2015 Diabetes Technology Meeting Abstracts

doi:10.1177/1932296816639698

. 2016 Mar 1;10(2):476–611. doi: 10.1177/1932296816639698

2015 Diabetes Technology Meeting Abstracts

PMCID: PMC4773984 PMID: 28264166

Acree	Assessment of a Patient-Level Diabetes Education Elective for First-Year…	A1
Ahn	Continuous Glucose Monitoring Using Near-Infrared Fluorescent Carbon and…	A2
Albanese-O’Neill	Ketone Monitoring Frequency and Methods in Type 1 Diabetes Subjects	A3
Aloi	eGlycemic Management System Provides Safe and Effective Glycemic Control for…	A4
Andreatta	Optimization of Core Sensor Components for an Osmotic Pressure Sensor for…	A5
Bailey	Accuracy and User Performance Evaluation of a New Blood Glucose Monitoring…	A6
Baram	Development of a New Glucose Dehydrogenase Mutant That Is Free of Xylose…	A7
Benesch	ClampArt: Improved Algorithm for Automated Glucose Clamps	A8
Blaabjerg	Development of a Hypoglycemia Alarm Based on Electroencephalography	A9
Bode	Patient-Level Assessment of Cough with Technosphere Inhaled Insulin in Patients…	A10
Bondia	Coordination of Insulin and Glucagon Delivery in Bihormonal Artificial Pancreas…	A11
Breton	Calibration of the Dynamical Tracking eA1c Algorithm with Reference HbA_1c…	A12
Carlson	Safety Analysis of a 90-Day Implantable Continuous Glucose Monitoring System…	A13
Cembrowski	Biologic Variation of Intensive Care Unit Patient Capillary Glucoses Monitored…	A14
Ching	Biodel’s Glucagon Emergency Management Autoreconstitution Device…	A15
Christiansen	Accuracy and User Performance Evaluation of a New Blood Glucose Monitoring…	A16
da Silva	Personalizing the Model Predictive Controller Aggressiveness for the Artificial…	A17
Dagliati	Metformin Exposure Patterns Are Related to Type 2 Diabetes Nephropathy	A18
Datye	Timing of Meal Insulin Delivery and Its Relation to Missed Insulin Doses	A19
David	Serum Total Homocysteine as a Potential Predictor for Increased Risk for…	A20
Daya	Planar Microwave Sensor for Noninvasive Blood Glucose Sensing	A21
de Bock	Performance of a Hybrid Closed-Loop Insulin Delivery Algorithm with Insulin…	A22
Docherty	Basal Glucose Is Just the First Measured Point . . . Right?	A23
Dungan	Messaging via Electronic-Medical-Record-Based Patient Portal: Effect on Glucose…	A24
Emilion	Screening Obese Women with Prediabetes from Classes of Serum…	A25
Esenaliev	Noninvasive Glucose Monitoring with Novel High-Resolution Ultrasound Technique	A26
Famulla	Recombinant Human Soluble Insulin Julphar Insulin R Is Bioequivalent to…	A27
Feldman	Pramlintide versus Liraglutide: Assessing Incremental Benefits of Adjunctive…	A28
Feng	Controller Performance Assessment of Model-Based Artificial Pancreas Control…	A29
Fischer	Glooko: Mobile, Unified Platform for Diabetes Management, Combined with Nurse…	A30
Frank	An Estimation of Glucose Uptake during Exercise	A31
Freckmann	ISO 15197:2013: Evaluation of Two Improved Blood Glucose Monitoring Systems	A32
Fuller	Evaluation of Glucometer Test Strips for the Presence of Potential Pathogens…	A33
Gal	Simplifying Calibration Practice for Home Use Noninvasive Glucose Monitoring…	A34
Gal	Simplifying Calibration Practice for Home Use Noninvasive Glucose Monitoring…	A35
Garcia	A Novel, Time-Invariant, Intuitive Metric for Assessing Glycemic Variability	A36
Godavarty	Optical Classification of Diabetic Wounds: Preliminary Sensitivity and Specificity…	A37
Gómez	Long-Term Follow-Up of Sensor-Augmented Pump Therapy with Low-Glucose…	A38
Groat	Characterizing Self-Management Behaviors of Type 1 Diabetes Patients on Insulin…	A39
Groat	Introducing a Method to Retrospectively Compare Insulin Dosing Recommendations	A40
Grosman	Hybrid Closed Loop in Monitored Outpatient Conditions	A41
Gu	In Vitro and In Vivo Performance of a Multianalyte Sensor	A42
Hajnsek	The SPIDIMAN Single-Port System: Combining Continuous Glucose Monitoring…	A43
Hansen	Insulin-to-Carbohydrate Ratio and 24-h Urinary Glucose Excretion to Assess…	A44
Haselwood	Toward a Sensor and Model for Cortisol as Feedback in the Bioartificial Pancreas	A45
Hauzenberger	Systematic Study to Optimize Insulin Delivery through a Continuous Subcutaneous…	A46
Henderson	Finger Stickin’ Good: Improved Glycemic Control in Cardiovascular Surgery…	A47
Herrero	In Silico Evaluation of Ultra-Fast-Acting Insulins Using the Bioinspired Artificial…	A48
Hodish	Scalable and Durable Optimal Glycemic Control in Insulin Users	A49
Howsmon	Infusion Set Fault Detection via Monitoring of Real-Time Data over Sliding Windows	A50
Huber	Performance of a Blood Glucose Monitoring System in a Point-of-Care Setting	A51
Jacobs	Use of a Bihormonal Artificial Pancreas System with Exercise Announcement in…	A52
Kastellorizios	Cytotoxicity of Photopatternable Hydrogels for Implantable Glucose Sensors	A53
Kastellorizios	Glucowizzard: An Injectable, Wireless Biosensor for Continuous Glucose Monitoring	A54
Kastellorizios	Multianalyte Monitoring via Implantable Biosensors Predicts Exhaustion	A55
Kircher	Automated Insulin Delivery “In the Wild”	A56
Knab	The Importance of Multiple Sensors in Continuous Glucose Monitoring	A57
Kovatchev	Hemoglobin A1c and Self-Monitored Average Glucose: Validation of the…	A58
Kudva	Acute Effect of Low-Dose Pramlintide on Islet and Gastrointestinal Peptides in…	A59
Laguna Sanz	Adaptive Blood Glucose Predictions Using Untrained Models: a Performance…	A60
Lajara	Evaluation of Insulin Delivery with V-Go Combined with a Glucagon-Like…	A61
Lajara	Using the V-Go Disposable Insulin Delivery Device in Patients Prescribed High…	A62
Lande	Widespread Clinical Experiences Validating a Blood-Based Artificial Pancreas	A63
Leal	Online Detection of Sensor Faults from Continuous Glucose Monitoring, Meal, and…	A64
Lee	MT-Diet: Automated Diet Assessment Using Myo Wristband and Thermal Images	A65
Little	The Use of Insulin Pump Technology to Advance Safe and Effective Care	A66
Lucisano	Late-Breaking Human Clinical Results from a Second-Generation Long-Term Fully…	A67
Mader	Glycemic Control by Using GlucoTab in Insulin-Naive and Insulin-Treated…	A68
Magee	mHealth Blood Glucose Monitoring Solution Enables Timely Targets Attainment	A69
Melish	Pharmacokinetic Study of Mixed-Diet Glucose Concentration Responses in Type 1…	A72
Mould	Pharmacokinetic and Pharmacodynamic Modeling Predict Concentrated Biphasic…	A73
Murase	Resistance to Acetaminophen Interference of the Direct Electron Transfer-type…	A74
Nguyen	U-Strip Transdermal Insulin Patch	A75
Nunlee-Bland	Technology Use Enhances Body Mass Index Reduction in African American…	A76
Oser	Using Blogs to Support Caregivers of Patients with Type 1 Diabetes: A Pilot Study	A77
Otto	Evaluating a Novel Glycemic Pattern Recognition Algorithm for Alerting to Risk of…	A78
Pery	Performance of a Smartphone-Linked Blood Glucose Assessment Platform	A79
Pesl	Clinical Pilot Study Assessing the Postprandial Outcome of an Advanced Bolus…	A80
Pesl	Performance of an Automated Revision Algorithm for a Personalized Bolus…	A81
Pohl	Mixture of Insulin Monomers, Dimers, and Hexamers Demonstrated by…	A82
Reiterer	Is There a Recipe for Comparability of Mean Absolute Relative Difference Values…	A83
Renard	Multicenter Assessment of Usability and Safety of a Model Predictive Control…	A84
Resalat	Incorporating an Exercise Detection, Grading, and Hormone Dosing Algorithm into…	A85
Rigla	Artificial-Intelligence-Augmented Telemedicine Applied to the Management of…	A86
Rollins Sr.	Automatic Insulin Delivery System Approach Capable of Tight Control	A87
Ruan	Night-to-Night Variability of Basal Insulin Requirements in Adults with Type 1…	A88
Sacchi	Predicting Microvascular Complications from Type 2 Diabetes Retrospective Data	A89
Samadi	Meal Detection and Carbohydrate Estimation Algorithm Based on Continuous…	A90
Sartori	Comparison of the Nova Biomedical StatStrip Glucose Meter to an Isotope Dilution…	A91
Schwartz	Augmenting the Detection and Prediction of Blood Glucose Control Problems with…	A92
Shehata	New Technology for Metformin Hydrochloride Mucoadhesive Microparticles…	A93
Shomali	Data Science Framework for Mobile Health Engagement and Outcomes	A94
Simon	MyDiaText: Short Message Service Technology for Behavioral Health in Patients…	A95
Skowronski	Circulating Levels of Protease Activity during a Meal Tolerance Test in Type 2…	A96
Smajada	Interference Test Results for an Osmotic Pressure Sensor for Interstitial Glucose…	A97
Sode	First Three-Dimensional Structure of Fungus-Derived Glucose Dehydrogenase…	A98
Stewart	High Compliance = Good Control? Compliance of the Stochastic Targeted…	A99
Sundarapperuma	Perspectives on Diet during Postpartum Periods in Mothers with Gestational…	A100
Terrazas	Improving Meal Detection Using a Smartwatch and Accelerometer	A101
Thavareesan	Smartphone-Based Patient Health Care Management System	A102
Tiangco	Progress toward Passive Transdermal Glucose Sensing Based on a Glucose-Binding…	A103
Uyttendaele	Insulin Sensitivity Profile as a Marker for Reduced Outcome in the Neonatal…	A104
Vahlsing	Combination of Microdialysis and Ex Vivo Glucose Sensing by Infrared…	A105
Valuenzuela	Early Results Support Efficacy and Clinical Efficiency of Diabetes Management…	A106
Veiseh	Long-Term Glycemic Control Using Polymer Encapsulated Human-Stem-Cell…	A107
Vettoretti	Simulation of Patient Decision Making to Test Safety and Effectiveness of…	A108
Vettoretti	Simulation of Patient Decision Making to Test Safety and Effectiveness of…	A109
Visentin	Modeling the Effect of Liraglutide in Type 2 Diabetes Mellitus with the Type 2…	A110
Voge	Usage and Perceptions of Type 1 Diabetes Devices	A111
Voskanyan	Feasibility of Use of Novel Stable Soluble Glucagon Analog ZP-GA-1 in Artificial…	A112
Waki	Telehealth System: Smartphone-Based Self-Management for Diabetes Patients	A113
Walker	Accuracy of a Blood Glucose Monitoring System as It Relates to the ISO…	A114
Walker	Evaluation of a Blood Glucose Monitoring System in Comparison with Global…	A115
Wang, G.	Autonomous Shape-Memory-Alloy-Based Actuator Reaching Subcutaneous Blood…	A116
Wang, Y.	Comparison of Point-of-Care Blood Glucose Monitoring System Performance…	A117
Wang, Y.	Development of a Primary Isotope Dilution Mass Spectrometry Traceable…	A118
Wilson	Glucosense: A Low-Cost Glucose Meter System for Resource-Poor Settings	A119
Woodley	Cytotoxic Effect of Three Commercial Rapid-Acting Insulin Analogs on Four…	A120
Xenakis	Characterization of the Information Technology Utilization and Experiences of…	A121
Yablonski	Human Kinetic Approach to Improving Conformability in PaQ, a 3-day Wearable…	A122
Yan	Simulation Study on Insulin Sensitivity Factor Changes for Patients with Illness/Stress	A123
Yavelberg	Accuracy of Select Wearable Technology for Varying Exercise Modalities and…	A124
Zhao	Pilot Study of Oral Insulin Enteric-Coated Soft Capsules Research and…	A125
Zhong	Continuous Glucose Monitoring Accuracy Improvement Using the Autocalibration…	A126
Zhong	Real-World Assessment of Undetected Glycemic Excursion Events in Type 2…	A127
Zhong	Real-World Performance of a Predictive Low-Glucose Management Algorithm	A128
Zijlstra	Dance 501 Inhaled Human Insulin Has a Dose-Linear Response and Similar…	A129
Zijlstra	Demonstration of Very High Compliance and HbA_1c Benefit of a Long-Term…	A130
Zijlstra	Faster-Acting Insulin Aspart Using Continuous Subcutaneous Insulin Infusion…	A131
Zijlstra	Particle Size Does Not Impact the Pharmacokinetic Response of a Novel Inhaled…	A132

Open in a new tab

Assessment of a Patient-Level Diabetes Education Elective for First-Year Pharmacy Students Utilizing Hands-On Diabetes and Coexisting Disease State Point-of-Care Screening Tools

Lindsay Acree, PharmD; Dean Reardon, PhD; Stephen Cook, PharmD; Marea Dodd, BS; David Latif,PhD; Aladin Siddig, PhD

University of Charleston School of Pharmacy

Charleston, West Virginia

lindsayacree@ucwv.edu

Objective:

Utilizing the Diabetes Empowerment Education Program intended for non-health-care providers, this course teaches first-year students the tools needed to understand and educate patients about diabetes and several related disease states. Through hands-on training and practice, students learn how to screen patients for diabetes, cholesterol, and blood pressure.

Methods:

Diabetes Education and the Patient is a three-credit-hour elective course that is open to first-year PharmD students as an elective course. The program addresses the AADE7 self-care behaviors through eight modules. In addition to the modules, students are required to participate in two health care screening events in which they are able to practice the skills and techniques learned throughout the course. Within the modules, students are trained to utilize screening tools for cholesterol, A1C, and blood glucose testing. In addition, students also master manual blood pressure skills through practice with other students.

Result:

Course assessment includes student evaluations, oral and written assignment rubrics, reflections, and informal verbal feedback. Students were given patient-level pre- and posteducation training tests. Test scores improved from an average of 67.5% on the pretest to 91.8% on the posttest. Through analysis of pretest data, the area of diabetes diagnostics was the weakest, with an average of only 40%. Diabetic pathophysiology was identified as having the highest score on the pretest, with an average of 75%. However, the posttest revealed a low score of 86% and a high score of 97% in all question categories.

Conclusion:

Typically, diabetes is not taught in the curriculum until the third-year pharmacotherapy course, in which they receive 6–8 h of didactic lecture and classroom activities. Although blood pressure and blood glucose skills are taught during the first year, A1C and cholesterol screening equipment is not utilized in the curriculum until the third year in the Introductory Pharmacy Practice Experience course. This innovative diabetes course provides students with the skills and tools needed to counsel patients on diabetes, cholesterol, and lifestyle changes during their early experiential education rotations. With the hands-on activities included in the diabetes elective course, students may also gain the confidence during the first year of pharmacy school to become more involved with health fairs and patient education activities.

Continuous Glucose Monitoring Using Near-Infrared Fluorescent Carbon and Corona Phase Molecular Recognition Molecular Sites for Glucose

Jiyoung Ahn, BS; Nicole M. Iverson, PhD; Gili Bisker, PhD; Michael S. Strano, PhD

Massachusetts Institute of Technology

Cambridge, Massachusetts

jyahn@mit.edu

Objective:

Developing a continuous glucose sensor with which individuals can monitor and control the potential life-threatening effects of diabetes mellitus remains a heavily researched challenge. Herein, we present our work developing a new class of glucose sensors for long-term continuous blood glucose monitoring based on glucose-responsive near-infrared fluorescent nanoparticles utilized as a subcutaneous tattoo, which enables external, noninvasive, optical signal detection.

Method:

A novel composition of phenylboronic acid polymer was complexed with near- infrared fluorescent single-walled carbon nanotubes (SWNT) to generate unique glucose recognition sites using a technique we term corona phase molecular recognition (CoPhMoRe). The resulting glucose nanosensors were encapsulated within two different form factors, a dialysis membrane (~1.5 cm) and liquid core -hydrogel shell particles (~200 µm), and implanted subcutaneously in mice for long-term stability test and continuous glucose monitoring. The modulation of the fluorescent signal was monitored noninvasively and compared with the blood glucose levels measured by the commercially available continuous glucose monitor.

Result:

Our in vitro studies demonstrated notable instantaneous and reversible fluorescence responses to varying glucose concentrations over the hypoglycemic to hyperglycemic range. In vivo studies confirmed the stability and biocompatibility of the sensors for over 4 weeks, with no signs of inflammation. The preliminary glucose challenge tests showed promising results of correlations between changes in glucose concentration and modulation of the fluorescence emission intensity.

Conclusion:

We developed a new class of glucose sensors using SWNT complexed with novel CoPhMoRe sites for glucose recognition. These sensors were successfully implanted subcutaneously in mice and enabled noninvasive continuous glucose monitoring over long time periods, providing great promise for long-term glucose sensing tattoo implants.

