Roman Hovorka

Continuous glucose monitoring and closed-loop systems

Background  The last two decades have witnessed unprecedented technological progress in the development of continuous glucose sensors, resulting in the first generation of commercial glucose monitors. This has fuelled the development of prototypes of a closed‐loop system based on the combination of a continuous monitor, a control algorithm, and an insulin pump.

A probabilistic approach to glucose prediction and insulin dose adjustment: description of metabolic model and pilot evaluation study

A model of carbohydrate metabolism has been implemented as a causal probabilistic network, allowing explicit representation of the uncertainties involved in the prediction of 24-h blood glucose profiles in insulin-dependent diabetic subjects. The parameters of the model were based on experimental data from the literature describing insulin and carbohydrate absorption, renal loss of glucose, insulin-independent glucose utilisation and insulin-dependent glucose utilisation and production. The model can be adapted to the observed glucose metabolism in the individual patient and can be used to generate predicted 24-h blood glucose profiles. A penalty is assigned to each level of blood glucose, to indicate that high and low blood glucose levels are undesirable. The system can be asked to find the insulin doses that result in the most desirable 24-h blood glucose profile. In a series of 12 patients, the system predicted blood glucose with a mean error of 3.3 mmol/l. The insulin doses suggested by the system seemed reasonable and in several cases seemed more appropriate than the doses actually administered to the patients.

ISEC: a program to calculate insulin secretion

ISEC (Insulin SECretion) is a computer program which calculates pre-hepatic insulin secretion from plasma C-peptide measurements. The program uses a regression (population) model to derive parameters of C-peptide kinetics from subjects gender, type (normal, obese, non-insulin dependent diabetes mellitus), age, weight, and height. Insulin secretion is calculated as a piece-wise constant (step) function with flexible step length allowing for a fine resolution of the secretion profile between measurements. A constrained regularisation method of deconvolution is employed to carry out the calculations. The calculated profile satisfies three properties: (i) it fits the measurement within the given level of the measurement error, (ii) it is non-negative, and (iii) it has a minimum value of a regularisation criterion (norm of second differences) which quantifies the degree of deviation of the secretion profile from a straight line. Both theoretical aspects and specific features related to ISEC are considered. To exemplify the use of ISEC, pre-hepatic insulin secretion is calculated during meal tolerance test, frequently sampled intravenous glucose tolerance test, hyperinsulinaemic euglycaemic glucose clamp, and basal conditions with frequent sampling.

A model-based approach to insulin adjustment

A differential equation model of carbohydrate metabolism was implemented in the form of a causal probabilistic network. This permitted explicit represen-tations of the uncertainties associated with model based predictions of 24-hour blood glucose profiles. In addition, the implementation gave automatic learning and adjustment of model parameters based on measured blood glucose profiles. Insulin therapy was adjusted using a decision theoretical approach. Losses were assigned to blood glucose values that deviated from normal, and the insulin therapy was adjusted to minimize the expected total loss. The system was tested retrospectively on cases from 12 insulin dependent patients and seemed to compare favourably with clinical practice.

A Comparison of Six Deconvolution Techniques

We present results for the comparison of six deconvolution techniques. The methods we consider are based on Fourier transforms, system identification, constrained optimization, the use of cubic spline basis functions, maximum entropy, and a genetic algorithm. We compare the performance of these techniques by applying them to simulated noisy data, in order to extract an input function when the unit impulse response is known. The simulated data are generated by convolving the known impulse response with each of five different input functions, and then adding noise of constant coefficient of variation. Each algorithm was tested on 500 data sets, and we define error measures in order to compare the performance of the different methods.

Measuring prehepatic insulin secretion using a population model of C-peptide kinetics: accuracy and required sampling schedule

Summary The accuracy of calculations of pre-hepatic insulin secretion were investigated, to provide independent validation of a population model of C-peptide kinetics. The effects of sampling frequency were also assessed. Five normal subjects (aged 28 to 43 years; BMI (kg/m2) 20.5 to 24.5) and five subjects with non-insulin-dependent diabetes mellitus (NIDDM) treated by diet alone (aged 34 to 57 years; BMI 22.6 to 25.6) were given a variable intravenous infusion of biosynthetic human C-peptide (BHCP) (t = –60 to 240 min) mimicking meal stimulated C-peptide secretion, with short-term oscillations (peak approximately every 12 min) superimposed on the infusion profile. Plasma C-peptide was measured every 5 min (t = 0 to 240 min). The BHCP infusion was reconstructed from C-peptide measurements using a population model of C-peptide kinetics and a deconvolution method. Bias, defined as the percentage difference between the total amount of calculated BHCP and the total amount of infused BHCP (t = 0 to 240 min), indicated that overall C-peptide secretion can be measured with 14 % [95 % confidence interval (CI) –11 to 39 %] and 21 % (95 % CI –3 to 45 %) accuracy in normal subjects and subjects with NIDDM respectively. Accuracy was not reduced by reducing the sampling frequency to every 30 min. The root mean square error, measuring the average deviation between the infused and normalised calculated BHCP profiles, was also independent of the sampling frequency [mean (95 % CI) 0.9 (0.3 to 1.6) pmol/kg per min in normal subjects; 1.0 (0.9 to 1.1) pmol/kg per min in subjects with NIDDM]. Deconvolution employing a population model of C-peptide kinetics can be used to estimate postprandial total C-peptide secretion with biases of 14 % and 22 % respectively in normal subjects and subjects with NIDDM. Plasma C-peptide samples need only be drawn every 30 minutes. [Diabetologia (1998) 41: 548–554]

