Abstract
Dose Accuracy of Reusable Insulin Pens in India: NovoPen 4, NovoPen 3, HumaPen Ergo, Glaritus Pen Royale, INSUPen, and AllStar
Mona Abdel-Tawab, PhD; Ingo Fritz, PhD; Volker Korger, PhD
Central Laboratory of German Pharmacists
Eschborn, Hessen, Germany
Objective:
Currently several reusable insulin pens from different manufacturers are available in India and in some other emerging countries in Asia and Latin America. We assessed the dose accuracy of the insulin pens NovoPen 4 (NP4) and NovoPen 3 (NP3) for NovoMix, HumaPen Ergo (HPE) for Humulin, Glaritus Pen Royale (WR) for Glaritus, INSUPen (IP) for Basalog, and AllStar (AS) for Lantus.
Method:
In total, 15 pens of each type were used for the 4-fold, nonrandomized delivery of low (5 units), midpoint (30 units), and high (60 units) dosage levels, respectively, generating 180 values per pen type. Dose accuracy was evaluated gravimetrically as recommended by DIN EN ISO 11608-1:2000.
Result:
All tested insulin pens delivered doses within the International Organization for Standardization limits at all dosage levels. However, WR and IP revealed decreased dose-accuracy performance at all dosage levels, reflected in higher coefficients of variation (6.5 and 4.6%, respectively, at the 5-unit level, compared with 3.6% for AS, 3.5% for HPE/NP3, and 3.3% for NP4). The mechanical last-dose-stop feature of NP4, NP3, IP, and AS, preventing the dialed dose from exceeding the remaining content and providing greater patient safety, is not available in HPE and WR.
Conclusion:
While all pens delivered doses within ISO limits, slight dose-accuracy variations, as well as last-dose-stop features, confer notable differences between some pens.
Duality of Interest:
This study was supported by Sanofi.
From Bench to Bedside: Conquering the Translational Gap by Humanizing Type 2 Diabetes Research
Zeeshan Ali, PhD; P. Charukeshi Chandrasekera, PhD; Neal D. Barnard MD
Physicians Committee for Responsible Medicine Washington DC
Background:
Obesity and type 2 diabetes (T2D) have reached epidemic proportions worldwide, and thus considerable research effort has been dedicated to developing strategies to understand and treat these complex diseases. The two hallmark features of T2D, insulin resistance and pancreatic dysfunction, have been studied extensively using various animal models. Despite the wealth of knowledge hitherto acquired from animal models, many details of human disease pathogenesis remain unknown, and pharmacotherapeutic options for humans remain limited, with adverse effects associated with many widely used T2D drugs. From gene regulation to pancreatic cytoarchitecture to glucose sensing and transport to regulation of insulin secretion, emerging human data have raised concern regarding the immutable species differences and the subsequent limitations to translatability.
Method:
This study addresses the challenges and opportunities in T2D research and how it can be transformed with existing and novel future methodologies using human-based data acquisition.
Results:
We discuss how state-of-the-art in vitro, in vivo, and in silico technologies can be used to delineate disease mechanisms and drug responses. Insulin resistance and pancreatic dysfunction can be studied at many levels from gene transcription/regulation to functional proteomics to cellular signaling and cell-cell interactions to organ cultures and whole-body glucose regulation to human population–based studies.
Conclusion:
T2D research is in dire need of a paradigm shift, with increased emphasis on human-based data acquisition and decreased reliance on animal models. With continued support from scientists and funding agencies alike, this is the clear-cut path to address the challenges in the translational barrier and create opportunities for diabetes research and treatment.
Development and Use of a 13C6-Glucose Isotope-Dilution Liquid Chromatography–Mass Spectrometry Method to Determine System Accuracy and Precision of the In Touch Cellular Glucose Meter System
Kimon Angelides, PhD, MBA; David Engler, PhD; Risë Matsunami, PhD
Office of Science and Technology
Livongo Health
Chicago, Illinois
Introduction:
We evaluated the precision and accuracy of the In Touch meter, a new color touch screen and cellular-enabled blood glucose meter, using a new high-throughput, highly precise and accurate 13C6 liquid chromatography (LC) isotope-dilution mass spectrometry (MS) method.
Methods:
An LC-MS assay was designed and validated against National Institute of Standards and Technology standard reference material to an uncertainty of <1%. Blood glucose measurements from the In Touch meter were then referenced to this standard. Glucose and [13C6]glucose were monitored at multiple reaction monitoring transitions of m/z 179→89 and 179→59 for glucose, and 185→92 and 185→61 for the [13C6]glucose used as an internal standard. The accuracy of the In Touch meter was determined across the range of 15–700 mg/dL using venous blood samples spiked with stock solutions of glucose for bias of the individual measured values against the MS-determined values and through a consensus error grid.
Results:
At blood glucose concentrations <75 mg/dL, 100% of the In Touch measurements are within ±8 mg/dL from the true MS reference standard; at blood glucose levels >75 mg/dL, 100% of the In Touch measurements are within ±10% of the true MS reference standard. One hundred percent of the results are within zone A of the consensus grid. Within-run and intermediate precision show coefficient of variation <3.72% between 20 and 40 mg/dL and coefficient of variation <2.21% between 500 and 600 mg/dL. The results show that the In Touch meter exceeds both the ISO 15197:2003 and ISO 15197:2013 standards.
Conclusion:
To our knowledge, these are the first studies of a new glucose meter where precision and accuracy have been measured against an LC-MS reference. The results show that the In Touch meter performs at a high standard and exceeds all current and proposed performance criteria using a stringent reference standard.
Predictors of Nocturnal Hypoglycemia in Patients with Type 1 Diabetes: Effects of Automatic-Threshold-Suspend Feature Activation
Timothy S. Bailey, MD, FACE, CPI; James Thrasher, MD; Frank Schwartz, MD; David C. Klonoff, MD, FACP; Ram Weiss, MD, PhD; Meng Mao, PhD; ASPIRE In-Home Study Group
AMCR Institute Inc.
Escondido, California
Objective:
We aim to 1) identify patient attributes associated with increased risk of nocturnal hypoglycemia (NH) and 2) test the effect of the threshold-suspend (TS) feature in subjects at highest risk for NH.
Method:
The TS feature automatically suspends the insulin pump for up to 2 h in response to crossing a set sensor glucose threshold. The ASPIRE In-Home study was designed to establish the safety and efficacy of the use of TS in patients prone to NH. We performed univariate analyses to identify risk factors for hypoglycemia during the run-in phase and used multivariate analysis to assess the impact of the TS feature.
Result:
A total of 10,984 events (5,027 nocturnal) occurred during the 3-month study phase. All 126 subjects in the control group (aged 16–70 years) and 118 of 121 subjects in the TS group (aged 16–69 years) had at least one NH event. In univariate analysis, baseline A1C <7% (P < 0.001) and >1 manual bolus administration per day (P = 0.04) were associated with higher NH rates, while age, diabetes duration, basal/bolus ratio, and boluses per day were not. In the multivariate model that included all risk factors and treatment assignment, TS was associated with a significantly reduced NH risk (P < 0.001). Baseline A1C <7% (P < 0.001) and delivering boluses manually (P = 0.02) remained significant predictors of NH. The adjusted relative NH event rate in the TS versus the control group was 0.70, indicating a 30% reduction attributable to the TS feature. Prolonged NH events (those longer than 2 h) were reduced by 60%.
Conclusion:
Patients with optimal metabolic control (A1C <7%) and those who deliver boluses without using the bolus calculator are at increased risk for NH, which is significantly ameliorated by using TS.
Accurate Prediction and Avoidance of Hypoglycemia in an Artificial Pancreas Using Spatiotemporal Models
Ayan Banerjee, PhD; Sandeep Gupta, PhD; Yi Zhang, PhD; Paul Jones, MS
Arizona State University
Tempe, Arizona
Objective:
An artificial pancreas (AP) induces nocturnal hypoglycemia with an average of seven hypoglycemic events in a 24-h period. Some of the primary causes include 1) a lack of accurate predictive models of glucose-insulin interactions in the control algorithm that take into account the delay in insulin action and 2) a lack of integration testing of control algorithms, glucose meters, and insulin pumps. In this work, we seek an accurate model of glucose-insulin action that can be used for integration testing.
Method:
We use a novel modeling construct of spatiotemporal hybrid automata (STHA) to model AP systems. In STHA, the discrete operating modes of an AP such as bolus, correction, meal, and braking are modeled using discrete states. In each state, the glucose-insulin interaction is modeled using a combination of 1) spatiotemporal diffusion model of a drug, which captures the variation of glucose meter readings with respect to the installation location of glucose meter, and 2) nonlinear models of glucose-insulin interaction. We tested our model for 10 different patient profiles using the type 1 diabetes simulator and evaluated the hypoglycemic excursions.
Result:
The STHA model was found to avoid all hypoglycemic excursions that were originally induced in the simulated patients. However, the STHA model still could not avoid hyperglycemic events.
Conclusion:
Our study results show that the accuracy of AP predictive model(s) can be improved by using STHA to model the nonlinear insulin-glucose interaction, which may result in reduced hypoglycemic episodes. The STHA model capturing insulin-glucose interaction can be constructed as a combination of diffusion equations and drug interaction models.
A Synthetic Fusion Protein for Epigenetic Control of Pancreatic Cell Function
David K. Barclay; Karmella A. Haynes, PhD
School of Biological and Health Systems Engineering
Arizona State University
Tempe, Arizona
Objective:
Type 1 diabetes is caused by the failure of beta cells to produce sufficient insulin in pancreatic tissue. Many recent treatment strategies have focused on replenishing functional beta cells. We focus on a strategy to manipulate the DNA-protein complex known as chromatin to transdifferentiate glucagon-producing alpha cells into insulin-producing beta cells. Chromatin modifications store a secondary level of information that determines the expression states of developmental genes that control cell fates.
Method:
Others have demonstrated that indirectly disrupting gene-silencing chromatin induces a beta-cell gene expression profile in alpha cells. In order to directly target chromatin-silenced genes for specific activation, we used a synthetic fusion protein called PcTF, which binds to the histone H3 lysine 27 trimethylation mark (H3K27me3). Previously, we have demonstrated that the PcTF protein reactivates epigenetically silenced target genes in human bone-derived cells. We aim to use PcTF to regulate genes in pancreatic cells in order to advance the treatment of diabetes.
Result:
We transfected mouse TC1 alpha cells with PcTF-expressing plasmid DNA and confirmed expression by quantitative reverse-transcription PCR (qRT-PCR). Thus far, we have observed upregulation of H3K27me3-associated genes MAFA and IAPP via qRT-PCR in PcTF-expressing cells, compared with nonexpressing cells. Our preliminary bioinformatics analysis has identified ~4,950 potential PcTF target genes in human alpha cells through the analysis of chromatin immunoprecipitation data.
Conclusion:
Taken together, our proof-of-concept results from mouse cells and bioinformatics analysis of human chromatin suggest the translational potential of the synthetic PcTF protein. PcTF could be an effective agent for the epigenetic conversion of human alpha cells to insulin-producing cells.
Service Evaluation of the Diabetes Insulin Guidance Service (DIGS)
Eran Bashan, PhD; Israel Hodish, MD, PhD
Hygieia Inc.
Ann Arbor, Michigan
Objective:
Insulin therapy has been available for almost a century. However, its success rate is still disappointing where the majority of users sustain destructively elevated A1C. The key element essential for effective and safe insulin therapy, even for simple regimens like basal or premixed insulin, is frequent dosage titration to overcome constant variations in insulin requirements. In reality, dosage titration is done sporadically during clinic visits. Hygieia Inc. has developed a scalable solution to this problem. A diabetes-nurse service improves glycemic control without overburdening the health system. The service relies on a device called d-Nav, which provides patients with an insulin dose recommendation for each injection while using the device to monitor glucose. By analyzing stored glucose trends, in a similar approach that providers use during clinical encounters, d-Nav titrates insulin dosage without care providers’ constant supervision. A service evaluation was conducted at the Diabetes Consult Ward, Ulster Hospital, Dundonald, Belfast, Northern Ireland.
Method:
Patients aged ≥21 years with A1C ≥7.0% (53 mmol/mol) who were receiving insulin therapy for at least 1 year were recruited to the service and were followed for 1 year. A1C and information on hypoglycemia were collected during scheduled 3-monthly clinic visits.
Results:
A total of 126 patients were recruited, while 96 completed the evaluation. The mean (±SD) A1C for the 96 active users decreased from 9.2 ± 1.4% (77 ± 15 mmol/mol) at baseline to 7.8 ± 1.2% (62 ± 13 mmol/mol) at the 3–5-month clinic visit and to 7.5 ± 1.2% (58 ± 13 mmol/mol) at the 6–12-month clinic visit (P < 0.05). The frequency of minor hypoglycemia (glucose <3.7 mmol/L or <65 mg/dl) was 0.6 ± 0.2 events per week.
Conclusion:
The Diabetes Insulin Guidance Service (DIGS) was shown to be an effective, scalable solution for the management of insulin-treated patients without overburdening the health care system.
Acetaminophen Interference with Subcutaneous Glucose Sensing
Ananda Basu, MD; Simmi Dube, MD; Barbara Norby, RN; Cheryl Shonkwiler, RN; Roy Dyer, PhD; Michael Slama, BS; Thomas Peyser, PhD; Rita Basu, MD
Division of Endocrinology, Diabetes, Metabolism, and Nutrition
Mayo Clinic
Rochester, Minnesota
Objective:
We evaluate the effect of Tylenol (acetaminophen) ingestion on subcutaneous glucose sensing by continuous glucose monitors (CGMs) during a fasting state in nondiabetic healthy humans.
Method:
Five healthy participants had a combination of multiple approved and available CGMs (Dexcom SEVEN PLUS, Medtronic Guardian, Dexcom G4 Platinum) placed 48 h prior to the study visit and calibrated per manufacturer instructions. On the morning of study day, four microdialysis catheters were inserted into subcutaneous abdominal fat adjacent to CGM sensors, and perfusate was started and continued until the end of study. Heated hand vein technique was used to obtain arterialized venous blood at periodic intervals. One gram of Tylenol was administered by mouth at ~8:00 AM. Blood samples and microdialysate collected during the study were analyzed for glucose and Tylenol measurements in plasma and interstitial fluid with commercial automated assays.
Result:
Following Tylenol ingestion, interference in glucose sensing was observed in all CGM devices tested. While plasma glucose concentrations remained at ~90 mg/dL (~5 mmol/L) throughout the study, CGM glucose measurements varied between ~162 and 252 mg/dL (~9 to 14 mmol/L). The temporal profile of CGM interference followed Tylenol concentrations measured in interstitial fluid.
Conclusion:
This is the first study to our knowledge to simultaneously measure the concentration of acetaminophen and glucose in the interstitial fluid using currently available CGM devices. We found that typical dosages of Tylenol do interfere with CGM performance. Future studies are intended to evaluate other potential pharmacologic interference with CGM performance. As CGM accuracy and reliability continue to improve overall, such data will be essential for regulatory approval of therapeutic dosing claims for standalone CGM devices and for artificial pancreas systems.
Acoustic Methods for Transmitting Blood Glucose Measurements
Yiftah Ben Aharon, MSc; Dov Moran, ENG; Roee Tuval, MR
GlucoMe Ltd.
Yarkona, Israel
Objective:
We aim to find a way to transfer blood glucose measurements from a blood glucose monitor to a mobile device without modifying the hardware of the blood glucose monitor or the mobile device.
Method:
We have developed an acoustic method that uses the existing hardware in order to generate acoustic signals. The method was tested in various environments and distances (total of 1,260 tests).
Result:
No faulty measurements were received. The reception rate below 6 feet was 100%. The reception rate below 80 db was 100% as well. In very loud environments (above 80 db) and at high distances (above 6 feet), the reception rate was ~95%. Using a retransmission mechanism improved the rate to 99%.
Conclusion:
The proposed acoustic method was found to be efficient, robust, and valid.
Improved Automated Glucose Clamps Using Algorithm Variations with the ClampArt Device
Carsten Benesch, PhD; Judith Haensler, PhD; Tim Heise, MD
Profil
Neuss, Germany
Objectives:
The pharmacodynamic effects of antidiabetes drugs are usually characterized in glucose clamp experiments where variable glucose infusion rates (GIRs) keep blood glucose (BG) concentrations as closely as possible to a predefined target level. For a high glucose-clamp quality, low BG and GIR variability are desirable. In automated clamps, GIRs are calculated by implemented algorithms such as the Biostator algorithm published in 1982. For ClampArt, a new, CE-marked state-of-the-art glucose clamp device, we improved this algorithm to reduce the oscillations of both BG and GIR.
Methods:
We first improved the algorithm by numerical simulations of glucose clamps (in silico). With the results of the simulations, we started in vitro experiments using a ClampArt device and a container with water and glucose as a “test subject.” Finally, we performed a small number of in vivo clamps with real subjects using ClampArt with the optimized algorithm.
Results:
All three approaches showed that the improved algorithm reduces the amplitude of oscillation of both BG and GIR. The in silico optimization of the algorithm showed a reduction in the amplitudes of the oscillations in BG (43%) and GIR (66%). The in vitro experiments show similar results, with a reduction of oscillations of both BG (61%) and GIR (76%). Finally, we observed in the in vivo experiments reduced amplitudes of oscillations for BG (35%) and GIR (65%).
Conclusions:
In silico, in vitro, and in vivo experiments showed that an improved algorithm reduced BG oscillations by 35– 61% and GIR oscillations by 65–76%. This optimized algorithm will now be implemented in ClampArt and used for future glucose clamp studies.
Effect of Local Warming on Regular Human Insulin Pharmacokinetics
Gabriel Bitton, PhD; Dmitry Feldman, MD; Tal Alon, RN; Ron Nagar, MSc; Andreas Pfützner, MD, PhD; Itamar Raz, MD, PhD
Insuline Medical
Petach Tikvah, Israel
Background:
Regular human insulin (RHI) is still widely used in many countries, although rapid-acting insulin is available. In Germany, ~40% of multiple daily injection patients use RHI, as well as 70% of the multiple daily injection patients in Russia and in many developing countries. It is therefore worth exploring methods to improve RHI insulin therapy. The InsuPad device was shown to improve insulin therapy when used with rapid-acting insulin analogs. The device applies local controlled heat to the skin in the vicinity of the injection site, which promotes local blood perfusion and enables faster and more consistent absorption of the insulin from the injection site. The pharmacokinetic (PK) and pharmacodynamic (PD) profiles of RHI were found to be accelerated when subjects were in a hot environment such as in a sauna after the injection. However, a recent study showed no effect of local heating on PK and PD properties of RHI. In this study, we report of the effect of local heating on the PKs and PDs of properties of RHI when the heating profile is adapted to RHI time contestants.
Method:
Type 2 diabetes patients on basal-bolus insulin therapy were admitted for a meal tolerance test after an overnight fast. Subjects consumed standardized liquid meal and injected 0.2 units/kg of RHI. The study was conducted twice: without the InsuPad device (control) and with the InsuPad device activated for 100 min of intermittent heating (test). Blood samples for insulin measurements were taken from a venous line during the study.
Result:
Seven type 2 diabetes subjects participated in the study, aged 62 ± 7.9 years, HbA1c of 8.2 ± 1.3, BMI of 31 ± 4.0 kg/m2, and diabetes duration of 20 ± 3.7 years. Maximum insulin concentration was higher (55.4 ± 23.6 vs. 85.3 ± 31.1 mU/L; P = 0.056) and area under the curve of the insulin concentration during the first hour postmeal was higher (21.6 ± 16.3 vs. 39.1 ± 16.7 mU/L/h; P = 0.048) when the InsuPad device was used.
Conclusion:
The results from this feasibility study shows that the RHI PK profile can be improved by using local warming of the injection site. We intend to study the effect of local heating on postmeal glucose levels under real-life conditions.
Early Results Suggest Effectiveness of an Online Diabetes Prevention Program
Gladys Block, PhD; Robert J. Romanelli, PhD, MPH; Kristen M.J. Azar, RN, MSN/MPH; Torin J. Block, BA; Clifford H. Block, PhD; Latha Palaniappan, MD, MS
NutritionQuest
Berkeley, California
Objective:
The objective of this research is to test the effectiveness of Alive-PreventDiabetes (Alive-PD), a 1-year e-mail web program for prediabetic patients, in a randomized controlled trial.
Method:
Alive-PD uses a variety of behavioral approaches, including both long-term and weekly individualized goal setting, logging, diabetes-related content, quizzes, team-based social support, automated coaching, and other activities. The program is entirely automated, with no human coaches. The trial is conducted in collaboration with the Palo Alto Medical Foundation Research Institute (PAMF). PAMF members were identified through electronic health records, and prediabetes was confirmed through point-of-care testing. Randomization was by computer code to immediate or 6-month delayed entry. Clinic visits are conducted at baseline and 3, 6, 9, and 12 months. Primary end points are change in HbA1c and glucose. Secondary end points are change in weight, lipids, and blood pressure. Treatment effects are analyzed using regression methods and controlling for baseline level. The final analysis will use intention to treat.
Result:
Treatment groups are balanced on baseline factors. Characteristics of the 348 subjects are as follows: male 68%, nonwhite 33%, mean age 55 years, BMI 31 (59.4% obese), fasting glucose 109.1 mg/dL, and HbA1c 5.6%. Among the 178 who have reached the 3-month time point, treatment arms differ significantly in weight change (P < 0.01; intervention −7.8 pounds; control −2.3 pounds) and glucose change (P < 0.05; intervention −2.2 mg/dL; control +0.5 mg/dL). Decrease in HbA1c was greater in intervention (P = 0.12 for treatment group comparison).
Conclusion:
These results suggest that Alive-PD is an effective program to help prediabetes patients slow or stop their progression to diabetes. It will be available through organizations and practitioners and to individuals.
Comparison of Insulin Pump Therapy and Multiple Daily Injection Therapy in Suboptimally Controlled Type 2 Diabetes: The OpT2mise Trial
Bruce W. Bode, MD; Ronnie Aronson, MD; Ignacio Conget, MD; Yves Reznik, MD; Sarah Runzis, MSc; John B. Welsh, MD, PhD; Larisa Yedigarova, MD, PhD; John Shin, PhD; OpT2mise Study Group
Atlanta Diabetes Associates
Atlanta, Georgia
Objective:
OpT2mise (NCT01182493) aimed to compare insulin pump therapy with multiple daily injection (MDI) therapy in suboptimally controlled type 2 diabetes (T2D).
Method:
Subjects with T2D using MDI were enrolled in a 2-month run-in period for insulin dose optimization. After this, subjects with A1C values≥8% and ≤12% were randomly assigned (1:1 ) to switch to pump therapy or continue MDI for 6 months. The primary end point was the between-group difference in mean A1C change from baseline to the end of the randomized phase in the intention-to-treat population. Continuous glucose monitoring (CGM) studies were conducted at baseline and 6 months.
Result:
A total of 331 subjects were randomized (163 to MDI and 168 to pump therapy). Randomized subjects were 45.6% female and were 56.0 ± 9.6 years old, and their A1C was 9.0 ± 0.8% (mean ± SD). At the end of the randomized phase, subjects in the pump therapy group had significantly greater A1C reductions than subjects in the MDI group (−1.1 ± 1.2 vs. −0.4 ± 1.1%, respectively; P < 0.001). CGM data showed that the pump therapy group had greater reductions in 24-h mean glucose (P = 0.006) and in time spent in hyperglycemia >180 mg/dL (P = 0.0007) than the MDI group. The percentage of subjects with A1C <8.0% was 57% in the pump therapy group versus 27% in the MDI group (odds ratio = 1.9; 95% CI 1.47 to 2.46; P < 0.001). At study’s end, total daily insulin dose was 20.4% lower in the pump therapy group than in the MDI group. No ketoacidosis occurred. One episode of severe hypoglycemia occurred in the MDI group.
Conclusion:
Compared with MDI, insulin pump therapy allowed subjects with suboptimally controlled T2D to achieve lower A1C values, with less insulin, without impacting safety or hypoglycemic exposure.
Buccal Delivery of the Glucagon-Like Peptide Agonist Exenatide Using ArisCrown Technology
Paolo Botti, BSc, MSc, PhD; Sylvie Tchertchian, BSc, MSc, PhD; Doriane Theurillat, BSc; Frédéric Preitner, BSc, MSc, PhD; Patrice Nury, BSc, PhD; Andrew Parker, BSc, MBA, PhD
Arisgen SA
Geneva, Switzerland
Objective:
Oral delivery of therapeutic peptides such as insulin and GLP-1 agonists represents a significant unmet need in the management of diabetes. Arisgen has developed a novel lipid-based formulation technology that can concomitantly promote permeation and overcome poor peptide oral bioavailability. This technology is called ArisCrown and is based on the selective and reversible masking of peptide functional groups by novel and proprietary biodegradable cyclic compounds. In this study, we evaluate the potential of ArisCrown technology to deliver the GLP-1 analog exenatide via the sublingual route of administration in mice and nonhuman primates.
Method:
Exenatide was administered at t -30 min to anesthetized mice or cynomolgus monkeys either via injection (intraperitoneal/subcutaneous) or formulated with ArisCrown (designated ARG011) and delivered under the tongue. At t-zero a glucose bolus was administered and glucose and insulin levels determined (glycometer and ELISA, respectively) over the following 2 h. In addition, food intake was monitored separately in mice for 3 h after waking.
Result:
We have demonstrated in mice that sublingual administration of ArisCrown Exenatide (ARG011) is able to control glycemia, as determined by glucose and insulin regulation, as well as food intake, in a manner equivalent to intraperitoneal injection of unformulated peptide. Dose response experiments demonstrate a >30-fold shift in EC50 when comparing sublingual to injected peptide. In addition, in monkey studies, we have demonstrated that ARG011 combined in a buccal patch is also able to control PD markers of glycemia equivalent to subcutaneously injected peptide.
Conclusion:
ARG011 is a novel formulation of exenatide combined with ArisCrown chemistry that is as effective as injected exenatide at regulating glycemia. ARG0011 represents a clinical candidate for the buccal delivery of exenatide.
Designing a Decision Support System to Prevent Postexercise Hypoglycemia in Type 1 Diabetes
Najib Ben Brahim, MSc; Jerome Place, MSc; Eric Renard, MD, PhD; Marc D. Breton, PhD
Center for Diabetes Technology
University of Virginia
Charlottesville, Virginia
Objective:
We develop a predictive hypoglycemia alert system for type 1 diabetes patients intending to exercise.
Method:
A metadata analysis was conducted over 59 type 1 diabetes patients from 4 different studies in the U.S. and France (37 men and 22 women; 47 adults, weight 71.41 ± 10.56 kg, age 42 ± 9.7 years; 12 adolescents, weight 60.7 ± 12.53 kg, age 14 ± 1.4 years). All participants had physical activity between 4:00 and 5:00 PM at a mild to moderate intensity for ~30 min. The data set was split two-thirds to one-third for training and testing, respectively. Main factors explaining the glucose dynamics in presence of exercise were identified using stepwise linear model regression. Those factors were then included in the development of a logistic regression model that served as the foundation for an exercise-induced glycemic state predictor.
Results:
Main factors explaining the glucose dynamics in presence of exercise at a 0.05 significance level were blood glucose at the beginning of exercise (BG0), the ratio of insulin on board over total daily insulin, and initial glycemic slope. The alert system was able to detect at 92% of postexercise low glucose (19% false alarms) on the training set and 86% (22% false alarms) on the validation set.
Conclusion:
The main factors explaining glucose dynamics in the presence of exercise have been identified. A classifier has been tested to predict the glycemic state induced by mild to moderate physical activity. A decision support system can be designed to better inform type 1 diabetes patients on their projected glycemic state after immediate exercise, allowing the patient to delay exercise (reducing insulin on board) or consume carbohydrate if at risk of hypoglycemia.
A Novel Imaging Technique for Monitoring Stability of Encapsulated Islets
Eric M. Brey, PhD; Alyssa A. Appel, MS; Jeffery C. Larson, BS; Alfred B. Garson III, PhD; Huifeng Guan, BS; Zhong Zhong, PhD; Emmanuel C. Opara, PhD; Mark A. Anastasio, PhD
Illinois Institute of Technology
Chicago, Illinois
Objective:
The objective of this research is to investigate techniques based on X-ray phase contrast (XPC) for quantitative three-dimensional (3D) imaging of encapsulated islets and evaluation of local tissue response to the implanted materials.
Method:
Alginate beads encapsulating islets were synthesized and implanted in a rodent omentum pouch model. Explanted samples were imaged with both synchrotron and benchtop XPC imaging systems. The synchrotron system is an analyzer method known as multiple image radiography. Computed tomography data were acquired at 500 tomographic views over 180° at X-ray energy of 20 keV at 11 angular analyzer positions over 8 µradians. The benchtop X-ray system used an in-line imaging scheme where the detector is placed 155 cm behind the sample in order to capture a single image with edge enhancement at material interfaces. For computed tomography scans, 186 projection views were collected over a 186° angular range.
Result:
The 3D structure of beads could be clearly identified in XPC images, allowing quantification of volume and shape. Both stable and failed beads could be identified, with failed beads showing tissue invasion into the bead structure. Soft tissue detail corresponded well with histological findings. Adipose tissue and inflammatory tissue were clearly identified. The foreign body encapsulation response could be identified around the beads and layer thickness quantified. These findings were translated to an in-laboratory system. Beads could be identified with identification of soft tissue structure similar to the synchrotron multiple image radiography system.
Conclusion:
Alginate samples are invisible in traditional absorption-based X-ray images. XPC techniques enable quantitative 3D imaging of material structure and soft tissue response. These results suggest the significant potential use of XPC techniques to monitor encapsulated islets.
Correlation of Lens Autofluorescence and Signs of Diabetic Retinopathy in Wide-Angle Retinal Photographs
Frederick Cahn, PhD; John McIntyre, OD; Jerome Sherman, OD, FAAO; John Burd, PhD; Paul Williams, MS; Shardendu Mishra, MBBS; Keith Ignotz, MBA
Freedom Meditech Inc.
San Diego, California
Objective:
The physiological basis of lens autofluorescence is the accumulation of advanced glycation end products (AGEs) in the lens protein as the subject ages. Because glycation is initiated by free reducing sugars, diabetic subjects have an accelerated accumulation of AGEs in tissues, which is postulated to be a key mechanism for the ocular and other complications of diabetes. This mechanism suggests that a correlation between lens autofluorescence and observable signs of diabetic retinopathy should exist in a patient population independently of whether the subject is known to have diabetes.
Method:
Patients were tested for lens autofluorescence by the ClearPath DS-120 Lens Fluorescence Biomicroscope and by wide-field retinal photography using a nonmydriatic digital fundus camera. Lens autofluorescence was adjusted for subject age. Anomalies attributable to diabetic retinopathy in the retinal photographs were classified as no retinopathy, nonproliferative diabetic retinopathy, and proliferative diabetic retinopathy. Nonproliferative diabetic retinopathy was further classified as mild, moderate, or severe.
Result:
The signs of diabetic retinopathy in this patient population appear to be correlated with lens autofluorescence measurements.
Conclusion:
A correlation between lens autofluorescence and signs of diabetic retinopathy supports the role of AGEs in diabetic retinopathy and suggests that lens autofluorescence may have clinical application in the management of diabetes.
Comparison of Distribution of Lens Autofluorescence in Patient Populations from Different Optometric Practices
Frederick Cahn, PhD; Agnes Palys, OD; Carmen Castellano, OD; John McIntyre, OD; Robert Bauman, OD; John Burd, PhD; Paul Williams, MS; Shardendu Mishra, MBBS; Keith Ignotz, MBA
Freedom Meditech Inc.
San Diego, California
Objective:
Lens autofluorescence is correlated with subject age and with diabetes status. The predictive value of a high autofluorescence measurement for a diagnosis of diabetes depends on the distribution of lens autofluorescence in a population of subjects without known clinical disease. Our objective is to better characterize the lens autofluorescence distributions of patient populations from optometric practices in different geographical locations.
Method:
Since commercial introduction of the ClearPath DS-120 Lens Fluorescence Biomicroscope, several commercial sites have collected data on more than 250 patients each. We obtained de-identified lens autofluorescence data for patients from four sites with autofluorescence data on 250 or more subjects and compared their fluorescence distributions. Data were age-adjusted by subtracting the expected lens autofluorescence according to the age of the subject. Data were also adjusted for a small apparent effect of sex on fluorescence; no correlation of fluorescence with height, weight, or BMI was observed in this data set.
Result:
The independent samples Kruskal-Wallis test showed significant differences among sites of the distribution in sex, height, weight, and BMI. However, a correlation with lens autofluorescence was observed only with sex in this data set. After adjustment for age and sex, lens autofluorescence distributions of fluorescence remained significantly different among the sites, by the independent samples Kruskal-Wallis test.
Conclusion:
Receiver-operating-characteristic curves and predictive values we previously estimated from our limited patient population may be generalizable to patients seen in the ordinary course of optometric practice, but positive and negative predictive values for diabetes can be different in different practices.