Ketone Monitoring Frequency and Methods in Type 1 Diabetes Subjects

Anastasia Albanese-O’Neill, PhD, ARNP, CDE; Tonya D. Riddlesworth, PhD; Laura Jacobsen, MD; Michael J. Haller, MD; Desmond Schatz, MD; Kellee M. Miller, MPH; Roy W. Beck, MD, PhD; for the T1D Exchange

University of Florida

Gainesville, Florida

t1dstats@jaeb.org

Objectives:

Clinical recommendations advise self-monitoring of blood or urine ketones during hyperglycemia and illness to allow for early diagnosis of ketonemia or ketonuria and to reduce the risk of diabetic ketoacidosis (DKA). The purpose of this study was to characterize ketone management in persons with type 1 diabetes (T1D) who completed an online questionnaire between February and August of 2014 as part of the T1D Exchange registry.

Methods:

We analyzed data from 3,737 participants or parents of participant’s ages 3–88 years (median 23 years; median T1D duration 11 years; 56% female; 90% non- Hispanic white). A linear regression model was used to analyze the association between ketone checking and age.

Results:

Thirty percent (1,129) reported checking ketones at least once in the last 30 days. Among these 1,129 subjects, checking ketones was more common among younger participants (59% <18 years vs. 15% ≥18 years; P < 0.001). Checking ketones most or all the time when glucose level is above 300 mg/dL was reported by 30% of 61 <6-year-olds, 36% of 644 6–12- year- olds, 36% of 580 13–17-year-olds, 14% of 808 18–25-year-olds, and 7% of 1,644 ≥26-year-olds and checking most or all the time when nauseated or vomiting by 87, 81, 73, 42, and 18%, respectively. Among participants who reported a method for checking ketones, 81% reported checking with urine, 13% blood, and 7% urine and blood.

Conclusions:

Checking ketones is infrequent when the blood glucose is high or when nausea or vomiting occur, particularly in adults. The reported rate of monitoring is unacceptably low and suggests a need for more robust diabetes education to improve adherence to ketone monitoring guidelines and to reduce the risk for DKA in patients with T1D.

eGlycemic Management System Provides Safe and Effective Glycemic Control for Stroke Patients Requiring Subcutaneous Insulin in the Hospital Setting

Joseph A. Aloi, MD; Amy B. Henderson, BSN; Raymie S. McFarland, PT, SSGB; Robby Booth; Jagdeesh Ullal, MD

Wake Forest Baptist Medical Center

Winston Salem, North Carolina

jaloi@wakehealth.edu

Objective:

Admission hyperglycemia (>130 mg/dL) and persistent inpatient hyperglycemia (>48 h) have been predictors of poor patient outcomes, including higher mortality, higher rates of in -hospital complications, and longer length of stay in patients with acute ischemic stroke. This study evaluated the glycemic outcomes of patients using eGlycemic Management System subcutaneous (eGMS-SQ) for subcutaneous insulin management versus standard subcutaneous orders (SSO) in patients with acute ischemic stroke.

Method:

The study evaluated 40 patients with type 2 diabetes mellitus at Sentara Health System who required subcutaneous insulin to manage hyperglycemia. Qualifying patients were treated with eGMS-SQ or SSO with a target of 120–160 mg/dL. The efficacy and safety of each was evaluated by the following: 1) percentage of blood glucose (BG) between 71 and 180 mg/dL, 2) percentage of BG >180 mg/dL, 3) percentage of hypoglycemic events <40 and <70 mg/dL, 4) average BG, and 5) average length of stay (LOS).

Result:

Patients (n = 20) treated with eGMS- SQ had a starting BG of 232 mg/dL, average BG 149 mg/dL, and 74.7% of readings in target, and hypoglycemia <70 mg/dL was 2.1% and hypoglycemia <40 mg/dL was 0.0%. Average LOS using eGMS- SQ was 5.3 days. Patients (n = 20) treated with SSO had a starting BG of 235 mg/dL, average BG on eGMS-SQ of 169 mg/dL, and 56.8% of readings in target, and hypoglycemia <70 mg/dL was 4.4% and hypoglycemia <40 mg/dL was 0.5%. Average LOS using SSO was 6.4 days.

Conclusion:

These results suggest that eGMS-SQ can effectively and, more importantly, safely control and maintain glucose control for patients with acute ischemic stroke. A higher percentage of patients reached target glucose levels and had over 50% less hypoglycemia <70 mg/dL, and a lower LOS was observed.

Optimization of Core Sensor Components for an Osmotic Pressure Sensor for Interstitial Glucose Assessment ANDR15123D

Gaelle Andreatta, PhD; Rita Smajada, PhD; Silvia Angeloni, PhD; Sanja Ramljak, PhD; Rune Frisvold, MBA; Andreas Pfützner, MD, PhD; Martha Liley, PhD

CSEM

Neufchatel, Switzerland

gaelle.andreatta@csem.ch

Background:

A new interstitial glucose sensor based on an osmotic pressure sensor technology is currently under development by Lifecare, Norway (Sencell). In the sensor core, competitive binding of glucose versus dextran to specific binding sites of the plant lectin concanavalin A (Con A) is used to induce an osmotic pressure signal, which is in close correlation to interstitial glucose content. Molecular weight of dextran and environmental temperature are supposed to influence the reproducibility and stability of the sensor signal. The purpose of this experiment was to identify the most suitable dextran size and to assess stability of the sensor signal over months at 21 and 37 °C.

Methods:

A working laboratory model of the core sensor technology was developed to perform the chemistry optimization experiments. After confirmation of proper sensor operation, three different molecular weight versions of dextran were applied (10, 40, and 70 kDa) in a standardized experiment, and sensor sensitivity was tested over a glucose range from 2 to 30 mmol/L. Influence of glucose concentrations on chamber fluid viscosity was tested at different Con A:dextran concentration ratios (6:1, 3:1, and 1:1). In addition, the stability of the sensor signal was tested after 3 months of storage at 21 and 37 °C.

Results:

Lowering of the Con A concentration decreased the influence of glucose on the viscosity of the active fluid most likely because the number of intermolecular bonds between dextran and Con A is reduced. In addition, decreasing the molecular Con A:dextran ratio from 6:1 to 1:1 resulted in a significant reduction of glucose on active fluid viscosity and enhanced the amplitude of the osmotic pressure sensor signal. The combination of 40 kDa dextran in a 1:1 molecular ratio to Con A provided the most optimal sensor signal and was chosen for further stability experiments. When experiments were carried out at 21 vs. 37 °C, they led to comparable results. Stability of the active fluid at 37 °C was confirmed by the experimental data. In addition, there was no signal of a loss of Con A integrity when assessed by ultraviolet-visible spectroscopy.

Conclusions:

The results of our experiments and numerical simulations enabled us to optimize the composition of the active chamber fluid and to calculate the absolute osmotic pressure that will be measured for a given concentration of glucose. In addition, our results indicate that the sensor is working with a stable performance when stored for 3 months at body temperature of 37 °C.

Accuracy and User Performance Evaluation of a New Blood Glucose Monitoring System in Development for Use with Contour Plus Test Strips

Timothy Bailey, MD, FACE, CPI; Jane Wallace, BA, CCRA; Scott Pardo, PhD, PStat; Mary Ellen Warchal-Windham, BS; Bern Harrison, BA; Robert Morin, MD; Mark Christiansen, MD

AMCR Institute Inc.

Escondido, California

tbailey@amcrinstitute.com

Objective:

We evaluate the accuracy of a new blood glucose monitoring system (BGMS) in development for use with Contour Plus test strips in the laboratory and clinical settings. The BGMS features an easy-to-use wireless-enabled blood glucose meter that links to a smart mobile device via Bluetooth connectivity.

Method:

In the laboratory study, fingertip blood samples from 100 subjects were tested in duplicate using each of three test-strip lots. BGMS results were compared with YSI reference results and assessed per ISO 15197:2013 accuracy criteria (i.e., ≥95% of results within ±15 mg/dL [glucose <100 mg/dL] or ±15% [glucose ≥ 100 mg/dL] of the reference result). In the clinical study, 134 subjects with diabetes, who had never used this BGMS previously, enrolled at two clinical sites. The primary objective was to evaluate BGMS accuracy using subject fingertip self-tests based on ISO 15197:2013 accuracy criteria (see above).

Result:

In the laboratory study (glucose range, 37–526 mg/dL), 99.0% (594/600) of BGMS results met the accuracy criteria; additionally, 96.3% (578/600) of results were within ±10 mg/dL or ±10% of the YSI reference result. In the clinical study (glucose range, 44.3–474.5 mg/dL), 99.3% (133/134) of subject fingertip self-test results met the accuracy criteria; moreover, 95.5% (128/134) of results were within ±10 mg/dL or ±10% of the YSI reference result.

Conclusion:

The BGMS exceeded ISO 15197:2013 accuracy criteria in the laboratory (section 6.3) and in a clinical setting when used by untrained subjects (section 8).

Development of a New Glucose Dehydrogenase Mutant That Is Free of Xylose Interference

David Baram, PhD; Shilo Ben Zeev, MBA; Chen Guttmann, MSc; Itamar Gofberg, MSc; Rachel Diamant, MSc; Liron Hadar, PhD; Andreas Pfützner, MD, PhD

Smartzyme Biopharma

Ness Ziona, Israel

david@smzyme.com

Background:

Current blood glucose (BG) meters using glucose dehydrogenase (GDH) have been reported to be influenced by xylose concentrations, resulting in a profound bias of the BG measurement result. With more xylose replacing glucose as sweetener in “functional” food, this interference may become more important in future clinical routine.

Methods:

We used our proprietary Provolution process to generate several variants of GDHs from a bacterial source. All variants were screened using not only biochemical assays, but also electrochemical assays in order to identify the most suitable candidates. Mutants were tested both in vitro and with physiological fluids. In tests that involved physiological fluid, heparinized whole blood was freshly drawn from healthy volunteers and diabetic patients and manipulated to contain different glucose concentrations (0, 1, 3, 5, 10, 20, and 30 mmol/L) and different xylose concentrations (0, 1, 3, 5, 10, 20, and 30 mmol/L). The samples were tested with SmartZyme electrodes, and the resulting signals were analyzed in order to conclude the interference percentage of the original sample.

Results:

The commercial GDH showed approximately 20% interference with xylose at all tested glucose ranges. The finally selected mutant candidates showed no xylose interference; the signal resulting from xylose with this mutant was below the signal-to-noise ratio.

Conclusions:

By means of the Provolution technology, we have been able to develop a GDH mutant that showed no interference with xylose in a laboratory setting. Clinical studies are now warranted to demonstrate the suitability of this GDH mutant for glucose sensing devices, like self-monitoring BG systems and continuous glucose monitors.

ClampArt: Improved Algorithm for Automated Glucose Clamps

Carsten Benesch, PhD; Mareike Kuhlenkötter, MSc; Tim Heise, MD

Profil

Neuss, Germany

carsten.benesch@profil.com

Objective:

In contrast to manual techniques, automated glucose clamps (e.g., using the Biostator) offer the advantage of a bias-free assessment of the pharmacodynamic properties of blood -glucose-lowering agents and a minute-by-minute adaptation of glucose infusion rates (GIRs), thereby keeping blood glucose (BG) concentrations very close to the clamp target level. However, the Biostator algorithm leads to pronounced oscillations of both BG and GIRs, requiring mathematical smoothing procedures for the determination of time-related parameters. As ClampArt, our novel automated clamp device, offers the option to modify the clamp algorithm, we compared the unmodified Biostator algorithm (UBA) and a potentially improved clamp algorithm (ICA) regarding GIR/BG oscillations and clamp quality.

Methods:

We analyzed 90 clamps using the UBA and 44 clamps with ICA performed in 67 patients with type 1 or type 2 diabetes. BG target was 5.5 mmol/L (100 mg/dL), and the clamp duration was 12 h after a single dose of 0.2 U/kg insulin aspart in all clamps. After applying a Fourier filter (band -pass filter 0.01 to 0.1), we compared the oscillations (coefficient of variation [CV] of GIRs and BG) and control deviation (mean absolute relative deviation of BG from the clamp level) as parameters of clamp quality between UBA and ICA.

Results:

ICA showed significantly lower oscillations in both GIRs and BG versus UBA (CV GIR 48.1 ± 21.9 vs. 85.1 ± 79.4%, P = 0.0029; CV BG 4.4 ± 1.2 vs. 6.6 ± 2.9%, P < 0.0001). Control deviation did not change (1.1 ± 0.6 vs. 1.1 ± 2.5%, not significant).

Conclusion:

ICA substantially improved clamp quality over UBA, in particular oscillations in both GIRs and BG, which were reduced by 43 and 33%, respectively, and variation of control deviation is reduced by 76%. This improved algorithm will be used in future ClampArt experiments.

Development of a Hypoglycemia Alarm Based on Electroencephalography

Lykke Blaabjerg, PhD; Line Sofie Remvig, MSc; Rasmus Elsborg, PhD; Jesper Johannesen, MD, DMSc; Anne-Sophie Sejling, MD, PhD; Henning Beck-Nielsen, MD, PhD; Claus Juhl, MD, PhD

Hypo-Safe A/S

Lyngby, Denmark

lb@hyposafe.com

Objective:

Hypoglycemia is associated with increased activity in the low-frequency bands in the electroencephalogram (EEG). The objective of this research is to develop a hypoglycemia alarm device with an automated EEG algorithm predicting hypoglycemia in patients with type 1 diabetes and hypoglycemia unawareness.

Methods:

Patients with type 1 diabetes were exposed to insulin-induced hypoglycemia terminated at nadir due to significant hypoglycemic symptoms or at blood glucose levels below 1.8 mmol/L for adults or 2.2 mmol/L for children. In addition, the alarm device was tested during everyday activities in adults with type 1 diabetes, enabling detection of spontaneous hypoglycemic events. EEG was recorded and analyzed real-time using an automated EEG algorithm.

Results:

For adults, the hypoglycemia-associated EEG changes remained stable across age groups, duration of diabetes, sleep stage, and hypoglycemia awareness status. Furthermore, the hypoglycemia-induced EEG changes were unaffected by recent antecedent hypoglycemia. Spontaneous events of hypoglycemia during everyday activities were detected in time for the patient to take appropriate action, avoiding progression to severe events. In children, the automated EEG algorithm developed in adults was able to detect hypoglycemia prior to blood glucose nadir in awake state during daytime. During sleep, the current version of the EEG algorithm was unable to distinguish hypoglycemia from deep sleep patterns.

Conclusions:

Due to the consistency of the hypoglycemia-induced changes in the EEG in adults, it seems possible to develop a hypoglycemia alarm device with a general algorithm for detection of hypoglycemia, based on EEG recordings and real-time data processing. This is currently being tested in clinical trials.

Patient-Level Assessment of Cough with Technosphere Inhaled Insulin in Patients with Type 1 Diabetes Mellitus

Bruce W. Bode, MD; Jasvinder Gill, PhD; Elena Nikonova, MD; John Stewart, MSc; Robert A. Wise, MD

Atlanta Diabetes Associates

Atlanta, Georgia

bbode001@aol.com

Objective:

Technosphere insulin inhalation powder (TI) is a dry powder formulation of regular human insulin adsorbed onto Technosphere microparticles for oral inhalation in patients with diabetes. Cough is an adverse event that is reported with the use of TI. Here we describe the characteristics and the outcomes of patients reporting cough.

Method:

This patient-level analysis of a 24-week, phase 3 study in patients with type 1 diabetes mellitus (T1DM) assessed the characteristics of those who self-reported cough while using TI delivered via the Gen2 inhaler. Details regarding cough were collected in a cough-specific case report form.

Results:

One hundred seventy-four patients (mean age 37 years; 56% female) were included in the analysis, and cough was reported by 31.6% (n = 55) of patients; it was considered mild (73%), dry (82%), intermittent (69%), and occurred within 10 min of inhalation (89%). Cough was reported to start in week 1 of initiating TI for a majority of patients reporting cough (62%) . There was no difference (P > 0.05 for all) in patients who reported cough versus those who did not in terms of age (39 vs. 36 years), sex (female, 56% vs. 56%), TI dose (average dinner dose at baseline, 0.45 vs. 0.61 U/kg), A1C at 24 weeks (7.90 vs. 7.75%), incidence of any confirmed hypoglycemia (76 vs. 83%), or pulmonary function at study end (forced expiratory volume in 1 s/forced vital capacity ratio, 83 vs. 81%). Ten patients (5.7%) discontinued due to cough, which resolved within 1–2 days of discontinuation.

Conclusion:

This analysis found that the characteristics of patients with T1DM who reported cough with TI but continued treatment were similar to those who did not report cough, with comparable safety and efficacy outcomes.

Coordination of Insulin and Glucagon Delivery in Bihormonal Artificial Pancreas through Habituating Control

Jorge Bondia, PhD; Pau Herrero, PhD; José Luis Díez, PhD; Pantelis Georgiou, PhD

Universitat Politècnica de València

Valencia, Spain

jbondia@isa.upv.es

Objective:

Current approaches to bihormonal control in an artificial pancreas consider independent controllers for insulin and glucagon delivery, whereas physiological secretion of both hormones is interlinked. Bihormonal control can be posed as a multiple-input single-output control problem, with inputs differing significantly in their dynamics and cost of manipulation. Habituating control addresses this class of control problems, leading to coordination schemes among the different hormones delivery. This work aims at analyzing the benefits of insulin and glucagon delivery coordination in the framework of habituating control.

