2016 Diabetes Technology Meeting Abstracts

Acciaroli	Good Accuracy of CGM-based Glucose Variability Indices for IGT and …	A1
Acciaroli	Reduced Calibrations and Maintained Accuracy on Next Generation CGM …	A2
Armbrecht	Mobile Text Messaging to Support Post-Discharge Transition Care and Reduce…	A3
Bailey	Insulin Glargine 300 U/mL (Gla-300) Provides More Stable and More Evenly …	A4
Benesch	A Different Design of Glucose Clamps: Fixed Rate of Blood Glucose Change…	A5
Bollyky	Nutrition Text Education Program Offered to Livongo Diabetes Patients …	A6
Bollyky	Use of Connected BG Meter and remote CDE Coaching Reduces eHbAlc by 0.8%…	A7
Carter	Software Modeling in Type 1 Diabetes - Real Patient Insights	A8
Cembrowski	Hemoglobin Alc Precision Profiles: Aid to Method Selection and Improvement	A9
Christiansen	User Performance Evaluation of the New Contour®Next ONE Blood Glucose …	A10
Corl	Mind the Glucose Gap: Validation of Point-of-Care Blood Glucose (POC BG)…	A11
Douglas	A New Tool for Monitoring in Acute Care Facilities and in Home Care Diabetes…	A12
DeJournett	Comparative Simulation Study of ICU Glucose Controllers	A13
Delbeck	Reliable Glucose Monitoring by ex-vivo Blood Micro-dialysis and Infrared…	A14
Erande	Monitor Intensely To Meet Intention	A15
Esenaliev	Noninvasive, Continuous Glucose Monitoring with Novel, High-resolution…	A16
Fabris	Prediction of Hypoglycemia from CGM Data from 1 to 3 Hours Ahead in Type 1…	A17
Felice	3rd Generation Biosensor Technology for Continuous Glucose Monitoring Systems	A18
Freckmann	Measurement Accuracy of Four Blood Glucose Monitoring Systems with Insulin…	A19
Freckmann	Post-marketing Evaluation of a Blood Glucose Monitoring System’s …	A20
Freckmann	System Accuracy of a Novel System for Blood Glucose Monitoring Linked to a…	A21
Gal	Non-Invasive Glucose Monitoring Device: Reducing Impact of Expected…	A22
Gautier	Defining Fasting Glucose from Continuous Glucose Monitoring in Patients with…	A23
Gill	Performance of the GLUCOCARD® Vital Blood Glucose Monitoring System…	A24
Greenwood	Exploring the Wisdom of the Diabetes Online Community: Themes from Social…	A25
Grosman	Virtual Patients with Circadian Rhythm for Type 1 Diabetes	A26
Harrison	Performance and Accuracy Capability of a New, Wireless-enabled Blood Glucose …	A27
Harrison	Performance and Accuracy Capability of a New, Wireless-enabled Blood Glucose …	A28
Heise	Insulin Degludec: Four-times Lower Pharmacodynamic Within-patient Variability …	A29
Herrero	A Novel Bi-hormonal Control Strategy Inspired on the Paracrine Interaction…	A30
Hompesch	Retrospective analysis of Impact on SMBG and Glycemic Control of Mobile…	A31
Ide	Results Support Efficacy and Clinical Efficiency of Diabetes Management…	A32
Inman	Continuous Glucose Monitoring in a Cystic Fibrosis patient with glucose…	A33
Kamecke	Dosing Accuracy of Different Insulin Pumps	A34
Kruger	Predictors of Glycemic Control in Adults with Type 2 Diabetes on Insulin	A35
Kudva	Unique Metabolomics profiles in adolescent Type 1 diabetes	A36
Kuhlenkoetter	Minimal Blood Loss with Automated Glucose Clamps: A new configuration of…	A37
La Belle	Rapid point of Care Insulin Sensor	A38
Lachal	Personalization of a Nonlinear Glucose-insulin system via a Markov Chain Monte…	A39
Lajara	Glycemic Efficacy and Insulin Requirements when Administering U-100 Regular…	A40
Lau	electronic Diabetes Artificial Pancreas Training (eDAPT) - Development and…	A41
Lee	Evaluation of Night Time Glucose and Insulin During a Pivotal Hybrid Closed…	A42
Liang	Blood Gas Analyzer Glucose Measurement Accuracy	A43
Lucisano	Update of Clinical Experience with a Long Term Fully-Implanted Continuous…	A44
Mader	PAQ® - 3 Month Observation Study in Adults with Type 2 Diabetes	A45
Maher	Performance Comparison of the GLUCOCARD® Shine and Accu-Chek® Aviva…	A46
Mauseth	Nocturnal Blood Glucose Control by Fuzzy Logic (FL) Dosing Algorithm	A47
McDonald	Diabetes Related Retinopathy Screening by Diabetes Educators: A Pilot Study	A48
Moriuchi	Realization of BGM Within ±10% Accuracy Based on Innovative Optical…	A49
Morrow	Accuracy and Utility of the GlucoScout Glucose Analyzer in Clinical Research	A50
Moscardo	Amperometric Glucose Sensors’ Background Current is a Confounding Factor…	A51
Mumpower	Relationship Between Glycemic Control Using eGMS and Readmission Rates in…	A52
Mumpower	Does Glycemic Control Using eGMS Reduce Readmission Rates for Hospitalized…	A53
Novak	In Silico Modeling of the Effects of Tissue Microenvironment on Subcutaneous …	A54
Offringa	Diabetes Management Application Improves Self-Care Behavior and Glycemic…	A55
Parikh	In-Silico Performance of 670G Hybrid Closed Loop (HCL) with Cumulative Error…	A56
Pesl	Analysis of Real-Time Data Transmission for Insulin Dosing Decision Support…	A57
Pfutzner	Evaluation of the Non-Invasive Glucose Monitoring Device GlucoTrack in…	A58
Pfutzner	Quantification of the Insulin Content in Insulin Pen Cartridges by Means of…	A59
Pleus	Accuracy of Blood Glucose Monitoring Systems: New Options for Graphical…	A60
Prastmark	Pen Needle Design Influences Ease of Insertion, Pain, and Skin Trauma in…	A61
Ramljak	In vivo Evaluation of an Osmotic Pressure-based Implantable Glucose Sensor…	A62
Rigla	Macrophage Accumulation Affects Sensor Accuracy in Humans	A63
Rittmeyer	Comparative Analysis of Handling Steps Necessary When Using Insulin Pumps	A64
Shomali	Mobile Health Interventions for Diabetes: From Taxonomy to Product	A65
Smith	A Review of Patient Factors Associated with Consistent CGM use in Patients…	A66
Steinberg	Pupillary Response Analysis as a means of Non-Invasive Glucose Monitoring	A67
Stewart	Can We Fix It? Yes We Can! Simplifying Nutrition in STAR Glycemic Control	A68
Stewart	Variability is a Constant! Insulin Sensitivity and its Variability in 4 ICU Cohorts	A69
Stormo	Stationary Lancing Surpasses Traditional Lancing in Both Pain and Complexity	A70
Tay	An Integrated Microdevice for Neutrophil Purification and Functional…	A71
Thomas	A Model of Endogenous Insulin Secretion During Exercise	A72
Tipnis	Composite Coatings for Long-term Prevention of Foreign Body Reaction During…	A73
Tsui	Hospital QC Rarely Out in Newer Whole Blood Glucose Meter: Opportunity for…	A74
Uyttendaele	Are Survivors Easier to Control? Why the Association of Glycemia and Mortality …	A75
Uyttendaele	When NICE is Not Nice: Performance of Two ICU Glycaemic Control Protocols	A76
Walker	Performance Comparison of the GLUCOCARD® Shine and Contour® Next…	A77
Walker	Performance of the Assure® Platinum Blood Glucose Monitoring System for…	A78
Wang, D…	Advanced Hydrogels for Implantable Electrochemical Glucose Sensors	A79
Wang, G…	In-Situ Blood Glucose Monitoring Using Autonomous Skin Lancing Device and …	A80
Wiltshire	CSII Catheter with Extended Lifetime and Rapid On-Off Pharmacodynamics	A81
Wu	Detection of Hypoglycemia that Requires Further Treatment in Addition to…	A82
Wuttke	Insulin Doses Prediction AI For Diabetes Type 1	A83
Wyman	Verifying Hospital Insulin Syringe Dose	A84
Yeisley	Qualitative Analysis of a Diabetes Alert Dog-Themed Blog: Pilot Study of…	A85
Zavitsanou	An Explicit and Verifiable Solution to Zone Model Predictive Control for…	A86
Zhang	Establishment of a Porcine Type I Diabetes Model to Evaluate Continuous …	A87
Zhong	Real World Assessment of MiniMed Connect	A88
Zijlstra	HbAlc Estimation from a Long-Term Continuous Glucose Monitoring System…	A89

Good Accuracy of CGM-based Glucose Variability Indices for IGT and T2D Classifications

Giada Acciaroli, MS; Alessandro Palombit, MS; Giorgio Maria Di Nunzio, PhD; Andrea Facchinetti, PhD; Giovanni Sparacino, PhD; Liisa Hakaste, MD, PhD; Tiinamaija Tuomi, MD, PhD; Rafael Gabriel, MD, PhD; Claudio Cobelli, PhD

Department of Information Engineering, University of Padova

Padova, Italy

giada.acciaroli@phd.unipd.it

Objective:

Many glucose variability (GV) indicators have been proposed in the literature, even more since the advent of continuous glucose monitoring (CGM) sensors. How to use the plethora of GV indicators is, however, to some extent, controversial, because several GV indices provide redundant information. In addition, the use of GV indicators to automatically recognize the metabolic status of a subject (e.g. when the subject has impaired glucose tolerance (IGT) or Type 2 diabetes (T2D) remains to be addressed. In the present work, we verify whether or not CGM-based GV metrics can be used for classification purposes in the quite simple task of distinguishing healthy subjects from subjects with either IGT or T2D.

Method:

The dataset consisted of 102 subjects, labelled into three classes through an oral glucose tolerance test (OGTT): 34 healthy, 39 IGT and 29 T2D subjects. A Guardian Real Time or the iPro CGM systems (Medtronic MiniMed, Inc., Northridge, CA) was used to produce a glucose trace from which 25 GV indices were extracted from each monitored subject. To classify each subject into a diagnostic group, we implemented a two-level binary logistic regression model. The first level distinguished healthy from unhealthy subjects and the second level classified the unhealthy subjects as either IGT or T2D.

Result:

The GV indicators distinguished healthy from unhealthy subjects with 90.48 ± 7.53% accuracy. The unhealthy subjects were subdivided into IGT or T2D with 79.18 ± 20% accuracy. The global classification into the three classes has an overall accuracy of 85.86 ± 12.79%.

Conclusion:

CGM-based GV indicators can be effectively used to accurately distinguish CGM traces of healthy and unhealthy patients. More critical, but still promising, is the subdivision of patients into those with IGT or T2D.

Reduced Calibrations and Maintained Accuracy on Next Generation CGM Compared to Dexcom G5: a Bayesian Approach

Giada Acciaroli, MS; Martina Vettoretti, MS; Andrea Facchinetti, PhD; Giovanni Sparacino, PhD; Claudio Cobelli, PhD

Department of Information Engineering, University of Padova

Padova, Italy

giada.acciaroli@phd.unipd.it

Objective:

The current generation of continuous glucose monitor (CGM) sensors need to be calibrated twice/day using self-monitored blood glucose (SMBG) samples. Recently, we demonstrated that an on-line Bayesian calibration algorithm could reduce calibrations to once-per-day without decreased accuracy. To further reduce, or even eliminate, the need for in vivo SMBG calibrations in next-generation CGM sensors, we propose the use of an algorithm based on a global calibration model that is defined over the entire monitoring period and a Bayesian strategy to estimate unknown parameters.

Method:

The dataset consisted of 55 diabetic subjects monitored for 10 days by a next-generation Dexcom CGM sensor prototype. The proposed calibration algorithm was evaluated, retrospectively, by simulating an on-line procedure using progressively fewer SMBG samples from one/day, one/two days, one/four days, etc. decreasing to zero blood glucose calibrations per day. Accuracy of the calibrated glycemic profiles was assessed against blood glucose references using mean absolute relative difference (MARD). For comparison, the same methodology was applied to 108 raw sensor traces using the current Dexcom G5 continuous sensor.

Result:

The once-per-day calibration scenario has accuracy of 9.29% MARD (compared to an 11.95% MARD for the Dexcom G5 sensor). The accuracy decreases slightly to 9.84% MARD (vs. 12.3% MARD for G5) and 10.18% MARD (vs. 12.46% MARD on G5) for calibrations once every two days or every four days, respectively. The zero-calibration scenario results in a 10.28% MARD (vs. 13.16% MARD on G5).

Conclusion:

Retrospective data analysis of the Dexcom G5 and the next-generation Dexcom CGM sensors show the potential of using multiple-day calibration and zero-calibration scenarios. Next generation CGM sensors appear better suited for this advancement than current sensors. In fact, sensor accuracy using our algorithm and only three calibrations over a period of 10 days is equal to the accuracy reported in the literature for the Dexcom G5 sensors calibrated twice/day.

Mobile Text Messaging to Support Post-Discharge Transition Care and Reduce 30-Day Hospital Readmission

Eric S. Armbrecht, PhD

Saint Louis University Center for Health Outcomes Research, School of Medicine

St. Louis, MO, United States

armbrees@slu.edu

Objective:

The objective of this evaluation is to measure the impact of a two-way mobile text messaging service that supports patients with chronic disease to stay healthy and avoid returning to the hospital.

Method:

A mobile health technology company (SanusEO) and 478-bed community hospital collaborated to provide a text message transition care support program for patients with diabetes discharged from the hospital for about four months. Tailored text messages were sent to enrolled participants for 30 days after discharge. The messages were designed to support selfmanagement behaviors in five key areas: medication adherence, vital signs monitoring (e.g., blood glucose), diet, physical activity, and communication with healthcare providers. The number, timing and content of messages depended on the profile and responses by participants to multiple-choice questions asked via text message.

Result:

After excluding 35 opt-out patients, 93% of the retained 148 patients had Type 2 diabetes. Mean age was 54.5 years, 55% were female, and mean BMI was 35.8 kg/m². The average number of co-morbidities per participant was 1.5 per patient. Six of 148 patients were readmitted within 30 days of discharge, yielding a readmission rate of 4.1%; which is 57% lower than the historical benchmark (10.3%) and statistically significant (p = 0.018). The calculated benchmark of 10.3% was based on 522 discharges of adults, ages 19-64 years, for the 4 months prior to the intervention. The historical benchmark excluded all subjects with a diagnosis of psychosis or with any procedure for hemodialysis because patients with these conditions would not have been invited to enroll in the program. Sixty-six percent of 148 patients responded to at least 15 of the questions sent to them, 19% of 148 patients had only a passive interaction (i.e., receiving messages but not replying), and 39% were super-participants (i.e., providing a response to 70% or more of the text messages requesting an answer).

Conclusion:

This program evaluation project was designed with adequate statistical power to measure reduction in readmission between text message support program participants and a historical benchmark based on similar patients. This observational approach to evaluation was employed to measure the impact of the program within the context of real-world implementation at a community hospital. Only 6 of the 148 patients (or 4.1%) who participated in this mobile textmessaging support program were readmitted within 30 days of discharge; this was 57% lower than the expected rate of 10.3%, which was determined by an analysis of the hospital’s historical data of a similar group of patients. The vast majority of patients (81%) interacted with the texting system by responding to at least one of the many multiple-choice questions asked during the 30-day period following discharge; 39% had a very high level of engagement, responding to 70% or more of the messages that requested an answer. Satisfaction was surveyed at the end and results indicated patients were very satisfied with the program — 93% of responding patients indicated the program was very helpful and 84% reported feeling more confident in the management of their diabetes.

Insulin Glargine 300 U/mL (Gla-300) Provides More Stable and More Evenly Distributed Steady-state PK/PD Profiles Compared with Insulin Degludec in Type 1 Diabetes

Timothy Bailey, MD; Raphael Dahmen, MD, PhD; Jeremy Pettus, MD; Ronan Roussel, MD; Karin Bergmann, DiplBiol; Magali Maroccia, MSc; Nassr Nassr, MD; Oliver Klein, MD; Geremia Bolli, MD; Tim Heise, MD

AMCR Institute

Escondido, CA, USA

tbailey@amcrinstitute.com

Objective:

To compare steady-state pharmacodynamic (PD) and pharmacokinetic (PK) profiles of insulin glargine 300U/mL (Gla- 300) with insulin degludec 100 U/mL (IDeg1) in two parallel cohorts with fixed once-daily dose regimens in Type 1 diabetes, in a multiple-dosing, crossover, euglycemic glucose clamp study.

Method:

For both insulins, participants received 0.4 U/kg/day (Cohort 1; n=24) or 0.6 U/kg/day (Cohort 2; n=24), before breakfast, for 8 days. Metabolic activity was then measured by glucose infusion rate (GIR) over 30 hours. Main endpoint: within-day variability (fluctuation) of smoothed GIR over the dosing interval (GIR-smFL0_-24). Insulin concentrations were measured using validated radioimmunoassays. GIR-smFL0_-24 treatment ratios were obtained using a linear mixed-effects model.

Result:

Within-day variability of smoothed GIR (GIR-smFL0_-24) was significantly lower with Gla-300 than IDeg1 at 0.4 U/kg/day (p=0.047; treatment ratio 0.7978 [90% CI: 0.6637 to 0.9591]) (LOESS smoothing 0.15), while it was comparable for Gla-300 and IDeg1 at 0.6 U/kg/day. Both doses of Gla-300 provided a plateau-like insulin exposure from 2 to 16 hours post-injection, with a slight decline afterwards, whereas both doses of IDeg1 caused total insulin to increase from ~1 hour to a T_max at ~10 hours after dosing, followed by a steady decline with no plateauing. Both insulins provided exposure and activity until 30 hours (end of clamp) and were generally well tolerated.

Conclusion:

This PK/PD analysis supports a superior glucodynamic profile of Gla-300 versus IDeg1 at a dose clinically relevant for Type 1 diabetes (0.4 U/kg/day) in terms of within-day variability. An overall more stable and more evenly distributed insulin exposure over the dosing interval was observed at both dose levels of Gla-300.

A Different Design of Glucose Clamps: Fixed Rate of Blood Glucose Change rather than Fixed Blood Glucose Levels

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

Profil

Neuss, North Rhine-Westphalia, Germany

carsten.benesch@profil.com

Objective:

Glucose clamps usually fix blood glucose concentrations (BG) to a defined target. However, for assessments of glucose- stimulated insulin secretion capacity or the accuracy and technical lag times of continuous glucose monitoring systems (CGMS), changing BG rather than fixed concentrations, are needed. Establishing precise rate-of-BG-changes (RoC) with graded glucose infusions is technically challenging. We therefore refined the software of ClampArt, Profil's proprietary modern clamp device, to implement automated clamps with precise RoC-settings.

Methods:

The ClampArt software allows investigators to pre-define a specific target glucose level as well as the rate-of-BG-change from that level over a given time period. The software was tested in-vitro with a 5 L container filled with glucose solution. For positive glucose slopes, ClampArt infused glucose into the container to raise glucose concentrations from 50 mg/dl by 0.5, 1.5, 2.5 and 3.5 mg/dl per minute to 400 mg/dl. For the simulation of falling glucose slopes, the effect of a short acting insulin was simulated through water infusion into the container and ClampArt glucose infusion to control the rate of glucose decline.

Result:

Precise rate-of-BG-changes were achieved. The deviation of glucose concentrations from various target levels was 0.3 ±1.0 mg/dl (mean ± SD) for positive slopes (N=2586 measurements) and 6.2 ±4.5 mg/dl for negative slopes (N=2934 measurements). No significant differences were observed between the different change rates. Time to reach target concentrations was close to the calculated time periods (mean deviation: 1.1 ± 2.3 and 4.9 ± 3.2 minutes for positive and negative slopes, respectively).

Conclusion:

The new RoC-software of ClampArt establishes precise and automatically controlled BG change rates. This software allows for investigations into the accuracy of CGMS at different rates of BG changes or the assessment of glucose-stimulated insulin secretion at high clamp quality.

Nutrition Text Education Program Offered to Livongo Diabetes Patients Increases Personalized Coaching Requests and BG checks

Jennifer Bollyky, MD; Jennifer Schneider, MD, MS; Anastasia Toles, MD, MPH; Jodi Pulizzi, RN, CDE; Michael Boulos

Livongo Health

Mountain View, CA, United States

jbollyky@livongo.com

Objective:

The Livongo for Diabetes program offers a cellular-enabled blood glucose monitoring system that measures blood glucose, captures contextual data, and stores these data in the cloud. Depending on the blood glucose value, personalized recommendations are delivered back to members through the glucose meter. We hypothesized that diabetes education offered to a targeted population would increase requests for coaching from a certified diabetes educator and encourage members to set and achieve American Association of Diabetes Educators (AADE7™) goals.

Method:

We examined a 15-week text message program offered to a subset of Livongo members with a calculated or self-reported HbA1c >7% to provide in-depth nutrition coaching. Nutrition topics included information on portion sizes, protein, calories, food labels, meal spacing, healthy fats and carbohydrates.

Result:

Two hundred fifteen members were invited to participate by text message, 89 members responded, 69 members opted into the program, and 59 members completed it. Participants were 42% male with mean age of 50 years, mean BMI of 31.2 kg/m², 88% with Type 2 diabetes, 61% on oral diabetes medications, and 22% using insulin. Text messaging content triggered 12 personalized telephone-coaching sessions with a certified diabetes educator (CDE) (i.e., 18% of enrolled members compared with baseline coaching rate of 4%). Average daily blood glucose checking frequency increased from 2.1 checks/day 30 days prior to program enrollment to 2.6 checks/day in the last 30 days of the program.

Conclusion:

Engaging people with diabetes through a cellular-enabled blood glucose meter with real-time, context-aware, actionable recommendations and a targeted and personalized text-message program, increases blood glucose checking frequency and CDE coaching requests.

Use of Connected BG Meter and remote CDE Coaching Reduces eHbAlc by 0.8% for up to 9 months

Jennifer Bollyky, MD; Jennifer Schneider, MD, MS

Livongo Health

Mountain View, CA, United States, 94041

jbollyky@livongo.com

Objective:

Self-insured employers, at-risk providers, and payers are motivated to empower people with diabetes to improve selfmanagement and decrease health costs. The Livongo for Diabetes Program leverages novel technology to offer: (1) a cellular enabled blood glucose meter with real-time personalized cloud-based analytics delivered to an individual through the meter, e-mail, or text, (2) unlimited glucose test strips, and (3) access to certified diabetes educator coaches. We hypothesized these offerings would improve blood glucose (BG) control for participants.

Methods:

People with a medical claim including a diagnosis of diabetes (ICD9 250.XX) were invited to join the program between October 2014 and March 2016 through their employers. In May 2016, mean BG values were used to estimate HbA1c using a published linear model (Nathan, Diabetes Care, 2008) for members with at least 90 BG values uploaded at Day 90 and compared with reported HbA1c values at registration. The estimated HbA1c (eHbA1c) was also calculated at Day 180 and Day 270 in a similar manner.

Results:

Seven hundred forty five members met study criteria. Participants were 51% male with mean age of 51.8 years, mean BMI of 32.9 kg/m², 70% with Type 2 diabetes, 71% on oral diabetes medications, and 42% using insulin. Mean reported HbA1c at registration was 7.7% (n=745, SD=1.8) compared to eHbA1c of 6.9% at Day 90 (n=745, SD=1.3, p<0.001). The eHbA1c at Day 180 was 7.0% (n=398, SD=1.2) and at Day 270,6.9% (n=217, SD=1.3) for participants with > 1 BG check per day. The likelihood of low BG <80 mg/dL did not increase for participants over this time.

