A model of glucose-insulin-pramlintide pharmacokinetics and pharmacodynamics in type I diabetes.

Type 1 diabetes mellitus (T1DM) complications are significantly reduced when normoglycemic levels are maintained via intensive therapy. The artificial pancreas is designed for intensive glycemic control; however, large postprandial excursions after a meal result in poor glucose regulation. Pramlintide, a synthetic analog of the hormone amylin, reduces the severity of postprandial excursions by reducing appetite, suppressing glucagon release, and slowing the rate of gastric emptying. The goal of this study is to create a glucose-insulin-pramlintide physiological model that can be employed into a controller to improve current control approaches used in the artificial pancreas. A model of subcutaneous (SC) pramlintide pharmacokinetics (PK) was developed by revising an intravenous (IV) pramlintide PK model and adapting SC insulin PK from a glucose-insulin model. Gray-box modeling and least squares optimization were used to obtain parameter estimates. Pharmacodynamics (PD) were obtained by choosing parameters most applicable to pramlintide mechanisms and then testing using a proportional PD effect using least squares optimization. The model was fit and validated using 27 data sets, which included placebo, PK, and PD data. SC pramlintide PK root mean square error values range from 1.98 to 10.66 pmol/L. Pramlintide PD RMSE values range from 10.48 to 42.76 mg/dL. A new in silico model of the glucose-insulin-pramlintide regulatory system is presented. This model can be used as a platform to optimize dosing of both pramlintide and insulin as a combined therapy for glycemic regulation, and in the development of an artificial pancreas as the kernel for a model-based controller.

Reduced hypoglycemia and increased time in target using closed-loop insulin delivery during nights with or without antecedent afternoon exercise in type 1 diabetes.

OBJECTIVE: Afternoon exercise increases the risk of nocturnal hypoglycemia (NH) in subjects with type 1 diabetes. We hypothesized that automated feedback-controlled closed-loop (CL) insulin delivery would be superior to open-loop (OL) control in preventing NH and maintaining a higher proportion of blood glucose levels within the target blood glucose range on nights with and without antecedent afternoon exercise. RESEARCH DESIGN AND METHODS: Subjects completed two 48-h inpatient study periods in random order: usual OL control and CL control using a proportional-integrative-derivative plus insulin feedback algorithm. Each admission included a sedentary day and an exercise day, with a standardized protocol of 60 min of brisk treadmill walking to 65-70% maximum heart rate at 3:00 p.m. RESULTS: Among 12 subjects (age 12-26 years, A1C 7.4±0.6%), antecedent exercise increased the frequency of NH (reference blood glucose<60 mg/dL) during OL control from six to eight events. In contrast, there was only one NH event each on nights with and without antecedent exercise during CL control (P=0.04 vs. OL nights). Overnight, the percentage of glucose values in target range was increased with CL control (P<0.0001). Insulin delivery was lower between 10:00 p.m. and 2:00 a.m. on nights after exercise on CL versus OL, P=0.008. CONCLUSIONS: CL insulin delivery provides an effective means to reduce the risk of NH while increasing the percentage of time spent in target range, regardless of activity level in the mid-afternoon. These data suggest that CL control could be of benefit to patients with type 1 diabetes even if it is limited to the overnight period.

Real-time continuous glucose monitoring in an intensive care unit: better accuracy in patients with septic shock.

OBJECTIVE: This study assessed the accuracy of real-time continuous glucose monitoring system (RTCGMS) devices in an intensive care unit (ICU) to determine whether the septic status of the patient has any influence on the accuracy of the RTCGMS. SUBJECTS AND METHODS: In total, 41 patients on insulin therapy were included. Patients were monitored for 72 h using RTCGMS. Arterial blood glucose (ABG) samples were obtained following the protocol established in the ICU. The results were evaluated using paired values (excluding those used for calibration) with the performance assessed using numerical accuracy. Nonparametric tests were used to determine statistically significant differences in accuracy. RESULTS: In total, 956 ABG/RTCGMS pairs were analyzed. The overall median relative absolute difference (RAD) was 13.5%, and the International Organization for Standardization (ISO) criteria were 68.1%. The median RADs reported for patients with septic shock, with sepsis, and without sepsis were 11.2%, 14.3%, and 16.3%, respectively (P<0.05). Measurements meeting the ISO criteria were 74.5%, 65.6%, and 63.7% for patients with septic shock, with sepsis, and without sepsis, respectively (P<0.05). CONCLUSIONS: The results showed that the septic status of patients influenced the accuracy of the RTCGMS in the ICU. Accuracy was significantly better in patients with septic shock in comparison with the other patient cohorts.