Predicting Sleep and Meals for Improved Artificial Pancreas Performance
Fraser Cameron, PhD; Bruce A. Buckingham, MD; Darrell M. Wilson, MD; David M. Maahs, MD; B. Wayne Bequette, PhD
Rensselaer Polytechnic Institute
Troy, New York
Objective:
Foregoing meal announcement introduces uncertainty into artificial pancreas–based insulin dosing. Better predictions are required to counter this uncertainty. These predictions can be improved to better predict meals and sleep using empirical data from the massive National Health and Nutrition Survey (473,000 meals) and the American Time Use Survey (137,000 sleep events). This work uses these data to develop five sleep and meal prediction methods and test them on 5,000 randomly selected days of sleep and meal patterns.
Method:
The methods are as follows: method 1, the time of day (TOD), the current sleep state (CSS), and the time of the last meal (TOLM); method 2, CSS; method 3, value and duration of the CSS and TOLM; method 4, TOD; and method 5, no information. Method 5 serves as the neutral comparator since it offers no advantage. Methods 2, 3, and 5, since they do not receive the TOD, are best able to adapt to shift work or abnormal schedules. Methods 2 and 5 additionally cannot predict impending sleep since they have no method of duration of CSS or TOD.
Results:
For sleep prediction, methods 1–4 greatly improve on method 5. Method 2 suffered for longer prediction horizons. Method 4 suffers for shorter predictions. Method 3 shows slightly better improvement for shorter prediction horizons. Methods 1–4 are beneficial in more than 99% of days. For meal prediction, methods 1–4 improve on method 5. Method 2 and 5 show slightly decreased accuracy. All but method 2 are beneficial for more than 98% of days.
Conclusion:
These methods, especially method 3, improve the veracity and speed of sleep and meal detection and prediction while also being able to adapt to abnormal schedules.
Influence of Glucose Set Point and Premeal Bolus on Proportional-Integral-Derivative–Based Closed-Loop System Performance
Casey Card, BA; Jennifer L. Sherr, MD, PhD; Eda Cengiz, MD, MHS; Camille I. Michaud, MD; Neha Patel, DO; Miladys M. Palau-Collazo, MD; Lori Carria, BS; Eileen Tichy, MS; William V. Tamborlane, MD; Stuart A. Weinzimer, MD
Department of Pediatrics
Yale University
New Haven, Connecticut
Objective:
Proportional-integral-derivative (PID)–based algorithms are employed in the design of closed-loop (CL) systems, but the optimal glucose set point and need for manual premeal boluses (PMBs) have not been established. We analyzed glucose profiles from five CL protocols using a PID controller with insulin feedback to explore these questions.
Method:
Data for 41 subjects (23 female; age 19 ± 4 years; A1C 7.3 ± 0.7%) were pooled to assess the effect of glucose set point (100 mg/dL [G100] vs. 120 mg/dL [G120]) and use of PMB (PMB+ or PMB–) on CL system performance. Mean blood glucose (BG); BG levels within (70–180 mg/dL), below (<70 mg/dL), or above (>180 mg/dL) target range; prandial BG excursion and prandial BG area under the curve; and hypoglycemia (BG <60 mg/dL) and hyperglycemia (BG >250 mg/dL) were calculated for each condition.
Result:
G100 subjects had a significantly lower average BG than G120 (132 ± 33 vs. 146 ± 33 mg/dL; P < 0.01). G100 subjects spent significantly more time in the target range (83.8 ± 9.8 vs. 74.9 ± 10.5%; P < 0.02) and less time above target range (13.0 ± 9.6 vs. 23.7 ± 11.4%; P < 0.01) than G120 subjects, with no significant increase in time below range (3.2 ± 3.5 vs. 1.4 ± 2.4%; P = 0.07) or hypoglycemia (1.1 vs. 0.3%; P = 0.11). Time in hyperglycemia was not different (1.7% G100 vs. 3.6% G120). PMB+ subjects had smaller prandial BG excursion
and BG area under the curve than PMB– subjects (94 vs. 105 mg/dL and 207.4 vs. 254.2 mg⋅h/dL, respectively), but these differences did not reach statistical significance.
Conclusion:
Using a sensor glucose set point of 100 mg/dL rather than 120 mg/dL in a PID-based CL system results in a lower average BG and improved time in target range without a significant increase in hypoglycemia. The use of a small PMB did not equivocally demonstrate benefit in mitigating prandial glycemic excursions, although further studies are needed to definitively answer this question.
Leveraging Glooko for Real-Time Glucometer Data Sharing Between Patients with Type 1 Diabetes and Their Providers: A “Smart” System
Mark A. Clements, MD, PhD, CPI, FAAP
Children’s Mercy Hospital
Kansas City, Missouri
Objective:
We aim is to improve data sharing and communication between patients with diabetes and their health care providers.
Method:
We leverage the Glooko web service to stream glucometer data into the Cerner Electronic Health Record. Provider-defined alerts are created to notify the diabetes care team when a clinically actionable pattern emerges in the blood glucose data stream. Alerts based on blood glucose data trends are created via event processing engines. Such engines permit the identification of complex trends in “in-flight” data without creating transactional burden on medical record databases.
Result:
Multiple custom alerts are demonstrated, including alerts for postprandial hyperglycemia at breakfast, lunch, and dinner; decreases in the frequency of blood glucose monitoring; mild to moderate hypoglycemia occurring repeatedly within the same 3-h window; and increasing average daily glucose. The number of days over which the event of interest must occur to create an alert can be customized by provider or by patient. For instance, an alert can be generated if postbreakfast blood glucose is persistently 40 mg/dL greater than the preprandial level for 4 consecutive days. Alternately, the percentage change in a specific metric can be used to trigger an alert (i.e., a 20% increase in the average daily glucose observed over a 14-day period compared with the previous 14-day period).
Conclusion:
Glooko is an enabling technology that supports real-time sharing of glucometer data between patients and health care providers. Health care providers can provide “value added” to Glooko by delivering Glooko data into the electronic health record and leveraging event processing engines to alert the diabetes care team to clinically actionable changes in the quality of glycemic control.
The SMART Pilot Project: The Utility of a Text-Messaging-Based Intervention for Adolescents with Type 1 Diabetes
Suzanne Collier, BA; Alexa Stern, BA; Catherine Gillespie, PhD; Linda Herbert, PhD; Celia Henderson, RN, CDE; Randi Streisand, PhD
Children’s National Medical Center
Washington DC
Objective:
Type 1 diabetes (T1D) is a lifelong condition that is particularly challenging during adolescence. Most adolescents use cell phones; thus mobile health interventions may be an effective method of T1D education and adherence promotion. The aims of this study are 1) to examine the utility of a text-message-based intervention program to promote self-care skills and 2) to determine which adolescents are most likely to benefit from the intervention.
Method:
Twenty-three participants (mean age = 15.13 years; range = 13–17 years) were enrolled in the SMART Project, a 6-week text-message program during which adolescents received blood glucose prompts and T1D education texts 2–3 times/day. Adolescents completed a demographic/medical history questionnaire, the Diabetes Behavior Rating Scale, the Self-Care Inventory, and the International Physical Activity Questionnaire at baseline and upon completion of the intervention. Glucometer data regarding the past 30 days were downloaded, and medical charts were reviewed for HbA1c from the clinic appointments immediately preceding and following the intervention.
Result:
Ninety-six percent of participants completed the follow-up questionnaires. Potentially due to a limited sample size, there were no statistically significant differences in any of the measures between the two time points. However, the intervention did appear to be useful among specific subgroups of adolescents, such as those who reported better self-care and physical activity levels at baseline, such that they had improved blood glucose and HbA1c levels after the intervention.
Conclusion:
A text -based intervention appears feasible for teens with T1D, yet more research is needed to understand for whom such interventions would be most beneficial. Mobile health interventions may be a useful method to provide health reminders and education to certain subgroups of adolescents with T1D.
Cardiovascular-Risk-Associated Qualitative Pathways in Type 2 Diabetes Patients
Arianna Dagliati, MS; Lucia Sacchi, PhD; Riccardo Bellazzi, PhD
Biomedical Informatics Laboratories “Mario Stefanelli”
Univeristy of Pavia
Pavia, Italy
Objective:
Clinical and administrative data can be integrated to monitor patients’ behavior and clinicians’ actions in order to show how clinical processes are actually executed. In this context, the main goal of our approach is to identify significant behavioral patterns; hence the best care pathways for a certain population can be recognized. To this end, we propose to jointly use temporal and process-mining techniques to extract the most frequent health care pathways. We implemented this framework on a 1.020 type 2 diabetes patient data set collected by the Pavia Healthcare Agency.
Method:
As the first task, we stratified patients according to their cardiovascular risk to mine the most frequent behaviors for each risk class. We identified three classes—high, moderate, and low—through an algorithm derived from the Framingham index. In the second phase, we exploited knowledge-based temporal abstractions on blood glucose control aimed at representing clinical information in the form of qualitative interval-based description. Within the last stage, we applied process-mining techniques to gain insights into patients’ histories. Our algorithm allows extracting the most frequent blood glucose paths for each risk class and consequently enriches the mined flow with administrative information about therapies related to diabetes.
Result:
One of the most interesting results comes from the comparison of high and low risk classes. When histories start with very high blood glucose values, in the low risk class, they mainly end in the same state or worse, while in the high risk class, it is possible to recognize improvement paths in more than half the cases.
Conclusion:
This work tackles the major challenges we faced managing complex clinical and administrative temporal data to identify and compare clusters of relevant health care pathways.
Pilot Baseline Results from DeviceChoices.com: A Diabetes Device Patient Registry
Aaron Davis, MBA; Damon Tanton, MD
Metabolic Markets LLC
San Clemente, California
Objective:
We assess the feasibility of a web-based patient registry to collect and evaluate data about diabetes device utilization in a real-world setting.
Methods:
A patient registry portal was built using a readily available, secure technology platform. Baseline and 90-day follow-up survey instruments were developed from the best practices of clinical trials and patient-reported outcomes studies. One pilot community-based clinic was selected to refer device-using patients to the registry website DeviceChoices.com. The clinic was provided materials to describe the registry project to patients. Patients enrolled via the secure interface and completed the baseline 36-question survey.
Results:
In the June 2014 pilot period, 10 patients completed the baseline survey (50% male and female; 80% type 1 diabetes diagnosis; average age across all users 38 years; average HbA1c across all users 7.6%, with a of range 5.4–10.3%; 30% reported a severe hypoglycemic event in the last 90 days). Pump utilization by manufacturer was Tandem t:slim 50%, Insulet OmniPod 30%, Asante Snap 10%, and Medtronic MiniMed 10%. Eighty percent of registrants had previous experience with an insulin pump, usually Medtronic (6/8). Registrants reported an average 5.3 in response to diabetes treatment satisfaction (scale 1 to 7).
Conclusions:
During the pilot period, the DeviceChoices diabetes device patient registry platform demonstrated the ability to collect important data about the real -world patient experience of those using insulin pumps and continuous glucose monitors. The pilot baseline data characterizes insulin-pump-using patients with data not previously available. Further expansion of the DeviceChoices platform will increase the understanding of the value specific devices contribute to diabetes management.
A Smartphone-Based Carbohydrate Estimation System Using Computer Vision Methods: A Feasibility Study
Joachim Dehais, BSc, MSc; Marios Anthimopoulos, PhD, MSc, BSc; Sergey Shevchik, PhD, MSc, BSc; Ransford Botwey, MSc, BSc; Prosper Fiave, MSc, BSc; David Duke, PhD, MSc, BSc; Alan Greenburg, PhD, MSc, BSc; Peter Diem, MD; Stavroula Mougiakakou, PhD, MSc, BSc
Diabetes Technology Research Group
Graduate School for Cellular and Biomedical Sciences
University of Bern
Bern, Switzerland
Objective:
Carbohydrate (CHO) counting is essential for individuals with type 1 diabetes. The scope of the GoCARB is the automatic and near-real-time estimation of a meal’s CHO for diabetic patients based on computer vision and smartphone technologies.
Method:
The system is developed given the requirements for error in CHO estimation less than 20 g and minimum user interactions. The user places a credit card–sized reference object next to the meal and acquires two images using a smartphone from different viewing angles. The graphical user interface guides the user to choose the optimal angles based on the built -in smartphone sensors (accelerometer and gravity sensor). Then a series of computer vision steps is executed providing the types and volumes of the foods present, which are used along with nutritional databases for the CHO estimation.
Result:
The Android prototype was evaluated using 20 real meals of known CHO content under controlled lighting conditions. The tested meals were included in the used databases, the plate used was circular and shallow, and the food items were of irregular shape and not occluded. The first image was acquired above the plate at a distance of 30–40 cm, while the second at 30 degrees from the vertical axis crossing the center of the dish. The system was able to estimate the CHO content with an average error of less than 10 g/meal and an average execution time below 15 s.
Conclusion:
The feasibility study in a laboratory environment using GoCARB and real meals was promising. Currently, the system is used by adult individuals with type 1 diabetes at the Bern University Hospital within the framework of a preclinical evaluation trial.
Gut Absorption Rates for Glucose in Term and Premature Neonatal Intensive Care Unit Infants Using Continuous Glucose Monitor Data
Jennifer Dickson, BENG (Hons); Matthew Signal, PhD; Balázs Benyó, PhD; József Homlok, MSC, Gábor Marics, MD; Geoff Chase, PhD
Department of Mechanical Engineering
University of Canterbury
Christchurch, Canterbury, New Zealand
Objective:
STAR is a physiological model-based glycemic control protocol adapted for use in neonatal intensive care, where hyperglycemia is a common complication of stress and prematurity. A two-compartment model that describes stomach-gut and gut-bloodstream glucose transfer of enteral nutrition is used to determine the rate constant for gut-bloodstream glucose absorption (d2) in this unique cohort.
Method:
Continuous glucose monitor data from three term infants and one preterm infant (36 weeks) (cohort median: weight 4 [3.9–4.1] kg, postnatal age 23 [19–26] days) receiving 3-hourly enteral breast milk feeds (one term patient, 20–50 mL; all others 70–100 mL) was used alongside a clinically validated glucose-insulin model to determine d2 using gradient descent parameter identification. The half-life of stomach-gut glucose transport was taken from literature to be 20 min (d1 = –log(0.5)/20 min−1). Statistical comparisons were carried out using the Mann-Whitney test. Patient data were obtained as a retrospective audit from the Semmelweis University, First Department of Pediatrics, Intensive Care Unit.
Result:
Of 160 enteral feeds, 77% converged to a positive d2 value. Median (interquartile range) d2 = 0.020 (0.01–0.03) min−1, corresponding to an absorption half-life of 34 (23–69) min. There was no significant difference in d2 between term infants (P ≥ 0.17). Between the preterm infant and two term infants receiving comparable milk volumes, there was variation in the significance of differences (P = 0.05; P = 0.24), and d2 was higher with higher milk volume (P = 0.01).
Conclusion:
Continuous glucose monitor data have been used with a simple gut model to estimate glucose absorption of enteral feeds in infants. Results suggest that glucose absorption by the gut could vary with milk volume and by prematurity. A larger cohort is required to evaluate trends across gestational/postnatal age and feed volume/type.
Assessing the Relationship between Patient Compliance to Blood Glucose Monitoring and Health-Related Quality of Life
Koenraad Dierick, MSc; Mary McBride, MBA
GfK
Brussels, Belgium
Objective:
The objective of our research was to evaluate whether there is a relationship between patient compliance to blood glucose monitoring (BGM) and health-related quality of life. Moreover, we wanted to understand what drives patients not to be compliant to BGM.
Method:
Data were taken from the GfK ROPER Diabetes program, which captures information direct from diabetes patients across 27 countries on a regular basis. For this specific study, data from a sample of 1,480 diabetes patients living in the U.S. were collected from June to August 2013. Each patient completed a questionnaire comprising some 2,000+ variables, which included the EQ-5D-5L instrument and accompanying visual analog scale. Patients were asked to state their recommended frequency of BGM and what their actual BGM frequency was during the past month. Moreover, patients were asked to explain why they were not compliant to the recommended BGM frequency.
Result:
Not being compliant explained 52% (R² = 0.52) of the variations in health-related quality of life. The main reasons for not being compliant to the recommended BGM frequency were no coverage of strips by the insurance, pain and discomfort related to blood testing, and unwillingness to know the test result. Other less important drivers of noncompliance were inconvenience, issues with food intake, and meter malfunction.
Conclusion:
Meter manufacturers are right when they reckon that BGM is crucial to diabetes patients’ disease management. Manufacturers have been innovating to make the blood testing as convenient as possible. Yet an important driver for not complying to the recommended BGM frequency remains to be pain and discomfort during blood testing. Manufacturers should also continue their efforts to insure coverage of strips by the different insurance providers in the U.S.
White Coat Adherence in Pediatric Insulin Pump Users
Kimberly A. Driscoll, PhD; Yuxia Wang, MPH; Suzanne Bennett Johnson, PhD; Elizabeth Gill, BA; Rebecca Lynch, MS; Haley Stephens, MS; Katelyn Willbur, BS; Nancy Wright, MD; Larry C. Deeb, MD
Florida State University College of Medicine
Tallahassee, Florida
Objective:
Assess occurrence of white coat adherence in pediatric patients with type 1 diabetes who use insulin pumps.
Method:
Data were downloaded from insulin pumps (N = 48; mean age = 13.1 ± 3.2 years; mean A1C = 8.3 ± 1.1%; mean diabetes duration = 8.0 ± 10.4 years; mean pump duration = 4.5 ± 2.7 years) at two clinic visits. Correlations were conducted between days prior to type 1 diabetes clinic visit and frequency of blood glucose monitoring (BGM), carbohydrate inputs, and insulin bolusing. Mixed linear models were used to analyze adherence patterns prior to each clinic visit.
Result:
BGM, carbohydrate inputs, and insulin bolusing were significantly negatively correlated with A1C (P < 0.0001) at visit 1; only BGM was significantly negatively correlated with A1C at visit 2. BGM, carbohydrate inputs, and insulin bolusing were significantly negatively correlated with participant age (P ≤ 0.001). A white coat adherence effect, with BGM frequency, carbohydrate inputs, and insulin bolusing increasing prior to both clinic visits, was found in the mixed models (P < 0.0001).
Conclusion:
This is the first study to use downloaded insulin pump data to demonstrate white coat adherence. Patients and parents may have been motivated to increase adherence to capitalize on clinical recommendations or gain physician approval. They should be encouraged to maintain a consistently high frequency of BGM, carbohydrate inputs, and insulin bolusing.
Adoption of Insulin Pumps and Continuous Glucose Monitors: Patient Perceptions of Utility and Usability
Isabelle Dutil, BASc; Linda Gonder-Frederick, PhD; Bruce Perkins, MD, MPH; Patricia Trbovich, PhD; Joseph Cafazzo, PhD, PEng
Institute of Biomaterials and Biomedical Engineering
University of Toronto
Toronto, Ontario, Canada
Objective:
The objective of this study was to determine the perceptions of adoption and use of insulin pumps and continuous glucose monitors (CGM) by patients through semistructured interviews. Based on these findings, the aim was to make recommendations as to how diabetes technology education can be tailored to lessen the learning curve and burden of care at the initial stages of pump or CGM adoption.
Method:
The study was comprised of two phases, each phase devoted to one of the two diabetes technologies being examined. In both the insulin pump and the CGM studies, two types of patients were recruited: the first group included patients currently using the technology, while the second group had considered adopting the technology previously but ultimately rejected it. These groups allowed the research team to amass a larger scope of perceived barriers and benefits of the technologies being evaluated.
Result:
Four themes were identified through the interviews conducted in the insulin pump phase (n = 12). One prominent theme among both users and nonusers was “value add.” Participants discussed their perception of the pump being less beneficial than expected, both from software and hardware perspectives. In the CGM phase (n = 12), three themes were identified through the interviews. The most-often articulated perception of CGM was that it would be a considerable source of frustration. Participants alluded to alarm fatigue and to a fear of unreliability.
Conclusion:
This two-phase study produced a number of themes that were shared among both users and nonusers of diabetes technology. The data gathered will later inform technology design and education offered at diabetes clinics as they are tailored to better fit the needs of type 1 diabetes patients.
Antihyperglycemic Effects of γ-Hydroxy-δ-Valerolactone Isolated from Ethanolic Extract of Adenia lobata Engl. (Passifloraceae)
Dominic Adotei Edoh, PhD, MPhil, BSc (Hons); Joseph Adusei Sarkodie, PhD
Department of Pharmacognosy and Herbal Medicine
University of Ghana
Accra, Ghana
Objective:
Adenia lobata Engl. (passifloraceae) is a woody climber that grows in most African countries, mainly the coastal belt. It is an important medicinal plant used to treat diabetes, hemorrhoids, malaria, fever, and gonorrhea. The present study aims to investigate the antihyperglycemic constituent of the stem of A. lobata in streptozocin-induced diabetic rats.
Method:
The powdered stem (650 g) was Soxhlet extracted using 70% ethanol. The doses for the study were fixed based on Irwin test. The antihyperglycemic effect of the extract (with doses 300 and 600 mg/kg) and the pure compound γ-hydroxy-δ-valerolactone (with doses 60 and 180 mg/kg) were studied in streptozotocin-induced diabetic rats.
Result:
The extract and its isolate, γ-hydroxy-δ-valerolactone, showed dose -dependent reduction of glucose levels after 6 h of administration in the streptozotocin-induced diabetic rats. At 600 mg/kg body weight, the extract caused 51.7% reduction, whereas at 180 mg/kg body weight, the constituent caused 50.9% reduction of blood glucose level after 6 h posttreatment of the diabetic rats. The findings demonstrated the potential of the ethanolic extract and γ-hydroxy-δ-valerolactone on type 1 diabetes, justifying the traditional use of the stem of A. lobata in the management of diabetes by traditional healers in Ghana.
Conclusion:
The compound γ-hydroxy-δ-valerolactone may be one of the active constituents of A. lobata responsible for the observed activities.
Computerization of Patient Records at a Public Herbal Clinic in Ghana
Dominic Adotei Edoh, PhD, MPhil, BSc (Hons); Alex Ocampo, BS; Kingsley Attuah, BSc; Thomford Prah, BSc; Alfred Appiah, PhD
Department of Pharmacognosy and Herbal Medicine
University of Ghana
Accra, Ghana
Objective:
The Center for Scientific Research Into Plant Medicine (CSRPM) is a research institute that produces herbal medicines and has a clinic where patients access health care. At the clinic, patients’ records are compiled and stored manually. The aim of the study was to computerize the recording of patient information at the CSRPM clinic.
Method:
The OpenMRS software was modified and adopted at the clinic. The software was modified to take up data on patient demographics, medical history, diagnosis, laboratory results, medicines given, and follow-ups. The data were keyed in, stored, and retrieved at various sections of the clinic, including records, nurses unit, consulting rooms, laboratories, and dispensary. The data were analyzed at certain intervals, and reports were generated.
Results:
During the 2-month study, patients’ information stored in the electronic medical recording system were as follows: 6,000 patients’ records; 8,000 visits; 3,612 laboratory tests; and 133,432 observations. Diabetes was the third most prevalent disease, after hypertension and malaria, at the CSRPM herbal clinic. Diabetes is managed with two herbal products at the clinic, namely, dibrana and bredilia; the latter product works better in combination with orthodox medicines. The waiting time of patients at the clinic was reduced by 45%.
Conclusion:
The computerization of patients’ records at the CSRPM herbal clinic reduced patient waiting time by 30 min and revealed diabetes as the third most common disease at the clinic.
Inclusion of Glucagon Dynamics into a Model of Glucose Homeostasis
Joseph El Youssef, MD, MCR; Jessica Castle, MD; Peter Jacobs, PhD; Navid Resalat, MS; W. Kenneth Ward, MD
Oregon Health and Science University
Portland, Oregon
Objective:
Many currently used models of glucose dynamics were developed with insulin action in mind, and only recently has glucagon action been considered for inclusion in such models. We describe the updating of an accepted glucose homeostasis model with absorption and action components for glucagon for use in a bihormonal artificial pancreas system, with identification of parameters for the model.
Method:
Based on data obtained from a recent physiology study, we identify six individual models of glucagon action on glucose homeostasis for identification. Using Matlab and Simulink, these models were devised and are being prepared for testing.
Result:
High levels of plasma insulin reduce the effect of glucagon action to elevate blood glucose. However, whether this reduction can be overcome by a sufficiently high dose of glucagon could not be determined. Therefore, two models of glucagon action (immediate and delayed actions) were developed with a threshold insulin level above which glucagon effect is nullified, while two were built with the ability of glucagon to override insulin. The final two use a derivative effect of glucagon on endogenous production.
Conclusion:
Identification and validation of the most appropriate model of glucagon action allows us to update a previously validated model of glucose homeostasis for use in a bihormonal artificial pancreas controller.
Are Risk Indices Derived from Continuous Glucose Monitoring Interchangeable with Self-Monitoring of Blood Glucose–Based Indices?
Chiara Fabris, MSc; Stephen Patek, PhD; Marc Breton, PhD
Center for Diabetes Technology
University of Virginia
Charlottesville, Virginia
Objective:
High blood glucose index (HBGI) and low blood glucose index (LBGI) are popular metrics used to quantify the risk of hyper - and hypoglycemia from sparse self-monitoring of blood glucose (SMBG) data and have been shown to be predictive of future glycemic events, e.g., LBGI of severe hypoglycemia. Continuous glucose monitoring (CGM) systems, however, allow almost continuous monitoring of blood glucose, providing a large amount of data. Using indices such as LBGI and HBGI, and their cutoff values, on CGM data stream has never been validated. The aim of this work is to model the relationship between HBGI/LBGI values obtained from SMBG and CGM, providing a correction to apply when these indices are computed from CGM time series.
Method:
Twenty-eight subjects with type 1 diabetes were monitored for up to 4 weeks with both SMBG and CGM systems. CGM time series were collected using the Dexcom G4 Platinum. HBGI and LBGI were calculated from SMBG and CGM, and a number of models were used to describe the relationship between them.
Result:
A linear model with unitary slope and no intercept is reliable for HBGI. Thus no correction is needed to compute this index from CGM time series. For LBGI, values obtained from CGM did not closely match SMBG-based values, with clear underestimation at low values; a nonlinear transform without saturation performed best to match CGM-based LBGI to SMBG-based LBGI.
Conclusion:
We propose alternate versions of the well-known HBGI and LBGI indices that are adapted to the characteristics of CGM time series. Using such transformations, we enable references to previous studies defining clinically relevant cutoffs for HBGI and LBGI.
Analytical and User Performance Evaluation of a Blood Glucose Monitoring System Following ISO 15197:2013
Guido Freckmann, MD; Manuela Link, ME; Annette Baumstark, PhD; Stefan Pleus, MS; Christina Schmid, PhD; Nina Jendrike, MD; Cornelia Haug, MD
Institut für Diabetes-Technologie Forschungs- und Entwicklungsgesellschaft mbH
Universität Ulm
Ulm, Germany
Objective:
The international standard ISO 15197 defines requirements and testing procedures for self-monitoring of blood glucose (SMBG) systems. In this study, analytical performance, i.e., system accuracy, precision (measurement repeatability, intermediate measurement precision), hematocrit influence, and user performance, was evaluated for a novel SMBG system (Medisafe Fit, Terumo Corporation) following testing procedures and acceptance criteria as described in ISO 15197:2013.
Method:
Analytical performance was evaluated with three different reagent lots by using capillary blood (system accuracy), venous blood (measurement repeatability, hematocrit influence), and control solutions (intermediate measurement precision). The accuracy of the system in the hand of intended users was investigated by using one reagent lot. Compliance with system accuracy criteria was analyzed by calculating the percentage of results within ±15% or ±15 mg/dL of the comparison method measurement results for glucose concentrations above or below 100 mg/dL, respectively.
Result:
The system showed 97 to 100% of results within ISO 15197:2013 accuracy limits and 100% of results within consensus error grid zones A and B. In addition, 100% of results were within the accuracy limits when measurements were performed by the intended users. Regarding measurement repeatability and intermediate precision, the coefficients of variation ranged from 1.8 to 3.2% and from 1.7 to 2.9%, respectively, for glucose concentrations ≥100 mg/dL and the SD ranged from 1.0 to 2.5 mg/dL and from 0.7 to 1.5 mg/dL, respectively, for glucose concentrations <100 mg/dl. In addition, ISO 15197:2013 acceptance criteria for hematocrit influence were met.
Conclusion:
In this evaluation, the Medisafe Fit SMBG system showed high analytical performance and complied with accuracy criteria of the recently revised standard ISO 15197:2013.
Low Pain Perception across Two Generations of Infusion Set Systems for Continuous Subcutaneous Insulin Infusion
Guido Freckmann, MD; Albrecht Fießelmann, MD; Gerhard Klausmann, MD; Stephan Arndt, MD; Kristina Pralle, MD; Linda Amstutz; Thomas Künsting, MD; Bettina Petersen, PhD
Institut für Diabetes-Technologie GmbH
Ulm, Germany
Objective:
As pain perception can impede therapy adherence, two infusion sets with soft cannulas were compared in a randomized controlled crossover trial.
Method:
We enrolled 80 adult patients (20–74 years, median 48 years; 41.3% female) who were divided into two groups. One used Accu-Chek FlexLink (FL) during a 4-week period followed by the novel Accu-Chek FlexLink Plus (FLP) for another 4 weeks. The other group used the inverted sequence: FLP first followed by FL. At the time of the application of each infusion set, patients had to rate their pain perception on a visual analog scale (no pain to maximum pain, values 0–10). As the most traumatic event best encountered with a device will be remembered best and considered exemplary, we did not consider each patient’s pain perception on average, but evaluated the worst pain perception patients reported for the respective infusion set system (FL or FLP).
Result:
Very low pain perception values were found for both systems (1.70 ± 1.86 for FL, 1.74 ± 1.73 for FLP) without significant difference (two-sided t test, P = 0.97).
Conclusion:
In this study in people with diabetes, only low pain perceptions were found for the novel infusion system FLP, comparable to its predecessor system.
Glucosense: A Low-Cost Glucometer System for Resource-Poor Settings
Kayla Gainey, BS; Tyler Ovington, BS; Delphine Dean, SB, MEng, PhD
Clemson University
Clemson, South Carolina
Objective:
In low-resource settings that rely on donated medical supplies, glucose monitoring supplies are not always available to patients, and even when they are, standard supplies can be prohibitively expensive. Our objective is to design a low-cost glucometer system that can be used in resource-poor settings.
Method:
Our system uses test strips made using a standard desktop inkjet printer. The empty printer cartridges are filled with enzyme and dye solutions. Precise amounts of the solutions are printed onto filter paper using a word processor template document. Glucose in the blood induces a color change on the printed strips. To read the strips, we designed a low-cost glucometer using light-emitting diodes, photodetectors, and an Arduino microcontroller, which determines the glucose concentration based on a preprogrammed standard curve.
Result:
The standard curve relating absorbance to glucose concentration showed that the strips were capable of accurately processing the glucose in the range of 0–350 mg/dL. The filter paper allows for controlled diffusion of the blood to the printed enzymes. The printed design incorporates an internal calibration to account for varying patient blood color and oxygen concentration that may affect results.
Conclusion:
The research thus far confirms that a printable test strip is feasible as a means for monitoring glucose levels. While these strips are not as fast as current commercially available strips (readings take ~30 s), they meet the International Organization for Standardization standard for accuracy in a glucometer. The desktop printing of test strips is easy to implement in clinics in resource-poor settings. The current design of the measurement device is inexpensive at roughly one-tenth to one-one-hundredth the cost of conventional glucometer and strip systems.
A Novel Noninvasive Glucose Monitor for Home Use: Assessing the Learning Curve of Use
Avner Gal, MSc, MBA; Ilana Harman-Boehm, MD; Andrew Drexler, MD; Eugene Naidis, MSc; Yulia Mayzel, MSc; Neta Goldstein BSc; Keren Horman, MSc; Shimrit Cohen, BSc
Integrity Applications Ltd.
Ashkelon, Israel
Objective:
The time a layperson requires to learn how to properly use and operate a new device, i.e., learning curve length, is an important factor in such device acceptance and initial utilization. GlucoTrack is a noninvasive, CE-mark-approved glucose monitoring device for home use. Assessment of GlucoTrack’s acceptance included evaluation of its learning curve of use.
Method:
Following individual calibration, measurements’ performance using GlucoTrack requires positioning the sensor-bearing personal ear clip (PEC) at the same earlobe location as during calibration. When wrongly positioned, the device instructs about location adjustment and disables measurement until the PEC is appropriately located. The device’s learning curve of use, based on the time users need to learn how to position the PEC successfully, was assessed in clinical trials. Following calibration and training, 42 subjects (variety of demography) conducted measurements by themselves for 3 nonsequential days within 1 month. The mean number of times to successful PEC positioning (MTSPP) per measurement at each day indicates users’ proficiency level of use.
Result:
The MTSPP per measurement is 1.6 on the first day of use (for comparison, proficient users’ MTSPP per measurement is 1.2). On day 2 of use, MTSPP reduces to 1.3 and on day 3 to 1.2. The same trend is found among all subjects in different demography groups.
Conclusion:
The results suggest that proficiency in conducting measurements for the majority of users is gained after ~3 days of use. Furthermore, the most significant improvement in use (in PEC positioning context) is obtained after only one day. Achieving a high proficiency level after such short experience is expected to encourage device acceptance among new users.
Enabling Frequent Blood Glucose Monitoring at Home Using a Truly Noninvasive Device
Avner Gal, MSc, MBA; Ilana Harman-Boehm, MD; Andrew Drexler, MD; Eugene Naidis, MSc; Yulia Mayzel, MSc; Neta Goldstein BSc; Keren Horman, MSc
Integrity Applications Ltd.