DIAS—the diabetes advisory system: an outline of the system and the evaluation results obtained so far

The present paper gives a description of the Diabetes Advisory System (DIAS), and the evaluation results obtained so far. DIAS is a decision support system for the management of insulin dependent diabetes. The core of the system is a compartment model of the human carbohydrate metabolism implemented as a causal probabilistic network (CPN or Bayesian network), which gives it the ability to handle the uncertainty, for example, in blood glucose measurements or physiological variations in glucose metabolism. The evaluation results suggest that, at least in our hands, DIAS can generate advice that is safe and of a quality that is at least comparable to what is available from experienced clinicians.

A consultation system for insulin therapy

This paper describes a computer system to advice on insulin therapy for diabetic in-patients. A mathematical model was developed to describe the effect of insulin on blood glucose (BG) level. The system uses an adaptive approach to analyse the response to an applied insulin dosage. It learns the patients individual parameters. All conventional injection and insulin pump regimens are supported. The individualised model is used to predict BG level of the proposed insulin dosage. The system uses a generate-reject strategy to output optimum insulin therapy in terms of optimum BG. The predictive capability of the system was tested and it is able to predict BG with a precision of 2.5 mmol/l after 3 days and 6 days of insulin pump treatment and conventional injection therapy, respectively.

Effect of growth hormone treatment on postprandial protein metabolism in growth hormone-deficient adults.

Growth hormone (GH) treatment of GH-deficient adults increases lean body mass. To investigate this anabolic effect of GH, body composition and postabsorptive and postprandial protein metabolism were measured in 12 GH-deficient adults randomized to placebo or GH treatment. Protein metabolism was measured after an infusion of [1-13C]leucine before and after a standard meal at 0 and 2 mo. After 2 mo, there was an increase in lean body mass in the GH group (P < 0. 05) but no change in the placebo group. In the postabsorptive state, there was increased nonoxidative leucine disappearance (NOLD; a measure of protein synthesis) and leucine metabolic clearance rate and decreased leucine oxidation in the GH group (P < 0.05) but no change in the placebo group. After the meal, there was an increase in NOLD and oxidation in all studies (P < 0.05), but the increase in NOLD, measured as area under the curve, was greater in the GH group (P < 0.05). This study clearly demonstrates for the first time that the increase in protein synthesis in the postabsorptive state after GH treatment of GH-deficient adults is maintained in the postprandial state.Growth hormone (GH) treatment of GH-deficient adults increases lean body mass. To investigate this anabolic effect of GH, body composition and postabsorptive and postprandial protein metabolism were measured in 12 GH-deficient adults randomized to placebo or GH treatment. Protein metabolism was measured after an infusion of [1-13C]leucine before and after a standard meal at 0 and 2 mo. After 2 mo, there was an increase in lean body mass in the GH group ( P < 0.05) but no change in the placebo group. In the postabsorptive state, there was increased nonoxidative leucine disappearance (NOLD; a measure of protein synthesis) and leucine metabolic clearance rate and decreased leucine oxidation in the GH group ( P < 0.05) but no change in the placebo group. After the meal, there was an increase in NOLD and oxidation in all studies ( P < 0.05), but the increase in NOLD, measured as area under the curve, was greater in the GH group ( P < 0.05). This study clearly demonstrates for the first time that the increase in protein synthesis in the postabsorptive state after GH treatment of GH-deficient adults is maintained in the postprandial state.

Casual probabilistic network modeling: an illustration of its role in the management of chronic diseases

This paper describes the role of the novel technique of causal probabilistic network (CPN) modeling as an approach to tackling control system problems typified by that of the administration of treatment to the patient suffering from a chronic disease such as diabetes. Three roles of a CPN are discussed. First, since diabetes arises as a consequence of impaired control of carbohydrate metabolism, the ability of a CPN to represent the uncertainty of a physiologically-based model is described. Second, its ability to make robust estimates of the parameters of the metabolic model is presented, and finally, in conjunction with decision theory approaches, its ability to compare alternative therapies and advise on insulin therapy for patients with insulin-dependent diabetes mellitus is illustrated.