Artificial pancreas: model predictive control design from clinical experience

A strategy based on integer programming for optimal dosing and timing of preventive hypoglycemic treatments in type 1 diabetes management

BACKGROUND AND OBJECTIVES

One of the major problems related to type 1 diabetes (T1D) management is hypoglycemia, a condition characterized by low blood glucose levels and responsible for reduced quality of life and increased mortality. Fast-acting carbohydrates, also known as hypoglycemic treatments (HT), can counteract this event. In the literature, dosage and timing of HT are usually based on heuristic rules. In the present work, we propose an algorithm for mitigating hypoglycemia by suggesting preventive HT consumption, with dosages and timing determined by solving an optimization problem.

METHODS

By leveraging integer programming and linear inequality constraints, the algorithm can bind the amount of suggested carbohydrates to standardized quantities (i.e., those available in "off-the-shelf" HT) and the minimal distance between consecutive suggestions (to reduce the nuisance for patients).

RESULTS

The proposed method was tested in silico and compared with competitor algorithms using the UVa/Padova T1D simulator. At the cost of a slight increase of HT consumed per day, the proposed algorithm produces the lowest median and interquartile range of the time spent in hypoglycemia, with a statistically significant improvement over most competitor algorithms. Also, the average number of hypoglycemic events per day is reduced to 0 in median.

CONCLUSIONS

Thanks to its positive performances and reduced computational burden, the proposed algorithm could be a candidate tool for integration in a DSS aimed at improving T1D management.

The importance of interpreting machine learning models for blood glucose prediction in diabetes: an analysis using SHAP

Machine learning has become a popular tool for learning models of complex dynamics from biomedical data. In Type 1 Diabetes (T1D) management, these models are increasingly been integrated in decision support systems (DSS) to forecast glucose levels and provide preventive therapeutic suggestions, like corrective insulin boluses (CIB), accordingly. Typically, models are chosen based on their prediction accuracy. However, since patient safety is a concern in this application, the algorithm should also be physiologically sound and its outcome should be explainable. This paper aims to discuss the importance of using tools to interpret the output of black-box models in T1D management by presenting a case-of-study on the selection of the best prediction algorithm to integrate in a DSS for CIB suggestion. By retrospectively "replaying" real patient data, we show that two long-short term memory neural networks (LSTM) (named p-LSTM and np-LSTM) with similar prediction accuracy could lead to different therapeutic decisions. An analysis with SHAP-a tool for explaining black-box models' output-unambiguously shows that only p-LSTM learnt the physiological relationship between inputs and glucose prediction, and should therefore be preferred. This is verified by showing that, when embedded in the DSS, only p-LSTM can improve patients' glycemic control.

An Adaptive Model Predictive Controller to Address the Biovariability in Blood Clotting Response During Therapy With Warfarin

OBJECTIVE

Effective dosing of anticoagulants aims to prevent blood clot formation while avoiding hemorrhages. This complex task is challenged by several disturbing factors and drug-effect uncertainties, requesting frequent monitoring and adjustment. Biovariability in drug absorption and action further complicates titration and calls for individualized strategies. In this paper, we propose an adaptive closed-loop control algorithm to assist in warfarin therapy management.

METHODS

The controller was designed and tested in silico using an established pharmacometrics model of warfarin, which accounts for inter-subject variability. The control algorithm is an adaptive Model Predictive Control (a-MPC) that leverages a simplified patient model, whose parameters are updated with a Bayesian strategy. Performance was quantitatively evaluated in simulations performed on a population of virtual subjects against an algorithm reproducing medical guidelines (MG) and an MPC controller available in the literature (l-MPC).

RESULTS

The proposed a-MPC significantly (p 0.05) lowers rising time (2.8 vs. 4.4 and 11.2 days) and time out of range (3.3 vs. 7.2 and 12.9 days) with respect to both MG and l-MPC, respectively. Adaptivity grants a significantly (p 0.05) lower number of subjects reaching unsafe INR values compared to when this feature is not present (8.9% vs.15% of subjects presenting an overshoot outside the target range and 0.08% vs. 0.28% of subjects reaching dangerous INR values).

CONCLUSION

The a-MPC algorithm improve warfarin therapy compared to the benchmark therapies.

SIGNIFICANCE

This in-silico validation proves effectiveness of the a-MPC algorithm for anticoagulant administration, paving the way for clinical testing.

Uncovering personalized glucose responses and circadian rhythms from multiple wearable biosensors with Bayesian dynamical modeling

Wearable biosensors and smartphone applications can measure physiological variables over multiple days in free-living conditions. We measure food and drink ingestion, glucose dynamics, physical activity, heart rate (HR), and heart rate variability (HRV) in 25 healthy participants over 14 days. We develop a Bayesian inference framework to learn personal parameters that quantify circadian rhythms and physiological responses to external stressors. Modeling the effects of ingestion events on glucose levels reveals that slower glucose decay kinetics elicit larger postprandial glucose spikes, and we uncover a circadian baseline rhythm for glucose with high amplitudes in some individuals. Physical activity and circadian rhythms explain as much as 40%-65% of the HR variance, whereas the variance explained for HRV is more heterogeneous across individuals. A more complex model incorporating activity, HR, and HRV explains up to 15% of additional glucose variability, highlighting the relevance of integrating multiple biosensors to better predict glucose dynamics.

