Modeling the glucose regulatory system in extreme preterm infants

Classification-based deep neural network vs mixture density network models for insulin sensitivity prediction problem

Model-based glycemic control (GC) protocols are used to treat stress-induced hyperglycaemia in intensive care units (ICUs). The STAR (Stochastic-TARgeted) glycemic control protocol - used in clinical practice in several ICUs in New Zealand, Hungary, Belgium, and Malaysia - is a model-based GC protocol using a patient-specific, model-based insulin sensitivity to describe the patient's actual state. Two neural network based methods are defined in this study to predict the patient's insulin sensitivity parameter: a classification deep neural network and a Mixture Density Network based method. Treatment data from three different patient cohorts are used to train the network models. Accuracy of neural network predictions are compared with the current model- based predictions used to guide care. The prediction accuracy was found to be the same or better than the reference. The authors suggest that these methods may be a promising alternative in model-based clinical treatment for patient state prediction. Still, more research is needed to validate these findings, including in-silico simulations and clinical validation trials.

Model-based subcutaneous insulin for glycemic control of pre-term infants in the neonatal intensive care unit

Hyperglycaemia is a common problem in neonatal intensive care units (NICUs). Achieving good control can result in better outcomes for patients. However, good control is difficult, where poor control and resulting hypoglycaemia reduces outcomes and confounds results. Clinically validated models can provide good control, and subcutaneous insulin delivery can provide more options for insulin therapy for clinicians. However, this combination has only been significantly utilised in adult outpatient diabetes, but could hold benefit for treating NICU infants. This research combines a well-validated NICU metabolic model with subcutaneous insulin kinetics models to assess the feasibility of a model-based approach. Clinical data from 12 very/extremely pre-mature infants was collected for an average study duration of 10.1 days. Blood glucose, interstitial and plasma insulin, as well as subcutaneous and local insulin were modelled, and patient-specific insulin sensitivity profiles were identified for each patient. Modelling error was low, where the cohort median [IQR] mean percentage error was 0.8 [0.3 3.4] %. For external validation, insulin sensitivity was compared to previous NICU cohorts using the same metabolic model, where overall levels of insulin sensitivity were similar. Overall, the combined system model accurately captured observed glucose and insulin dynamics, showing the potential for a model-based approach to glycaemic control using subcutaneous insulin in this cohort. The results justify further model validation and clinical trial research to explore a model-based protocol.

Gene Variants Associated with Transient Neonatal Diabetes Mellitus in the Very Low Birth Weight Infant

BACKGROUND

Transient and permanent neonatal diabetes mellitus (NDM), usually defined as diabetes diagnosed within the first 6 months of life, are rare conditions occurring in 1:90,000-260,000 live births. The origin of NDM is rarely related to type 1 diabetes, but rather to single gene defects.

METHODS

Genetic analysis was performed using targeted parallel sequencing including 323 diabetes genes. Data were filtered by a locally developed program.

RESULTS

A very low birth weight neonate born at 28 weeks postmenstrual age developed diabetes 13 days after birth. The patient was treated with continuous subcutaneous insulin infusion. After 1 month, insulin treatment could be stopped. At 18 months of age, the child was normoglycemic and developing normally. Genetic analysis revealed a novel variant (p.Pro190Leu) in HNF4A, which is located in the ligand binding domain of the transcription factor, and the p.Glu23Lys variant in KCNJ11, which is associated with type 2 diabetes.

CONCLUSION

Here, we describe a novel HNF4A variant associated with transient NDM in a premature infant. We hypothesize that the neonatal phenotype previously described in carriers of HNF4A mutations was modified by the additional variant in KCNJ11 and prematurity.

Insulin kinetics and the Neonatal Intensive Care Insulin-Nutrition-Glucose (NICING) model

BACKGROUND

Models of human glucose-insulin physiology have been developed for a range of uses, with similarly different levels of complexity and accuracy. STAR (Stochastic Targeted) is a model-based approach to glycaemic control. Elevated blood glucose concentrations (hyperglycaemia) are a common complication of stress and prematurity in very premature infants, and have been associated with worsened outcomes and higher mortality. This research identifies and validates the model parameters for model-based glycaemic control in neonatal intensive care.

METHODS

C-peptide, plasma insulin, and BG from a cohort of 41 extremely pre-term (median age 27.2 [26.2-28.7] weeks) and very low birth weight infants (median birth weight 839 [735-1000] g) are used alongside C-peptide kinetic models to identify model parameters associated with insulin kinetics in the NICING (Neonatal Intensive Care Insulin-Nutrition-Glucose) model. A literature analysis is used to determine models of kidney clearance and body fluid compartment volumes. The full, final NICING model is validated by fitting the model to a cohort of 160 glucose, insulin, and nutrition data records from extremely premature infants from two different NICUs (neonatal intensive care units).