A simple robust method for estimating the glucose rate of appearance from mixed meals.

BACKGROUND: Estimating the rate of glucose appearance (R(a)) after ingestion of a mixed meal may be highly valuable in diabetes management. The gold standard technique for estimating R(a) is the use of a multitracer oral glucose protocol. However, this technique is complex and is usually not convenient for large studies. Alternatively, a simpler approach based on the glucose-insulin minimal model is available. The main drawback of this last approach is that it also requires a gastrointestinal model, something that may lead to identifiability problems. METHODS: In this article, we present an alternative, easy-to-use method based on the glucose-insulin minimal model for estimation of R(a). This new technique avoids complex experimental protocols by only requiring data from a standard meal tolerance test. Unlike other model-based approaches, this new approach does not require a gastrointestinal model, which leads to a much simpler solution. Furthermore, this novel technique requires the identification of only one parameter of the minimal model because the rest of the model parameters are considered to have small variability. In order to account for such variability as well as to account for errors associated to measurements, interval analysis has been employed. RESULTS: The current technique has been validated using data from a United States Food and Drug Administration-accepted type 1 diabetes simulator [root mean square error (RMSE) = 0.77] and successfully tested with two clinical data sets from the literature (RMSE = 0.69). CONCLUSIONS: The presented technique for the estimation of R(a) showed excellent results when tested with simulated and actual clinical data. The simplicity of this new technique makes it suitable for large clinical research studies for the evaluation of the role of R(a) in patients with impairments in glucose metabolism. In addition, this technique is being used to build a model library of mixed meals that could be incorporated into diabetic subject simulators in order to account for more realistic and varied meals.

The effect of insulin feedback on closed loop glucose control.

CONTEXT: Initial studies of closed-loop proportional integral derivative control in individuals with type 1 diabetes showed good overnight performance, but with breakfast meal being the hardest to control and requiring supplemental carbohydrate to prevent hypoglycemia. OBJECTIVE: The aim of this study was to assess the ability of insulin feedback to improve the breakfast-meal profile. DESIGN AND SETTING: We performed a single center study with closed-loop control over approximately 30 h at an inpatient clinical research facility. PATIENTS: Eight adult subjects with previously diagnosed type 1 diabetes participated. INTERVENTION: Subjects received closed-loop insulin delivery with supplemental carbohydrate as needed. MAIN OUTCOME MEASURES: Outcome measures were plasma insulin concentration, model-predicted plasma insulin concentration, 2-h postprandial and 3- to 4-h glucose rate-of-change following breakfast after 1 d of closed-loop control, and the need for supplemental carbohydrate in response to nadir hypoglycemia. RESULTS: Plasma insulin levels during closed loop were well correlated with model predictions (R = 0.86). Fasting glucose after 1 d of closed loop was not different from nighttime target (118 ± 9 vs. 110 mg/dl; P = 0.38). Two-hour postbreakfast glucose was 132 ± 16 mg/dl with stable values 3-4 h after the meal (0.03792 ± 0.0884 mg/dl · min, not different from 0; P = 0.68) and at target (97 ± 6 mg/dl, not different from 90; P = 0.28). Three subjects required supplemental carbohydrates after breakfast on d 2 of closed loop. CONCLUSIONS/INTERPRETATION: Insulin feedback can be implemented using a model estimate of concentration. Proportional integral derivative control with insulin feedback can achieve a desired breakfast response but still requires supplemental carbohydrate to be delivered in some instances. Studies assessing more optimal control configurations and safeguards need to be conducted.

Physiologic insulin delivery with insulin feedback: a control systems perspective.