Synthetic biomedical data generation in support of In Silico Clinical Trials

Living in the era of Big Data, one may advocate that the additional synthetic generation of data is redundant. However, to be able to truly say whether it is valid or not, one needs to focus more on the meaning and quality of data than on the quantity. In some domains, such as biomedical and translational sciences, data privacy still holds a higher importance than data sharing. This by default limits access to valuable research data. Intensive discussion, agreements, and conventions among different medical research players, as well as effective techniques and regulations for data anonymization, already made a big step toward simplification of data sharing. However, the situation with the availability of data about rare diseases or outcomes of novel treatments still requires costly and risky clinical trials and, thus, would greatly benefit from smart data generation. Clinical trials and tests on animals initiate a cyclic procedure that may involve multiple redesigns and retesting, which typically takes two or three years for medical devices and up to eight years for novel medicines, and costs between 10 and 20 million euros. The US Food and Drug Administration (FDA) acknowledges that for many novel devices, practical limitations require alternative approaches, such as computer modeling and engineering tests, to conduct large, randomized studies. In this article, we give an overview of global initiatives advocating for computer simulations in support of the 3R principles (Replacement, Reduction, and Refinement) in humane experimentation. We also present several research works that have developed methodologies of smart and comprehensive generation of synthetic biomedical data, such as virtual cohorts of patients, in support of In Silico Clinical Trials (ISCT) and discuss their common ground.

Development platform for artificial pancreas algorithms

BACKGROUND AND AIMS

Assessing algorithms of artificial pancreas systems is critical in developing automated and fault-tolerant solutions that work outside clinical settings. The development and evaluation of algorithms can be facilitated with a platform that conducts virtual clinical trials. We present in this paper a clinically validated cloud-based distributed platform that supports the development and comprehensive testing of single and dual-hormone algorithms for type 1 diabetes mellitus (T1DM).

METHODS

The platform is built on principles of object-oriented design and runs user algorithms in real-time virtual clinical trials utilizing a multi-threaded environment enabled by concurrent execution over a cloud infrastructure. The platform architecture isolates user algorithms located on personal machines from proprietary patient data running on the cloud. Users import a plugin into their algorithms (Matlab, Python, or Java) to connect to the platform. Once connected, users interact with a graphical interface to design experimental protocols for their trials. Protocols include trial duration in days, mealtimes and amounts, variability in mealtimes and amounts, carbohydrate counting errors, snacks, and onboard insulin levels.

RESULTS

The platform facilitates development by solving the ODE model in the cloud on large CPU-optimized machines, providing a 62% improvement in memory, speed and CPU utilization. Users can easily debug & modify code, test multiple strategies, and generate detailed clinical performance reports. We validated and integrated into the platform a glucoregulatory system of ordinary differential equations (ODEs) parameterized with clinical data to mimic the inter and intra-day variability of glucose responses of 15 T1DM patients.

CONCLUSION

The platform utilizes the validated patient model to conduct virtual clinical trials for the rapid development and testing of closed-loop algorithms for T1DM.

Blood glucose concentration control for type 1 diabetic patients: a multiple-model strategy

In this study, a multiple-model strategy is evaluated as an alternative closed-loop method for subcutaneous insulin delivery in type 1 diabetes. Non-linearities of the glucose-insulin regulatory system are considered by modelling the system around five different operating points. After conducting some identification experiments in the UVA/Padova metabolic simulator (accepted simulator by the US Food and Drug Administration (FDA)), five transfer functions are obtained for these operating points. Paying attention to some physiological facts, the control objectives such as the required settling time and permissible bounds of overshoots and undershoots are determined for any transfer functions. Then, five PID controllers are tuned to achieve these objectives and a bank of controllers is constructed. To cope with difficulties of the presence of delays in subcutaneous blood glucose (BG) measuring and in administration of insulin, a glucose-dependent setpoint is considered as the desired trajectory for the BG concentration. The performance of the obtained closed-loop glucose-insulin regulatory system is investigated on the in silico adult cohort of the UVA/Padova metabolic simulator. The obtained results show that the proposed multiple-model strategy leads to a closed-loop mechanism with limited hyperglycemia and no severe hypoglycemia.

Preventing undesirable behavior of intelligent machines

Intelligent machines using machine learning algorithms are ubiquitous, ranging from simple data analysis and pattern recognition tools to complex systems that achieve superhuman performance on various tasks. Ensuring that they do not exhibit undesirable behavior-that they do not, for example, cause harm to humans-is therefore a pressing problem. We propose a general and flexible framework for designing machine learning algorithms. This framework simplifies the problem of specifying and regulating undesirable behavior. To show the viability of this framework, we used it to create machine learning algorithms that precluded the dangerous behavior caused by standard machine learning algorithms in our experiments. Our framework for designing machine learning algorithms simplifies the safe and responsible application of machine learning.