Conclusion:

Participants in the Livongo program showed 0.8% eHbA1c reduction after 90 days without increased frequency of hypoglycemia. Improvements are sustained up to Day 270 thus far.

Software Modeling in Type 1 Diabetes – Real Patient Insights

Simon A. Carter, BEng, BS

ManageBGL Pty Ltd

Melbourne, Victoria, Australia

simon.carter@managebgl.com

Objective:

The objective is to investigate real patient data flows after patients provide medical data to the PredictBGL Diabetes App in non-clinical settings.

Method:

After patients downloaded, installed, and used the PredictBGL Diabetes App, the time-based data flows and software settings were examined.

Result:

The data show that patients are over-confident in their understanding of the software feature descriptions. Future applications should provide appropriate safeguards to protect users.

Conclusion:

Commercial software should be designed to hide advanced features and also better educate users after they become familiar with the software system.

Hemoglobin A1c Precision Profiles: Aid to Method Selection and Improvement

George S. Cembrowski MD, PhD; Yuelin Qiu; Trefor Higgins, MS

Department of Laboratory Medicine and Pathology, University of Alberta Hospital, Edmonton, Alberta, Canada

george.cembrowski@albertahealthservices.ca

Introduction:

When a series of quality control products are repeatedly analyzed over short or extended durations, the standard deviations (imprecision) of the quality control products are generally linear when graphed against the level of the analyte. An instrument’s targets for imprecision are not only important for quality control but also are very useful in instrument comparison and selection. Generally, the instrument with the lowest imprecision will be the most desirable system for result reporting. Between-analyzer imprecision provides an excellent indicator of instrument variation as it provides a measure of average system operation. Different instruments will display unique standard deviation (imprecision) profiles that can be used to simplify instrument acquisition. These precision profiles provide a measure of analytical imprecision over a range of mean analytical values.

Method:

The College of American Pathologist proficiency test (PT) summaries were obtained for two hemoglobin A1c (HbAlc) PT surveys: GH2 (2 cycles yearly [2013-2014] with 3 unknowns per cycle) and GH5 (3 cycles [2015] with 5 unknowns). When summary statistics were available for a minimum of 100 like analyzers, the analyzer model, the HbA1c group mean, and standard deviation were abstracted for each unknown PT sample. Standard deviation vs. mean (precision profile) graphs were generated for each analyzer model.

Results:

The precision profile graphs are largely linear with the high performance chromatography analyzers generating tighter HbA1c values compared to immunoassay.

Conclusions:

These precision profiles facilitate the acquisition of HbA1c analyzers with the best reproducibility. The availability of these profiles should motivate the manufacturers to relentlessly improve their HbA1c test performance.

User Performance Evaluation of the New Contour^®Next ONE Blood Glucose Monitoring System in Subjects With and Without Diabetes

Mark Christiansen, MD; Carmine Greene, MS, CCRA; Scott Pardo, PhD, PStat®; Robert Morin, MD; Timothy Bailey, MD, FACE, CPI

Diablo Clinical Research

Walnut Creek, California, United States

mchristiansen@Diabloclinical.com

Objective:

The objective is to evaluate the accuracy of the new Contour®Next ONE blood glucose monitoring system (BGMS) in clinical settings when used by subjects with or without diabetes. The BGMS utilizes Contour®Next test strips and features an easy-to-use, wireless-enabled blood glucose meter that links to a smart mobile device via Bluetooth® technology and can sync with the Contour™ Diabetes app on a smartphone or tablet.

Method:

This two-center clinical study enrolled 375 subjects with (n=332) or without (n=43) diabetes, who had never used this BGMS previously. Secondary objectives included accuracy per FDA Draft SMBG Guidance 2014 Section C for all subjects with or without diabetes (i.e., 95% of results within ±15% and 99% within ±20% of the laboratory method across the entire tested range).

Result:

Considering all subjects with or without diabetes, 99.5% (370/372) of subject fingertip self-test results were within ±15% and 99.7% (371/372) were within ±20% of YSI reference results. For study staff tests of subject fingertip blood, 99.7% (374/375) of results were within ±15% and 100% (375/375) were within ±20% of YSI reference results. At least 95% of BGMS results for all subjects were within 9.5% and 8.9% of the YSI reference result for subject and study staff-obtained fingertip tests, respectively. Regression analysis of results for all subjects demonstrated a strong correlation between BGMS and YSI reference results (adjusted R² >0.98 for subject and study staff-obtained fingertip results). By Parkes-Consensus Error Grid analysis, 100% (372/372) of all subject-obtained fingertip results and 100% (375/375) of study staff-obtained fingertip results were within Zone A.

Conclusion:

The BGMS exceeded FDA Draft 2014 criteria in a clinical setting when used by subjects with or without diabetes.

Mind the Glucose Gap: Validation of Point-of- Care Blood Glucose (POC BG) Measurement in Patients Admitted to Intensive Care Units (ICU)

Dawn Corl, RN, MN, CDE, CTDC; Lynn Graf, PhD; Lucy Greenfield, RN, MN, CCRN; Ronald Pergamit, MPA; Mark Wener, MD; Brent Wisse, MD

University of Washington (UW) Medicine/Harborview Medical Center

Seattle, WA, United States

corld@uw.edu

Objective:

Impaired glucose homeostasis is common in ICU patients as is the potential for POC BG inaccuracy. POC BG measurement in ICU patients remains common even though this practice is controversial. We developed and evaluated a program for timely identification of ICU patients with inaccurate POC BG testing.

Method:

A Quality Improvement (QI) project at a single center with ~90 ICU beds compared POC BG (Roche Inform II) measurements from capillary, arterial, or venous sample types to venous laboratory glucose values for patients admitted to an ICU service (+/-10% was considered accurate). An electronic medical record (EMR) tool identifying eligible patients and displaying their results was developed. Compliance was communicated to ICU nursing leadership. Frequency of testing and accuracy of data were assessed.

Result:

Over 2 months, 1,118 patients were hospitalized in the ICU; 283 (21%) met inclusion criteria for POC BG validation (1,155 patient days). A total of 1,329 validations were performed (69% compliance). Overall 82% of POC BG vs. lab BG comparisons were found to be accurate. However, 49% of patients had at least one 24-hour period when one or more POC BG samples were inaccurate. Overall mean values for accurate and inaccurate POC BG were not different (i.e., mean± SD 159.5 ± 48.7 vs. 153.2 ± 57.3 mg/dL, p = NS). For inaccurate POC BG values, the mean difference from lab BG values was 17.5 ± 18.3% (range 10.5-238%). Capillary POC values were inaccurate more often than arterial/venous POC values (N=529; 24.2 vs. 13.4%, p = 0.02).

Conclusion:

This project demonstrates a process to identify inaccurate POC BG values in ICU patients. Our data suggest that POC BG testing is inaccurate more frequently than expected in ICU patients. Future studies aim to identify ICU patient factors that predict POC BG inaccuracy.

A New Tool for Monitoring in Acute Care Facilities and in Home Care Diabetes to Address the Hospital Readmission Issue

Joel S. Douglas, MS, Michael Curtis, PhD, Keith D. Ignotz, MBA

Loon Medical, Inc.

Tolland, CT, United States

jdouglas@loonmedical.com

Background:

The lack of preventive and on-going care for patients with chronic conditions is a major cause of hospital readmissions. Senior patients with a chronic condition are more than 100 times more likely to experience a preventable hospitalization due to a lack of preventive care and monitoring for their chronic conditions .

Methods:

Meetings and interviews were held with caregiver organizations to determine their opinions regarding the cause of hospital readmissions.

Results:

It was found that patients and caregivers were using manual or standalone systems to collect data on their patients that lacked timeliness, consistency, and trending information.

Conclusion:

The manual collection of patient data and patient monitoring in chronic care appears to be a root cause of increased readmission rates. There is a need for new low cost, easy to implement, tools such as the Loon Medical CareCom Hub. CareCom can provide relatively frequent and accurate data about the patients’ blood glucose levels from a variety of the most widely used meters sent to the Cloud allowing real time analysis and intervention prior to an event. CareCom also addresses the co-morbidities of diabetes providing the collection of weight change over time (for congestive heart failure monitoring) giving the healthcare team the opportunity to intervene to prevent a hospital readmission. The technology addresses both the needs of Acute Care Facility Operators and Family Home Care. With the CareCom technology now available, studies should be conducted to validate the hypothesis that accurate real time technology can impact hospital readmission rates by facilitating early intervention.

Comparative Simulation Study of ICU Glucose Controllers

Jeremy DeJournett; Leon DeJournett, MD

Ideal Medical Technologies

Asheville, NC, United States

jeremy@idealmedtech.com

Objective:

The objective is to compare widely used ICU glucose controllers to a new knowledge-based method employed by Ideal Medical Technologies [IMT].

Methods:

The Yale paper protocol, a version of the Glucommander glucose control algorithm, and IMT’s artificial intelligence based controller were implemented in a new simulation suite in LabVIEW. Each controller was interfaced with the patient validated Van Herpe ICU Minimal Model and tested against 10 different clinically based exogenous dextrose infusions. All simulations had an initial glucose of 200 mg/dl and controlled to a range of 100 - 140 mg/dl. The model parameters were not altered in a time variant fashion as done in a previous publication by the authors.

Results:

The results, presented as Mean (SD), of the ten simulations were calculated for each controller. We applied standard metrics of time in control range (100 - 140 mg/dl), coefficient of variation, moderate hypoglycemia (40-69 mg/dl), and hyperglycemia (>140 mg/dl). The results for the Yale, Glucommander, and IMT simulations are presented in order: Time in Range 100 - 140mg/dl: 75.1(14.1), 76.3 (13.9), 92.8 (5.0); Coefficient of Variation: 16.0 (4.9), 14.6 (4.2), 10.2 (1.6); Moderate Hypoglycemia (40 - 69 mg/dl): 0.81 (0.89), 0.97 (1.08), 0.01 (0.03); Hyperglycemia (>140 mg/dl): 19.2 (9.9), 16.7 (8.2), 4.3 (2.6).

Conclusion:

This small-scale simulation shows promising results when comparing accepted protocols to an artificial intelligence approach, when interfaced with the same model, initial conditions, and tested against the same exogenous dextrose infusions. These results will be further validated in a larger scale simulation study, which will also include standard PID and MPC controllers, to be conducted by the end of the year.

Reliable Glucose Monitoring by ex-vivo Blood Micro-dialysis and Infrared Spectrometry for Patients in Critical Care

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

South-Westphalia University of Applied Sciences, Interdisciplinary Centre for Life Sciences, Iserlohn, NRW, Germany

delbeck.sven@fh-swf.de

Objective:

Blood glucose monitoring has been realised by biosensors in combination with micro-dialysis using either intravascularly or subcutaneously implanted catheters. Another alternative is ex-vivo micro-dialysis of continuously sampled heparinised whole blood available from ICU patients. A drawback is variable recovery rates that can be observed for all devices. Infrared spectrometry has been suggested for analyte detection and quantification because, besides glucose, other clinically relevant analytes can be simultaneously determined that are important for intensive care patients.

Method:

Perfusates with either acetate or mannitol were investigated as recovery markers. Despite the overlap of mannitol and glucose spectra, their simultaneous accurate quantification by infrared spectrometry was successful, which included also low-mass components such as lactate and urea, contributing to the dialysate spectra.

Result:

In contrast to the previously used acetate marker substance, an almost linear dependency between mannitol loss and glucose recovery was realised for micro-dialysis catheters and two flat and blood-compatible membranes tested within a custom- made micro-dialysator, which provides a straightforward compensation 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. Using this system, blood glucose concentration values can be reliably and continuously monitored, and these measurements can be considered to be the gold standard in glycemic control of critically ill patients.

Monitor Intensely To Meet Intention

Suhas Erande MD

Akshay Hospital Pune, Maharashtra, India

drsse@rediffmail.com

Objective:

The objective is to use an ambulatory glucose profile to attain glycemic targets. In outpatient diabetic clinics, the proportion of patients reaching and maintaining their ADA glycemic targets remains poor. Infrequent or haphazard laboratory- or selfmonitoring remains an obstacle for physician decision-making. I monitored and performed intensive continuous glucose monitoring on patients in order to observe if it helps getting to ADA targets expeditiously.

Method:

The following patients wore a sensor on their arm for 14 days [Abbott India, Ambulatory Glucose Profile (AGP)]: 1) Patients on multi dose insulin regimen, self-monitoring (~4 readings/day) and not reaching pre/postprandial goals or HbA1c; 2) Patients on 1 or 2 insulin injections plus oral anti-diabetic drugs (OADs), self-monitoring (~2 readings/day) and not at glycemic target; and 3) Patients who found it difficult to reach HbA1c goals despite diet/exercise and multiple OADs at full doses (>3 classes of OADs) and monitoring in laboratory. All patients were adults (ages 30 to 82 years) and Type 2 diabetes (duration of Diabetes ~6 months to 25 years). All patients continued their ongoing medication to start with and after every 3 days they visited my clinic for interim downloading and analysis of AGP records. Patients maintained their own dietary record for corroboration. Insulin doses or OAD regimen were modified in tandem with AGP graphs. Patients on insulin and self-monitoring transmitted their sugar values by telephone before each insulin injection so that doses (units) could be suitably titrated. Patients doing physical exercise maintained those records for analysis.

Result:

1) A shorter duration of hyperglycemia was observed in 90% of patients. 2) Reasons for hyperglycemia were identified and addressed (e.g., need for increased titration of insulin or OAD dose, change in frequency of dose, changes in quantity/quality of dietary carbohydrates, timing of exercise or different patterns of insulin resistance- (predominantly nocturnal/late night/post breakfast). 3) Reasons for patient’s complaints were identified in relation to hypoglycemia (e.g., feeling hungry/sweating/trembling or feeling dizzy/exhausted). Hypoglycemic events were, identified and managed and the duration of hypoglycemia was reduced in all patients. 4) The time spent in the euglycemic range (80-140 mg/ld.) improved in all patients. Strategies to maintain euglycemia were discussed and explained to the patients and family (wherever possible). 5) HbA1c (done 12 weeks after the last reading and following an AGP) improved in all patients. In 35% of patients, HbA1C reached the ADA target (<7%). In 50% of patients, HbA1c improved by >2%. 6) Adherence to insulin/OAD doses, the timing, and the regularity of self-monitoring and telephone reporting also improved in all patients.

Conclusion:

Intensive glycemic monitoring using an AGP in patients with diabetes is a helpful and appropriate tool. AGP helps to improve the amount of time spent in the euglycemic range, controls hyperglycemia expeditiously, and simultaneously avoids hypoglycemia. AGP serves as an important way to impart knowledge to patients on insulin/OAD dosing and the timing of both diet and physical exercise. Because India is home to the second largest number of people with diabetes and is a developing economy, patient empowerment is a challenging idea. AGP may prove to be a small important step towards the goal of better health outcomes.

Noninvasive, Continuous Glucose Monitoring with Novel, High-resolution Ultrasound Systems

Rinat O. Esenaliev, PhD

Glucowave, Inc.

League City, Texas, United States

rinesenal@gmail.com

Background:

Many university groups and companies have proposed and tested various noninvasive glucose monitoring techniques. However, limited success has been achieved and noninvasive glucose monitoring remains one of the most challenging and important biomedical problems.

Method:

We proposed, developed, and tested a novel technique for noninvasive, continuous glucose monitoring. It is based on high- resolution detection of ultrasound waves reflected from tissues (e.g. skin). Glucose-induced changes in tissue can be detected with specially designed ultrasound systems that can be used for noninvasive glucose measurements. We developed and built high-resolution, highly compact ultrasound systems with unique, ultra-sensitive probes for detecting ultrasound waves from the outer wrist area. We tested the systems in diabetic (both Type 1 and Type 2) and non-diabetic subjects using glucose and meal intake. The ultrasound signals recorded from the skin were used to continuously measure blood glucose concentrations that were compared to invasively measured reference blood glucose concentrations.

Result:

The noninvasively measured glucose concentrations were in good agreement with the reference glucose concentration in the hypoglycemic, euglycemic, and hyperglycemic ranges. The accuracy of the noninvasive glucose monitors closely approximated that of invasive glucose meters.

Conclusion:

The obtained results indicate that the high-resolution ultrasound technique may provide noninvasive, continuous blood glucose monitoring in diabetic and non-diabetic patients in the physiologic range.

Prediction of Hypoglycemia from CGM Data from 1 to 3 Hours Ahead in Type 1 Diabetes

Chiara Fabris, PhD; Boris P. Kovatchev, PhD; Marc D. Breton, PhD

Center for Diabetes Technology, University of Virginia

Charlottesville, VA, United States

cf9qe@virginia.edu

Objective:

On-demand assessment of the risk for hypoglycemia could support patients with Type 1 diabetes (T1D) in making treatment and behavioral decisions. Here, we propose a method to predict instances of hypoglycemia up to 3h ahead from information-rich continuous glucose monitoring (CGM) data.

Method:

Assessments of future hypoglycemia below 70mg/dl were run at 30min time increments. For each 30min window, minimum blood glucose (BG), CGM slope, and CGM curvature were calculated and included in a logistic regression classifier. Logistic regression coefficients were obtained using bootstrap to compute the probability for upcoming hypoglycemia for 1h, 2h, and 3h prediction time horizons. The method was developed from 1 month of CGM data from 54 T1D subjects, and then applied, without any further modification, to an independent test dataset consisting of 1 month of blinded CGM from 12 T1D subjects.

Result:

At 10% false alarms, prediction rates over 1h, 2h, and 3h reached 93%, 80%, and 70% on the test dataset. When an alert for hypoglycemia was observed (vs. no alert), minimum BG across the prediction horizons was 68, 66, and 63 mg/dl (vs. 170, 158, and 147mg/dl), percent time below 70 mg/dl was 29%, 27%, and 25% (vs. 0.2, 0.9, and 1.8%), and low blood glucose index was 6.7, 6.1, and 5.6 (vs. 0.1, 0.3, and 0.5); all differences were statistically significant (p<0.001). Average glucose at the time of alert was 101, 105, and 107 mg/dl.

Conclusion:

We present a novel CGM-based method predicting more than 9 out of 10 upcoming hypoglycemic events, 1h ahead, and 7 out of 10, 3h ahead, allowing significant time for the patient to avoid low BG. Further analysis will be needed to associate optimal decisions with these alerts.

3rd Generation Biosensor Technology for Continuous Glucose Monitoring Systems

Alfons Felice, MSc; Aleksandra Pinczewska, PhD; Christopher Schulz, PhD; Roman Kittl, PhD; Roland Ludwig, PhD

DirectSens GmbH

Klosterneuburg, Lower Austria, Austria

alfons.felice@directsens.com

Objective:

Glucose sensors used for diabetes management rely not only on specific and sensitive enzymes but also on an efficient transmission of electrochemical information. The DirectSens team has leveraged the unique property of direct electron transfer of a specifically designed glucose-sensing enzyme for development of 3^rd generation continuous glucose monitoring systems. The main advantage of the presented technology is its simplicity; a high amperometric glucose response can be reached at very low working potentials without the need for mediators (including oxygen) or nano-structures rendering the system very precise and reproducible.

Methods:

Sensor prototype properties were carefully tuned in order to meet requirements of CGM systems with a focus on simple chemistry and performance. The sensors were operated at the low potential of -0.1 V (vs. an Ag|AgCL reference electrode) and key performance parameters were evaluated.

Results:

Very high sensitivities (60 na/mM/mm²), a low limit of detection (0.24 mM) over a reasonable measurement range, and linearity up to 10 mM were recorded. The engineered sensor matrix was stable over a period of days during continuous glucose measurements. Signals from common interferences, including ascorbic acid and acetaminophen, were found to be within the < 10% range without employing a selective polymer coating. Data from in vivo animal trials will also be reported.

Conclusion:

The presented prototype of a 3^rd generation glucose sensing technology is a cost efficient and precise system with a high potential for implantable sensor platforms. The focus will now be: (1) to show in vivo proof of concept and (2) to identify potentially complementary technologies.

Measurement Accuracy of F our Blood Glucose Monitoring Systems with Insulin Dose Advisors

Guido Freckmann, MD; Thomas Leucht, PhD; Nina Jendrike, MD; Annette Baumstark, PhD; Stefan Pleus, MSc; Alexandra Beer, PhD; Frank Flacke, PhD; Cornelia Haug, MD

Institute for Diabetes Technology Research and Development Corporation at the University of Ulm

Ulm, Germany

guido.freckmann@idt-ulm.de

Objective:

Some blood glucose monitoring systems (BGMS) have built-in insulin dose advisors that provide recommendations for the next insulin dose. Adequacy of the suggested dose depends, among other factors, on the measurement quality of the BGMS.

Method:

The BGMS Accu-Chek Aviva Expert [1], FreeStyle InsuLinx [2], FreeStyle Precision Neo [3] and MyStar DoseCoach [4] were evaluated in this study. Blood glucose levels in capillary blood samples from 100 subjects were determined with three lots of each BGMS and two different comparison methods (hexokinase [HK]- and glucose oxidase [GOD]-based methods) and deviations between the results were calculated. According to ISO 15197:2013, at least 95% of the blood glucose results measured with the BGMS have to be within ±15 mg/dl from comparison method results at glucose concentrations <100 mg/dl and within ±15% at concentrations >100 mg/dl.

Result:

When evaluated against the HK method, BGMS 1 had 92% to 99.5% of measurement results of the three lots within the allowed limits, BGMS 2, 83.5% to 98%, BGMS 3,73.5% to 83.5% and BGMS 4, 99% to 100%. When evaluated against the GOD method, BGMS 1 had 98.5% to 100% of measurement results of the individual lots within the allowed limits, BGMS 2, 97% to 99.5%, BGMS 3, 86% to 96% and BGMS 4,98% to 99%. All BGMS had 100% of the results within zones A and B of the Consensus Error Grid.

Conclusion:

Especially for BGMS used by patients to directly calculate insulin doses, a high level of accuracy is required to ensure safety of the patients. For adequate therapy adjustment, it is essential that the results be precise with no relevant systematic measurement deviation,.

Post-marketing Evaluation of a Blood Glucose Monitoring System’s Measurement Accuracy

Guido Freckmann, MD; Annette Baumstark, PhD; Nina Jendrike, MD; Stefan Pleus, MSc; Delia Rittmeyer, MSc; Ulrike Kamecke, MEng

Institute for Diabetes Technology Research and Development Corporation at the University of Ulm

Ulm, Germany

guido.freckmann@idt-ulm.de

Objective:

In Germany, blood glucose monitoring systems (BGMS) are usually purchased from pharmacies or expert stores, but BGMS are also available from other suppliers, such as supermarkets. Accuracy of the available test strip lot of a BGMS from a German discounter chain was tested following ISO 15197.

Method:

Test strips from 2 separate orders (A and B) with identical lot numbers were tested with 2 meters each. Capillary blood glucose samples from 100 subjects distributed into glucose concentration categories stipulated by ISO 15197:2013 were measured with 2 meters for each of the 2 orders to achieve 400 BGMS measurements. Measurement results were compared to those obtained with the glucose oxidase-based YSI 2300 STAT Plus and a second hexokinase-based comparison method. Trueness and precision of the comparison measurement methods were verified during the test procedures using NIST SRM 965. Accuracy of the BGMS was evaluated applying ISO 15197:2013 criterion A: > 95% of results have to be within ±15 mg/dl from comparison values at glucose concentrations <100 mg/dl and within ±15% at higher concentrations. In addition, less stringent limits of ISO 15197:2003 were applied.

