Glucose variability and survival in critically ill children: Allostasis or harm?

Objectives: To assess whether individual blood glucose variability in critically ill children is associated with increased mortality and to define the temporal patterns of blood glucose variability during critical illness in children. Design: Retrospective cohort study. Setting: A 20-bed pediatric intensive care unit in a children's hospital. Patients: Patients aged 0-20 yrs and with at least 12 blood glucose measurements taken within the first 72 hrs of pediatric intensive care unit admission. Interventions: None. Measurements and Main Results: A total of 101 eligible patients had 3,144 measured blood glucose concentrations with 16% mortality. Nonsurvivors had higher median blood glucose concentrations (129 mg/dL vs. 118 mg/dL, p < .01), more hyperglycemia (blood glucose >200 mg/dL) (88% vs. 59%, p < .05), and more hypoglycemia (blood glucose <60 mg/dL) (56% vs. 15%, p < .01) than survivors. The mean blood glucose range (257 mg/dL vs. 185 mg/dL, p < .01) and the blood glucose variability (63 mg/dL vs. 45 mg/dL, p = .02) were greater in nonsurvivors compared with survivors. Blood glucose variability tertiles were proportionately associated with increasing mortality: 6% vs. 15% vs. 27% (p = .07). Compared with survivors, daily blood glucose variability was significantly higher in nonsurvivors during the first 48 hrs of admission and after 1 wk of admission. After controlling for confounders, individual blood glucose variability was associated with higher pediatric intensive care unit mortality for each mg/dL of blood glucose concentration (adjusted odds ratio, 1.03; 95% confidence interval, 1.01-1.05). Conclusions: Glucose variability is common in critically ill children and is associated with increased mortality. Whereas early alterations in blood glucose may represent allostasis, later fluctuations in blood glucose may represent an alteration of autoregulation with resulting higher mortality. Control of variability may need to be incorporated into glycemic control regimens. (Pediatr Crit Care Med 2010; 11: 707-712)