A Technological Revolution: The Integration of New Treatments to Manage Type 1 Diabetes

Intensive insulin treatment and frequent self-monitoring of blood glucose (SMBG) have been recognized as pillars of diabetes treatment. Many patients with type 1 diabetes (T1D) struggle to achieve targeted glycemic control. Technology has vastly changed how these tenets to treatment can occur. Continuous subcutaneous insulin infusion (CSII) pumps and continuous glucose monitoring (CGM) can be used in place of their counterparts, multiple daily injections and SMBG. We present a review of CSII, CGM, and of different levels of integration among these two therapies, ranging from low glucose suspension devices to hybrid closed loop insulin delivery. Analysis of the various tools, their effect on glycemic control, and a guide to integrate them into pediatric clinical practice is presented. Although a cure for T1D remains the ultimate goal, technology holds the promise of keeping youth with T1D in targeted control and minimize the burden of this chronic medical condition. [Pediatr Ann. 2019;48(8):e311-e318.].

Model individualization for artificial pancreas

BACKGROUND AND OBJECTIVE

The inter-subject variability characterizing the patients affected by type 1 diabetes mellitus makes automatic blood glucose control very challenging. Different patients have different insulin responses, and a control law based on a non-individualized model could be ineffective. The definition of an individualized control law in the context of artificial pancreas is currently an open research topic. In this work we consider two novel identification approaches that can be used for individualizing linear glucose-insulin models to a specific patient.

METHODS

The first approach belongs to the class of black-box identification and is based on a novel kernel-based nonparametric approach, whereas the second is a gray-box identification technique which relies on a constrained optimization and requires to postulate a model structure as prior knowledge. The latter is derived from the linearization of the average nonlinear adult virtual patient of the UVA/Padova simulator. Model identification and validation are based on in silico data collected during simulations of clinical protocols designed to produce a sufficient signal excitation without compromising patient safety. The identified models are evaluated in terms of prediction performance by means of the coefficient of determination, fit, positive and negative max errors, and root mean square error.

RESULTS

Both identification approaches were used to identify a linear individualized glucose-insulin model for each adult virtual patient of the UVA/Padova simulator. The resulting model simulation performance is significantly improved with respect to the performance achieved by a linear average model.

CONCLUSIONS

The approaches proposed in this work have shown a good potential to identify glucose-insulin models for designing individualized control laws for artificial pancreas.

Considering Plasma Insulin Concentrations in Adaptive Model Predictive Control for Artificial Pancreas Systems

Adaptive and personalized model predictive control (MPC) algorithms that explicitly consider insulin dosing constraints are necessary to prevent overdose-induced hypoglycemia in type 1 diabetes. A personalized plasma insulin concentration (PIC) estimator is integrated with adaptive models to characterize the temporal dynamics of PIC and blood glucose concentration (BGC). The dynamic profile trajectories of the PIC and BGC are used to adapt the control problems and constraints on-line to accurately reflect the concurrent metabolic state of the individuals and improve glucose regulation. The adaptive MPC algorithm, explicitly considering the PIC in the insulin dose computations, is demonstrated to effectively control BGC without any manual user input on meal information while avoiding excessive insulin administration that can cause hypoglycemia. Simulation results demonstrate the ability of the proposed approach with an average 71.14 % of time spent in the target range (BGC ϵ [70, 180]mg/dL).

Computer Aided Clinical Trials for Implantaule Cardiac Devices

In this paper we aim to answer the question, "How can modeling and simulation of physiological systems be used to evaluate life-critical implantable medical devices?" Clinical trials for medical devices are becoming increasingly inefficient as they take several years to conduct, at very high cost and suffer from high rates of failure. For example, the Rhythm ID Goes Head-to-head Trial (RIGHT) sought to evaluate the performance of two arrhythmia discriminator algorithms for implantable cardioverter defibrillators, Vitality 2 vs. Medtronic, in terms of time-to-first inappropriate therapy, but concluded with results contrary to the initial hypothesis- after 5 years, 2,000+ patients and at considerble ethical and monetary cost. In this paper, we describe the design and performance of a Computer-aided Clinical Trial (CACT) for Implantable Cardiac Devices where previous trial information, real patient data and closed-loop device models are effectively used to evaluate the trial with high confidence. We formulate the CACT in the context of RIGHT using a Bayesian statistical framework. We define a hierarchical model of the virtual cohort generated from a physiological model which captures the uncertainty in the parameters and allow for the systematic incorporation of information available at the design of the trial. With this formulation, the estimates the inappropriate therapy rate of Vitality 2 compared to Medtronic as 33.22% vs 15.62% $(\mathrm{p}\lt 0.001)$, which is comparable to the original trial. Finally, we relate the outcomes of the computer- aided clinical trial to the primary endpoint of RIGHT.