Individuals with type 1 diabetes mellitus must effectively manage glycemia to avoid acute and chronic complications related to aberrations of glucose levels. Because optimal diabetes management can be difficult to achieve and burdensome, research into a closed-loop insulin delivery system has been of interest for several decades. This paper provides an overview, from a control systems perspective, of the research and development effort of a particular algorithm--the external physiologic insulin delivery system. In particular the introduction of insulin feedback, based on ß-cell physiology, is covered in detail. A summary of human clinical trials is provided in the context of the evolution of this algorithm, and this paper outlines some of the research avenues that show particular promise.

Closed-loop insulin delivery using a subcutaneous glucose sensor and intraperitoneal insulin delivery: feasibility study testing a new model for the artificial pancreas.

OBJECTIVE: Attempts to build an artificial pancreas by using subcutaneous insulin delivery from a portable pump guided by an subcutaneous glucose sensor have encountered delays and variability of insulin absorption. We tested closed-loop intraperitoneal insulin infusion from an implanted pump driven by an subcutaneous glucose sensor via a proportional-integral-derivative (PID) algorithm. RESEARCH DESIGN AND METHODS: Two-day closed-loop therapy (except for a 15-min pre-meal manual bolus) was compared with a 1-day control phase with intraperitoneal open-loop insulin delivery, according to randomized order, in a hospital setting in eight type 1 diabetic patients treated by implanted pumps. The percentage of time spent with blood glucose in the 4.4-6.6 mmol/l range was the primary end point. RESULTS During the closed-loop phases, the mean +/- SEM percentage of time spent with blood glucose in the 4.4-6.6 mmol/l range was significantly higher (39.1 +/- 4.5 vs. 27.7 +/- 6.2%, P = 0.05), and overall dispersion of blood glucose values was reduced among patients. Better closed-loop glucose control came from the time periods excluding the two early postprandial hours with a higher percentage of time in the 4.4-6.6 mmol/l range (46.3 +/- 5.3 vs. 28.6 +/- 7.4, P = 0.025) and lower mean blood glucose levels (6.9 +/- 0.3 vs. 7.9 +/- 0.6 mmol/l, P = 0.036). Time spent with blood glucose <3.3 mmol/l was low and similar for both investigational phases. CONCLUSIONS: Our results demonstrate the feasibility of intraperitoneal insulin delivery for an artificial beta-cell and support the need for further study. Moreover, according to a semiautomated mode, the features of the pre-meal bolus in terms of timing and amount warrant further research.

In silico evaluation platform for artificial pancreatic beta-cell development--a dynamic simulator for closed-loop control with hardware-in-the-loop.

BACKGROUND: A critical step in algorithm development for an artificial beta-cell is extensive in silico testing. Computer simulations usually involve only the controller software, leaving untested the hardware elements, including the critical communication interface between the controller and the glucose sensor and insulin pump. METHODS: An in silico simulation platform has been developed that uses all of the components of the clinical system. At the core is a comprehensive in silico population model that covers the variability of principal metabolic parameters observed in vivo, to replace the human subject, with the ability to use historical clinical data. A continuous glucose monitor, in this case either the Abbott Diabetes Care (Alameda, CA) FreeStyle Navigator or the DexCom (San Diego, CA) STS7, is supplied with a glucose signal provided by the simulator. The Insulet (Bedford, MA) OmniPod insulin pump is also interfaced with the simulator to provide insulin delivery data. These hardware elements are an integral part of the system under testing, which also includes the algorithm components. RESULTS: The system is unique in that it uses the same hardware components for simulations as are required in clinical trials, allowing for full-system level verification and validation. With a detailed mathematical model, a suite of patients can be simulated to reflect various conditions. Because all hardware is used, their related limitations are automatically included. CONCLUSIONS: A complete artificial beta-cell evaluation platform was realized with the flexibility to interface various algorithms and patient models, allowing for the systematic analysis of monitoring and control algorithms. The system facilitates a variety of tests and challenges to the software and the component devices, streamlining preclinical validation trials.

Clinical update on optimal prandial insulin dosing using a refined run-to-run control algorithm.