Ashkelon, Israel
Objective:
A blood glucose (BG) monitor that is painless, easy and simple to use, with acceptable accuracy is likely to promote frequent self-monitoring. GlucoTrack, a CE-mark-approved noninvasive BG monitoring device for home use, offers a painless way of measurement. It requires individual calibration (once every 6 months), afterward enabling virtually unlimited utilization. To evaluate fulfillment of essential requirements for frequent self-monitoring, GlucoTrack usability, ease of use, and accuracy level maintenance (along the entire calibration validity period) were assessed.
Method:
Maintenance of GlucoTrack accuracy (throughout 6 months) and device usability were evaluated in 198 subjects (12,956 data points). Following calibration, subjects participated in 2–19 nonconsecutive full-day sessions. To demonstrate consistency of accuracy, interdaily performances were analyzed across subjects. Device usability and ease of use were assessed based on users’ feedback analysis.
Result:
Throughout all trial days, up to 6 months from calibration, results remain similar: 95.9 ± 3.1% of the points within Clarke error grid A+B zones and 32.9 ± 4.2% mean absolute relative difference. Eighty-six percent of the subjects are willing to use the device regularly, 97% declared they will use the device more often than their invasive one, 71% claimed the device is easy to use, and 81% found performing measurement easy and simple.
Conclusion:
The similar accuracy demonstrated interdaily during a 6-month period suggests the device accuracy is maintained at any home or home-like environment over its entire calibration validity period. The results demonstrate positive user feedback, high satisfaction of use, and willingness to increase frequency of self-monitoring. These findings can support frequent use of GlucoTrack for enhanced BG monitoring and tighter glycemic control.
MDiasite Smart Website for Diabetes Education
Salvador González-Pérez, PhD; Javier Fernández-Cañete, PhD; Elena Alenina, PhD; Victor Grigorenko, PhD
Department of System Engineering and Automation
University of Malaga
Malaga, Spain
Objective:
MDiasite is an update of a web educational platform (Diasite) for course management system (CMS), a free, open-source software package designed using sound pedagogical principles to help diabetes educators and health care professionals in order to create effective online learning communities. It is an initiative that involves public and private partners to improve the treatment and outcomes for people with diabetes, to promote early diagnosis, and to prevent or delay the onset of diabetes.
Method:
This SW is based in the Moodle platform as open-source software under the GNU General Public License and has the following features: promotes a social constructionist pedagogy (e.g., collaboration, activities, critical reflection); suitable for 100% online classes as well as supplementing face-to-face learning; simple, lightweight, efficient, compatible, low-tech browser interface; and easy to install on almost any platform that supports PHP. Nowadays, mobile devices are very important for patients, so this update is related to use MDiasite on mobile devices: users can open MDiasite on their mobile web browsers and can download native applications for their mobile devices. MDiasite is mobile accessible through server extensions in IOS, Android, Windows Mobile, and Blackberry.
Result:
This platform can show all descriptions for every course on the server, including accessibility to guests. Contents can be edited using an embedded WYSIWYG HTML editor. Goals are to reduce administrator involvement to a minimum, while retaining high security. It supports a range of authentication mechanisms through plug-in authentication modules, allowing easy integration with existing systems. It offers standard e-mail methods where patients can create their own login accounts. For security, educators can add an “enrollment key” to their courses to keep out nonpatients. They can give out this key face-to-face or via personal e-mail. Educators can enroll and unenroll patients manually if desired even after a certain period of inactivity. Patients are encouraged to build an online profile. Personal data can be protected from display if required. Every user can specify their own time zone. Every user can choose the language used for the MDiasite interface. The modules include a chat module, choice module, forum module, quiz module, resource module, survey module, and workshop module.
Conclusion:
Use of this SW improves the range of diabetes educators, bringing content to mobile devices so that patients can take courses from anywhere.
Transcutaneous Glucose Sensor Performance in a Diabetic Animal Model
Bing Gu, MS; Sagar Vaddiraju, PhD; Fotios Papadimitrakopoulos, PhD; Diane J. Burgess, PhD
Department of Pharmaceutical Sciences
University of Connecticut
Storrs, Connecticut
Objective:
The purpose of this work is to test amperometric glucose biosensor performance in a diabetic animal model.
Method:
Each streptozotocin-induced diabetic rat was transcutaneously implanted with one polyvinyl alcohol hydrogel/poly(lactic- co-glycolic acid) microsphere composite–coated amperometric glucose sensor. Six units/kg insulin was injected intravenously via the tail vein to induce hypoglycemia. Meanwhile, the blood glucose concentration was determined periodically. Two dextrose injections were then performed intraperitoneally under hypoglycemic conditions to test the sensor response. In order to test the effect of our composite coating (designed to prevent inflammation and fibrosis while still allowing glucose diffusion toward the sensor) on glucose diffusion, polyvinyl alcohol hydrogel/poly(lactic-co-glycolic acid) microsphere composite–coated microdialysis probes were also implanted into diabetic rats. Relative recovery of glucose from the microdialysis probe was determined at different days.
Result:
The rats were induced to be diabetic with blood glucose concentrations of ~400 mg/dL. Amperometric sensor output closely followed the trend of the blood glucose concentration from hyperglycemic to hypoglycemic region. Under the hypoglycemic conditions, the glucose sensor was able to differentiate minor glucose concentration changes within a short time period. This indicates the amperometric glucose sensor has a fast response time and good sensitivity. The relative recovery of glucose from composite-coated microdialysis probes can be maintained at ~12% for up to 1 week.
Conclusions:
The amperometric glucose biosensor can perform at both hyper- and hypoglycemic conditions in diabetic animals. The sensor has a fast response time and good sensitivity. The composite coating was able to inhibit the foreign body reaction to transcutaneously implanted microdialysis probes over a 1-week test period.
Funding:
This study was supported by U.S. Army Medical Research (W81XWH0910711, W81XWH0710688), the National Institutes of Health (1R21HL09045801, R43EB011886, 9R01EB014586), and the National Science Foundation/Small Business Innovation Research (1046902, 1230148).
Continuous Exenatide Infusion Reduces Glycemic Variability and Improves Glucose Control in Cardiac Surgery Patients: The EXECUTIVE Trial
Martin Haluzik, MD, DSc; Michal Lipš, MD; Jana Drapalova, MSc; Milos Mraz, MD, PhD; Milos Dobias, MD; Petr Kopecky, MD; Jaroslav Lindner, MD, PhD; V. Burda, BSc; Daniel Novak, Ing, PhD; Michaela Diamant, MD, DSc; Stepan Svacina, MD, DSc
Third Department of Medicine
Department of Endocrinology and Metabolism
General University Hospital
Charles University in Prague
First School of Medicine
Prague, Czech Republic
Objective:
We performed a randomized control trial with the GLP-1 receptor agonist exenatide as add-on to standard perioperative insulin therapy in subjects undergoing elective cardiac surgery to intensify perioperative glucose control while minimizing the risk of hypoglycemia.
Method:
Thirty-eight subjects (63.2% diabetes patients) with decreased left ventricular systolic function (ejection
fraction ≤ 50%) scheduled for elective coronary artery bypass grafting were randomized to receive either exenatide or placebo in a continuous 72-h intravenous infusion on top of standard perioperative insulin therapy. Parameters of glucose control and glycemic variability together with indices of cardiac function assessed by transthoracic echocardiography, perioperative hemodynamic parameters, and the need of antiarrhythmic and inotropic treatment were collected as primary end points.
Result:
Compared with the placebo group, subjects receiving exenatide showed improved perioperative glucose control (average glycemia 6.4 ± 0.1 vs. 7.2 ± 0.01 mmol/L, P < 0.001; time in target range of 4.5–6.5 mmol/L 53.6 ± 3.3 vs. 38.6 ± 3.3%, P < 0.01; time above target range 40.8 ± 3.1 vs. 52.8 ± 3.5%, P < 0.05) without an increased risk of hypoglycemia (0.11 ± 0.08 vs. 0.21 ± 0.1 episodes of hypoglycemia ≤3.3 mmol/ L in exenatide versus placebo group; P = 0.62). Exenatide infusion also reduced glycemic variability (SD 1.4 ± 0.5 vs. 2.0 ± 0.6, P < 0.01; mean amplitude of glycemic excursion 2.5 ± 1.1 vs. 3.3 ± 0.9, P < 0.01) and decreased the need of temporary pacing (11.1 vs. 47.4% of subjects, P < 0.05), while no significant difference in perioperative hemodynamics, postoperative echocardiographic parameters, and inotropic medication dosage was found between the groups.
Conclusion:
Perioperative administration of intravenous exenatide in subjects undergoing elective coronary artery bypass grafting improved glucose control and decreased glycemic variability without increasing the risk of hypoglycemia. Except of decreased need of temporary pacing, exenatide did not significantly affect parameters of cardiac function.
Funding:
This work was supported by SVV260019/2014 and MZOVFN64165.
Glycemic Variability and Glycemic Goals in a Pilot of Young Children with Type 1 Diabetes
Celia Henderson, BSN, CDE; Linda Herbert, PhD; Eleanor Mackey, PhD; Maureen Monaghan, PhD; Fran R. Cogen, MD, CDE; Randi Streisand, PhD, CDE
Children’s National Medical Center
Washington DC
Objective:
We explore the feasibility of meeting the American Diabetes Association–recommended A1C goal of 7.5% for children/adolescents.
Method:
This pilot study assesses glycemic variability, nutrition, and physical activity in 10 young children (YC) with type 1 diabetes (80% female; mean age = 5.88 years; mean duration = 2.69 years; mean A1C = 7.3%; 60% Caucasian). Participants wore blinded a continuous glucose monitor (CGM) for 5 days. Parents completed daily 24-h diabetes management interviews, including meter blood glucose (BG) levels, meal/snack content, and physical activity. Participants were offered the opportunity to review data and discuss changes to improve glycemic control with a Certified Diabetes Educator (CDE).
Result:
Findings indicate that YC experienced daily BG excursions (<100 or >200 mg/dl per the American Diabetes Association’s previous recommendations for YC) that were captured by a CGM. Hyperglycemic episodes were more common, spending a daily average of 39.88% of the 5-day trial (SD = 17.89%) with a BG level above 200 mg/dl and 14.37% of the trial (SD = 11.24%) with a BG level below 100 mg/dl. An average of 54.25% of the 5-day trial was spent out of range. Five out of 10 families met with a CDE to review the data and discuss changes in regimen, nutrition, and activity to improve glycemic control. Average A1C prior to CDE intervention was 7.6% (minimum 6.9% and maximum 8.2%), and despite specific recommendations to improve control, the average A1C after the intervention increased to 8.1% (minimum 7.2% and maximum 9.5% ).
Conclusion:
Only 5 out of 10 participants took advantage of the opportunity to review the data, indicating a possible disinterest in using CGM technology. Of those who participated in this phase, a measurable decrease in A1C levels was not attained despite specific recommended changes. Achieving goals of A1C <7.5% may be challenging in YC with type 1 diabetes.
Use of the eGlycemic Management System by Glytec Provides Safe and Effective Transition from Intravenous to Subcutaneous Insulin Therapy
Amy B. Henderson, RN, BSN; Lori Hubbard, RN, BSN; Raymie S. McFarland, PT
Glytec
Greenville, South Carolina
Objective:
It is well established that hyperglycemia is associated with poor outcomes for hospitalized patients. The transition from intravenous (IV) to subcutaneous (SubQ) insulin therapy can present a challenge for maintaining glycemic control. This study evaluated the glycemic outcomes of patients transitioning from IV to SubQ insulin using either the eGlycemic Management System (eGMS) by Glytec or a paper protocol (PPT), and posttransition glycemic outcomes on SubQ insulin therapy were measured.
Method:
The study evaluated 21 patients at Johnson City Medical Center that transitioned from IV to SubQ insulin. Qualifying patients were treated on IV insulin with eGMS. Ten patients transitioned to SubQ using eGMS, and 10 patients transitioned using a PPT. The efficacy and safety of each was evaluated by the following: 1) percentage of blood glucose (BG) between 71 and 180 mg/dL and 2) percentage of hypoglycemic events <40, <60, and <70 mg/dL.
Result:
Patients who transitioned using eGMS had 90.16% of BG values between 71 and 180 mg/dL in the first 24 h following transition, compared with PPT at 83.33%. Patients who transitioned using eGMS had 92.09% of BG values between 71 and 180 mg/dL overall following transition, compared with PPT at 79.41%. The percentage of overall hypoglycemic events <40, <60, and <70 mg/dL for eGMS were 0, 0.56, and 3.33% and for PPT were 0%, 0%, and 1.18%, respectively.
Conclusion:
The patients who transitioned from IV to SubQ insulin using eGMS had significantly more BG values within the target range, with a low incidence of hypoglycemia (<60 mg/dL) and no incidence of critical hypoglycemia (<40 mg/dL). The results suggest that eGMS can safely transition patients from IV to SubQ insulin therapy while maintaining optimal glycemic control.
Bolus Calculator Autotuning
Pau Herrero, PhD; Peter Pesl, MSc; Monika Reddy, MRCP; Nick Oliver, FRCP; Desmond Johnston, FMedSci; Pantelis Georgiou, PhD; Christofer Toumazou, FRS
Center for Bio-Inspired Technology
Institute of Biomedical Engineering
Imperial College London
London, United Kingdom
Objective:
Insulin bolus calculators are formula-based decision support tools using subject-specific parameters for insulin dose calculation. Although clinical benefit has been demonstrated, performance remains suboptimal due to technical and human factors. Accurate tuning of the parameters is important and is currently done manually. We present, and in silico validate, a novel algorithm for automatic tuning of a standard bolus calculator.
Method:
A bolus calculator incorporating insulin-to-carbohydrate ratio (ICR), correction factor (CF), and estimated insulin on board was employed. The proposed method uses continuous glucose monitoring for glycemic outcomes. The glycemic target range is the only required input. The algorithm adjusts ICR and CF (assumed linearly correlated) to position the minimum postprandial glucose value within the glycemic target. For validation, the University of Virginia-Padova simulator was employed. Variability of parameters such as ingested carbohydrates, meal composition, ingestion time, capillary blood glucose, and continuous glucose monitoring data were accounted for. ICR and CF were randomly initialized to suboptimal values. The algorithm was tested over a 2-week simulation period in 20 virtual subjects, with standard glucose metrics measured.
Result:
Studied metrics are presented as mean ± SD (day 1 vs. day 14): blood glucose, 143.4 ± 48.1 vs. 138.5 ± 14.1 mg/dL (P = 0.651); percentage of time in 70–180 mg/dL, 57.0 ± 24.3 vs. 90.8 ± 11.9 (P < 0.001); percentage of time below 70 mg/dL, 15.7 ± 15.7 vs. 0.2 ± 0.9 (P < 0.001); percentage of time above 180 mg/dL, 27.1 ± 29.6 vs. 8.9 ± 12.0 (P < 0.001); and risk index, 10.6 ± 5.7 vs. 2.7 ± 1.7 (P < 0.001). Mean relative change in ICR (and CF) with respect to the initialization value was 0.85 ± 0.44.
Conclusion:
A new method for the automatic tuning of a bolus calculator is able to safely converge toward its optimal parameters within 2 weeks. This method is currently being used in combination with case-based reasoning to cope with real-life variability.
Use of Diabetes Data Management Software Reports by Health Care Providers Improves Accuracy and Efficiency of Data Analysis and Interpretation Compared With Traditional Logbook Data: First Results of the Accu-Chek Connect Reports Utility and Efficiency Study (ACCRUES)
Deborah A. Hinnen, APRN; Ann Buskirk, BSN; Maureen Lyden, MS; Linda Amstutz, BSN; Tracy Hunter, BS; Robin Wagner, DVM, PhD
University of Colorado Health Center
Colorado Springs, Colorado
Objective:
Utilization of diabetes management software programs has been shown to enhance glycemic control; however, the accuracy in interpreting data presented by these programs and time required to use them effectively have not been well studied. The Accu-Chek Connect Reports Utility and Efficiency Study (ACCRUES) assessed the proficiency and efficiency of clinicians in identifying and interpreting self-monitored blood glucose, insulin, and carbohydrate intake data using data management software reports compared with standard logbook use.
Method:
This prospective, self-controlled, randomized study enrolled clinicians with no diabetes specialization who were naïve to computer diabetes data management software. Diabetes specialists reviewed four paired clinical cases (eight total) based on data from diabetes patient histories (two multiple daily injections, two continuous subcutaneous insulin infusion) and determined the correct multiple-choice responses regarding identification of meaningful diabetes information, recognition of glycemic patterns, and appropriate management decisions. Cases and questions were presented to participants via a secure web portal. Response time and accuracy were documented and assessed. Participants completed a preference questionnaire at study completion.
Results:
All (n = 33) clinicians completed the patient cases. Participants achieved significantly greater accuracy (assessed by percentage of accurate answers) using the software reports versus logbooks: 89.5 (8.0) vs. 66.4 (12.3)%, P < 0.0001. The average time required to complete each case was significantly less using the software reports versus logbooks: 6.7 (2.9) vs. 16.0 (12.0) min, P < 0.0001. All clinicians expressed agreement (24.2%) or strong agreement (75.8%) that they preferred using the software reports versus logbook data.
Conclusion:
Clinicians naïve to diabetes data management software identified, interpreted, and used key diabetes information with greater accuracy and efficiency using the software reports compared with traditional logbook information. Use of software reports was preferred over logbooks.
Use of Diabetes Data Management Software Reports by Patients with Diabetes and Caregivers Improves Accuracy and Efficiency Data Analysis and Interpretation Compared With Traditional Logbook Data: First Results of the Accu-Chek Connect Reports Utility and Efficiency Study (ACCRUES)
Deborah A. Hinnen, APRN; Ann Buskirk, BSN; Maureen Lyden, MS; Linda Amstutz, BSN; Tracy Hunter, BS; Robin Wagner, DVM, PhD
University of Colorado Health Center
Colorado Springs, Colorado
Objective:
The Accu-Chek Connect Reports Utility and Efficiency Study (ACCRUES) assessed the proficiency and efficiency of patients with diabetes (PWD) and caregivers of PWD under 18 years of age treated with multiple daily insulin injections (MDI) and continuous subcutaneous insulin infusion (CSII) in identifying and interpreting self-monitored blood glucose, insulin, and carbohydrate intake data using data management software reports compared with standard logbook use.
Method:
This prospective, self-controlled, randomized study enrolled MDI- and CSII-treated PWD and caregivers who were naïve to computer diabetes data management software. Diabetes specialists reviewed four paired clinical cases (eight total) based on data from diabetes patient histories (two MDI, two CSII) and determined the correct multiple-choice responses regarding identification of meaningful diabetes information, recognition of glycemic patterns, and appropriate management decisions. Cases and questions were presented to participants via a secure web portal. Response time and accuracy were documented. All participants completed a preference questionnaire at study completion.
Results:
A comparable number of MDI-treated (n = 27) and CSII-treated (n = 27) patients and caregivers (n = 24) completed the patient cases. Participants achieved significantly greater accuracy (assessed by percentage of accurate answers) using the software reports versus logbooks: PWD 80.3 (13.2) vs. 63.7 (15.0)%, P < 0.0001; caregivers 84.4 (12.3) vs. 63.9 (14.2)%, P < 0.0001. The average time required to complete each case was significantly less using the software reports versus logbooks: PWD 8.6 (4.3) vs. 19.9 (12.2) min, P < 0.0001; caregivers 7.0 (3.5) vs. 15.5 (11.8) min, P = 0.0005. Most PWD (88.9%) and caregivers (95.7%) expressed agreement or strong agreement that they preferred using the software reports versus logbook data.
Conclusion:
PWD and caregivers identified, interpreted, and used key diabetes information with greater accuracy and efficiency using software reports versus traditional logbook information and preferred software reports over logbooks.
Detection of Peripheral Arterial Disease in Diabetic Patients Using Diffuse Optical Tomography in Feet
Jennifer W. Hoi, MS; Michael A. Khalil, PhD; Hyun K. Kim, PhD; Rajeev Dayal, MD; Gautam Shrikhande, MD; Andreas H. Hielscher, PhD
Department of Biomedical Engineering
Columbia University
New York, New York
Objective:
Early and accurate evaluation of peripheral arterial disease (PAD) is key to preventing future cardiovascular events. Detection of PAD can be difficult in diabetic patients due to arterial calcification, which falsely elevates the ankle-brachial index. We investigate the use of diffuse optical tomography (DOT) in the foot.
Method:
Thirty-six subjects were enrolled into healthy (n = 16), PAD without diabetes (n = 10), or PAD with diabetes (n = 10) cohorts. Two-dimensional reconstructions of the total hemoglobin (HbT) within the foot during rest and 60 mmHg venous occlusions were collected using a DOT system. The difference in average HbT from regions associated with blood transport during the two phases (∆[HbT]) were calculated.
Result:
There were significant differences between the three cohorts (P = 0.002) and between the healthy and PAD cohorts (with diabetes P = 0.006, without diabetes P = 0.004). However, there was no difference between PAD cohorts (P = 0.367), indicating the measure is independent of diabetic comorbidities. A preliminary receiver-operating-characteristic curve analysis achieved a sensitivity of 0.82, specificity of 0.70, and positive predictive value of 0.73.
Conclusion:
The DOT ∆(HbT) measure may be a good detector of PAD, irrespective of diabetic comorbidities.
The Diabetes Snapshot: Using the Electronic Medical Record to Improve Glycemic Control among Hospitalized Patients
Karen R. Horowitz, MD; Saira Khan, MD; Loren Carpenter, MPH
Louis Stokes Cleveland Veteran Affairs Medical Center
Cleveland, Ohio
Objective:
We evaluate the effectiveness of the diabetes snapshot in combating physician inertia with regard to inpatient glycemic control.
Method:
The diabetes snapshot is a tool in the computerized patient record system that charts point-of-care (POC) blood glucose and doses of insulin administered in a time-wise fashion for the previous 72 h. It can be imported directly into the daily progress note. It was developed by the investigator to facilitate inpatient insulin titration. We were interested in whether the diabetes snapshot would facilitate insulin titration by medicine interns, particularly for inpatients whose blood glucose was difficult to control. Internal medicine interns were introduced to the diabetes snapshot and instructed in its use in July 2013. Using a retrospective chart review of computerized patient record system from July to October 2013, we identified patients with POC glucose >250 on day 3 of hospitalization who were under the care of internal medicine postgraduate year 1 residents. We assessed how often the snapshot was used and how often changes in insulin regimen were made.
Result:
Sixteen patients were identified with difficult-to-control blood glucose (POC glucose >250 on day 3 of hospitalization) admitted under an internal medicine intern. For the care of these patients, the diabetes snapshot was used 69% of the time. One hundred percent of medicine interns who used the diabetes snapshot made changes to the insulin regimen of their patients compared with 40% of the interns who did not use the snapshot.
Conclusion:
The electronic medical record is a valuable tool for synthesizing data and presenting trends. We have demonstrated that use of the diabetes snapshot to identify inpatient hyperglycemia prompted interns to act on poor glycemic control more often than interns who did not use the tool.
Application of Fuzzy Anti-Reset Windup for Proportional-Integral-Derivative Control in an Implantable Artificial Pancreas
Lauren M. Huyett, BS; Eyal Dassau, PhD; Francis J. Doyle III, PhD
Department of Chemical Engineering
University of California, Santa Barbara
Sansum Diabetes Research Institute
Santa Barbara, California
Objective:
Proportional-integral-derivative (PID) control is a popular choice for the artificial pancreas (AP). While the integral component allows the controller to adapt to changing insulin sensitivity, it has also been shown to cause postprandial undershoot of the set point. Previous researchers have imposed an upper limit on the integral component; however, this method does not guarantee set point tracking. In this study, fuzzy anti-reset windup was investigated as an alternative solution to minimize undershoot without compromising on set point tracking in a fully implantable AP.
Method:
A PID controller was tuned using internal model control rules and evaluated in the 10-subject University of Virginia/Padova metabolic simulator. The protocol included one meal (100 g carbohydrates) followed by a 30% change in insulin sensitivity, with a set point of 110 mg/dL. The intravenous insulin and glucose-sensing ports were used to approximate an intraperitoneal system. The rate of change of the integral action component was attenuated when the glucose deviation became large according to a fuzzy logic scheme.
Result:
Without anti-reset windup, the PID controller resulted in a postprandial undershoot to 78 ± 7 mg/dL. Elimination of the integral component improved undershoot to 106 ± 2 mg/dL but allowed a final offset of 15 ± 2 mg/dL after a 30% decrease in insulin sensitivity. When the fuzzy anti-reset windup with optimal tuning was applied, it resulted in a postprandial undershoot to 104 ± 2 mg/dL and had no steady-state offset following the sensitivity change. The peak blood glucose was unchanged at 218 ± 12 mg/dL for all cases.
Conclusion:
The application of fuzzy anti-reset windup to the PID controller reduced the postprandial undershoot by 26 mg/dL while retaining set point tracking. In addition, this method did not affect the quality of meal control, making it an attractive improvement to the AP controller.
Insulin Sensitivity Changes Assessed by Continuous Glucose Monitoring and Insulin Pump System During Menstrual Cycles in Women with Type 1 Diabetes
Boyi Jiang, MA; Sue Brown, MD; Marc D. Breton, PhD
Center for Diabetes Technology
University of Virginia
Charlottesville, Virginia
Objective:
We quantify the insulin sensitivity (SI) during menstrual cycles in type 1 diabetic women using continuous glucose monitoring and pump data, targeting a description of pattern in SI changes.
Method:
We introduce a mathematical description of glucose homeostasis in log space based on the minimal model of glucose kinetics using feed-forward models for meal and insulin appearance. Using a Kalman filter, we are able to extract a physiologically sound estimate of SI. Twelve women with type 1 diabetes participated in a 3-month clinical trial during which they were asked to wear a Dexcom G4 sensor and track phases of their menstrual cycle (e.g., menstruation and ovulation). Self-monitoring of blood glucose, continuous glucose monitoring, and pump were recorded for all 3 months. For each subject, we computed the average nocturnal SI (1:00 A.M. to 6:00 A.M.) to minimize the impact of the model discrepancy on the Kalman filter performance around meals. Phases of the menstrual cycle were based on menstruation and ovulation for each cycle and confirmed a posteriori by blood draws. The SI in different cycle phases was then compared after being averaged across the three cycles.
Result:
The SI (e-4/min per mU/L) in the early follicular phase, the early-mid luteal phase, and the late luteal phase were 4.6 ± 0.5, 4.0 ± 0.9, and 4.0 ± 1. In the standard t test, early follicular phase versus early-mid luteal phase P = 0.05 and early follicular phase versus late luteal phase P = 0.03.
Conclusion:
The SI in the early-mid luteal phase and the late luteal phase deceases significantly compared with that in the early follicular phase in women with type 1 diabetes. The perimenstrual glycemic control is expected to be improved by applying the developed SI tracking technique.
Noninvasive Continuous Glucose Monitoring in Postsurgical Diabetic and Nondiabetic Intensive Care Unit Patients
Jeffrey I. Joseph, DO; Marc Torjman, PhD; John Reich, MD; Anthony Furnary, MD; Stanley A. Nasraway, MD; Mary McNamara-Cullinane, BA, RAC; Christine Olimpio, MS; David Olson, PhD; David Walton, MBA
Thomas Jefferson University Hospital
Philadelphia, Pennsylvania
Objective:
We assess the performance of a noninvasive, transdermal continuous glucose monitoring (CGM) system (Symphony CGM System, Echo Therapeutics, Philadelphia, PA) in the intensive care unit (ICU) setting.
Method:
Adult surgical patients with planned ICU admission of ≥24 hours at four medical centers were enrolled. A CGM sensor containing glucose oxidase was applied to a 6-mm diameter skin site that was prepared with controlled microabrasion using the Symphony SkinPrep device. Up to 30 blood samples from each patient were obtained from a radial artery catheter and measured using a YSI 2300 STAT Plus Glucose Analyzer as a reference. CGM was prospectively calibrated every 4 h. Safety was assessed by visual inspection of the site.
Result:
Thirty-two subjects completed the study. The study cohort was 19% female and 28% diabetes patients with a mean age of 65 ± 13 years. Overall mean absolute relative difference between CGM and reference blood glucose was 12.5% overall (12.2% for diabetic patients and 12.6% for nondiabetic patients). The post–warm-up, 8-h interval mean absolute relative differences were 13.2, 12.2, and 12.1%. Glucose values ranged from 49 to 324 mg/dL. No device/study-related adverse events were reported. Ninety-seven percent of subjects experienced no or mild irritation from the microabrasion skin preparation process, and 94% of subjects experienced no or mild irritation following the CGM sensor removal.
Conclusion:
Symphony CGM demonstrated clinically relevant accuracy consistently throughout the CGM session and excellent safety in diabetic and nondiabetic patients. Future studies are needed to determine whether Symphony CGM can be used to improve blood glucose control.
A Novel Focal Drug Delivery System to the Hypothalamus to Assess the Neuronal Role of GLP-1 Receptor in Energy Homeostasis
Katrin Kaineder, MSc; Thomas Birngruber, PhD; Thomas R. Pieber, MD
Center for Basic Medical Research
Department of Endocrinology
Graz, Styria, Austria
Objective:
Obesity is a main risk factor in the development of type 2 diabetes. GLP- 1 analogs are currently used for the systemic treatment of type 2 diabetes, as they are known to reduce nutrient ingestion and body weight. We aim to investigate the pharmacological effects of systemic GLP-1 application versus GLP-1 application directly into the brain in terms of metabolic changes and GLP-1 receptor expression. Therefore, we developed a focal delivery system for drug administration directly into the brain, specifically into the hypothalamus of rats.
Method:
We assessed body weight changes in response to the implanted probe. The distribution pattern of the injected liquid was analyzed using dextran blue as a marker to verify precise application of high-molecular-weight substances into the hypothalamus. GLP-1 receptor distribution in the hypothalamus was examined via in situ hybridization and fluorescently labeled GLP-1 peptide. We assessed metabolic changes after systemic and intrahypothalamic GLP-1 administration by using the PhenoMaster cage system.
Result:
We found no changes in body weight in response to probe implantation and only minimal histological tissue reaction in the immediate vicinity of the probe. The results indicated GLP-1 receptor expression in the hypothalamus and also metabolic changes triggered by those. We examined difference between intrahypothalamic and intraperitoneal-administered GLP-1 in terms of energy homeostasis.
Conclusion:
We were able to successfully establish a focal delivery system in a preclinical pilot study. First PCR results are promising and will further be used to investigate precise signal transduction pathways triggered by intrahypothalamic GLP-1. Current experiments include the analysis of the signal transduction pattern with real-time PCR array analysis in situ after intrahypothalamic GLP-1 administration.
Design of the ENDO Trial: A Randomized, Sham-Controlled, Double -Blind Trial of the EndoBarrier for the Treatment of Type 2 Diabetes in Subjects with Obesity
Lee Kaplan, MD, PhD; John Buse, MD, PhD; Kenneth Malomo, BS; Keith Gersin, MD; Jennifer Cormier, BS; Steven Edmundowicz, MD; Louis Aronne, MD; Paul Visintainer, PhD; Lauren Baker, PhD; David Maggs, MD
Massachusetts General Hospital
Boston, Massachusetts
Objective:
The EndoBarrier (EB) is a 2-foot-long polymer liner endoscopically anchored in the duodenal bulb. Previous trials in 119 subjects demonstrated an average 16% total weight loss in obese subjects and a 1.5% decrease in HbA1c in those with type 2 diabetes mellitus (T2DM). Because of the greater manipulation and subject expectation in device trials, special consideration is needed to minimize placebo effects and unblinding of subjects and study staff.
Methods:
The ENDO Trial is a randomized, double-blind, sham-controlled trial of a 52-week EB implantation for T2DM treatment in obese subjects at 25 U.S. sites. Five hundred subjects will be enrolled in a 2:1 EB:sham ratio. Inclusion requires an HbA1c of 7.5–10% on up to two oral T2DM agents and BMI of 30–55 kg/m2.
Results:
The trial design includes two study teams at each site: a blinded endocrinology team responsible for recruitment and follow-up evaluation and an unblinded endoscopy team responsible for randomization, device (or sham) implantation and removal, and management of severe gastrointestinal events. To minimize the risk of unblinding, the EB and sham device, packaging, and disposal are indistinguishable. Protocols and scripts were prepared to ensure that device and sham procedures are similar in duration, personnel, use of endoscopy and fluoroscopy equipment, and clinical documentation. Specific protocols are included to avoid subject unblinding during routine care and airport screening.
Conclusion:
The design of an effectively blinded, randomized, sham-controlled trial of an endoscopically implanted device requires complex design strategies distinct from those used in placebo-controlled medication trials. The ENDO Trial incorporates several novel features to facilitate sensitive yet unbiased assessment of EB efficacy and safety.
A Semi-Implantable Biosensor for Continuous Lactate Monitoring
Michail Kastellorizios, MSc; Sagar Vaddiraju, PhD; Allen Legassey, BS; Rob Croce, PhD; Fotios Papadimitrakopoulos, PhD; Diane J. Burgess, PhD
Department of Pharmaceutical Sciences
University of Connecticut
Storrs, Connecticut
Objective:
Lactate accumulation is a biomarker of anaerobic metabolism. Continuous lactate monitoring, in tandem with glucose monitoring, can be used to assess the performance of diabetic athletes. We have previously presented this concept using microdialysis as a monitoring method. The objective of this study was to develop and test a miniaturized, transcutaneous sensor for continuous lactate monitoring.
