Glucose controlThe Glucosafe system for tight glycemic control in critical care: A pilot evaluation study
Introduction
Hyperglycemia in critical care has been associated with increased mortality [1], [2], severe infection [3], sepsis and septic shock [4], myocardial infarction [5], and multiple-organ failure [6]. Tight glycemic control (TGC) to concentrations less than 6.1 to 7.75 mmol/L and reduced blood glucose (BG) variance can benefit survival and clinical outcomes [7], [8], [9], [10]. A great variety of TGC protocols has been published; common glycemic targets fall into the tight range of 4.4 to 6.1 mmol/L used in the Leuven study from 2001 [7] or the slightly wider range of 4.4 to 7.75 used by other studies [8], [9]. However, consistent control of mean BG concentrations and variance has proven elusive [11], [12], [13]. More important, TGC protocols have been shown to significantly increase the frequency of mild (BG <3.5 mmol/L) and severe (BG <2.2 mmol/L) hypoglycemic episodes, putting patient safety at risk [14].
Model-based methods can adapt to a patient's individual response to insulin and nutrition, thus enabling patient-specific treatment. “Glucosafe” is a metabolic model that calculates the time-evolving, patient-specific insulin sensitivity from glucose measurements, insulin treatment, and nutritional inputs [15]. The model takes insulin saturation effects and reduced rates of gut absorption of glucose in critical care patients into account. Glucosafe's accuracy in predicting BG several hours ahead was studied retrospectively using data from a neuro and trauma intensive care cohort and a general intensive care cohort [16]. The model was also studied in comparison to other metabolic models [17], including the minimal model. Using this model to build a system, prospective clinical testing is needed to evaluate system safety, achievement of target BG and variance, and convenience of use on a routine basis.
This article presents a decision support system (DSS) for TGC based on the Glucosafe model and a pilot study of the system's use on a small critical care cohort for a limited amount of time. Four penalty functions were developed to balance different combinations of insulin, carbohydrate input, and route of feeding against model-predicted BG outcome. The recommended intervention is the combination with the lowest sum of penalty scores. For convenience of the clinical caretakers, the system provided a graphical interface for data input and output and an interactive function that allowed modulating the nutrition part of the advice. The objectives of this pilot study were to assess the suitability of the penalty functions to achieve TGC, to ascertain the system's safety against hypoglycemia, and to gain valuable feedback from the hands-on experience of the nurses with the system.
Section snippets
Decision support system
The Glucosafe model was implemented in a computerized DSS using the equations specified in Ref. [15], with 2 exceptions: (1) the posthepatic insulin clearance was assumed to be constant at 1.57 mL/min per kilogram body mass, and (2) the endogenous insulin appearance rate after the first pass through the liver was assumed to be constant at 1.46 U/h for patients not diagnosed with type 1 diabetes.1 Reduced
Cohort
Table 1 lists baseline characteristics, hospital admission diagnoses, and clinical variables of this cohort. Sexes were equally distributed. Medians and interquartile ranges (IQRs) of APACHE II score and Glasgow Coma Score were 12.5 (IQR, 7.5-16.25) and 10.0 (IQR, 6.75-11.25), respectively. Of 10 patients, 7 were included in the study within 4 days after admission to intensive care. Six patients received steroids during the trial. The cohort included 2 patients with type 2 diabetes and 1
Discussion
Glucosafe significantly improved glycemic control in terms of low BG variance, which has been shown to be an important factor for clinical outcomes [10]. The mean BG values of 7.0 ± 1.1 mmol/L overall and 6.7 ± 1.2 mmol/L after the mean time to target compared well with the 6.0 ± 1.5 mmol/L achieved by the Specialised Relative Insulin Nutrition Tables (SPRINT) study [9] and are lower than the 7.3 ± 3.4 mmol/L reported by Krinsley [28], which were studies that have shown a reduction in
Acknowledgments
This work was partially supported by a grant awarded by the Programme Commission on Nanoscience, Biotechnology, and IT under the Danish Council for Strategic Research.
References (32)
Association between hyperglycemia and increased hospital mortality in a heterogeneous population of critically ill patients
Mayo Clinic Proceedings
(2003)- et al.
Stress hyperglycaemia and increased risk of death after myocardial infarction in patients with and without diabetes: a systematic overview
Lancet
(2000) Effect of an intensive glucose management protocol on the mortality of critically ill adult patients
Mayo Clin Proc
(2004)- et al.
A glucose-insulin pharmacodynamic surface modeling validation and comparison of metabolic system models
Biomed Signal Process Control
(2009) - et al.
Prediction validation of two glycaemic control models in critical care
- et al.
A probabilistic approach to glucose prediction and insulin dose adjustment: description of metabolic model and pilot evaluation study
Comput Methods Programs Biomed
(1994) - et al.
A model of the endogenous glucose balance incorporating the characteristics of glucose transporters
Comput Methods Programs Biomed
(2000) - et al.
A new predictive equation for resting energy expenditure
Am J Clin Nutr
(1990) - et al.
Caloric intake in medical ICU patients: consistency of care with guidelines and relationship to clinical outcomes
Chest
(2003) - et al.
Hyperglycemia: an independent marker of in-hospital mortality in patients with undiagnosed diabetes
J Clin Endocrinol Metab
(2002)
Hyperglycaemia and infection: which is the chicken and which is the egg?
J Parenter Enteral Nutr
Stress-hyperglycemia, insulin and immunomodulation in sepsis
Intensive Care Med
Intensive insulin therapy protects the endothelium of critically ill patients
J Clin Invest
Intensive insulin therapy in the critically ill patients
N Engl J Med
Implementation and evaluation of the SPRINT protocol for tight glycaemic control in critically ill patients: a clinical practice change
Critical Care
Variability of blood glucose concentration and short-term mortality in critically ill patients
Anesthesiology
Cited by (24)
Digital twins in critical care: What, when, how, where, why?
2021, IFAC-PapersOnLineGlycemic control in the intensive care unit: A control systems perspective
2019, Annual Reviews in ControlCitation Excerpt :It is more physiologically relevant than the ICU Minimal Model with specific pharmacodynamics terms for each form of glucose removal. However, it demonstrated limited performance in pilot GC trials in regular and neuro ICU patients (Pielmeier, Andreassen, Juliussen, et al., 2010, Pielmeier et al., 2012). The Cambridge Model is highly physiologically relevant (Hovorka et al., 2008).
Stochastic Model Predictive (STOMP) glycaemic control for the intensive care unit: Development and virtual trial validation
2015, Biomedical Signal Processing and ControlCitation Excerpt :This type of controller was chosen due to the flexibility of cost functions in allowing the dynamic response of the controller to be easily tuned. MPC has also been used for glycaemic control with a different model [32–35]. This article presents a Stochastic Model Predictive (STOMP) GC protocol that uses a low error, infrequently measured, BG signal to control the BG levels in adult ICU patients while providing greater flexibility than STAR.
Evaluating modifications to the Glucosafe decision support system for tight glycemic control in the ICU using virtual patients
2014, Biomedical Signal Processing and ControlCitation Excerpt :The rationale for the shapes of the functions has previously been described [6]. The shape and scaling of the penalty functions initially used in this study are identical to the penalty functions used during the data collection [10,11]. The virtual patients are based on insulin sensitivity profiles from patients from a previous clinical study of Glucosafe [11].
Survey about diffusion and adoption of glycaemic controller in European intensive care units
2014, IFAC Proceedings Volumes (IFAC-PapersOnline)Virtual testing of decision support for tight glycemic control in the ICU
2012, IFAC Proceedings Volumes (IFAC-PapersOnline)