Result:

With test strips from order A, 63% of 200 results were within the ISO 15197:2013 limits and with test strips from order B, 50.5% of results were within the limits. Regarding ISO 15197:2003, 73.5% and 62.5%, respectively, fell within the limits. The relative bias was +14% and +16.6%, respectively.

Conclusion:

The BGMS showed on average higher results than the comparison methods with the evaluated test strip lot. These data demonstrate that post-marketing evaluations of available BGMS are important to ensure safe therapy of patients, especially when they are treated with insulin.

System Accuracy of a Novel System for Blood Glucose Monitoring Linked to a Smartphone

Guido Freckmann, MD; Nina Jendrike, MD; Thomas Leucht, PhD; Annette Baumstark, PhD; Delia Rittmeyer, MSc; Stefan Pleus, MSc; Cornelia Haug, MD

Institute for Diabetes Technology Research and Development Corporation at the University of Ulm

Ulm, Germany

guido.freckmann@idt-ulm.de

Objective:

Recently, a novel blood glucose monitoring system coupled to a smartphone (PixoTest™, iXensor Co. Ltd.) has been developed. In this system, the test unit, containing the test field as well as a lancet, can be attached to an iPhone. Blood glucose values are determined colorimetrically via the camera and displayed in the associated app. In this study, measurement accuracy of this system was evaluated following ISO 15197:2013.

Method:

For the evaluation, capillary blood samples from 100 different subjects were tested with three test strip lots of the blood glucose monitoring system. Comparison measurements were performed with the glucose oxidase-based laboratory method YSI 2300 STAT Plus. The percentage of results within ±15 mg/dl from the comparison method at glucose concentrations <100 mg/dl and within ±15% at concentrations >100 mg/dl was calculated. ISO 15197:2013 specifies that at least 95% of the results of each test strip lot have to be within these limits and that 99% of all results have to be within zones A and B of the consensus error grid.

Result:

For the three test strip lots 99%, 97% and 98.5% of the respective measurement results were within the limits of ISO 15197:2013. In addition, 100% of the values were within zone A of the consensus error grid. The relative bias was between -0.6% and -1.0%.

Conclusion:

The new blood glucose monitoring system linked to a smartphone achieved the system accuracy requirements of ISO 15197:2013. Convenient and discrete measurements with the all-in-one-system might contribute to regular self-monitoring of blood glucose, as recommended for patients with diabetes.

Non-Invasive Glucose Monitoring Device: Reducing Impact of Expected Postprandial Time Lag During Measurement

Avner Gal, MSc, MBA; Ilana Harman-Boehm, MD; Andrew Drexler, MD; Yulia Mayzel, MSc; Keren Horman, MSc, Michal Markovits, BSc, Karnit Bahartan, PhD, Tamar Lin, MSc

Integrity Applications Ltd.

Ashdod, Israel, Israel

AvnerG@integrity-app.com

Objective:

GlucoTrack® is a non-invasive glucose monitoring device that measures glucose in the tissue, rather than in the blood, thus it is subject to postprandial lagging effects that may affect accuracy. In order to improve postprandial tracking, an algorithm was developed and evaluated that compensates for delays relative to blood-glucose changes by accounting for the time of meal intake and glycemic load.

Method:

We compared device performance using new versus previous algorithms, during pre- and post-prandial times. Both algorithms were applied to the same dataset of 34 Type 2 subjects (2,033 paired invasive/non-invasive readings) who consumed various meals with different glycemic loads throughout the day. Results were analyzed at pre- and 30-90 minutes post-prandial, using Clarke Error Grid (CEG), Surveillance Error Grid (SEG), Median Absolute Relative Difference (MARD), and Mean Absolute Difference (MAD).

Result:

Previously, pre-prandial measurements showed better accuracy than post-prandial measurements (100.0% vs. 96.5% in CEG A+B zones, 58.9% vs. 49.3% in A zone, 15.8% vs. 20.3% MARD, and 26.3 vs. 37.3 mg/dLin MAD). Using the new algorithm, 98.1% of post-prandial points were in CEG A+B zones, 57.0% in zone A, MARD was 16.9% and MAD was 32.9 mg/dL. SEG analysis showed that the measurements obtained with the new algorithm were mostly within the deep green zone (“no risk”; 63.6%) or within the light green zone (“slight, lower risk"; 27.6%).

Conclusion:

Incorporating information regarding recent meals, significantly improves GlucoTrack accuracy at different post-prandial times. These results point to better tracking of postprandial glucose excursions that may enable optimized post-prandial control. Users may also benefit from the stored meal-related information to better understand their diabetes and actively participate in its management.

Defining Fasting Glucose from Continuous Glucose Monitoring in Patients with Type 1 Diabetes

Thibault P. Gautier, BSc; Chiara Fabris, PhD; Boris P. Kovatchev, PhD; Marc D. Breton, PhD

Center for Diabetes Technology, University of Virginia

Charlottesville, VA, United States

mb6nt@virginia.edu

Objective:

Assessment of fasting blood glucose (BG) from self-monitoring of BG (SMBG) is a key clinical index, but no indication about the fasting state is provided in continuous glucose monitoring (CGM) data. We present a novel algorithm for defining fasting BG from CGM in Type 1 diabetes (T1D).

Method:

BG drawn from the CGM profile at the time of the first SMBG after 5 AM was considered as the reference fasting value. Kernel density estimation was used to compute the probability of fasting BG across the day using the fasting reference and CGM slope, both from the entire population and the specific subject. CGM slope and curvature were used to build a day- specific probability law. Using a Bayesian update, population-level data, and patient- and day-specific information the CGM-based fasting BG was derived as the CGM value at the mode (maximum) of the posterior distribution. The method was developed on training data collected from 54 T1D subjects over 4-week monitoring, and then tested on an independent test set of 12 subjects monitored for 30 days. Pearson’s correlation and linear regression were computed between CGM- based and reference fasting BG.

Result:

Correlation for the mean-per-patient fasting BG of 0.87 (p< 0.001) and 0.79 (p< 0.001) were obtained in training and test set, respectively, with coefficient of determination (R²) from linear regression of 0.76 and 0.62. The same analysis from 4- day averaging produced correlations of 0.80 (p<0.001) and 0.82 (p < 0.001), and R² of 0.64 and 0.67.

Conclusion:

We propose a novel approach that enables the determination of fasting BG from CGM in T1D without additional information, potentially enabling accurate enough CGM to be used to track treatment effectiveness.

Performance of the GLUCOCARD^® Vital Blood Glucose Monitoring System against the ISO 15197:2013 Accuracy Criteria

Patricia Gill, BA, MLT; Julie Walker, RN, BSN, PHN; Danielle Maher, BS; John Gleisner, BS, PhD

ARKRAY USA, Inc.

Minneapolis, Minnesota, United States

gillp@arkrayusa.com

Objective:

The purpose of this study is to determine the performance of the GLUCOCARD® Vital Blood Glucose Monitoring System (BGMS) against accuracy boundaries of ISO 15197:2013 in ongoing trending studies. This standard requires that 95% of BGMS results be within ±15mg/dL of the reference analyzer at glucose concentrations <100 mg/dL and within ±15% of the reference analyzer at glucose concentrations >100 mg/dL. Furthermore, 99% of all results are required to be within the A and B zones of the Consensus Error Grid.

Method:

Fingerstick testing was collected by trained laboratory professionals from subjects with diabetes (n=240) on eight lots of GLUCOCARD® Vital at the ArKrAY USA factory in Minneapolis, MN. Reference values were obtained by using the YSI Model 2300 Analyzer. The data were analyzed against the accuracy boundaries of the ISO 15197:2013 standard and the percentage of the results in the A plus B zones of the Consensus Error Grid.

Result:

The data show that 100% of the results <100 mg/dL (27/27) were within ±15 mg/dL of the YSI and 99.5% of the results >100 mg/dL (212/213) fell within the ±15% of the YSI. All data were within the A and B zones of the Consensus Error Grid. The overall bias to YSI was -1.9%. The correlation coefficient (r) was 0.98, which demonstrates a strong linear relationship between GLUCOCARD® Vital and the YSI reference method.

Conclusion:

The data acquired on the GLUCOCARD® Vital BGMS by laboratory professionals was within the accuracy boundaries of the ISO 15197:2013 standard.

Exploring the Wisdom of the Diabetes Online Community: Themes from Social Networking Conversations

Deborah A. Greenwood, PhD, RN, BC-ADM, CDE; Perry M. Gee, PhD, RN; Corinna Cornejo, BA; Mila Ferrer, BA; Kathy Nelson, PhD; Melissa Lee, BA; Mariana Gomez, BA; James Ferrer, BA; Lisa Miller, PhD.

Research, Development and Dissemination, Office of Patient Experience, Sutter Health, Sacramento, CA, United States

Greenwd@sutterhealth.org

Objective:

Diabetes requires 24/7 self-management, 99% of the time outside of the healthcare system. The diabetes online community (DOC) creates a safe place where people can share experiences. The objective of this research project was to identify themes relating to improved health outcomes when participating in the DOC.

Methods:

A verbatim transcript of publicly available Twitter data (Symplur LLC, Pasadena, CA), during two Tweet Chats (781 tweets from 85 individual participants in November 2015), were evaluated using a qualitative content analysis, completed first manually and then by the use of MAXQDA Version 11.1 qualitative analysis software (VERBI GmbH software Berlin, Germany). The analysis began with open coding to analyze the data as a whole. The data were then grouped into broad categories and sub-categories within the context of two questions posed on Twitter. Two independent researchers coded the data and discrepancies were resolved through discussion and mutual reevaluation of the data. Last, the findings were shared with Twitter chat participants to verify accuracy.

Results:

The major themes identified from the individual questions are as follows:

1. What health outcomes (clinical and quality of life) do you think can be improved by being a member of the DOC?

   • Improved social connectedness (support and compassion for others)
   • Enhanced quality of life
   • Empowerment of self-management skills
2. How can we encourage people with Type 2 diabetes to engage in social support via social media?

   • Overcoming stigma
   • Encouraging provider engagement
   • Managing barriers (false information, privacy and negativity)

Conclusion:

We found that DOC participants identified quality of life, empowerment, social connectedness and not feeling alone as primary themes. Being connected and not stigmatized is an important benefit of the DOC.

Virtual Patients with Circadian Rhythm for Type 1 Diabetes

Benyamin Grosman, PhD; Di Wu, PhD; Anirban Roy, PhD; Neha Parikh, PhD; Rebecca Gottlieb, PhD

Medtronic

Northridge, California, United States

benyamin.grosman@medtronic.vom

Background:

Simulations using mathematical models are important for studying, developing and improving insulin delivery therapies for people with Type 1 diabetes (T1DM). Medtronic CareLink® database has a large pool of patients’ sensor augmented pump (SAP) uploads. A method is introduced to transform this data and create a large number of virtual patients with a variety of insulin sensitivities including circadian rhythms, pharmacokinetics, age, and gender.

Method:

The CareLink® database is used to generate a large number of virtual patients. For this purpose, a mathematical model was developed along with a dedicated parameter optimization method to overcome the intrinsic identification problem of the SAP data. Each virtual subject has been trained on 10 days of CareLink® uploads and was validated using a mean of 14 ± 11 days. The virtual patients passed the null hypothesis that was conducted on the therapy outcomes (i.e., percentage of time above 180 and below 70 mg/dL).

Results:

A total of 404 virtual patients have been developed. Circadian rhythm in the insulin gain (the glucose drop as a result of an increase of 1U/h of insulin delivery) was identified in 71 of them. In addition, each virtual patient was associated with specific insulin sensitivity, specific meal absorption rates (i.e., breakfast, lunch, and dinner), age group, and total daily insulin. The percent time between 70 and 180 mg/dL for the CareLink® uploads and the simulated glucose is 76 ± 12 and 79 ± 11%, respectively. The percent of time below 70 mg/dL for the virtual patients and the real CGM data is 4 ± 4% and 5 ± 4%, respectively.

Discussion:

The Medtronic CareLink® database was demonstrated to be a rich pool for producing a large number of virtual patients with a variety of time-dependent insulin sensitivities, pharmacokinetics, and meal absorption rates. This new simulation environment can be used to further improve and develop the insulin delivery techniques for people with T1DM and especially the development of artificial pancreas algorithms and systems.

Performance and Accuracy Capability of a New, Wireless-enabled Blood Glucose Monitoring System (Contour®Next ONE) That Links to a Smart Mobile Device: Laboratory and Clinical Sample Reapplication Studies

Bern Harrison, BA; Daniel Brown, PhD; Miho Takeshima, MSc

Ascensia Diabetes Care

Parsippany, New Jersey, United States

bern.harrison@ascensia.com

Objective:

The objective is to evaluate the performance and accuracy of the Contour®Next ONE blood glucose monitoring system (BGMS) during Second-Chance® sampling. The BGMS features an easy-to-use, wireless-enabled BG meter that links to a smart mobile device via Bluetooth® technology and syncs with the Contour™ Diabetes app on a smartphone or tablet.

Method:

In the laboratory study, blood samples were tested in 3 temperature environments (16°C, 22°C, 34°C) with blood adjusted to 3 BG levels (70, 300, 500 mg/dL) at 3 hematocrit levels (20%, 42%, 55%). Two sample reapplication volumes (0.28 and 0.46 pL) were added to the initial blood sample volume and each sample was tested with 3 delay times between the initial and second inoculation (5, 30, 55 seconds). For each sample and test condition, 10 replicate BGMS readings were obtained with each of 3 test strip lots. In the fingerstick/clinical study, 52 subjects with diabetes intentionally performed self-tests with an initially insufficient application that required a second blood sample application in order to produce an adequate number of Second-Chance® sampling tests. Results were compared with YSI reference results and assessed per the following acceptance criterion: 95% of results within ±15 mg/dL (BG <100 mg/dL) or ±15% (BG > 100 mg/dL) of reference result.

Result:

In the laboratory study, results met protocol-specified acceptance criteria. In the clinical study (BG range, 65-347 mg/dL; hematocrit range, 35%-55%), 100% (85/85) of subject fingertip self-test results were within ±15 mg/dL or ±15% of the YSI reference result. Moreover, 92.9% (79/85) of self-test results were within ±10 mg/dL or ±10% of the YSI reference result.

Conclusion:

In the laboratory and in a clinical setting when used by subjects with diabetes, BGMS sample reapplication results met acceptance criteria.

Performance and Accuracy Capability of a New, Wireless-enabled Blood Glucose Monitoring System (Contour®Plus ONE) That Links to a Smart Mobile Device: Laboratory and Clinical Sample Reapplication Studies

Bern Harrison, BA; Daniel Brown, PhD; Miho Takeshima, MSc

Ascensia Diabetes Care

Parsippany, New Jersey, United States

bern.harrison@ascensia.com

Objective:

The objective is to validate the Contour®Plus ONE blood glucose monitoring system (BGMS) performance and accuracy capability during Second-Chance® sampling. This new BGMS features an easy-to-use, wireless-enabled BG meter that links to a smart mobile device via Bluetooth® technology and syncs with the Contour™ Diabetes app on a smartphone or tablet.

Method:

In the laboratory study, blood testing was conducted in 3 temperature environments (16°C, 22°C, 34°C) with blood adjusted to 3 BG levels (70, 300, 500 mg/dL) at 3 hematocrit levels (20%, 42%, 55%). Two sample reapplication volumes (0.28 and 0.46 pL) were added to the initial blood sample volume and each sample was tested with 3 delay times between initial and second inoculation (5, 30, 55 seconds). For each sample and test condition, 10 replicate BGMS readings were obtained with each of 3 test strip lots. In the fingerstick/clinical study, 52 subjects with diabetes intentionally performed self-tests with an initially insufficient application that required a second blood sample application in order to produce an adequate number of Second-Chance® sampling tests. Results were compared with YSI reference results and assessed per the following acceptance criterion: 95% of results within ±15 mg/dL (BG <100 mg/dL) or ±15% (BG >100 mg/dL) of reference result.

Result:

In the laboratory study, results met protocol-specified acceptance criteria. In the clinical study (BG range, 65-347 mg/dL; hematocrit range, 35%-55%), 97.7% (84/86) of subject fingertip self-test results were within ±15 mg/dL or ±15% of the YSI reference result. Moreover, 90.7% (78/86) of self-test results were within ±10 mg/dL or ±10% of the YSI reference result.

Conclusion:

The BGMS sample reapplication results met acceptance criteria in the laboratory and in a clinical setting when used by subjects with diabetes.

Insulin Degludec: Four-times Lower Pharmacodynamic Within-patient Variability Compared to Insulin Glargine U300 in Type 1 Diabetes

Tim Heise, MD; Marianne Norskov, PhD; Leszek Nosek, MD; Kadriye Kaplan, MS; Susanne Famulla, PhD; Hanne L. Haahr, PhD

Profil

Neuss, Germany

tim.heise@profil.com

Objective:

To compare within-patient variability for the pharmacodynamic (PD) effects of insulin degludec (IDeg) and insulin glargine 300U/mL (IGlarU300) in patients with Type 1 diabetes.

Method:

In this double-blind, randomized, crossover study, 57 completers received 0.4 U/kg of IDeg or IGlarU300 once daily for 12 days. Day-to-day PD variability (within-patient variances) was investigated from the glucose-infusion rate (GIR) profiles of three 24-hour (h) euglycemic glucose clamps performed at steady state (SS) (day 6, 9 and 12).

Result:

Day-to-day PD variability was four-times lower with IDeg compared to IGlarU300 (variance ratio IGlarU300/IDeg was 3.70, 95%CI [2.42; 5.67] for AUC_GIR,_tSS [p<0.001] and 2.48, 95%CI [1.42; 4.31] for GIR_maxSS [p=0.002]). Variability was consistently low with IDeg over the entire 24-h period, but increased with IGlarU300 to a maximum at 12-14h after dosing (variance IGlarU300/IDeg AUC_GIRi0-₂h,ss, 0.19/0.12; AUC_GIR,i_2-i_4h,_SS,2.01/0.18). The PD effect of IDeg was evenly distributed over 24-h with 24-26% of the total effect seen in the intervals of 0-6h, 6-12h, 12-18h and 18-24h. In contrast, IGlarU300 had a greater effect in the first and the last 6h-interval (35% and 28%) and lesser activity at 6-12h (20%) and 12- 18h (17%). The potency (total glucose-lowering effect, AUC_GIR,_t,_SS) of IGlarU300 was 30% lower compared to IDeg (estimated ratio 0.70, 95%CI [0.61; 0.80], p<0.001).

Conclusion:

IDeg has a lower within-patient variability in its glucose-lowering effect and a more even distribution of its PD activity over 24h than IGlarU300. Furthermore, IGlarU300 has a lower potency than IDeg. These results suggest that, with IDeg, patients can achieve lower fasting glucose targets at lower risk of hypoglycemia compared with IGlarU300. Further clinical investigation into the two insulins is warranted.

A Novel Bi-hormonal Control Strategy Inspired on the Paracrine Interaction Between Beta Cell and Alpha Cell

Pau Herrero, PhD; Jorge Bondia, PhD; Nick Oliver, FRCP; Pantelis Geogiou, PhD

Imperial College London

London, United Kingdom

pherrero@imperial.ac.uk

Objective:

To date, bihormonal closed-loop systems for glucagon and insulin delivery in Type 1 diabetes have been based on two independent controllers (e.g. MPC+PD, PID+PID, Bio-inspired+PD). However, it is well known that the secretion of these two hormones in the body is closely interconnected. In this work, we present a novel biologically-inspired glucose control strategy that accounts for such coordination.

Method:

To account for the potentiation of insulin secretion by plasma glucagon levels, a positive feedback loop was incorporated into a standard bi-hormonal control strategy. Specifically, the Imperial College Artificial Pancreas controller was employed in this work. The UVa-Padova T1DM v3.2 system was used to carry out a two-week in silico study on 11 adult subjects taking into account inter- and intra-subject variability of insulin requirements and uncertainty of carbohydrate intake. The controller tuning was purely based on the insulin sensitivity factor (i.e. correction factor) and a newly defined concept referred to as glucagon sensitivity factor. Using the same controller tuning, the coordinated bi-hormonal control strategy (CO) was compared against its non-coordinated counterpart (NC).

Result:

Mean blood glucose (NC vs. CO): 136.03 ±6.84 vs. 130.02 ± 5.97 (p<0.01); percentage time in target [70,140] mg/dl: 61.37 ±6.52% vs. 66.75 ± 5.84% (p <0.01); percentage time below 70 mg/dl: 1.53 ± 1.02% vs. 1.79±1.50%(p = 0.16); percentage time above 140 mg/dl: 37.09 ± 6.46% vs. 31.44 ± 6.04% (p < 0.01); percentage time in target [70,180] mg/dl: 84.53 ± 5.76%vs. 88.17 ± 4.73% (p <0.01); percentage time above 180mg/dl: 13.93 ± 5.87%vs. 10.02 ± 4.97% (p < 0.01); daily average of insulin delivered (U): 43.61 ± 9.92 U vs. 44.99 ± 10.25 U (p<0.01); and daily average of glucagon delivered (mg): 0.64 ± 0.52 mg vs. 0.70 ± 0.53 mg (p < 0.01).

Conclusion:

Compared to a standard bihormonal control approach, the proposed coordinated control strategy significantly reduces hyperglycemia without increasing hypoglycemia. Although statistically significant, the increases in insulin and glucagon delivery were clinically marginal.

Retrospective analysis of Impact on SMBG and Glycemic Control of Mobile Health (MHealth)-Application for Diabetes Management

Marcus Hompesch, MD; Lindsey Hergesheimer, BSc; Klaudius Kalcher, PhD; Roland Boubela, PhD; Fredrik Debong

Profil Institute for Clinical Research

Chula Vista, CA, United States

marcus.hompesch@profilinstitute.com

Objective:

To investigate the potential impact of the mySugr Logbook app usage on blood glucose (BG) control. The mySugr Logbook (registered class I medical device application) was developed to make logging of metabolic control data appealing and useful in day-to-day life, and is one of the market leading apps with over 800,000 registered users. Using mHealth tools for diabetes self-management may have a beneficial impact on the quality of metabolic control; however relevant and sufficient real-world data is lacking, as many mHealth projects never go beyond pilot stage.

Method:

A randomly selected group of 2,104 highly engaged users (logging >5 days/week for >6 months) were included (T1D, aged 34.5 ± 16.13 years, 45.77% female). The change of BG results [mean, standard deviation (SD), coefficient of variability (CV)] within the group was analyzed at baseline (to), month 1 (f), and month 2-6 (t₂), using R software. Baseline data was based on an intercept of regression model of all data from f, which is more stable than data from day/s 1/1-3.

Result:

Baseline BG results (t₀) were 162.10 ± 59.25 mg/dl, dropping to 156.41±55.67 mg/dl at f and decreasing further to 155.33±52.96 mg/dl at t₂ - a stable reduction in mean of 4.1%, SD of 11% and CV of 6.8% (P<10^-10).

Conclusion:

The reduction of parameters indicative of BG variability, SD and CV, demonstrate that BG logging alone with the mySugr Logbook app may have positively impacted the quality of BG control. These findings highlight the necessity for a prospective, controlled clinical study. We hypothesize that the addition of upcoming features to the application (coaching, bolus calculator and more) will result in further improvements of self-monitoring behavior and glycemic control for highly engaged users.