Method:
Lactate sensors were modeled after our previously developed coil-type glucose sensors. Lactate oxidase was deposited on a platinum coil (working electrode), and silver was used as the reference electrode. A proximity communicator transmitted the sensor output wirelessly to a nearby computer. The sensors were tested in vitro in PBS and in vivo in normal rats via transcutaneous implantation. Blood lactate concentrations were used as a reference.
Result:
Lactate sensors showed a linear response over a wide range (0.6–5.5 mmol/L) in vivo and in vitro. Interestingly, the sensor sensitivity increased after implantation (0.21 vs. 0.10 mV/mmol/L) while the baseline was similar (0.55 mV in vivo, 0.63 mV in vitro). Implanted lactate sensors were successfully used to monitor lactate trends in the subcutaneous tissue of exercising rats.
Conclusion:
This study proved the feasibility of using coil -type electrochemical sensors for continuous lactate monitoring in the subcutaneous tissue. The sensors showed a linear response over a wide lactate concentration range both in vivo and in vitro. Moreover, this study has provided proof of concept that electrochemical lactate sensors can be used to monitor lactate trends during exercise.
A U.K. Patient’s Perspective on Graphical Display of Glycemic Control Data
Daniel Kay, PGDip
Worcester, Worcestershire, United Kingdom
Objective:
Despite advances in graphics and the availability of graphic software, multiple-point glucose profiles are often suboptimally displayed in practice. The proposed graphical display focuses on providing the person with diabetes with accurate statistical information regarding both HbA1c and self-monitored blood glucose (SMBG) data through a clearer visual display. This commentary was written from the perspective of a person with type 1 diabetes and might add value to the existing understanding about relevant ways to display glycemic control data.
Method:
Actual quarterly HbA1c and SMBG data sets were recorded from a person who has lived with type 1 diabetes for 27 years (29-year-old male), for retrospective analysis across a 12-month period. Two graphs were constructed so that it was possible to visually interpret relevant sources of information with regard to glycemic control. Graph 1 was based on a more traditional line graph (typical of that depicted during clinical appointments), which portrayed four HbA1c results over a 12-month period and illustrates a target line that intersects 57 mmol/mol. Graph 2 takes graph 1 a step further through inclusion of distributional information. Tukey-style boxplots were used to display relevant SMBG statistics: maximum values; 25th, 50th, and 75th percentiles; and minimum values at four preprandial times (0600, 1200, 1600, and 2300 h). In acknowledgement of the fact that the central tendency is a frequently used and relevant statistic to better understand glycemic patterns, a variation on boxplots that displays modal information (indicated by dark plus symbols) was chosen. HbA1c data were presented directly underneath the respective distributional information.
Result:
The data density of graph 1 was low compared with graph 2, as it portrayed one value at each quarter, that is, the HbA1c results that were 86, 64, 75, and 53 mmol/mol, respectively. For the first three quarters, they reveal differences in HbA1c results (a reduction from 3 to 6 months and an increase from 6 to 9 months) above the target 57 mmol/mol. In addition, graph 2 shows a reduction in the interquartile ranges (IQRs) at 1200 and 1600 h from 3 to 6 months (7.8 to 4.2 vs. 7.0 to 3.0 mmol/L, respectively). The second to third quarter sees an increased proportion of hyperglycemic events. This is affirmed by increased IQRs at 1200, 1600, and 2300 h, as well as increased modal values of 15 and 10 mmol/L (at 1200 and 2300 h, respectively). These changes are reflected by an overall increase in HbA1c at the third quarter. Although the target HbA1c parameter is surpassed with a 12-month result of 53 mmol/mol, SMBG data depict room for further action. In particular, IQRs are higher at 0600 and 2300 h than at 1200 and 1600 h (5.2 and 3.8 vs. 2.0 and 2.4 mmol/L, respectively). Increased variability in glycemia of 9.5 mmol/L occurred at 0600 h for this final quarter. Features of effective graph design principles, such as the smallest effective difference, were applied to graph 2 in order for key information (such as modal data, axis units, and labels) to be made prominent and more easily discriminant. Finally, graph 2 provides further statistical information and indicates the extraneous requirement for arbitrary labels (such as poor and out of control) used in some clinical appointments as descriptors of HbA1c category dispersions.
Conclusion:
Overall, Tukey-style boxplots presented together with a line graph afford a person with diabetes the opportunity and expectation to visually interpret and explain their glycemic control.
Dose-dependent Rise in Plasma Insulin Levels Following Preprandial Treatment of Type 1 Diabetes Patients with Oral Insulin Capsules (ORMD-0801)
Miriam Kidron, PhD; Ehud Arbit, MD; Daniel Schurr, MD; Camil Fuchs, PhD; Shoshi Shpitzen, MSc; Avram
Hershko, MD, PhD
Oramed Pharmaceuticals
Jerusalem, Israel
Objective:
We determine whether oral insulin (ORMD-0801) has a dose-dependent effect in type 1 diabetes mellitus (T1DM) patients.
Method:
ORMD-0801 was preprandially administered (t = −15 min) to 10 fasting T1DM patients, to final doses of 8, 16 (two 8-mg capsules), and 24 (one 8-mg and one 16-mg capsule) mg insulin, at three independent visits. Plasma glucose and insulin concentrations were monitored throughout the ensuing 5-h period. Pairwise comparisons of the responses to the various treatments were performed.
Result:
No adverse events were reported throughout the study period. Two patients were irresponsive to treatment and were excluded from all analyses. A dose-dependent response to ORMD-0801 was consistently observed, with significantly greater responses to the 24-mg dose. Mean insulin levels, Cmax, and area under the curve were, respectively, 47, 11.3, and 46 higher following 24-mg treatment when compared with the 8-mg dose (P ≤ 0.05) and, respectively, 35.9, 10.3, and 35.7% higher when compared with the 16-mg dose (P ≤ 0.01). In addition, the rate of decline in plasma insulin was slower following the 24-mg dose, with mean end-of-session insulin levels (1.6 ± 1.6 mU/mL) well above baseline values when compared with the other doses (−1.0 ± 1.1 and 0.7 ± 1.6 mU/mL, for 8 mg and 16 mg, respectively). Peak insulin levels were obtained within 105–120 min of dosing and coincided with peak glucose concentrations. In all three sessions, glucose concentrations gradually declined to baseline or near-baseline levels by the end of the monitoring period.
Conclusion:
The basic ORMD-0801 formulation provides for a dose-dependent bioavailability and bioefficacy of insulin preprandially delivered to T1DM patients.
Improved Prediction Accuracy for Dynamic Bayesian Network Glucoregulatory System Model
Robert Kircher, MS; Justin Lee, BS; Don Matheson, MS; Richard Mauseth, MD
Dose Safety Inc.
Redmond, Washington
Objective:
We aim to improve the prediction accuracy of the statistical glucoregulatory system model. The prediction algorithm will be used within the hypoglycemia prevention module of the Dose Safety artificial pancreas controller.
Method:
We described the prediction algorithm and initial prediction accuracy results in a 2014 Advanced Technologies and Treatments for Diabetes poster. Changes to the algorithm included new blood glucose input signal-smoothing filter and revised feature vectors (FVs) for improved pattern matching of the glucose and insulin on board input signals and normalizing of the lambda and gamma model vectors to numerically bound the probability calculations. The blood glucose signal-smoothing filter uses a weighted average of the previous six glucose values to better reflect the trend of the glucose signal. The inherent delay caused by the filter is acceptable. The glucose and insulin on board 4-tuple (45 min) glucose FVs were revised to better conform to the rates and accelerations inherent in the input signals. The combined effects of the smoothing filter and the revised FVs produced improved FV matches within the perceptron-learning algorithm. The FV matching is reflected in the 1 × 15,000 real-valued lambda and gamma vectors representing the statistical model combining glucose and insulin dynamics. The improved FV matching caused the probability calculations to be unbounded, which were corrected by normalizing the values.
Result:
The revised algorithm produced mean prediction errors (standard deviation) for the 10 test subjects of 5.2% (4.8), 8.0% (6.9), and 10.7% (9.4) for predictions of 30, 45, and 60 min, respectively. This represents prediction accuracy improvements of 20.5, 22.4, and 21.5%, respectively, over the data previously presented.
Conclusion:
The prediction accuracy of the statistical glucoregulatory system model was substantially improved.
Results from an Interview-Based Survey of Gla-300 SoloSTAR Compared with Three Commercialized Disposable Insulin Pens
David Klonoff, MD; Irina Nayberg, RN, CDE, CDTC; Frank Erbstein, PhD; Anna MG Cali, MD, MSc; Claire Brulle-Wohlhueter, MD; Thomas Haak, MD
Diabetes Research Institute
Mills-Peninsula Health Services
San Mateo, California
Objective:
We evaluate perceptions of people with diabetes (users) and health care providers (trainers) on usability of Gla-300 SoloSTAR disposable insulin pen, compared with Lantus SoloSTAR, FlexPen, and KwikPen.
Method:
A 75-min quantitative face-to-face interview was conducted with trainers (64 [34%] nurses, 63 [33%] endocrinologists, 59 [31%] primary care physicians, and 4 [2%] pharmacists) with experience using≥1 of the devices tested and users (228 [90%] type 2 diabetes patients [of whom 128 (56%) were pen naïve] and 26 [10%] type 1 diabetes patients [of whom 3 (11%) were pen naïve]). Following moderator demonstration, participants assessed pen usability and ranked the devices in 10 different categories.
Result:
Four hundred forty-four participants (users, n = 254; trainers, n = 190) from six countries (France, Germany, Spain, U.K., U.S., and Japan) were interviewed. More users ranked Gla-300 SoloSTAR first for “ease of use” (44%), “ease of injection” (49%), and “least effort to press down plunger” (55%) versus Lantus SoloSTAR (22, 20, and 17%, respectively), KwikPen (18, 17, and 18%), and FlexPen (17, 14, and 10%). More trainers also ranked Gla-300 SoloSTAR first for “ease of use” (49%), “ease of injection” (59%), and “least effort to press down plunger” (69%) versus Lantus SoloSTAR (26, 18, and 13%), KwikPen (8, 13, and 11%), and FlexPen (17, 10, and 8%). Fewer users ranked Gla-300 SoloSTAR first for “visualizing how much insulin is left” versus all tested pens and “hearing the dial turning” versus KwikPen and FlexPen. The majority experienced no problems preparing Gla-300 SoloSTAR and delivering the dose.
Conclusion:
Gla-300 SoloSTAR was ranked first by more participants as being the easiest to use and inject and requiring least effort to depress the plunger than the three other disposable pens tested.
Duality of Interest:
This market research project was conducted by Research Partnership and funded by Sanofi. F.E., A.M.G.C., and C.B.-W are employees of Sanofi.
Sensor Fusion and Zone Model Predictive Controller for Targeted Glucose Control in Critical Care
Timothy Knab, BS; Gilles Clermont, MD; Samantha Weiss, BS; Matthew Urich, BS; Robert Parker, PhD; Ari Pritchard-Bell, BS
Department of Chemical and Petroleum Engineering
Swanson School of Engineering
University of Pittsburgh
Pittsburgh, Pennsylvania
Objective:
Targeted glucose control and reduced glycemic variability have been associated with reduced morbidity and mortality in critical care patients. To minimize the frequency and magnitude of glycemic excursions from a target zone, we developed virtual patients from high-frequency continuous glucose monitoring (CGM) data for use in testing a linear zone model predictive controller to automate administration of intravenous insulin and glucose in the presence of glucose sensor signals that display considerable drift and variability.
Methods:
An adaptive multirate Kalman filter was developed to fuse CGM data from two subcutaneous sensors, producing a more reliable and less noisy single signal. The filter was evaluated on over 1,200 h of CGM data (5-min sample interval) from 24 patients. Virtual patients were constructed by fitting a dynamic model of glucose-insulin dynamics to the data and adjusting insulin sensitivity and basal glucose production.
Results:
The filtering algorithm produces a single composite signal exhibiting an increased signal-to-noise ratio from interstitial glucose measurements from two sensors. Using the filtered signals, we were able to construct virtual patients with a normalized root mean squared deviation from the CGM data of 0.05 mg/dL. Controller evaluation in silico on these virtual patients, with performance based quantitated as percentage of time within the normoglycemic zone (110–130 mg/dL) and the number of hypoglycemic events, is ongoing.
Conclusion:
Filtering and fusion of noisy data is crucial for making better decisions. A zone model predictive controller using a fused signal is expected to reduce glycemic variability and hypoglycemic incidents in simulated patients and is an important step toward the realization of automatic glucose control in critical care patients.
Effects of the Application of Whole- Body Vibration on the Balance of Type 2 Diabetes Patients with Peripheral Neuropathy
Amin Kordi Yoosefinejad, PhD; Azadeh Shadmehr, PhD; Ghloamreza Olyaei, PhD
Shiraz University of Medical Sciences
Shiraz, Fars, Iran
Background:
The human being has always encountered diabetes. In the 21st century, diabetes seems to be uncontrollable. It has several secondary complications, such as peripheral neuropathy, with a prevalence of 32/3%. The importance of physical activity in the treatment of diabetes has been known for many years, but patients are not willing to do exercises. Whole-body vibration (WBV) might be a suitable substitution. The aim of this study was to evaluate the effects of WBV on the balance of diabetic patients.
Methods and Materials:
Twenty type 2 diabetes patients matched on age, sex, degree of neuropathy, and BMI were assigned either to the treatment (n = 10) or control (n = 10) group. Balance and strength parameters were measured at baseline. The treatment group received WBV for 6 weeks and twice per week with a frequency of 30 Hz and the amplitude of 2 mm. The control group did not receive any special treatment during this 6-week period.
Results:
The balance that was evaluated with TUGT and UST showed improvement in comparison with the control group. Mean velocity decreased in the treatment group postexercise, and the pattern was different between groups. Muscle strength increased in the treatment group in comparison with the control group.
Conclusion:
The application of WBV with a frequency of 30 Hz and the amplitude of 2 mm for 6 weeks twice per week can improve the balance and the strength of type 2 diabetes patients with mild or moderate peripheral neuropathy.
Multimarker Electrochemical Sensor for Artificial Pancreas
Jeffrey T. La Belle, PhD; Curtiss B. Cook, MD
School of Biological and Health Systems Engineering
Arizona State University
Tempe, Arizona
Objective:
Current diabetes management is limited, measuring only one marker, glucose, which can be monitored rapidly and continuously. No similar technology exists for detection of glucagon, insulin, or a biomarker panel. Current artificial pancreases are based on measurements of a system output (glucose) rather than inputs (insulin, glucagon). Incorporating all biomarker panels as parameters in nonlinear models of closed-loop insulin delivery systems would improve accuracy and precision.
Method:
Electrochemical impedance spectroscopy (EIS) requires the covalent immobilization to attach a molecular recognition element to the sensor. Then the appropriate target can be introduced in a concentration gradient while EIS is run. Data are analyzed to determine responsivity and degree of fit at each frequency, and the optimal frequency is selected per marker.
Result:
A variety of purified markers pertinent to diabetes have been studied (replicated) and pertinent parameters taken, namely, dynamic range, optimal binding frequency, and lower limits of detection. The current biomarkers under study include 1,5-anhydroglucitol, glucose, insulin, glycated albumin, and HbA1c at their physiological levels.
Conclusion:
Currently, a biomarker panel has been measured via EIS in purified form. Expectations include deepening knowledge of the dynamic behavior and feedback control of insulin (and its analogs) for glucagon-glucose-insulin regulation models and improving point-of-care test strips for multibiomarker diabetes management.
Practical Lifelong A/V Access for a Forearm-Worn, Closed-Loop, Blood-Based Artificial Pancreas
Arnold J. Lande, MD; John Erickson, MS
Northport Navigable Waters Institution
Northport, Michigan
Objective:
We reference several clinically proven components of a blood-based artificial pancreas.
Method:
1) The bedside, blood-based (not subcutaneous) Biostator has since 1976 (Food and Drug Administration approved in 1983) demonstrated the ability to place a clamp on blood glucose level. 2) When a gold standard hemodialysis arteriovenous fistula (AVF) fails to mature or a matured AVF fails, emergency obstruction of flow-and-pressure-diverting accessory veins is routinely performed to route the highest available blood flow through a single outflow vein. This demonstrated, efficacious response to an emergency predicts that an identical elective in situ debranched vein fistula graft (VFG) will result in lifelong (with revisions) durable blood access. 3) Intermittent external compressions, over the VFG and between the in and outflow cannulae, make available the A/V pressure differential and thus eliminate the need for bulky, dangerous blood pumps and their necessary safety adjuncts. Turbulent flow-through at the arteriovenous anastomosis is interrupted frequently, reducing tissue reaction. There is no downstream anastomosis, so no downstream scarring and constriction. 4) Like the spider-web -appearing AVF, which is widely accepted by hemodialysis patients, technicians, nurses, and physicians, a more cosmetic single outflow VFG is even more likely to be tolerated. The blood-contacting disposable, with medication cartridges manufactured in, is easily replaced twice weekly by the patient at home to avoid biofilm, clotting, thrombosis, and the looming threat of drug-resistant septicemia.
Result:
Lifelong access to extracorporeal blood flow and A/V pressure differential is desirable for a forearm-wearable artificial pancreas, kidney, liver, congestive heart failure aquapheresis, and a variety of other phereses.
Conclusion:
Lifelong vascular access is already available in the form of several clinically proven components. Immense personal and public benefits stand to accrue from the integration and implementation of these components.
Artificial Pancreas Database
Gaven Larson, BS; Lauren Huyett, BS; Joon Bok Lee, BS; Eyal Dassau, PhD; Frank Doyle, PhD
Department of Chemical Engineering
University of California, Santa Barbara
Santa Barbara, California
Objective:
Research for an artificial pancreas (AP) has accelerated throughout the past decade. The variety of protocols and reported metrics has led to challenges in comparing results across different AP configurations, which is critical to improve future studies. To address this problem, a database and search engine of published AP trials from 2004 to present was created, allowing efficient scrutiny across multiple publications and providing a uniform structure for data reporting and study comparisons.
Methods:
Key information from peer-reviewed AP clinical trial publications (47 to date) was recorded into a MySQL database to support a searchable framework that provides user query capabilities with SQL flexibility. An accompanying website interface to find publications satisfying a combination of customizable criteria was also created. The website is augmented with tools that allow for deeper review of individual entries and incorporates administrative capabilities to update the database over time.
Results:
The AP database (http://thedoylegroup.org/apdatabase/) allows users to efficiently search relevant publications and extract the data for further comparison. This framework provides the capability to search for multiple key details simultaneously (e.g., number of subjects, meal bolus strategy, and times in range), the results of which can be viewed in a succinct, tabulated format.
Conclusion:
The AP database presents many complementary features to preexisting search engines by providing the ability to query and extract study design and results when searching for and analyzing clinical trial publications. Use of this database will provide significant streamlining of clinical trial design for novel AP designs. Newly published studies will be added to the database regularly, ensuring that the website contains the most up-to-date information in AP research.
A Novel Run-to-Run Algorithm for Enhanced Continuous Glucose Monitor Accuracy and Improved Calibration Based on Continuous Wear
Joon Bok Lee, BS; Eyal Dassau, PhD; Francis J. Doyle III, PhD
University of California, Santa Barbara
Santa Barbara, California
Objective:
Continuous glucose monitor (CGM) accuracy is critical in artificial pancreas design for blood glucose (BG) regulation. We present a run-to-run (R2R) CGM calibration algorithm that incorporates accumulated sensor data prior to each sensor reinsertion for enhanced CGM accuracy.
Method:
Raw sensor signals were simulated for 10 in silico subjects based on published patent data. Personalized calibration parameter sets (calibration curve slope and intercept, sensitivity drift curve, mean reference error) that convert week-long simulated raw signals to CGM values were identified to minimize summed square deviation from simulated finger sticks. Likelihood of each set was evaluated against known population variability of each parameter (truncated normal distribution), and the personalized set with highest combined likelihood was incorporated in the following week with simulated new sensor insertion.
Result:
The weekly R2R adjustment of calibration parameters was evaluated over six iterations for 10 in silico subjects. Each iteration significantly improved the CGM’s capability in capturing actual BG values, decreasing the mean summed square of residuals between actual BG and sensor values by 23% in the second week and converging to 77% improvement by sixth week, compared with standard CGM operation. Moreover, 40% additional values were within the A zone of the Clarke error grid analysis after one iteration, which improved to 47% more values after 6 weeks for a total of 95% of all values placed within the A zone.
Conclusion:
This novel approach exploits the wealth of personalized calibration data gathered over the course of each week that is currently discarded under standard reinsertion procedures and gives significant iterative improvements in the CGM’s ability to capture BG concentrations. Application of this R2R algorithm on clinical data is projected in the near future.
Microfabricated Microporous Membranes Prolong the Functional Lifetime of a Closed-Loop Insulin Delivery Implant in a Type 1 Diabetic Rat Model by Passively Modulating the Host Immune Response
Jason Li, MASc; Michael K. Chu, BA; Claudia R. Gordijo, PhD; Adria Giacca, MD, PhD; Oliver Plettenburg, PhD; Xiao Y. Wu, PhD
Department of Pharmaceutical Science
Faculty of Pharmacy
University of Toronto
Toronto, Ontario, Canada
Objective:
Closed-loop insulin delivery systems offer diabetic patients improved glycemic control, compliance, and quality of life; however, long-term efficacy has yet to be demonstrated due to poor device biocompatibility. Sustained administration of anti-inflammatory agents such as dexamethasone or nitric oxide may improve device lifetime; however, such strategies are accompanied with side effects and are difficult to implement. We demonstrate prolonged functional lifetime of a chemically driven closed-loop insulin delivery implant by passively impeding cell migration to the implant surface using a microfabricated microporous membrane.
Method:
Polydimethylsiloxane membranes featuring micron-sized pores were constructed using soft lithography and affixed overtop of a chemically driven closed-loop insulin delivery system comprised of an insulin reservoir sealed with a variable-porosity glucose-responsive albumin plug. Devices with and without the microporous membrane were implanted subcutaneously in healthy Sprague-Dawley rats and assessed histologically for tissue biocompatibility over 30 days. Device efficacy was assessed in a streptozotocin-induced diabetic rat model. Blood insulin, glucose, and C-peptide levels were measured for 30 days.
Result:
Microporous membranes minimized accumulation of cells at the glucose-responsive plug surface and subsequent cell-mediated degradation compared with unprotected devices. The membranes further resulted in resolution of the inflammatory process to the implant as evidenced by minimal residual perivascular lymphocytic and histocytic inflammation, absence of neutrophils at the implant site, and a thinner fibrous capsule by day 30 following implantation. Finally, use of the microporous membranes successfully prolonged the in vivo efficacy duration by up to three-fold (28 days) compared with unprotected devices.
Conclusion:
Microfabricated microporous membranes act as a passive barrier to cell migration to the immunogenic surface of a chemically driven closed-loop insulin delivery device, thereby improving implant biocompatibility and prolonging implant lifetime.
Photo-Patternable Hydrogels for Implantable Glucose Sensors
Zhe Li, MS; Liangliang Qiang, PhD; Sagar Vaddiraju, PhD; Diane J. Burgess, PhD; Fotios Papadimitrakopoulos, PhD
University of Connecticut
Storrs, Connecticut
Objectives:
The function and lifetime of implantable electrochemical-based continuous glucose monitoring (CGM) devices are intimately linked to the stability of the glucose oxidase (GOx) enzyme responsible for glucose detection. Enzyme denaturation and leaching, as well as biofouling-induced pore clogging, gradually decrease device 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 the use of a novel PEGylated hydrogel, which is capable of being photocrosslinked in place in its fully hydrated state. This hydrogel allows for CGM sensors with long-term stability and optimal performance.
Methods:
A PEGylated copolymer was synthesized via a free-radical polymerization using an azobisisobutyronitrile initiator. The polymer was mixed with GOx enzyme and photopatterned on the electrode (via ultraviolet exposure) followed by 24-h incubation to allow chemical coupling of GOx to the PEGylated hydrogel.
Results:
Glucose sensors using GOx-grafted hydrogels showed excellent stability during continuous testing for 30 days. Control experiments indicated no enzyme leaching over a test period of 30 days. The high molar ratio of PEG functionality affords good biofouling resistance, as investigated using BSA as model protein. The glucose sensors exhibited sensitivities as high as 320 nA/(mmol/L−1)/mm−2 with linearity beyond 25 mmol/L of glucose and limit of detection less than 1 µmol/L.
Conclusions:
Photopatternable PEGylated hydrogels were demonstrated as ideal matrices for covalent immobilization of GOx enzyme with high spatial tolerances, enabling the facile fabrication of CGM devices. Long-term in vivo studies are currently underway to determine the ultimate stability of these photopatternable CGM devices.
Measurement Accuracy of Two Blood Glucose Monitoring Systems with Built-In Bolus Calculators
Manuela Link, ME; Annette Baumstark, PhD; Cornelia Haug, MD; Stefan Pleus, MS; Christina Schmid, PhD; Nina Jendrike, MD; Guido Freckmann, MD
Institut für Diabetes-Technologie Forschungs- und Entwicklungsgesellschaft mbH
Universität Ulm
Ulm, Germany
Objective:
Systems for self-monitoring of blood glucose (BG) with built-in automated bolus calculators may facilitate the calculation of insulin doses by patients. The systems’ BG measurement accuracy is one important aspect for appropriate insulin bolus calculations. We evaluated the accuracy of two self-monitoring of BG systems with built-in bolus calculators following the standard ISO 15197.
Method:
Accuracy of system 1 (Accu-Chek Aviva Expert [Roche Diagnostics GmbH]) and system 2 (FreeStyle InsuLinx [Abbott Diabetes Care Inc.]) was evaluated with three reagent system lots each. Measurement results obtained on 100 capillary blood samples were analyzed applying ISO 15197 criteria by calculating the percentage of results within ±20% or ±15 mg/dL of the comparison method measurement results for BG concentrations above or below 75 mg/dL (ISO 15197:2003), respectively, and within ±15% or ±15 mg/dL above or below 100 mg/dL (ISO 15197:2013), respectively.
Result:
Both systems showed between 99 and 100% of results within accuracy limits of ISO 15197:2003. Applying the more stringent criteria of ISO 15197:2013, system 1 showed between 99 and 99.5% and system 2 between 88 and 100% of results within the limits. Both systems showed 100% of results within consensus error grid zones A and B. The relative bias according to Bland-Altman ranged from −3 to −1.6% for system 1 and from −8.4 to +0.1% for system 2.
Conclusion:
Both systems fulfilled accuracy criteria of ISO 15197:2003 with the three tested lots. ISO 15197:2013 accuracy criteria were fulfilled by all lots of system 1 and by two lots of system 2. Mandatory compliance with the recently published ISO 15197:2013 is recommended after a transition period of 36 months.
Technology-Enabled Type 1 Diabetes Education and Support System
Terri H. Lipman, PhD; Tina L. Harralson, PhD; Tammy Mawson, MA; Ashley Hooker, BA; Laura Dietzen, MS; Kathryn M. Murphy, PhD
University of Pennsylvania
Philadelphia, Pennsylvania
Objective:
The purpose of this phase 1 study was to develop a web-based diabetes management (DM) interface between patient and clinicians to support DM in adolescents with type 1 diabetes (T1D). Phase 2 of this study will include electronic health record integration and interactive patient-provider messaging.
Method:
Forty patients (13–18 years) with T1D and 40 caregivers were enrolled in the pilot study to test the system. This included patients completing the T1D Education and Support (T1DES) assessment on a tablet computer, with patients and clinicians receiving real-time reports highlighting patient knowledge deficits, psychosocial issues, and barriers to successful DM management. Patients received daily e-messages tailored to their selected diabetes goals. Patients and parents also had access to e-learning tools and an interactive webinar with a registered dietitian. Clinician reports included information on patient insulin and sick-day management, diet and exercise, communication with parents, depression, distress, barriers to DM, and DM knowledge and goals.
Results:
Based on the T1DES assessment, 33% of the youth reported diabetes-related distress and 13% reported symptoms of depression. Of pump patients, 57% reported missing bolus doses at least once a week; 25% of injection users reported missing an injection at least once a week. Reasons and barriers for missing injections and bolus doses and not adhering to diet or exercise recommendations included lack of knowledge and frustration with DM and were listed on the clinician reports.
Conclusion:
The automated T1DES system was well accepted and demonstrated the ability to identify psychosocial issues and barriers to DM. Once topics of concern were identified, clinicians reported improved ability to focus patient visits and provide needed support. Data from phase 1 will inform the development of the phase 2 system.
Patient Preference Assessment of a Long-Term Fully Implantable Continuous Glucose Monitoring System
Joseph Lucisano, PhD; Robert Engler, MD; Timothy Routh, MEng
GlySens Inc.
San Diego, California
Objective:
The objective of this study was to assess perceptions of potential users regarding a novel long-term fully implantable (no skin-attached elements) continuous glucose monitoring (CGM) system (the GlySens ICGM system). Such perceptions can be important determinants in influencing eventual user adoption.
Method:
A survey (fielded by dQ&A Market Research Inc.) of 701 insulin-using diabetic subjects, randomly sampled from a panel of 10,000 U.S. patients, included three cohorts: 1) CGM current users (n = 282); 2) CGM ex-users (n = 206); and 3) people who have never used CGM (n = 213). After reviewing a complete system description with device photograph (discussing brief outpatient surgical implant, yearly sensor replacement, once or twice monthly calibration, and other system features), subjects indicated if they would definitely, likely, unlikely, or definitely not adopt the ICGM system. Additional questions to identify most and least attractive system characteristics were also included.
Result:
Definite or likely adoption of the ICGM system among the three cohorts was reported as 1) 90%, 2) 88%, and 3) 91%. The most user-attractive features were related to freedom from device maintenance/nuisance (e.g., replacement only annually, no skin-attached element). The feature most significant in limiting adoption for those indicating unlikely or definitely not (72 of 701) was the need for surgical implant (n = 42). Adoption assessments were consistent regardless of age or type of diabetes. A selection bias toward subjects who practice good glucose control is inherent in the methodology, but the surveyed population nonetheless represents a very large and growing segment of people with diabetes.
Conclusion:
The study results indicate that the ICGM system could potentially achieve broad and deep adoption with patients motivated to achieve improved glucose control. The requirement for annual minor surgical replacement of the ICGM sensor would pose only a minimal barrier to adoption.
Introducing an Exercise Model for the Artificial Pancreas
Dayu Lv, PhD; Marc D. Breton, PhD
Center for Diabetes Technology
University of Virginia
Charlottesville, Virginia
Objective:
Exercise has been a great challenge to the tight control of glycemia because of its multiple and variable effects on glucose homeostasis. An accurate model of exercise would provide a useful tool to predict the impact of exercise and to improve artificial pancreas design. As similar efforts have been addressed on in silico populations, the aim here is to 1) establish exercise impacts on insulin-dependent/independent utilization and 2) validate the model.
Method:
A bout of mild exercise stimulates the translocation of GLUTs to muscle membranes, increasing both insulin-independent and insulin-dependent glucose uptake with different time and magnitude scales. These effects of exercise were incorporated into a model of glucose/insulin dynamics as actions at the sites of insulin-independent glucose utilization and insulin-dependent maximal glucose transport rate. Simulations were performed using the so modified University of Virginia/Padova type 1 diabetes mellitus (T1DM) simulator for 100 adult in silico subjects. The distribution of glycemic drop during exercise was then compared with glucose concentration in 32 data sets of frequent YSI measurements in T1DM during mild to moderate exercise.
Result:
Simulation results were obtained for 30 min of mild exercise for each subject. A metric was created that used the slope of the blood glucose (BG) values during the 1 h before exercise, the starting BG, the BG drop after exercise, and insulin on board at the start of exercise. We were able to demonstrate similar distributions of this metric between the simulations and the clinical data (two-sample Kolmogorov-Smirnov test = 0; P = 0.79831).
Conclusion:
A new model component of exercise has been developed for simulation of mild to moderate exercise and validated on clinical data. This could facilitate the design and analysis of decision support system for exercise in T1DM.
Glycemic Control and Risk of Hypoglycemia in Hospitalized Patients with Type 2 Diabetes Mellitus at Different Wards Using the GlucoTab System
Julia K. Mader, MD; Katharina M. Neubauer, MSc, BSc; Lukas Schaupp, PhD; Felix Aberer, MD; Klaus Donsa, MSc; Stephan Spat, MSc; Bernhard Höll, MSc; Peter Beck, PhD; Johannes Plank, MD; Thomas R. Piener, MD
Medical University of Graz
Graz, Styria, Austria
Objective:
The aim of this study was to compare glycemic control and risk of hypoglycemia using the GlucoTab system in patients with type 2 diabetes mellitus (T2DM) hospitalized at different general wards.
Method:
The GlucoTab system is a tablet-based decision support system for glycemic management in the hospital setting based on a basal-bolus algorithm. According to the GlucoTab, blood glucose (BG) measurements were performed four times daily (prebreakfast, prelunch, predinner, and bedtime); insulin injections were given according to the GlucoTab suggestion. Advice for total daily dose (50% basal insulin, 50% premeal bolus insulin) was generated once daily.