Method:

An habituating control scheme was implemented consisting of a discretized central controller designed from a given plant factorization. This central controller computes the “control effort” required for tight glucose control (virtual control action), which is then distributed among insulin and glucagon signals following a given coordination law. Different coordination laws were obtained based on different plant factorizations. In silico performance of the resulting control schemes were evaluated using the University of Virginia/Padova simulator and compared to an independent proportional-derivative controller for glucagon delivery, as currently used.

Result:

Current bihormonal control approach efficiently avoided hypoglycemia. However, coordinated action significantly improved postprandial control, with a reduction of time in hyperglycemia in approximately 50% while avoiding hypoglycemia. For equivalent closed-loop performance, factorization of faster input-output dynamics led to more efficient hormone coordination in terms of amount delivered.

Conclusion:

Bihormonal artificial pancreas performance may benefit from the coordination of insulin and glucagon delivery. These results justify the need for an increased understanding of the insulin and glucagon interplay that will translate to a more faithful bioinspired approach for bihormonal control.

Calibration of the Dynamical Tracking eA1c Algorithm with Reference HbA_1c Values: Performance in Type 1 and Type 2 Diabetes

Marc D. Breton, PhD; Jochen Sieber, MD; Frank Flacke, PhD; Boris P. Kovatchev, PhD

Center for Diabetes Technology

University of Virginia

Charlottesville, Virginia

mb6nt@eservices.virginia.edu

Background:

We previously introduced the eA1c, a new approach to real-time tracking of average glycemia and estimation of HbA _1c from infrequent self-monitoring of blood glucose data. Tested in both type 1 diabetes mellitus (T1DM) and type 2 diabetes mellitus (T2DM), the eA1c procedure yielded mean average relative deviations (MARDs) between 6.5 and 8%. We now present improved performance using calibration with one or two reference HbA_1c values.

Methods:

Reanalysis of previously published 12-month data from 120 patients with T1DM and 3-month data from 375 patients with T2DM, age (T1DM/T2DM) 39/55 years, baseline HbA_1c 8.0/7.5%, duration of diabetes 20.3/8.7 years. In the unstructured type 1 data surrogate fasting blood glucose and seven points profiles were obtained by time-of-day bins. Initial calibration was obtained using an additive offset to account for 33% of the previously published hemoglobin glycation index. Secondary calibration was performed by minimizing the weighted sum of square between previous two references and linear transform of uncalibrated eA1c values. Accuracy was assessed using MARD, days of reference measurements were excluded, and calibration was applied causally (no a posteriori calibration).

Results:

With one-point calibration, correlation between eA1c and HbA_1c improved from (T1DM/T2DM) r = 0.75/0.78 to 0.81/0.84, and its deviation from reference from MARD = 8.1/6.6 to 6.6/6.1%. Two-point calibration further improved performance to correlation r = 0.82/0.85 and MARD = 5.4/6.0%. Using the longer T1DM data, we assessed eA1c performance 3, 6, and 9 months following the second calibration; accuracy remained improved with MARD = 3.6, 4.0, and 4.7%, respectively.

Conclusions:

The eA1c procedure can be significantly improved by one- or two-point calibration schemes in both T1DM and T2DM. Accuracy was improved nearly twofold and slowly deteriorated during the 9 months following a second calibration, but MARD remained below 5%.

Safety Analysis of a 90-Day Implantable Continuous Glucose Monitoring System in the PRECISE Study

Grace Carlson, MD, MBA; Xiaoxiao Chen, PhD; Colleen Mdingi, MS; Andrew Dehennis, PhD; Eric Zijlstra, PhD

Alta Bates Summit Medical Center

Oakland, California

carlsong@sutterhealth.org

Objective:

A new implantable continuous glucose monitoring (CGM) system consisting of a fluorescence-based glucose sensor, body-worn smart transmitter, and smartphone application has been developed. The small cylindrical sensor has been designed to be inserted subcutaneously and provides 90 days of continuous measurements. Besides demonstrating functionality of the CGM system, a pivotal study was designed to demonstrate safety of the system through 90-day postinsertion use by measuring incidence of device or procedure related adverse events (AEs).

Method:

The full 81-subject study was evaluated at 90 days with a total of 44 subjects crossing the 90-day time point. In each subject, two sensors were inserted bilaterally in the upper arm, and a smart transmitter was placed over each sensor for communication. During the investigation, AEs in the clinic and during home use were documented. The primary investigator and the medical monitor independently evaluated each AE.

Result:

A total of 17 AEs in 12 (14.8%) subjects were identified as device or procedure related by the medical monitor. They included four cases of transmitter adhesive patch skin reactions, three cases of site infections, two cases of insertion procedure problem during training, two cases of insertion site pain, and one case each of depression, ingrown hair, pain in forearm muscle, numbness in hand, nausea, and headache. None of these AEs were serious or unanticipated. All AEs were resolved or stable at the study analysis point.

Conclusion:

Clinical data from a 90-day pivotal study of an implantable fluorescence-based CGM system have demonstrated that the system is safe for use in subjects with type 1 or type 2 diabetes mellitus.

Biologic Variation of Intensive Care Unit Patient Capillary Glucoses Monitored with Hematocrit-Independent Strips (2012–2015) Compared to That Monitored with Hematocrit-Dependent Whole Blood Glucose Strips (2009–2010)

Metformin Exposure Patterns Are Related to Type 2 Diabetes Nephropathy

Planar Microwave Sensor for Noninvasive Blood Glucose Sensing

K. S. Daya, PhD

Department of Physics and Computer Science

Faculty of Science

Dayalbagh Educational Institute

Uttar Pradesh, India

sdayak@gmail.com

Objective:

This paper presents a proof of concept study of a planar stripline-type microwave sensor as a point-of-care system for continuous monitoring of blood glucose.

Method:

The microwave planar spiral stripline sensor is designed at 1,800 MHz and is based on the principal of resonance. On pressing the sensor with a finger or thumb, the overall capacitance of the resonator changes, which changes the resonance frequency. This change in resonance is the measure of change in permittivity of blood, which is related to the concentration of glucose in blood. The sensor was calibrated by running parallel testing in the pathology laboratory of a local hospital on 70 patients. To increase the accuracy, a pressure sensor was incorporated in the sensor, and the glucose level was predicted using a fuzzy c-means clustering algorithm.

Result:

The sensor was tested on 70 patients and was found to have accuracy of +20%. The sensing range of the sensor is from 80 to 300 mg/dL. Efforts are in progress to increase the accuracy and sensing range.

Conclusion:

Glucose sensor presented here is microwave- integrated- circuit compatible and can be integrated into the cell phones. Results of this proof of concept testing phase promise a paradigm shift in the pathology for diabetes with further optimization in the design.

Performance of a Hybrid Closed-Loop Insulin Delivery Algorithm with Insulin Limits Designed for Home Use, Under Serial Hypoglycemic Challenges

Martin de Bock, PhD; Anirban Roy, PhD; Julie Dart; Ray Davey, PhD; Carolyn Berthold, BNurs; Benyamin Grosman, PhD; Natalie Kurtz, PhD; Fran Kaufman, MD; Elizabeth Davis, FRACP; Timothy Jones, FRACP

Department of Endocrinology of Diabetes

Princess Margaret Hospital for Children

Perth, Western Australia, Australia

martin.debock@health.wa.gov.au

Objective:

We assess a hybrid closed-loop algorithm with maximum basal insulin limits in scenarios likely to induce hypoglycemia.

Method:

Type 1 diabetes patients were commenced on a Medtronic hybrid closed-loop system (HCL) for a 4-day in-clinic study: day 1, carbohydrate ratio 25–35% more aggressive; day 2, overreading glucose sensor by 20% + aggressive carbohydrate ratio; day 3, exercise (45 min stationary cycling at 55% VO₂ max minutes) + overreading glucose sensor + aggressive carbohydrate ratio; and day 4, aggressive carbohydrate ratio. Primary outcome was hypoglycemic events (plasma glucose <63 mg/dL or symptomatic hypoglycemia with plasma glucose <70 mg/dL).

Result:

Seven studies are completed. Data are shown as median and interquartile range, or mean ± standard deviation. Twelve hypoglycemic events occurred. Four on day 1, four on day 2, two on day 3 (one pre- and one postexercise). The remaining two events occurred in one participant: one on day 3 after 130 min of open loop (HCL exited due to sensor failure) and one during exercise shortly after. There was no overnight hypoglycemia. All hypoglycemia was preceded by no algorithm-calculated insulin delivery (median 60 [55, 71] min). Median time hypoglycemia followed an insulin bolus for carbohydrate was 138 (74, 118) min. Mean plasma glucose was 144 ± 12.6 mg/dL. Time in target plasma glucose range (60–180 mg/dL) was 77.1 ± 10.7%. Time spent <60, 180–270, and >270 mg/dL was 1.8 ± 1.4, 20.0 ± 11.3, and 1.0 ± 1.5%, respectively.

Conclusion:

There was no overnight hypoglycemia despite several hypoglycemia stimuli. Although it is hard to generalize since sensors were overcalibrated, hypoglycemic risk appears to be related to food boluses and an aggressive carbohydrate ratio. It is likely carbohydrate counting accuracy will still be important in HCL.

Basal Glucose Is Just the First Measured Point . . . Right?

Paul D. Docherty, PhD; Nor Azlan Othman, PhD; J. Geoffrey Chase, PhD; Damon A Bell, MD; Jeremy D Krebs, MD, PhD

University of Canterbury

Christchurch, New Zealand

paul.docherty@canterbury.ac.nz

Objective:

Assuming measured fasting glucose (G₀) equals true basal glucose (G_B) is common in model-based insulin sensitivity (SI) tests. This study determines the impact of a model-identified G_B versus the first sampled fasting, G₀.

Method:

Fourteen participants with established type 2 diabetes mellitus in a 24-week dietary intervention study were given an insulin-modified intravenous glucose tolerance test at weeks 0, 12, and 24. The validated dynamic insulin sensitivity and secretion test (DISST) model was used to identify SI, V_G, and, additionally, G_B (three-parameter identification) versus a typical two-parameter approach. SI values and identified G_B versus measured G₀ are compared using nonparametric statistics.

Result:

SI from the three- parameter DISST identification was median (interquartile range) 3.36 × 10⁻⁴ (2.30–4.95 × 10⁻⁴) L·mU⁻¹·min⁻¹ and significantly lower than the typical two-parameter identified values: 6.38 × 10 ⁻⁴ (4.87–9.39 × 10⁻⁴) L·mU⁻¹ ·min⁻¹ (P < 0.001). Identified G _B values were lower and more consistent across all weeks than measured fasting G ₀ for all tests and by week (P < 0.001), although they were well correlated (r = 0.70). The three-parameter approach provided significantly better fit with lower error and yielded lower SI values more in-line with expected values based on participant physiology. The changes in identified S_I also tracked changes in participant state over the 24-week trial better.

Conclusion:

This analysis suggests model- identified G _B is a stable and important metric for the assessment and modeling of glucose data from individuals with type 2 diabetes.

Messaging via Electronic-Medical-Record-Based Patient Portal: Effect on Glucose Control

Kathleen Dungan, MD, MPH; Byron Noble, BS; Brian Petullo, MD

Department of Internal Medicine

Ohio State University

Columbus, Ohio

kathleen.dungan@osumc.edu

Objective:

The objective is to determine if the use of electronic messaging through an electronic medical record (EMR)-based patient portal is associated with improved glycemic control among patients with diabetes.

Method:

Patients who were seen over a 1-year period between January 1, 2013, and December 31, 2013, at a university endocrinology clinic with a diagnosis of diabetes (diagnosis code 250.xx) were identified using the institution’s Information Warehouse. The association between electronic messaging and HbA_1c was further characterized using multivariable linear regression models.

Result:

A total of 3,616 patients were identified, 1,207 who had activated the patient portal but were not using the messaging function (portal group), 867 who were active and using electronic messaging (message group), and 1,542 patients who had not activated the patient portal (inactive group). The message group was younger and more likely to be Caucasian and have type 1 diabetes compared with the inactive group. The HbA_1c was 7.7 ± 1.5, 8.4 ± 1.9, and 8.2 ± 1.8% among the message, portal, and inactive groups, respectively (P < 0.0001). After controlling for age, gender, race, and type of diabetes, electronic messaging was associated with a 0.19% (standard error 0.04%) lower HbA_1c compared with the inactive group and a 0.25% (standard error 0.04%) lower HbA_1c compared with the portal group (P < 0.0001 for both).

Conclusion:

Electronic messaging was associated with better glycemic control, even after controlling for electronic portal access and other variables.

Screening Obese Women with Prediabetes from Classes of Serum 25-Hydroxyvitamin D and Serum Parathormone Levels among African Migrants Living in Paris

Ernest Emilion, MD; Richard Emilion, PhD

Community Clinic, Paris

Paris, France

eemilion@gmail.com

Background:

It has been postulated that vitamin D may affect glucose homeostasis. We hypothesized that taking into account fasting plasma glucose concentrations in estimation of a serum 25-hydroxyvitamin D threshold below which parathyroid hormone concentrations increase could be useful to identify part of our sample usually called outliers.

Methods:

Measurements of serum 25-hydroxyvitamin D and serum parathormone were taken between February and June 2008 among 165 adult African migrant women living in Paris. All were calcium sufficient. We used receiver operating characteristic analysis to identify serum 25- hydroxyvitamin D/parathormone threshold. Analysis of variance was performed with Wilcoxon test. We used machine learning model.

Results:

A threshold of serum 25-hydroxyvitamin D of 65 nmol/L and serum parathormone of 44 ng/L was found with a sensitivity of 86%, a specificity of 83%, and a total predictive value of 98%. We identified 15% of the sample as a particular class of obese and prediabetic women with high levels of serum 25-hydroxyvitamin D and serum parathormone.

Conclusions:

Estimation with capillary glucose measurement instead of fasting plasma glucose could be a less costly method to screen glucose and vitamin D status among African migrant women. Estimating vitamin status and glucose status in other samples of the population might be of interest.

Noninvasive Glucose Monitoring with Novel High-Resolution Ultrasound Technique

Rinat O. Esenaliev, PhD

Glucowave Inc.

League City, Texas

rinesenal@gmail.com

Background:

Noninvasive, continuous blood glucose monitoring is very important for management of diabetes. For the last four decades, a number of university groups and companies have proposed and tested various noninvasive glucose monitoring techniques. Despite these enormous efforts, limited success has been achieved and noninvasive glucose monitor development remains one of the most challenging and important biomedical problems. We proposed, patented, and tested a novel technique for noninvasive continuous glucose monitoring. It is based on detection of high-resolution ultrasound signals reflected from tissues including skin.

Method:

Glucose-induced (mostly osmotic) changes in tissue can be detected with the high-resolution ultrasound technique that can be used for noninvasive glucose measurements. We developed and built high-resolution ultrasound systems with specially designed, ultrasensitive probes. We used oral glucose tolerance test or meal intake to test the systems in diabetic (both type 1 and type 2) and nondiabetic subjects. Blood glucose reference values were measured with commercially available invasive glucose meters. The probes were placed on the outer wrist area, and ultrasound signals were recorded continuously. System calibration was performed with one or two blood samples to provide real-time, continuous, noninvasive blood glucose concentrations.

Result:

High-resolution ultrasound measurements allowed for precise (1 ns) measurement of skin signal position and shift, which was linearly dependent on glucose concentration in the range of 56 to 400 mg/dL. Calibrated systems provided noninvasive glucose concentration values that closely followed the reference glucose concentration in the physiologic range. The accuracy of the noninvasive glucose monitors was approaching that of invasive glucose meters.

Conclusion:

The obtained results suggest that the proposed high-resolution ultrasound technique may provide noninvasive continuous blood glucose monitoring in diabetic and nondiabetic patients.

Recombinant Human Soluble Insulin Julphar Insulin R Is Bioequivalent to Huminsulin Normal, a European Union– Marketed Soluble Insulin

Susanne Famulla, PhD; Tim Heise, MD; Birgit Kronshage, PhD; Talal S. Al Zaher, PhD; Ulrike Hövelmann

Profil

Neuss, Germany

tim.heise@profil.com

Objective:

We compared the pharmacokinetic (PK) and pharmacodynamic (PD) properties of the recombinant human soluble insulin Julphar Insulin R (test) and Huminsulin Normal (reference) in a randomized, double-blind, crossover study performed in full compliance with the European guidelines for biosimilar insulins.

Method:

Twenty-six healthy subjects (24 male/2 female, age 38.4 ± 9.9 years (mean ± standard deviation), body mass index 24.5 ± 2.3 kg/m²) received a single subcutaneous dose (0.3 U/kg) of test or reference under automated euglycemic glucose clamp conditions (ClampArt, clamp level 5 mg/dL below fasting blood glucose, clamp duration 12 h postdose) with a washout period of 3–14 days between dosings.

Result:

Test and reference showed superimposable serum insulin (ins) and glucose infusion rate (GIR) profiles. Bioequivalence was proven for the primary PK end points (AUC_{ins 0–12 h} 1.02 [0.99; 1.04], geometric least squares means ratio test/reference [90% confidence interval (CI)]; C_{ins max} 1.02 [0.96; 1.09]) as well as for the primary PD end points (AUC_{GIR 0–last} 1.03 [0.96; 1.11], treatment ratio [95% CI]; GIR_max 1.04 [0.97; 1.12]), as the CIs did not exceed the prespecified range of 0.8 to 1.25. Bioequivalence criteria were also fulfilled for a number of additional end points (AUC_{ins 0–4 h} 1.04 [0.98; 1.12]; AUC_{ins 0–6 h} 1.02 [0.97; 1.08]; AUC_{GIR 0–4 h} 1.05 [0.95; 1.17]; AUC_{GIR 0–6 h} 1.05 [0.96; 1.14]; AUC_{GIR 6–last} 1.00 [0.86; 1.13]). Both insulins were well tolerated, and no injection-site reactions were observed.