Adaptive and Personalized Plasma Insulin Concentration Estimation for Artificial Pancreas Systems

BACKGROUND

The artificial pancreas (AP) system, a technology that automatically administers exogenous insulin in people with type 1 diabetes mellitus (T1DM) to regulate their blood glucose concentrations, necessitates the estimation of the amount of active insulin already present in the body to avoid overdosing.

METHOD

An adaptive and personalized plasma insulin concentration (PIC) estimator is designed in this work to accurately quantify the insulin present in the bloodstream. The proposed PIC estimation approach incorporates Hovorka's glucose-insulin model with the unscented Kalman filtering algorithm. Methods for the personalized initialization of the time-varying model parameters to individual patients for improved estimator convergence are developed. Data from 20 three-days-long closed-loop clinical experiments conducted involving subjects with T1DM are used to evaluate the proposed PIC estimation approach.

RESULTS

The proposed methods are applied to the clinical data containing significant disturbances, such as unannounced meals and exercise, and the results demonstrate the accurate real-time estimation of the PIC with the root mean square error of 7.15 and 9.25 mU/L for the optimization-based fitted parameters and partial least squares regression-based testing parameters, respectively.

CONCLUSIONS

The accurate real-time estimation of PIC will benefit the AP systems by preventing overdelivery of insulin when significant insulin is present in the bloodstream.

Accuracy of a CGM Sensor in Pediatric Subjects With Type 1 Diabetes. Comparison of Three Insertion Sites: Arm, Abdomen, and Gluteus

BACKGROUND

Patients with diabetes, especially pediatric ones, sometimes use continuous glucose monitoring (CGM) sensor in different positions from the approved ones. Here we compare the accuracy of Dexcom® G5 CGM sensor in three different sites: abdomen, gluteus (both approved) and arm (off-label).

METHOD

Thirty youths, 5-9 years old, with type 1 diabetes (T1D) wore the sensor during a clinical trial where frequent self-monitoring of blood glucose (SMBG) measurements were obtained. Sensor was inserted in different sites according to the patient habit. Accuracy metrics include absolute relative difference (ARD) and absolute difference (AD) of CGM with respect to SMBG. The three sites were compared with ANOVA. If the test detected a difference, an additional pair-wise comparison was performed.

RESULTS

Overall, no accuracy difference was detected: the mean ARD was 13.3% (SD = 13.5%) for abdomen, 13.4% (12.9%) for arm and 12.9% (20.2%) for gluteus ( P value = .83); the mean AD was 17.0 mg/dl (17.2 mg/dl) for abdomen, 17.2 mg/dl (17.1 mg/dl) for arm and 18.3 mg/dl (18.5 mg/dl) for gluteus ( P value = .30). In hypo- and euglycemia ARD ( P value = .87 and .15, respectively), and AD ( P value = .68 and .37, respectively) were not statistically different. At variance, in hyperglycemia, a significant difference was detected between the two approved sites, abdomen and gluteus (ΔARD = -2.2% [CI = -4.2%, -0.1%], P value = .04), whereas the comparisons with the off-label location, arm-abdomen, and arm-gluteus were not significant.

CONCLUSIONS

These results suggest that the accuracy of the sensor placed on the arm was not significantly different with respect to the two approved insertion sites (abdomen and gluteus). Larger, randomized trials are needed to draw final conclusions.

Fractional calculus in pharmacokinetics

We are witnessing the birth of a new variety of pharmacokinetics where non-integer-order differential equations are employed to study the time course of drugs in the body: this is dubbed "fractional pharmacokinetics". The presence of fractional kinetics has important clinical implications such as the lack of a half-life, observed, for example with the drug amiodarone and the associated irregular accumulation patterns following constant and multiple-dose administration. Building models that accurately reflect this behaviour is essential for the design of less toxic and more effective drug administration protocols and devices. This article introduces the readers to the theory of fractional pharmacokinetics and the research challenges that arise. After a short introduction to the concepts of fractional calculus, and the main applications that have appeared in literature up to date, we address two important aspects. First, numerical methods that allow us to simulate fractional order systems accurately and second, optimal control methodologies that can be used to design dosing regimens to individuals and populations.

Prediabetes in youth - mechanisms and biomarkers

Obesity has been estimated to decrease life expectancy by as little as 0.8 to as much as 7 years being the second leading cause of preventable death in the United States after smoking. Along with the increase in the prevalence of obesity, there has been a dramatic rise of the prevalence of prediabetes and type 2 diabetes among adolescents. Despite that, very little is known about the pathogenesis of these conditions in pediatrics and about how we could detect prediabetes in an early stage in order to prevent full blown diabetes. In this review we summarize the current knowledge on the pathophysiology of prediabetes and type 2 diabetes in adolescents and describe how biomarkers of beta-cell function might help identifying those individuals who are prone to progress from normal glucose tolerance towards prediabetes and overt type 2 diabetes. To better understand and fight this disease, we will need to explore and develop novel therapeutic strategies and individuate more sensitive and specific biomarkers that can allow an earlier detection of the disease.

