Faye Cameron

Application of Zone Model Predictive Control Artificial Pancreas During Extended Use of Infusion Set and Sensor: A Randomized Crossover-Controlled Home-Use Trial

OBJECTIVE As artificial pancreas (AP) becomes standard of care, consideration of extended use of insulin infusion sets (IIS) and continuous glucose monitors (CGMs) becomes vital. We conducted an outpatient randomized crossover study to test the safety and efficacy of a zone model predictive control (zone-MPC)–based AP system versus sensor augmented pump (SAP) therapy in which IIS and CGM failures were provoked via extended wear to 7 and 21 days, respectively. RESEARCH DESIGN AND METHODS A smartphone-based AP system was used by 19 adults (median age 23 years [IQR 10], mean 8.0 ± 1.7% HbA1c) over 2 weeks and compared with SAP therapy for 2 weeks in a crossover, unblinded outpatient study with remote monitoring in both study arms. RESULTS AP improved percent time 70–140 mg/dL (48.1 vs. 39.2%; P = 0.016) and time 70–180 mg/dL (71.6 vs. 65.2%; P = 0.008) and decreased median glucose (141 vs. 153 mg/dL; P = 0.036) and glycemic variability (SD 52 vs. 55 mg/dL; P = 0.044) while decreasing percent time <70 mg/dL (1.3 vs. 2.7%; P = 0.001). AP also improved overnight control, as measured by mean glucose at 0600 h (140 vs. 158 mg/dL; P = 0.02). IIS failures (1.26 ± 1.44 vs. 0.78 ± 0.78 events; P = 0.13) and sensor failures (0.84 ± 0.6 vs. 1.1 ± 0.73 events; P = 0.25) were similar between AP and SAP arms. Higher percent time in closed loop was associated with better glycemic outcomes. CONCLUSIONS Zone-MPC significantly and safely improved glycemic control in a home-use environment despite prolonged CGM and IIS wear. This project represents the first home-use AP study attempting to provoke and detect component failure while successfully maintaining safety and effective glucose control.

Predictive Hyperglycemia and Hypoglycemia Minimization: In-Home Evaluation of Safety, Feasibility, and Efficacy in Overnight Glucose Control in Type 1 Diabetes

OBJECTIVE The objective of this study was to determine the safety, feasibility, and efficacy of a predictive hyperglycemia and hypoglycemia minimization (PHHM) system compared with predictive low-glucose insulin suspension (PLGS) alone in overnight glucose control. RESEARCH DESIGN AND METHODS A 42-night trial was conducted in 30 individuals with type 1 diabetes in the age range 15–45 years. Participants were randomly assigned each night to either PHHM or PLGS and were blinded to the assignment. The system suspended the insulin pump on both the PHHM and PLGS nights for predicted hypoglycemia but delivered correction boluses for predicted hyperglycemia on PHHM nights only. The primary outcome was the percentage of time spent in a sensor glucose range of 70–180 mg/dL during the overnight period. RESULTS The addition of automated insulin delivery with PHHM increased the time spent in the target range (70–180 mg/dL) from 71 ± 10% during PLGS nights to 78 ± 10% during PHHM nights (P < 0.001). The average morning blood glucose concentration improved from 163 ± 23 mg/dL after PLGS nights to 142 ± 18 mg/dL after PHHM nights (P < 0.001). Various sensor-measured hypoglycemic outcomes were similar on PLGS and PHHM nights. All participants completed 42 nights with no episodes of severe hypoglycemia, diabetic ketoacidosis, or other study- or device-related adverse events. CONCLUSIONS The addition of a predictive hyperglycemia minimization component to our existing PLGS system was shown to be safe, feasible, and effective in overnight glucose control.

Model-based falsification of an artificial pancreas control system

We present a model-based falsification scheme for artificial pancreas controllers. Our approach performs a closed-loop simulation of the control software using models of the human insulin-glucose regulatory system. Our work focuses on testing properties of an overnight control system for hypoglycemia/hyperglycemia minimization in patients with type-1 diabetes. This control system is currently the subject of extensive phase II clinical trials. We describe how the overall closed loop simulator is constructed, and formulate properties to be tested. Significantly, the closed loop simulation incorporates the control software, as is, without any abstractions. Next, we demonstrate the use of a simulation-based falsification approach to find potential property violations in the resulting control system. We formulate a series of properties about the controller behavior and examine the violations obtained. Using these violations, we propose modifications to the controller software to improve its performance under these adverse (corner-case) scenarios. We also illustrate the effectiveness of robustness as a metric for identifying interesting property violations. Finally, we identify important open problems for future work.