Conclusion:

Julphar Insulin R and Huminsulin Normal show very similar PK/PD properties. This study demonstrated bioequivalence between both insulins.

Pramlintide versus Liraglutide: Assessing Incremental Benefits of Adjunctive Therapy to Closed-Loop Insulin Delivery

Rachel E. Feldman, BA; Jennifer L. Sherr, MD, PhD; Eda Cengiz, MD; Lori Carria, MS; Eileen Tichy, RD; Neha Patel, DO; Camille Michaud, MD; Miladys Palau-Collazo, MD; William V. Tamborlane, MD; Stuart A. Weinzimer, MD

Department of Pediatrics

Yale University

New Haven, Connecticut

stuart.weinzimer@yale.edu

Objective:

Closed-loop (CL) insulin delivery effectively maintains blood glucose (BG) overnight but struggles when challenged with meals. Therefore, pramlintide and liraglutide were assessed as adjuncts to CL insulin delivery. The objective of this analysis was to compare the relative effect of each adjunctive agent on glycemic excursions and insulin requirements during CL.

Research Design:

Two studies, one evaluating adjunctive pramlintide, the other liraglutide, were compared. Ten subjects (six female; age 16–23 years; A1C 7.2 ± 0.6%) completed two 24-hour sessions, one on CL alone and one on CL plus 60 µg pramlintide (CLP), after a 3–4 week outpatient dose escalation. Eight subjects (seven female; age 19–27 years; A1C 7.4 ± 1.0%) underwent the identical protocol with 1.8 mg liraglutide (CLL) after similar 3–4 dose escalation. Timing and content of meals during CL were identical within experiments; meals were not announced.

Results:

Mean BG levels were lower during CLL than CLP (131 ± 27 vs. 145 ± 28; P < 0.005) . Overall reduction in magnitude of glycemic excursions was greater with liraglutide than pramlintide (24 ± 4 vs. 13 ± 5 mg/dL; P = 0.09). However, differences by individual meal were noted, with breakfast and dinner favoring CLL and lunch favoring CLP. Pramlintide delayed the time to peak BG excursion (CLP 2.6 ± 1.5 vs. CLL 1.6 ± 0.8 h; P < 0.005) . Reductions in meal-related insulin requirements were similar in both studies (CLP 1.5 ± 1.9 vs. CLL 2.1 ± 2.1 U/meal; P = 0.51).

Conclusions:

Adjunctive liraglutide during CL was associated with lower mean BG and greater reduction in magnitude of prandial BG excursions than pramlintide and had less effect on delaying postprandial glucose appearance. Differences in lunchtime BG excursion may be due to carryover effect of pramlintide. Given its once-daily dosing regimen, liraglutide appears to be the preferable adjunct to CL insulin delivery.

Controller Performance Assessment of Model-Based Artificial Pancreas Control Systems

Lars Hansen, MD, PhD; Christian Sonesson, PhD; Fredrik Thoren, MD, PhD; Agata Ptaszynska, MD; Nayyar Iqbal, MD; John Xu, BS, MS, PhD; Eva Johnsson, MD, PhD

Bristol-Myers Squibb

Princeton, New Jersey

lars.hansen@bms.com

Objective:

We hypothesized that 1 ) individual insulin-to -carbohydrate ratios (I/C), adjusted with 24-h urinary glucose excretion (UGE) data, can derive an insulin dose effect equivalent to the glucose amount excreted in urine after 7 days treatment with the sodium-glucose cotransporter-2 inhibitor dapagliflozin (DAPA) in subjects with type 1 diabetes mellitus (T1DM) and 2) the UGE:insulin dose equivalent can guide insulin dose adjustment during treatment with DAPA.

Method:

Day 7 data on total daily insulin dose (TDD), UGE, and fasting serum beta-hydroxybutyrate (BOHB) were obtained from study NCT01498185, a 14-day study of DAPA (1, 2.5, 5, and 10 mg/day; n − 15/arm) as add on to insulin in subjects with T1DM. The insulin prediction factor 450 = TDD × I/C (for adults on short-acting insulin) was applied to derive I/C. UGE:insulin dose equivalent was calculated as UGE:I/C and expressed as percentage of TDD.

Result:

At day 7, mean observed reductions in TDD were 16, 11, 19, and 17% for the 1, 2.5, 5, and 10 mg DAPA groups, respectively. There was good agreement with the calculated mean UGE:insulin dose equivalents, which were 11, 13, 18, and 22% across groups, respectively. At day 7, BOHB slightly correlated with the observed reduction in TDD (r = 0.19) across groups. In subjects with ≤20% reduction in TDD (n=41), the maximal value of BOHB was 1.00 mmol/L. In subjects with >20% reduction in TDD (n=26), the maximal value of BOHB was 2.15 mmol/L.

Conclusion:

Insulin dose reduction in T1DM at/after the initiation of DAPA treatment should probably not exceed 20%. Further reduction may offset the glycemic benefit of UGE and elevate BOHB and could potentially increase the risk of ketoacidosis.

Toward a Sensor and Model for Cortisol as Feedback in the Bioartificial Pancreas

In Silico Evaluation of Ultra-Fast-Acting Insulins Using the Bioinspired Artificial Pancreas

Pau Herrero, PhD; Lena Kirschneck, MSc; Nick Oliver, FRCP; Desmond Johnston, FMedSci; Pantelis Georgiou, PhD; Christofer Toumazou, FRS

Center for Bio-Inspired Technology

Institute of Biomedical Engineering

Imperial College London

London, United Kingdom

pherrero@imperial.ac.uk

Objective:

Currently available rapid-acting insulin analogs (RAIs) are not fast enough to provide good postprandial glycemic control in a fully automated artificial pancreas (AP) system (no user intervention required) . Therefore, most of the existing AP systems incorporate a premeal insulin bolus to minimize postprandial glucose peaks. The advent of ultra-fast-acting insulin (UFI), showing an even more rapid onset of action and a shorter duration of action after subcutaneous administration, opens the door to a fully automated closed-loop AP system. This in silico study evaluates the benefits of using UFIs in a closed-loop glucose control setting.

Method:

For validation, the University of Virginia-Padova simulator was used. The bioinspired artificial pancreas (BiAP) controller without meal announcement was used over a 1- week simulation period in 10 adult and 10 adolescent virtual cohorts. For comparison, standard glucose metrics were used: mean glucose (G), percentage of time in glucose target 70–180 mg/dL (%inT), percentage of time below target (%<T), percentage of time above target (%>T), and risk index (RI) . The parameters of the insulin absorption model of the University of Virginia-Padova simulator were adjusted to model the dynamics of a UFI with a peak action time of 30 min.

Result:

When compared with an RAI, the utilization of a UFI within the BiAP controller improved all the evaluated glycemic metrics (P < 0.05) but %<T in the adult cohort. In adults, G 133.8 ± 5.9 vs. 127.5 ± 6.1; %inT 88 ± 7 vs. 92.8 ± 5.6; %<T 0.85 ± 1.2 vs. 0.59 ± 0.94; %>T 11 ± 6 vs. 6.5 ± 4.8; and RI 3.1 ± 1.1 vs. 2.1 ± 1.0. In adolescents, G 153.8 ± 16.3 vs. 139.7 ± 11.9; %inT 68.9 ± 12.6 vs. 80.9 ± 11.6; %<T 2.3 ± 1.7 vs. 1.2 ± 1.3; %>T 28.8 ± 11.3 vs. 17.9 ± 10.3; and RI 7.4 ± 3.5 vs. 4.6 ± 2.7.

Conclusion:

UFIs have the potential to significantly improve glycemic control in a fully automated AP system.

Scalable and Durable Optimal Glycemic Control in Insulin Users

Israel Hodish, MD, PhD; Raman Minhas, BMBS; Eran Bashan, PhD

Department of Internal Medicine

Division of Metabolism, Endocrinology, and Diabetes

University of Michigan Medical Center

Ann Arbor, Michigan

Glucose concentrations ranged from 2.2 to 28.4 mmol/L. The overall accuracy of XT BGMS according to ISO limits was 98.41%. For samples with glucose levels <5.551 mmol/L (n = 42) and ≥5.551 mmol/L (n = 63), 96.3 and 100% of BGMS measurements satisfied ISO accuracy criteria, respectively. Parkes error grid analysis revealed 99.4% of results for the BGMS to be in zone A, while 0.6% fell in zone B. Using the newly developed surveillance error grid analysis, 97.46% of measurements fell within the no -risk zone. The relative differences of measurements using control solutions (26 control tests per strip lot) were within the limits defined by QUALAB (errors within ±10% of control solution target values) and Rilibäk (errors within ±11% of target value) POC guidelines.

Conclusion:

This study demonstrates that XT BGMS fulfills the ISO 15197:2013 accuracy limits under routine hospital conditions, as well as stringent Swiss and German quality control requirements for the POC setting.

Use of a Bihormonal Artificial Pancreas System with Exercise Announcement in Type 1 Diabetes to Prevent Hypoglycemia

Peter G. Jacobs, PhD, MSEE; Joseph El Youssef, MD, MCR; Navid Resalat, MSEE; Ravi Reddy, MSEE; Deborah Branigan, BA; Nicholas Preiser, BSEE; John Condon, MSBE; Jessica R. Castle, MD

Department of Biomedical Engineering

Oregon Health and Science University

Portland, Oregon

jacobsp@ohsu.edu

Objective:

We developed and tested a wireless bihormonal artificial pancreas system that adjusts dosing after an exercise announcement to reduce exercise-related hypoglycemia.

Method:

Adult subjects with T1D underwent three 20 h sessions: open loop (OL), closed loop (CL), and CL with exercise announcement (CLX). Exercise announcement stopped insulin for 30 min, then 50% reduction for 60 min, and increased glucagon by twofold for 1.5 h. Parameters were selected via in silico testing. G4 sensor values were pushed to a Google Nexus phone that sent delivery commands to t:slim pumps for subcutaneous insulin and glucagon infusion. After an overnight stay, subjects exercised for 45 min at 60% of maximum heart rate. Subjects were given two meals and wore a Zephyr heart rate monitor and accelerometer.

Result:

Twelve subjects participated in 36 20 -h studies. Hypoglycemic events were reduced from 10 during OL, to 6 during CL, to 1 during CLX (P < 0.01, CLX vs. OL). The CLX reduced the number of hypoglycemia events versus CL (P < 0.05). Seven, five, and one subjects had a hypoglycemic event that required carbohydrates during the OL, CL, and CLX (P < 0.05 for CLX vs. OL). Mean glucose was 156 mg/dL for OL, 145 mg/dL for CL, and 149 mg/dL for CLX. Mean insulin delivery was similar across conditions. The mean amount of glucagon delivered was higher during the CLX (286 µg) compared with CL (180 µg; P < 0.01).

Conclusion:

Automated insulin and glucagon delivery effectively controlled glucose levels and prevented hypoglycemia. Incorporating exercise announcement that adjusted insulin and glucagon delivery significantly reduced hypoglycemic events compared with standard CL. Incorporating an exercise announcement or automatic exercise detection into the design of the artificial pancreas is advantageous for preventing hypoglycemia.

Multianalyte Monitoring via Implantable Biosensors Predicts Exhaustion

Michail Kastellorizios, PhD; Sagar Vaddiraju, PhD; Allen Legassey, BS; Fotios Papadimitrakopoulos, PhD; Diane J. Burgess, PhD

Biorasis Inc.

Storrs, Connecticut

michail@bio-orasis.com

Objective:

Our aim is to develop a biosensor platform capable of continuous real-time monitoring of metabolic level changes to predict exhaustion during physical exercise. Monitoring of more than one analyte simultaneously is necessary to provide a holistic view of the metabolic changes that lead to exhaustion. This approach to metabolic monitoring is of great importance to diabetic patients, especially when undertaking intense physical activity since it can allow prediction of imminent metabolic changes and enable the individual to take preventive measures.

Methods:

Our previously developed coil-type electrochemical sensors were implanted transcutaneously in normal and diabetic rats, and a wireless dual transmitter collected data continuously from glucose and lactate sensors. Microdialysis samples were taken from both the jugular vein and subcutaneous tissue and analyzed for glucose and lactate via a YSI analyzer were used as a control. Forced exercise was performed on a treadmill (IITC Life Sciences).

Results:

Glucose and lactate trends indicated two major metabolic events following commencement of exercise: 1) rest to exercise transition and 2) exhaustion. Individually, glucose and lactate monitoring revealed only one of these two events. When glucose and lactate changes over time are combined, the result is a multianalyte biomarker, which is more informative than independent assessment of the single analytes. Additionally, the observed trends are more sensitive in the subcutaneous tissue (an implantation -friendly peripheral tissue) compared with the blood. This unexpected observation was confirmed in normal as well as type 1 diabetic rats.

Conclusions:

With the current work, we have extended the capabilities of our biosensor platform by incorporating lactate monitoring and identifying an appropriate multianalyte biomarker. These findings can be implemented in new multianalyte continuous monitoring technologies for more accurate insulin dosing, as well as for exhaustion prediction studies based on objective data rather than the subjects’ perception.

Glucowizzard: An Injectable, Wireless Biosensor for Continuous Glucose Monitoring

Michail Kastellorizios, PhD; Sagar Vaddiraju, PhD; Allen Legassey, BS; Diane J. Burgess, PhD; Faquir Jain, PhD; Fotios Papadimitrakopoulos, PhD

Biorasis Inc.

Storrs, Connecticut

michail@bio-orasis.com

Objective:

Biorasis Inc. and the University of Connecticut are developing a totally implantable continuous glucose monitoring device. Glucowizzard is engineered at the smallest possible footprint (0.5 × 0.5 × 0.5 mm). This miniaturization is made possible by utilizing light for powering as well as wireless communication. The objective of this study was to optimize sensor parameters and provide in vitro and in vivo proof of concept.

Methods:

Glucowizzard uses a first-generation Clark-type enzymatic sensor based on three -electrode (working, reference, and counter) electrochemical transduction for detecting glucose levels. An integrated circuit is used to interact with the glucose sensing element and transmit data wirelessly through the skin to a proximity communicator. The sensors were tested both in vitro (in phosphate-buffered saline) and in vivo in normal rats. Blood glucose concentrations were used as a reference.

Results:

Glucowizzard showed a linear response over a wide glucose range in vivo and in vitro. Interestingly, the sensor sensitivity increased after 4 days of implantation. Implanted sensors were successfully used to monitor glucose trends in the subcutaneous tissue of rats, which closely followed the changes in blood glucose levels.

Conclusions:

This study provided proof of concept for the in vivo functionality of Glucowizzard, an injectable, wireless glucose biosensor. The sensors showed a linear response over a wide glucose concentration range both in vivo and in vitro. Moreover, blood glucose levels correlated with trends obtained from Glucowizzard, which is paramount for tight glucose control and diabetes management.

Cytotoxicity of Photopatternable Hydrogels for Implantable Glucose Sensors

Michail Kastellorizios, PhD; Dongliang Wang, MS; Zhe Li, PhD; Diane J. Burgess, PhD; Faquir C. Jain, PhD; Fotios Papadimitrakopoulos, PhD

Biorasis Inc.

Storrs, Connecticut

michail@bio-orasis.com

Objectives:

The function and lifetime of implantable electrochemical-based continuous glucose monitoring (CGM) devices depend on the stability of glucose oxidase (GO_x), the enzyme responsible for glucose detection. Polyethylene glycol (PEG)-ylated hydrogels offer an opportune venue to minimize biofouling, while retaining their highly hydrated state to prevent enzyme denaturation. Here we report the cytotoxicity studies of a novel PEGylated hydrogel capable of being photocrosslinked in its fully hydrated state. This hydrogel allows for CGM sensors with long-term stability and optimal performance.

Methods:

A PEGylated copolymer was synthesized via a free-radical polymerization using an azobisisobutyronitrile initiator. This PEGylated copolymer was crosslinked into hydrogel nanoparticles and suspended in cell culture media with concentrations from 0.01 to 100 µg/mL. Subsequently, mouse dermal fibroblasts (cell line L929) were incubated for 72 h with various nanoparticle concentrations. Cell viability was assessed via CellQuanti-Blue Cell Viability Assay Kit from BioAssay Systems (Hayward, CA).

Results:

Typically, nanoparticle cytotoxicity studies are conducted at very low concentrations (down to parts-per- billion levels) and for only 24- or 48-h incubation periods. In this study, cells were exposed to high nanoparticle concentrations for 72 h. It was determined that these PEGylated copolymer nanoparticles show no cytotoxicity for up to 72 h for the entire concentration range.

Conclusions:

Our studies indicate that photopatternable PEGylated hydrogels are ideal matrices for covalent immobilization of GO_x enzyme, enabling the fabrication of CGM devices. In this study we demonstrate that these photocrosslinked hydrogel nanoparticles are not cytotoxic and therefor pose no threat to the local tissue. Long-term in vivo studies are currently underway to determine the ultimate stability of these photopatternable CGM devices.

Automated Insulin Delivery “In the Wild”

Robert Kircher, MS; Don Matheson, MS; Richard Mauseth, MD

Dose Safety Inc.