Result:
The study was performed at four different general wards (A–D) in 99 T2DM patients. Patient characteristics were as follows: ward A, 42 patients (age 67 ± 11 years, HbA1c 70 ± 24 mmol/mol, BMI 31 ± 7 kg/m2, diabetes duration 15 ± 10 years); ward B, 30 patients (age 70 ± 12 years, HbA1c 65 ± 21 mmol/mol, BMI 29 ± 7 kg/m2, diabetes duration 11 ± 8 years); ward C, 15 patients (age 64 ± 8 years, HbA1c 57 ± 10 mmol/mol, BMI 31 ± 5 kg/m 2, diabetes duration 13 ± 8 years); and ward D, 12 patients (age 65 ± 10 years, HbA1c 55 ± 13 mmol/mol, BMI 29 ± 6 kg/m2, diabetes duration 14 ± 10 years). Mean BG was 150 ± 35 mg/dL (A), 163 ± 33 mg/dL (B), 162 ± 34 mg/dL (C), and 134 ± 31 mg/dl (D). The overall risk of getting a BG value <70 mg/dL during the first 10 treatment days was 2.4% (A) vs. 0.2% (B) vs. 1.0% (C) vs. 1.2% (D). No severe hypoglycemic event (BG <40 mg/dL) occurred in any of the patients.
Conclusion:
The GlucoTab system could establish glycemic control when used in patients with T2DM hospitalized at different general wards. The risk of getting a BG value <70 mg/dL is not uniquely associated with level of glycemic control but seems to be influenced by other factors (comorbidities, diabetes duration, severity of disease).
Funding:
This study was supported by the European Commission (Project REACTION, FP7 248590).
Exploring Continuous Glucose Monitoring on the Frequency Domain to Identify Risk Factors in Type 2 Diabetes
Miguel Maria Isabel, BS candidate; Jorge Cancela, PhD candidate; Giuseppe Fico, PhD candidate; Andrea Facchinetti, PhD; Chiara Fabris, PhD candidate; Claudio Cobelli, PhD; Maria Teresa Arredondo, PhD; on behalf of the MOSAIC Consortium
Life Supporting Technologies
Universidad Politecncia de Madrid
Madrid, Spain
Objective:
Glucose variability (GV) is believed to be a key indicator of risk factors in both individuals with type 1 diabetes mellitus and individuals with type 2 diabetes mellitus (T2DM). Continuous glucose monitoring (CGM) devices allow a better characterization of GV thanks to the almost continuous monitoring of glucose concentration. A review of the literature revealed that all GV indices derived from CGM data are calculated on the time domain, while the frequency domain is still unexplored. This work is aimed to assess the CGM signal on frequency domain trying to identify new features for a better characterization of T2DM.
Method:
The proposed frequency analysis consists of transforming CGM signals to the frequency domain using the fast Fourier transform and mining several spectrum parameters: the maximum power spectral density and the frequency where it is located, power spectral density of the second spectrum peak and its relative frequency, and the difference between the maximum and the second peak of the spectrum.
Result:
The method was applied on 11 T2DM subjects recruited during the METABO research project monitored with CGM for 3 days in two sessions 1 month apart, before and after the intervention. Comparison of power spectra evidenced narrower and higher spectra after the intervention. This is in line with expectations, being the spectrum of a regular signal (i.e. with reduced GV) is less wide.
Conclusion:
This introductory study evidenced that frequency analysis of CGM data is a candidate tool for the evaluation GV. Further work should be carried out on the validation of frequency parameters as GV indexes, e.g., exploiting the upcoming CGM data sets that will be collected within the MOSAIC project funded by the European Union under the FP7 framework.
A Multivariate Data-Driven Model to Investigate the Arising of Complications in Type 2 Diabetes Patients
Simone Marini, PhD; Milad Malavolti; Arianna Dagliati, MS; Riccardo Bellazzi, PhD
Biomedical Informatics Laboratories
Univeristy of Pavia
Pavia, Italy
Objective:
We present a data-driven model based on a combination of multivariate clinical time series to simulate the arising of complications in type 2 diabetes (T2D) patients. The framework is based on continuous time Bayesian networks (CTBN). The CTBN model describes stochastic processes evolving in continuous time. Our model in particular exploits the combination of medical knowledge with data-driven information. On the one hand, network structure and clinical state thresholds are based on previous medical knowledge; on the other hand, the probabilities of transaction between the network states are learned from data. We tested our framework on a data set of 500 T2D patients followed by the Fondazione Salvatore Maugeri for periods between 5 and 10 years.
Method:
The network structure is conceived out of medical guidelines. The model aims to represent how changes over time of a simple set of clinical variables (cholesterol, systolic pressure, HbA1c) can influence the onset of T2D complications (retinopathy, neuropathy, nephropathy). While clinical variables are represented by dichotomies (in/out of range), complications are depicted as multistate nodes, their state depending on the cumulative number of adverse events. Raw data were preprocessed with temporal abstractions techniques, which allow detecting time intervals within a variable as a qualitative state (HbA1c “out-target” for 3 months).
Result:
The model correctly fits the data and replicates their trajectories in simulations. Both clinical variables and complications have an average error of less than 10% over a 10-year period.
Conclusion:
In this work, we have jointly applied CTBN and temporal data mining techniques to discover how changes in patients’ clinical states might influence complications onset. Future efforts will be dedicated to applying the methods on a larger population, also stratified on the basis of cardiovascular risk.
GlucoCare-140: A Target, Not a Target Range, for Glycemic Control in the Intensive Care Unit
Michael R. Marvin, MD; Brian Besterman, MD; Silvio Inzucchi, MD
University of Louisville
Pronia Medical Systems
Louisville, Kentucky
Objective:
GlucoCare-140 is designed to target blood glucose (BG) to 140 mg/dL. Our objective is to compare the results of GlucoCare-140 to the GlucoCare-directed Yale 100–140 mg/dL protocol and a modified-Yale 120–140 mg/dL protocol.
Method:
All intensive care unit patients at a single institution whose BG levels were managed by GlucoCare (a Food and Drug Administration–cleared insulin dosing calculator; Pronia Medical Systems) were included in this study. Three protocols were studied with the following targets: 100–140, 120–140, and 140 mg/dL. BG was excluded if it was clearly entered in error (e.g., BG of 1 mg/dL) or was >75% delayed (e.g., 45-min late for hourly readings) or if patients were terminally ill and support was withdrawn unrelated to hypoglycemia.
Result:
Overall, there were 44 patients/1,657 readings, 1,375 patients/55,952 readings, and 113 patients/4,263 readings in the three protocol groups, respectively. Mean BG was 130.2 ± 38.2, 135.7 ± 33.0, and 146.2 ± 35.0 mg/dL, respectively. Hypoglycemia as a proportion of all BG determinations defined as <40 mg/dL decreased from 0.06 to 0.01 to 0.00% and as <70 mg/dL from 1.5 to 0.34 to 0.03%, respectively. Hypoglycemia as a proportion of patients <40 mg/dL decreased from 2.33 to 0.29 to 0.00% (P = 0.18, Fisher exact test) and those experiencing <70 mg/dL from 30.23 to 8.08 to 0.90% (P < 0.001). If BG was performed within 50% of the recommended interval (e.g., 30 min for hourly checks), no episode of hypoglycemia <70 mg/dL occurred with GlucoCare-140.
Conclusion:
Use of GlucoCare-140 completely eliminated hypoglycemia <40 mg/dL and nearly eliminated hypoglycemia <70 mg/dL, while achieving a mean BG of 146 mg/dL. These results are in keeping with the goals of the American Association of Clinical Endocrinologists/American Diabetes Association (140–180 mg/dL) and Society of Critical Care Medicine (<150 mg/dL) recommendations.
Smart Low Glucose Suspend Capabilities of the Dose Safety Fuzzy Logic Dosing Controller
Richard Mauseth, MD; Robert Kircher, MS; Don Matheson, MS
Dose Safety Inc.
Redmond, Washington
Objective:
We develop a capability that automatically suspends insulin delivery when hypoglycemia is likely and then resumes delivery when the blood glucose response shows it is safe to do so.
Method:
Other low glucose suspend (LGS) systems automatically stop insulin delivery when a predetermined glucose value is reached or is predicted. The pump suspends insulin infusion for an arbitrary period of time in the absence of user intervention. The fuzzy logic dosing module (FLDM) dosing rules matrix within the Dose Safety artificial pancreas controller attenuates dosing based on absolute glucose value, rate, and acceleration. Instead of using an arbitrary duration of suspension, our FLDM resumes dosing when the blood glucose level, rate, and acceleration indicate that it is safe to do so. For example, if the rise is slow, then the dosing is minimal, and if the rise is more rapid, the dosing is increased. These dosing capabilities are embedded into the overall design of the FLDM, i.e., dosing matrix.
Result:
Seventeen 9-h daily living studies were conducted in the Benaroya Research Institute clinical research center. The smart LGS feature of the FLDM shut off insulin in five of the studies. During the studies, insulin was shut off at 172, 117, 100, 193, and 198 mg/dL, thereby automatically preventing blood glucose from going below 60 mg/dL. In each case, insulin dosing automatically resumed when the glucose dynamics indicated the need for insulin.
Conclusion:
The smart LGS capability built into the Dose Safety FLDM design is effective at preventing hypoglycemia and reinstituting insulin dosing in clinical studies involving real-life daily living activities.
Telephone-Based Insulin Management Program
Chhavi Mehta, MD; Shetal Desai Rautela, RD, CDE, Steven Sun, MS
Palto Alto Medical Foundation
Mills-Peninsula Division
Burlingame, California
Objective:
We assess the short-term effects on patients’ glycemic control and disease management behaviors of increasing the frequency of contact (face-to-face and telephone) in patients with insulin-dependent type 2 diabetes in a primary care setting.
Method:
Patients with inadequately controlled (HbA1c >7.5) insulin-dependent type 2 diabetes (n = 31) were contacted by letter and, as needed, phone calls from a patient service representative or the patient’s primary care physician. A dietician (RD, CDE) held an initial 1-h face-to-face visit with each patient, followed by weekly telephone calls over 3 months to counsel the patient, as needed, on nutrition, weight loss, and set/monitor behavioral goals and to review patients’ self-monitoring of blood glucose logs. Dietician recommendations for insulin titration were reviewed and approved by the primary care physician through the Epic electronic health record system. HbA1c values were measured at enrollment and the 3-month follow-up. The dietician used an electronic health record–embedded numerical scale to track progress in meeting the behavioral goals. Some patients could choose more than one goal over the period of the program. A score of 7 or more was marked as successful.
Results:
Of 31 approached patients, 17 (55%) participated in both the face-to-face appointment and subsequent telephone calls and contributed baseline and follow-up blood samples. The preprogram HbA1c mean was 9.7% (range 7.6–14.5%), and the postprogram mean was 8.56% (range 6.0–11.1%), a mean (SD) decrease of 1.16% (1.61) with a range from −5.7 to +0.7 (P = 0.01; n = 17; paired t test). Patients were able to reach the behavioral goal they picked 75% of the time.
Conclusion:
Adoption of a primary care–embedded, telephone-based insulin management program as a quality improvement project was well received and provided evidence of improved patient disease management behavior and clinically beneficial reduction in mean HbA1c.
Inhaled Insulin: Recovered from Deep Coma
Sultan Ayoub Meo, MBBS, PhD, FRCP (London-Dublin-Glasgow-Edinburgh)
Department of Physiology
College of Medicine
King Saud University
Riyadh, Saudi Arabia
Background:
Diabetes is a rapidly growing major public health problem with increasing incidence and long-term complications. In spite of marvelous advancement in medical sciences, it is still an incurable and lifelong disease. It is swiftly increasing in all age groups, among both sexes, and in developing as well as developed countries. Presently, the new estimates of prevalence of diabetes are reaching sky-high; globally 382 million people are diabetes patients, and the number is expected to rise to 592 million by 2035.
Method:
Multiple routes for insulin administration have been experimented. In 2006, inhaled insulin (Exubera) was approved by the Food and Drug Administration. However, in October 2007, Pfizer announced it would be taking Exubera off the market, citing that the drug had failed to gain market acceptance. The main reasons for taking it off the market was adverse effects, including cough, shortness of breath, sore throat, dry mouth, lung function impairment, hypoglycemia, and lung cancer among ex-smokers. More recently, in June 2014, once again the Food and Drug Administration approved an inhaled insulin, Afrezza, a rapid-acting inhaled insulin, but did not recommended it for patients who smoke or who have chronic respiratory disease.
Result:
The most common adverse reactions associated with Afrezza are hypoglycemia, cough, and throat pain or irritation.
Conclusion:
Inhaled insulin has major challenges; the research community did not investigate the impact of inhaled insulin on lung functions and highly sensitive lung inflammatory biomarkers, namely, fractional exhaled nitric oxide. Although inhaled insulin has recovered from deep its coma, respiratory physiologists still doubt its survival.
MyDiaText: Feasibility of a Text Messaging System for Youth with Type 1 Diabetes
Kathleen A. Montgomery, MSN, CRNP; Reid Simon, BSE; Kelly Lord, RD, CDE, LDN; Terri H. Lipman, PhD, CRNP
Division of Endocrinology and Diabetes
The Children’s Hospital of Philadelphia
Philadelphia, Pennsylvania
Objective:
Text messaging (short message service [SMS]) can promote healthy behavior change. The purpose of this study was to determine the feasibility and functionality of receiving SMS educational and motivational messages to support behavior change in youth with type 1 diabetes.
Method:
Patients with type 1 diabetes aged 10 to 17 years, diagnosed within 6 months, were recruited by phone or at a diabetes education class. A web user profile was created and a behavioral goal selected. A test text message was sent to verify registration. Subjects received a daily text message related to their goal, and twice weekly responded to a “rate your progress” text using a numeric scale 1–10. Progress was tracked in graph format on the subject’s profile page. Participants could complete optional quizzes and view additional resources to earn points. Certificates were provided to those who accumulated 100 points or more. After 1 month, participants completed an online questionnaire to determine their level of satisfaction.
Results:
Nine subjects have been recruited (five male, four female), ranging in age from 10 to 17 years. A variety of behavioral goals were selected. All have received daily text messages, responded to “rate your progress” texts, and accrued points during the participation period. Five subjects have completed a satisfaction survey, stating they enjoyed receiving texts and earning points and found the information helpful. However, one respondent wanted to receive fewer text messages.
Conclusion:
SMS is a satisfying, cost- effective, and feasible methodology to promote behavioral change in youth with type 1 diabetes. Further study is needed to determine the effectiveness of SMS in sustaining positive diabetes behavior change.
Measurement of Glucose Concentrations in Saliva via Infrared Spectroscopy
David A. Mucci, BS, MD; William A. Petit Jr., AB, MD; Ronald Clark, BS, MD; Jeanne T. Mucci; J. Scott Fox, BA
Quick LLC
Farmington, Connecticut
Objective:
The objective is to develop a bloodless, noninvasive, accurate glucose monitoring system
Method:
The methodology of this glucose testing system uses state-of-the-art advanced capability infrared technology combined with a uniquely designed polymer saliva holding system. Infrared spectroscopy is used to directly measure the glucose concentrations in the saliva of patients with diabetes. A saliva collection device (Draw Wick) uses a microfluidic technique to obtain a saliva specimen filling the window of an infrared neutral strip. The Draw Wick is then inserted into the infrared analyzer, in a similar fashion as inserting strips into a blood glucose meter, to directly measure the glucose concentration in the sample. The result is available within seconds. Testing to date indicates diabetes patients must refrain from food and nonwater consumption prior to attaining a saliva sample with a Draw Wick.
Result:
Initial patient testing of saliva samples versus existing glucometer technology has achieved an R value of 0.94 when compared with the existing monitoring technology.
Conclusion:
Infrared spectroscopy combined with saliva sampling can be used as a bloodless, noninvasive means to determine glucose levels in diabetes patients. Further testing on larger populations of patients with diabetes is required to refine saliva analyzer calibration.
Continuous Glucose Monitoring Greatly Improves Sensitivity for Detection of Reduced Prandial Excursions in Response to Infusion-Site Pretreatment with Recombinant Human Hyaluronidase (rHuPH20)
Douglas B. Muchmore, MD; Simon R. Bruce, MD; Ya Huang, MS; Xionghua W. Wu, PhD; David Rodbard, MD
Halozyme Therapeutics Inc. San Diego, California
Objective:
We conducted a clinical trial comparing recombinant human hyaluronidase (rHuPH20) pretreatment of continuous subcutaneous insulin infusion (CSII) sites to usual CSII in subjects with type 1 diabetes. We sought to compare postprandial glucose (PPG) excursions obtained with the ultrafast insulin profile associated with rHuPH20 pretreatment (pre-Rx) to that of usual CSII with rapid-acting analog insulin alone. Here we compare the sensitivity of CGM profiles obtained in a 25% subset of subjects to a self-monitored blood glucose (SMBG) data set, including every SMBG value with >261,000 SMBG records from the 6-month study to assess improvement in PPG using rHuPH20.
Method:
Four hundred fifty-six subjects were randomized 3:1 to rHuPH20 pre-Rx at infusion set change or usual CSII; infusion set change was every 3 days. SMBG was mandated before meals and ~90 min after≥2 meals/day. SMBG values were uploaded to a central server with annotations to identify pre- versus postprandial values. Mean excursions were calculated for each subject. CGM data obtained from whatever system the subject was using prior to study were collected by direct uploads and merged with meal timing data obtained during 7 days prior to each of six visits.
Results:
CGM demonstrated a 20 mg/dL improvement in peak PPG occurring within 4 h after breakfast (P < 0.001), and mean postbreakfast excursion PPG0–4 h was reduced by 17 mg/dL with rHuPH20 pre-Rx (P < 0.001) as compared with usual CSII. In contrast, mean breakfast PPG excursion by SMBG showed minimal (<4 mg/dL) differences between treatments (P = 0.4).
Conclusion:
In type 1 diabetes, CGM was far more sensitive than SMBG in detecting reduced PPG excursions after pre-Rx of CSII infusion sites with rHuPH20 in the ambulatory care setting.
Health Care Professionals’ Adherence and Satisfaction Using the GlucoTab, a Computerized Workflow and Decision Support System for In-Hospital Glucose Management
Katharina Maria Neubauer, BSc, MSc; Lukas Schaupp, PhD; Julia Katharina Mader, MD; Stephan Spat, MSc; Bernhard Höll, MSc; Felix Aberer, MD; Peter Beck, PhD; Thomas Augustin, PhD; Johannes Plank, MD; Thomas Rudolf Pieber, MD
Medical University of Graz
Graz, Austria
Objective:
The development and evaluation of point-of-care computerized decision support systems are recommended to guide health care professionals (HCPs) in implementing evidence-based guidelines. The objective of this study was to investigate user adherence and satisfaction of the GlucoTab system.
Method:
The GlucoTab is a computerized workflow and decision support system for nurses and physicians with an integrated basal -bolus insulin algorithm for hospitalized noncritically ill patients with type 2 diabetes mellitus (T2DM). After testing the GlucoTab in 99 patients with T2DM at four different wards (cardiology, endocrinology, nephrology, plastic surgery), user adherence and satisfaction were assessed. Sixty-five HCPs (54 female, 85% nurses, 15% physicians, mean age 36 ± 11 years) participated in an anonymous questionnaire.
Result:
HCPs accepted 96.7 and 96.5% of the GlucoTab suggested basal and bolus insulin dosages. More basal insulin corrections by the users occur on treatment day 1. HCPs did not perform 4.2% of scheduled blood glucose measurements, 5.3% of scheduled bolus insulin injections, and 1.4% of basal insulin injections. Questionnaire data show that 53 HCPs (82%) believed that the GlucoTab is useful in daily routine; 54 HCPs (83%) stated that using the GlucoTab could prevent medication errors; 51 HCPs (78%) believed that they can handle the GlucoTab correctly; 49 HCPs (75%) answered that by using the GlucoTab, physicians have to be consulted less often; and 53 HCPs (82%) stated that by applying the GlucoTab, glucose control is more efficient.
Conclusion:
HCP adherence with the GlucoTab suggestions was very high, and the GlucoTab was used constantly in clinical practice. HCP satisfaction was high, which supports the implementation of the GlucoTab in other hospitals for evidence-based glucose management.
Funding:
The study was supported by the European Commission, Project REACTION (FP7 248590).
Evaluation of a Fuzzy Inference Model for Continuous Regimen Alterations in Type 2 Diabetes
Nonso Nnamoko, MPhil; Farath Arshad, PhD; David England, PhD; Jiten Vora, MD
School of Computing and Mathematical Science
Liverpool John Moores University
Liverpool, Merseyside, United Kingdom
Objective:
Conventional diabetes care delivery adopts a long-term feedback system, involving structured educational programs and care plan adjustments during routine checks (usually 3–6-month intervals). Here, goodness of fit evaluation was carried out on a model designed to provide short-term (daily) feedback with recommendations to compensate for any regimen disturbance.
Methods:
The prediction model is a multiple input–multiple output fuzzy inference system calibrated through expert knowledge and diabetes management algorithms; with seven inputs (age, sex, weight, height, average daily glucose, average heart rate, and calorie intake) and three output assessment predictions (glycated hemoglobin [A1C], exercise, and diet level). One day observation data sample was collected for 35 patients from Abbott Diabetes Care, which includes continuous blood glucose measurement, body weight, height, average daily heart rate, daily calorie intake, and most recent A1C result along with blind assessments of individual exercise and diet performance. The data sample consists of 13 male and 22 female patients aged between 18 and 65 years. Goodness of fit evaluation was conducted using mean square error, normalized mean square error (NMSE), and normalized root mean square error (NRMSE) between observed/advised and predicted data sets. A total of 35 × 3 output data pairs were compared.
Results:
Overall mean square error was 0.0899, showing good fit with little variation. NRMSE and NMSE evaluation allowed for individual results for each of the output assessment variables: A1C, exercise, and diet levels. NRMSE (0.7387, 0.7881, and 0.3716) and NMSE (0.9317, 0.9551, and 0.6051) results shows good fit for A1C and exercise level prediction but poor fit for diet level prediction.
Conclusions:
The model shows great potential in providing short-term interpretation of daily monitoring data, ultimately leading to continuous regimen alteration to improve diabetes management. As expected, prediction error was slightly high in some areas, but this could be due to a number of reasons, e.g., sample data veracity and assessment criteria used by Abbott. Testing in a real user environment is underway to generate more data sets that will guide future evaluations and relevant modifications to the model.
Reducing Risk with Continuous Glucose Monitoring in Intensive Care Unit Patients
John Norrie, MSc; Nicholas Barwell, PhD; Barry Crane
SumStats Ltd.
Edinburgh, Midlothian, United Kingdom
Objective:
We assess benefits in reducing risk of missed clinically adverse episodes (hyper- and/or hypoglycemia or maintaining tight glycemic control) using a new continuous glucose monitor (GlySure-CGM) in comparison with a reference intermittent monitor (Yellow Springs, YSI).
Method:
We measured blood glucose in 34 post–cardiac surgery intensive care unit (ICU) patients (India 2003) both continuously (sampling frequency 15 s) using the novel GlySure-CGM (via a central venous catheter) and intermittently (YSI monitor every 2–4 h) during 48 h postsurgery. We first calculated a global summary of risk difference of missed/delayed glucose excursions (relative difference in area under the curve for the two methods), projecting forward intermittent values using a step function. We then considered the performance of the two methods for three types of glucose behaviors : 1) hypoglycemic episodes, 2) hyperglycemic episodes, and 3) excursions outside tight glycemic control, using different approaches to interpolating the intermittent measures and varying sampling frequency. As well as quantifying the overall risk difference in terms of areas between the curves, we assessed time to detecting these clinically significant excursions, characterizing the difference between the two methods as delay in initiating appropriate treatment and time spent incorrectly treating.
Result:
The presentation will quantify advantages of continuous over intermittent glucose monitoring as the risk of missing or delaying identifying important glucose excursions in ICU patients in several real clinical situations (hypo- and hyperglycemic episodes and maintaining tight glycemic control). In addition, we prepare a new limit for continuous monitoring that maintains the benefits of continuous monitoring over intermittent monitoring.
Conclusion:
Continuous glucose monitoring offers better information to reduce risk by informing optimal clinical decision making on treatment strategies for controlling blood glucose in the ICU.
Efficacy of Continuous Glucose Monitoring in Children with Type 1 Diabetes
Irina Osokina, MD, PhD; Gregory Strelnikov
Endocrinology Department
State Research Institute for Medical Studies of the North
Russian Academy of Medical Science
Krasnoyarsk, Krasnoyarsk, Russia
Objective:
The introduction of real-time continuous glucose monitoring (CGM) may have the potential to increase the proportion of patients who are able to maintain target HbA1c values and limit the risk of severe hypoglycemia. The aim of this study was to evaluate the efficacy of pump insulin therapy with CGM in children.
Method:
The study included 30 children aged 6–14 years with type 1 diabetes treated by insulin pumps with the possibility of real-time glucose monitoring (Paradigm Real-Time, Medtronic) and 15 patients aged 6–15 years who received multiple insulin injection by pens. We estimated HbA1c level, the frequency of acute diabetic complications (ketoacidosis, severe hypoglycemic events), and satisfaction of the treatment during the year. All patients and parents were trained insulin therapy and glucose monitoring.
Result:
CGM was used on a continuous basis by only 6 patients (20%), by 18 (60%) two times per month, and by 6 (20%) one time per month. In addition, all patients had glucose monitoring by glucometers from two up to eight times per day. The level of HbA1c was in the first group 7.6 ± 0.8% and was in the control group 8.7 ± 1.4% (P < 0.05). At the end of the study, the rate of severe hypoglycemic episodes and hypoglycemia with blood glucose levels ≤70 mg/dL were significantly lower in the CGM group than in the control group. Mean HbA1c at the end of the study was significantly better in the CGM group. In the insulin pump group, there have been no cases of ketoacidosis; in the control group, there has been one case of ketoacidosis. The treatment satisfaction on a 10-point scale was 7 ± 2 in the pump group and 6 ± 2 in the control group. The patients with pumps called for larger size of transmitter, the divergence of glycemia levels by sensor and glucometer, and additional transmitter fixing. In the control group, the problems were multiple injections and fear of hypoglycemia .
Conclusion:
The study demonstrated that the use of CGM in pediatric patients with type 1 diabetes helped them to keep a good HbA1c level or even improve it with increased time within target glucose levels (70–180 mg/dL) and decreased time with hypoglycemic levels (<70 mg/dL).
Using Radar Plots to Evaluate Five Blood Glucose Monitoring Systems for Accuracy and Precision
Scott Pardo, PhD, PStat; Brian Pflug, PharmD; Nancy Dunne, RN, BSN, CDE; David A. Simmons, MD
Bayer HealthCare LLC, Diabetes Care
Whippany, New Jersey
Objective:
In a previous study, fingertip capillary blood samples from 106 subjects were evaluated using five different blood glucose monitoring systems (BGMSs; CONTOUR PLUS [CP], Accu-Chek Active [ACA], Accu-Chek Performa [ACP], FreeStyle Freedom [FF], and OneTouch SelectSimple [OTSS]) and a YSI glucose analyzer as the reference. In this analysis, radar plots provide a visual representation of the accuracy and precision of the blood glucose measurements obtained in the previous study.
Method:
The radar plots are composed of concentric circles radiating out from the center, similar to a target. Each circle represents a contour of constant error; errors are in mg/dL or percentage for YSI values <100 or ≥100 mg/dL, respectively. A red circle represents ±15 mg/dL or ±15% error, and points within the circle satisfy ISO 15197:2013 accuracy limits.
Result:
For YSI values <100 mg/dL (n = 123), the mean differences of BGMS results from the reference results were as follows: CP, 0.3736 mg/dL; ACA, −0.9142 mg/dL; ACP, 0.8622 mg/dL; FF, −11.1272 mg/dL; and OTSS, −7.1467 mg/dL. For YSI values ≥100 mg/dL (n = 191), the mean percentage differences between BGMS and reference results were as follows: CP, −1.3515%; ACA, −1.4441%; ACP, −0.2587%; FF, −13.7614%; and OTSS, −6.6422%. Furthermore, 99.0% of CP, 96.5% of ACA, 97.5% of ACP, 74.2% of FF, and 86.3% of OTSS results were within ISO 15197:2013 accuracy limits. The data for each BGMS will be presented in radar plots at the meeting.
Conclusion:
Radar plots encompass a variety of information, including accuracy, precision, and whether results satisfy International Organization for Standardization accuracy criteria in a single, accessible presentation and thus may be a useful format for visually assessing multiple BGMSs.
Evaluating the Accuracy and Precision of Six Blood Glucose Monitoring Systems Using Radar Plots
Scott Pardo, PhD, PStat; Brian Pflug, PharmD; Nancy Dunne, RN, BSN, CDE; David A. Simmons, MD
Bayer HealthCare LLC, Diabetes Care
Whippany, New Jersey
Objective:
Previously, fingertip capillary blood samples from 146 subjects were evaluated using six different blood glucose monitoring systems (BGMSs; CONTOUR NEXT [CN], Accu-Chek Aviva Nano [ACAN], FreeStyle Lite [FL], OneTouch Ultra2 [OTU2], OneTouch Verio Pro [OTVP], and TRUEtrack [TT]) and a reference YSI glucose analyzer. In this analysis, radar plots are used to visually demonstrate the accuracy and precision of the six BGMSs using blood glucose measurements from the previous study.
Method:
The radar plots resemble a target with concentric circles emanating out from the center, with each circle representing a contour of constant error. Errors are in mg/dL or percentage for YSI values <100 or ≥100 mg/dL, respectively. Points within a red circle representing ±15 mg/dL or ±15% error satisfy ISO 15197:2013 accuracy limits.
Result:
For YSI values <100 mg/dL (n ≥ 172), the mean differences of BGMS results from the reference results were as follows: CN, 1.0145 mg/dL; ACAN, 2.4972 mg/dL; FL, −2.2572 mg/dL; OTU2, −9.5694 mg/dL; OTVP, 3.8828 mg/dL; and TT, −8.9971 mg/dL. For YSI values ≥100 mg/dL (n ≥ 364), the mean percentage differences between BGMS and reference results were as follows: CN, 1.7958%; ACAN, 0.5105%; FL, −11.3312%; OTU2, −6.4371%; OTVP, −1.7167%; and TT, −6.4362%. Moreover, 99.6% of CN and ACAN, 84.0% of FL, 91.6% of OTU2, 99.8% of OTVP, and 83.6% of TT results were within ISO 15197:2013 accuracy limits. The data for each BGMS will be presented in radar plot format at the meeting.
Conclusion:
Radar plots may provide a useful format for visually assessing multiple BGMSs, as they allow for the inclusion of a variety of information, including accuracy, precision, and whether the results meet International Organization for Standardization accuracy criteria, all in a single, user-friendly presentation.
CardioChek Plus: A Wireless Point-of-Care Analyzer Enabling Rapid, Simultaneous Blood Glucose and Lipid Profile Analysis
Aniruddha Patwardhan, PhD; Christopher A. Dailey, PhD; Gary Hughes, BS; Jim Miller, AS; Kathy Rogers, BS; Jonathan Broadwell, BS; Ryan Jesswein, MS; Valerie Bader, BA; Christine Casterline, BS; William Benedict, BS, MBA; Keith A. Moskowitz, PhD
Polymer Technology Systems Inc.
Indianapolis, Indiana
Objective:
Persons with diabetes are twice as likely to suffer ischemic events as persons without. Point -of-care results for glucose and lipid profiles significantly improve outcomes, but obtaining both often requires multiple analyzers and finger sticks. Here we report the development and performance evaluation of the CardioChek Plus (CC Plus), a wireless, handheld analyzer providing glucose measurements and complete lipid profiles in <2 min from a single finger stick (~40 µL capillary blood required). Simultaneous measurements are accomplished by using both electrochemical and reflectance technologies.
Method:
Each assay was evaluated across the dynamic range (glucose 20–600 mg/dL; total cholesterol 100–400 mg/dL; HDL 15–100 mg/dL; and triglycerides 50–500 mg/dL). Method correlations compared the CC Plus analyzer (whole blood) and the Roche COBAS Integra Plus analyzer with matched plasma. Within-run precision was evaluated using whole blood at three concentrations (n = 50 at each level). Linearity was evaluated according to current Clinical and Laboratory Standards Institute standards.
Result:
Method comparisons produced slopes of 0.93–1.02 and correlation coefficients (R2) exceeding 0.96 for all analytes as compared with the Roche COBAS Integra Plus analyzer. Precision (n = 50) at low, middle, and high concentrations ranged from 3.9–6.3% coefficient of variation. Each test strip demonstrated linearity across the respective measurement range.
Conclusion:
The CC Plus analyzer demonstrated accurate, precise, and linear total cholesterol, HDL cholesterol, triglycerides, glucose, and calculated LDL values from blood, statistically equal to a central laboratory plasma method in <2 min while enabling wireless data transfer aiding in the monitoring of patients with diabetes at risk of ischemia.
