Development of blood glucose control for extremely premature infants

Abstract

Extremely premature neonates often experience hyperglycaemia, which has been linked to increased mortality and worsened outcomes. Insulin therapy can assist in controlling blood glucose levels and promoting needed growth. This study presents the development of a model-based stochastic targeted controller designed to adapt insulin infusion rates to match the unique and changing metabolic state and control parameters of the neonate. Long-term usage of targeted BG control requires successfully forecasting variations in neonatal metabolic state, accounting for differences in clinical practices between units, and demonstrating robustness to errors that can occur in everyday clinical usage. Simulation studies were used to evaluate controller ability to target several common BG ranges and evaluate controller sensitivity to missed BG measurements and delays in control interventions on a virtual patient cohort of 25 infants developed from retrospective data. Initial clinical pilot trials indicated model performance matched expected performance from simulations. Stochastic targeted glucose control developed using validated patient-specific virtual trials can yield effective protocols for this cohort. Long-term trials show fundamental success, however clinical interface design appears as a critical factor to ensuring good compliance and thus good control.

Introduction

Metabolic homeostasis in the extremely premature infant is often compromised by immaturity of control systems. Up to 32–80% of low birth weight infants experience hyperglycaemia with glucose levels exceeding thresholds of 6.9–13.9 mmol/L during the neonatal period [1], [2], [3], [4], [5]. The risk of hyperglycaemia is at least 18 times greater in infants with birth weight less and 1000 g compared to infants weighing greater than 2000 g [6]. Hyperglycaemia has been linked to worsened outcomes. Associated morbidities include osmotic diuresis, electrolyte imbalance, intraventricular haemorrhage, sepsis, and increased ventilator dependence, retinopathy of prematurity, hospital length of stay and mortality [2], [3], [4], [5], [7], [8], [9]. High rates of proteolysis are also common in low birth weight infants, reducing muscle mass and inhibiting growth [10].

The known mechanisms responsible for hyperglycaemia specific to extremely premature infants are related to their reduced ability to produce insulin [11]; defective beta-cell processing of pro-insulin (which is 10–16 times less active than insulin) to insulin [12]; an inability to suppress hepatic glucose production in response to glucose infusion [13]; and, finally, a decreased uptake of glucose secondary to a limited mass of insulin-sensitivity tissues (e.g. muscle and adipose tissue) [14]. In addition to these factors, it has been shown that preterm infants can mount a hormonal response to stress similar to older critically ill patients [15]. Inhibiting the physiological response to reduce increased glycaemic levels are factors such as increased insulin resistance, absolute or relative insulin deficiency, and drug therapy [16], [17], [18], [19].

Blood glucose control for the neonate poses several challenges that differ from the adult critical care case. Blood volumes in preterm infants are relatively small [6]. Thus, the number of blood glucose measurements must be optimised to a minimum useful number to conserve volume and restrict opening incubator doors, which may affect the neonate's hydration status [20]. Endogenous energy substrates are very low in preterm infants at birth [21]. Thus, these infants must be constantly fed to provide enough energy for basal requirements in addition to growth. In contrast, adults can tolerate periods of reduced caloric intake. Less saturation of the insulin-stimulated glucose uptake pathway has also been reported in neonates [22], unlike the saturation in adults [23]. Finally, also unlike the adult case, growth is a major goal of neonatal care. Thus, the anabolic effects of insulin are of relatively higher importance that in adults [10].

A small number of prospective trials have used insulin infusions to treat hyperglycaemia and/or promote growth [16], [24], [25], [26], [27], [28], [29], [30], [31], [32]. All reported insulin infusion trials have used either protocols that fixed insulin dosing to weight or other factors [25], or clinician judgment to determine insulin infusion rates. Positive outcomes of insulin infusion have been reduced proteolysis [10], [33], [34], improved glucose tolerance, improved caloric intake and weight gain [16], [24], [26], [27], [28], [29], [31]. Negative reports of hyperinsulinaemia include hypoglycaemia and possible metabolic acidosis due to excessive carbohydrate oxidation [34], [35].

Persistent low blood glucose concentrations can reduce cerebral development and lead to long-term neurological deficiencies [36]. The upper limit for clinically desirable blood glucose concentration is also subject to debate [11]. Thus, glucose management goals vary widely between insulin therapy studies [30], [35] and it is likely the desired target range for glucose may change in the future. Similarly, Alsweiler et al. [37] demonstrated a wide range of responses from different clinical units when questioned on glycaemic control targets. Hence, a control system needs to handle different glucose targets to comply with local practices.

It is well known that neonatal response to insulin and glucose infusions exhibits great heterogeneity [11], [24], which would render fixed protocols ineffective because they do not adopt well to different patients or the evolution of metabolic response over time within an individual patient. Model-based blood glucose control may provide more optimised care by adapting in real-time to identified parameters representing the current metabolic state of the infant, and using this information to drive insulin dosing. This metabolic information can be combined with a controller utilising model predictions to achieve targeted blood glucose control. This approach has been validated in adult critical care studies [38]. However, sudden changes in patient condition independent of metabolic state indicate limits on model-based controller actions are required to maximise safety and control performance [39]. Insulin sensitivity changes can be captured and characterised using stochastic models of insulin sensitivity variability [40], [41], [42], specifically to quantify the level of hypoglycaemia risk and adjust control actions accordingly. This approach creates a targeted, model-based control system that uses stochastic forecasting to guarantee the risk of hypoglycaemia for any intervention.

Virtual trials offer the opportunity to explore control strategies in simulation to achieve the goals of maximising time within a desired glycaemic control band, which may vary between clinical units, whilst minimizing the number of hypoglycaemic episodes [38], [39]. Long-term clinical usage also requires robustness to missed BG measurements, control interventions, and delays in adjusting insulin dosing. The model-based controller used in this study tracks insulin sensitivity in real-time by fitting to available clinical data, and these errors may thus cause insulin dosing to be driven by an estimate of insulin sensitivity based on artificial effects rather than physiology. The effect of these errors can be evaluated in a clinically validated simulation environment to assess performance prior to clinical implementation. Short-term clinical trials are used to confirm model prediction accuracy and control efficacy. Pilot long-term clinical trials are presented to assess the agreement between simulation results and real-world outcomes, as well as to highlight some of the difficulties experienced in transitioning from the fully compliant simulation environment to real-world clinical usage.