BACKGROUND: This article provides a clinical update using a novel run-to-run algorithm to optimize prandial insulin dosing based on sparse glucose measurements from the previous day's meals. The objective was to use a refined run-to-run algorithm to calculate prandial insulin-to-carbohydrate ratios (I:CHO) for meals of variable carbohydrate content in subjects with type 1 diabetes (T1DM). METHOD: The open-labeled, nonrandomized study took place over a 6-week period in a nonprofit research center. Nine subjects with T1DM using continuous subcutaneous insulin infusion participated. Basal insulin rates were optimized using continuous glucose monitoring, with a target fasting blood glucose of 90 mg/dl. Subjects monitored blood glucose concentration at the beginning of the meal and at 60 and 120 minutes after the start of the meal. They were instructed to start meals with blood glucose levels between 70 and 130 mg/dl. Subjects were contacted daily to collect data for the previous 24-hour period and to give them the physician-approved, algorithm-derived I:CHO ratios for the next 24 hours. Subjects calculated the amount of the insulin bolus for each meal based on the corresponding I:CHO and their estimate of the meal's carbohydrate content. One- and 2-hour postprandial glucose concentrations served as the main outcome measures. RESULTS: The mean 1-hour postprandial blood glucose level was 104 +/- 19 mg/dl. The 2-hour postprandial levels (96.5 +/- 18 mg/dl) approached the preprandial levels (90.1 +/- 13 mg/dl). CONCLUSIONS: Run-to-run algorithms are able to improve postprandial blood glucose levels in subjects with T1DM.

Experimental evaluation of a recursive model identification technique for type 1 diabetes.

BACKGROUND: A model-based controller for an artificial beta cell requires an accurate model of the glucose-insulin dynamics in type 1 diabetes subjects. To ensure the robustness of the controller for changing conditions (e.g., changes in insulin sensitivity due to illnesses, changes in exercise habits, or changes in stress levels), the model should be able to adapt to the new conditions by means of a recursive parameter estimation technique. Such an adaptive strategy will ensure that the most accurate model is used for the current conditions, and thus the most accurate model predictions are used in model-based control calculations. METHODS: In a retrospective analysis, empirical dynamic autoregressive exogenous input (ARX) models were identified from glucose-insulin data for nine type 1 diabetes subjects in ambulatory conditions. Data sets consisted of continuous (5-minute) glucose concentration measurements obtained from a continuous glucose monitor, basal insulin infusion rates and times and amounts of insulin boluses obtained from the subjects' insulin pumps, and subject-reported estimates of the times and carbohydrate content of meals. Two identification techniques were investigated: nonrecursive, or batch methods, and recursive methods. Batch models were identified from a set of training data, whereas recursively identified models were updated at each sampling instant. Both types of models were used to make predictions of new test data. For the purpose of comparison, model predictions were compared to zero-order hold (ZOH) predictions, which were made by simply holding the current glucose value constant for p steps into the future, where p is the prediction horizon. Thus, the ZOH predictions are model free and provide a base case for the prediction metrics used to quantify the accuracy of the model predictions. In theory, recursive identification techniques are needed only when there are changing conditions in the subject that require model adaptation. Thus, the identification and validation techniques were performed with both "normal" data and data collected during conditions of reduced insulin sensitivity. The latter were achieved by having the subjects self-administer a medication, prednisone, for 3 consecutive days. The recursive models were allowed to adapt to this condition of reduced insulin sensitivity, while the batch models were only identified from normal data. RESULTS: Data from nine type 1 diabetes subjects in ambulatory conditions were analyzed; six of these subjects also participated in the prednisone portion of the study. For normal test data, the batch ARX models produced 30-, 45-, and 60-minute-ahead predictions that had average root mean square error (RMSE) values of 26, 34, and 40 mg/dl, respectively. For test data characterized by reduced insulin sensitivity, the batch ARX models produced 30-, 60-, and 90-minute-ahead predictions with average RMSE values of 27, 46, and 59 mg/dl, respectively; the recursive ARX models demonstrated similar performance with corresponding values of 27, 45, and 61 mg/dl, respectively. The identified ARX models (batch and recursive) produced more accurate predictions than the model-free ZOH predictions, but only marginally. For test data characterized by reduced insulin sensitivity, RMSE values for the predictions of the batch ARX models were 9, 5, and 5% more accurate than the ZOH predictions for prediction horizons of 30, 60, and 90 minutes, respectively. In terms of RMSE values, the 30-, 60-, and 90-minute predictions of the recursive models were more accurate than the ZOH predictions, by 10, 5, and 2%, respectively. CONCLUSION: In this experimental study, the recursively identified ARX models resulted in predictions of test data that were similar, but not superior, to the batch models. Even for the test data characteristic of reduced insulin sensitivity, the batch and recursive models demonstrated similar prediction accuracy. The predictions of the identified ARX models were only marginally more accurate than the model-free ZOH predictions. Given the simplicity of the ARX models and the computational ease with which they are identified, however, even modest improvements may justify the use of these models in a model-based controller for an artificial beta cell.

