, SaadMF. Feasibility of automating insulin delivery forthe treatment of type 1 diabetes. Diabetes2006;55:3344–3350

Figure 3—Glucose control over 24 h. Results are reported as median (IQR).

care.diabetesjournals.org Ly and Associates 1211