Comparative Accuracy of Insulin Dosing Based on Results from Two Clinical Trials of Blood Glucose Monitoring Systems
Brian Pflug, PharmD; Scott Pardo, PhD, PStat; David A. Simmons, MD
Bayer HealthCare LLC, Diabetes Care
Whippany, New Jersey
Objective:
In two previous clinical trials (N = 146 subjects per trial), fingertip capillary blood samples were evaluated using a panel of blood glucose monitoring systems (BGMSs) and a reference YSI glucose analyzer. In trial 1, samples were tested with CONTOUR NEXT EZ (EZ), Accu-Chek Aviva (ACA), FreeStyle Freedom Lite (FFL), OneTouch Ultra2 (OTU2), and TRUEtrack (TT). In trial 2, samples were tested with CONTOUR NEXT (CN), Accu-Chek Aviva Nano (ACAN), FreeStyle Lite (FL), OTU2, OneTouch Verio Pro (OTVP), and TT. In this analysis, insulin dosing was calculated using these glucose measurements; dosing error was compared across each panel of BGMSs.
Method:
For each blood glucose measurement, premeal bolus insulin dosing was determined using a computer model, assuming a 60-g carbohydrate meal and 100 mg/dL target glucose level. The model adjusted inputs of 1/25 insulin sensitivity and 1:15 insulin:carbohydrate ratio to account for BGMS measurement error. Dosing error was the difference between doses calculated using the meter and YSI results.
Result:
In trial 1, 95% dose error ranges (in units of insulin) were as follows: EZ, −0.7 to 0.2; ACA, −0.3 to 1.2; FFL, −3.0 to −0.3; OTU2, −5.4 to −0.1; and TT, −3.2 to 0.6 (negative error signifies underdosing, and positive error signifies overdosing). In trial 2, 95% dose error ranges were as follows: CN, −0.4 to 1.0; ACAN, −0.7 to 1.2; FL, −4.1 to 0.0; OTU2, −3.5 to 0.1; OTVP, −1.3 to 1.1; and TT, −4.4 to 0.5. Within each trial, EZ and CN had statistically significantly less dosing error than all other meters (P<0.0001).
Conclusion:
The model predicted that 95% of insulin dose errors with EZ and CN, compared with all other meters in their respective panels, would be within a relatively narrow range. Differences in meter accuracy could result in clinically important differences in insulin dosing.
Comparative Accuracy of Insulin Dosing Based on Results from Five Blood Glucose Monitoring Systems
Brian Pflug, PharmD; Scott Pardo, PhD, PStat; David Simmons, MD
Bayer HealthCare LLC, Diabetes Care
Whippany, New Jersey
Objective:
For insulin-treated people with diabetes, blood glucose monitoring system (BGMS) accuracy is important because test results may be used to adjust insulin dosing. In a previous clinical trial, fingertip capillary blood samples from 106 subjects were evaluated using five BGMSs (CONTOUR PLUS [CP], Accu-Chek Active [ACA], Accu-Chek Performa [ACP], FreeStyle Freedom [FF], and OneTouch SelectSimple [OTSS]) and a reference YSI glucose analyzer. In this analysis, those glucose measurements were used to calculate insulin dosing; dosing error was compared across meters.
Method:
An appropriate premeal bolus insulin dose was determined for each blood glucose measurement (assuming a 60-g carbohydrate meal and 100 mg/dL target glucose level). Dose was calculated using an insulin dosing application with inputs of 1/25 insulin sensitivity and 1:15 insulin:carbohydrate ratio; dose was independently calculated using a computer model, which adjusted these inputs to account for BGMS measurement error. Using each method, insulin dosing error was the difference between the doses determined using the meter and YSI results.
Result:
The 95% dose error ranges (in units of insulin) using the application were as follows: CP, −0.6 to 0.3; ACA, −1.1 to 0.4; ACP, −1.4 to 0.4; FF,−2.5 to −0.3; and OTSS, −1.8 to 1.2 (negative dose error denotes underdosing, and positive dose error denotes overdosing). Using the computer model, 95% dose error ranges were as follows: CP, −1.2 to 0.4; ACA, −1.5 to 0.7; ACP, −1.9 to 0.4; FF, −5.5 to −0.4; and OTSS, −3.5 to 1.8. For both methods, CP dose error was statistically significantly less than FF (P < 0.0001) and OTSS (P = 0.0199) dose error.
Conclusion:
Both methods predicted that 95% of insulin dose errors with CP will be within a relatively narrow range compared with dose errors with the other meters. Thus differences in meter accuracy could result in clinically important differences in insulin dosing.
Preliminary Results of a Crossover Study on Usage Time for Insulin Pump Infusion Systems
Andreas Pfützner, MD, PhD; Marco Grenningloh, PhD; Lene Walther-Johannesen, RN; Rabi Gharabli, BSc; David Klonoff, MD
Pfützner Science and Health Institute
Diabetes Center and Practice
Mainz, Germany
Background:
It is generally recommended that infusion sets for use with insulin pumps should be used for 2 to 3 days. Potential reasons for limited usage time are local skin reactions to the constantly infused insulin formulation with its preservatives (e.g., metacresol) or limited biocompatibility of the catheter material. However, many patients use the catheters longer for economical reasons, risking adverse events and skin reactions. A study investigating the efficacy and tolerability of different usage time for infusion sets has never been performed. To fill this gap, we investigated the tolerability of regular catheter use (2 days) in comparison with extended use (4 days) in a real-world setting.
Methods:
This was a prospective randomized controlled crossover study with 2 × 3 months observation periods with 24 type 1 diabetes patients. At baseline, patients were trained on the use of the infusion system, which was compatible with their pump type (Medtronic/Mio or inset II). Thereafter, they were randomized to any of the two treatment sequences. Observation parameters included HbA1c, frequency and nature of device-related and procedure-related adverse events, and patient preference.
Results:
Here we report on an interim analysis that was performed with 12 patients (4 men, 8 women; age 47 ± 11 years; BMI 27.4 ± 3.2 kg/m²). In this group, the overall number of treatment-related adverse events was 189 with 2-day use versus 201 with 4-day use (not significant). However, the number of catheter-related events was 42 with 2-day use versus 130 with 4-day use (P < 0.001). The combination of catheter- and treatment-related events was significant, voting for 2-day use (231 vs. 331; P < 0.001). Several patients starting with 2-day use reported a major increase in infusion-site problems when extending the usage time to 4 days. However, few individual patients did not experience any difference between the treatment groups. Glycemic variability was less favorable with extended use (e.g., hypoglycemic events, 238 vs. 341 events; P < 0.001).
Conclusions:
This interim analysis demonstrates that using the infusion sets for a longer-than-recommended usage period of 2–3 days resulted in a clinically relevant increase in treatment-related tolerability problems and impaired glycemic control already in a small number of patients. Patients should be encouraged not to use insulin pump infusion sets for a longer than the recommended time period.
Use of InsuPad in Daily Practice in Patients with Type 1 Diabetes: Post Hoc Analysis from the BARMER Study
Andreas Pfützner, MD, PhD; Thomas Behnke, MD; Klaus Funke, MD; Norbert Hermanns, PhD; Gabriel Bitton, PhD; Ron Nagar; Thomas Haak, MD
Pfützner Science and Health Institute
Mainz, Germany
Background and Aims:
The InsuPad device improves insulin absorption by standardized modulation of the injection-site temperature after insulin administration. This leads to better postprandial glycemic control and/or lower insulin requirements. The primary objective of this post hoc analysis of the prospective controlled BARMER study was to investigate the impact of InsuPad use on prandial rapid-acting insulin dose and glycemic control in the participating patients with type 1 diabetes.
Material and Methods:
This post hoc analysis was performed with data from 13 patients with type 1 diabetes (2 female, 11 male; age [mean ± SD] 49.9 ± 12.5 years; HbA1c 7.2 ± 0.5%, body weight 91.5 ± 12.7 kg). All patients were treated with multiple daily injections with insulin glargine and a short-acting insulin analog (aspart, glulisine, lispro). After a run-in treatment optimization and basal insulin stabilization period of up to 4 weeks, the patients were randomized to continue therapy for 3 months without (control; n = 6) or with (n = 7) InsuPad. Only three visits at the site (screening, baseline, end point) were performed to ensure real -world conditions. Observation parameters included HbA1c, insulin dose, frequency of hypoglycemia, and body weight.
Results:
During the run-in period, HbA1c decreased in the type 1 group from 7.2 ± 0.5 to 6.8 ± 0.5% (P < 0.05), and further improved in both arms until study end (control group 6.5 ± 0.4%; InsuPad 6.5 ± 0.8%; both not significant versus baseline; not significant between the groups). To achieve this glycemic control, patients in the control group needed an increase in the daily prandial insulin dose from baseline by 9.3% (from 46 ± 24 to 50 ± 30 units; not significant) with stable basal insulin requirements (39 ± 10 vs. 40 ± 11 units; not significant). Patients in the InsuPad group required significantly less prandial insulin to reach these HbA1c results (34 ± 7 to 28 ± 7 units; −17%; P < 0.05) and a slight decrease in the basal insulin dose (from 32 ± 8 to 30 ± 6 units; not significant). In consequence, total daily insulin dose increased in the control group (+6.1%) and decreased with InsuPad (−11.6%; P < 0.01 between the groups). The number of hypoglycemic events (blood glucose readings <63 mg/dL) during the observation period was higher in the control group (22.7 ± 5.0/patient) than in the InsuPad group (5.8 ± 2.3/patient; −74%; P = 0.06).
Conclusion:
When treating type 1 diabetes patients to target with intensified insulin therapy, use of the InsuPad device for 3 months resulted in a substantially lower frequency of hypoglycemic events and a significant reduction in insulin analog requirements as compared with a control group not using the device under real- world conditions. These post hoc pilot results need to be confirmed in an appropriately designed study with a larger patient cohort. InsuPad may be useful to achieve treatment targets with a safer and more efficient basal-bolus therapy in insulin-treated patients with type 1 diabetes.
Glycemic Variability Is Associated with Frequency of Blood Glucose Testing and Bolus Numbers: Post Hoc Analysis Results from the Real-World ProAct Study
Andreas Pfützner, MD, PhD; Jörg Weissmann, MD; Stavroula Mougiakakou, PhD; Elena Daskalaki, PhD; Norbert Weis, PhD; Ralph Ziegler, MD
Pfützner Science and Health Institute
Mainz, Germany
Background:
The ProAct study has shown that transition from older pump systems to the Accu-Chek Combo system in a large patient population resulted in stable glycemic control with significant improvements in HbA1c, especially in patients with unsatisfactory baseline HbA1c and shorter pump use.
Methods:
In this post hoc analysis of the ProAct database, we investigated the glycemic control and glycemic variability at baseline (i.e., independent from the Accu-Chek Combo use) by determination of several established parameters and proposed glycemic variability scores (HbA1c, hypoglycemia frequency, J-score, hypo- and hyperglycemia index, index of glycemic control) in patients with different daily bolus numbers and different blood glucose measurement frequencies (<3/day, 3–5/day, and >5/day, in both cases). The data were derived from 299 patients (172 female, 127 male; age [mean ± SD] 39.4 ± 15.2 years; continuous subcutaneous insulin infusion duration 7.0 ± 5.2 years) enrolled by 61 European sites.
Results:
Patients with frequent daily blood glucose readings (>5/day) were better controlled than patients with few blood glucose readings (<3/day; HbA1c 7.2 ± 1.1 vs. 8.0 ± 0.9%; mean daily blood glucose 151 ± 22 vs. 176 ± 30 mg/dL; percentage of readings/month >300 mg/dL 10 ± 4 vs. 14 ± 5; percentage of readings in euglycemia [80– 180 mg/dL] 59 vs. 48%, P < 0.05 in all cases; percentage of readings/month <70 mg/dL 4 ± 2 vs. 4 ± 3, not significant) and had a lower glycemic variability (J-score 49 ± 13 vs. 71 ± 25; hyperglycemia index 0.9 ± 0.5 vs. 1.9 ± 1.2; index of glycemic control 1.9 ± 0.8 vs. 3.1 ± 1.6, P < 0.05 in all cases; hypoglycemia index 0.9 ± 0.8 vs. 1.2 ± 1.3, not significant). Frequent testing was associated with a higher number of bolus applications (6.1 ± 2.2 vs. 4.5 ± 2.0 boli/day; P < 0.05). Therefore a similar but less pronounced effect on glycemic variability in favor of more daily bolus applications was observed (>5 vs. <3 boli/day; J-score 57 ± 17 vs. 63 ± 25, not significant; hypoglycemia index 1.0 ± 1.0 vs. 1.5 ± 1.4, P < 0.05; hyperglycemia index 1.3 ± 0.6 vs. 1.6 ± 1.1, not significant; index of glycemic control 2.3 ± 1.1 vs. 3.1 ± 1.7, P < 0.05).
Conclusions:
This post hoc analysis indicates that CSII patients who perform frequent daily blood glucose readings have a better glycemic-control-associated lower glycemic variability as assessed by a variety of proposed glycemic variability indices.
Long-Term Use of InsuPad Is Associated with High Treatment Adherence and Results in Maintenance of Excellent Glycemic Control with Further Reduction of Prandial Insulin Requirements
Andreas Pfützner, MD, PhD; Klaus Funke, MD; Thomas Behnke, MD; Gabriel Bitton, PhD; Ron Nagar
Pfützner Science and Health Institute
Mainz, Germany
Background:
The InsuPad device enhances insulin absorption by standardized injection-site modulation and was shown to reduce the frequency of hypoglycemia by 46% and prandial insulin requirements by approximately 30% in a controlled trial with 3- month duration (BARMER study). This follow-up investigation aimed to explore the effect of using InsuPad over a period of more than 12 months.
Material and Methods:
After the BARMER study, patients were supplied with the device and disposables and were allowed to continue with device use for at least 1 year. Patients in the previous control group were also allowed to use the device. This long-term-use follow-up investigation was performed in 52 patients (22 female, 30 male; age [mean ± SD] 65 ± 8 years; HbA1c 7.1 ± 0.7% at start of the BARMER study) who participated at one study site and who could be recontacted after a minimum of 13 months. A standardized questionnaire was completed with each patient.
Results:
The mean usage time was 17.8 ± 2.5 months (range 13–21 months). During the whole time, only two patients (3.8%) had stopped using the device because of persistent skin reaction to the adhesive. In the remaining patients, body weight had remained stable (3 months 100 ± 23 kg vs. 18 months 100 ± 18 kg; not significant), HbA1c was stable (baseline 7.2 ± 0.7%; after 3 months 6.4 ± 0.7% vs. 18 months 6.3 ± 0.6%; not significant), and total daily insulin dose was even further reduced as compared with BARMER study start (change versus baseline, 3 months −16.5% [result with all patients on InsuPad] vs. 18 months −25.3%; P < 0.001)
Conclusions:
These results indicate that the excellent glycemic control achieved in the BARMER study was maintained when all patients used the device over a period of 18 months or longer with further reduction of prandial insulin dose requirements and with a high treatment adherence.
Clinical Development of XeriSol Glucagon: A Stable, Nonaqueous Liquid Glucagon Formulation
Steven J. Prestrelski, PhD, MBA
Xeris Pharmaceuticals Inc.
Austin, Texas
Objective:
The objective of our program is the clinical development and ultimately commercialization of a novel, nonaqueous, stabilized liquid glucagon formulation for several indications in treating hypoglycemia.
Method:
Xeris has completed a phase 2 safety-efficacy study of XeriSol glucagon, as compared with the Lilly Glucagon Emergency Kit, in healthy volunteers. There are currently three additional clinical studies ongoing that use Xeris’s nonaqueous glucagon formulation for various hypoglycemia-related indications. In these studies, glucagon doses are administered using various syringes or pump devices, though all studies use the same glucagon formulation, and in each case, the drug is administered into the subcutaneous tissue.
Results:
XeriSol glucagon (G-Pen [glucagon injection]) 1.0 mg demonstrated therapeutic equivalence to Lilly glucagon (glucagon for injection [recombinant DNA origin]) 1.0 mg in terms of area under the curve, Cmax, and tmax, at significantly lower levels of glucagon exposure (glucagon area under the curve). Ongoing clinical studies are progressing, and topline data will be reported at the conference.
Conclusion:
These phase 2 clinical study results support further clinical development of G-Pen (glucagon injection) as a premixed, autoinjectable, room-temperature-stable treatment for severe hypoglycemia and XeriSol glucagon in general for additional hypoglycemia indications.
Illustration of Potential Cost-Saving Implications of Lower-Extremity Nerve Decompression for Diabetic Foot Ulceration
Timothy M. Rankin, MD; John D. Miller, BS; Angelika Gruessner, PhD; D. Scott Nickerson, MD
Department of Surgery
University of Arizona
Tuscon, Arizona
Background:
The U.S. diabetic foot ulcer (DFU) incidence is 3 to 4% of 22.3 million diagnosed diabetes cases plus 6.3 million undiagnosed cases, 858,000 cases total. Risk of recurrence after healing is 30% annually. Lower-extremity multiple nerve decompression (ND) surgery reduces neuropathic DFU (nDFU) recurrence risk by >80%. Cost-effectiveness of hypothetical ND implementation to minimize nDFU recurrence is compared to the current annual $6.171 billion DFU expense.
Methods:
A literature review identified best estimates of annual incidence, recurrence risk, medical management expense, and noneconomic costs for DFU. Illustrative cost/benefit calculations were performed assuming widespread application of bilateral ND after wound healing to the nDFU problem, using Center for Medicare Services mean expense data of $1,143/case for unilateral lower-extremity ND. Calculations use conservative, evidence-based cost figures that are contemporary (2012) or adjusted for inflation.
Results:
Widespread adoption of ND after nDFU healing could reduce annual DFU occurrences at least 21% by the third program year, representing annual cost avoidance savings of $1.296 billion. This scenario proffers significant expense reduction and societal benefit and represents a minimum 1.9× return on the investment cost for surgical treatment. Further large cost savings would require reductions in initial DFU incidence, which ND may also achieve in advanced diabetic sensorimotor polyneuropathy.
Conclusions:
By minimizing the contribution of recurrences to yearly incidence rate of DFUs, ND has potential to significantly reduce the annual cost of treatment for nDFU in the U.S.
Improved Glycemic Management with an Implantable Continuous Glucose Monitoring System in a 90-Day Feasibility Study
Ravi Rastogi, PhD; Xiaolin Wang, PhD; Carrie Lorenz, PhD
Senseonics Inc.
Germantown, Maryland
Objective:
The Senseonics continuous glucose monitoring (CGM) system is composed of a fluorescence-based long-term implantable sensor and a smart wearable transmitter, which wirelessly powers the sensor and transmits data to a mobile medical application (MMA). The MMA displays the real-time glucose levels and rate of change and has features like logging calorie consumption and insulin dosage for improved glycemic management. For assessment of accuracy and clinical benefits of using a CGM system, a 90-day pilot study was performed.
Method:
A total of 30 sensors were implanted on the upper arms of 21 subjects with type 1 or type 2 diabetes. During the study, the CGM system measured glucose every 2 min. In-clinic sessions were performed every 14 days, where laboratory reference measurements were collected every 15 min over a duration of 8 h. Throughout the study, subjects followed the therapeutic regimen as prescribed by their doctor. Self-monitoring of blood glucose measurements were taken as per the protocol and entered in the MMA when requested for calibration (two per day). HbA1c levels of the patients were also monitored before and after the study.
Result:
The mean absolute relative difference was found to be 11.0%, and the mean absolute difference was 10.3 mg/dL. According to the Clarke error grid analysis, 99% of the data were in clinically accurate zone A (85.8%) and benign zone B (13.2%). We also found that the HbA1c levels of the patients on average decreased by 0.8% (SD = 0.7%) after the study.
Conclusion:
The results of this study showed that the Senseonics CGM system got a mean absolute relative difference of 11.0% and a reduction in HbA1c levels of 0.8%, showing that subjects were able to manage their therapeutic regimen better using CGM.
iSense Mastery Utility Demonstrated in 7-Day Clinical Study
Mihailo V. Rebec, PhD; Richard G. Sass, MBA
iSense CGM Inc.
Wilsonville, Oregon
Introduction:
iSense CGM is developing a continuous glucose monitoring (CGM) system that is called the Mastery CGM. The product development has advanced to the point that several clinical studies were completed. A 7-day clinical study was completed recently, and the results of that study will be presented in this poster.
Methods:
Sixteen subjects were enrolled in a clinical study. The CGM sensors were inserted by each subject on the right and left side of the abdomen. Each CGM system’s initial calibration was done 1 h postinsertion on day 1. Following calibration, Yellow Springs Instrument (YSI) venous plasma glucose assays were performed at 15-min intervals. Subjects returned home with instructions to calibrate their CGM system every 12 h and to keep a daily diary of living activities. Sensors were removed at the end of the seventh day. The accuracy of the sensors was determined with a comparison of paired senor glucose and YSI glucose values.
Results:
Fifteen subjects completed the 7-day study. A total of 22/34 sensors functioned for the entire 7 days. There was a total of 1,377 YSI reference measurements with valid CGM values associated with them. The overall (mean) absolute relative difference and (median) absolute relative difference for these was 18 and 15%, respectively. When optimal lag corrections were applied, the mean absolute relative difference was reduced to 14.8%.
Discussion:
The Mastery CGM has demonstrated performance that indicates its potential as an adjunct device for the treatment and control of diabetes.
Integrating Exercise Compensation into a Bihormonal Artificial Pancreas System
Seyed Navid Resalat, MSEE; Joseph El Youssef, MD, MCR; Deborah Branigan, BA; Nicholas Preiser, BSEE; John Condon, MSBE; Jessica Castle, MD; Peter G. Jacobs, PhD, MSEE
Department of Biomedical Engineering
Oregon Health and Science University
Portland, Oregon
Objective:
Enabling an appropriate response to external disturbances such as exercise remains a challenge within artificial pancreas control systems. Both non–insulin-mediated glucose uptake and increased insulin sensitivity during exercise can lead to hypoglycemia in people with type 1 diabetes if insulin dosing is not adjusted in anticipation of or in response to an exercise event.
Method:
An adaptive personalized exercise (APE)-sensing algorithm to identify the onset and duration of exercise, grade the intensity of the exercise, and adjust bihormonal dosing of insulin and glucagon based on exercise estimation is presented. The APE algorithm uses the heart rate and three-axis accelerometers and gyroscopes mounted on the body to estimate exercise. A total of eight subjects without diabetes were asked to exercise on a treadmill and a stationary bicycle for 5 min each at 25, 50, and 75% VO2max. We converted these physical estimates of human movement to metabolic equivalent of task (MET) using an artificial neural network. Within an in silico model, we show how insulin and glucagon dosing are adjusted based on MET during exercise to prevent hypoglycemia.
Result:
We present results from the human exercise trial and show how our algorithm accurately estimates MET using body-worn accelerometer, gyroscope, and heart rate sensors. We also discuss which of these sensors is most relevant for grading exercise. Finally, we show results from in silico testing using the University of Virginia-Padova simulator to assess how our APE algorithm can be used to compensate for higher glucose uptake and higher insulin sensitivity during exercise to avoid hypoglycemia.
Conclusion:
Insulin and glucagon dosing can be adjusted during closed-loop control systems to compensate for increased glucose uptake and higher insulin sensitivity during exercise.
The I-KAN Study: Internet Initiation of Insulin for Type 2 Diabetes in Kansas
David C. Robbins, MD; Ann C. Walker, MS, RDN, LD; Virginia Lewis, MA; Debbie Griffin, RN, CDE
KU Diabetes Institute
Kansas City, Kansas
Objective:
We determine the safety and effectiveness of an Internet initiation of insulin intervention.
Methods:
Type 2 diabetes patients failing on oral agents were taught to inject and titrate basal insulin, using a simple treat-to-target algorithm in four synchronous group Internet sessions with 6-month follow-up. Frequency-severity of hypoglycemia and changes in A1C were outcome measures. Participants were eligible if their provider had recommended insulin treatment and they 1) were willing to start, 2) had started within the last 6 months, or 3) were ambivalent but willing to learn. Fifty-one participants consented, with 39 completing the classes. The average age was 58 ± 10 years; 52% were female.
Result:
The primary outcome (safety)—frequency of hypoglycemia (blood glucose <70 mg/dL or symptomatic) from baseline to 6 months—was 6.88 ± 11.20 (95% CI 2.50 to 11.27) per person/year. All but one were grade 1 (mild). One event required assistance (grade 2). Forty percent experienced one or more hypoglycemic events. Change in A1C from baseline to 6 months for those starting insulin was −1.33 ± 1.44% (95% CI −1.89 to 0.77%). Forty percent achieved an A1C <7%. Sample sizes were insufficient to achieve significance; however, both safety and effectiveness measures were within expected ranges extrapolated from large treat-to-target randomized controlled trials to a real-world environment.
Conclusion:
Safety and effectiveness were comparable to clinical studies where patients were engaged in person-to-person teaching. Internet classes were well received. Challenges included lack of Internet fluency in some, psychological resistance to insulin treatment by both providers and patients, and evolving technology. While insulin is among the most difficult drugs to teach/administer, Internet management appears to be safe and effective.
Graphical Display of Quality of Glycemic Control: New Methods
David Rodbard, MD
Biomedical Informatics Consultants
Potomac, Maryland
Background and Aims:
There is need for better graphical displays of self-monitoring of blood glucose and continuous glucose monitoring data to facilitate clinical interpretation.
Materials and Methods:
We propose several new methods. 1) One can plot percentage of hypoglycemia versus percentage of hyperglycemia. 2) To evaluate therapeutic interventions, one can plot change in percentage of hypoglycemia versus change in percentage of hyperglycemia. The four quadrants represent increases in both hyper- and hypoglycemia, decreased risk of hyperglycemia but increased hypoglycemia, decreased risk of both hypo- and hyperglycemia, and decreased hypoglycemia but increased hyperglycemia. 3) We propose use of a two-dimensional triangular plot where three axes show the percentage of hyperglycemic, percentage of hypoglycemic, and percentage in target range, simultaneously. We also propose displays of 4) SD versus mean glucose and 5) percentage of hypoglycemia versus mean glucose.
Results:
These new methods have been applied successfully to multiple data sets and studies: they facilitate simultaneous display, analysis, and hypothesis testing of risks of hypo- and hypoglycemia before and after interventions (therapies, sensors, devices, artificial pancreas, lifestyle, educational, psychosocial) for individual patients and for groups of subjects. They provide compact, readily learned and applied graphical displays of quality of glycemic control.
Conclusions:
Each of these methods is a useful addition to previously available methods. A plot of percentage of hyperglycemia versus percentage of hypoglycemia, the corresponding plot of changes in each of these two parameters, and of percentage of hypoglycemia versus mean glucose are the easiest to learn and interpret.
A Continuous Glucose Estimation System on a Smart Phone
Derrick Rollins, BS, MS, MS, PhD; Yong Mei, BS, MS; Queenster Nartey
Iowa State University
Ames, Iowa
Objective:
It is well-known that disturbances such as eating, stress, and activity can affect the level of blood sugar. However, how they specifically affect an individual can vary widely from person to person. As a result, this research has focused on the development of a continuous glucose estimation system for non-insulin-dependent people that gives personalized and continuous estimated blood glucose concentration (BGC) using a novel mathematical model that is developed from infrequent lancet BGC measurements.
Method:
This device is not used to predict future BGC and is thus not a continuous glucose monitoring system. It is a tool that gives the user personalized information, i.e., education, on how their BGC is impacted by the size of meals, type and duration of activities, and stress.
Results:
A prototype of this system will be shown that is basically an application on a smart phone. Lancet data are collected automatically, stress and activity data are uploaded automatically from an activity armband, and the user enters meal size via the time stamp button on the armband. Up to the most recent lancet measurement, the user can change a combination of meal sizes, activity and stress over some period and see an estimation of how their glucose would have changed continuously over the specified period.
Conclusion:
By providing convenient, informative, personalized information on how one’s behavior affects their BGC, users will be motivated to change their lifestyles in the ways that are most effective for them.
Evaluating Safety and Efficacy of a Closed-Loop Glucose Control System with Biased Sensors Induced by Forced Calibration Error
Anirban Roy, PhD; Benyamin Grosman, PhD; Neha Parikh, PhD; Di Wu, PhD; Natalie Kurtz, PhD; Jino Han, BS; Francine Kaufman, MD; Barry Keenan, PhD
Medtronic Diabetes
Northridge, California
Objective:
Closed-loop glucose control is directly affected by sensor accuracy. Sensor accuracy, or inaccuracy, in return, is partly dependent on the calibration with meter blood glucose (MBG) measurements. It is well-known that calibrating a sensor with an erroneous MBG value can introduce significant bias. In this work, the safety and efficacy of Medtronic’s overnight closed-loop system was evaluated by inducing the maximum permissible sensor-calibration error allowed by a calibration-check safeguard algorithm at the initiation of closed loop.
Method:
A total of 16 type 1 diabetes subjects were recruited for overnight closed-loop control at two different research centers (eight per center). Closed-loop was initiated right after 10:00 P. M., and data were collected until 8:00 A.M. the next-day. At the closed-loop initiation point, the sensor was either overcalibrated (in case it was overreading) or undercalibrated (in case it was underreading) to a maximum allowable error of 35% (between sensor and MBG).
Result:
Out of the 16 experiments, in 13 cases, the sensor was undercalibrated, which resulted in an average (SD) sensor bias of −17.3 (±6.6)%. In the remaining three cases, the sensor was overcalibrated, which resulted in an average (SD) sensor bias of +28.5 (±15.1)%. The average (SD) mean absolute relative difference of all the 16 sensors was 21.2 ± 8%. Mean (SD) YSI glucose value and the time in range (70–180 mg/dL) were recorded to be 137 ± 36 mg/dL and 88.9% for all the 16 subjects, respectively. There were no reported hypoglycemic events (<60 mg/dL).
Conclusion:
Medtronic’s overnight closed-loop system is safe and effective in controlling blood glucose levels even with sensor-calibration error up to 35% at the time of initiation.
A Method for Safe Real-Time Retuning of General Artificial Pancreas Systems
Daniel Ruiz; José Luis Díez, PhD; Jorge Bondia, PhD
Universitat Politècnica de València
Valencia, Spain
Objective:
The domiciliary use of an artificial pancreas will face new scenarios and disturbances. It cannot be expected that the initial tuning of the controller will remain acceptable along time. A real-time retuning module is thus a necessary component. A method for controller retuning with minimal risk to the patient is presented here. The method is independent of the controller implemented and thus can be integrated into any artificial pancreas system.
Method:
Sliding mode reference conditioning (SMRC) was used as an outer feedback loop limiting insulin on board (IOB) by acting on the controller’s reference. IOB constraints were dependent on the meal size and day/night periods. They were adapted according to the controller’s performance. A multiplicative factor for insulin infusion was also considered and adapted according to the degree of activation of SMRC. The method was tested in combination with a detuned proportional-integral -derivative (PID) controller in a cohort of 10 virtual patients from the University of Virginia/Padova simulator. Ten simulations per patient of 100 days of use were considered. Intrapatient variability and variations in patient eating behavior were considered.
Result:
With a conservative initial tuning of the IOB constraints for patient’s safety, SMRC without adaptation in combination with a detuned PID (PID-SMRC) managed to reduce hypoglycemia as compared with the detuned PID alone from 16.25 to 3.23% h/day. However, this was done at the expense of higher hyperglycemia (>180 mg/dL), reaching 26.86%. With the adaptation module, performance was significantly improved as compared with PID-SMRC with 0.48% h/day in hypoglycemia (0.08 hypoglycemic events per day), 15.74% in hyperglycemia, and 83.79% in euglycemia.
Conclusion:
The proposed method can efficiently manage variability challenges in long-term use of AP systems in in silico studies.
Effect of Diary Milk Powder as a Source of Vitamin D on Total Cholesterol Levels in White Male Rat Type 2 Diabetes Models
Nindy Sabrina
Brawijaya University
Malang, East Java, Indonesia
Objective:
In patients with diabetes, impaired insulin causes fat metabolism disorder, resulting elevated total cholesterol levels of decreased fat breakdown. Milk powder is a food that is high in vitamin D that can repair insulin resistance. The aim of this study was determine the effect of milk powder on total cholesterol levels in a model of type 2 diabetes (Rattus novergicus strain Wistar).
Methods:
This was an experimental study with posttest-only control group design. The study was conducted for 90 days using 30 male Wistar rats randomly divided into five groups: P0 (normal diet), P1 (high-fat diet), P2 (high-fat diet + 0.9 g milk powder), P3 (high-fat diet + 1.8 g milk powder), and P4 (high-fat diet + 2.7 g milk powder). The parameters measured were total cholesterol levels of rats.
Result:
The graph results showed that cow milk powder can increase total cholesterol levels of the negative control group. Kruskal-Wallis test showed no difference between groups (P = 0.232 ).
Conclusion:
Giving a dairy milk powder with doses such as 0.9, 1.8, and 2.7 g/day can increase total cholesterol levels of rats with type 2 diabetes.
Engineering of Fungi-Derived Flavin Adenine Dinucleotide Glucose Dehydrogenase for Continuous Glucose Monitoring
Genki Sakai, BS; Kazushige Mori, MS; Katsuhiro Kojima, PhD; Stefano Ferri, PhD; Aimi Suzuki, BS; Seiya Tsujimura, PhD; Koji Sode, PhD
Department of Industrial Technology and Innovation
Graduate School of Engineering, Tokyo University of Agriculture and Technology
Koganei, Tokyo, Japan
Objective:
Fungi-derived flavin adenine dinucleotide (FAD)-dependent glucose dehydrogenase (GDH) has received much attention for its glucose sensor application, focusing their narrow substrate specificity. GDH-FAD uses a variety of external electron acceptors, but not oxygen, making a sensor element for the blood glucose measurement that is not affected by the partial pressure of oxygen in the blood. However, stability of GDH-FAD is lower compared with glucose oxidase, which is the gold standard enzyme used in the sensor for the continuous glucose monitoring (CGM). This study aims to develop stable GDH-FAD suitable for CGM applications.