Seattle, Washington

robert.kircher@dosesafety.com

Objective:

Automated insulin delivery “in the wild” is more challenging than in hospital clinical research centers, diabetes camps, or hotels where health care professionals and engineers are closely standing by. Challenges include sensor changes/calibrations, system start-up/shutdown, continuous glucose monitor (CGM) communication errors, infusion set changes, insulin refills, and battery changes. Each of these challenges could occur for arbitrarily long periods of time. Furthermore, a normally operating CGM can produce physiologically impossible glucose values, such as following sensor calibration.

Method:

Our artificial pancreas autodosing mode control module enables closed-loop insulin dosing only when interfacing subsystems are operating normally and the CGM signal is physiological. The software module turns off automated insulin dosing and reverts to the user’s preprogrammed basal when something is wrong. The system resumes automated dosing when the problems are resolved.

Result:

The automated insulin doses at any time depend only on current CGM, rate, and acceleration and do not depend on prior insulin administration. This means that our fuzzy logic dosing controller can tolerate arbitrarily long challenges and time gaps in the CGM data input stream. Insulin on board (IOB) is taken into account 1 ) by the expertise model of the fuzzy logic dosing controller and 2) in the glucose prediction module. The prediction module uses the IOB and CGM histories to calculate the glucose predictions. If low blood glucose is predicted, automated insulin dosing is suspended until CGM begins rising.

Conclusion:

The modular system architecture and design of the dose safety controller enables automated insulin delivery “in the wild.”

The Importance of Multiple Sensors in Continuous Glucose Monitoring

Timothy Knab, BS; Ari Pritchard-Bell, BS; Gilles Clermont, MD; Robert Parker, PhD

University of Pittsburgh

Pittsburgh, Pennsylvania

tdk17@pitt.edu

Objective:

Targeted glucose control (TGC) may improve outcomes in critically ill patients by reducing the incidence of hypoglycemia and the severity of stress hyperglycemia. Automated TGC requires high-frequency intervention and rapid measurement of glucose, achievable via continuous glucose monitors (CGMs). The efficacy of TGC is directly related to the measurement quality of the CGM. We explored whether a single sensor is sufficient to achieve TGC in the intensive care unit.

Methods:

Composite blood glucose (CBG) profiles, dynamically weighted linear combinations of duplicate CGM signals from critically ill patients (n = 8), were regressed against low -frequency reference finger sticks, from which a distribution of CGM error derivatives was calculated. Integration of random samples from the distribution generates error trajectories, which, when added to CBG profiles, simulate a CGM. Simulated CGMs are used singly and in duplicate for moving horizon estimation (MHE) of blood glucose.

Results:

CBG profiles follow reference finger stick measurements within ±5%. Stochastic simulations show that 88% of actual CGM error trajectories fall within the 95% confidence envelope of the simulations. MHE using a single simulated noisy sensor results in blood glucose estimates with a mean absolute relative deviation (MARD) of 13.80%, but the use of two sensors recovers MARD to a clinically acceptable 10.29%.

Summary:

Blood glucose profiles are reconstructed from noisy CGM data and used to create a model of sensor error. Blood glucose estimates from simulated noisy single sensor measurements result in a MARD beyond an acceptable clinical level, whereas the use of two CGMs allows for estimation of blood glucose with a clinically acceptable MARD. TGC with a single CGM is insufficient for high-quality control in critically ill patients.

Hemoglobin A_1c and Self-Monitored Average Glucose: Validation of the Dynamical Tracking eA1c Algorithm in Type 1 Diabetes

Boris P. Kovatchev, PhD; Jochen Sieber, PhD; Frank Flacke, PhD; Marc D. Breton, PhD

Center for Diabetes Technology

University of Virginia

Charlottesville Virginia

bpk2u@eservices.virginia.edu

Background:

We previously introduced the eA1c, a new approach to real-time tracking of average glycemia and estimation of HbA_1c from infrequent self -monitoring of blood glucose (SMBG) data, which was developed and tested in type 2 diabetes. We now validate eA1c in type 1 diabetes and assess its relationship to the hemoglobin glycation index (HGI), an established predictor of complications and treatment effect.

Methods:

Reanalysis of previously published 12-month data from 120 patients with type 1 diabetes, age 39.15 (14.35) years, 51/69 males/females, baseline HbA_1c 7.99% (1.48%), duration of diabetes 20.28 (12.92) years, and number SMBG/day 4.69 (1.84) . Surrogate fasting blood glucose was generated by selecting the average of SMBG values between 5:00 a.m. and 8:00 a.m. on any given day; seven-point daily profiles were derived from these unstructured SMBG data similarly in seven time-of-day bins over the past 2 weeks. We then applied the previously reported eA1c method without any changes. Following the literature, we calculated HGI = HbA_1c − (0.009 * Fasting Blood Glucose + 6.8).

Results:

The correlation of eA1c with reference HbA _1c was r = 0.75, and its deviation from reference was mean absolute relative deviation = 7.98%; 95% of all eA1c values fell within ±20% from reference. The HGI was well approximated by a linear combination of the eA1c calibration factors: HGI = 0.007552 * q1 + 0.007645 * q2 − 3.154 (P < 0.0001); 73% of low- versus moderate high HGIs were correctly classified by the same factors as well.

Conclusions:

The eA1c procedure developed in type 2 diabetes to track in real time changes in average glycemia and present the results in HbA_1c- equivalent units has now been validated in type 1 diabetes. The eA1c calibration factors are highly predictive of the HGI, thereby explaining partially the biological variation causing discrepancies between HbA_1c and its linear estimates from SMBG data.

Acute Effect of Low-Dose Pramlintide on Islet and Gastrointestinal Peptides in Type 1 Diabetes Mellitus

Yogish C. Kudva, MBBS; Vikash Dadlani, MBBS; Ling Hinshaw, MD, PhD; Adil E. Bharucha, MBBS, MD; Ananda Basu, MBBS, MD

Mayo Clinic

Rochester, Minnesota

kudva.yogish@mayo.edu

Introduction:

Postprandial hyperglycemia continues to be a challenge even with closed-loop control systems for the management of type 1 diabetes (T1D). Pramlintide improves postprandial hyperglycemia by delaying gastric emptying and decreasing postprandial glucagon. To better understand the mechanism of its effects, we explored the impact of prandial pramlintide administration on other islet hormones and gastrointestinal peptides.

Methods:

We conducted a mixed-meal study including 75 g of labeled glucose given as Jell-O in our clinical research unit in 12 (6 M) C-peptide-negative T1D patients (age 44.4 ± 14.1 years, HbA_1c 7.5 ± 1.1%, body mass index 28.9 ± 5.7 kg/m²) on continuous subcutaneous insulin infusion (CSII) investigating glucose kinetics, islet hormones, and gastrointestinal (GI) peptides with and without 30 µg pramlintide subcutaneous administration at time zero. Patients with active GI disease were excluded from the study. We measured integrated area under curve (iAUC) for plasma glucagon-like peptide-1 (GLP-1), ghrelin (total and active), somatostatin, and leptin up to 120 min.

Results:

Whereas pramlintide decreases postprandial glucose excursion from 0–120 min (P < 0.001), it had no effect on it from 0–360 min. iAUC 0–120 min for GLP-1 (mean 163.6 ± 1208.7 vs. 200.4 ± 1922.0 pmol/L*min), total and active ghrelin (mean 4584.4 ± 4531.5, −506.7 ± 2042.9 vs. −8,943.5 ± 9,713.9, −1,621.4 ± 5,185.3 [pg/mL]*min) somatostatin (mean 14.2 ± 345.0 vs. −181.9 ± 463.1 pmol/L), and leptin (mean −250.3 ± 358.1 vs. −66.4 ± 202.1 [ng/mL]*min) were not different between the two study conditions.

Conclusions:

Low -dose pramlintide acutely decreases postprandial glucose excursion in T1D on CSII with no effect on any of the islet and GI peptides measured.

Adaptive Blood Glucose Predictions Using Untrained Models: a Performance Comparison on the Continuous Glucose Monitor Rate of Change

Alejandro J. Laguna Sanz, PhD; Eyal Dassau, PhD; Francis J. Doyle III PhD

University of California, Santa Barbara

Santa Barbara, California

allasan@engineering.ucsb.edu

Objective:

Accurate model predictions are a key requirement for efficient glucose control and ultimately for a fully functional artificial pancreas for type 1 diabetes. It is common to evaluate the prediction capabilities of trained models by comparing them to the performance of a zero order holder (ZOH) predictor. Reports in literature show that most current glucose prediction models perform only marginally better than a ZOH in average. In this work, we assess the performance of an adaptive physiology-based untrained prediction model for longer prediction horizons and its dependence on the rate of change (ROC) of a continuous glucose monitor (CGM).

Method:

A three-state physiology-based minimal bilinear model is used in combination with an unscented Kalman filter (UKF) for glucose prediction based on data from a CGM. The UKF corrects the models state estimation and simultaneously adapts one of its parameters. The predictions are tested on a previously available clinical data set (n = 64). Prediction horizons tested range from one to nine samples ahead.

Result:

Glucose periods with ROC between −1 and 1 mg/(dL·min), accounting for more than 80% of the data samples, are better predicted using the ZOH. For ROC greater than 1 mg/(dL·min) or less than −1 mg/(dL·min), the combination of model + UKF yields superior predictions than the ZOH. The residuals are 8 and 10% smaller than the ZOH residuals for the four- and eight-steps-ahead predictions, respectively.

Conclusion:

The untrained combination of model + UKF surpasses the predictions of a ZOH for distant prediction horizons and rapidly changing CGM signals. High gradient CGM periods are critical stages, usually involving great uncertainty, in which to assess a controller’s performance. Improving glucose predictions in these periods is crucial for improved glycemic outcomes under closed loop.

Evaluation of Insulin Delivery with V-Go Combined with a Glucagon-Like Peptide-1 Receptor Agonist in Patients with Suboptimally Controlled Type 2 Diabetes Mellitus

Rosemarie Lajara, MD; Carla Nikkel, BS

Diabetes America

Plano, Texas

rlajara@diabetesamerica.com

Objective:

Combining insulin therapy with glucagon- like peptide-1 (GLP-1) has been associated with improvements in A1C and lower insulin requirements in patients with type 2 diabetes mellitus. No data have been published to evaluate if switching insulin delivery to V -Go, a disposable insulin delivery device, would impact A1C results in patients not achieving glycemic goals with insulin therapy + GLP-1.

Method:

Electronic medical records were queried to identify patients with an A1C >7.0% switched from previous insulin therapy + GLP-1 to insulin delivery with V-Go + GLP-1.

Result:

Thirty patients were identified: 13 patients administering basal insulin only + GLP-1 (mean A1C 9.1 ± 1.3%, weight 104 ± 25 kg, insulin total daily dose [TDD] 65 ± 28 U/day, duration of diabetes mellitus [DM] 13 ± 7 years) and 17 patients administering basal- bolus insulin + GLP-1 (mean A1C 9.1 ± 1.5%, weight 111 ± 17 kg, TDD 147 ± 63 U/day, duration of DM 13 ± 6 years). When switched to V -Go both groups demonstrated significant reductions in A1C (P = 0.002) . Mean changes in A1C (95% confidence interval) for those on basal insulin at baseline and basal-bolus at baseline were −1.9 (–0.8, −3.0) and −1.4 (−0.6, −2.2), respectively. Insulin was significantly reduced for those in the basal-bolus group, 147 to 70 U/day ( P < 0.0001). Redistribution of the TDD with V-Go proved beneficial in the basal-insulin-only group, where the TDD was reduced from 65 to 59 U/day.

Conclusion:

Delivering insulin with V-Go + GLP-1 resulted in significant reductions in A1C and lower insulin requirements in patients previously administering insulin using traditional injection methods. Improvements contributed to convenience and efficiency of insulin delivery and adherence to regimen; however, further studies are needed.

Using the V-Go Disposable Insulin Delivery Device in Patients Prescribed High Insulin Doses: A Comparison of Clinical Results between High and Moderate Insulin Dose Groups

Rosemarie Lajara, MD; Carla Nikkel, BS

Diabetes America

Plano, Texas

rlajara@diabetesamerica.com

Objective:

We evaluate the change in insulin requirements and glycemic control for patients with uncontrolled type 2 diabetes mellitus prescribed high insulin total daily doses (TDDs) compared with moderate doses when switched to V-Go. No data have been published evaluating what clinical impact baseline TDD has when using V-Go.

Method:

Electronic medical records queried to identify patients switched from previous insulin therapy to insulin delivery with V-Go. Patients (n = 104) were stratified by moderate baseline doses (<100 U/day) or high baseline doses (≥100 U/day) and compared using units per kilogram.

Result:

Sixty-six patients with a baseline insulin TDD <100 U/day (0.70 ± 0.27 U/kg) (mean A1C 9.3 ± 1.4%, weight 92 ± 20 kg) and 38 patients with a baseline TDD ≥100 U/day (1.37 ± 0.37 U/kg) (mean A1C 9.5 ± 1.6%, weight 106 ± 19 kg) were evaluated. Baseline characteristics were similar for all variables including age (57 years) and diabetes duration (14 years). Baseline insulin dose was significantly greater for those prescribed ≥100 U/day (P < 0.0001). Following a 6-month average duration using V-Go, both groups had significant reductions in A1C and units per kilogram insulin requirements. Mean changes in A1C (95% confidence interval) were similar for those on <100 and ≥100 U/day at baseline −1.5 (−1.2, −1.9) and −1.7 (−1.1, −2.3), respectively. Despite significant differences in baseline insulin doses, both groups had almost identical TDD using V-Go: 0.59 vs. 0.64 U/kg (P = 0.25). Those prescribed ≥100 U/day at baseline decreased TDD from 143 to 67 U/day (P < 0.0001).

Conclusion:

Patients prescribed high doses of insulin before switching to V- Go experienced similar A1C reductions administering similar insulin rates to those prescribed moderate insulin doses at baseline.

Widespread Clinical Experiences Validating a Blood-Based Artificial Pancreas

Arnold J. Lande, MD

Northport Navigable Waters Institution

Northport, Michigan

alande1@juno.com

Objective:

Told since the 1980s that directly and continuously accessing blood over a lifetime is impossible, research and development teams worldwide have been attempting to plug their artificial pancreases into remote subcutaneous or other time-delaying tissues. In light of the difficulties encountered, including having to predict the future of a volatile diabetes patient, a direct blood-based artificial pancreas (BBAP) deserves another look.

Method:

This presentation reviews widespread clinical data that reflect on a potential wearable BBAP.

Result:

The Biostator bedside blood-based console, Food and Drug Administration–approved since the 1980s, has demonstrated platinum standard capability to “place a clamp” on blood glucose. Arteriovenous fistula (AVF) is the gold standard for intermittent hemodialysis access. In case of AVF failure, accessory vein obliteration, which routes all fistula flow through one least damaged outflow vein, has been demonstrated to markedly improve that vessel’s function and durability. It is not a great stretch to project that a virgin in situ debranched vein fistula graft (VFG), revised or replaced as needed, might serve a lifetime. It is also not a stretch to recognize that point compression over the VFG makes atrioventricular differential pressure accessible to perfuse an extracorporeal device. Widespread physician and patient acceptance of AVF for hemodialysis and similar acceptance for the more cosmetic VFG, for a BBAP, is likely.

Conclusion:

The diabetes community has agreed for 95 years on the desirability of a practical, reliable, accurate means to automatically assess and administer insulin and other medications, to thereby eliminate life-threatening hypoglycemia and reduce the devastating and costly complications of hyperglycemia and also reduce costs through steady, nonsequestering, and nondegrading medications administration directly into the blood.

Online Detection of Sensor Faults from Continuous Glucose Monitoring, Meal, and Insulin Delivery Information

Yenny Leal, PhD; Simone Del Favero, PhD; Luis Gonzalez-Abril, PhD; Andrea Facchinetti, PhD; Giovanni Sparacino, PhD; Claudio Cobelli, PhD

University of Padova

Padova, Italy

yenny@dei.unipd.it

Objective:

Continuous glucose monitoring (CGM) faults, even if of limited time duration, can expose diabetic patients to severe risks and deteriorate effectiveness of automated insulin therapies. Typical faults originated by mechanical sources include occasional spikes and temporary attenuations induced by, e.g., pressure made by the patient lying or sleeping on the sensor. In this work, an online method based on a geometric mean support vector machine (GSVM) is used to detect such failures by exploiting CGM, meal, and insulin delivery information.

Method:

One hundred virtual patients were generated by using the University of Virginia/Padova type 1 diabetic patient simulator. For each patient, 7 fault-free days of closed-loop control with three meals per day were simulated starting from midnight. Then several artificially generated fault episodes were added at random time instants. For each fault type, different amplitudes (−7.5, −10, −15, −20, −25, and −30 mg/dL) and durations (5, 10, 20, 30, 40, 50, and 60 min) were considered. An individualized GSVM model was trained offline using past CGM sensor readings, rate of appearance of glucose and insulin. Then the model and inputs were used online to detect sensor faults. The first 4 data days were used for training. The remaining 3 days were used for testing, simulating an online scenario.

Result:

GSVMs with radial basis function kernel were tuned and validated using fivefold cross-validation. From all patients, 75,940 fault-free readings and 10,160 failures readings were obtained for testing, and 92.8% accuracy, 94.0% specificity, and 83.4% sensitivity were reported. Regarding amplitude, sensitivity was between 89.0 and 93.0%, except for 7.5 mg/dL (57.6%) and 10 mg/dL (72.3%). Regarding duration, sensitivity was around 83.0%, except for 5 min (94.1%), 10 min (92.2%), and 60 min (78.0%).