Results Support Efficacy and Clinical Efficiency of Diabetes Management Decision Support Software for Blood Glucose Control

Lucienne Ide MD, PhD; Lindsey Valenzuela, PharmD, BCACP; Jade Le, PharmD, BCACP

Desert Oasis Healthcare

Palm Springs, CA, United States

jle@mydohc.com

Introduction:

In the United States, there are approximately 29 million people, or 9.3% of the population, estimated to have diabetes and almost 2 million new diagnoses made yearly. In addition, 86 million adults aged 20 years and older have prediabetes.

Diabetes not only negatively impacts patients’ quality of life and increases the risk of death by 50%, but also accounts for $176 billion in direct medical costs and $69 billion in indirect costs form lost workdays, restricted activity, disability, and early death. Published literature has established the value of disease management programs led by pharmacists can improve clinical parameters such as A1c, lipid levels, blood pressure, better quality of life, and significantly lower health care costs. Pharmacist-run clinics improve patient outcomes and reduce healthcare-associated costs partly through a decrease in hospital utilizations for patients with chronic conditions like diabetes. Statistically significant improvement in patient outcomes and reductions in hospital utilization have been demonstrated by pharmacists at Desert Oasis Healthcare. The value added services serve to preserve patient satisfaction, physician satisfaction, and reduce healthcare costs by helping to reduce hospitalizations. Pharmacists have consistently been proven to improve surrogate markers such as A1c, lipids and blood pressure measures through high impact disease management programs. To better scale our current pharmacist-led program to include the ever growing population of diabetics, while preserving cost pressures in healthcare, Rimidi software was implemented. The impact of decision support software (Diabetes + Me ^TM) was used to model the anticipated effect of medication changes to the treatment of diabetes on clinical markers such as A1c as well as the time required to achieve meaningful clinical A1c goals. The primary objective of the study is to show that the use of Diabetes + Me software is superior to standard care as assessed by improvements in glycemic control at 12 weeks as compared to baseline. In addition, the decrease in frequency of hypoglycemia (# events per week), increase in percentage of patients with blood pressure in the normotensive range, decrease in BMI, healthcare provider acceptance of the software technology, time to A1c goal attainment, decrease in medication titrations needed to reach A1c goal, improvement in patient satisfaction with diabetes management, will be measured.

Methods:

Patients were randomly stratified to a Rimidi Arm using the Diabetes + Me tool (standard pharmacist intervention in addition to the use of the tool) versus the Non-Rimidi arm (standard of care pharmacist intervention). The study lasted from September 2014 to August 2016 and sampled two cohorts of 43 comparative cases. Members who were eligible for the study included Type 1 or 2 diabetes, HbAlc > 9% within 3 months prior to screening, age 18-80, and patients not currently managed by clinicians. Patient knowledge and self-efficacy were measured using a Rand paired survey tool.

Results:

Average absolute A1c, from baseline at week 0 to week 12, was reduced with Rimidi Arm versus non-Rimidi arm (5.74% versus 7.2%). BMI, blood pressure, lipid panel were also reduced during the study period for both arms. Patient satisfaction with the use of the technology tool and patient self-efficacy improved following the 12-week intervention. The final results will be available for presentation in October 2017.

Conclusion:

Previous studies have demonstrated that every 1% drop in A1c translates into a 14% reduction in acute myocardial infarction and a 33% reduction in the incidence of microvascular complications from diabetes. Results from this study indicate that the Diabetes + Me tool helps to ensure a safe but meaningful reduction in A1c and therefore reduction in event rates as well as overall healthcare costs. Diabetes + Me has not only led to improved benefits to patients but has also allowed Desert Oasis Healthcare to expand the scalability of its already successful diabetes management program without having to make additional investments in supplementary healthcare providers. Patient feedback related to improved confidence in care plan implementation, and the enhanced engagement in their health and disease control, further support Rimidi’s benefit in addressing part of the “triple aim” as a modern healthcare goal.

Continuous Glucose Monitoring in a Cystic Fibrosis patient with glucose intolerance: A possible tool to predict pulmonary exacerbation?

Taylor Inman, MD; James Proudfoot, MSc; Carla Demeterco-Berggren, MD, PhD

Department of Pediatric Endocrinology, University of California San Diego

San Diego, CA, United States

cdemeterco@rchsd.org

Objective:

Patients with cystic fibrosis experience a significant decline in pulmonary status two years before a diagnosis of cystic fibrosis-related diabetes. We hypothesized that hyperglycemia may play a role in the increase in pulmonary exacerbations and in the decline in pulmonary status. Long term continuous glucose monitoring has not been reported in patients with cystic fibrosis and impaired glucose tolerance.

Method:

We performed continuous glucose monitoring for three months on a patient with cystic fibrosis and impaired glucose tolerance to evaluate changes in blood glucose levels prior to diagnosis of a pulmonary exacerbation with decline in pulmonary function tests.

Results:

Results revealed that the patient had elevated overnight, fasting and post-prandial blood glucose levels up to one week prior to the diagnosis of a pulmonary exacerbation compared to when he was well. It was also found that he had elevated mean glucose and spent a greater percentage of time with interstitial glucose >140 mg/dL up to one week prior to the diagnosis of a pulmonary exacerbation.

Conclusion:

This study suggests that hyperglycemia in this population may contribute to pulmonary exacerbations. Blood glucose values could be an additional marker for pulmonary exacerbation and allow for earlier intervention prior to a significant decline in pulmonary function. This study stresses the importance of larger longitudinal studies to help understand the impact of glycemic control and pulmonary function in patients with cystic fibrosis and glucose intolerance.

Dosing Accuracy of Different Insulin Pumps

Ulrike Kamecke, MEng; Ralph Ziegler, MD; Stefan Pleus, MSc; Guido Freckmann, MD

Institute for Diabetes Technology Research and Development, Ulm University

Ulm, Germany

ulrike.kamecke@idt-ulm.de

Objective:

In continuous subcutaneous insulin infusion (CSII) therapy, bolus doses are delivered to cover meals and to correct high blood glucose values. Precision and trueness of insulin pump bolus dosing may have an influence on clinical outcomes.

Method:

In an experimental setting following EN 60601-2-24, four different pumps with different insulin infusion sets (IIS) were evaluated (Accu-Chek® Spirit Combo with Accu-Chek® FlexLink and Accu-Chek® Rapid-D Link; Accu-Chek® Insight with Accu-Chek® Insight Flex and Accu-Chek® Insight Rapid; Paradigm® VeoTM with MiniMed® MioTM, MiniMed® Sure-T®; and MiniMed® Quick-set®; mylife™ OmniPod® with its infusion set). Precision and trueness of bolus deliveries of 1 U and 0.1 U were assessed. Pumps were installed with the tip of the cannula in a water-filled beaker placed on an electronic balance. To avoid evaporation an oil film was applied. After a run-in period, 25 successive boluses were delivered and weighed individually. Each combination of pump and IIS was tested 9 times with each bolus volume.

Result:

There were considerable differences in the scattering among single boluses and between the different pump models. Dosing accuracy was better and more precise for bolus doses of 1.0 U compared to 0.1 U. Inter-quartile range was between 6% and 28% for 0.1 U and between 2% and 21% for 1 U for the different pumps. For 0.1 U, the maximal deviation from the target value was 64% whereas for 1 U it was 42%.

Conclusion:

In conclusion, dosing precision and trueness was adequate for larger bolus doses for all tested systems. However, dosing precision for small bolus doses showed considerable differences. These differences can be clinical significant.

Predictors of Glycemic Control in Adults with Type 2 Diabetes on Insulin

Davida Kruger, MSN, APRN-BC, BC-ADM; Robert Morlock, PhD; Richard Wood, MBA; Christine Kuerschner, MBA; Jay Warner, MBA

Henry Ford Health System

Detroit, MI, United States

dkruger1@hfhs.org

Objective:

Type 2 diabetes mellitus (T2DM) is a progressive disease, and despite new medications, insulin is still needed. Unfortunately, previous studies have concluded that the majority of people on insulin are uncontrolled and improvements are needed. In order to understand this population, this study examines variables predictive of glycemic control.

Method:

An online survey for people with diabetes was fielded 1Q-2016 using the dQ&A Patient Panel. Types of insulin treatment, years using insulin, age at diagnosis, comorbidities, use of glucose monitoring, self-reported adherence, insurance, and patient sociodemographics were collected. Multivariate and descriptive statistics were used to describe patients achieving Hemoglobin A1c (HbA1c) < 7.

Result:

Adults with T2DM, using insulin and reporting HbA1c (n=1,440) had a mean age of 61.28 (9.64) years, 60.25% were female, and the mean duration of insulin was 9.09 years. Less than 40% reported achieving HbA1c < 7%. Participants reaching target were older (62.64 vs. 60.38 years; p <0.001), diagnosed with T2DM at an older age (45.85 vs. 44.21 years; p=0.004), male (45.01% vs. 36.29%; p = 0.001), married (64.34% vs. 58.53%; p=0.027), less likely to take an oral medication (63.11% vs. 69.24%; p = 0.016), and had a lower BMI (34.71 vs. 36.27 kg/m²; p=0.001). A positive linear relationship with income was identified. A backward stepwise multivariate model found married men (OR 1.52; p = 0.003) were more likely to achieve HbA1c control while not currently working (OR 0.514; p = 0.031), not having insurance (OR 0.706; p = 0.009), and those using multiple daily injections (OR 0.455;p=0.016) were less likely to achieve HbA1c control.

Conclusion:

This study identifies factors predicting glycemic control. Understanding these factors may provide direction on best practices to help patients improve and achieve treatment goals.

Unique Metabolomics profiles in adolescent Type 1 diabetes

Yogish Kudva, MD; Heather D Wadams, MD; Tumpa Dutta, PhD; Vikash Dadlani, MBBS; Shelly McCrady-Spitzer, MS; Dhananjay Sakrikar, PhD; Aida Lteif, MD; Ravinder J. Singh, PhD

Mayo Clinic

Rochester MN, United States

kudva.yogish@mayo.edu

Objective:

Children with Type 1 diabetes (T1D) may experience significant hyperglycemia and glucose variability. Improvement in glycemic control reduces, but does not eliminate, the risk of microvascular complications and cardiovascular diseases in T1D. Metabolomics and the understanding of metabolic pathways may lead to identification of new metabolites and development of novel therapies that may target these pathways.

Methods:

We performed untargeted metabolic profiling of plasma using ultra performance liquid chromatography coupled with Quadruple time of flight mass spectrometry (UPLC-QToF-MS) to study the metabolomics perturbations in T1D adolescents and matched healthy controls (HC). Samples were collected after a period of 1 week during which equally physically active T1D and HC were provided calories to match their energy needs.

Results:

Study participants were matched for age, sex, BMI, and pubertal stage (nfor T1D, HbA1c 8.6 ± 0.7%, age 15.8 ± 1.3 yrs, n for HC, HbA1c 5.0 ± 0.2%, age 15.7 ± 2.0 yrs). The principle component analysis demonstrated distinct group separation based on their inherent metabolic differences. We identified 59 known metabolites involved in many pathways such as amino acids and tricarboxylic acid (TCA) cycle that were significantly altered from HC. Our T1D adolescents showed down regulations in creatinine, creatine, L-pipecolic acid, malate and L-fucose 1-phosphate while a-D-glucose pyruvate and butyrate were increased. 1,5-Anhydrosorbital was decreased in T1D as expected. Two fatty acids (FA) were significantly different from HC; hydroxy FA -Mevalonate-P was decreased and unsaturated FA -2-hendecenoic acid was increased. 1Z- alkenylglycerophosphoethanolamine and monoacylglycerophosphoethanolamine were decreased in T1D.

Conclusion:

The study identified 127 unknown metabolites (p < 0.05) that could be of potential importance to understand complications in T1D. Our results differ from previous studies because of younger age or uniformly elevated HbA1c or both.

Minimal Blood Loss with Automated Glucose Clamps: A new configuration of ClampArt

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

Profil

Neuss, NRW, Germany

mareike.kuhlenkoetter@profil.com

Objective:

Automated glucose clamps are the gold stand for the assessment of the pharmacodynamic (PD) properties of anti-diabetic agents. However, the continuous measurement of blood glucose concentrations (BG) for minute-by-minute adaptions of glucose infusion rates (GIR) is associated with substantial blood loss -especially for long-lasting clamps. We, therefore, implemented a minimal blood loss (MBL) configuration in ClampArt, Profil's modern clamp device, to reduce blood loss to only 1 ml/hour.

Methods:

As the MBL configuration leads to a greater time delay between blood sampling and BG measurements (total delay time approximately 3 minutes), the algorithm settings for the GIR-calculations were optimized in-silico using numerical simulations. Glucose measurements and clamp quality were validated in in-vitro experiments simulating the effects of GIR and insulin infusions in an artificial patient. Finally, the MBL configuration was tested in-vivo with euglycemic clamps assessing the PD effect of 0.6 U/kg of insulin glargine in healthy subjects. Clamp quality in the MBL-clamps was compared with that of 302 previous clamps with the standard configuration.

Results:

Technical downtime in MBL clamps (< 2.5%) compared favorably to the mean downtime of 4.6% in the standard configuration. Deviation from clamp target was -0.3 ± 0.1 mg/dl (mean ± SD) in MBL-clamps and was similar to the standard configuration (-0.2 ± 0.2 mg/dl). Precision (coefficient of BG variation) was comparable between the two configurations (MBL: 3.1 ± 0.6%; standard configuration: 3.8 ± 1.3%).

Conclusion:

The new MBL configuration reduces the blood loss of automated glucose clamps with ClampArt to only 1 ml/hour without compromising clamp quality. This option is particularly attractive for long-lasting clamps, e.g. for the PD evaluation of long-acting basal insulins.

Rapid point of Care Insulin Sensor

Jeffrey T. La Belle,PhD; David Probst, MS; Chi Lin, PhD; Aldin Malkoc, MS; Connor Beck, BS; Mukund Kahnwalker, BS; Curtiss Cook, MD

Biomedical Engineering at Arizona State University

Tempe, AZ, United States

jeffrey.labelle@asu.edu

Objective:

The capability of a patient to measure their own insulin concentration in real time could better inform decisions about subsequent dosing. We provide preliminary data from a rapid, point of care, insulin sensor capable of direct, label free detection of human insulin.

Method:

The point of care platform uses electrochemical impedance spectroscopy (EIS) technology. Using both real and imaginary impedance as separate entities, a calibration curve was derived with greater sensitivity and less variance. The method utilized to accomplish the objective ran EIS from 1 Hz to 100K Hz on a gold disk electrode with insulin specific antibody immobilized to a gold surface. The immobilization process was accomplished by binding insulin antibody via the carboxylic bond formed between the self-assembling monolayer complex (16-MHDA/EDC/NHS) to the surface.

Result:

Using human recombinant insulin, the team conducted preliminary experiments on the insulin antigen using the method described above. The initial data showed that insulin is detectable optimally from the range of 966 Hz up to 3125 Hz showing a high slope and strong correlation in both imaginary and complex spectra.

Conclusion:

The preliminary data for insulin detection using EIS has shown positive results. The goal is to use this method to detect insulin in a rapid, label free, point of care system. Together with glucose monitoring, insulin detection should “close the loop” on the diabetes treatment algorithms giving better patient care, and a more individualized blueprint to manage patients on insulin therapy.

Personalization of a Nonlinear Glucose-insulin system via a Markov Chain Monte Carlo algorithm for Model Predictive Control purposes

Sylvain Lachal, MSc; Celine Franco, PhD; Maeva Doron, PhD; Erik Huneker, MSc; Sylvia Franc, MD; Guillaume Charpentier, MD; Pierre Jallon, PhD

University of Grenoble Alpes - CEA Leti

Grenoble, Rhone Alpes, France.

sylvain.lachal@cea.fr

Objective:

Model Predictive Control is a widespread control design approach particularly suitable for long time delay systems such as glucose-insulin. Such control design techniques strongly rely on a model that needs to be accurate enough to predict patient glycaemia several hours later. To address the issue of model identification, a Bayesian approach has been developed and is compared to a conventional global nonlinear optimization approach.

Method:

The model used is the one developed by Hovorka et al. (2002) that is a high order, complex nonlinear model containing 16 parameters, from which only 7 are estimated. The algorithm has been evaluated on 25 real patients, over an 8-hour period containing at least one meal and one bolus. The identification method used is a custom MCMC based algorithm called Metropolis within Gibbs that aims at finding the Expectation A Posteriori (EAP) estimator for model parameters. These choices were made to ensure convergence. Mean Square Error (MSE) was used as a performance index for comparison with the benchmark nonlinear optimization method.

Result:

Better reconstruction of real data was obtained using the Bayesian approach (MSE = 0.59) instead of the conventional one (MSE = 0.89), leading to a MSE reduction of 33% on average (p<0.01).

Conclusion:

These results support the use of Bayesian methods to further improve model personalization. Furthermore, the obtained parameter distributions could also be used as prior information for subsequent model updates.

Glycemic Efficacy and Insulin Requirements when Administering U-100 Regular Human Insulin with V-Go® in Patients with Type 2 Diabetes

Rosemarie Lajara, MD; Leo Jeng, MD; Carla Nikkel, BS; Tracy Morris, PhD

Diabetes America

Plano, Texas, United States

rlajara@Diabetesamerica.com

Objective:

While insulin analogs are often associated with improved quality of life and reductions in severe hypoglycemia, they have substantially higher pharmacy costs with debated benefit over their predecessors. V-Go is a disposable insulin delivery device that delivers basal-bolus insulin therapy and is currently cleared for use with U-100 fast acting insulin (i.e., insulin lispro and insulin aspart). The objective is to evaluate the administration of U-100 regular human insulin (RHI) with V-Go in a real-world setting. Primary endpoints were change in A1C and change in insulin total daily dose (TDD) from baseline.

Method:

A retrospective analysis was conducted using data from electronic records to evaluate the delivery of RHI with V- Go. Hierarchical linear models were developed for statistical evaluations.

Result:

Ten patients with T2DM (mean age 65 y; duration of DM 15 y; A1C 9.3%; weight 96 kg) were evaluated of which 9 were using insulin (mean TDD: 99 units/day) at baseline. V-Go was initiated in 6 patients using RHI and 4 transitioned to RHI following the use of rapid acting insulin (RAI) first in V-Go. Model adjusted A1C and TDD decreased significantly from baseline (p<0.001), with a mean change of -1.8% in A1C and -46 units/day in TDD following a mean of 194 + 159 days on V-Go therapy.

Conclusion:

Currently V-Go is recommended for the delivery of U-100 RAI. This real-world analysis suggests the feasibility of additional uses for V-Go and the potential for a reduction in the costs of insulin regimen. Regardless of initiating V-Go therapy with RHI or transitioning to RHI from RAI, significant reductions in both A1C and insulin were observed. Administering RHI with V-Go provides an effective alternative for basal-bolus insulin therapy.

electronic Diabetes Artificial Pancreas Training (eDAPT) - Development and User Study

Nathan Lau, PhD; Elaine Schertz, BS; Jessica Robic, BS; Charlotte Barnett, BA; Christian Wakeman, BS; Siddarth Ponnala, BS; Sue Brown, MD

Grado Department of Industrial and Systems Engineering, Virginia Tech

Blacksburg, VA, United States

nathan.lau@vt.edu

Objective:

The objective is to develop a training program - eDAPT - that educates diabetes patients and medical professionals on the basic concept and operations of the Diabetes Assistant (DiAs) component of the Artificial Pancreas (AP) before in-person training. By educating AP users on the basics, eDAPT enables professional trainers to focus their interaction time with users on more complex operations and the potential misunderstandings of the system. Further, eDAPT can educate the public about AP technology.

Method:

The eDAPT consisted of 11 training modules; each module contained a video and a knowledge test on a specific topic of the DiAs. eDAPT was implemented in an open-source virtual learning environment maintained by the University of Virginia. In a two-part study, 18 participants were recruited for eDAPT evaluation. For the first part of the study, which was a two-hour session using eDAPT, the participants received an introduction and a pre-training knowledge test on the AP before receiving any training. Afterwards, participants trained themselves on the DiAs by going through the videos and knowledge tests in the 11 modules of eDAPT. Upon completion of the training, the participants completed a post-training knowledge test and an exit survey. The second part of the study was a focus group in which participants expressed their opinions on their experience with eDAPT.

Result:

Post-training knowledge test scores (M=6.64, SD=2.58) are significantly better than pre-training test scores (M=12.63; SD=1.63; t(10)=9.08, p < .001). The exit survey indicates that the participants generally had a positive experience with eDAPT (M=1.92, 1=best and 5=worst; SD=0.46).

Conclusion:

The results indicate that eDAPT can effectively train users on the basic concept and operations of the DiAs, enabling professional trainers to target teaching on more complex operations.

Evaluation of Night Time Glucose and Insulin During a Pivotal Hybrid Closed Loop Trial in Type 1 Diabetes

Scott W. Lee, MD; Timothy S. Bailey, MD; Richard M. Bergenstal, MD; Bruce W. Bode, MD; Ronald L. Brazg, MD; Trang Thao Ly, MD; Jennifer L. Sherr, MD, PhD; Jacob Ilany, MD; Robert H. Slover, MD; John Shin, PhD, MBA

Medtronic PLC

Northridge, California, United States

scott.w.lee@medtronic.com

Objective:

A hybrid closed-loop (HCL) insulin delivery system, consisting of the MiniMed 670G pump, 4th-generation sensors and transmitters, and a control algorithm, was evaluated in a pivotal study in Type 1 diabetes. Mean sensor glucose (SG), percent SG (%SG) below and within target (71-180mg/dL), and insulin delivery at night (00:00hrs-08:00hrs or 12:00 AM- 8:00 AM) were examined.

Method:

After a 2-week run-in period, 124 subjects (n=30, 14-21 years; n=94, 21-75 years) used the HCL system for a 3-month study period. Subjects were in closed-loop for 80.6 ± 14.2% of the time at night, during the study period.

Result:

For subjects < 21years, during run-in, mean night time SG was 151.8 ±28.6mg/dL, %SG in target 65.9 ± 15.6, <70 mg/dL 6.3 ± 6.3, <50 mg/dL 1.1 ± 1.5, insulin dosage 8.9 U/night, time with no insulin delivery 0 minutes. During the study period, mean night time SG was 148.9 ± 15.0 mg/dL, %SG in target 74.8 ± 10.7, <70 mg/dL 2.8 ± 1.8, <50 mg/dL 0.6 ± 0.7, insulin dosage 9.3 U/night, and time with no insulin delivery 84.0 minutes (p<0.0001, versus run-in). For subjects >21 years, during run-in, mean night time SG was 144.6 ±24.1mg/dL, %SG in target 68.5 ± 13.4, <70 mg/dL 7.0 ± 5.8, <50 mg/dL 1.1 ± 1.6, insulin dosage 7.9 U/night, time with no insulin delivery 3.6 minutes. During the study period, mean night time SG was 142.1 ± 12.9 mg/dL, %SG in target 80.0 ± 9.1, <70 mg/dL 3.0 ± 2.4, <50 mg/dL 0.6 ±0.8, insulin dosage 8.0 U/night, and time with no insulin delivery 98.3 minutes (p<0.0001, versus run-in).

Conclusion:

Improvements in overall glucose profiles were previously reported. This analysis in both age groups confirms that the HCL system improves time in target and reduces nighttime hypoglycemia. This was likely due to algorithm-appropriate cessation of insulin delivery accomplished by a small increase in insulin dosage, suggesting HCL appropriately adjusts the timing of insulin delivery at night.