Event-Triggered Model Predictive Control for Embedded Artificial Pancreas Systems

OBJECTIVE

The development of artificial pancreas (AP) technology for deployment in low-energy, embedded devices is contingent upon selecting an efficient control algorithm for regulating glucose in people with type 1 diabetes mellitus. In this paper, we aim to lower the energy consumption of the AP by reducing controller updates, that is, the number of times the decision-making algorithm is invoked to compute an appropriate insulin dose.

METHODS

Physiological insights into glucose management are leveraged to design an event-triggered model predictive controller (MPC) that operates efficiently, without compromising patient safety. The proposed event-triggered MPC is deployed on a wearable platform. Its robustness to latent hypoglycemia, model mismatch, and meal misinformation is tested, with and without meal announcement, on the full version of the US-FDA accepted UVA/Padova metabolic simulator.

RESULTS

The event-based controller remains on for 18 h of 41 h in closed loop with unannounced meals, while maintaining glucose in 70-180 mg/dL for 25 h, compared to 27 h for a standard MPC controller. With meal announcement, the time in 70-180 mg/dL is almost identical, with the controller operating a mere 25.88% of the time in comparison with a standard MPC.

CONCLUSION

A novel control architecture for AP systems enables safe glycemic regulation with reduced processor computations.

SIGNIFICANCE

Our proposed framework integrated seamlessly with a wide variety of popular MPC variants reported in AP research, customizes tradeoff between glycemic regulation and efficacy according to prior design specifications, and eliminates judicious prior selection of controller sampling times.

Parental evaluation of a telemonitoring service for children with Type 1 Diabetes

Introduction In the past years, we developed a telemonitoring service for young patients affected by Type 1 Diabetes. The service provides data to the clinical staff and offers an important tool to the parents, that are able to oversee in real time their children. The aim of this work was to analyze the parents' perceived usefulness of the service. Methods The service was tested by the parents of 31 children enrolled in a seven-day clinical trial during a summer camp. To study the parents' perception we proposed and analyzed two questionnaires. A baseline questionnaire focused on the daily management and implications of their children's diabetes, while a post-study one measured the perceived benefits of telemonitoring. Questionnaires also included free text comment spaces. Results Analysis of the baseline questionnaires underlined the parents' suffering and fatigue: 51% of total responses showed a negative tendency and the mean value of the perceived quality of life was 64.13 in a 0-100 scale. In the post-study questionnaires about half of the parents believed in a possible improvement adopting telemonitoring. Moreover, the foreseen improvement in quality of life was significant, increasing from 64.13 to 78.39 ( p-value = 0.0001). The analysis of free text comments highlighted an improvement in mood, and parents' commitment was also proved by their willingness to pay for the service (median = 200 euro/year). Discussion A high number of parents appreciated the telemonitoring service and were confident that it could improve communication with physicians as well as the family's own peace of mind.

Predictive Control of the Blood Glucose Level in Type I Diabetic Patient Using Delay Differential Equation Wang Model

Because of increasing risk of diabetes, the measurement along with control of blood sugar has been of great importance in recent decades. In type I diabetes, because of the lack of insulin secretion, the cells cannot absorb glucose leading to low level of glucose. To control blood glucose (BG), the insulin must be injected to the body. This paper proposes a method for BG level regulation in type I diabetes. The control strategy is based on nonlinear model predictive control. The aim of the proposed controller optimized with genetics algorithms is to measure BG level each time and predict it for the next time interval. This merit causes a less amount of control effort, which is the rate of insulin delivered to the patient body. Consequently, this method can decrease the risk of hypoglycemia, a lethal phenomenon in regulating BG level in diabetes caused by a low BG level. Two delay differential equation models, namely Wang model and Enhanced Wang model, are applied as controller model and plant, respectively. The simulation results exhibit an acceptable performance of the proposed controller in meal disturbance rejection and robustness against parameter changes. As a result, if the nutrition of the person decreases instantly, the hypoglycemia will not happen. Furthermore, comparing this method with other works, it was shown that the new method outperforms previous studies.

Randomized Summer Camp Crossover Trial in 5- to 9-Year-Old Children: Outpatient Wearable Artificial Pancreas Is Feasible and Safe

OBJECTIVE

The Pediatric Artificial Pancreas (PedArPan) project tested a children-specific version of the modular model predictive control (MMPC) algorithm in 5- to 9-year-old children during a camp.

RESEARCH DESIGN AND METHODS

A total of 30 children, 5- to 9-years old, with type 1 diabetes completed an outpatient, open-label, randomized, crossover trial. Three days with an artificial pancreas (AP) were compared with three days of parent-managed sensor-augmented pump (SAP).