RESULTS

Six model parameters related to insulin kinetics were identified. The resulting NICING model is more physiologically descriptive than prior model iterations, including clearance pathways of insulin via the liver and kidney, rather than a lumped parameter. In addition, insulin diffusion between plasma and interstitial spaces is evaluated, with differences in distribution volume taken into consideration for each of these spaces. The NICING model was shown to fit clinical data well, with a low model fit error similar to that of previous model iterations.

CONCLUSIONS

Insulin kinetic parameters have been identified, and the NICING model is presented for glycaemic control neonatal intensive care. The resulting NICING model is more complex and physiologically relevant, with no loss in bedside-identifiability or ability to capture and predict metabolic dynamics.

A C-Peptide-Based Model of Pancreatic Insulin Secretion in Extremely Preterm Neonates in Intensive Care

BACKGROUND

Model-based glycemic control relies on sufficiency of underlying models to describe underlying patient physiology. In particular, very preterm infant glucose-insulin metabolism can differ significantly from adults, and is relatively unstudied. In this study, C-peptide concentrations are used to develop insulin-secretion models for the purposes of glycemic control in neonatal intensive care.

METHODS

Plasma C-peptide, insulin, and blood glucose concentrations (BGC) were retrospectively analyzed from a cohort of 41 hyperglycemic very preterm (median age 27.2 [26.2-28.7] weeks) and very low birth-weight infants (median birth weight 839 [735-1000] g). A 2-compartment model of C-peptide kinetics was used to estimate insulin secretion. Insulin secretion was examined with respect to nutritional intake, exogenous and plasma insulin concentration, and BGC.

RESULTS

Insulin secretion was found to be highly variable between patients and over time, and could not be modeled with respect to age, weight, or protein or dextrose intake. In 13 of 54 samples exogenous insulin was being administered, and insulin secretion was lower. However, low data numbers make this result inconclusive. Insulin secretion was found to increase with BG, with a stronger association in female infants than males (R(2) = .51 vs R(2) = .13, and R(2) = .26 for the combined cohort).

CONCLUSIONS

A sex-based insulin secretion model was created and incorporated into a model-based glycemic control framework. Nutritional intake did not predict insulin secretion, indicating that insulin secretion is a complex function of a number of metabolic factors.

Brain mass estimation by head circumference and body mass methods in neonatal glycaemic modelling and control

INTRODUCTION

Hyperglycaemia is a common complication of stress and prematurity in extremely low-birth-weight infants. Model-based insulin therapy protocols have the ability to safely improve glycaemic control for this group. Estimating non-insulin-mediated brain glucose uptake by the central nervous system in these models is typically done using population-based body weight models, which may not be ideal.

METHOD

A head circumference-based model that separately treats small-for-gestational-age (SGA) and appropriate-for-gestational-age (AGA) infants is compared to a body weight model in a retrospective analysis of 48 patients with a median birth weight of 750g and median gestational age of 25 weeks. Estimated brain mass, model-based insulin sensitivity (SI) profiles, and projected glycaemic control outcomes are investigated. SGA infants (5) are also analyzed as a separate cohort.

RESULTS

Across the entire cohort, estimated brain mass deviated by a median 10% between models, with a per-patient median difference in SI of 3.5%. For the SGA group, brain mass deviation was 42%, and per-patient SI deviation 13.7%. In virtual trials, 87-93% of recommended insulin rates were equal or slightly reduced (Δ<0.16mU/h) under the head circumference method, while glycaemic control outcomes showed little change.

CONCLUSION

The results suggest that body weight methods are not as accurate as head circumference methods. Head circumference-based estimates may offer improved modelling accuracy and a small reduction in insulin administration, particularly for SGA infants.

Nasogastric aspiration as an indicator for feed absorption in model-based glycemic control in neonatal intensive care

BACKGROUND

STAR (stochastic targeted) is a glycemic control model-based framework for critically ill neonates that has shown benefits in reducing hypoglycemia and hyperglycemia. STAR uses a stochastic matrix method to forecast future changes in a patient's insulin sensitivity and then applies this result to a physiological model to select an optimal insulin treatment. Nasogastric aspiration may be used as an indicator to suggest periods of care when enteral feed absorption is compromised, improving the performance of glycemic control. An analysis has been carried out to investigate the effect of poorly absorbed feeds on glycemic control.

METHOD

Clinical data were collected from eight patients on insulin therapy and enteral feed, which included large or significantly milky aspirates. Patients had a median gestational age of 25 weeks and postnatal age of 5.5 days. Virtual patients were created using the NICING model, and insulin sensitivity (SI) profiles were fit. Alternative feed profiles were generated whereby enteral feed absorption was redistributed with time to account for poor feed absorption. The effect of poor feed absorption, as indicated by aspirates, is investigated.

RESULTS

The average percentage change of SI 4 h before a significant aspirate was 1.16%, and 1.49% in the 4 h following the aspirate. No distinct relationship was found between the fractional change in SI and the volume of the aspirate. Accounting for aspirates had a clinically negligible impact on glycemic control in virtual trials.