Blood Gas Analyzer Glucose Measurement Accuracy

Yafen Liang, MD; Jonathan Wanderer, MD, MPhil; James H. Nichols, PhD; David Klonoff, MD, FACP, FRCP (Edin), Fellow AIMBE; Mark J. Rice, MD

Department of Anesthesiology, Vanderbilt University Medical Center

Nashville, TN, United States

yafen.liang@vanderbilt.edu

Objective:

The objective is to retrospectively investigate glucose accuracy with a blood gas analyzer (BGA) compared to a central lab device (CLD).

Method:

Glucose measurements between June 2007 and March 2016 at Vanderbilt Medical Center were reviewed. The agreement between CLD (Abbott Architect C8000 and C16000 analyzers) and BGA (IL-GEM 4000 analyzer) were assessed using Bland-Altman (BA), Consensus Error Grid (CEG), and Surveillance Error Grid (SEG) analysis. We further analyzed the BGA performance against the FDA 2014 draft guidance and ISO 15197-2013 standards.

Result:

A total of 2,671 paired glucose measurements, including 50 pairs of hypoglycemic values, were analyzed. BA analysis showed a mean bias of -3.1 mg/dL with 98.1% paired values meeting the 95% limits of agreement (LOA). For CEG, 99.7% of the paired values fell within the no risk zone. For SEG analysis, 96.6% BGA values were within the no risk zone, 2.4% were within the slight risk, 0.08% were within the moderate risk zone, and no values were within the great or extreme risk zones. For FDA guidance, 90.5% of data pairs were +/-10% of the reference for glucose > 70 mg/dL and 90.0% of data pairs were within +/- 7 mg/dL for glucose < 70 mg/dL. For ISO comparison, 96.2% of the values > 100 mg/dL were within the LOA and 97.8% of the values < 100 mg/dL were within the LOA.

Conclusion:

We demonstrate that the glucose measurement agreement between a commonly used BGA compared to a laboratory reference method meets the 2013 ISO 15197 standard. However, BGA accuracy did not meet the stricter requirements of recent FDA guidance. Fortunately, plotting these results on either the CEG or the SEG revealed no worrisome values in either the great or extreme risk zones.

Update of Clinical Experience with a Long Term Fully-Implanted Continuous Glucose Monitoring System

Joseph Y. Lucisano, PhD

GlySens Incorporated

San Diego, California, United States

joelucisano@glysens.com

Objective:

The second generation long term fully-implanted (no skin-attached elements) continuous glucose monitoring system (the GlySens Eclipse™ ICGM® System) included the launch of same-pocket re-implantations of new sensors in six adult human subjects following completion of a 12-month initial sensor implant period.

Method:

At the end of an initial 12-month subcutaneous implant period, the GlySens Model 100 ICGM Sensor in each of six human subjects was exchanged for a new replacement sensor in a minor outpatient surgical procedure utilizing local anesthesia. Following sensor replacement, subjects were to self-monitor blood glucose four times per day and meter-stored fingerstick values were to be downloaded during monthly clinic visits that also included meal-based glucose excursions with YSI plasma glucose measurements and, in some cases, Dexcom G4 CGM recordings. Monthly subject interviews including a standardized survey questionnaire were conducted to assess the tolerability of the device.

Result:

There were no significant adverse events associated with the sensor replacements. All sensors were easily extracted with no significant adherent capsular tissue. Early performance measurements of the replacement sensors suggest feasibility of the approach for sensor replacement/renewal as part of an extended long term monitoring regimen.

Conclusion:

Use of the fully implanted ICGM Sensor requires an annual user decision regarding sensor implantation (or reimplantation) and recalibration. No maintenance of body-worn components or other regular user intervention is required to receive glucose readings. This combination of features offers minimal barriers for adherence to treatment modalities requiring continuous glucose monitoring.

PAQ® - 3 Month Observation Study in Adults with Type 2 Diabetes

Julia K Mader, MD; Leslie C Lilly, RN; Felix Aberer, MD; Tina Pottler; Sebastian Becvar; Christian Lanz; Michael Trautmann, MD

Medical University of Graz

Graz, Styria, Austria

julia.mader@medunigraz.at

Objective:

PAQ® (CeQur SA) is a simple 3-day continuous subcutaneous insulin delivery device that provides set basal rates and bolus insulin on demand. In this single-arm study, Type 2 diabetes (T2D) patients (A1c > 7.0 -< 11.0%) on > 2 insulin injections/day were enrolled to assess the performance and safety of PAQ®.

Method:

The study comprised 3 periods: baseline (1 week current insulin therapy), transition to PAQ® (1-2 weeks), and PAQ® treatment (12 weeks). Insulin dose adjustments were made to safely achieve glycemic targets. Performance was assessed by change in: A1c, venous fasting blood glucose (VFBG), 7-point self-monitored blood glucose (SMBG), total daily insulin dose (TDD) and body weight after 12 weeks of PAQ®. Safety endpoints included hypoglycemia (BG < 70 mg/dL) and adverse device effects.

Result:

Twenty adults (age 63 ± 7 y, 15% female, BMI 32.2 ± 3.7 kg/m², diabetes duration 15 ± 7 years, A1c 8.6 ± 1.1%) were enrolled and 17 completed (2 terminated early for personal reasons, 1 due to protocol violation). Transition to PAQ® with the first basal rate selected occurred in 80% of patients. After 12 weeks of PAQ®, A1c was reduced by 1.4 ± 0.9% (p < 0.0001), VFBG by 30 ± 54 mg/dL (p=0.03) and all SMBG 7-point values were significantly reduced from baseline values (p < 0.03). Compared to baseline, TDD increased by 14.3 U (p < 0.01) while body weight was stable. Five patients had mild to moderate catheter site reactions and 1 patient had a mild skin irritation. No patient experienced severe hypoglycemia.

Conclusion:

Patients were safely transitioned from multiple insulin injections to PAQ®. Continuous basal infusion and effective delivery of meal time dosing with PAQ® resulted in significantly improved glycemic control without severe hypoglycemia.

Performance Comparison of the GLUCOCARD® Shine and Accu-Chek® Aviva to the ISO 15197:2013 Accuracy Criteria

Danielle Maher, BS; Julie Walker, RN, BSN, PHN; Patricia Gill, BA, MLT; John Gleisner, BS, PhD

ARKRAY USA, Inc.

Minneapolis, Minnesota, United States

maherd@arkmyusa.com

Objective:

This study compared the performance of the GLUCOCARD® Shine to the Accu-Chek® Aviva against the accuracy boundaries of ISO 15197:2013. This standard requires that 95% of BGM results be within ±15 mg/dL of the reference analyzer at glucose concentrations <100 mg/dL and within ±15% of the reference analyzer at glucose concentrations >100 mg/dL. Furthermore, 99% of all results are required to be within the A and B zones of the Consensus Error Grid.

Method:

Three lots of test strips were evaluated for performance for each BGMS at ARKRAY Inc. factory. All testing was conducted under the same IRB approved protocol and used the same group of participants. To further reduce variables, samples were drawn directly from the fingertip of confirmed diabetics (n=104) by laboratory professionals. Reference values were obtained using the YSI Model 2300 Analyzer. The data were evaluated against the accuracy boundaries of the ISO 15197:2013 Standard and Consensus Error Grid.

Result:

For GLUCOCARD® Shine, 100.0% of the glucose <100 mg/dL samples results (6/6) were within ±15 mg/dL and 99.0% of the glucose >100 mg/dL samples (97/98) fell within ±15% of the reference YSI. Overall bias was -1.3% and the correlation coefficient was (r) =0.99. For the Accu-Chek® Aviva, 100.0% of the <100 mg/dL samples (5/5) gave values that were within ±15 mg/dL and 99.0% of the >100 mg/dL samples (98/99) fell within ±15% of the reference. Overall bias was -1.3% and the correlation coefficient was (r) = 0.99. All data for both BGMS were within the A and B zones of the Consensus Error Grid.

Conclusion:

The GLUCOCARD® Shine and the Accu-Chek® Aviva had equivalent performance when assessed against the ISO 15197:2013 accuracy boundaries.

Nocturnal Blood Glucose Control by Fuzzy Logic (FL) Dosing Algorithm

Richard Mauseth MD; Donald Matheson MS; Robert Kircher MS

Dose Safety, Inc.

Seattle, WA, United States

Rick@Dosesafety.com

Objective:

The objective is to compare the nocturnal glucose control of our closed loop fuzzy logic (FL) dosing algorithm with the subject’s own care at home.

Method:

We reviewed data from several of our JDRF and NIH funded studies to compare overnight results from 10 pm to 6am to patients at home results. We collected data for 48 hours prior to CRC admission and this was compared to the same time period in the CRC. A total of 17 subjects had preadmission and overnight data that met study criteria . Statistics were calculated from Dexcom G4 CGMS data.

Result:

The percent time < 70 mg/dL at home was 6.4% versus <1.0% for the FL algorithm; the percent time in 70-180 mg/dL was 58.5% at home versus 85% for the FL algorithm, and the average glucose value was 157 mg/dL at home versus 145 mg/dL for the FL algorithm. The FL dosing algorithm produced no blood glucose values less than 60 mg/dL and no treatment was required to avoid hypoglycemia.

Conclusion:

The Dose Safety FL dosing algorithm produced superior overnight control when compared to the participants' own care.

Diabetes Related Retinopathy Screening by Diabetes Educators: A Pilot Study

John E. McDonald, OD

Teachers College, Columbia University

New York, New York, United States

jem2249@tc.columbia.edu

Objective:

This study reviewed the personal perspectives of diabetes educators after participating in a certification process to grade images of diabetes-related retinopathy (DRR). A qualitative narrative and quantitative questionnaire were used to present and discuss the feedback, views, and opinions of diabetes educators about integrating DRR screening within the practice of diabetes education.

Method:

Five volunteer diabetes educators who were current students or graduates of the Diabetes Education and Management Master’s program at Teachers College, Columbia University agreed to participate in the project, which consisted of two phases. In the first phase, the participant educators enrolled in an online, asynchronous DRR consultant certification program offered by EyePACS LLC. In the second phase, participant educators answered an online Likert type survey (DRR survey) and were individually interviewed via phone during the project in addition to the exit interview.

Result:

Participants indicated strong agreement to increasing DRR education, agreed that DRR screening could be clinically helpful to patients, were in slight agreement that DRR screening is within a diabetes educator’s scope of practice, and indicated neutral to slight agreement about institutional acceptance of DRR screening and cost sustainability. Participants expressed strong agreement that the certification process increased their knowledge about DRR, felt certification was very challenging, expressed dissatisfaction with the current certification process and indicated neutral to slight agreement about recommending the certification process in its current format to other diabetes educators.

Conclusion:

Further research and development of a DRR certification process for diabetes educators could increase patient access to DRR screening, mitigate increasing demands for health care resources and could potentially reduce the devastating complications of vision loss.

Realization of BGM Within ±10% Accuracy Based on Innovative Optical Transmission Absorbance System

Takeyuki Moriuchi, PhD; Hiroya Satou, MS; Yusuke Komata, MS; Ryokei Aikawa, MS; Koji Sode, PhD

Terumo Corporation

Nakakoma-gun, Yamanashi, Japan

takeyuki_moriuchi@terumo.co.jp

Objective:

Since blood glucose monitoring (BGM) is essential in diabetes care and management, further improvement in the accuracy of BGM is mandatory to achieve better glycemic control for diabetics. In this study, we report on an innovative single optical transmission absorbance system to simultaneously measure glucose levels and hematocrit. In addition, a simple and inexpensive test strip was developed to measure the blood glucose levels of whole blood accurately and rapidly.

Method:

To realize the high-accuracy measurement of blood glucose levels, we have developed a highly sensitive reagent, a brand new enzyme, an original absorption dye, and an accurate hematocrit detection technology. We employed a new generation GDH (FAD) enzyme that shows a high catalytic efficiency, with a low Km value, to measure reaction endpoints. Together with the original high absorption dye, the optical transmission absorbance principle achieved an increase in the resolution of low glucose levels. Additionally, high-accuracy hematocrit compensation was achieved using multi-wavelength detection.

Result:

Accuracy of our new BGMs was evaluated with in-house blood samples, adjusted to three blood glucose levels (0 mg/dL, 100 mg/dL, and 400 mg/dL) with three hematocrits (Hct 20%, Hct 40%, and Hct 60%). From the measurement results of N=135, accuracy for 2SD was ±3 mg/dL (for glucose: 0 mg/dL), ± 5.7 mg/dL (for glucose: 100 mg/dL), and ± 5.9% (for glucose: 400mg / dL) versus the standard system.

Conclusion:

We have developed a novel BGMs based on an innovative optical transmission absorbance system and achieved accuracy < ±10% (± 5.9%).

Accuracy and Utility of the GlucoScout Glucose Analyzer in Clinical Research

Linda Morrow, MD; Lindsey Hergesheimer, BSc; Christian Weyer, MD; Marcus Hompesch, MD

Profil Institute for Clinical Research

Chula Vista, CA, United States

linda.morrow@profilinstitute.com

Objective:

The objective is to assess the accuracy of a FDA 510k reviewed intravenous blood glucose (BG) analyzer.

Method:

In a retrospective analysis of 14,926 paired data points from 108 healthy, non-diabetic, male subjects (aged 28.56 ± 6.5 years), we assessed BG results from the GlucoScout relative to the gold standard Yellow Spring Instrument (YSI) as a reference method. Intravenous BG measurements were performed during a 24-hour euglycemic clamp procedure (target BG level 80 mg/dl). Clinical accuracy was assessed using the Parke’s Error Grid Analysis (EGA, T1D version). Analytical accuracy was assessed by calculating the mean absolute relative difference (MARD) between GlucoScout and YSI measurements, precision absolute relative difference (PARD), and the Bland Altman method.

Result:

In the Parke’s-EGA, 99.9% of data fell into Zone A, and the remaining 0.1% into Zone B. There were no data in Zones C, D, or E. BG results averaged 83.2 ± 6.1 mg/dl for all GlucoScout samples and 79.4 ± 6.2 mg/dl for all YSI samples (mean ± SD), with resulting MARD and PARD values of 5.5% and 5.3%, respectively. The mean difference (bias) between measuring devices was 3.8 mg/dl with a SD of the difference of 3.6 mg/dl and 95% limits of agreement of -3.2 to 10.8 mg/dl.

Conclusion:

These results indicate that the GlucoScout is highly accurate within the euglycemic range. Its advantages over the YSI, and other glucose analyzers in clinical research settings, are that it can measure venous blood glucose automatically, without blood loss, as frequently as every 5 minutes, with auto calibration after each measurement.

Amperometric Glucose Sensors’ Background Current is a Confounding Factor of Plasma-interstitium Relationship Studies During Hypoglycaemia

Vanessa Moscardo, MSc; Paolo Rossetti, MD, PhD; F. Javier Ampudia-Blasco, MD, PhD; Jorge Bondia, PhD

Polytechnic University of Valencia

Valencia, Spain

vamosgar@gmail.com

Objective:

Glucose sensors measure glucose concentration in the remote interstitial fluid compartment instead of plasma. Several studies have analyzed if hypoglycaemia influences glucose transport between the plasma and interstitial fluid compartments by means of amperometric glucose sensors. However, these sensors produce a background current (BC) in absence of glucose that could negatively influence the signal-to-noise ratio for low glucose concentrations. The aim of this work is to better understand the influence of BC in glucose measurements during hypoglycaemia.

Method:

Glucose concentrations of fourteen subjects with T1DM were measured with two different sensors (Medtronic, Northridge, CA) during two eu-hypoglycaemic clamp studies with different levels of insulinemia (0.3 mU/Kg/min vs. 1 mU/Kg/min). The raw current signal (nA) of both sensors were used to estimate glucose concentration and sensor sensitivity at each clamp phase (euglycaemic feedback, euglycaemia, hypoglycaemia and hypoglycaemia recovery) by fitting a first-order glucose transport model considering that: (1) there is no BC and (2) there is BC (“one-point” vs. “two- point” calibration). Paired Wilcoxon Signed-Rank test was used for comparison.

Result:

Method 1 showed no statistically significant differences between phases (p > 0.05). In contrast, results obtained with Method 2 showed that feedback and euglycaemic phases were different from hypoglycaemic phases in both insulinemias (p < 0.05) and that that differences exist in hypoglycaemic recovery phase in low insulin studies (p =0.00005).

Conclusion:

The magnitude of the sensors background current had an important influence in our sensitivity estimations during hypoglycaemia. Thus, glucose estimations could be altered, depending on the method used to calibrate the sensors during hypoglycaemia, and this alteration represents a confounding factor of the study’s conclusions.

Relationship Between Glycemic Control Using eGMS and Readmission Rates in Cardiovascular Patients Hospitalized with AMI, CHF or Undergoing CABG During the Implementation of a System Wide Glycemic Initiative

April Mumpower, BSHI; Tamera Parsons, CPHQ

Mountain States Health Alliance

Johnson City, TN, United States

mumpoweraj@msha.com

Objective:

The CMS Hospital Readmission Reduction (HRR) program was initiated in 2012 to encourage hospitals to implement quality improvement program to reduce the rate of readmissions through a hospital financial penalty. It has been shown that uncontrolled inpatient hyperglycemia leads to an increased readmission rate and further that proper management of hyperglycemia has reduced the rates of poor patient outcomes, including mortality, complications, longer length of stay and, importantly, readmissions. This study evaluated the readmission outcomes of patients using an Electronic Glycemic Management System (eGMS) Glucommander (GM) for inpatient insulin management versus Standard Care (SC) in patients with Acute Myocardial Infarction (AMI), Congestive Heart Failure (CHF) and those undergoing Coronary Artery Bypass Grafting (CABG) procedures.

Method:

This retrospective study evaluated 3,198 patients with AMI, CHF or undergoing CABG procedures who were admitted to a 13 hospital health system over a 12-month timeframe from January 2015 through December 2015. Qualifying patients were treated with eGMS IV, and/or SubQ insulin management, or with SC. The main outcome measure was: Risk Adjusted Readmission Rates.

Result:

Patients (n=281) treated with eGMS had a Risk Adjusted Readmission rate of 0.75 for AMI, 0.34 for CHF and 0.65 for CABG. Patients (n=2,917) treated with SC had a Risk Adjusted Readmission rate of 1.17 for AMI, 0.97 for CHF and 2.04 for CABG. The p values from the t-tests of the observed cardiac patient groups (p=0.019 SD 2.990, 2.618) determined statistical 95% significance p< 0.005, particularly for CHF (p=0.009 SD 1.993, 0.905). Patients treated with eGMS had a Risk Adjusted reduction of 36% for AMI, 65% for CHF, and 68% for CABG.

Conclusion:

The evidence presented suggests that eGMS can effectively reduce the rate of readmission for patients with cardiovascular disease who are in need of insulin management. There is sufficient evidence supporting a decrease in Risk-Adjusted readmission rates among groups treated with eGMS management.

Does Glycemic Control Using eGMS Reduce Readmission Rates for Hospitalized Patients Undergoing CABG?

April Mumpower, BSHI; Tamera Parsons, CPHQ; Raymie McFarland, CSSGB, PT; Amy Henderson, RN, BSN

Mountain States Health Alliance

Johnson City, TN, United States

mumpoweraj@msha.com

Objective:

In 2016, the CMS Hospital Readmission Reduction (HRR) program will be adding Coronary Artery Bypass Grafting (CABG) to their list of admitting categories that will be measured and will attach a hospital financial penalty if targets are not reached. It has been shown that unrecognized and uncontrolled inpatient hyperglycemia in the cardiovascular disease population leads to an increased readmission rate and that proper management of hyperglycemia has shown trends toward reducing rates of readmissions. This study evaluated the readmission outcomes of patients using an Electronic Glycemic Management System (eGMS) Glucommander (GM) for inpatient insulin management versus insulin managed by Standard Care (SC) in patients undergoing Coronary Artery Bypass Grafting (CABG) procedures.

Method:

This retrospective study evaluated 448 patients undergoing CABG procedures who were admitted to a 525-bed teaching hospital over a 32-month timeframe from May 2013 through December 2015. Qualifying patient’s insulin regimens were managed to a glucose target of 100-140 mg/dL using the eGMS IV and SubQ program GM or clinician lead SC. The primary outcome measure was risk adjusted readmission rates over the 3 years.

Result:

Patients (n=267) treated with eGMS had a risk adjusted readmission rate of 0.53 (M=0.53, SD=2.560) and a readmission rate of 5.24% undergoing a CABG procedure. Patients (n=181) treated with SC had a risk adjusted readmission rate of 1.11 (M=1.11, SD=3.237) and a readmission rate of 12.71% for CABG. The eGMS had a significantly lower readmission rate compared to SC (t(376)=2.028, p = 0.043).

Conclusion:

These results suggest that eGMS can significantly reduce the rate of readmission for patients who are in need of insulin management while undergoing a CABG procedure compared to patients managed with Standard Care.

In Silico Modeling of the Effects of Tissue Microenvironment on Subcutaneous Insulin Absorption

Matthew T. Novak, PhD; Christopher Basciano PhD; Marcus Rademacher, MS; Patrick Downie, BS; Ronald J. Pettis, PhD

BD Technologies

Research Triangle Park, NC, United States

matthew_novak@bd.com

Objective:

Human pharmacokinetic/pharmacodynamic (PK/PD) models have been used extensively in the development of advanced insulin delivery systems such as the artificial pancreas. However, conventional time-dependent PK models are limited by offering little physiological context detailing the physical factors influencing insulin absorption. As a novel in silico complement to traditional in vivo PK/PD studies, we are developing a diffusion-reaction transport model that incorporates relevant physiological tissue processes to both accurately model subcutaneous insulin transport and investigate the role that the local tissue microenvironment has in affecting vascular absorption. This diffusion-based absorption model may be able to provide a mechanistic understanding around how the local tissue microenvironment impacts insulin uptake, potentially adding valuable knowledge for insulin delivery system design.

Method:

Insulin transport from a subcutaneously injected depot through the surrounding tissue was modeled as a two compartment, transient diffusion-reaction problem in one-dimensional spherical coordinates with terms to account for vascular uptake and cellular degradation. Governing partial differential equations for insulin transport were discretized in both space and time into a coupled system of ordinary differential equations and solved in MATLAB (The MathW orks, Natick, MA). The model output (i.e., plasma insulin concentration) was compared against experimental swine insulin PK data for accuracy. Computational sensitivity studies were conducted to investigate how changes in characteristic tissue properties would affect resultant insulin PK outputs (AUC, C_max, t_max, AUC₆₀).

Result:

Simulated plasma insulin levels accurately tracked experimental results (R² = 0.96), comparing favorably in fit to current state-of-the-art subcutaneous insulin PK models such as those from Wilinska, et al. Moreover, the model is able to determine spatial changes in tissue insulin levels over time. Insulin degradation terms were found to have a negligible influence on every PK metric. Both Cmax and AUC₆₀ were highly influenced (> 60% influence) by the bulk vascular permeability to insulin. Similarly, t_max was largely affected by bulk permeability (~45%). Diffusive effects were found to highly influence both AUC (~75%) and tmax (~45%).

Conclusion:

Here we present a new insulin transport model as a tool for analyzing how the physiology surrounding an insulin injection impacts its absorption. Results demonstrate the ability of the model to accurately track in vivo data and how the characteristics of the adjacent subcutaneous tissue space affect PK. This in silico approach complements traditional PK methods while facilitating mechanistic investigation into the role of tissue microenvironment on insulin delivery.

Diabetes Management Application Improves Self-Care Behavior and Glycemic Control

Reid Offringa, PhD; Tanushree Bose, PhD, MBA; Michael S. Greenfield, MD

Glooko Inc.