Conclusion:

The method shows good sensitivity by detecting CGM faults for mid and large amplitudes (above 10 mg/dL) and for duration faults lower than 60 min.

MT-Diet: Automated Diet Assessment Using Myo Wristband and Thermal Images

Junghyo Lee; Ayan Banerjee, PhD; Prajwal Paudyal; Sandeep Gupta, PhD

Arizona State University

Tempe, Arizona

jlee375@asu.edu

Objective:

Diet monitoring and control is an important factor for individuals with diabetes. Manual self-monitoring techniques for diet suffer from drawbacks such as low adherence, underreporting, and recall error. Camera-based applications that automatically extract type and quantity of food from an image of the food plate can potentially improve adherence and accuracy. However, state-of- the-art systems have fairly low accuracy for identifying cooked food (only 63%). To overcome these drawbacks, we introduce an automated diet assessment system, MT-Diet, that can identify cooked food with an accuracy of 88.5%.

Method:

MT-Diet is a smartphone-based system that interfaces a thermal sensor and a Myo wristband with a smartphone. Using this system, a user can take both thermal and visual images of her food plate with just one click. While eating the Myo device monitors hand movements and estimates number of bites taken by the user. We used a database of 80 frozen foods to evaluate three core components: 1) food segmentation, separating food items from the plate, 2) food identification, determining type of food, and 3) food intake measurement.

Result:

MT-Diet food segmentation methodology is fully automatic and requires no user input as opposed to recent works, and the accuracy was 92.5%. The accuracy of food identification using a support vector machine with color, texture and histogram of gradients features is 88.5%. The accuracy of detecting a bite using accelerometer data from Myo device was 91.37%.

Conclusion:

MT-Diet can potentially be an inexpensive, real-time, easy -to-use, and fully automated diet monitoring system. The tool can also be used to conduct clinical studies to develop models of meal patterns that can be incorporated to design a better artificial pancreas.

The Use of Insulin Pump Technology to Advance Safe and Effective Care

Kathy Little, BSN, RN, CDE

Missouri Baptist Medical Center

St. Louis, Missouri

klittle@bjc.org

Objective:

Traditionally insulin pump download software was developed for endocrinologist or patient use in assessing glycemic patterns. Upon admission, patients are fearful of losing the control they have achieved with the pump if the hospital directs them to remove it due to prescriber lack of insulin pump knowledge. Our institution developed a process for obtainment of pump settings using pump software for home medication list and appropriate physician medication order entry. Our goal was to use the technology to support more precise use of pump dosing by providers versus discontinuing the pump and ordering basal-bolus dosing.

Method:

A multidisciplinary team revised current policy to support the new practice. Diabetes champions were educated on the use of the software to obtain individual patient insulin dosing data. A reference manual was created for each inpatient unit and the emergency department. Education was provided for all prescribers, and dedicated computers for software download were identified.

Result:

Data were collected on 80 patients over 12 months. The average blood glucose level during hospitalization improved over the average preadmission blood glucose level. Hypoglycemic and hyperglycemic events decreased with the continued use of the insulin pump. Preadmission insulin dosing was maintained throughout the hospitalization for most patients. Patient lack of knowledge related to insulin pump management was identified and provided opportunities for targeted education.

Conclusion:

The ability to review preadmission insulin pump data resulted in more accurate insulin dosing and improved glycemic control and provided baseline information for the purposes of medication reconciliation.

Late-Breaking Human Clinical Results from a Second -Generation Long-Term Fully Implanted Continuous Glucose Monitoring System

Detectability of Confounding Effects in Glycemic Diary Data from In Silico Patients with Diabetes

Erin J. Mansell, BE(Hons); Paul D. Docherty, BE(Hons), PhD; J. Geoffrey Chase, BS, MS, PhD

Center for Bioengineering

University of Canterbury

Christchurch, New Zealand

erin.mansell@pg.canterbury.ac.nz

Objective:

There are a number of influences that can have a significant impact on glycemia but have sufficiently low incidence rates to be ignored by most attempts to model the glucose excursions of individuals with diabetes. This research aims to determine how much data are required to model specific parameters that capture these confounding effects.

Method:

A type 1 diabetes patient model was used to simulate glucose for a year and included weekly instances of exercise and monthly instances of high emotional stress. The data were sampled at a rate of six samples per day and mimicked finger-prink error. Identified values for insulin sensitivity and the magnitude of glycemic effect of exercise and stress were measured in increasing ranges of data in a Monte Carlo analysis.

Result:

As the range of data increased, the identified model parameters became more accurate to the parent model values. Since insulin sensitivity had a consistent effect on glycemia, it achieved a coefficient of variation (CV) <10% prior to 7 days of data. The confounding effects of stress and exercise had CVs within 10% after approximately 100 and 150 days, respectively.

Conclusion:

This research has shown the length of diary data required to quantify patient-specific effect sizes for some confounding glycemic effects in diabetes. However, this in silico analysis must be repeated in heterogeneous cohort of diabetes to capture the interpatient effect-size variability.

Perceived Knowledge and Attitudes of Certified Diabetes Educators Regarding Genetic Testing for Type 2 Diabetes Mellitus

Jacqueline R. Martz, DNP, ACNS, BC, RN, CDE

Rhodes State College

Lima, Ohio

martz.j@rhodesstate.edu

Objective:

The purpose of this study is to identify the relationship between knowledge and attitudes of genetic testing for type 2 diabetes mellitus among registered nurses that are Certified Diabetes Educators (CDEs).

Method:

Data were collected from 2,000 registered nurses who are CDEs using a mailed survey. Three instruments were used to collect data: demographics and background data, perceived knowledge of genetic testing scale, and attitudes toward genetic testing instrument.

Result:

Participants were predominantly females aged 50 years and older with a bachelor’s to a master’s degree. Years as a CDE were evenly spread from 1–30 years. Participants were also predominantly used in diabetes education 75– 100% of the time. Those reporting to have had no previous genetic training were greater than those who did. Means of the complete perceived knowledge tool and mean of the favorable and reserved attitudes subscale were correlated. There were no significant correlations found.

Conclusion:

The purpose of this study was to examine the perceived knowledge of CDEs regarding genetic testing for type 2 diabetes mellitus. The lack of significant findings suggests that future research involves more investigation into nursing education regarding the types, amounts, and specifics of genetic training that nurses receive. Other recommendations would be for more longitudinal studies of CDE knowledge and attitudes as chronic disease and genetics information grows. Also, further examination of the independent study variable may be warranted to examine what types of relationships exist between the statements.

Pharmacokinetic Study of Mixed-Diet Glucose Concentration Responses in Type 1 Diabetes Patients on Closed-Loop Pump Therapy

Derrick K. Rollins Sr., PhD; Yong Mei, MS

Iowa State University

Ames, Iowa

drollins@iastate.edu

Objective:

An automatic insulin delivery system (AIDS) must be capable of tight glucose control for major disturbances. Thus our objective is to present our novel AIDS approach that seeks to directly nullify the effects of eating, activity, and stress, resulting in a substantial reduction on the variability of blood glucose concentration (BGC).

Method:

Feedforward control (FFC) has the potential to completely cancel the effects of any disturbance that is measured and effectively modeled. This work applies a novel FFC modeling method and control algorithm to nullify critical disturbances that cause large variation in BGC. By removing the effects of these disturbances, the AIDS has the potential of controlling BGC as tightly as the human pancreas.

Result:

In this work, 11 cases of 2 weeks of free-living data collection on subjects with type 1 diabetes are modeled with 13 input variables. The activity variables were collected using the SenseWear Pro3 Body Monitoring System. For a FFC model, the critical performance measure is the correlation of the measured BGC and the fitted BGC (r_fit ). The average r_fit values of the 11 cases for training, validation, and testing were 0.64, 0.63, and 0.61, respectively. The closeness of the values for each of the three sets is an indication that the models did not significantly overfit the data on unmeasured disturbances. The potential impact of this fit is a reduction of approximately 90% the variability of BGC as measured by the standard deviation from the target value.

Conclusion:

Thus an AIDS based on this FFC approach has the potential to reduce the variation of BGC to the level of those without diabetes.

Night-to-Night Variability of Basal Insulin Requirements in Adults with Type 1 Diabetes during 12-Week Closed-Loop Insulin Delivery

Yue Ruan, MEng; Hood Thabit, MD; Sibylle Dellweg, MD; Julia K. Mader, MD; Carsten Benesch, PhD Manuel Holzer, MSc; Mark Evans, MD; Thomas R. Pieber, MD; Sabine Arnolds, MD; Roman Hovorka, PhD; on behalf of the AP@home consortium

Wellcome Trust-MRC Institute of Metabolic Science

University of Cambridge

Cambridge, United Kingdom

yr233@cam.ac.uk

Objective:

We evaluate night-to-night variability of basal insulin requirements (BR) in adults with type 1 diabetes (T1D) durin closed-loop insulin delivery.

Method:

A multicenter, randomized crossover study involved 33 subjects with T1D (15 females, age 40 ± 9.5 years, body mass index [BMI] 25.4 ± 4.4 kg/m², HbA _1c 8.5 ± 0.6%) who applied sensor-augmented pump therapy on two occasions, with or without day-and -night closed loop, in free-living home settings over 12 weeks. We retrospectively analyzed overnight (23:00 to 07:00) insulin delivery during the closed-loop period. The BR for each closed-loop night were expressed as the percentage of the total amount of basal insulin administered on that night over what was preprogrammed on the insulin pump and delivered during the open-loop period. Coefficient of variation of BR (CV_BR) was calculated to represent individual variability of overnight BR. Correlations were sought between subjects’ baseline characteristics (gender, age, BMI, HbA_1c, duration of diabetes, duration of insulin pump use) and CV_BR.

Result:

We analyzed data from 2,065 closed -loop nights. The CV_BR was 31% (22, 39%; n = 32, mean [95% confidence interval]) with the lowest variability CV_BR of 24% (with the corresponding BR of 137% [73, 201%]) and the highest CV_BR of 40% (BR 124% [28, 221%]). Significant correlation was found between gender and CV_BR, which suggests a larger variability in BR in male subjects (33 vs. 28%, male vs. female; P = 0.014). No other significant correlations were observed.

Conclusion:

Overnight insulin requirements vary considerably between nights in adults with T1D during long-term use of closed loop insulin delivery. Male subjects appear to have more variable overnight BR than females.

Predicting Microvascular Complications from Type 2 Diabetes Retrospective Data

First Three-Dimensional Structure of Fungus-Derived Glucose Dehydrogenase: Knowing Enzyme for Glucose Monitoring

Progress toward Passive Transdermal Glucose Sensing Based on a Glucose-Binding Protein

Cristina Tiangco, MS; Abhay Andar, PhD; Xudong Ge, PhD; Fortunato Sevilla III, PhD; Govind Rao, PhD; Audra Stinchcomb, PhD; Leah Tolosa, PhD

Center for Advanced Sensor Technology

University of Maryland, Baltimore County

Baltimore, Maryland

tina11@umbc.edu

Objective:

The ability to measure the levels of glucose in premature neonates is a critical part of good clinical care. Commercially available glucose sensing methods require breaking the skin and collecting blood for testing. On the other hand, there are no Food and Drug Administration–approved noninvasive glucose testing methods in the US market today. The purpose of this work is to study the possibility of passively collecting transdermal glucose and then measuring the extracted glucose with a fluorescent glucose biosensor. We foresee the use of this noninvasive technology in the neonatal intensive care unit.

Method:

Skin samples from Yucatan female pigs (9 months old) were used for the skin diffusion experiments. The dermatomed skin ranging from 200–350 µ m was placed in a PermeGear in-line flow-through diffusion cell (maintained at 37 °C). Phosphate-buffered saline (PBS) and glucose (2, 5, and 20 mmol/L to mimic hypo -, normo-, and hyperglycemic conditions, respectively) were allowed to flow in a bottom chamber at a constant rate of 0.8 mL/min. Skin is mounted in the diffusion cell apparatus with the outer surface of the stratum corneum facing the sample chamber. The chamber was loaded with 250 µL of PBS. Glucose in the flow solution was allowed to diffuse to the PBS for 15 min. The extracted glucose was then assayed with the fluorescent glucose- binding protein (GBP). The fluorescence intensities were measured using a 96-well plate on a SpectraMax M5 plate reader.

Result:

The fluorescence intensity of GBP labeled with the fluorophore badan increases with increasing glucose concentration. The linear range for the determination of glucose is up to 20 µmol/L. The concentration of glucose in the transdermal samples range from 1.30 ± 0.05 to 4.0 ± 0.10 µmol/L. As expected, the cumulative amount of glucose increases with the concentration of glucose in the flow- through. This shows that glucose passively diffused from the skin can be measured using the GBP over the hypo- to hyperglycemic range.

Conclusion:

Results show that the flow-through diffusion system is a promising tool to develop the noninvasive glucose sensing system. Future experiments will be dedicated to the optimization of the transdermal glucose sensing system and modeling.

Insulin Sensitivity Profile as a Marker for Reduced Outcome in the Neonatal Intensive Care Unit

Vincent Uyttendaele, MS; Jennifer L Dickson, PhD; Adrienne Lynn, MD; J. G. Chase, PhD; T. Desaive, PhD

University of Liège

Liège, Belgium

v_uyttendaele@hotmail.com

Objective:

Hyperglycemia in neonatal intensive care units is associated with mortality and morbidity. This trial aims to use machine learning methods to provide a prediction of outcomes in hyperglycemic neonates based on model-based metabolic (glycemic control) data as a noninvasive marker.

Method:

Glycemic control data from 44 patients (4,499 h) under the STAR-NICU or STAR-GRYPHON model-based glycemic controllers from Christchurch Women’s Hospital were used. Predictive models were built using attributes from hourly, patient-specific, model-based insulin sensitivity. Among these patients, 12 contracted sepsis, 8 suffered from intraventricular hemorrhage (IVH), and 8 died. The methods used were classification trees and K-nearest neighbors. The efficacy of the models was assessed evaluating sensitivity, specificity, and accuracy.

Result:

Mean insulin sensitivity was different among different subgroups: 7.51 × 10⁻⁴, 5.47 × 10⁻⁴, 2.42 × 10⁻⁴, and 7.50 × 10⁻⁴ L/mU/min for patients who were septic, had IVH >grade 1, nonsurvivors, and survivors, respectively. Variability assessed as interquartile range was also different between groups, with 1.00 × 10⁻⁴, 4.99 × 10⁻⁵, 4.22 × 10⁻⁵, and 9.09 × 10⁻⁵ L/mU/min, respectively. It was possible to predict mortality with 85% sensitivity after the first 15 h and (later proven) sepsis with sensitivity of 80% within 20 h.

Conclusion:

A clinically validated model-based insulin sensitivity measure and its variability may provide information about patient condition and possible outcome, despite modeling limitations. This study emphasized the potential of machine learning to provide information on degrading patient condition and worsened outcome as an alert to provide more intensive care.

Combination of Microdialysis and Ex Vivo Glucose Sensing by Infrared Spectrometry with Online Recovery Rate Determination

Thorsten Vahlsing, Dipl-Ing; Sven Delbeck, BSc; Janpeter Budde, BSc; Lars Cocchieri, BSc; Dieter Ihrig, PhD; H. Michael Heise, PhD

Interdisciplinary Center for Life Sciences

South-Westphalia University of Applied Sciences

Iserlohn, Germany

vahlsing.thorsten@fh-swf.de

Objective:

Blood glucose monitoring has been realized by biosensors in combination with microdialysis, using either intravascularly or subcutaneously implanted catheters. However, a drawback is variable recovery rates, which can be observed especially for the latter devices. For continuous sensing, infrared spectrometry has been suggested, since besides glucose, other clinically relevant analytes can be simultaneously determined that are important, e.g., for intensive care patients.

Method:

Perfusates with either acetate or mannitol have been investigated as recovery markers. The latter substance is suggested for application with tunable quantum cascade lasers, rendering only a limited wavenumber interval when compared with Fourier transform infrared spectrometers. Despite the overlap of mannitol and glucose spectra, their simultaneous accurate quantification was successful.

Result:

In contrast to the previously used acetate, an almost linear dependency of mannitol loss and glucose recovery was realized for all microdialysis catheters tested, which provides a straightforward correction of any dialysis recovery rate variation during patient monitoring.

Conclusion:

The combination of microdialysis with infrared spectrometry provides a calibration- free assay for accurate continuous glucose monitoring, as reference spectra of dialysate components can be a priori allocated.

Telehealth System seems to improve HbA_1c in diabetic and prediabetic patients. We plan to develop feedback functions for the system.

Accuracy of a Blood Glucose Monitoring System as It Relates to the ISO 15197:2013 Requirements in the Management of Diabetes Mellitus

Julie Walker, RN, BSN, PHN; Tricia Gill, BA, MLT; John Gleisner, BS, PhD

ARKRAY USA Inc.

Minneapolis, Minnesota

walkerj@arkrayusa.com

Objective:

The purpose of this study was to demonstrate that GLUCOCARD Shine meets the ISO 15197:2013 blood glucose monitoring system (BGMS) accuracy performance requirements.

Method:

Three lots of GLUCOCARD Shine test strips were analyzed for performance and bias comparison (n = 104 data points). Samples were collected from the fingertip of confirmed diabetic subjects by trained personnel at the ARKRAY Factory Inc. in Minneapolis, MN. Reference values were obtained using the YSI 2300 analyzer. Data were analyzed using the system accuracy performance criteria published in the ISO 15197:2013.