A Run-to-Run Control Strategy to Adjust Basal Insulin Infusion Rates in Type 1 Diabetes.

Maintaining good glycemic control is a daily challenge for people with type 1 diabetes. Insulin requirements are changing constantly due to many factors, such as levels of stress and physical activity. The basal insulin requirement also has a circadian rhythm, adding another level of complexity. Automating the adjustment of insulin dosing would result in improved glycemic control, as well as an improved quality of life by significantly reducing the burden on the patient. Building on our previous success of using run-to-run control for prandial insulin dosing (a strategy adapted from the chemical process industry), we show how this same framework can be used to adjust basal infusion profiles. We present a mathematical model of insulin-glucose dynamics which we augment in order to capture the circadian variation in insulin requirements. Using this model, we show that the run-to-run framework can also be successfully applied to adjust basal insulin dosing.

Use of continuous glucose monitoring to estimate insulin requirements in patients with type 1 diabetes mellitus during a short course of prednisone.

BACKGROUND: Insulin requirements to maintain normoglycemia during glucocorticoid therapy and stress are often difficult to estimate. To simulate insulin resistance during stress, adults with type 1 diabetes mellitus (T1DM) were given a three-day course of prednisone. METHODS: Ten patients (7 women, 3 men) using continuous subcutaneous insulin infusion pumps wore the Medtronic Minimed CGMS (Northridge, CA) device. Mean (standard deviation) age was 43.1 (14.9) years, body mass index 23.9 (4.7) kg/m(2), hemoglobin A1c 6.8% (1.2%), and duration of diabetes 18.7 (10.8) years. Each patient wore the CGMS for one baseline day (day 1), followed by three days of self-administered prednisone (60 mg/dl; days 2-4), and one post-prednisone day (day 5). RESULTS: Analysis using Wilcoxon signed rank test (values are median [25th percentile, 75th percentile]) indicated a significant difference between day 1 and the mean of days on prednisone (days 2-4) for average glucose level (110.0 [81.0, 158.0] mg/dl vs 149.2 [137.7, 168.0] mg/dl; p = .022), area under the glucose curve and above the upper limit of 180 mg/dl per day (0.5 [0, 8.0] mg/dl.d vs 14.0 [7.7, 24.7] mg/dl.d; p = .002), and total daily insulin dose (TDI) , (0.5 [0.4, 0.6] U/kg.d vs 0.9 [0.8, 1.0] U/kg.d; p = .002). In addition, the TDI was significantly different for day 1 vs day 5 (0.5 [0.4, 0.6] U/kg.d vs 0.6 [0.5, 0.8] U/kg.d; p = .002). Basal rates and insulin boluses were increased by an average of 69% (range: 30-100%) six hours after the first prednisone dose and returned to baseline amounts on the evening of day 4. CONCLUSIONS: For adults with T1DM, insulin requirements during prednisone induced insulin resistance may need to be increased by 70% or more to normalize blood glucose levels.

Prandial insulin dosing using run-to-run control: application of clinical data and medical expertise to define a suitable performance metric.