Method:
Based on the structural model of GDH-FAD, several amino acid substitutions were introduced. Thus, constructed engineered GDH-FADs were recombinantly produced using Escherichia coli as the host microorganisms, resulting in the nonglycosylated protein molecules, which is preferable for the combination with electrode-immobilized mediators. Stability of enzymes were evaluated based on their thermal stability as well as the stability of FAD binding.
Result:
The stability was greatly increased by introducing amino acid substitutions. The engineered GDH-FAD where one disulfide bond was introduced showed a 10-fold increase in the half-life of thermal inactivation compared with the wild-type enzyme. In addition, the holoenzyme stability, that is, the enzyme binding with its cofactor FAD, was also increased. Contrary to such drastic alteration of protein stability, the catalytic activity and the substrate specificity remained unchanged in the engineered GDH-FAD.
Conclusion:
The engineered GDH-FAD was constructed by amino acid substitutions without affecting its catalytic activity and substrate specificity. Thus, constructed molecules showed high thermal stability and holoenzyme stability, promising its application to the sensors for future CGM systems.
Remote Education and Management of Patients with Diabetes
Gary Scheiner, MS, CDE
Integrated Diabetes Services LLC
Wynnewood, Pennsylvania
Objective:
The aim is to demonstrate the cost savings, practicality, and effectiveness of 1) managing patients’ glucose levels and 2) teaching self-management skills via remote means.
Method:
Our practice consists of a multidisciplinary team of Certified Diabetes Educators with personal and professional expertise in intensive insulin therapy for children and adults with type 1 and type 2 diabetes. Working in collaboration with patients’ primary care providers and endocrinologists, our team consults patients (and their care providers) on strategies for optimizing glucose control by applying self-management skills. Services are provided through a variety of remote means, including telephonic consultation; video chat consultation; live chat (text only) consultation; e-mail correspondence; text messaging; website-based record keeping; electronic transmission of educational materials; sharing of downloaded data from glucose meters, insulin pumps, and continuous glucose monitors; and shared documentation via smartphone applications. All services are assessment-based, individualized, and delivered in accordance with American Diabetes Association Standards of Care and the AADE-7 Self-Care Behaviors initiative. In addition to traditional fee-for-service consultations, our practice offers ongoing retainer-type services and specialized programs for those who are newly diagnosed, use insulin pumps, are contemplating pregnancy, have weight loss needs, or are training for athletic events. Those engaged in a retainer/program service are offered monthly consultations consisting of pertinent self-management education and a detailed review of current glucose management strategies. Between appointments, patients are encouraged to download/submit data, convey questions, and obtain assistance whenever troubleshooting is required. In some instances, hands-on training/assessment is traditionally preferred. For example, when training on a self-management apparatus, checking for lipohypertrophy, or performing a foot examination. In most cases, the lack of face-to-face contact can be overcome by using video chat or training the patient’s partner to perform simple procedures. When more detailed visualization or palpation is necessary, the patient is referred to their local physician.
Result:
Patients report a very high level of satisfaction with remote care for their diabetes. They experience cost and time savings and are able to receive care in a more comfortable/relaxed setting than the traditional health care environment. No-show rates are near zero for scheduled remote appointments. Data collected over several years indicate a significant improvement in glycemic control based on reduced A1C levels and incidents of severe hypoglycemia. No significant complaints or negative events have been reported as a result of the remote form of care.
Conclusion:
Virtually every aspect of diabetes self-management education and glucose control can be achieved remotely. This approach to care is safe, practical, cost-effective, and results-oriented for all parties. It behooves health care providers to offer remote services to their patients and third-party payors to recognize remote care as a reimbursable service.
Modeling the Effect of Physical Activity on Postprandial Glucose Turnover in Healthy Subjects
Michele Schiavon, PhD; Chiara Dalla Man, PhD; Yogish C. Kudva, MD; Ananda Basu, MD; Claudio Cobelli, PhD.
Department of Information Engineering
University of Padova
Padova, Italy
Objective:
The effect of physical activity on glucose kinetics, although qualitatively well established, has been difficult to quantify. This represents a significant knowledge gap, especially in type 1 diabetes, since this information could be incorporated into currently available artificial pancreas control algorithms. However, such tools will need to be developed and tested in healthy subjects before validating in those with diabetes. Here a new model quantitatively describing the effect of physical activity on postprandial glucose turnover is presented.
Method:
Twelve healthy subjects underwent a triple tracer mixed meal containing 75 g of carbohydrate and performed moderate physical activity (50% V O2max) for 60 min during a 75-min period starting 120 min after the meal. Plasma glucose, insulin, and tracer concentrations were frequently measured and glucose fluxes calculated. A two-compartment model was used to describe glucose kinetics in absence of physical activity and appropriately modified to describe the effects of physical activity on glucose utilization. Model identification was performed using a Bayesian maximum a posteriori technique.
Result:
The model well fit the data and provides precise parameter estimates. In particular, we found that exercise increases both glucose transport from the plasma to the tissue compartment (0.56 ± 0.22) and insulin-independent glucose disposal in the tissue compartment (13.2 ± 7.9).
Conclusion:
A mathematical model quantifying the effect of physical activity on postprandial glucose turnover was developed and tested against data of healthy subjects. The proposed model is able to describe the effect of physical activity on glucose kinetics mainly by an increase in insulin-independent glucose disposal. If the model will be validated in type 1 diabetes, these results could be used to inform the next-generation control algorithms.
A Novel Approach toward Noninvasive Glucose Monitoring
Maarten J. Scholtes-Timmerman, MSc; Sabina Bijlsma, PhD; Marion J. Fokkert, BSc; Robbert Slingerland, PhD; Sjaak J.F. van Veen, MSc
TNO ELSS
Zeist, The Netherlands
Objective:
Noninvasive glucose diagnostics for home use is, understandably, a highly sought-after solution for diabetes patients. Increased health care costs forces a focus on diagnostics at home rather than in a professional care facility; easy-to-use and cheap glucometers are one way of reducing diabetes-related health care costs. We have assessed the power of a custom-made optical instrument, based on Raman spectroscopy, for noninvasive determination of glucose.
Method:
Raman spectra from the skin of the forearm of 111 hospitalized patients were taken. Also, capillary blood samples were taken for glucose reference levels. These Raman spectra were, retrospectively, correlated to reference glucose values using ten-fold double- cross partial least squares validation. Further stratification of the cohort by sex was used to ameliorate prediction quality.
Result:
Using male and female subjects, a correlation (R2) of 0.83 was found between Raman data and glucose references. However, further stratification gave correlations of 0.88 and 0.94 for female only and male only, respectively. Mean absolute prediction errors were as low as 0.9 mmol/L (16 mg/dL). This equals a relative prediction error <10%.
Conclusion:
Using Raman spectroscopy, and an optimized optical instrument, noninvasive glucose predictions with errors as low as 16 mg/dL are achievable. The system is designed in such a way that further development into an affordable, stable, and easy-to-use point-of-care glucometer is the next step.
Simulating the Effects of Lifestyle Choices on Blood Glucose Control: Implications of Predictions for mHealth and Artificial Pancreas Design
Sofie W. Schunk, BA; Jack M. Winters, PhD
Marquette University
Milwaukee, Wisconsin
Objective:
Many factors contribute to proper endogenous and exogenous blood glucose regulation. These include mode (and intensity) of exercise, daily movement activities, and meal type distinction, each challenging to model, but done so in an eight-state Matlab simulation framework.
Method:
Features include a separate muscle tissue compartment, addressing aerobic and anaerobic workload. Workload results in immediate and longer-term effects on glucose levels due to increased demand in muscle fuel consumption, nonlinear flux modulation for stored glucose (glycogen) breakdown, and insulin sensitivity. Fluctuating daily activity is anaerobic, consuming glucose. Foodstuff is partitioned into slower and faster digestive breakdown paths by glycemic index (with low glycemic meal composition including an additional, nonlinear, slower, saturating pathway). These phenomena use saturating Hill kinetics. The eight -state model includes a compartmental flow/mass balance (three for glucose, including muscle and nonmuscle, and two for digestive) and hormones as unidirectional modulation signals (two for insulin and one for glucagon).
Result:
Effort was focused on replicating meal-response predictions of models encouraged by the Food and Drug Administration for preclinical trials and available data on the effects of physical activity and workload. Adding the aforementioned practical everyday features could help real-time mHealth controllers take advantage of ongoing model predictions within a framework enabling personalized adaptive tuning. Simulation results focus on type 1 and type 2 versions.
Conclusion:
The proposed framework complements well-developed past models of blood glucose and insulin kinetics and, more recently, glucagon dynamics. It is intended to inform artificial pancreas design and mHealth-based disease management.
Temporal Association Rules for Stratification of Type 2 Diabetes Patients
Daniele Segagni, MD; Lucia Sacchi, PhD; Arianna Dagliati, MD; Paola Leporati, MD; Pasquale De Cata, MD; John H. Holmes, PhD; Carlo Cerra, MD; Luca Chiovato, PhD; Riccardo Bellazzi, PhD
IRCCS Fondazione Salvatore Maugeri
Pavia, Italy
Objective:
In this work, we present a system, based on temporal data mining methods, able to guide caregivers identifying the best diagnostic and care pathways for type 2 diabetes mellitus (T2DM) patients. Clinical data of the IRCCS Fondazione Salvatore Maugeri hospital have been enhanced with administrative data coming from the local health care agency of the Pavia area and with environmental data provided at a regional level.
Method:
The core of the system relies on the i2b2 clinical data warehouse. The system encloses a module able to extract specific abstract concepts from raw quantitative data through a temporal abstraction framework and store them in the data warehouse itself. Through a dashboard integrated in the EMR, users can perform analysis during their daily practice. Temporal aspects related to the evolution of the disease have been analyzed, showing the most frequent care flows in terms of hospitalizations and visits.
Result:
We focused our attention on T2DM patients with clinical follow-up between 5 and 10 years. For these patients, the most frequent temporal patterns have highlighted an intervention on diet, whereby 34% of patients had achieved improved eating habits. Furthermore, there was evidence of improved HbA1c control in 39% of patients, while good HbA1c control was reached through metformin use in 48%.
Conclusion:
Considering the temporal aspects of the evolution of a disease and its complications, the implemented system offers a novel perspective, giving a dynamic picture of the treated population. The proposed system offers the opportunity to evaluate the most frequent temporal patterns of a T2DM population, grouping patients with similar disease progressions and giving doctors the possibility of managing critical situations, and may help health care managers in the allocation of hospital resources.
In Vitro/In Vivo Characterization of the Continuous Glucose Monitoring Sensor Using Direct Electron Transfer Principle
Shinjiro Sekimoto, MS; Yosuke Murase, MS; Kotaro Shinozaki, MS; Akihiro Yamamoto, ME; Yuki Matsuura, MS; Yasuhide Kusaka, ME; Wakako Tsugawa, PhD; Koji Sode, PhD
ARKRAY Inc.
Kyoto, Japan
Objective:
The continuous glucose monitoring (CGM) sensor based on direct electron transfer (DiET) technology was characterized. DiET technology uses a unique glucose dehydrogenase chemistry capable of DiET from the enzyme redox center to electrode. Therefore the sensor does not require any artificial electron mediator in order to transfer electrons to the electrode. This study reports the detailed features and the improvements of the performance of this novel CGM sensor. Data are presented to show in vitro performance of the sensor operating at various potentials to provide information on accuracy improvement of the sensor and in vivo studies to evaluate the continuous operation of the sensor with modified conditions.
Method:
A sensor consisting of three electrodes was used as the reference platform. Sensor reagent containing glucose dehydrogenase was dropped onto the working electrode. The sensor performance, as well as its stability, was evaluated by operating under various potentials (50–400 mV; phosphate buffer, pH 7.4) to the working electrode for the Ag/AgCl electrode. The in vivo study was performed at 200 mV by inserting the sensor into the subcutaneous area of the abdomen of a healthy internal volunteer subject. Capillary blood glucose values were also obtained.
Result:
Little change was observed in the sensitivity after 7-day in vitro continuous measurement at various measuring voltages (50–400 mV). CGM operation of the in vivo human study showed good correlation to capillary blood glucose values for 7 days.
Conclusion:
A novel CGM sensor was developed using DiET technology. The sensor showed stable CGM response when measurements were performed at lower voltages. Improvement of accuracy is expected by the measurement at lower voltage due to reduction of interference materials.
Evaluation of a Methodology for Estimating HbA1c Values by a New Glucose Meter
Jochen Sieber, MD, PhD; Frank Flack, PhD; Bonnie Dumais, BS, AS; Casey C. Peters, BA, MA; Erin B. Mallery, MS; Liz Taylor, MS, RD, LD, CDE, CCRC, CES
Sanofi
Frankfurt, Germany
Objective:
Performance of HbA1c estimation (~A1C) with an algorithm built into a blood glucose meter (MyStar Extra) has been demonstrated by in silico testing, but prospective data were missing. We evaluated the performance of ~A1C and the ease of use in a clinical setting.
Methods:
One hundred thirty-three subjects (mean ± SD age, 60.0 ± 1.3 years; 69 male; 104 patients with diabetes, 29 without) used the blood glucose meter for 4 months to obtain ~A1C in this open-label, single-center study. Laboratory measurement of HbA1c was done every 2 weeks at the investigators’ site and the corresponding ~A1C documented. Subjects completed a questionnaire at the end of the study.
Results:
Mean HbA1c at study entry was 7.0 ± 0.1% (minimum 5.5%, maximum 11.9%); 1,008 pairs of ~A1C and laboratory HbA1c were available; 97.5% of the ~A1C results in diabetic patients fell within ±20% of the laboratory HbA1c, 95.0% within ±18%, and 90.7% within ±15%. Mean slope of the scatter plot of ~A1C versus laboratory HbA1c was 0.53 (95% CI 0.51–0.55), mean intercept 0.03 (95% CI 0.03–0.04). Seventy-two percent of subjects agreed or strongly agreed that the ~A1C section in the user guide and the flash cards provided with the meter were easy to follow. Seventy percent of subjects agreed or strongly agreed that they would use the system to track their ~A1C. Seventy-nine percent agreed or strongly agreed that they find the ~A1C tool helpful. Eighty-three percent agreed or strongly agreed that the tool may motivate to better manage diabetes.
Conclusion:
Accuracy of the ~A1C in this clinical setting was similar to the performance in in silico studies. The majority of subjects found this tool helpful and agreed that it may motivate to better manage diabetes.
Duality of Interest:
This study was supported by AgaMatrix.
Hypoglycemia Reduction with Threshold Suspend and Its Economic Impact in Insulin-Treated Patients with Diabetes in a U.S. Health Plan
Joseph A. Sierra, BA; Matthew Sussman, MA; Richard M. Bergenstal, MD; Satish K. Garg, MD; Bruce W. Bode, MD; Molly Frean, BA; Mark Friedman, MD; Max S. Gill, MS, MBA; Francine R. Kaufman, MD; Joseph Menzin, PhD
Medtronic Diabetes
Northridge, California
Objective:
Patients with diabetes requiring insulin therapy are at higher risk of hypoglycemia, which may result in dizziness, confusion, seizure, or coma. The objective of this study was to assess the cost of hypoglycemic events among insulin-treated patients with diabetes and the potential financial impact to a U.S. health plan of reducing hypoglycemic events.
Method:
A cost-calculator model was developed to estimate the direct financial burden of hypoglycemic events, accounting for diabetes type, age, and event severity. Inputs were derived from published rates of hypoglycemic events and direct medical costs. Assumed intervention efficacy was based on published studies of insulin pumps with a threshold-suspend feature that yielded 72.2 and 31.8% reductions in severe and nonsevere hypoglycemic events, respectively. The extrapolated number of severe (requiring medical assistance) and nonsevere events, and associated direct medical costs, were evaluated over a 1-year period.
Result:
In a health plan with 10 million enrollees, we estimated that 256,000 patients would have insulin-treated diabetes (both type 1 and type 2). Without the intervention, patients would have experienced 0.09 and 14.60 severe and nonsevere hypoglycemic events per patient per year (PPPY), respectively, leading to direct medical costs of $245 PPPY. Using threshold suspend, patients would have experienced 0.02 and 9.96 severe and nonsevere events PPPY, respectively, leading to a 72% reduction ($68 PPPY) in direct medical costs, translating into annual savings of $45 million for the health plan.
Conclusion:
Insulin-treated patients are at high risk of hypoglycemia associated with significant financial burden. Payors and providers should consider medical interventions such as insulin pump therapy with a threshold-suspend feature, which reduces hypoglycemic events.
Intermittent Versus Continuous Glucose Monitoring to Detect Hypoglycemia in Neonates
Matthew Signal, PhD; Deborah L. Harris, PhD; Philip J. Weston, FRACP; Felicity Thomas, BE (Hons); Jane E. Harding, FRACP, DPhil; J. Geoffrey Chase, PhD; on behalf of The CHYLD Study Group
Department of Mechanical Engineering
University of Canterbury
Christchurch, Canterbury, New Zealand
Objective:
Continuous glucose monitoring (CGM) devices have been used in newborn infants to detect hypoglycemia, which is thought to cause negative long-term developmental outcomes. Several studies report that CGM detects hypoglycemia at a higher rate than intermittent blood glucose (BG) measurements, but CGM data are typically not available for at least 1 h after sensor insertion. This study compares hypoglycemia detection via intermittent BG monitoring and CGM in neonates.
Method:
CGM and BG data from 161 at -risk infants were used to investigate hypoglycemia (BG <47 mg/dL). All BG measurements were determined using a blood-gas analyzer (Radiometer ABL800FLEX) and CGM (Medtronic CGMS Gold) sensors were inserted as soon as was practical after birth. A previously published recalibration algorithm was applied to CGM data before median filtering. The incidence of hypoglycemia detected by BG and CGM was compared.
Result:
Of the 2,271 BG measurements recorded, 399 detected hypoglycemia, and 198 of these occurred during CGM. However, CGM reported 337 hypoglycemic episodes, suggesting that 139 episodes were missed and occurred between intermittent BG measurements. Conversely, during the first 6 h after birth, when hypoglycemia was most prevalent, a total of 213 episodes were detected by BG measurements, and only 65 of these were detected by CGM due to the ~3-h delay after birth before CGM data were available. After this 6-h period, CGM consistently detected more episodes than intermittent BG monitoring.
Conclusion:
This investigation showed that CGM detects more episodes of hypoglycemia than BG measurement, but only after the first 6 h. Thus an optimal measurement protocol for detecting the episodes might consist of CGM with regular intermittent BG measurements for the first 6 h after birth.
Clinical Experience with the V-Go Disposable Insulin Delivery Device
John H. Sink II, PA-C; Barry R. Johns, MD; Thomas C. Jones, MD; Catherine E. Cooke, PharmD
The Jones Center for Diabetes and Endocrine Wellness Macon, Georgia
Objective:
We describe the benefits and challenges of incorporating a new insulin delivery device into clinical practice.
Method:
A query of electronic medical records identified patients who had initiated therapy with V-Go, a disposable insulin delivery device, at a community-based diabetes center. Demographic (age, sex, race) and clinical data (type of diabetes, year diabetes diagnosed, BMI, glycated hemoglobin [A1C], antidiabetes medication regimen) were abstracted from the electronic medical records at V-Go initiation and for up to 1 year of follow-up.
Result:
The electronic query identified 91 adults (type 2 diabetes n = 86; type 1 diabetes n = 5) who had initiated V-Go. Clinical and demographic parameters for type 2 diabetes were as follows: 40% African American, 73% diagnosed ≥5 years ago, 76% had A1C ≥8%, and 93% were using insulin with 85% (68/80) using total daily insulin doses >40 units. At first follow-up after V-Go initiation (n = 69), the mean (SD) difference in A1C was −0.8 (0.01). Additionally, 65% of patients reduced their total daily insulin doses when switching to V-Go. For patients with type 1 diabetes, all had diabetes of ≥5 years duration and A1C >8%. After V-Go initiation, three of four patients with at least one follow-up had A1C reductions. Overall, 79% were persistent with V-Go as of the last recorded follow-up, with 14% experiencing gaps in therapy and 11% noting difficulty obtaining V-Go because of cost/lack of insurance coverage.
Conclusion:
V-Go, a disposable insulin delivery device, was mostly initiated in patients with type 2 diabetes with elevated A1C levels who were already using insulin. Improvements in A1C and a decrease in total daily insulin dose were found. The greatest challenge was remaining on therapy because of cost/lack of insurance coverage.
The Use of Big Data and Online Crowdsourcing to Improve Outcomes and Reduce Treatment Costs
Michael Slage, MA
HealthEngage Inc.
Arlington, Virginia
Objective:
Because of the global growth of online health tools and communities, research technologies and methods must be developed to effectively access the users and data being generated by these sites.
Method:
We examine the assessment, development, and implementation processes associated with using an online social community’s existing aggregate user and data assets for research and data applications.
Result:
Addressed will be the many technical, research, and business issues at play and what the implications are for the industry overall. Case studies will be discussed during the presentation to describe what works and what does not when using this type of data for research purposes.
Conclusion:
Given the instant nature of the web, a whole new category of products and services for research can be tapped. Researchers can access and export real-time and retrospective aggregate data sets in seconds or view a real-time dashboard to be able to see developing trends. Using these tools can improve outcomes for individual patients as well as generating global aggregate data that can be used to help all diabetes patients.
Performance of Systems for Self-Monitoring of Blood Glucose Compared with the Analytical Quality Specifications in ISO 15197:2013
Una Ørvim Sølvik, PhD; Marianne Risa, BLS; Camilla Eide Jacobsen, MS; Grete Monsen, BLS; Sverre Sandberg, MD, PhD
Department of Global Health and Primary Care
Faculty of Medicine and Dentistry
University of Bergen
Bergen, Norway
Objective:
The aim of this study was to assess the accuracy of 10 different self-monitoring of blood glucose (SMBG) systems against the quality goal specified in ISO 15197:2013 under optimal conditions performed by biomedical laboratory scientists (BLSs) in a hospital environment and by persons with diabetes. In addition, the SMBG systems were evaluated against the American Diabetes Association quality goal for imprecision (CV ≤5%).
Method:
Data from 10 SMBG systems evaluations performed by the Scandinavian evaluation of laboratory equipment for primary health care (www.skup.nu) were used. Approximately 90 persons with diabetes participated in each evaluation. All measurements on each SMBG system were compared with the glucose hexokinase method. Standard reference material from the National Institute of Standards and Technology was used to show traceability, and controls with target values determined at a reference laboratory were used to verify trueness of the comparison method. The imprecision of the SMBG systems was calculated using duplicate results.
Result:
Out of the 10 SMBG systems, 9 and 6 fulfilled the requirements for accuracy when a BLS and a person with diabetes performed the measurements, respectively. Out of the 10 SMBG systems, 10 and 8 were below the recommended limit for repeatability when a BLS and a person with diabetes performed the measurements, respectively.
Conclusion:
Nearly all the SMBG systems fulfill the analytical quality specification in ISO 15197:2013 and from American Diabetes Association, especially when the BLS performs the measurements. All SMBG systems fulfilled the requirement in ISO 15197:2003. It appears that the quality of SMBG systems has improved in past years. This contributes to more accurate and reliable results, which is essential for therapeutic decisions.
Individualized Rapid-Acting Insulin Action Estimation from Insulin Pump and Continuous Glucose Measurement Data
Fredrik Ståhl, MSc; Mona Landin-Olsson, MD, PhD; Rolf Johansson, MD, PhD
Department of Automatic Control
Lund University
Lund, Sweden
Objective:
Knowledge of the glucose-lowering effect of insulin and the dynamics thereof—the insulin action (IA)—is important to optimize insulin therapy. In modern insulin pumps, generic estimates of the IA are used as components in bolus guides, acting as a decision support system to determine meal and correction boluses. The shape of these generic estimates are based on glucose clamp experiments and do not reflect the dynamic response of the individual user. The total glucose-lowering effect estimate is either drawn from individual experience or by use of generic formulas as the 100 rule. As an alternative to the above, could individualized estimates be achieved using home-monitored continuous insulin infusion and continuous glucose measurement data?
Method:
Twenty-nine patients on continuous insulin infusion therapy were monitored with Medtronic Sof (6 patients) and Enlite (8 patients) and Dexcom G4 (15 patients) sensors for more than 36 days. Using the fasting overnight data series, a combined pharmacokinetic/pharmacodynamic model, together with the basal insulin levels needed to maintain a steady-state fasting glucose value, were estimated for each patient.
Result:
The individualized action profiles had a glucose-lowering effect of 2.4 (1.1–3.7) mmol/(L∙IU), and half of the effect was achieved after 135 (105–165) min. When used for overnight glucose prediction, assuming the basal level to be known, leave-one-out cross validation implied an estimation error corresponding to a mean absolute relative difference of 11.2% (5.3–18.4%).
Conclusion:
The results imply that the individualized IA can be estimated from a limited data set with useful accuracy. Updating the pump decision support system with these estimates could imply improved capabilities to assess the risk of nocturnal hypoglycemia and to achieve better tailored bolus doses.
A Device to Support the Use of a Glucose Monitor by Individuals Who Are Blind or Visually Impaired
Matthew Standard, BS; Jacob Park, BS; Dianne Pawluk, PhD; John Clore, MD; Amber Spain, MSW, LCSW, CDE; Linda Thurby Hay, MS, RN, ACNS-BC, BC-ADM
Virginia Commonwealth University
Richmond, Virginia
Objective:
Diabetes patients who are blind or visually impaired find it difficult to properly place their blood sample on a test strip because of their lack of sight to guide the procedure. This leads to frustration, anger, and/or stress as well as either increased costs or skipped tests because of wasted test strips. Our objective was to develop a simple-to-use, low-cost, sanitary, and effective device to aid with procurement of a blood sample on a test strip for a talking glucometer.
Method:
The device developed is comprised of two parts: a constrained space where the glucometer is placed from which the test strip extends and a hollow cylinder with a side slit aligned with the test strip. Haptic cues in the form of ridges guide the user from the lancing position on the top of the cylinder to the test strip via the rotation of their finger inside the cylindrical hollow. The device can be used with different fingers as well as different locations on the finger. It is also dishwasher safe.
Result:
Pilot studies with blindfolded sighted subjects indicated that the device reliably guides the blood from the lancing site to the test strip. An institutional review board–approved longitudinal study involving 24 users who are blind or visually impaired is underway. The study is recording subjective measures of frustration, anger, and stress as well as quantitative measures of number of test strips used and finger pricks performed.
Conclusion:
The device has shown promise to be a simple and low-cost assistive device to address the frustration individuals who are blind or visually impaired feel while taking their daily blood glucose measurements.
Checking In: Feasibility of a Physician-Delivered Intervention to Improve Blood Glucose Monitoring
Alexa Stern, BA; Lauren Clary, PhD; Priya Mehta, BA; Christina Sharkey; Fran Cogen, MD, CDE; Priya Vaidyanathan, MD; Celia Henderson, RN, CDE; Randi Streisand, PhD, CDE; Maureen Monaghan, PhD
Children’s National Health System
Washington DC
Objective:
There is a significant need to develop brief, low-cost interventions to improve adherence and glycemic control in teens with type 1 diabetes (T1D). This study evaluated a physician-delivered and text-message intervention designed to increase parent-teen communication about blood glucose (BG) monitoring.
Method:
Thirty teens (mean age = 13.7 years; 46.7% female) enrolled in Checking In and received the physician-delivered intervention, consisting of print materials and encouraging parents and teens to schedule 3-min meetings three times per week to review BG meters. Participants received weekly text-message boosters for 12 weeks after intervention delivery and completed a satisfaction survey at 12 weeks. A1C and glucometer data (30 days) were abstracted from medical charts.
Result:
Fifty-one percent of eligible families agreed to participate, 63% of dyads completed satisfaction surveys, and follow-up medical data were obtained for 93% of participants. While changes were in the expected direction, there were not statistically significant changes from baseline to follow-up for A1C (8.83 vs. 8.61%), BG monitoring (4.07 vs. 4.52 checks/day), or mean BG level (216.28 vs. 209.77 mg/dL); P > 0.05 for all. Participants were satisfied with the program. All parents reported a desire to continue 3-min meetings with their teen, and 55% felt their child’s T1D management improved because of Checking In. Parents were enthusiastic about text reminders, with 86% of parents and 36% of teens rating the texts as helpful.
Conclusion:
Results demonstrate feasibility of a low-cost physician-delivered intervention targeting parent- child communication about BG values. Parents may benefit from text-message reminders about diabetes care as they support and monitor their teen’s T1D management. This intervention should be conducted with a larger sample to evaluate program impact and potential areas for improvement.
Stochastic Model Predictive Glycemic Control for the Intensive Care Unit
Kent W. Stewart; Christopher G. Pretty, PhD; Hamish Tomlinson, BE (Hons); Liam Fisk, BE (Hons); Geoffery M. Shaw, MbChB, FJFICM; J. Geoffrey Chase, PhD
Department of Mechanical Engineering
Centre for Bio-Engineering
University of Canterbury
Christchurch, Canterbury, New Zealand
Objective:
Critically ill patients often experience stress-induced hyperglycemia, which has been shown to result in increased morbidity and mortality. Stochastic model predictive (STOMP) control is a new model-based glycemic control protocol that implements the probabilistic, stochastic forecasting methods of its predecessor, stochastic targeted (STAR). STOMP formalizes the control methodology using model predictive control theory to improve the ability to tune the dynamic response of the controller.
Methods:
The STOMP controller uses a model of the glucose-insulin system combined with a model of insulin sensitivity variability to predict the response to a proposed insulin/nutrition intervention. An optimal intervention is selected through minimization of the weighted sum of 6-h forecasted performance metrics derived from penalty functions, enabling safe 4-hourly measurement intervals. STOMP was evaluated using the validated virtual trial method with 149 virtual patient profiles, comprising 17,610 h of clinical data collected from the Christchurch Hospital intensive care unit.
Results:
The glycemic performance of STOMP was very similar to STAR. Both protocols maintain 85% of time within the 4.4–8.0 mmol/L glycemic target band and have very low incidence of hypoglycemia (0.06% time <2.2 mmol/L). STOMP simultaneously maintained nutrition targets, with a median patient nutrition rate of 90% of the desired goal of 25 kcal/kg/day. Thus STOMP maintained equivalent glycemic and nutritional performance to STAR, with the advantage of 35% fewer measurements.
Conclusions:
STOMP formalizes the heuristic algorithm of STAR, allowing customization to patient-specific requirements and variation in clinical practices. STOMP’s robust long-term forecasting prioritizes glycemic stability, allowing fewer interventions, which reduces clinical burden. STOMP’s strong glycemic and nutrition delivery performance in this virtual trial demonstrates its potential as a valuable enhancement to the STAR glycemic control framework.
Screening for Glucose Intolerance Using Area under the Curve after Glucose Loading
Chihiro Suminaka, MSc; Toshihiro Watanabe, MSc; Aya Morimoto, MSc; Samiko Hosoya, BSc; Toshiyuki Sato, MSc
Central Research Laboratories
Sysmex Corporation
Kobe, Japan
Objective:
We developed a glucose area under the curve (AUC) measurement system using minimally invasive interstitial fluid extraction technology (MIET) that allows convenient monitoring of postprandial glucose levels without blood sampling. Although the glucose AUC is believed to reflect whole information about glycemic excursion after glucose loading, there are few reports regarding its application to screening. Here we evaluated the performance of screening for glucose intolerance using AUC.
Method:
The subjects were 570 patients who underwent oral glucose tolerance tests (75 g). In 50 patients, we performed simultaneous measurement of interstitial fluid glucose AUC (IG-AUC) using MIET. Plasma glucose (PG) levels were measured at predefined points for 2 h after glucose loading to calculate PG-AUC. MIET included a pretreatment step using a plastic microneedle array to enable painless transdermal interstitial fluid glucose extraction and an accumulation step with a hydrogel patch placed on the pretreated area for a predefined period.
Result:
We defined the cutoff level for PG-AUC during oral glucose tolerance tests as 290 mg·h/dL according to the receiver-operating-characteristic curve (sensitivity, 0.90; specificity, 0.93). The receiver-operating-characteristic curve showed that PG-AUC could identify glucose intolerance better than HbA1c, fasting PG, or 2-h PG levels. Moreover, IG-AUC measured by MIET could identify diabetes using the same cutoff value. Both PG-AUC and IG-AUC accurately reflected peak PG levels and could identify patients with high peak PG ≥180levels ( mg/dL).
Conclusion:
PG-AUC is a reliable index of glucose intolerance, and IG-AUC measured by MIET performed as well as PG-AUC without the need for blood sampling. We are conducting a feasibility study on the simultaneous measurement of insulin AUC using MIET and plan to report the results at the annual meeting.