Result:

Results revealed that 100% of <100 mg/dL samples (n = 6/6) were within ±15 mg/dL, meeting the 95% accuracy criteria, and 99.0% of the ≥100 mg/dL samples ( n = 97/98) fell within the predetermined 15% meeting the 95% performance criteria. All data were within the A and B zones of the consensus error grid. Overall bias was −1.3%, demonstrating strong agreement between the GLUCOCARD Shine and YSI reference analyzer results. Correlation coefficient (r) = 0.99 demonstrated a strong linear relationship between the YSI reference method and BGMS results.

Conclusion:

Data acquired on the GLUCOCARD Shine meet the ISO 15197:2013 system accuracy performance criteria.

Evaluation of a Blood Glucose Monitoring System in Comparison with Global Accuracy Performance Requirements

Julie Walker, RN, BSN, PHN; Tricia Gill, BA, MLT; John Gleisner, BS, PhD

ARKRAY USA Inc.

Minneapolis, Minnesota

walkerj@arkrayusa.com

Objective:

The purpose of this study was to demonstrate that the GLUCOCARD Vital meets the global blood glucose monitoring system (BGMS) accuracy performance requirements of 95% of results within ±15 mg/dL for glucose samples <100 mg/dL and 95% of results within 15% for glucose samples ≥100 mg/dL.

Method:

Three lots of GLUCOCARD Vital test strips were analyzed for performance and bias comparison (n = 104 data points). Samples were collected from the fingertip of confirmed diabetic subjects by trained personnel at the ARKRAY Factory Inc. in Minneapolis, MN. Reference values were obtained using the YSI 2300 analyzer. Data were analyzed using the accuracy performance criteria of 95% of results within ±15 mg/dL for glucose samples <100 mg/dL and 95% of results within 15% for glucose samples ≥100 mg/dL.

Result:

Results revealed that 100% of <100 mg/dL samples (n = 6/6) were within ±15 mg/dL, meeting the 95% accuracy criteria, and 99.0% of the ≥100mg/dL samples (n = 97/98) fell within the predetermined 15%, meeting the 95% performance criteria. All data were within the A and B zones of the consensus error grid. Overall bias was 0.5%, demonstrating strong agreement between the GLUCOCARD Vital and YSI reference analyzer results. Correlation coefficient (r) = 0.98 demonstrated a strong linear relationship between the YSI reference method and BGMS results.

Conclusion:

Data collected and analyzed on the GLUCOCARD Vital meet the global accuracy performance criteria of 95% of results within ±15 mg/dL for glucose samples <100 mg/dL and 95% of results within 15% for glucose samples ≥100 mg/dL.

Autonomous Shape-Memory-Alloy-Based Actuator Reaching Subcutaneous Blood Capillaries: Results from In Vitro Characterization and Two Pilot Human Studies

Gang Wang, BEng; Michael Poscente, BSci; Simon Park, PhD; Orly Yadid-Pecht, DSci; Martin Mintchev, PhD

University of Calgary

Calgary, Alberta, Canada

gawang@ucalgary.ca

Objective:

Despite numerous claims for less invasive solutions, reliable glucose monitoring is still based on blood samples drawn by finger pricking, which can be tedious and painful and affect patients’ adherence to disease management. A novel shape-memory alloy (SMA)-based actuator has been developed to provide a reliable, autonomous, wearable solution for skin lancing, blood sample collection, and in situ blood glucose analysis. This study offers in vitro characterization and in vivo human testing of this realistic approach for glucose maintenance.

Method:

Eight SMA- based actuator prototypes were successfully fabricated using compression molding. Two of them were prepared on a mechanical test station to characterize the output force and penetration depth, which are the major factors facilitating skin penetration. The other six prototypes were integrated on watch straps that were worn on the forearms of two volunteers to examine the blood extraction capability.

Result:

In vitro tests demonstrated that the actuator can produce a maximum penetration force of 160 gf and a maximum depth of 2.45 mm, which both exceeded the minimum requirements for blood-capillary-reaching skin penetration in humans. In subsequent human studies, visual observation of the testing site was performed after the actuation. In three tests, a blood drop autonomously emerged at the skin surface within 1 min. In the other three tests, a blood drop was formed by mildly squeezing the skin around the testing site.

Conclusion:

The designed SMA actuator successfully reached capillaries in all six human tests. However, the width and the depth of skin penetration should be expanded in order to reliably extract a sufficient amount of whole blood for further analysis. The method has the potential to become a core technology for reliable autonomous blood glucose monitoring.

Development of a Primary Isotope Dilution Mass Spectrometry Traceable Reference Procedure in China for Blood Glucose Testing Is a Key Step for Evaluating Point-of-Care Glucose Methodologies

Yufei Wang, MD; Junling Tang; Weixiang Wang; Lu Dong; Chengchen Gu; Weiping Jia, MD; Xiaofang Hu; Wei Qiu

Shanghai Clinical Center for Diabetes, Shanghai Clinical Center for Metabolic Diseases, Shanghai Key Laboratory of Diabetes Mellitus, Shanghai Diabetes Institute, Department of Endocrinology and Metabolism, Shanghai Jiao Tong University–Affiliated Sixth People’s Hospital

Shanghai, China

wang1fei@126.com

Background:

There is considerable discussion about the accuracy of hospital blood glucose monitoring systems (BGMS) and self-monitoring of blood glucose (SMBG) meters. Studies undertaken with the same point -of-care (POC) glucose meters have shown different results, and this is often related to the different comparative methods used. With new Food and Drug Administration, International Organization for Standardization, and Clinical and Laboratory Standards Institute guidelines emerging, it is important that evaluations are standardized to a true and traceable definitive reference method. We advocate the use of an isotope dilution liquid-chromatography mass spectrometry (ID-LCMS) aligned hexokinase (HK) method on a central laboratory analyzer (Roche Modular) for undertaking evaluations combined with pretreatment of samples with perchloric acid (PCA) to hemolyze and deproteinate specimens, and we describe our validation method for achieving this traceability.

Methods:

The definitive ID-LCMS reference method was calibrated using a bracketing technique using varying concentrations of glucose in water (primary standards, National Institute of Standards and Technology [NIST] 917a). NIST 965b secondary protein-based standards were then evaluated to demonstrate the performance of the ID-LCMS definitive method and its agreement with the NIST -assigned target values as compared with the primary standards. The alignment of the laboratory HK and PCA (PCA-HK) methods to the ID-LCMS method was undertaken with NIST primary (aqueous) and secondary (serum albumin based) standards and plasma samples. The correlation of the ID-LCMS aligned PCA-HK to the routine HK method was further assessed with patient whole blood specimens and subsequent plasma samples.

Results:

Excellent linearity was obtained and correlation was obtained between all three methods: plasma HK, PCA -HK, and ID-LCMS. For example the R² for NIST 917a and 965b on the HK laboratory reference analyzer method were 0.999 with 95% confidence level 1.015 to 1.048, standard error = 7.7143 E-03, and a P value < 0.0001. PCA-HK and ID-LCMS demonstrated comparable results.

Conclusion:

In China, SMBG meters are routinely used in hospitals, and POC with BGMS is now beginning to be used more frequently. Consequently, the missing element for proper assessment of these devices to glucose performance standards is defining an acceptable primary reference method. In our facility, we have chosen a PCA-HK method that is closely aligned to ID-LCMS method for plasma glucose measurements, and in turn, the routine plasma HK method is closely aligned to the PCA-HK method. This type of method traceability approach is essential in evaluations of POC glucose methods.

Comparison of Point-of-Care Blood Glucose Monitoring System Performance Isotope Dilution Mass Spectrometry Aligned Reference Methods: An Extensive Evaluation

Yufei Wang, MD; Junling Tang; Weixiang Wang; Lu Dong; Chengchen Gu; Weiping Jia, MD; Xiaofang Hu; Wei Qiu

Shanghai Clinical Center for Diabetes, Shanghai Clinical Center for Metabolic Diseases, Shanghai Key Laboratory of Diabetes Mellitus, Shanghai Diabetes Institute, Department of Endocrinology and Metabolism, Shanghai Jiao Tong University–Affiliated Sixth People’s Hospital

Shanghai, China

wang1fei@126.com

Background:

Hospital glycemic management protocols help reduce hyperglycemia and subsequent mortality in critically ill patients. Handheld blood glucose monitoring systems (BGMSs) are ideal for guiding intensive insulin therapy and maintaining safe and effective glycemic control. However, while many of these BGMSs are used in hospitals, these have not been fully validated for use in critical care patient settings in China. Complicating matters, BGMSs are validated against various methods. Our approach was to evaluate BGMS devices with several comparative methods. This includes plasma glucose measurements via hospital chemistry analyzers serving as a comparative/reference method or hexokinase (HK)- based definitive reference method calibrated using National Institute of Standards and Technology isotope dilution mass spectrometry (IDMS) traceable standards and further verified with perchloric acid (PCA) deproteinated samples on both the HK laboratory reference laboratory method and isotope dilution liquid - chromatography mass spectrometry definitive method. The objective our study was to evaluate the performance of existing BGMSs against IDMS, PCA, and plasma glucose methods.

Methods:

One hundred remnant venous whole blood specimens were collected and tested on four BGMSs, two from manufacturer A (StatStrip, Nova Biomedical, Waltham, MA) and two from manufacturer B (AccuChek Inform II, Roche Diagnostics, Indianapolis, IN). BGMS performance was compared against IDMS (LC-MS/MS, Thermo Fisher TSQ Quantum Access Max, Waltham, MA), hospital laboratory reference analyzer HK method (Roche Diagnostics, HK generation 3, Mannheim, Germany), and the subsequently aligned PCA/HK (Roche Diagnostics, Modular P module, Mannheim, Germany). We used several statistical methods to evaluate the concordance of the methods evaluated.

Results:

Data on method precision, effect of interferences, and comparison of BGMS to central laboratory reference HK and PCA/HK aligned definitive methods too numerous to include in this abstract will be presented in the poster.

Conclusions:

BGMSs from both manufacturers showed differences when compared with the reference methods. These data illustrate differences among BGMSs and reference methods. To this end, there is a critical need in China to standardize performance measurements for BGMSs and hospital laboratory analyzers.

Glucosense: A Low-Cost Glucose Meter System for Resource-Poor Settings

Kayla Wilson, BS; Delphine Dean, BS, MEng, PhD

Clemson University

Clemson, South Carolina

kgainey@g.clemson.edu

Objective:

In low-resource settings that rely on donated medical supplies, glucose monitoring supplies are not always available to patients, and even when they are, standard supplies can be prohibitively expensive. Our objective is to design a low-cost glucose meter system that can be used in resource-poor settings.

Method:

Our system utilizes test strips made using a desktop inkjet printer. Empty printer cartridges are filled with enzyme and dye solutions. Precise amounts of the solutions are printed onto filter paper using a word processor template document. Glucose in the blood induces a color change on the printed strips. To read the strips, we designed a low-cost glucose meter, which determines the glucose concentration based on a preprogrammed standard curve. However, the strips can double as a visual indicator if a meter is not available.

Result:

Both pipetted and printed test strips have shown accuracy in processing glucose standards and bovine blood with added glucose in the range of 0–350 mg/dL. The design incorporates an internal calibration to account for varying patient blood color and oxygen concentration that may affect results.

Conclusion:

The research thus far confirms that a printable test strip is feasible as a means for monitoring glucose levels. While these strips are not as fast as current commercially available strips (readings take ~30 s), they meet the standard for accuracy in a glucose meter. The desktop printing of test strips is easy to implement in clinics in resource-poor settings.

Cytotoxic Effect of Three Commercial Rapid-Acting Insulin Analogs on Four Human Immune Cell Types

Wendy D. Woodley, BS, BA; Suvadra S. Gerth, BS; Didier R. Morel, BS, PhD; Kristen Borchert, MS; Lynn Godfrey, BA; Nicholas Herrmann, BS; Ronald J Pettis, BS, PhD; Natasha G Bolick, BS,

BD Technologies

Research Triangle Park, North Carolina

wendy_woodley@bd.com

Objective:

Local tissue inflammation from continuous subcutaneous insulin infusion is well documented, but causation and overall cellular effects are poorly characterized. The present study characterizes the in vitro cytotoxic effects of insulin analog formulations (lispro, glulisine, and aspart) on four human immune inflammatory cell types active the subcutaneous (SC) tissue: peripheral blood mononuclear cells (PBMC), polymorphonuclear leukocytes (PM human macrophages (THP-1), and human mast cells (HMC-1).

Method:

THP- 1 and HMC-1 cells were cultured in vitro; primary PBMC and PMN were freshly isolated from human blo Each cell type was incubated with serial dilutions of each insulin analog in culture media (0.26–33 U/mL concentration range). Cell viability was evaluated at 24 h via CyQUANT Direct Cell Proliferation Assay (Life Technologies).

Result:

Regardless of analog, insulin dilutions exhibited a dose-responsive direct cytotoxic effect on each cell type teste Cytotoxicity curves for individual cell types were nearly superimposable between insulin formulations, excepti HMC-1 at 16.7 U/mL, where glulisine reduced viability >50% but aspart and lispro did not. PMN exhibited the overall cytotoxicity with viability of 70% (glulisine and lispro) to 90% (aspart) at 33 U/mL and recovery to 100 16.7 U/mL for all analogs. Fifty percent reduction in cell viability, EC₅₀, across the three insulin formulations occurred at 13 U/mL (lispro and glulisine) and 15 U/mL (aspart) in PBMC, 7–8 U/mL (all formulations) in THP and 28 U/mL (aspart and lispro) and 16 U/mL (glulisine) in HMC-1.

Conclusion:

All three insulin analogs demonstrated direct cytotoxicity against SC immunoreactive cells at exposures at or b expected therapeutic concentrations. Additional exploration of specific insulin formulation components and in confirmation is warranted to obtain a better mechanistic understanding and develop inflammation prevention methods.

Characterization of the Information Technology Utilization and Experiences of Diabetes Patients Insured Under the Affordable Care Act

Jason Xenakis, BS; Mary McBride, MBA; Susan Garfield, DrPH

GfK Market Access

Wayland, Massachusetts

jason.xenakis@gfk.com

Objective:

In 2014, the Affordable Care Act (ACA) was implemented extending insurance coverage to nine million previously uninsured patients. These patients represent a population with unique demographics, lifestyles, and socioeconomic status. In this study, we sought to characterize how differential access to technology and information sources correlate with the health outcomes of diabetes patients covered by the ACA compared with those with other insurance types.

Method:

Using dual data collection methods, involving telephone-based interviews and Internet-based questionnaires, a total of 2,282 diabetes patients in the United States were sampled. Questions focused on demographic, lifestyle, treatment, access to information, technology/device ownership, and socioeconomic status and were stratified by insurance coverage (ACA, Medicare, Medicaid, and private insurance).

Result:

ACA patients have high rate of Internet access at home (98.3%); however, when asked about utilized information sources, ACA patients most frequently listed no regularly used source of diabetes education, whereas Medicare and private insurance patients listed the Internet as their most common source. ACA patients showed poor control over their diabetes, with 20.9% claiming “little or no control” (Medicare 6.56%, Medicaid 6.49%, private 5.67%). Additionally, ACA patients are not adhering to personal monitoring, with 42.61% complying with the frequency of blood glucose testing recommended by their physician (Medicare 73.93%, Medicaid 65.51%, private 63.37%)

Conclusion:

Our findings show that despite having access to Internet-capable devices, ACA patients insured are not using the available education sources that allow them to properly self-manage their diabetes. Additionally, these patients are not taking the required precautions to make sure their diabetes is properly controlled. It is necessary that programs targeting these patients understand these behaviors in order to develop future intervention strategies.

Human Kinetic Approach to Improving Conformability in PaQ, a 3-day Wearable Insulin Delivery Device

Michelle M. Yablonski, BS, MS; Joseph H. Arsenault; Kenneth J. Barnett, BS, MS; Craig L. Brodeur, BSME; Anne M. Cropp, BS; Jessica S. Shelsky, BS, MS

CeQur Corporation

Marlborough, Massachusetts

michelle.yablonski@cequr.net

Objective:

Adhesive tape conformability can depend on tape-to- device fixation and material stretch properties for body-worn devices such as PaQ (CeQur SA), an insulin delivery device adhered to the abdomen for 3 days, providing continuous subcutaneous insulin infusion. The purpose of this work was to identify an adhesive tape fixation and orientation that improves conformability, emphasizing freedom of movement and security while optimizing adherence around the cannula to maintain insulin delivery.

Method:

Ten people of varying height, weight, and gender were recruited to evaluate the natural stresses imposed on the abdomen during physical activity. Four dots were drawn in a rectangular pattern on the abdomen at a specified distance from the navel. Participants then performed a series of rotational and elongating movements in which dot displacement was measured. Based on these results, a test system was designed to assess conformability by mimicking the shear and stretch experienced by the adhesive tape during normal daily activities. PaQ devices were tested in a crossed design consisting of four fixation methods and two stretch orientations for 1 h, approximating 1,000 cycles of movement. Conformability was assessed by quantifying the nonadhered surface area.

Result:

An optimal combination of fixation method and orientation was determined and compared with a control. The optimal test group exhibited an average adherence loss of only 8%, outperforming the control group, which had an average loss around 25% of total surface area.

Conclusion:

The data from the test system demonstrated that improved conformability can be achieved by optimizing the combination of adhesive tape fixation method and material stretch orientation. Additional studies will be conducted to assess these improvements on human subjects.

Simulation Study on Insulin Sensitivity Factor Changes for Patients with Illness/Stress

Jin Yan, PhD; Dima Sokolovskyy, BS; Salman Monirabbasi, PhD

Medtronic Inc.