OBJECTIVE: We propose a novel algorithm to adjust prandial insulin dose using sparse blood glucose measurements. The dose is adjusted on the basis of a performance measure for the same meal on the previous day. We determine the best performance measure and tune the algorithm to match the recommendations of experienced physicians. RESEARCH DESIGN AND METHODS: Eleven subjects with type 1 diabetes, using continuous subcutaneous insulin infusion, were recruited (seven women and four men, aged 21-65 years with A1C of 7.1 +/- 1.3%). Basal insulin infusion rates were optimized. Target carbohydrate content for the lunch meal was calculated on the basis of a weight-maintenance diet. Over a period of 2-4 days, subjects were asked to measure their blood glucose according to the algorithm's protocol. Starting with their usual insulin-to-carbohydrate ratio, the insulin bolus dose was titrated downward until postprandial glucose levels were high (180-250 mg/dl [10-14 mmol/l]). Subsequently, physicians made insulin bolus recommendations to normalize postprandial glucose concentrations. Graphical methods were then used to determine the most appropriate performance measure for the algorithm to match the physician's decisions. For the best performance measure, the gain of the controller was determined to be the best match to the dose recommendations of the physicians. RESULTS: The correlation between the clinically determined dose adjustments and those of the algorithm is R2 = 0.95, P < 1e - 18. CONCLUSIONS: We have shown how engineering methods can be melded with medical expertise to develop and refine a dosing algorithm. This algorithm has the potential of drastically simplifying the determination of correct insulin-to-carbohydrate ratios.

Calculating the insulin to carbohydrate ratio using the hyperinsulinaemic-euglycaemic clamp-a novel use for a proven technique.

BACKGROUND: In patients with type 1 diabetes, three main variables need to be assessed to optimize meal-related insulin boluses: pre-meal blood glucose (BG), insulin to carbohydrate ratio (I : C), and basal insulin. We are presenting data for a novel use of the hyperinsulinaemic-euglycaemic clamp (HEC) in patients with type 1 diabetes that minimizes the impact of these variables and can be used to determine the I : C. METHODS: Ten subjects (six men and four women) using continuous subcutaneous insulin infusion (CSII) pumps were recruited for this study [24-65 years; BMI 27.1 +/- 4.9 kg/m(2); A1C 7.2 +/- 1.4% (mean +/- SD)]. The HEC used a primed continuous intravenous insulin infusion of 40 mU/m(2)/min and a variable infusion of 20% glucose to maintain BG at 90 mg/dL. After subjects were in steady state (SS) for 50 min, a standardized meal (40% of total calories/day - 30% carbohydrate, 30% protein, 40% fat) was consumed. Subjects gave the insulin bolus with their CSII pump. No changes were made in the glucose infusion rate. RESULTS: Mean BG at SS was 85.7 +/- 10.4 mg/dL. Peak BG was 115.0 +/- 12.7 mg/dL at 68.5 +/- 8.8 min after the meal. Mean I : C was 1 : 9.3 +/- 1.7 (range 1 : 7-1 : 12). Insulin sensitivity varied from 1.9 to 9.1 mg/kg/min. CONCLUSIONS: The HEC can be used to reduce confounding factors and to determine the I : C. As a first estimate of the I : C in patients with type 1 diabetes, it is recommended to start with a ratio of 1 : 9.3 and to measure post-prandial BG at 70 min.

Hypoglycemia detection and prediction using continuous glucose monitoring-a study on hypoglycemic clamp data.

MOTIVATION: The fear of hypoglycemia remains an important limiting factor in the ability of an individual with type 1 diabetes to tightly regulate glycemia. Continuous glucose monitors provide important feedback to improve glycemic control, but there remains a need for these devices to better alarm of possible impending hypoglycemia, particularly overnight or other periods when the individual is engaged in activities that take their focus away from glucose monitoring. METHODS: We have previously proposed an algorithm, based on the use of real-time glucose sensor signals and optimal estimation theory (Kalman filtering), to predict hypoglycemia; the algorithm was validated in simulation-based studies. In this article we further refine and validate the prediction algorithm based on the analysis of clinical hypoglycemic clamp data from 13 subjects. The sensitivity and specificity of the predictions are calculated with respect to reference blood glucose values obtained at the same sampling rate of the sensor. RESULTS: For a 30-minute prediction horizon and alarm threshold of 70 mg/dl, the sensitivity and specificity were 90 and 79%, respectively, indicating that a 21% false alarm rate must be tolerated to predict 90% of the hypoglycemic events 30 minutes ahead of time. Shorter prediction horizons yield a significant improvement in sensitivity and specificity. DISCUSSION: Sensitivity and specificity data as a function of prediction horizon and alarm threshold enable an individual to adjust the alarm to best meet their needs. Such decisions can be made depending on the subject's risk for hypoglycemia, for example.