RESULTS

Overnight time-in-hypoglycemia was reduced with the AP versus SAP, median (25(th)-75(th) percentiles): 0.0% (0.0-2.2) vs. 2.2% (0.0-12.3) (P = 0.002), without a significant change of time-in-target, mean: 56.0% (SD 22.5) vs. 59.7% (21.2) (P = 0.430), but with increased mean glucose 173 mg/dL (36) vs. 150 mg/dL (39) (P = 0.002). Overall, the AP granted a threefold reduction of time-in-hypoglycemia (P < 0.001) at the cost of decreased time-in-target, 56.8% (13.5) vs. 63.1% (11.0) (P = 0.022) and increased mean glucose 169 mg/dL (23) vs. 147 mg/dL (23) (P < 0.001).

CONCLUSIONS

This trial, the first outpatient single-hormone AP trial in a population of this age, shows feasibility and safety of MMPC in young children. Algorithm retuning will be performed to improve efficacy.

Day-and-Night Closed-Loop Glucose Control in Patients With Type 1 Diabetes Under Free-Living Conditions: Results of a Single-Arm 1-Month Experience Compared With a Previously Reported Feasibility Study of Evening and Night at Home

OBJECTIVE

After testing of a wearable artificial pancreas (AP) during evening and night (E/N-AP) under free-living conditions in patients with type 1 diabetes (T1D), we investigated AP during day and night (D/N-AP) for 1 month.

RESEARCH DESIGN AND METHODS

Twenty adult patients with T1D who completed a previous randomized crossover study comparing 2-month E/N-AP versus 2-month sensor augmented pump (SAP) volunteered for 1-month D/N-AP nonrandomized extension. AP was executed by a model predictive control algorithm run by a modified smartphone wirelessly connected to a continuous glucose monitor (CGM) and insulin pump. CGM data were analyzed by intention-to-treat with percentage time-in-target (3.9-10 mmol/L) over 24 h as the primary end point.

RESULTS

Time-in-target (mean ± SD, %) was similar over 24 h with D/N-AP versus E/N-AP: 64.7 ± 7.6 vs. 63.6 ± 9.9 (P = 0.79), and both were higher than with SAP: 59.7 ± 9.6 (P = 0.01 and P = 0.06, respectively). Time below 3.9 mmol/L was similarly and significantly reduced by D/N-AP and E/N-AP versus SAP (both P < 0.001). SD of blood glucose concentration (mmol/L) was lower with D/N-AP versus E/N-AP during whole daytime: 3.2 ± 0.6 vs. 3.4 ± 0.7 (P = 0.003), morning: 2.7 ± 0.5 vs. 3.1 ± 0.5 (P = 0.02), and afternoon: 3.3 ± 0.6 vs. 3.5 ± 0.8 (P = 0.07), and was lower with D/N-AP versus SAP over 24 h: 3.1 ± 0.5 vs. 3.3 ± 0.6 (P = 0.049). Insulin delivery (IU) over 24 h was higher with D/N-AP and SAP than with E/N-AP: 40.6 ± 15.5 and 42.3 ± 15.5 vs. 36.6 ± 11.6 (P = 0.03 and P = 0.0004, respectively).

CONCLUSIONS

D/N-AP and E/N-AP both achieved better glucose control than SAP under free-living conditions. Although time in the different glycemic ranges was similar between D/N-AP and E/N-AP, D/N-AP further reduces glucose variability.

Continuous Glucose Monitoring, Future Products, and Update on Worldwide Artificial Pancreas Projects

The development of accurate and easy-to-use continuous glucose monitoring (CGM) improved diabetes treatment by providing additional temporal information on glycemia and glucose trends to patient and physician. Although CGM enables users to lower their average glucose level without an increased incidence of hypoglycemia, this comes at the price of additional patient effort. Automation of insulin administration, also known as closed-loop (CL) or artificial pancreas treatment, has the promise to reduce patient effort and improve glycemic control. CGM data serve as the conditional input for insulin automation devices. The first commercial product for partial automation of insulin administration used insulin delivery shutoff at a predefined glucose level. These systems showed a reduction in hypoglycemia. Insulin-only CL devices show increased time spent in euglycemia and a reduction of hypo- and hyperglycemia. Improved glycemic control, coinciding with a minor decrease in hemoglobin A1c level, was confirmed in recent long-term home studies investigating these devices, paving the way for pivotal studies for commercialization of the artificial pancreas. Although the first results from dual-hormone CL systems are promising, because of increased cost of consumables of these systems, long-term head-to-head studies will have to prove superiority over insulin-only approaches. Now CL glucose control for daily use might finally become reality. Improved continuous glucose sensing technology, miniaturization of electrical devices, and development of algorithms were key in making this possible. Clinical adoption challenges, including device usability and reimbursement, need to be addressed. Time will tell for which patient groups CL systems will be reimbursed and whether these devices can deliver the promise that they hold.