Factors Influencing Overnight Closed-Loop Performance during Free Living in Children and Adults with Type 1 Diabetes
Martin Tauschmann, MD; Hood Thabit, MD; Daniela Elleri, PhD; Alexandra Lubina-Solomon, PhD; Marietta Stadler; PhD; Simon R. Heller, FRCP; Stephanie A. Amiel, FRCP; Mark L. Evans, FRCP; David B. Dunger, MD; Roman Hovorka, PhD
Wellcome Trust-MRC Institute of Metabolic Science
University of Cambridge
Cambridge, United Kingdom
Objectives:
Overnight closed-loop (OCL) insulin delivery is feasible, safe, and effective in the home setting, but outcomes vary between individuals. Understanding factors influencing glucose control may help target those likely to benefit most and propose appropriate use characteristics.
Methods:
We combined data collected during two randomized at home studies in 16 adolescents and 24 adults with type 1 diabetes on insulin pump therapy. Participants underwent, in random order, two periods of sensor-augmented insulin pump therapy or sensor-augmented insulin pump therapy combined with OCL insulin delivery, each lasting 3 (adolescents) or 4 (adults) weeks. Associations between baseline characteristics (age, HbA 1c, duration of diabetes), utility (time of OCL start, duration of OCL application per night), and performance of OCL (time in target [TT], improvement in TT) were examined using Spearman correlation.
Results:
We analyzed data on 866 closed-loop nights. Longer OCL application was associated with greater improvement in TT between 3.9 to 8.0 mmol/L (r = 0.39; P = 0.013). Baseline HbA1c was not associated with TT, improvements in TT, or duration of OCL application. Being adolescent as opposed to being adult as well as shorter duration of diabetes was associated with higher TT (age group r = 0.47, P < 0.01; duration r = −0.46, P < 0.01). Higher age and longer diabetes duration was associated with later OCL start (age r = 0.58, P < 0.01; duration r = 0.47, P < 0.01) and shorter OCL application (age r = −0.58, P < 0.01; duration r = −0.44, P < 0.01).
Conclusions:
In adolescents and adults with type 1 diabetes, the extent of glucose improvements during OCL is seen irrespective of baseline HbA1c. Improved outcomes may be observed if OCL is started earlier and applied for longer, particularly in adults.
Antihyperglycemic Potential of Psidium guajava Leaf in Streptozotocin-Induced Diabetic Rats
Toluwani Tella, BSc, MSc; Samson Mukaratirwa, BVSc, MVSc, PhD; Bubuya Masola, BSc, MSc, PhD
Discipline of Biochemistry
School of Life Sciences
University of KwaZulu-Natal (Westville Campus)
KwaZulu-Natal, South Africa
Objectives:
Psidium guajava (PG) leaf is known to have a blood-glucose-lowering effect in diabetic rats. The objectives of the present study were to carry out a phytochemical study of PG leaf extract and investigate its protective effect on the pancreas and its effect on muscle glycogen synthase and phosphorylase activities in streptozotocin-induced diabetic male Sprague-Dawley rats.
Methods:
Diabetes was induced in male Sprague-Dawley rats with a single dose of 40 mg/kg body weight streptozotocin. The aqueous extract of PG leaves was used to treat both normal and diabetic animals (400 mg/kg body weight) for 2 weeks while control animals were treated with the vehicle.
Results:
After 2 weeks of treatment, PG lowered blood glucose and protected pancreatic tissue from diabetic damage. The treatment restored glycogen synthase activity depressed by diabetes and decreased glycogen phosphorylase activity in skeletal muscle. Gas chromatography–mass spectrometry analysis of the aqueous extract of PG indicated the presence of phenolic compounds and triterpenoids.
Conclusions:
We conclude that PG has a significant antihyperglycemic effect and that this effect may be associated with the presence of phenolic compounds and triterpenoids. PG also protects the pancreas against diabetic damage and modulates the activity of enzymes in the insulin-signaling pathway.
Accuracy and Precision of Three Common Glucose Meters in the Intensive Care Unit
Felicity Thomas, BE (Hons); Matthew Signal, PhD; Chris Pretty, PhD; Geoffrey M. Shaw, MbChB FJFICM; J. Geoffrey Chase, PhD
Department of Mechanical Engineering
University of Canterbury
Christchurch, Canterbury, New Zealand
Objective:
Handheld glucometers have become standard in most intensive care units (ICUs) for monitoring patients with stress-induced hyperglycemia. Inaccuracies in these devices can lead to reduced glycemic control performance. This research quantifies and compares the performance of three glucometers (Abbott Optium Xceed, Roche Accu-Chek Inform II, and Nova StatStrip) in the ICU setting.
Methods:
Blood samples from 13 critically ill patients were analyzed for blood glucose (BG) concentration using a blood-gas analyzer (BGA; Radiometer ABL90 Flex). Aliquots from each sample were also distributed across up to five glucometers of each model, resulting in 724 Abbott BGA, 432 Nova BGA, and 481 Roche BGA paired measurements. Bias and precision were used to quantify performance. Bias was defined as the median of the discrepancies between glucometer and BGA for a given blood sample. Precision was characterized by the difference between the maximum and minimum glucometer values for a given blood sample.
Results:
The mean (bias) results for the glucometers were Abbott 0.4 (0.6) mmol/L, Nova −0.1 (0.5) mmol/L, and Roche −0.2 (0.3) mmol/L.
Conclusion:
The Nova and the Roche glucometers adjust for hematocrit when calculating plasma glucose concentration. The lack of patient-specific hematocrit adjustment may account for the increased bias and reduced precision of the Abbott meter. Both the Nova and Roche glucometers are designed for point-of-care testing in the hospital environment. Their negative bias means they will underestimate BG level, which is safe. In contrast, the Abbott, an inexpensive device designed for diabetes management, can overestimate BG. This overestimation may not be significant for many cases of diabetes management. However, in an ICU setting, with very variable patients, underestimation is preferred, as it would result in lower insulin doses and increased safety.
Glycemic and Nutrition Delivery: Performance of the Stochastic Targeted Protocol
Hamish Tomlinson, BE (Hons); Liam Fisk, BE (Hons); James Geoffrey Chase, PhD; Christopher Pretty, PhD;
Geoffrey Shaw, MbChB, FJFICM
Department of Mechanical Engineering
University of Canterbury
Christchurch, Canterbury, New Zealand
Objective:
Stress-induced hyperglycemia is a significant issue in the intensive care unit (ICU). Accurate glycemic control aims to reduce episodes of hyperglycemia to mitigate negative outcomes while simultaneously eliminating the risk of hypoglycemia. The stochastic targeted (STAR) protocol provided safe and effective glycemic control in pilot trials. This research evaluates the long-term glycemic and nutrition delivery performance of STAR in the Christchurch Hospital ICU.
Method:
STAR is tablet based and uses a model of the glucose-insulin system, coupled with stochastic-model-derived risk intervals, to compute optimal nutrition and insulin interventions. The software lets clinical staff enter patient-specific requirements, which constrain the optimization algorithm. Patient data from June 2011 to December 2013 were retrieved from the Christchurch ICU. A total of 183 patients were analyzed, after exclusion of patients who spent less than 24 h on STAR.
Result:
Patients on STAR spent 80.6% of time within recommended blood glucose (BG) limits of 4.4–8.0 mmol/L, with an average of 13.6 BG measurements/day. Only three cases of BG less than 2.2 mmol/L occurred (1.6% by patients), all caused by clinical error. The median patient received over 90% of the 25 kcal/kg/day caloric goal by day 4 and 1 g/kg/day of protein by day 2. Over 60% of patients received 1 g/kg/day of protein by day 3.
Conclusion:
The STAR protocol has consistently maintained safe glycemic levels in the Christchurch ICU and provides flexibility to patient-specific requirements while reducing clinical burden. It reduces feed to combat high insulin resistance, which is prevalent in early days of stay, but exceeds most published reports in delivering total and protein nutrition. This research supports STAR as an important companion to regular clinical practice for optimized glycemic control and nutrition delivery.
Detection of Continuous Glucose Monitor Faults for Patients with Type 1 Diabetes
Kamuran Turksoy, BS; Lauretta Quinn, PhD; Elizabeth Littlejohn, MD; Ali Cinar, PhD
Department of Biomedical Engineering
Illinois Institute of Technology
Chicago, Illinois
Objective:
Recent progress in artificial pancreas (AP) development efforts yielded various AP system concepts that can automatically regulate the blood glucose concentrations of patients with type 1 diabetes. Even though the performance of these studies has been very encouraging, they have also indicated the need for improving the reliability of APs in routine use by patients during their daily activities. An AP includes several components such as glucose sensors, continuous glucose monitors (CGMs), pumps, and infusion sets that are prone to failures/faults. The development of algorithm-based approaches to detect CGM-related faults are necessary to maintain euglycemia.
Methods:
Multivariate statistical methods such as principal component analysis are used to develop multivariable statistical models. Hotelling T2 and squared prediction error control charts are used for detection of faults. Two sensors (Medtronic Enlite) are used to collect glucose concentration information. The SenseWear Pro3 armband (BodyMedia Inc., Pittsburgh, PA) is worn for the collection of metabolic and physical activity data.
Results:
Two different faults were performed during one clinical experiment. Glucose and physical activity information were collected every 5 min. The first fault (sensor value was decreased by 15% [30 mg/dL]) was detected 5 min after performing. The second fault (15% [16 mg/dL)] increase in CGM value) was detected without any delays.
Conclusion:
The preliminary results show that CGM-related faults could be detected using multivariable statistical models. More complex faults, such as drift changes or noise amount in CGM readings, are being investigated in current studies.
Evaluation of Luminescent Copolymer Hydrogels as Fully Implantable Sensors
Rachel Unruh, BS; Jason Roberts, PhD; Scott Nichols, PhD; Mary K. Balaconis, PhD; Natalie Wisniewski, PhD; Mike McShane, PhD
Department of Biomedical Engineering
Texas A&M University
College Station, Texas
Objective:
Current continuous glucose monitoring systems feature transcutaneous electrochemical probes with limited duration of use and accuracy. A fully implantable, optically interrogated sensor has been proposed as a paradigm shift toward long-term monitoring. In this study, we evaluated the potential of various copolymer hydrogels as hosts for glucose-sensing chemistry by measuring their glucose-tracking performance in vitro and in vivo.
Method:
Porous and solid copolymer hydrogels of poly(2-hydroxyethyl methacrylate) (pHEMA), poly(acrylamide) (67.2 wt% in deionized water), and poly(dimethylacrylamide) containing 90:10 and 75:25 HEMA:comonomer v/v% ratios were used. Each formulation contained equivalent concentrations of glucose oxidase and phosphorescent oxygen indicator, a palladium benzoporphyrin. Sensor response was evaluated in vitro in a benchtop flow-through system. In vivo performance of subcutaneous implants was measured on a porcine model challenged with glucose and insulin. A custom luminescence lifetime measurement system was used for all measurements.
Result:
The hydrogels exhibited measurable response ranges (R) and linearized sensitivities (S) as follows. pHEMA-co-acrylamide 90:10 R = 4.31–398.23 mg/dL, S = 0.43 ± 0.02 µs(mg/dL)−1; 75:25 R = 0.34−187.76 mg/dL, S = 1.15 ± 0.13 µs(mg/dL) −1. pHEMA-co-dimethylacrylamide 90:10 R = 8.49–398.88 mg/dL; S = 0.50 ± 0.04 µs(mg/dL) –1; 75:25 R=0-194 mg/dL, S = 0.94 ± 0.10 µs(mg/dL) −1. All formulations tracked rising and falling glucose in vivo while monitoring 2.5 h after implantation.
Conclusion:
The in vitro and in vivo results demonstrate sensor response is highly dependent on material type. Gels with more HEMA have a lower sensitivity, wider analytical range, and respond slower in vivo while gels with less HEMA content respond more quickly and saturate at lower glucose concentrations. This is likely due to the increased gel hydrophilicity and decreased diffusion barrier offered by the added comonomer. Long-term implant monitoring and parallel in vitro stability studies are currently in progress.
Noninvasive Glucose Measurement at the Finger With Mid-Infrared Spectroscopy and Photoacoustic Detection Is Not Influenced by Body Water Content
Hermann v.Lilienfeld-Toal, MD; Alexander Bauer, BA; Otto Hertzberg, BA; Arne Küderle; Miguel Pleitez, PhD; Werner Mäntele, PhD
Elté Sensoric GmbH
Gelnhausen, Hesse, Germany
Objective:
The noninvasive glucose measurement by means of mid-infrared (IR) spectroscopy combined with photoacoustic (PA) detection is based on the quantification of glucose within the epidermis. The composition of the different parts of the epidermis (such as stratum corneum and the stratum spinosum) is probably of critical importance for the result with this approach. Therefore we systematically analyzed the influence of different easily accessible locations with different thickness of stratum corneum for measurement to find an optimal position.
Method:
During oral glucose tolerance tests (OGTTs) in five volunteers we compared blood glucose with noninvasively predicted values based on a mid-IR PA system designed to evaluate the glucose concentration within the epidermis. A test was performed in four different locations on the hand (index finger, thumb, hypothenar, and arm inside). In order to study body water loss, two OGTTs, before and after physical exercise (jogging for 45 min), were performed. The root mean square error of cross validation was calculated as an indicator of the quality of prediction.
Result:
The results of these experiments show the lowest error in the OGTTs performed was at the index finger (16 mg/dL), and the physical exercise did not influence the quality of prediction (18 mg/dL before vs. 16 mg/dL after).
Conclusion:
Our data suggest that regardless of differences in stratum corneum composition, the mid-IR PA gives best results in the optimal perfused parts of the hand. Changes in body water content as it is produced by physical activity do not influence the quality of the results.
A Fuzzy Adaptive Strategy for Blood Glucose Controllers
Josep Vehi, PhD; Aleix Beneyto, BEng
University of Girona
Girona, Girona, Spain
Objective:
A major challenge of any blood glucose control is to tune the controller and to keep it tuned. A strategy to adjust the parameters of the controller to an specific patient is presented. Main parameters of amount of bolus, limits of insulin on board, proportional gain, and basal profile are tuned using the information given by the controller itself and also the postprandial performance.
Method:
A fuzzy inference engine has been implemented. Both internal variables of the controller and postprandial performance indexes are taken as input variables. The output variables are the parameters of the controller. Inference rules have been designed using clinical knowledge.
The adaptive and robust performance of the control algorithm has been intensively assessed in silico on a cohort of virtual patients under challenging realistic scenarios considering mixed meals, circadian variations, time-varying uncertainties, discrete measurement and actuation, sensor errors, and other disturbances.
Result:
From a cohort of 10 virtual patients, starting with a detuned controller (25% underestimation the inulin action, 25% overestimation of the insulin on board, 25% error in the estimation of carbohydrate intake), all the controllers have achieved optimal performance after less than 10 iterations (meals)
Conclusion:
Fuzzy systems are a natural way to include clinical knowledge into blood glucose controllers. In this work, it has been proven that they are also efficient to tune the controller and keep it tuned.
Physician Acceptance of an Innovative Digital Feedback System Depends on Their Perceived Role in Medication Adherence
Naunihal Virdi, MD, MBA; Celine Pering, BA; Katie Peters, BA
Proteus Digital Health
Redwood City, California
Objective:
Medication adherence is a major issue in managing chronic diseases such as diabetes; over half of patients do not take their medications as prescribed. Proteus Digital Health developed a digital feedback system that provides insights on medication taking to help guide treatment plans. The system consists of an ingestible sensor in the form of a pill that sends a signal to a wearable sensor once swallowed. The wearable sensor also measures activity, rest patterns, and heart rate. Several studies have demonstrated accuracy and patient acceptance, but there is a lack of data on physician acceptance.
Method:
Six geographically dispersed primary care physicians were interviewed individually by phone, and six were interviewed in person at a research facility. Each participant was asked for feedback on the Proteus system.
Results:
All 12 physicians viewed medication nonadherence as a major issue. They cited that on average, 30% or more of their patients were not medication adherent. While all but one physician agreed the Proteus system may provide value to their patients, the average “likelihood to pursue” score was 3.25 out of 5. Seven physicians felt medication adherence was at least partly their responsibility and rated the system 4 out of 5. The other five physicians viewed medication adherence as the patient’s responsibility; they rated the Proteus system as 2.5 out of 5.
Conclusion:
Although the physicians agreed nonadherence is a problem, their perceptions regarding responsibility for medication adherence varied. Attitudes toward medication adherence may have influenced their “likelihood to pursue” score.
Use of iPro2 in Real-Life Management of Type 2 Diabetes Patients in India
Mohan Viswanathan, MD, PhD; Sunil M. Jain, MD, DM; Jothydev Kesavadev, MD, FRCP; Manoj Chawla, D.Diabetology; Abhay Mutha, M.D; Vijay Viswanathan, MD, PhD, FRCP; Banshi Saboo, MD; Rajiv Kovil, D.Diabetology; Ambrish Mithal, MD, DM; Dharmen Punatar, D.Diabetology
Dr.Mohans Diabetes Specialties Centre
Chennai, Tamilnadu, India
Objective:
Retrospective continuous glucose monitoring studies (iPro2, Medtronic) may provide health care professionals (HCPs) with better understanding of glycemic patterns in patients with type 2 diabetes and thereby support appropriate therapeutic interventions.
Method:
The study is a 3-month, interventional, postmarket, prospective study of 181 adults aged 19–70 years with baseline A1C values >8–10.0% across India. Patients were scheduled for five visits that included iPro2 evaluations at baseline and month 2 (visits 1 and 3), data review and therapy modifications at visit 2, A1C determinations at baseline and month 3 (visits 1 and 5), and questionnaire completions at visits 1–5. Questionnaires were completed by HCPs to evaluate the utility of iPro2 results and by patients to evaluate understanding of the importance of compliance with HCP recommendations. Glycemic control and variability were to be estimated from iPro2 data.
Result:
iPro2 was found to be safe and was not associated with any serious adverse device effects. Most subjects (91.2%) had one therapy change after the iPro2 data review at visit 2. Mean A1C decreased from 8.6% (baseline) to 8.0% (month 3; P < 0.0001). All HCP and patient questionnaires were consistent with the acceptability and utility of iPro2 studies and results (mean responses 5 on a 7-point Likert scale). The average self-monitoring of blood glucose value decreased (168 to 155 mg/dL), and there were consistent decreases of SG-based glycemic variability metrics (SD 48.3 to 47.7; mean amplitude of glycemic excursion 100.4 to 99.3) between the two iPro2 evaluations.
Conclusion:
iPro2 evaluations provided HCPs with insights and opportunities for initiating changes to treatment regimens and to diet and exercise behaviors, resulting in favorable reductions of A1C and glycemic variability metrics. Patients showed improved knowledge of the importance of therapy compliance.
Artificial Pancreas: A Novel Coping Strategy for Big Dietary Ingestion
Youqing Wang, PhD; Chaofeng Yan, MS
Beijing University of Chemical Technology
Beijing, China
Objective:
The existing AP control algorithms are generally defective when there exist large diet disturbances. The security measures are inclined to guiding the current control action, while this kind of guidance role is different from person to person. Using standard clinical parameters, a model predictive control (MPC)-based algorithm with built-in security mechanism was proposed in this study.
Method:
The concept of insulin on board (IOB), as a term of MPC cost function, is used to design the insulin delivery rate. In other words, the proposed method takes the influence of insulin accumulation into consideration. The set points for IOB are related to subject-specific correction factor and the difference between the current blood glucose (BG) level and the expected BG value. Because correction factor and IOB are involved, the proposed method is named clinical-parameter-based model predictive controller (CPMPC). The weights in the cost function of CPMPC are tuned by using a so-called expected insulin calibration mode, which was proposed by our research group previously.
Result:
CPMPC has been evaluated in silico using the Food and Drug Administration–accepted University of Virginia/Padova simulator. Ten virtual adult subjects followed the same 7-day protocol, where in each day, three meals are at 7:00 A.M., noon, and 6:00 P.M. of 60, 200, and 60 g of carbohydrates, respectively. In the first day, the optimal basal and bolus treatment is applied to each patient. As a reference, CPMPC is compared with the standard MPC from the second to seventh days. CPMPC can evidently decrease the BG index value day by day, from 16.33 on average in the first day down to 5.70 on the last day. All subjects have significant improved closed-loop control performance compared with the open-loop therapy.
Conclusion:
Because of using IOB and expected insulin calibration mode, CPMPC is superior to standard MPC in terms of achieving normoglycemia, excellent robustness to large meal disturbances, avoiding hypoglycemia, and reducing prandial hyperglycemia.
Assessing Diabetic Foot Ulcer Healing at Wound Clinics: Development of a Tracking System Using Support Vector Machine–Based Classification
Lei Wang, PhD Candidate; Peder Pedersen, PhD; Diane Strong, PhD; Bengisu Tulu, PhD; Emmanuel Agu, PhD; Qian He, PhD
Worcester Polytechnic Institute
Worcester, Massachusetts
Objective:
A system for assessing the healing progress of diabetic foot ulcers using machine learning to analyze visual images was developed and tested in a wound clinic. The system detects the wound area and tracks the healing of the foot ulcer over consecutive visits.
Method:
We applied a novel object recognition algorithm for determining the area of diabetic wounds. The system uses a smartphone camera placed on an image capture box to facilitate the image capture of the sole of the foot. A support vector machine–based binary classifier is trained on the bag-of-words histogram representation of local dense scale-invariant feature transform features in each super-pixel. Three experienced wound clinicians manually marked the wound areas on collected wound images to provide labeled samples required for supervised machine learning. We merged these labeled samples into one ground truth corpus. In addition, we improved a new color segmentation algorithm by adopting the bag-of-words idea. Based on area trend and color analysis, we have proposed a healing score to quantitatively assess the wound healing status.
Result:
We have evaluated our system on 40 foot-ulcer images collected over 8 months from 10 patients at the University of Massachusetts Wound Clinic. The Matthews correlation coefficient score of the leave-one-out cross-validation result is nearly 0.7, which is an improvement over previously applied recognition solutions. The healing score is still being evaluated for diagnostic validity.
Conclusion:
Our system provides a promising quantitative method for objective wound assessment. The experimental results indicate that the feature descriptor used in our solution can reflect the unique nature of foot ulcers, and the support vector machine binary classification model is well suited for wound recognition tasks.
Benefit of an Implantable Continuous Glucose Monitoring System for Nocturnal Hypoglycemic Alerts in a 90-Day Feasibility Study
Xiaolin Wang, PhD; Andrew DeHennis, PhD
Senseonics Incorporated Germantown, Maryland
Objective:
The Senseonics continuous glucose monitoring system is composed of a long-term implantable glucose sensor and a wearable transmitter, which wirelessly communicates with a smartphone-based mobile medical application. In order to study the clinical benefit of the nocturnal hypoglycemic alerts of the system, a 90-day study was designed and conducted, and the data were analyzed.
Method:
In this study, 10 subjects with type 1 or type 2 diabetes were enrolled. Throughout the study, the system was unblinded to enable full access to the real-time glucose display. The hypoglycemic alerts were also enabled to help the subjects be aware of the hypoglycemic episodes. The subjects managed their therapeutic regimen as prescribed by their doctor during the study. Self-monitoring of blood glucose (SMBG) measurements were taken by the subjects as per the protocol.
Result:
The data analysis checked the occurrences of nocturnal signal attenuation. A sudden decrease in glucose levels that violates physiological limits on rate of change (the start of the nocturnal signal attenuation) was not seen in this study. The subjects all used the hypoglycemia alarm setting of 70 mg/dL. The data analysis shows that the average percentage of the nights with a hypoglycemic alert being triggered was 18%. After confirmation with SMBG, an average of 92% of the time the subjects showed a recovery into euglycemia within 30 min from the timestamp of SMBG. The true alert rate by using SMBG as references was 75%.
Conclusion:
This study has shown the clinical benefit of using the unblinded Senseonics continuous glucose monitoring system. The nocturnal attenuation is not seen in this study. In addition, the nocturnal hypoglycemic alert helped the subjects to be aware of a hypoglycemic episode.
Effect of Dumpling Cooking Method on Postprandial Glucose Level
Liu Yanping, BA; Zhao Weigang, MD; Wang Zhu, MD; Dong Yinyue, BA; Fuyong, MD
Department of Nutrition
Peking Union Medical College Hospital
Beijing, China
Objective:
The purpose of this study is to evaluate the reliability of using glycemic index alone to select suitable diet for type 2 diabetes. The continuous glucose monitoring (CGM) data analysis method is used to compare two different dumpling cooking methods (fried versus boiled) on postprandial glucose level.
Method:
Ten type 2 diabetes inpatients were enrolled, and their fasting and postprandial glucose levels were controlled below 8 and 10 mmol/L, respectively. All other diet conditions were strictly controlled the same for all subjects, only on days 2 and 4, fried dumpling and boiled dumpling were taken for lunch, respectively. Four-hour postprandial CGM data are used to evaluate glycemic effect of those two dumpling cooking methods. Nine glucose values at different time points (0, 15, 30, 60, 90, 120, and 150 min and 3 and 4 h) and four area under the curve values at different time points (1, 2, 3, and 4 h) are used for data analysis.
Result:
Comparing the CGM data taken from those two dumpling diets, higher glucose values from all nine time points and higher AUC values from all four time windows were observed from the fried dumpling diet. The average time reached to the peak glucose value obtained from the fried dumpling diet is lower compared with the boiled dumpling diet (120 and 150 min, respectively).
Conclusions:
As demonstrated by the different CGM data, the fried dumpling diet raises the glucose level faster than the boiled dumpling diet, even though in theory the glycemic index of the fried food is lower than the boiled food. This study also demonstrates CGM is a useful tool to evaluate the direct impact of different diets on diabetes.
Clinical Accuracy Evaluation of San MediTech’s Real-Time Continuous Glucose Monitoring Product
Huashi Zhang, PhD; Gang Hao, EMBA; Qingfeng Wu, BS; Dan Wang, MS; Jie Li, MS; Jing Huang, BS
San MediTech
Beijing, China
Objective:
The purpose of this study is to evaluate the accuracy of the San MediTech (SMT) real-time continuous glucose monitoring (RT-CGM) device with the comparison of the capillary blood glucose (self-monitoring of blood glucose [SMBG]) reference values.
Method:
Forty-seven subjects (16 with type 1 diabetes and 31 with type 2 diabetes) were enrolled into four different studies from April 2013 to April 2014. The majority of the subjects (38) wore one SMT RT-CGM device for 5 days, and one study (9 subjects) wore two SMT RT-CGM devices on their right and left upper arms to evaluate the reproducibility of the SMT sensor. The SMT RT-CGM device measures the subcutaneous interstitial glucose and was calibrated with the SMBG reference, twice on the first day and one calibration per day for the rest of the 5-day study period. Each day, four to seven capillary glucose values were measured with the glucose meter as the reference glucose values. The total number of the SMBG reference values used for the accuracy study are 1,265 pairs, and 80 pairs (6.3%) are in the hypoglycemic range (<4.4 mmol/L).
Result:
All the subjects finished the 5-day study, and no adverse safety event was observed. The overall mean absolute relative difference of SMT’s RT-CGM device compared with the capillary and venous references values is 14.8%. The overall system agreement percentage within 20/20% SMBG reference values is 78.0%. The Clarke error grid analysis plot of the SMT RT-CGM values versus the SMBG references is also carried out. The combined A zone and B zone percentage is 97.3%. The paired absolute relative difference calculated from the sensor study is 6.1%.
Conclusion:
The SMT RT-CGM device demonstrates safety and effectiveness when compared to the SMBG reference. Good in vivo sensor reproducibility was also observed.
Online Hyper/Hypoglycemia Alert Based on Adaptive Kalman Filter for Type 1 Diabetes
Chunhui Zhao, PhD; Hong Zhao, BEng; Yongji Fu, PhD
Zhejiang University
Hangzhou, Zhejiang, China
Objective:
For online hyper/hypoglycemia alert, measurement noises must be filtered before the signals can be used for real-time display. An improved Kalman filter (KF) method is proposed to denoise continuous glucose monitoring (CGM) signals. To overcome the problem that KF parameters are difficult to define, an estimation method is proposed by using a stochastically based smoothing criterion. Also, it is demonstrated that the filtering performance is not influenced by the initial conditions. The performance of alert is then evaluated by redefining two indices (sensitivity and specificity). Considering that a longer hyper/hypoglycemic event has larger effects on diabetes management, the two evaluation indices are redefined by considering the duration of different glucose statuses.
Method:
The noisy time series are generated for 30 subjects with a 5-min sampling period using the Food and Drug Administration–accepted University of Virginia/University of Padova metabolic simulator. The noisy CGM signals are generated by adding to the blood glucose (noise-free signals) a zero-mean white Gaussian noise sequence with variance 4. The hypoglycemic threshold is set to be 70 mg/dL and the hyperglycemic threshold to be 180 mg/dL. An alert is triggered when the concerned signals are lower than hypoglycemic threshold or higher than hyperglycemic threshold. For the concerned subjects, we use the KF method to reduce the noise and then use the filtered signals to generate hyper/hypoglycemia alert. Also, the performance is compared with that of the noise-free signals and the moving average (MA) filter method.
Result:
The denoised CGM signals by the proposed KF algorithm can give similar performance with that of noise-free CGM. Then the proposed KF method is compared with the MA method. For the hypoglycemia alert, the time lag using KF is only ~1.46 min, which is reduced ~79.3% from that of MA method. The sensitivity, i.e., the percentage of true alerts for all the hyper/hypoglycemic events using KF is ~98.44%, which is higher than that (96.86%) using MA. For the hyperglycemia alert, the time lag of KF is ~3.71 min, which is approximately reduced by 46.5% from that using MA. The sensitivity is also higher using the proposed KF method, which is 96.94% in comparison with 91.67% using MA method. The specificity is not significantly different between the two methods.
Conclusion:
The performance of hyper/hypoglycemia alert using KF is better than using MA. Also, the alert performance using KF is comparable with that using noise-free glucose signals, revealing that the added noises are well filtered.
New Three-Step Clamp Method for the Evaluation of Blood Glucose Meters
Eric Zijlstra, PhD; Lutz Heinemann, PhD; Annelie Fischer; Christoph Kapitza, MD
Profil
Neuss, Germany
Objective:
We compare the performance (in terms of accuracy, precision, and bias) of six CE-certified and commercially available blood glucose (BG) meters using an innovative clinical experimental design with a three-step glucose clamp and frequent capillary sampling.
Method:
Seventeen subjects with type 1 diabetes participated in this open-label, single-center trial. BG was clamped at 60–100–200 mg/dL by variable rate infusions of glucose and insulin. Medical staff performed regular finger pricks (up to 10 per BG level) to obtain capillary blood samples for paired BG meter and YSI reference measurements.
Result:
Accu-Chek Aviva Nano (4.8%), BGStar (4.6%), iBGStar (4.8%), and MyStar Extra (4.1%) displayed significantly lower mean absolute relative deviations than FreeStyle InsuLinx (7.6%) and OneTouch VerioIQ (9.9%). The measurement precision was similar for all meters, but bias was also lower for the Accu-Chek (1.3%), BGStar (−0.9%), iBGStar (1.0%), and MyStar Extra (−0.3%) compared with FreeStyle (−7.1%) and OneTouch (9.0%) meters.
Conclusion:
A new three-step clamp method with frequent capillary sampling was introduced, which provides valuable data for simultaneous investigations of BG meter accuracy, precision, and bias. In this trial, the random error of the tested BG meters is comparable, but a lower systematic error for Accu-Chek, BGStar, iBGStar, and MyStar Extra gives these meters a highly accurate performance at low, normal, and high BG levels.
Duality of Interest:
This investigator-initiated study was supported by Sanofi.
Blood Glucose Meter Performance: A Comparison between Two Test Procedures
Eric Zijlstra, PhD; Annette Baumstark, PhD; Cornelia Haug, MD; Lutz Heinemann, PhD; Guido Freckmann, MD; Christoph Kapitza, MD
Profil
Neuss, Germany
Objective:
We compare the measurement accuracy and bias of the BGStar blood glucose (BG) meter evaluated by a reduced scale standard test protocol and a new three-step clamp method.
Method:
Three BGStar systems were evaluated in two clinical trials. The first used a reduced scale protocol of the ISO 15197:2013 accuracy evaluation, including 35 subjects with type 1 or type 2 diabetes, two measurement series per subject, and no defined distribution of BG concentrations (instead of 100 subjects with one measurement series per subject and with a defined BG distribution). BG concentrations ranged from 56 to 397 mg/dL. The second trial, conducted with the same test materials at a different test center, evaluated meter accuracy and bias using a new three-step (60–100–200 mg/dL) clamp method with frequent capillary sampling (10 samples per clamp level). Nineteen subjects with type 1 diabetes participated in this trial. Reference samples were analyzed using the YSI2300 STAT Plus glucose analyzer at both sites.
Result:
Results are presented for the three BGStar systems compared between the two procedures, i.e., reduced scale standard test protocol versus clamp method. The mean absolute relative deviation was 4.3, 4.6, 5.3 vs. 4.6, 5.4, 4.8%. Results within accuracy limits of ISO 15197:2013 were 99.3, 98.6, 94.3 vs. 100, 99.1, 98.2%. The Clarke error grid zone A results were 100, 100, 98.6 vs. 99.4, 100, 100%. Bias was −2.4, −1.9, 3.7 vs. −0.9, −2.6, 2.9%.
Conclusion:
Similar performance results for the BGStar systems were obtained with the three-step clamp method and with a reduced scale standard test protocol.
Duality of Interest:
This investigator-initiated study was supported by Sanofi.