Predictive hyperglycemia and hypoglycemia minimization: In-home double-blind randomized controlled evaluation in children and young adolescents

The primary objective of this trial was to evaluate the feasibility, safety, and efficacy of a predictive hyperglycemia and hypoglycemia minimization (PHHM) system vs predictive low glucose suspension (PLGS) alone in optimizing overnight glucose control in children 6 to 14 years old.

Continuous Glucose Monitoring Enables the Detection of Losses in Infusion Set Actuation (LISAs)

Reliable continuous glucose monitoring (CGM) enables a variety of advanced technology for the treatment of type 1 diabetes. In addition to artificial pancreas algorithms that use CGM to automate continuous subcutaneous insulin infusion (CSII), CGM can also inform fault detection algorithms that alert patients to problems in CGM or CSII. Losses in infusion set actuation (LISAs) can adversely affect clinical outcomes, resulting in hyperglycemia due to impaired insulin delivery. Prolonged hyperglycemia may lead to diabetic ketoacidosis—a serious metabolic complication in type 1 diabetes. Therefore, an algorithm for the detection of LISAs based on CGM and CSII signals was developed to improve patient safety. The LISA detection algorithm is trained retrospectively on data from 62 infusion set insertions from 20 patients. The algorithm collects glucose and insulin data, and computes relevant fault metrics over two different sliding windows; an alarm sounds when these fault metrics are exceeded. With the chosen algorithm parameters, the LISA detection strategy achieved a sensitivity of 71.8% and issued 0.28 false positives per day on the training data. Validation on two independent data sets confirmed that similar performance is seen on data that was not used for training. The developed algorithm is able to effectively alert patients to possible infusion set failures in open-loop scenarios, with limited evidence of its extension to closed-loop scenarios.

Sensor-based detection and estimation of meal carbohydrates for people with diabetes

Abstract People with type 1 diabetes (T1D) must estimate the carbohydrate (CHO) content in meals to compute the bolus insulin correctly. To release T1D patients from the cumbersome task of counting CHO, we develop a method for detecting meals that can be used in blood glucose (BG) control. The algorithm detects a meal and estimates the meal onset and the amount of CHO. The inputs of the meal detector are the continuous glucose monitoring (CGM) data and the insulin infusion rate. We use second-order linear input-output models for insulin to subcutaneous glucose dynamics and for CHO to subcutaneous glucose dynamics. The models are converted to a linear discrete-time state-space model. A white noise double integrator models the unknown meal disturbances. The state-space model is augmented with the unknown meal disturbance (CHO ingestion rate) and a Kalman filter (KF) estimates the CHO rate (g/min). The algorithm uses two tests to announce a meal. The first test is a cumulative sum algorithm that detects changes in the KF innovation and estimates the onset of change. The second test is comparison of the estimated CHO rate with a threshold to detect a change in the rate. If both tests simultaneously detect a change, an optimal smoother estimates the meal-size. If the estimated meal-size reaches a certain amount, the algorithm announces a meal. Furthermore, we integrate a bolus calculator (BC) with the meal detector. We test the algorithm for nine virtual T1D patients. In total, the patients eat 45 meals in 13.5 days. The detection sensitivity is 93% and the detection delay has a median of 40 min. The median of the meal onset estimation bias is 5 min. Out of 42 detected meals, the algorithm underestimates 26 meals with a median bias of −19 g, and it overestimates 16 meals with a median bias of 21 g. The meal detector with the BC reduces the BG postprandial peak from 274 mg/dL (unbolused meals) to 207 mg/dL, and it increases the mean time in euglycemia from 50% to 79%. The meal detector combined with the BC improves glycemia for the virtual patients in this study.

Erratum. Application of Zone Model Predictive Control Artificial Pancreas During Extended Use of Infusion Set and Sensor: A Randomized Crossover-Controlled Home-Use Trial. Diabetes Care 2017;40:1096–1102

In the above-mentioned article, the supplementary data had the incorrect figure listed for …

Algorithms for a Single Hormone Closed-Loop Artificial Pancreas: Challenges Pertinent to Chemical Process Operations and Control

The development of a closed-loop artificial pancreas to regulate the blood glucose concentration of individuals with type 1 diabetes has been a focused area of research for over 50 years, with rapid progress during the past decade. The daily control challenges faced by someone with type 1 diabetes include asymmetric objectives and risks, and one-sided manipulated input action with frequent relatively fast disturbances. The major automation steps toward a closed-loop artificial pancreas include (i) monitoring and overnight alarms for hypoglycemia (low blood glucose); (ii) overnight low glucose suspend (LGS) systems to prevent hypoglycemia; and (iii) fully closed-loop systems that adjust insulin (and perhaps glucagon) to maintain desired blood glucose levels day and night. We focus on the steps that we used to develop and test a probabilistic, risk-based, model predictive control strategy for a fully closed-loop artificial pancreas. We complete the paper by discussing ramifications of lessons learned for chemical process systems applications.