CONCLUSION

Accounting for aspirates by manipulating enteral feed profiles had a minimal influence on both modeling and controlling glycemia in neonates. The impact of this method is clinically insignificant, suggesting that a population constant for the rate of glucose absorption in the gut adequately models feed absorption within the STAR framework.

Pilot study of a model-based approach to blood glucose control in very-low-birthweight neonates

Background

Hyperglycemia often occurs in premature, very low birthweight infants (VLBW) due to immaturity of endogenous regulatory systems and the stress of their condition. Hyperglycemia in neonates has been linked to increased morbidities and mortality and occurs at increasing rates with decreasing birthweight. In this cohort, the emerging use of insulin to manage hyperglycemia has carried a significant risk of hypoglycemia. The efficacy of blood glucose control using a computer metabolic system model to determine insulin infusion rates was assessed in very-low-birth-weight infants.

Methods

Initial short-term 24-hour trials were performed on 8 VLBW infants with hyperglycemia followed by long-term trials of several days performed on 22 infants. Median birthweight was 745 g and 760 g for short-term and long-term trial infants, and median gestational age at birth was 25.6 and 25.4 weeks respectively. Blood glucose control is compared to 21 retrospective patients from the same unit who received insulin infusions determined by sliding scales and clinician intuition. This study was approved by the Upper South A Regional Ethics Committee, New Zealand (ClinicalTrials.gov registration NCT01419873).

Results

Reduction in hyperglycemia towards the target glucose band was achieved safely in all cases during the short-term trials with no hypoglycemic episodes. Lower median blood glucose concentration was achieved during clinical implementation at 6.6 mmol/L (IQR: 5.5 – 8.2 mmol/L, 1,003 measurements), compared to 8.0 mmol/L achieved in similar infants previously (p < 0.01). No significant difference in incidence of hypoglycemia during long-term trials was observed (0.25% vs 0.25%, p = 0.51). Percentage of blood glucose within the 4.0 – 8.0 mmol/L range was increased by 41% compared to the retrospective cohort (68.4% vs 48.4%, p < 0.01).

Conclusions

A computer model that accurately captures the dynamics of neonatal metabolism can provide safe and effective blood glucose control without increasing hypoglycemia.

Trial Registration

ClinicalTrials.gov registration NCT01419873

Characterisation of the iterative integral parameter identification method

Parameter identification methods are used to find optimal parameter values to fit models to measured data. The single integral method was defined as a simple and robust parameter identification method. However, the method did not necessarily converge to optimum parameter values. Thus, the iterative integral method (IIM) was developed. IIM will be compared to a proprietary nonlinear-least-squares-based Levenberg-Marquardt parameter identification algorithm using a range of reasonable starting values. Performance is assessed by the rate and accuracy of convergence for an exemplar two parameters insulin pharmacokinetic model, where true values are known a priori. IIM successfully converged to within 1% of the true values in all cases with a median time of 1.23 s (IQR 0.82-1.55 s; range 0.61-3.91 s). The nonlinear-least-squares method failed to converge in 22% of the cases and had a median (successful) convergence time of 3.29 s (IQR 2.04-4.89 s; range 0.42-44.9 s). IIM is a stable and relatively quick parameter identification method that can be applied in a broad variety of model configurations. In contrast to most established methods, IIM is not susceptible to local minima and is thus, starting point and operator independent.

Physiological modeling, tight glycemic control, and the ICU clinician: what are models and how can they affect practice?

Critically ill patients are highly variable in their response to care and treatment. This variability and the search for improved outcomes have led to a significant increase in the use of protocolized care to reduce variability in care. However, protocolized care does not address the variability of outcome due to inter- and intra-patient variability, both in physiological state, and the response to disease and treatment. This lack of patient-specificity defines the opportunity for patient-specific approaches to diagnosis, care, and patient management, which are complementary to, and fit within, protocolized approaches.Computational models of human physiology offer the potential, with clinical data, to create patient-specific models that capture a patient's physiological status. Such models can provide new insights into patient condition by turning a series of sometimes confusing clinical data into a clear physiological picture. More directly, they can track patient-specific conditions and thus provide new means of diagnosis and opportunities for optimising therapy.This article presents the concept of model-based therapeutics, the use of computational models in clinical medicine and critical care in specific, as well as its potential clinical advantages, in a format designed for the clinical perspective. The review is presented in terms of a series of questions and answers. These aspects directly address questions concerning what makes a model, how it is made patient-specific, what it can be used for, its limitations and, importantly, what constitutes sufficient validation.To provide a concrete foundation, the concepts are presented broadly, but the details are given in terms of a specific case example. Specifically, tight glycemic control (TGC) is an area where inter- and intra-patient variability can dominate the quality of care control and care received from any given protocol. The overall review clearly shows the concept and significant clinical potential of using computational models in critical care medicine.