2 month evening and night closed-loop glucose control in patients with type 1 diabetes under free-living conditions: a randomised crossover trial

BACKGROUND

An artificial pancreas (AP) that can be worn at home from dinner to waking up in the morning might be safe and efficient for first routine use in patients with type 1 diabetes. We assessed the effect on glucose control with use of an AP during the evening and night plus patient-managed sensor-augmented pump therapy (SAP) during the day, versus 24 h use of patient-managed SAP only, in free-living conditions.

METHODS

In a crossover study done in medical centres in France, Italy, and the Netherlands, patients aged 18-69 years with type 1 diabetes who used insulin pumps for continuous subcutaneous insulin infusion were randomly assigned to 2 months of AP use from dinner to waking up plus SAP use during the day versus 2 months of SAP use only under free-living conditions. Randomisation was achieved with a computer-generated allocation sequence with random block sizes of two, four, or six, masked to the investigator. Patients and investigators were not masked to the type of intervention. The AP consisted of a continuous glucose monitor (CGM) and insulin pump connected to a modified smartphone with a model predictive control algorithm. The primary endpoint was the percentage of time spent in the target glucose concentration range (3·9-10·0 mmol/L) from 2000 to 0800 h. CGM data for weeks 3-8 of the interventions were analysed on a modified intention-to-treat basis including patients who completed at least 6 weeks of each intervention period. The 2 month study period also allowed us to asses HbA1c as one of the secondary outcomes. This trial is registered with ClinicalTrials.gov, number NCT02153190.

FINDINGS

During 2000-0800 h, the mean time spent in the target range was higher with AP than with SAP use: 66·7% versus 58·1% (paired difference 8·6% [95% CI 5·8 to 11·4], p<0·0001), through a reduction in both mean time spent in hyperglycaemia (glucose concentration >10·0 mmol/L; 31·6% vs 38·5%; -6·9% [-9·8% to -3·9], p<0·0001) and in hypoglycaemia (glucose concentration <3·9 mmol/L; 1·7% vs 3·0%; -1·6% [-2·3 to -1·0], p<0·0001). Decrease in mean HbA1c during the AP period was significantly greater than during the control period (-0·3% vs -0·2%; paired difference -0·2 [95% CI -0·4 to -0·0], p=0·047), taking a period effect into account (p=0·0034). No serious adverse events occurred during this study, and none of the mild-to-moderate adverse events was related to the study intervention.

INTERPRETATION

Our results support the use of AP at home as a safe and beneficial option for patients with type 1 diabetes. The HbA1c results are encouraging but preliminary.

FUNDING

European Commission.

Multicenter outpatient dinner/overnight reduction of hypoglycemia and increased time of glucose in target with a wearable artificial pancreas using modular model predictive control in adults with type 1 diabetes

AIMS

To test in an outpatient setting the safety and efficacy of continuous subcutaneous insulin infusion (CSII) driven by a modular model predictive control (MMPC) algorithm informed by continuous glucose monitoring (CGM) measurement.

METHODS

13 patients affected by type 1 diabetes participated to a non-randomized outpatient 42-h experiment that included two evening meals and overnight periods (in short, dinner & night periods). CSII was patient-driven during dinner & night period 1 and MMPC-driven during dinner&night period 2. The study was conducted in hotels, where patients could move around freely. A CGM system (G4 Platinum; Dexcom Inc., San Diego, CA, USA) and insulin pump (AccuChek Combo; Roche Diagnostics, Mannheim, Germany) were connected wirelessly to a smartphone-based platform (DiAs, Diabetes Assistant; University of Virginia, Charlottesville, VA, USA) during both periods.

RESULTS

A significantly lower percentage of time spent with glucose levels <3.9 mmol/l was achieved in period 2 compared with period 1: 1.96 ± 4.56% vs 12.76 ± 15.84% (mean ± standard deviation, p < 0.01), together with a greater percentage of time spent in the 3.9-10 mmol/l target range: 83.56 ± 14.02% vs 62.43 ± 29.03% (p = 0.04). In addition, restricting the analysis to the overnight phases, a lower percentage of time spent with glucose levels <3.9 mmol/l (1.92 ± 4.89% vs 12.7 ± 19.75%; p = 0.03) was combined with a greater percentage of time spent in 3.9-10 mmol/l target range in period 2 compared with period 1 (92.16 ± 8.03% vs 63.97 ± 2.73%; p = 0.01). Average glucose levels were similar during both periods.

CONCLUSIONS

The results suggest that MMPC managed by a wearable system is safe and effective during evening meal and overnight. Its sustained use during this period is currently being tested in an ongoing randomized 2-month study.