Mountain View, CA, United States

reid@glooko.com

Objective:

Glooko is a diabetes management mobile application that allows patients to download and visualize their blood glucose (BG) readings to their smartphone and annotate their BG readings with diet, physical activity, and medication information. Our objective was to determine whether the consistent use of Glooko increases the frequency of self-monitoring of blood glucose (SMBG) and improves glycemic control.

Method:

Glooko users who had at least 4 months of BG readings after their first sync were included in this analysis (n = 2,607). A three-month average before the first sync with Glooko served as baseline data. All changes over time were evaluated with a mixed-effects generalized linear model, considering the subject effect as a random effect and with change over time and diabetes type as fixed effects. The monthly count of BG readings was modeled with a quasipoisson distribution while estimated HA1c was modeled using a log-Gaussian distribution.

Result:

After syncing with Glooko, patients with diabetes significantly increased their frequency of SMBG by 27.3%. These patients also decreased their estimated HbAlc by 0.4% within the first month of Glooko use and sustained HbAlc improvement over 3 months.

Conclusion:

Patients with diabetes who consistently use diabetes management mobile application increase their frequency of SMBG. Improved self-care behavior, prompted by access to contextual blood glucose readings, might be responsible for the reduction in HbAlc levels. The mobile application allows patients to see the effect of diet, exercise and medication on their glycemic targets and modify their behavior and therapy.

In-Silico Performance of 670G Hybrid Closed Loop (HCL) with Cumulative Error in Meter BG and Sensor Measurements

Neha Parikh, PhD; Anirban Roy, PhD; Benyamin Grosman, PhD; Di Wu, PhD; Andrea Varsavsky, PhD; Cesar Palerm, PhD; Lou Lintereur, MS; Weydt Patrick, MBA; Rebecca Gottlieb, PhD

Medtronic Diabetes

Northridge, CA, United States

neha.j.parikh@medtronic.com

Objective:

Calibration with erroneous meter blood glucose (MBG) measurements, causing positively or negatively biased sensor readings, can result in over or under-delivery of insulin by a closed-loop system. In-silico modeling of the 670G HCL system with improbable scenarios (probability <1/1,000,000,000) of cumulative MBG and sensor bias was performed to evaluate safety. Meter BG and sensor readings were persistently biased at ± 25% (greater than EN ISO 15195:2015 limits) and ± 35% (system limits) respectively.

Method:

In-silico modeling was conducted using the University of Virginia/Padova T1DM simulator with 10 virtual adult subjects each for the positive and negative bias scenarios. Simulations were conducted for 32 hours (10:00 PM Day 1 until 6:00 AM Day 3) with breakfast, lunch and dinner on Day 2 at 7:00 AM, 12:00 PM and 6:00 PM respectively. Persistent positive and negative biases were simulated for both the MBG and sensor readings. MBG readings were persistently biased by ± 25% with respect to the YSI readings with additional persistent sensor bias of ± 35% with respect to the MBG readings. These errors are at system limits with greater errors expected to trigger safety features and an exit from closed loop delivery.

Results:

For positive bias simulations, mean ± SD of YSI (mg/dL), percentage time in range (70-180 mg/dL), % time <70 mg/dL were 108.7 ± 7.9 mg/dL, 95.9 ± 5.9%, 3.7 ± 5. >r negative bias simulations, mean ± SD of YSI (mg/dL), percentage time in range (70-180 mg/dL), % time >180 mg/dL, %time >300 mg/dL were 222.6 ± 23.8 mg/dL, 18 ± 12.4%, 82 ± 12.4%, 2.5 ± 3.3% respectively.

Conclusion:

Medtronic’s 670G HCL system demonstrated safety in improbable scenarios with no severe hypoglycemia <50 mg/dL and limited time in severe hyperglycemia >300 mg/dL.

Analysis of Real-Time Data Transmission for Insulin Dosing Decision Support and Remote Monitoring

Peter Pesl, PhD; Pau Herrero, PhD; Monika Reddy, PhD; NickOliver, FRCP; Desmond Johnston, FMedSci; Christofer Toumazou, FRS; PantelisGeorgiou, PhD

Imperial College London

London, United Kingdom

peterp@imperial.ac.uk

Objective:

Remote access to glucose related data (e.g. glucose measurements, meal insulin boluses, carbohydrates, exercise) is essential for the safety of insulin decision support systems with capability to automatically adapt insulin therapy in Type 1 diabetes (T1D). Decision support tools running on smartphones have the potential to send data through WiFi or mobile data connectivity to a secure web-server for approval of therapy changes or remote monitoring. However, WiFi unavailability or loss of connection to the mobile network can cause significant delays in data transmission. Therefore, we assessed data transmission time as part of a safety evaluation of the smartphone-based insulin advisory system - the 'Advanced Bolus Calculator for Diabetes '(ABC4D).

Method:

Ten people with T1D used ABC4D over a period of two months. The time difference between glucose events (n=1,629) entered into ABC4D and the time the event was received by the web-server was analysed. For half of the participants, WiFi (when available) was used to send data, while for the other half, in addition to WiFi, mobile data (i.e. 3G) was enabled.

Result:

In the cohort where both 3G and WiFi were used for data transmission, 84.6% of all data were delivered to the server within one minute after entry, 88.0% within the first hour, and 99.4% within one day. When relying on WiFi only, these numbers were 51.3%, 61.1% and 94.5% for <1 min, <1 hour and <24 hours, respectively.

Conclusion:

Mobile data connectivity (e.g. 3G), in addition to WiFi, is a reliable way to transmit glucose related data for real-time monitoring, while the time delay for sending data through WiFi alone is still acceptable for periodic revisions of insulin therapy.

Evaluation of the Non-Invasive Glucose Monitoring Device GlucoTrack in Patients with Type 2 Diabetes and Subjects with Pre-diabetes

Andreas Pfutzner, MD, PhD; Daniela Sachsenheimer, MD; Alexander Lier, MD; Filiz Demircik, PhD; Sanja Ramljak, PhD

Pfutzner Science & Health Institute

Mainz, RLP, Germany

andreas.pfuetzner@pfuetzner-mainz.com

Objective:

GlucoTrack® (Integrity Applications, Ashdod, Israel) is a CE-approved, non-invasive glucose monitoring device for use at home and indoor environments. The device measures three physiological conditions at the earlobe that are correlated with tissue glucose concentrations by employing ultrasound, electromagnetic, and thermal measurement technologies. GlucoTrack is intended for use in adult Type 2 diabetic patients and pre-diabetic patients. We performed this study to evaluate the performance of the device during a standardized meal test.

Method:

A total of 27 participants were enrolled into this prospective, open-label trial (N=20 Type 2 patients, 4 female, mean age: 68 ± 8 yrs (HbAlc: 7.2 ± 1.0%, BMI: 32.1 ± 4.7 kg/m²), N=7 pre-diabetic subjects, 2 women, (HbAlc: 5.8 ± 0.3%, BMI: 30.4 ±5 .9 kg/m2). After calibration of the device on the day before by using comparator values from a HemoCue meter, the patients ingested a standardized breakfast at the site during the next visit. Blood glucose was measured every 30 min over 180 min with HemoCue, AccuChek Performa, and the YSI Stat 2300. Mean absolute relative difference and a Consensus Error Grid analysis were performed against the YSI reference method.

Result:

For the Consensus Error Grid, 100% of the GlucoTrack results were within the clinically-accepted zones A and B (62.4 % and 37.6 %, respectively). Mean absolute relative difference of the GlucoTrack devices, when compared to YSI Stat2300, was found to be 19.7% (17.5 % vs. HemoCue). The performance was similar between patients and pre-diabetic subjects.

Conclusion:

The current data confirms the clinically-acceptable performance of the GlucoTrack device among its intended users, including pre-diabetic patients, for pain-free non-invasive monitoring of glucose levels.

Accuracy of Blood Glucose Monitoring Systems: New Options for Graphical Presentation

Stefan Pleus, MSc; Frank Flacke, PhD; Jochen Sieber, PhD; Cornelia Haug, MD; Guido Freckmann, MD

Institute for Diabetes-Technology Research and Development Corporation at the University of Ulm, Ulm, Germany

stefan.pleus@idt-ulm.de

Objective:

System accuracy evaluations of blood glucose monitoring systems (BGMS) commonly include graphical presentations such as traditional difference plots (DPs), in which differences between BGMS results and results from a comparison method are displayed. Recently, three new approaches were presented: radar plots (RPs), rectangle target plots (RTPs), and surveillance error grids (SEGs).

Method:

Data from four system accuracy evaluations (2x BGStar, 1x MyStar Extra, 1x MyStar DoseCoach; Agamatrix Inc., Salem, NH) were analyzed in RPs, RTPs, and SEGs. In RPs, individual data are plotted in a circular graph, with higher differences between BGMS and comparison results being farther from the center. RTPs show one rectangle per data set instead of individual data. The size and location of the rectangle allows visualization of the extent of random and systematic differences, respectively. SEGs also display individual data, and a risk score is attributed to each data point. Strengths and limitations of these new plots were established by comparison with DPs.

Result:

Plots with individual data points (DPs, RPs, SEGs) allow for more detailed assessment of data than RTPs, in which data are sorted into 1 of 2 categories (below or above 100 mg/dL) and then averaged. DPs and RPs become harder to read with increasing numbers of data points, whereas RTPs are not affected. RTPs also provide more easily accessible information about trueness and precision (location and size of the rectangle). SEGs have the advantage of having many data points and also provide for risk estimations.

Conclusion:

The four types of data plots have different strengths and limitations. The selection of a specific type depends mostly on the kind of information, e.g. assessment of lot-to-lot variability, being sought,

Pen Needle Design Influences Ease of Insertion, Pain, and Skin Trauma in Subjects with Type 2 Diabetes

Kezia Ann Postmark, PhD; Morten Lind Jensen, MD, PhD; Nils Berg Madsen, PhD; Jonas Kildegaard, PhD; Bente Merete Stallknecht, MD, PhD, DMSc

University of Copenhagen, Department of Biomedical Sciences, Copenhagen, Denmark & Novo Nordisk A/S, Device R&D, Hillerod, Denmark

kezj@novonordisk.com

Objective:

Pen needles used for subcutaneous injections have gradually become shorter, thinner, more thin walled, and therefore less robust for patient reuse. Thus, different needle sizes, alternative tip designs, and needles allowing for reuse were tested to explore how needle design influences ease of insertion, pain, and skin trauma.

Method:

Thirty subjects with injection-treated Type 2 diabetes and BMI 25-35 kg/m² were included in the single-blinded study. Each subject received abdominal insertions with 18 different types of needles. All needles were tested twice per subject and in random order. Penetration force (PF) through the skin, pain perception using a 100 mm visual analog scale (VAS), and change in skin blood perfusion (SBP) were quantified after the insertions.

Result:

Needle diameter positively related to PF, SBP (p < 0.05) and pain. Lack of needle lubrication and small “needle hooks” increased PF and SBP (p < 0.05) but did not affect pain. Short-tip, obtuse needle grinds affected PF and SBP, but pain was only significantly affected in extreme cases. PF using skin and polyurethane rubber were linearly related and pain outcome was dependent on increases of SBP.

Conclusion:

The shape and design of the needle and the needle tip affect ease of insertion, pain, and skin trauma. Relations are seen across different data acquisition methods and across species, enabling needle performance testing outside of clinical trials.

In vivo Evaluation of an Osmotic Pressure- based Implantable Glucose Sensor Technology

Sanja Ramljak, PhD; Bernd Lecher, MD; Jan Baumann, PhD; Anja Meier, MD; Eugene Van Wyk, MSc; Shoan Watson, BSc; Rune Frisvold, MBA; Andreas Pfutzner, MD, PhD

Pfutzner Science & Health Institute

Mainz, RLP, Germany

andreas.pfuetzner@pfuetzner-mainz.com

Objective:

A novel implantable glucose affinity biosensor (Sencell, LifeCare, Norway) is currently under development, that is based on competitive and reversible binding of glucose and the polysaccharide dextrane to the glucose specific lectin concavalin A (ConA).

Method:

The Sencell technology uses osmotic pressure differences arising between a reagent chamber (containing active fluid with ConA and dextrane) and a diffusion chamber (in direct contact with interstitial fluid) to determine interstitial glucose concentrations. Both compartments are separated by a nanoporous membrane permeable to glucose and water, but not to ConA or dextrane. A first study in pigs has initially demonstrated the capability of the technology to track interstitial glucose. It has also led to specific sensor modifications for signal noise reduction, e.g. the introduction of a second measurement chamber to measure mechanical pressure signals and deduct them from the signals. This device iteration was now evaluated in another animal experiment. Implantation of four modified Sencell sensors and one Dexcom® G4 control sensor was performed in the back and the neck area of three pigs (1 female, 2 male), respectively. After two days of equilibration and signal documentation, they received an intravenous glucose load of up to 100 mg/kg of dextrose. Reference measurements from capillary blood samples were performed using the YSI 2300 STAT Plus glucose analyzer every 15 min for 5 h.

Result:

Several Sencell prototype devices were able to track glucose changes paralleling the results of the Dexcom Sensor over extended time periods. In comparison to the previous results, deduction of movement artifacts, as captured by the reference pressure chambers, substantially reduced the noise level of the signals. The magnitude of the signals was in a predicted range and working sensors matched G4 results. Although the body temperatures in the animals within three experimental days partially exceeded 39°C, the activity of ConA was preserved. No external signs of inflammation were observed in the histology examinations.

Conclusion:

After recent design modifications, improved performance of the Sencell sensor technology has beendemonstrated in vivo.

Macrophage Accumulation Affects Sensor Accuracy in Humans

Mercedes Rigla, PhD, MD; Belen Pons, MsC; Pere Rebasa, PhD, MD; Alexis Luna, PhD, MD; Francisco Javier Pozo, MD; Assumpta Caixas, PhD, MD; Maria Villaplana, MD; David Subi'as, MD; Maria Rosa Bella, PhD, MD; Neus Combalia, PhD, MD

Endocrinology and Nutrition Department, Parc Tauli Sabadell University Hospital

Sabadell, Barcelona, Spain

mrigla@tauli.cat

Objective:

Subcutaneous glucose sensors have become a key component in Type 1 diabetes management. However, their use is limited by the impact of foreign body response to their duration, reliability and accuracy. The aim of our study is to give the first description of human acute and subacute subcutaneous response to sensors, describing the changes observed in sensors surface, and the inflammatory cells involved including their relationship to sensor performance.

Method:

Fourteen obese patients (7 DM) underwent two abdominal biopsies comprising the surrounding area were they had worn glucose sensors: the first one inserted 7 days before and the second one 24 hours before bariatric surgery procedure. Samples were processed and studied to describe tissue changes by optic microscopy by two independent pathologists (blinded to sensor duration) and transmission electron microscopy (TEM). Macrophage quantification (N/gm²) was studied by immunohistochemical (IHQ) methods in the area surrounding the sensor (CD 68, CD 163). Sensor surface changes were studied in a subgroup of cases by scanning electron microscopy. Seven-day CGM records were subdivided according to their accuracy (MARD < 10%: accurate; MARD > 10%: inaccurate).

Result:

Pathologists were able to correctly classify the sensor duration for all of the tissue samples. Acute response (24 h) was characterized by the presence of neutrophils whereas macrophages were the main cells involved in subacute inflammation. The number of macrophages around the insertion hole was higher for inaccurate sensors than for those performing accurately (32.6 ± 14 vs. 10.6 ± 1 cells/gm2;p <0.05)

Conclusion:

The accumulation of macrophages at the sensor-tissue interface appears to decrease the accuracy of sensor glucose measurements.

Comparative Analysis of Handling Steps Necessary When Using Insulin Pumps

Delia Rittmeyer, MSc; Antje Westhoff, BSc; Cornelia Haug, MD; Guido Freckmann, MD

Institute for Diabetes Technology Research and Development Corporation at the University of Ulm

Ulm, Germany

delia.rittmeyer@idt-ulm.de

Objective:

The ease of use of insulin pumps might contribute to the effectiveness of insulin pump therapy. Complex handling may be associated with potential errors when using the device. In previous investigations, handling steps required to perform usual tasks with different insulin pumps were evaluated. The present study investigated handling steps of the recently introduced insulin pump MiniMed® 640G including an update regarding the Accu-Chek® Insight system that is now available with a prefilled cartridge.

Method:

The MiniMed 640G insulin pump (Medtronic MiniMed) and its consumables were evaluated with regard to the number of handling steps required to perform different routine tasks (e.g. bolus delivery, basal rate setting, filling and changing of the insulin cartridge, replacing the battery, and pairing the pump with the blood glucose meter). The number of steps was compared to the steps determined for the Accu-Chek Insight system (Roche Diabetes Care) with the new prefilled cartridge and to the Accu-Chek Combo system (Roche Diabetes Care).

Result:

MiniMed 640G required more handling steps for seven of the evaluated tasks when compared either to the Accu-Chek Insight or the Accu-Chek Combo system (where 2 and 4 tasks, respectively, could be performed with a reduced number of steps). The availability of the prefilled cartridge for the Accu-Chek Insight system contributed to the reduction of associated handling steps.

Conclusion:

For most of the tasks that users have to perform frequently, the Accu-Chek Insight and the Accu-Chek Combo systems required fewer handling steps than the MiniMed 640G system. Simple and intuitive handling might reduce potential sources of error and thus increase compliance and patient safety.

Mobile Health Interventions for Diabetes: From Taxonomy to Product

Mansur Shomali, MD, CM; Matti Prasad Rao, MS; Malinda Peeples, MS, RN

WellDoc

Baltimore, Maryland, United States

mshomail@welldoc.com

Objective:

Mobile health products for chronic conditions such as diabetes can be thought of as digital therapeutics. Unlike drugs, however, their mechanisms of action have to do with user engagement with their disease in a way that improves selfmanagement and facilitates behavioral and lifestyle change. We characterized the various types of messaging interventions that a mobile health product can provide as a first step in understanding the mechanisms for behavioral change. In this analysis, we used the interventions of WellDoc’s BlueStar® app.

Method:

The first mobile prescription therapy for patients with Type 2 diabetes, BlueStar®, was designed to collect data such as blood glucose (BG), exercise, food choices, and sleep for coaching patients and providing decision support to providers. Blue Star messaging interventions were classified as one of five message types: prompts, touchpoints, real-time feedback, longitudinal, and summary messages.

Result:

We reviewed the behavioral interventions of all five types of messages. For example, the longitudinal message starts with an effect and then the user is asked to evaluate the cause: “Your last 2 post meal BGs have been high; was it something you ate?” On the other hand, the summary message starts from the cause and then goes to the effect: “On Tuesday, you recorded 75 grams of carbs at dinner, and your BG went up to 251 mg/dL.”

Conclusion:

In randomized controlled studies, the BlueStar app has been associated with significant reductions in A1C that cannot be explained by changes in pharmacotherapy alone. We propose that messaging interventions drive improvement in the user's diabetes self-management. In this qualitative analysis, the mechanisms of delivering insights to users about how behavior affects BG have become apparent.

A Review of Patient Factors Associated with Consistent CGM use in Patients with Diabetes

Madison B. Smith, RN, BSN; Anastasia Albanese-O’Neill, PhD, ARNP, CDE; Joseph M. Abbatematteo, PharmD; Debra Lyon, RN, PhD; Yao Yingwei, PhD; Gail Keenan, PhD, RN

University of Florida, College of Nursing, PhD Program

Gainesville, FL, United States

mbriokl6@ufl.edu

Objective:

The purpose of this systematic review is to summarize the state of science for the patient factors associated with consistent CGM use to inform training methods and best practices and to support adherence to CGM and improve glycemic control.

Method:

A literature search was conducted in MEDLINE, CINAHL, The Cochrane Library, and PsychInfo databases using PRISMA guidelines. Manuscripts included use quantitative or mixed methods study designs to report measures of satisfaction, usability, or quality of life (QOL) as an outcome measure related to CGM use or sensor-augmented pump therapy (SAPT) in patients with diabetes.

Result:

Twenty manuscripts published between 2010 and 2015 met inclusion criteria. Eleven studies were primarily designed to evaluate patient factors with the remaining 9 conducted as ancillary studies. Six primary studies received a “weak” quality assessment score, while 2 ancillary studies received a “weak” score. Standardized questionnaires were used exclusively in 11 studies, while the remaining 9 used non-standardized questionnaires. Patient factors associated with consistent CGM use are treatment satisfaction, barriers, self-efficacy, distress, and quality of life. Eight patient-reported barriers and reasons cited for inconsistent/discontinued use were identified: alarm fatigue, pain/discomfort, unmet expectations, too bulky/body image issues, skin/adhesion issues, interference with activities, technical issues, and cost. Studies that evaluated patient factors had less rigorous methodology and often lacked ability to prove correlation.

Conclusion:

To date, studies on patient factors associated with consistent CGM use have generally lacked sufficient methodological rigor to establish correlation. A more robust understanding of how identified patient factors influence consistent CGM use is necessary. Additionally, differences between the perceived barriers of youth and adults should be explored. Future methodologically rigorous studies should target modifiable patient factors as an opportunity to improve consistent use of this technology.

Pupillary Response Analysis as a means of Non-Invasive Glucose Monitoring

Ken Steinberg, BSCS

GlucoSight

Nashua, New Hampshire, United States

ken@glueosight.com

Objective:

The objective is to derive a pervasive and portable method for determining the blood sugar level and/or the glycemic range of a TD1/TD2/Gestational diabetic using true, low-power, non-invasive methods.

Method:

The subject’s pupil is placed in a light isolated IR environment where it is subsequently stimulated and then re-isolated. Digital video imagery is captured during this sub-ten second stimuli cycle and analyzed in near real-time. The datum derived is quantitatively analyzed resulting in a glucose range determination.

Result:

Reliable outcomes are desired for use in daily glucose monitoring and potential screening applications.

Conclusion:

Current and ongoing research shows deterministic correlations between a diabetic’s blood sugar level and a number of data points derived from image analysis and comparative analytics. Research has also yielded compelling supporting data for the requirement for individualized metrics.

Can We Fix It? Yes We Can! Simplifying Nutrition in STAR Glycemic Control

Kent W. Stewart, BE(Hons); J. Geoffrey Chase, PhD; Jennifer Dickson, PhD; Christopher G. Pretty, PhD; Geoffrey M. Shaw, MbChB, FJFICM

Department of Mechanical Engineering, Centre for Bio-Engineering, University of Canterbury, Christchurch, Canterbury, New Zealand

kent.stewart@pg.canterbury.ac.nz

Objective:

Critically ill patients often experience stress-induced hyperglycemia, with increased morbidity and mortality. STAR is a proven tablet-based glycemic control (GC) protocol that varies insulin and nutrition to control glycaemia. However, varying nutrition rates can be contentious despite STAR achieving internationally leading nutrition levels. This study evaluates constant and time-varying nutrition protocols for STAR.

Method:

STAR currently modulates caloric nutrition to between 30-100% of American College of Chest Physician (ACCP) goals (ACCP Guidelines). This study investigates constant feed rates of: 70%, 90%, 100%, and fixed rates that step nutrition to 100% over the first three days of GC (70%, 90% and 100%). In silico virtual trials of 33 patients (6,924 hours) from the Christchurch ICU compared safety, performance and nutrition-insulin delivery across all protocols and clinical STAR data.