Glucose estimation and prediction through meal responses using ambulatory subject data for advisory mode model predictive control.

BACKGROUND: A primary challenge for closed-loop glucose control in type 1 diabetes mellitus (T1DM) is the development of a control strategy that will be applicable during all daily activities, including meals, stress, and exercise. A model-based control algorithm requires a mathematical model that has the simplicity for online glucose prediction, yet retains the complexity necessary to cope with variations in insulin sensitivities and carbohydrate ingestion. METHODS: A modified Bergman minimal model was linearized for Kalman filter (KF) state estimation on data from T1DM subjects, and multiple methods of parameter augmentation were developed for online adaptation. In addition, model deterioration for glucose prediction was assessed to determine an appropriate prediction horizon for model predictive control (MPC). Furthermore, MPC strategies were validated using advisory mode simulations. RESULTS: Twenty days of continuous glucose data, which included 97 meals, were evaluated for three subjects. A constant parameter minimal model was used to predict glucose levels for normal days with meal announcement and with a maximum prediction horizon of approximately 45 minutes. In order to attain this prediction horizon in the absence of meal announcement, parameter adaptation was necessary to capture the glucose disturbance. Evaluation of advisory mode MPC permitted effective tuning for a moderately aggressive controller that responded well to meal disturbances. CONCLUSIONS: Estimation and prediction of glucose were accomplished using a KF based on a modified Bergman model. For a model with no meal announcement, parameter adaptation provided the means for closed-loop implementation. This state estimation and model validation scheme established the necessary framework for advisory mode MPC.

A dual-rate Kalman filter for continuous glucose monitoring.

A dual-rate Kalman filter is developed for realtime continuous glucose monitoring. Frequent (5 minute) sampling of a noisy, continuous glucose sensor is used for estimation of glucose and its rate-of-change. Infrequent (8 hour intervals) reference glucose meter samples enable the sensor gain and its rate-of-change to be updated. The dual-rate Kalman filter formulation accounts for uncertainty in both the continuous glucose sensor and the reference glucose meter. The method is tested on simulated and experimental data, confirming its superiority to simple one-point calibration.

Robust parameter estimation in a model for glucose kinetics in type 1 diabetes subjects.

There is a significant push to develop closed-loop control systems to deliver insulin for type 1 diabetic subjects. As part of this process, mathematical models are required to test and validate the proposed algorithms. There are several published physiology-based models of glucose and insulin dynamics in the literature, however, all of them were derived using data from subjects without diabetes. For this particular study we have selected one of the recently published models, by Hovorka et al., replacing the subcutaneous insulin infusion model with the one described by Wilinska et al., Five subjects with type 1 diabetes underwent a hyperinsulinemic-euglycemic clamp with a meal challenge and corresponding subcutaneous insulin bolus. The data collected were used to fit the model parameters using global optimization methods. Our results show that the model is capable of describing the observed dynamics for type 1 subjects under the experimental conditions, and as such can be used to simulate subject behavior under the experimental conditions.

Hypoglycemia prediction and detection using optimal estimation.

Patients with diabetes play with a double-edged sword when it comes to deciding glucose and A1c target levels. On the one side, tight control has been shown to be crucial in avoiding long-term complications; on the other, tighter control leads to an increased risk of iatrogenic hypoglycemia, which is compounded when hypoglycemia unawareness sets in. Development of continuous glucose monitoring systems has led to the possibility of being able not only to detect hypoglycemic episodes, but to make predictions based on trends that would allow the patient to take preemptive action to entirely avoid the condition. Using an optimal estimation theory approach to hypoglycemia prediction, we demonstrate the effect of measurement sampling frequency, threshold level, and prediction horizon on the sensitivity and specificity of the predictions. We discuss how optimal estimators can be tuned to trade-off the false alarm rate with the rate of missed predicted hypoglycemic episodes. We also suggest the use of different alarm levels as a function of current and future estimates of glucose and the hypoglycemic threshold and prediction horizon.