Designing an artificial pancreas architecture: the AP@home experience

The latest achievements in sensor technologies for blood glucose level monitoring, pump miniaturization for insulin delivery, and the availability of portable computing devices are paving the way toward the artificial pancreas as a treatment for diabetes patients. This device encompasses a controller unit that oversees the administration of insulin micro-boluses and continuously drives the pump based on blood glucose readings acquired in real time. In order to foster the research on the artificial pancreas and prepare for its adoption as a therapy, the European Union in 2010 funded the AP@home project, following a series of efforts already ongoing in the USA. This paper, authored by members of the AP@home consortium, reports on the technical issues concerning the design and implementation of an architecture supporting the exploitation of an artificial pancreas platform. First a PC-based platform was developed by the authors to prove the effectiveness and reliability of the algorithms responsible for insulin administration. A mobile-based one was then adopted to improve the comfort for the patients. Both platforms were tested on real patients, and a description of the goals, the achievements, and the major shortcomings that emerged during those trials is also reported in the paper.

Model of glucose sensor error components: identification and assessment for new Dexcom G4 generation devices

It is clinically well-established that minimally invasive subcutaneous continuous glucose monitoring (CGM) sensors can significantly improve diabetes treatment. However, CGM readings are still not as reliable as those provided by standard fingerprick blood glucose (BG) meters. In addition to unavoidable random measurement noise, other components of sensor error are distortions due to the blood-to-interstitial glucose kinetics and systematic under-/overestimations associated with the sensor calibration process. A quantitative assessment of these components, and the ability to simulate them with precision, is of paramount importance in the design of CGM-based applications, e.g., the artificial pancreas (AP), and in their in silico testing. In the present paper, we identify and assess a model of sensor error of for two sensors, i.e., the G4 Platinum (G4P) and the advanced G4 for artificial pancreas studies (G4AP), both belonging to the recently presented "fourth" generation of Dexcom CGM sensors but different in their data processing. Results are also compared with those obtained by a sensor belonging to the previous, "third," generation by the same manufacturer, the SEVEN Plus (7P). For each sensor, the error model is derived from 12-h CGM recordings of two sensors used simultaneously and BG samples collected in parallel every 15 ± 5 min. Thanks to technological innovations, G4P outperforms 7P, with average mean absolute relative difference (MARD) of 11.1 versus 14.2%, respectively, and lowering of about 30% the error of each component. Thanks to the more sophisticated data processing algorithms, G4AP resulted more reliable than G4P, with a MARD of 10.0%, and a further decrease to 20% of the error due to blood-to-interstitial glucose kinetics.

Multicenter closed-loop/hybrid meal bolus insulin delivery with type 1 diabetes

BACKGROUND

This study evaluated meal bolus insulin delivery strategies and associated postprandial glucose control while using an artificial pancreas (AP) system.

SUBJECTS AND METHODS

This study was a multicenter trial in 53 patients, 12-65 years of age, with type 1 diabetes for at least 1 year and use of continuous subcutaneous insulin infusion for at least 6 months. Four different insulin bolus strategies were assessed: standard bolus delivered with meal (n=51), standard bolus delivered 15 min prior to meal (n=40), over-bolus of 30% delivered with meal (n=40), and bolus purposely omitted (n=46). Meal carbohydrate (CHO) intake was 1 g of CHO/kg of body weight up to a maximum of 100 g for the first three strategies or up to a maximum of 50 g for strategy 4.

RESULTS

Only three of 177 meals (two with over-bolus and one with standard bolus 15 min prior to meal) had postprandial blood glucose values of <60 mg/dL. Postprandial hyperglycemia (blood glucose level >180 mg/dL) was prolonged for all four bolus strategies but was shorter for the over-bolus (41% of the 4-h period) than the two standard bolus strategies (73% for each). Mean postprandial blood glucose level was 15.9 mg/dL higher for the standard bolus with meal compared with the prebolus (baseline-adjusted, P=0.07 for treatment effect over the 4-h period).

CONCLUSIONS

The AP handled the four bolus situations safely, but at the expense of having elevated postprandial glucose levels in most subjects. This was most likely secondary to suboptimal performance of the algorithm.

Monitoring Artificial Pancreas Trials Through Agent-based Technologies: A Case Report

The increase in the availability and reliability of network connections lets envision systems supporting a continuous remote monitoring of clinical parameters useful either for overseeing chronic diseases or for following clinical trials involving outpatients. We report here the results achieved by a telemedicine infrastructure that has been linked to an artificial pancreas platform and used during a trial of the AP@home project, funded by the European Union. The telemedicine infrastructure is based on a multiagent paradigm and is able to deliver to the clinic any information concerning the patient status and the operation of the artificial pancreas. A web application has also been developed, so that the clinic staff and the researchers involved in the design of the blood glucose control algorithms are able to follow the ongoing experiments. Albeit the duration of the experiments in the trial discussed in the article was limited to only 2 days, the system proved to be successful for monitoring patients, in particular overnight when the patients are sleeping. Based on that outcome we can conclude that the infrastructure is suitable for the purpose of accomplishing an intelligent monitoring of an artificial pancreas either during longer trials or whenever that system will be used as a routine treatment.