Result:

Fixing the nutrition by 90% of goal feed showed most benefit. All protocols achieved similar time-in-band data (4.4- 8.0 mmol/L: STAR: 91.9%, Fixed: 93.2-91.6%, Stepped: 92.1%). Slight reductions in hyperglycemia (BG > 10 mmol/L) were achieved using 70% and 90% fixed nutrition (STAR: 1.6%, Fixed: 1.0-1.7%, Stepped: 1.5%). However, carbohydrate nutrition intake in this fixed and stepped approach was slightly reduced vs. STAR’s patient-specific approach (STAR: 5.3 g/hr, Fixed: 3.7-5.3 g/hr, Stepped: 4.7 g/hr). Hypoglycemia was reduced the most when 90% fixed nutrition was used (BG <2.2 mmol/L: STAR: 4 episodes, Fixed: 2, 1 and 3 episodes, Stepped: 3 episodes). Median insulin delivery was very similar (STAR: 2.5 U/hr, Fixed: 2.0-2.5 U/hr, Stepped: 2.5 U/hr) for all 3 nutrition protocols.

Conclusion:

All protocols provided equally safe and effective GC. The current patient-specific nutrition protocol provides only a small overall benefit. Fixed nutrition protocols should be used with the STAR to reduce clinical burden. The 90% goal feed fixed approach provides world leading nutrition delivery, best performance, and easiest clinical implementation.

Variability is a Constant! Insulin Sensitivity and its Variability in 4 ICU Cohorts

Kent W. Stewart, BE(Hons); Jennifer Dickson, PhD; Christopher G. Pretty, PhD; Geoffrey M. Shaw, MbChB, FJFICM; J. Geoffrey Chase, PhD

Department of Mechanical Engineering, Centre for Bio-Engineering, University of Canterbury, Christchurch, Canterbury, New Zealand

kent.stewart@pg.canterbury.ac.nz

Objective:

Glycemic control (GC) in the intensive care unit (ICU) has been shown to be associated with reduced morbidity and mortality, but it is contentious because safe and effective GC is difficult to achieve consistently. Insulin sensitivity (SI) describes a patient’s metabolic state and changes with time. STAR is a proven GC protocol, which uses a stochastic model of SI to provide safe and effective GC based on risk and variability in patient response. This study evaluates the SI value and its hour-to-hour variation across four different cohorts.

Method:

Clinically validated model-based SI was identified using clinical data from: SPRINT and STAR in the Christchurch ICU, New Zealand (N=436, N=549); STAR in the Gyula ICU, Hungary (N=48) and CHU, Liege, Belgium (N=20). Each cohort uses different GC protocols and/or clinical practices. Fitted SI values and their hour-to-hour variation were compared for each cohort using cumulative density functions (CDF’s) and cohort specific stochastic models.

Result:

SI CDFs showed absolute SI values between all the cohorts to be significantly different (P < 0.05). However, the hour-to- hour SI variation in the stochastic models for each cohort had almost identical median, 5^th, and 95^th percentile profiles. This result suggests the equivalent GC achieved using STAR in both Gyula and Christchurch (P = 0.6) was due to almost identical SI variability across cohorts, even though absolute SI values were significantly different (P < 0.05).

Conclusion:

The absolute value of SI is cohort specific. However, hour-to-hour variation for each cohort is almost identical. Thus, stochastic SI models are a robust and effective way to provide safe and effective GC across multiple different cohorts because dosing is specific to future risk rather than current response and dosing is much more consistent across cohorts than SI level.

Stationary Lancing Surpasses Traditional Lancing in Both Pain and Complexity

Chandler Stormo; Curt Christensen, BS; Dan Davis, BS; Erik Davis, BS; Lonny Stormo, BS, MBA

POPS! Diabetes Care, Inc.

Oak Park Heights, MN, United States

chandler.stormo@popsdiabetes.com

Objective:

Obtaining an accurate blood glucose measurement depends on lancing the skin. For people with diabetes, this step proves to be painful and tedious, leading them to test their blood glucose less frequently and reuse lancets, both of which pose significant health risks for the patient. A new lancing innovation, which keeps the needle stationary to reduce pain and includes far less steps than traditional lancing to reduce complexity, was developed to make blood glucose tests much easier for people with diabetes.

Method:

While traditional lancing uses a spring to accelerate a lance into the finger, stationary lancing accelerates the finger onto the lance by pressing a spring layer. With the stationary lancing technique, there is no need to assemble the device before lancing; the patient only has to press down with their finger onto the stationary lancet. A study was conducted with 143 people who use traditional lancing regularly. The study asked users to lance their finger with a stationary lancet, and then answer survey questions comparing the two techniques.

Result:

In the study, 93% of participants said that stationary lancing was “much less” or “less” painful and 86% said that it was “much easier” or “easier” than traditional lancing.

Conclusion:

Stationary lancing demonstrated to be far more preferable to traditional lancing, due to reduced pain and complexity associated with it. Many participants in the study said they would rather use this lancing method when testing their blood glucose level. Both the preferable user experience, and the improved size, will enable future integration of lancing and blood glucose testing to improve blood glucose management.

An Integrated Microdevice for Neutrophil Purification and Functional Phenotyping from a Drop of Blood

Hui Min Tay, BS; Rinkoo Dalan, MBBS, MRCP, FRCP, FAMS; Bernhard Otto Boehm, MD, FRCP; Han Wei Hou, PhD

Lee Kong Chian School of Medicine, Nanyang Technological University

Singapore

tayhuimin@ntu.edu.sg

Objective:

Neutrophils are key effector cells of the innate immune system and neutrophil dysfunction has been previously described in diabetic subjects. We aim to develop an integrated microfluidics biochip for single step neutrophil purification and functional phenotyping using small blood volume (fingerprick) for point-of-care diabetes testing.

Method:

Whole blood (~10 pL) was pumped through a polydimethylsiloxane (PDMS) microdevice consisting of a straight microchannel connected to two side chambers. Larger neutrophils were efficiently separated from red blood cells based on biomimetic cell margination and further affinity-enriched (CD66b+) at the side chambers. Purified neutrophils were subsequently exposed to a stable diffusion-based gradient of the preloaded chemoattractant fMLP and calcium ionophore (AZ23187) to study chemotaxis and formation of neutrophil extracellular traps (NETosis), respectively. Time-lapsed imaging and single cell image analysis was performed to probe chemotaxis (migrating cells and chemotactic velocity) and NETosis (nuclear membrane degradation) phenotypes.

Result:

Cell migration and chemotatic velocity were significantly lower (P < 0.05) in inflammatory conditions, as defined by high glucose (30 mM) or TNF-a, when compared to untreated neutrophils. Exposure to metformin (1 mM) also led to dynamic changes in neutrophil chemotaxis behavior. Glucose-treated neutrophils (30 mM) exhibited increased NETosis (P < 0.05) compared to untreated or mannitol-treated (30 mM) neutrophils. A significant suppression of neutrophil chemotaxis was observed in T2DM patients which correlates with higher cardiovascular risk and which might be mitigated by in vitro metformin treatment.

Conclusion:

Our preliminary results suggest that neutrophil chemotaxis and NETosis are novel functional biomarkers for risk stratification of T2DM patients. The unique strategy of integrating microfluidic neutrophil sorting with cellular functional assays facilitates user operation. The developed diagnostic platform could be further advanced for point-of-care inflammatory profiling.

A Model of Endogenous Insulin Secretion During Exercise

Felicity Thomas, BE(Hons); Chris Pretty, PhD; Kent Stewart, BE(Hons); Geoffrey M Shaw, MbChB FJFICM; Thomas Desaive, PhD; J Geoffrey Chase, PhD

Department of Mechanical Engineering, University of Canterbury

Christchurch, Canterbury, New Zealand

felicity.thomas@pg.cantebury.ac.nz

Objective:

Modeling endogenous insulin secretion can play a significant role in model-based glycemic management. The accuracy of modeled endogenous insulin secretion can significantly impact other model-identified parameters, like insulin sensitivity (SI), and thus dosing. However, many models of endogenous secretion are based exclusively on sedentary individuals (healthy and diabetic), and are not appropriate during and immediately after exercise, due to the impact on secretion of postexercise hyperglycemia and the increase of counter-regulatory hormones. This study develops an endogenous insulin secretion model for healthy individuals during exercise.

Method:

Ten subjects (resting heart rate < 60bpm, training 6-15 hrs/week) underwent a fasted continuous exercise test to exhaustion (EX). Oral glucose doses were given at 30 mins (0.5g/kg) and exhaustion ~85 mins (1 g/kg). Reference capillary BG measurements were measured every 10 mins. Venous blood samples were assayed for plasma insulin and C-peptide at 0, 30, 45min, EX, EX+15min, EX+30min (5/10 subjects), EX+60min Insulin secretion was identified from C-Peptide and plasma insulin, and fitted as a function of BG.

Result:

A constrained linear function of BG provided the best model of pre-hepatic insulin secretion during and immediately after exercise R²=0.53. The glucose coefficient 2559 mU.L/mmol.hr identified was comparable to but at the upper end of a wide range (2646-485 mU.L/mmol.hr) of data published in a number of studies using healthy sedentary individuals. Additionally, compared to data from intensive care patients, modeled equivalent subjects exhibit increased responsiveness despite the known increase of stress hormones during intense exercise.

Conclusion:

The proposed model of endogenous insulin secretion, based on physiological measurements, provides a simple estimate of insulin secretion with comparable physiological parameters to the existing literature. Overall, the endogenous insulin secretion model provides a valuable addition to glucose-insulin modeling; specifically for healthy individuals during exercise.

Composite Coatings for Long-term Prevention of Foreign Body Reaction During Continuous Glucose Monitoring

Namita Tipnis, MS; Michail Kastellorizios, PhD; Fotios Papadimitrakopoulos, PhD; Diane J. Burgess, PhD

University of Connecticut

Storrs, CT, United States

Namita.tipnis@uconn.edu

Objective:

To develop three-month releasing PLGA microspheres/PVA hydrogel composite coatings to prevent the foreign body reaction during continuous glucose monitoring. In this study, the microsphere formulation containing dexamethasone was optimized to achieve high drug loading and low burst release using the drug/polymer co-precipitation method.

Method:

The drug/polymer co-precipitation method was optimized for the following formulation parameters: i) dimethyl sulfoxide (DMSO) concentration in the external aqueous phase (0-50% v/v); ii) temperature of aqueous phase (premixed at 4°C, premixed at room temperature (RT) and just-mixed); and iii) solvent ratios in organic phase [DCM: DMSO (8:1, 4:1, 2:1)]. Polymer blends were dissolved in dichloromethane (DCM) in a 50 mL Teflon vial. The microspheres were then used to prepare composite coatings in pre-made grooved molds. The microsphere formulations and composite coatings were evaluated for drug loading and burst release.

Results:

The optimized process for three-month releasing dexamethasone microspheres was: 40% v/v DMSO in external aqueous phase, aqueous phase mixed just before addition to the organic phase, and a solvent ratio (DCM: DMSO) of 4:1. The optimized formulation had a drug loading of 11.52 ± 0.104% w/w and burst release of 7.65 ± 0.287%. PVA composite coatings made with this formulation (~10 mg) released approximately 150pg dexamethasone within the first 24 hrs. The dose thereafter was anticipated to be approximately 6.67 pg/day for three months.

Conclusion:

The drug/polymer co-precipitation method was successfully used to prepare three-month releasing dexamethasone microspheres and composite coatings using the optimized formulation showed potential to prevent foreign body reaction for up to 3 months.

Hospital QC Rarely Out in Newer Whole Blood Glucose Meter: Opportunity for QC Reduction?

Albert KY Tsui, PhD; Sharon Redel, MLT; Tara Mair, MLT; Rachelle Tessier, MLT; Maria Naval, MLT; Sonya Benwood, MLT; Joshua Fernandes; Anna K Fuzery, PhD; Martha Lyon PhD, George Cembrowski, MD, PhD

Department of Laboratory Medicine and Pathology, University of Alberta, Alberta Health Services, Edmonton, Alberta, Canada

albert.tsui@ahs.ca

Objective:

While quality control (QC) frequency is dictated by accreditation standards, local customs, and the meter manufacturer, hospital blood glucose meter (BGM) QC generally comprises the daily analysis of two different QC levels. Over the last 3 years we have observed a marked decrease in BGM testing at University Hospital using our Roche Inform II (maltose independent, hematocrit independent) coupled with less variation in serial patient measurements. While the use of 2 controls per day seemed appropriate for our previous BGM, we studied the utility of this QC strategy for the Roche Inform II.

Methods:

Two control levels were used, with means of 45 and 308 mg/dL (± 3.5 SD QC limits). We summarized one year of quality control data for 900 meters used in 25 different hospitals and operated by 12,000 personnel.

Results:

Of the 436,908 strips used for QC in 2015, only 0.88% provided error signals with the majority (83%) being false alarms associated with low and high specimens being incorrectly designated as high and low specimens, respectively. Only 741 strips provided error signals that could be investigated. The high and low controls were out of limits 352 and 389 times, respectively. Reanalysis of these controls resulted in 40 apparently persistent errors. Following a third subsequent analysis, 16 were out of control. Interestingly, no meters were returned to Regional Point-of-Care (most of the returns were associated with faulty strip insertion and lesser numbers with impaired connectivity and battery issues).

Conclusion:

The exceedingly low error rate helps to explain the decreased utilization of blood glucose testing in our region. This low error rate forms the rationale for beginning to study more efficient QC strategies including running 1 control per day.

Are Survivors Easier to Control? Why the Association of Glycemia and Mortality in Critical Care is Real

Vincent Uyttendaele, MS(Ing); Jennifer L. Dickson, PhD; Kent Stewart, BE; Geoffrey M Shaw, MBChB; Thomas Desaive, PhD; J. Geoffrey Chase, PhD

University of Canterbury Christchurch, Canterbury, New Zealand

vincent.uyttendaele@pg.canterbury.ac.nz

Objective:

Glycemic control of hyperglycemia to improve outcomes in critical care has proven difficult to achieve consistently. The association of glycemia and outcome is hypothesized to occur because survivors are essentially easier to control, thus having lower glycemia, while non-survivors are harder to control. This study compares patient-specific insulin sensitivity (SI) level and variability (difficulty of control) in survivors and non-survivors.

Method:

Data were analyzed from Days 1-3 of N=145, Specialized Relative Insulin and Nutrition Tables (SPRINT) protocol patients (N=119 survivors; N=26 non-survivors) who started GC within 12 hours of ICU admission and spent >24 hours on GC. Hourly SI was determined for each patient using a clinically-validated physiological model. SI variability was defined as hourly percentage change in SI (%ASI). SI level and variability are compared over 6-hour blocks to 72 hours.

Result:

Patients on the SPRINT protocol achieved equal GC to both groups, with 77% of patients (76% survivors; 81% nonsurvivors; p=0.80, Fisher Exact) achieving >50% of time in a 72-126 mg/dL band by 24 hours (91% and 93% by days 2-3). SI level increased over time and variability decreased. SI distributions were similar (p=0.11-0.25, Wilcoxon) between survivors and non-survivors hours, except hours 7-12 and 30-72 (p < 0.05). The %ASI were similar (p=0.12-0.77, Kolmogorov-Smirnov) except hours 37-42 (p=0.04). Results suggest SI and %ASI are drawn from the same distribution for both groups for the first 30 and 72 hours, respectively. Results did not change considering only patients who stayed >72 hours.

Conclusion:

Lack of difference between survivors and non-survivors in SI level and variability (which govern control of difficulty, risk, and outcome glycemic level and variability) suggest neither cohort is easier/harder to control, particularly in the acute first 30 hours. Glycemic level and variability are therefore a function of the GC protocol, rather than eventual patient outcome.

When NICE is Not Nice: Performance of Two ICU Glycaemic Control Protocols

Vincent Uyttendaele, MSc; Jennifer Dickson, PhD; Kent Stewart, BEng; Geoffrey Shaw, MBChB; J. Geoffrey Chase, PhD

University of Canterbury, Dept of Mechanical Engineering

Christchurch, New Zealand

vincent.uyttendaele@pg.canterbury.ac.nz

Objective:

Hypoglycemia, hyperglycemia and blood glucose (BG) variability are associated with suboptimal outcomes in critical care. However, the NICE-SUGAR trial (unexpectedly?) showed no clinical benefit from intensive insulin therapy. This study compares the table-based NICE-SUGAR and model-based STAR protocols to assess their relative capability to achieve safe and effective control for all patients.

Method:

Validated virtual patients (n=443) were used to simulate glycemic outcomes of the NICE-SUGAR and STAR protocols. Key outcomes assessed tightness and safety of control for all patients: %BG in 80-144 mg/dL range (PTR); per-patient mean BG (PPM_BG); and incident hypoglycemia (BG<40 mg/dL). These metrics assessed overall performance and safety for each patient. Results were assessed for NICE-SUGAR measuring per-protocol (~24/day) and at reported average rate (~3-hourly; ~8/day). STAR was assessed with measures 1-3-hourly at an average rate ~12/day.

Result:

Per-protocol, STAR provided tight control, with higher PTR (90.7% vs. 78.3%) and tighter median [IQR] PPM_BG (112 [106-119] vs. 117 [106-137] mg/dL), and greater safety from hypoglycemia (N=5 (1%) vs. N=10 patients (2.5%)). The 595^th percentile range PPM_BG for NICE-SUGAR (97-185 mg/dL) shows ~5% of NICE-SUGAR patients had mean BG above 180 mg/dL matching clinically reported performance. STAR’s 90^th percentile PPM_BG range was (97-146 mg/dL). Measuring as recorded clinically, NICE-SUGAR had PTR of 77%, PPM_BG of 122 [110-140] mg/dL and 24 (6%) of patients experienced hypoglycemia. These results match clinically reported values well (mean BG: 115 vs. 118 mg/dL clinically vs. simulation) and clinically 7% of patients had a hypoglycemic event.

Conclusion:

Glycemic control protocols need to be both safe and effective for all patients before potential clinical benefits can be assessed. NICE-SUGAR (measured ~24/day or as reported ~8/day) was unable to achieve this outcome for all patients.

Performance Comparison of the GLUCOCARD® Shine and Contour® Next Against the ISO 15197:2013 Accuracy Criteria

Julie Walker, RN, BSN, PHN; Patricia Gill, BA, MLT; Danielle Maher, BS; John Gleisner, BS, PhD

ARKRAY USA, Inc.

Minneapolis, Minnesota, United States

walkerj@arkrayusa.com

Objective:

This study compared the performance of the GLUCOCARD® Shine to the Contour® Next.

Method:

Three lots of test strips were evaluated for performance for each BGMS at ARKRAY Inc. factory. All testing was conducted under the same IRB approved protocol and used the same group of participants. To reduce variability, samples were drawn directly from the fingertip of confirmed diabetics (n=104) by laboratory professionals. Reference values were obtained using the YSI Model 2300 Analyzer. The data was evaluated against the accuracy boundaries of the ISO 15197:2013 Standard and the Consensus Error Grid.

Result:

The GLUCOCARD® Shine demonstrated that 100.0% of the <100 mg/dL samples (6/6) were within ±15 mg/dL and 99.0% of the > 100 mg/dL samples (97/98) fell within ±15%. Overall bias was -1.3% and the correlation coefficient was (r) = 0.99. For the Contour® Next, 100.0% of the <100 mg/dL samples (7/7) gave values that were within ±15 mg/dL and 100.0% of the > 100 mg/dL samples (97/97) fell within ±15%. Overall bias was -2.1% and the correlation coefficient was (r) = 0.99. All data for both BGMS were within the A and B zones of the Consensus Error Grid.

Conclusion:

The GLUCOCARD® Shine and Contour® Next had equivalent performance when assessed against the ISO 15197:2013 boundaries.

Performance of the Assure® Platinum Blood Glucose Monitoring System for Multi- Resident Use in the Long Term Care Setting Against the ISO 15197:2013 Accuracy Criteria

Julie Walker, RN, BSN, PHN; Patricia Gill, BA, MLT; Danielle Maher, BS; John Gleisner, BS, PhD

ARKRAY USA, Inc.

Minneapolis, Minnesota, United States

walkerj@arkrayasa.com

Objective:

The purpose of this study is to determine in ongoing trending studies the performance of the Assure® Platinum Blood Glucose Monitoring System (BGMS) against accuracy boundaries of ISO 15197:2013. This standard requires that 95% of BGMS results be within ±15 mg/dL of the reference analyzer at glucose concentrations <100 mg/dL and within ±15% of the reference analyzer at glucose concentrations >100 mg/dL. Furthermore, 99% of all results are required to be within the A and B zones of the Consensus Error Grid.

Method:

Fingerstick testing was performed by trained laboratory professionals from subjects with diabetes (n=240) on eight lots of Assure^® Platinum test strips at the ARKRAY Inc. factory, in Minneapolis, MN. Reference values were obtained by using the YSI Model 2300 Analyzer. The data were analyzed against the accuracy boundaries of the ISO 15197:2013 standard and the percentage of the results in the A plus B zones of the Consensus Error Grid.

Result:

The data show that 100% of the results <100 mg/dL (35/35) were within ±15 mg/dL of the YSI and 99.5% of the results >100 mg/dL (204/205) fell within the ±15% of the YSI. All data were within the A and B zones of the Consensus Error Grid. The overall bias to YSI was -1.9%. The correlation coefficient was (r) = 0.98 which demonstrates a strong linear relationship between Assure^® Platinum and the YSI reference method.

Conclusion:

The data acquired on the Assure® Platinum BGMS by laboratory professionals was within the accuracy boundaries of the ISO 15197:2013 standard.

Advanced Hydrogels for Implantable Electrochemical Glucose Sensors

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

University of Connecticut and Biorasis Inc.

Storrs, CT, United States

papadim@mail.ims.uconn.edu

Objectives:

Glucose oxidase (GO_x)-based electrochemical sensors are commonly used for continuous glucose monitoring (CGM) devices. Enzyme denaturation and leaching, as well as biofouling-induced pore clogging, gradually impede their performance. 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 on a novel PEGylated hydrogel with low cytotoxicity that is capable of being photo-crosslinked in place in its fully hydrated state.

Methods:

A PEGylated copolymer was synthesized via a free radical polymerization using azobisisobutyronitrile as the initiator. This copolymer was mixed with GO_x enzyme and photo-patterned on the electrode (via UV exposure) followed by 24-hour incubation to allow chemical coupling of GO_x to the PEGylated hydrogel. Mouse dermal fibroblasts (cell line L929) were incubated for 72 hour with various concentrations of this hydrogel (0.01 to 100 u g/ml). Cell viability was assessed via CellQuanti-Blue™ Cell Viability Assay Kit fromBioAssay Systems (Hayward, CA).

Results:

Glucose sensors utilizing GO_x-grafted hydrogels showed excellent stability during continuous testing for 30 days. The glucose sensors exhibited sensitivities as high as 320 nA mM^-1mm^-2 with linearity beyond 25 mM of glucose and limit of detection lesser than 1 pM. Cytotoxicity studies indicate that this PEGylated copolymer shows no cytotoxicity for up to 72 hours for the entire concentration range.

Conclusions:

Photo-patternable PEGylated hydrogels were demonstrated as ideal matrices for covalent immobilization of GO_x enzyme with high spatial tolerances, enabling the facile fabrication of CGM devices. In this study, we demonstrate that this photocrosslinked hydrogel is not cytotoxic and therefore poses no threat to the local tissue. This hydrogel allows the fabrication of CGM sensors with long-term stability and optimal performance.

In-Situ Blood Glucose Monitoring Using Autonomous Skin Lancing Device and Needle- Type Sensor: Results from In-Vitro Testing

Gang Wang, BEng; Calena R. Marchand, BEng.; Oliver F. Bathe, MD; Orly Yadid-Pecht, DSc; Martin P. Mintchev, PhD

Biomedical Engineering Graduation Program, University of Calgary

Calgary, Alberta, Canada

gawang@ucalgary.ca

Objective:

Our team previously proposed a wearable microsystem called “eMosquito”. Human testing clearly demonstrated the feasibility of this shape memory alloy (SMA) microactuator design for skin lancing. This paper presents an integration of a needle-type glucose sensor with the SMA-based actuator of eMosquito for in-situ glucose monitoring to improve the reliability of blood sampling.