Northridge, California

jin.yan@medtronic.com

Objective:

The hybrid closed-loop (HCL) system is an integrated system that calculates the insulin dose based on continuous glucose monitor sensor data to achieve glycemic control throughout the day. The HCL system simulation was composed of the HCL algorithm, the NGP pump model, physiology model, blood glucose meter model, and a glucose sensor. The closed-loop algorithm used proportional-integral-derivative controller with insulin feedback, where this algorithm is designed with additional safeguards that includes maximum insulin delivery safeguards and moving window optimized patient -specific parameters so as to mitigate the potential risks of closed-loop control throughout the day and night. Our goal is to investigate the system performance under the circumstances when the patients are either sick or under stress with preprogrammed carbohydrate ratios (user provided carbohydrate ratios).

Method:

We first simulated the reference case, where the gains of insulin sensitivity factor (ISF) are static. Next we simulated the experiment case, where the gains of ISF are dynamic. Our goal is to explore the patients’ glucose with various food boluses amounts. We consider the 30/70 case, where the ratios of basal delivery total and food boluses daily delivery total is 30:70, together with 50:50 ratio. In order to implement insulin sensitivity with alternate basal-bolus ratio, the system is required to feature two optimized patient parameters: total daily insulin dose (TDD) and carbohydrate ratio. We use preprogrammed carbohydrate ratios for all three meals for each patient. Next the meal profiles for various basal-bolus ratios can be set as follows. There are no unannounced meals. All the patients have three meals per day and the amounts of meals are identical. More specifically, for 30:70 ratio, breakfast occurred at 7:00 a.m. with an amount of 100 carbohydrates, lunch occurred at 12:00 p.m. with an amount of 120 carbohydrates, and supper occurred at 6:00 p.m. with an amount of 130 carbohydrates. For 50:50 ratio, breakfast occurred at 7:00 a.m. with an amount of 50 carbohydrates, lunch occurred at 12:00 p.m. with an amount of 100 carbohydrates, and supper occurred at 6:00 p.m. with an amount of 100 carbohydrates. SafeBasal = TDD/48 (2.1), SafeBasalLow = TDD/166 (2.2), U_max = TDD/33 (2.3), and carbohydrate ratio = 500/TDD (2.4).

Result:

Sensor glucose is high and insulin delivery is insufficient due to U_max limit when patients are sick or under stress.

Conclusion:

Insulin was delivered at the maximum allowable rate, and preprogrammed carbohydrate ratio prevented patients from hyperglycemia when they were recovered from stress/sickness. Future research will focus on discussing relaxing the maximum allowable insulin infusion rate constraints.

Accuracy of Select Wearable Technology for Varying Exercise Modalities and Intensities in Persons with and without Type 1 Diabetes

Loren Yavelberg; Dessi Zaharieva; Robert Gumieniak; Ali Cinar, PhD; Michael Riddell, PhD; Veronica Jamnik, PhD

York University Kinesiology and Health Science

Toronto, Ontario, Canada

loreny@yorku.ca

Objective:

We examine the accuracy of two wearable devices that will be used in future studies involving the artificial pancreas. The BioHarness by Zephyr and the Metria Armband by Vancive were used during varying exercise intensity ranges plus modalities in persons with and without type 1 diabetes. The variables of interest included heart rate (HR), breathing frequency (BF), and caloric expenditure (EE).

Method:

The subjects (24.4 ± 6.1 years, four males and three females) were outfitted with a Metria, BioHarness, and Polar HR monitor. The subjects completed an incremental to maximal oxygen consumption (VO₂max) protocol twice and were randomly assigned to complete 40 min of continuous aerobic exercise and callisthenic-based circuit, once or twice. Oxygen consumption and BF were measured throughout each exercise session using the indirect calorimetry unit (Cosmed Fitmate or discrete system).

Result:

The results of the incremental to VO₂max protocol were categorized into the following exercise intensity ranges: light-to-moderate, moderate-to -vigorous, and vigorous-to-maximum. Paired t test and Bland-Altman analyses were performed. Within all exercise intensities, there were no differences in HR between the BioHarness and Polar monitor ( P = 0.713, 0.347, 0.785, respectively) . Significant differences in BF were observed in both moderate-to-vigorous (P = 0.01) and vigorous-to- maximum ( P = 0.000) exercise intensity ranges between the BioHarness and indirect calorimetry unit. Significant differences were also observed in the derived EE between the indirect calorimetry unit and the Metria during the vigorous-to-maximum exercise intensity range (P = 0.000).

Conclusion:

The preliminary findings indicate that the 1) Bioharness HR is accurate at all exercise intensity ranges, 2) Bioharness BF is only valid during light-to -moderate exercise intensities, and 3) Metria provides an acceptable estimation of EE for light-to-moderate and moderate-to-vigorous but not vigorous-to-maximum exercise intensities.

Pilot Study of Oral Insulin Enteric-Coated Soft Capsules Research and Development in Chinese

Weigang Zhao, MD; Tao Yuan, MD; Qi Sun; Shenyuan Yuan; Zhimin Liu; Guoliang Liu; Li Chen; Shaomei Han; Tao Xu, PhD

Department of Endocrinology

Key Laboratory of Endocrinology

Ministry of Health

Peking Union Medical College Hospital

Chinese Academy of Medical Sciences

Beijing, China

xiehezhaoweigang@163.com

Objective:

We study the pharmacodynamics and pharmacokinetics of oral insulin enteric-coated soft capsules and then evaluate the clinical efficacy and safety of capsules in type 2 diabetes patients.

Method:

Twenty healthy volunteers received insulin enteric-coated soft capsules 50 IU or regular insulin 15 IU after a baseline period of 2 h, with hyperinsulinemia euglycemic clamp technique. Glucose infusion rates (GIRs) were determined to calculate pharmacodynamics and pharmacokinetic parameters. Two hundred sixty type 2 diabetes patients were enrolled in a multicenter, randomized, open, parallel-controlled clinical trial. Patients were orally administered with enteric-coated soft capsules (n = 135) or subcutaneously injected neutral protamine Hagedorn insulin ( n = 125) 1 h before breakfast and dinner for 12 weeks. Prior to the trial, at the end of weeks 6 and 12, patients were served a standard test meal. Fasting blood glucose, 2 h postprandial blood glucose, and glycosylated hemoglobin A_1c (HbA _1c) were tested. Before the end of week 8, patients were followed up every 2 weeks. The dose of trial drugs was titrated to reach the goal of glycemic control based on the blood glucose monitoring results. At the end of week 12, physical and biochemistry examinations were performed the same as before.

Result:

The maximal concentration (C_max) of insulin was 22.1 ± 8.0 and 118.6 ± 25.2 mIU/L (oral versus regular insulin); the time for reaching C_max (T_max) was 225.8 ± 142.2 vs. 115.5 ± 43.4 min. The maximal GIR (GIR_max) was 3.56 ± 0.85 vs. 4.87 ± 1.26 mg·kg⁻¹·min⁻¹; the time for reaching GIR_max (TGIR_max) was 166.3 ± 75.9 vs. 148.0 ± 40.8 min. The relative bioavailability and bioefficacy of insulin capsules were 7.42 ± 3.25 and 24.78 ± 0.08%. HbA_1c were significantly decreased in both groups (P < 0.05; in control group, from 8.42 ± 1.53 to 7.23 ± 1.17%; in capsule group, from 8.39 ± 1.34 to 7.42 ± 1.00%), and the decreasing amplitude of HbA _1c in capsule group (0.98 ± 0.94) was remarkably smaller than control group (1. 19 ± 1.03%; P < 0.05). Decreasing of HbA_1c ≥0.5% was considered clinically effective. The effective rate in capsule and control group was 72.22 and 77.87%, respectively (P = 0.305). In terms of the percentage of patients achieving HbA_1c target recommended by American Diabetes Association (≤7.0%) was 38.9 and 45.1% in capsule and control groups, respectively (P = 0.323). The incidence of adverse effect was not different in two groups (P = 0.618). The satisfaction score was significantly higher in capsule group (P = 0.000).

Conclusion:

The action profile of the oral insulin enteric- coated soft capsules is similar to intermediate-acting insulin injection preparation. The difference between relative bioavailability and bioefficacy indicates the oral insulin capsules may mimic physiological procedure of insulin endosecretion. It has similar effectiveness and safety with insulin

Continuous Glucose Monitoring Accuracy Improvement Using the Autocalibration Feature of MiniMed Veo

Alex Zhong, MS; Pratik Agrawal, MS; John B. Welsh, MD, PhD; Michael Stone, MS; Ubi Van den Hombergh, MD; Beatrice Nurdin, MS, MPh, MBA; Brian Kannard, BS, MBA; Rajiv Shah, MS

Medtronic Diabetes

Northridge, California

alex.zhong@medtronic.com

Objectives:

The autocalibration feature of the MiniMed Veo system (Medtronic MiniMed Inc., Northridge, CA), allows for the immediate calibration of continuous glucose monitoring sensors based on blood glucose (BG) measurements. The purpose of this analysis is to evaluate the impact of autocalibration on sensor accuracy.

Methods:

CareLink personal data from 6,435 patients worldwide from December 1, 2013, to February 25, 2015, were analyzed. Sensor accuracy was compared between patients who enabled the autocalibration feature 100% of the tim (On) versus those who never enabled it (Off) using a Kruskal-Wallis nonparametric test. Sources of inaccuracy were further considered by comparing accuracy with calibration frequency and with the percentage of calibrations performed more than 2 min after a BG reading (delayed calibration).

Results:

Sensors of autocalibration On users (n = 276) had a mean absolute relative difference (MARD) value that was 1.52 percentage points less than sensors of autocalibration Off users (n = 5,567; P < 0.001). Sensor accuracy following delayed calibrations from autocalibration Off users had a 3 percentage point greater MARD as compared with timely calibrations (P < 0.001). The median calibration frequency for users with autocalibration On was 4.3/day compared with 3.4/day for autocalibration Off. Patients with any autocalibration use (20–80% of use) had a 0.85 percentage point MARD improvement as compared with patients with autocalibration Off (P = 0.03).

Conclusions:

The autocalibration feature, when enabled, significantly improves the sensor accuracy. Furthermore, the frequency of autocalibration use directly correlates to sensor accuracy. Accuracy improvement is primarily associated with immediate calibration and greater calibration frequency. The autocalibration feature allows for both benefits withou the burden of additional finger sticks.

Real-World Performance of a Predictive Low-Glucose Management Algorithm

Alex Zhong, MS; Kevin Velado, BS; Wilfredo Rodezno, MS; Michael Stone, MS; Pratik Agrawal, MS; Chantal McMahon, PhD; John B. Welsh, MD, PhD; Rajiv Shah, MS; Francine Kaufman, MD

Medtronic Diabetes

Northridge, California

alex.zhong@medtronic.com

Objectives:

The MiniMed 640G system includes the SmartGuard feature and the new -generation Enlite sensor. This system allows for the automatic suspension of insulin delivery using a predictive low-glucose management (PLGM) algorithm. We evaluated the real-world implications of routine use of SmartGuard by analyzing sensor data that were voluntarily uploaded to the CareLink personal database (Medtronic MiniMed Inc., Northridge, CA).

Methods:

The PLGM algorithm includes suspend before low, a feature that is triggered when the sensor glucose (SG) value is predicted to reach or fall below a preset low- glucose limit within 30 min. We evaluated biochemical hypo- and hyperglycemia at various thresholds for time intervals in which the feature was enabled versus intervals when it was not.

Results:

The feature was used by 597 people from January 13, 2015, to May 14, 2015 (a total of 12,012 patient-days of use), and resulted in 25,937 pump suspension events (2.2 per patient- day). With respect to hypoglycemia, SG values ≤50 mg/dL occurred 0.4% of the time when the feature was enabled versus 0.8% of the time when it was disabled. SG values ≤80 mg/dL occurred 5.3% of the time when the feature was enabled versus 7.0% of the time when it was disabled. With respect to hyperglycemia, SG values ≥180 mg/dL occurred 31.9% of the time when the feature was enabled versus 33.7% of the time when it was disabled. SG values ≥300 mg/dL occurred 3.4% of the time when the feature was enabled versus 3.5% of the time when it was disabled.

Conclusions:

Routine use of the PLGM algorithm and its suspend before low feature appears to lessen severe and moderate hypoglycemia with no detrimental effects on hyperglycemia.

Real-World Assessment of Undetected Glycemic Excursion Events in Type 2 Diabetes

Alex Zhong, MS; Wilfredo Rodezno, MS; Chantal McMahon, PhD; John B. Welsh, MD, PhD; Pratik Agrawal, MS; Brian Kannard, BS, MBA; Rajiv Shah, MS

Medtronic Diabetes

Northridge, California

alex.zhong@medtronic.com

Objectives:

This study used continuous glucose monitoring (CGM) data from patients with type 2 diabetes to quantify the frequency and severity of glycemic excursions otherwise missed by finger sticks. We also compared excursion frequency between therapy regimens to gauge the potential utility of CGM studies in different populations.

Methods:

This study included noninterventional professional CGM iPro2 data from the CareLink database (Medtronic MiniMed Inc., Northridge, CA). Hypoglycemia (sensor glucose [SG] ≤55 mg/dL) and hyperglycemia (SG ≥240 mg/dL) event frequency and duration were calculated. Patient recognition of an event was determined by a meter blood glucose (BG) value within ±15 min of a CGM-detected excursion.

Results:

Data from 99,102 patients and 367,080 patient-days from December 14, 2011, to May 27, 2015, were analyzed. A total of 70,024 hypoglycemic events lasting >20 min (69 events/patient-year) were detected, 74.4% of which were not accompanied by a BG reading. A total of 192,471 hyperglycemic events lasting >2 h (189 events/patient-year) were detected, 40% of which were not accompanied by a BG reading. Insulin users (either multiple daily injection or pump) had lower rates of hypoglycemia than patients treated with oral medications or lifestyle interventions (0.146 vs. 0.158 events/day) and lower rates of hyperglycemia (0.40/day) than either oral medication users (0.43 events/day) or those using lifestyle interventions (0.46 events/day).

Conclusions:

Over 70% of hypoglycemic and 40% of hyperglycemic events were missed by contemporaneous BG measurements in patients with type 2 diabetes. Notably, insulin users had lower rates of glycemic excursions than those treated with lifestyle modification therapy or oral medications. CGM studies provide important additional information for managing type 2 diabetes.

2015 Diabetes Technology Meeting Abstracts

Assessment of a Patient-Level Diabetes Education Elective for First-Year Pharmacy Students Utilizing Hands-On Diabetes and Coexisting Disease State Point-of-Care Screening Tools

Objective:

Methods:

Result:

Conclusion:

Continuous Glucose Monitoring Using Near-Infrared Fluorescent Carbon and Corona Phase Molecular Recognition Molecular Sites for Glucose

Objective:

Method:

Result:

Conclusion:

Ketone Monitoring Frequency and Methods in Type 1 Diabetes Subjects

Objectives:

Methods:

Results:

Conclusions:

eGlycemic Management System Provides Safe and Effective Glycemic Control for Stroke Patients Requiring Subcutaneous Insulin in the Hospital Setting

Objective:

Method:

Result:

Conclusion:

Optimization of Core Sensor Components for an Osmotic Pressure Sensor for Interstitial Glucose Assessment ANDR15123D

Background:

Methods:

Results:

Conclusions:

Accuracy and User Performance Evaluation of a New Blood Glucose Monitoring System in Development for Use with Contour Plus Test Strips

Objective:

Method:

Result:

Conclusion:

Development of a New Glucose Dehydrogenase Mutant That Is Free of Xylose Interference

Background:

Methods:

Results:

Conclusions:

ClampArt: Improved Algorithm for Automated Glucose Clamps

Objective:

Methods:

Results:

Conclusion:

Development of a Hypoglycemia Alarm Based on Electroencephalography

Objective:

Methods:

Results:

Conclusions:

Patient-Level Assessment of Cough with Technosphere Inhaled Insulin in Patients with Type 1 Diabetes Mellitus

Objective:

Method:

Results:

Conclusion:

Coordination of Insulin and Glucagon Delivery in Bihormonal Artificial Pancreas through Habituating Control

Objective:

Method:

Result:

Conclusion:

Calibration of the Dynamical Tracking eA1c Algorithm with Reference HbA1c Values: Performance in Type 1 and Type 2 Diabetes

Background:

Methods:

Results:

Conclusions:

Safety Analysis of a 90-Day Implantable Continuous Glucose Monitoring System in the PRECISE Study

Objective:

Method:

Result:

Conclusion:

Biologic Variation of Intensive Care Unit Patient Capillary Glucoses Monitored with Hematocrit-Independent Strips (2012–2015) Compared to That Monitored with Hematocrit-Dependent Whole Blood Glucose Strips (2009–2010)

Objective:

Method:

Results:

Conclusion:

Biodel’s Glucagon Emergency Management Autoreconstitution Device Demonstrates Superior Usability Compared to Marketed Glucagon Kits in Human Factors Study

Objective:

Method:

Result:

Conclusion:

Accuracy and User Performance Evaluation of a New Blood Glucose Monitoring System in Development for Use with Contour Next Test Strips

Objective:

Method:

Calibration of the Dynamical Tracking eA1c Algorithm with Reference HbA_1c Values: Performance in Type 1 and Type 2 Diabetes

Glooko: Mobile, Unified Platform for Diabetes Management, Combined with Nurse Coaching, Lowers HbA_1c and Improves Quality of Life