Method:

Five glucose solutions of different concentrations were prepared using distilled water and concentrated glucose tablets in beakers. A soft cannula sensor was first calibrated in a glucose solution before use. The sensor was then inserted into a 25G hypodermic needle and leveled with the needle tip. The new needle-type sensor was integrated with the skin lancing mechanism of the eMosquito device. Upon activation, the needle tip was submerged ~2.5mm deep in a glucose solution on a 35°C hotplate. Immediately after a glucose concentration value was obtained from the Dexcom receiving device, the concentration was measured five times with a standard glucose meter and test strips to obtain an average reading of each solution as a reference. This process was repeated five times to obtain measurements in each of the five glucose solutions.

Result:

A linear correlation is shown between the standard meter measurements and those from the integrated needle-type sensor on the eMosquito actuator.

Conclusion:

These preliminary results show potential for a needle-type sensor to be combined with the eMosquito device for continuous glucose monitoring. A customized glucose sensor is currently being developed to replace the commercialized biosensor used in this test and it will allow for the eMosquito device to be miniaturized into a wearable size in the future.

CSII Catheter with Extended Lifetime and Rapid On-Off Pharmacodynamics

Matthew Wiltshire, BS; Alek Dinesen, MS; Jasmin Hauzenberger, MS; Jessica Vidas, BS; Jason McGavin, BS; Channy Loeum; Alex Ebelin, BS; Yuri Trimba, BS; Talha Kaner, mD; Mayur Pate, BS; Mark Joseph, MD; Brooke Torjman; Anthony Pantoja, MD; Marc Torjman, PhD; Peter McCue, MD; and Jeffrey Joseph, DO.

Sidney Kimmel Medical College, Thomas Jefferson University

Philadelphia, PA, United States

jeffrey.joseph@jefferson.edu

Introduction:

Our goal is to understand the mechanisms that cause variable insulin absorption when infused through a commercial CSII catheter into the subcutaneous tissue. A pilot observational canine study was performed to evaluate the PK-PD of insulin lispro absorption and the tissue histology surrounding CSII catheters with a 6 mm Teflon cannula implanted for 7 days. Insertion of a CSII cannula through the skin into the subcutaneous tissue damages cells, connective tissue and extracellular matrix. Insertion trauma and ongoing trauma produces a layer of acute inflammatory tissue- consisting of thrombus, tissue debris, neutrophils, macrophages, and fibroblasts. This layer of inflammatory tissue may change over time from thin, sparse, and discontinuous to thick, dense, and continuous. A bolus of insulin needs to travel through this inflammatory tissue layer into adjacent vascular subcutaneous tissue, along the path of least resistance. The variability of insulin absorption into the circulation may be related to the surface area of insulin in close proximity to capillary and lymph vessels, capillary blood flow, insulin transport across the endothelium, lymph flow, degradation by proteases within the wound and/or lymph nodes, and insulin transport upward onto the skin surface. Clinicians recommend insertion of a new CSII catheter at an alternate location every 2 to 3 days to minimize the risk for hyperglycemia, hypoglycemia, DKA and infection.

Methods:

Thirteen adult mongrel female non-diabetic canines were studied for a maximum of 7 days. Two CSII catheters (Quickset, Medtronic MiniMed) and 2 CGM (DexCom G4 with Share) were inserted into the soft underbelly of each canine using general anesthesia. One CSII catheter was continuously infused with insulin lispro (basal rate 0.083 units/hour = 2 units/day x 7 days) while the other catheter was filled with saline and capped. Hyperinsulinemic-euglycemic glucose clamp experiments were performed using the insulin infused catheter on days 1, 3, 6, & 7 after implantation. Glucose clamps used a 0.1 unit/kg bolus of insulin lispro only (n= 5), insulin lispro plus tissue plasminogen activator (n=3), insulin lispro plus streptokinase (n=3), insulin lispro plus streptokinase plus mechanical vibration (n=1), and insulin lispro plus tissue plasminogen activator plus mechanical vibration (n=1). The skin and subcutaneous tissue surrounding the CSII catheters were excised, fixed, sectioned, and stained to produce tissue histology slides. The PK-PD and histology data were evaluated to determine a relationship between the location and degree of tissue damage, thrombus, and inflammation with the rate and consistency of insulin absorption.

Results:

Implantation of commercial Teflon CSII catheters for 7 days produced a variable amount of damage to adipose cells, connective tissue, capillary vessels, lymph vessels and skeletal muscle cells. All of the CSII cannulas were partially or completely surrounded by a layer of inflammatory tissue. This layer varied in thickness, composition, density, and continuity. The tissue histology surrounding CSII catheters infused with insulin for 7 days was similar to the tissue histology surrounding cannula filled with saline and capped. The tissue surrounding the CSII catheters infused with lispro only (control group), lispro plus tissue plasminogen activator, and lispro plus streptokinase had similar tissue histology and minimal thrombus. The absorption of insulin into the circulation (PK) on day 1 (shortly after awakening from general anesthesia) was blunted in all animals. In general, insulin absorption was highly variable from day to day and from canine to canine.

Discussion:

Conclusions from this pilot study are limited because the histology specimens were obtained only from catheters implanted for 7 days (lack of histology data from days 1, 3 and 5), and the catheters were infused with dilute insulin only at a basal rate to minimize the risk for hypoglycemia (no bolus infusion until day 7). The limited absorption of insulin on day 1 may have been caused by the residual effects of Isoflurane anesthesia. In general, the damage caused by CSII cannula insertion and maintenance produced a layer of inflammatory tissue that varied in thickness, density, composition and continuity. We hypothesize that the variable amount of tissue damage and immune response leads to the variable insulin PK and PD commonly observed in clinical practice. Insulin may easily travel into adjacent vascular subcutaneous tissue through a layer of inflammatory tissue that is thin, sparse, and discontinuous- leading to rapid onset/offset of insulin PK-PD. A layer of inflammatory tissue that is thicker, denser, and more continuous may slow or inhibit the movement of insulin into adjacent vascular tissue- leading to slower and more variable insulin absorption. In general, we observed similar variability in the layer of inflammatory tissue around CSII cannula that were infused with lispro insulin only, insulin plus tissue plasminogen activator, insulin plus streptokinase, or filled with saline and capped. Fibrinolysis did not seem to significantly affect the tissue histology or rate/consistency of insulin absorption. Further research is needed to better define the

Detection of Hypoglycemia that Requires Further Treatment in Addition to PLGM Suspension

Di Wu, PhD; Benyamin Grosman, PhD; Anirban Roy, PhD; Neha Parikh, PhD; Ohad Cohen, MD; Rebecca Gottlieb, PhD

Medtronic Diabetes

Northridge, CA, United States

di.wu@medtronic.com

Objective:

Medtronic’s predictive low glucose management (PLGM) system automatically suspends basal insulin delivery when sensor glucose levels are predicted to approach a low limit. The PLGM system reduces hypoglycemic events as well as time spent in the hypoglycemic range. However, severe hypoglycemia induced by excessive exercise or large amount of insulin on board needs further treatment in addition to insulin suspension. To determine the conditions by which PLGM deployment will be unable to mitigate hypoglycemia, glucose trend and insulin on board as conditions of detection algorithms were investigated. These additional insights can help guide users to reduce their risks of hypoglycemia.

Method:

Glucose trend and insulin on board are two dominant factors for the severity of impending hypoglycemia. Therefore, the glucose trend and insulin on board can be utilized to further aid the PLGM system in detecting missed hypoglycemic events. In-silico PLGM simulations were performed on 218 virtual patients generated using the Medtronic Carelink database. By optimizing the algorithm conditions utilizing glucose trend and insulin on board, specificity and sensitivity of detecting hypoglycemia (defined as glucose below 70 mg/dl) after PLGM suspension were maximized.

Result:

Through in-silico simulations, a total of 188 suspension events were triggered by PLGM. Of the 57 PLGM suspensions that were not able to prevent hypoglycemia, 43 events were truly detected (75% true positive) whereas 14 events were falsely missed (25% false negative) by the new hypoglycemia detection utilizing glucose trend and insulin on board. Of the rest 131 PLGM suspension events that successfully prevented hypoglycemia, 107 non-hypoglycemia events were truly detected (82% true negative) whereas 24 non-hypoglycemia events were falsely detected as hypoglycemia (18% false positive).

Conclusion:

Based on in-silico studies, glucose trend and insulin on board information, readily available to users, can detect the hypoglycemia not prevented by PLGM suspension. The rules formulated from the simulation demonstrated encouraging 82% specificity rate and 75% sensitivity rate in detecting the non-prevented hypoglycemia by PLGM suspension.

Insulin Doses Prediction AI For Diabetes Type 1

Thomas Wuttke, Dipl. Ing.

Diafyt MedTech

Leipzig, Saxonia, Germany

thomas@diafyt.com

Objective:

Estimating the optimal amount of insulin at the right time is challenging for people with diabetes. From a control viewpoint, the main challenges are time delays, constraints, meal disturbances, and nonlinear dynamics. Developing an easy to use and inexpensive way of suggesting insulin doses to patients will simplify diabetes therapy for Type 1 diabetes.

Method:

A minimally-invasive system using subcutaneous continuous glucose monitoring and a smartphone app is needed. The tested method is used a stochastic regression method in combination with real-time sensory data and big data analysis. The software prototype is running on a smartphone where data are collected from continuous BCL metering, a cellphone movement sensor, and manual data entry for food intake (e.g., carbohydrates) and insulin doses. Simulation and tests have been run for 6 months for data collection and algorithm tuning.

Result:

Test person/proband: adult, Type 1 diabetes: BCL variance decreased from 90% out-of-range to 40% out-of-range (range: 4.5-7.5 mmol/L). The average number of hypoglycemia events (< 2.8 mmol/L) changed from 4 events per week to 2 events per week and the average number of hyperglycemia (>11 mmol/L) changed from 10 events per week to 3 events per week. HbA1c decreased from a pre-test result of 8.3 to 6.8 after 6 months.

Conclusion:

The trial demonstrated that the system achieves satisfactory glycemia regulation in a Type 1 diabetic patient. Simulation tests have shown that the algorithm has the potential to introduce further improvements. The proposed scheme is also robust with respect to false or missing manual data entry and with imperfect knowledge of the amount of ingested glucose.

Verifying Hospital Insulin Syringe Dose

Phil Wyman, BA

CEO - Digital Hospital, Inc.

San Jose, CA, United States

wyman@digitalhospitalinc.com

Objective:

Digital Hospital, Inc. has as its goal the elimination of subcutaneous insulin medication errors in hospitals by confirming the proper type of insulin in a syringe (1 ml or smaller) and the proper dosage of insulin in a syringe (1 ml or smaller) before administering the insulin to the patient.

Method:

Digital Hospital, Inc. is developing a device for hospital nurses to use as a “double-check” system for when they fill a syringe from a vial of insulin. Upon verification of proper type and dose of insulin, our device will print an attachable syringe barcode. The barcode will be scanned at the patient’s bedside, effecting a record of the verified insulin dosage information into the patient’s EHR (Electronic Health Record).

Result:

Alpha software and hardware for the “double-check” insulin device have been created. Digital Hospital, Inc. is now poised to start the pre-manufacturing stage of Design Research and Industrial Design.

Conclusion:

The need to eliminate insulin medication errors in hospitals is addressed by this insulin confirmation device. Insulin dosage errors cause hypo or hyperglycemia resulting in prolonged length of stay, morbidity and mortality. Currently there is no true confirmation of the dose that is administered by the nurse. The development of this “double-check” device offers true confirmation of the correct type and dosage of insulin in a syringe before injection. The documentation of dose is accurately recorded in the EHR.

Qualitative Analysis of a Diabetes Alert Dog- Themed Blog: Pilot Study of Content and Utility of a Niche Blog

Christopher Yeisley; Tamara K. Oser, MD; Sean M. Oser, MD, MPH

Penn State College of Medicine

Hershey, PA, United States

cyeisley@hmc.psu.edu

Objective:

The utility of blogs and other social media sources as support systems for intensively self-managed conditions including Type 1 diabetes (T1D) has been a recently popular topic of study. This pilot study analyzes a blog specifically focused on Diabetic Alert Dogs (DADs), how users perceive DADs’ value, changes in quality of life, and how blogging provides support for other DAD users.

Method:

The blog “Black Dogs Rule” was qualitatively analyzed via grounded theory methodology to identify emergent themes. Blog posts and associated comments were imported into the qualitative data management program NVivo 11. The team reviewed the blog content, created a codebook, and established intercoder reliability with a kappa of 0.908. Coding will proceed until no new themes are uncovered and a detailed analysis to determine emergent themes will be conducted.

Result:

Using a predetermined number of blog posts, intercoder reliability was established between two coders with a kappa of 0.908. Upon completion of coding to saturation, thematic analysis will determine emergent themes. Clinical review of content will determine the frequency of medically incorrect or unsafe information.

Conclusion:

In addition to complications of physical health, significant psychological effects borne by patients and their families may result in restriction of daily activities or lifestyles, affecting mental well-being and quality of life. The increased sense of comfort provided by DADs might provide a major positive impact based on analysis of this blog and associated comments from numerous patients/families. Moreover, blogs themselves offer a venue for patients/families to connect and build a smaller community within the diabetes online community, and to offer help and support to others in similar situations.

An Explicit and Verifiable Solution to Zone Model Predictive Control for Glycemic Control in T1DM

Stamatina Zavitsanou, PhD; Ankush Chakrabarty, PhD; Eyal Dassau, PhD; Francis J. Doyle III, PhD

Harvard John A. Paulson School of Engineering and Applied Sciences, Harvard University, Cambridge, MA, United States

szavitsanou@g.harvard.edu

Objective:

Zone Model Predictive Control (ZMPC) is a clinically viable decision-making algorithm for glycemic control in Type 1 Diabetes Mellitus. A potential limitation of ZMPC is the computational expenditure involved in obtaining appropriate insulin infusions by iteratively solving a quadratic programming (QP) problem. The objective of this work is to develop an explicit and verifiable solution to the ZMPC that aims to enhance the online computational efficacy while preserving the clinical advantages demonstrated by the ZMPC.

Methods:

We leveraged multi-parametric programming techniques to transform the solutions of the online QP into a look-up table that is available offline. The resultant controller was referred to as a multi-parametric ZMPC (mp-ZMPC). The control action was obtained as a function of the states of the internal linear model and the IOB constraint. Specifically, given the current glucose measurement the system parameters were estimated, which were then used to compute the optimal insulin infusion by performing simple function evaluations.

Results:

In-silico testing of the mp-ZMPC was performed on 10 adult subjects using the UVA/Padova metabolic simulator with unannounced meal disturbances of 50 and 70 g of carbohydrate. The resulting insulin profile and glucose trajectories are observed to be identical from a clinical perspective to a ZMPC. In detail, out of 30 simulated hours, the mp-ZMPC control resulted in glucose trajectories spending 65.7 +/-11.0 % within 80—140 mg/dl, 16.5 +/-12.0 % within 140—180 mg/dl, with no hypoglycemia (<70 mg/dl) events.

Conclusion:

The proposed mp-ZMPC is an attractive decision-making algorithm for the artificial pancreas (AP) as it provides identical insulin profiles as a ZMPC. Furthermore, the availability of explicit solutions provides engineers and clinicians the opportunity for rigorous offline verification necessary for safety-critical biomedical devices such as the AP. Complexity reduction methods are currently under investigation for the development of a simpler and faster mp-ZMPC.

Establishment of a Porcine Type I Diabetes Model to Evaluate Continuous Subcutaneous Insulin Infusion Site Loss

Gina Zhang, PhD; Anuradha Bhatia, PhD; Rahul Ghosh, BS; Sarnath Chattaraj, PhD

Medtronic

Northridge, CA, United States

gina.zhang@medtronic.com

Background:

“Site-loss” in the context of continuous subcutaneous insulin infusion (CSII) refers to the decrease in insulin absorption at the infusion site over time. Since this is a barrier to effective glycemic control, patients are currently advised to replace their infusion sets every 2-3 days. This study describes the development and assessment of an animal model of site-loss for downstream use in testing potential mitigation strategies.

Methods:

Type 1 diabetes was chemically induced in hairless Yucatan miniature pigs using alloxan. Diabetic and non-diabetic pigs were then administered insulin, placebo, or no solution through subcutaneously implanted infusion set catheters. Blood glucose levels and insulin delivery were monitored over 7 days in order to detect site-loss. Additionally, tissue samples around the sites were biopsied and inspected for inflammation. The presence of specific inflammatory cell types and stimulation of cellular responses were scored.

Results:

Site loss at ~3 days was observed in diabetic pigs administered insulin. Localized tissue inflammation trended to be more severe in the diabetic pigs than in the non-diabetic pigs. Catheters used to infuse insulin were associated with increased presence of inflammatory cells and fibrosis compared to catheters without infusion or catheters used for placebo infusion.

Conclusions:

Based on data from these studies, the porcine Type I diabetes model exhibits site-loss behavior comparable to that in humans. The delivery of insulin appears to aggravate the foreign body response, possibly contributing to localized fibrosis around the catheter. Consequently, this model may be used to study the mechanisms leading to site-loss associated with CSII and evaluate potential solutions for overcoming them.

Real World Assessment of MiniMed Connect

Alex Zhong, MS; Chantal McMahon, PhD; Pratik Agrawal, MS; Siddharth Arunachalam, MS; Toni Cordero, PhD; Scott Tyson, BA, MBA; Huzefa Neemuchwala, PhD, MBA; Francine R. Kaufman, MD

Medtronic, Inc.

Northridge, CA, United States

alex.zhong@medtronic.com

Objective:

The MiniMed Connect (MC) system works with the MiniMed 530G sensor augmented pump system. System updates can be sent from the pump to an app installed on an iPhone® or iPod Touch® every five minutes. The app can display sensor and pump statistics including: sensor glucose data, active insulin level, calibration time, insulin in reservoir, and sensor life. MC can also send automatic text messages when excursion alerts occur. We retrospectively examined MC use from data voluntarily uploaded by 2,794 patients.

Method:

De-identified data from users who had a minimum 60 days of pump data (in the CareLink Personal Database) before and after starting MC use were used for the analyses. Thresholds for extreme hypoglycemia, hypoglycemia, euglycemia, hyperglycemia and extreme hyperglycemia were <50 mg/dL, <70 mg/dL, 70-180 mg/dL, >180 mg/dL and >300 mg/dL respectively. Excursion event frequency was calculated.

Result:

There was a reduction by 3.8% (1.6 per year) for excursions below 50 mg/dL, 8.1% (19.7 per year) for excursions below 70 mg/dL, 10.6% (56.7 per year) for excursions above 240 mg/dL, and 18.5% (40.6 per year) for excursions above 300 mg/dL observed with MC use. Nineteen percent of users (N=531) spent less time in both hypoglycemia and hyperglycemia using MC. Thirty-two percent of users (N=894) spent less time in hyperglycemia, among which 95% (N=849) spent similar or more time in range. Thirty-one percent of users (N=866) spent less time in hypoglycemia, among which 55% (N=476) spent similar or more time in range.

Conclusion:

Users of the MiniMed Connect system and its remote monitoring features may benefit from more convenient access to glucose data and more timely notifications of excursion events.

HbAlc Estimation from a Long-Term Continuous Glucose Monitoring System: Comparison of Different Time Windows

Eric Zijlstra, PhD; Xiaoxiao Chen, PhD; James Masciotti, MS

Profil

Neuss, Germany

Eric.Zijlstra@profil.com

Objective:

A new, long-term, implantable continuous glucose monitoring (CGM) system consisting of a fluorescence-based glucose sensor, wearable smart transmitter and smartphone app has been developed. The PRECISE clinical study was performed to investigate the CGM system’s performance over a 180-day period. The analysis presented here evaluates the different time windows for estimating HbAlc from the continuous glucose measurements.

Method:

In this prospective, single-arm investigation, 71 adult subjects with diabetes used the CGM system for up to 180 days. HbAlc of each subject measured in-clinic at the Day 90 and end of study visits were compared to an estimated HbAlc from average CGM glucose (AG_mg/dL) over the past 7, 14, 30, 60, and 90 days using a linear relationship: eHbAlc (%) = (AG_mg/dL + 46.7)/28.7. Forty-four HbA1c measurements, from 33 subjects with CGM glucose available for at least 80% of the time in all time windows, were included in this analysis.

Result:

The mean (SD) measured HbA1c was 6.93% (0.89%) while the mean eHbA1c was 7.03% (1.04%) from the 7-day window, 7.03% (0.97%) from the 14-day window, 7.01% (0.91%) from the 30-day window, 7.01% (0.87%) from the 60-day window, and 7.00% (0.85%) from the 90-day window. The mean absolute deviation (MAD) decreased as the time window increased (slope = -0.0010, p = 0.0034): 0.39%, 0.36%, 0.34%, 0.32%, and 0.30%, respectively. The percentage of eHbA1c within 0.5% of the laboratory value increased as the time window increased (slope = 0.1011, p = 0.0058): 68%, 70%, 73%, 75%, and 77%, respectively.

Conclusion:

Continuous glucose measurements from a long-term CGM system can be used to estimate HbA1c levels. The estimation error decreases as the time window used for estimation increases from 7 to 90 days.

Good Accuracy of CGM-based Glucose Variability Indices for IGT and T2D Classifications

Giada Acciaroli, MS; Alessandro Palombit, MS; Giorgio Maria Di Nunzio, PhD; Andrea Facchinetti, PhD; Giovanni Sparacino, PhD; Liisa Hakaste, MD, PhD; Tiinamaija Tuomi, MD, PhD; Rafael Gabriel, MD, PhD; Claudio Cobelli, PhD

Department of Information Engineering, University of Padova

Padova, Italy

giada.acciaroli@phd.unipd.it

Objective:

Many glucose variability (GV) indicators have been proposed in the literature, even more since the advent of continuous glucose monitoring (CGM) sensors. How to use the plethora of GV indicators is, however, to some extent, controversial, because several GV indices provide redundant information. In addition, the use of GV indicators to automatically recognize the metabolic status of a subject (e.g. when the subject has impaired glucose tolerance (IGT) or Type 2 diabetes (T2D) remains to be addressed. In the present work, we verify whether or not CGM-based GV metrics can be used for classification purposes in the quite simple task of distinguishing healthy subjects from subjects with either IGT or T2D.

Method:

The dataset consisted of 102 subjects, labelled into three classes through an oral glucose tolerance test (OGTT): 34 healthy, 39 IGT and 29 T2D subjects. A Guardian Real Time or the iPro CGM systems (Medtronic MiniMed, Inc., Northridge, CA) was used to produce a glucose trace from which 25 GV indices were extracted from each monitored subject. To classify each subject into a diagnostic group, we implemented a two-level binary logistic regression model. The first level distinguished healthy from unhealthy subjects and the second level classified the unhealthy subjects as either IGT or T2D.

Result:

The GV indicators distinguished healthy from unhealthy subjects with 90.48 ± 7.53% accuracy. The unhealthy subjects were subdivided into IGT or T2D with 79.18 ± 20% accuracy. The global classification into the three classes has an overall accuracy of 85.86 ± 12.79%.

Conclusion:

CGM-based GV indicators can be effectively used to accurately distinguish CGM traces of healthy and unhealthy patients. More critical, but still promising, is the subdivision of patients into those with IGT or T2D.