High Glucose Variability Is Associated with Mortality in Critically Ill Trauma Patients

Introduction: Normoglycemia maintained by tight glucose control (TGC) protocols are reported to reduce morbidity and mortality in critically ill surgical patients. Glucose variability (GV) has been used to characterize glycemic control and efficacy of TGC, but the definition and methods for determining GV have not been standardized. We inquired how various descriptions of glucose variability were associated with mortality in our trauma population, and hypothesized that these descriptions of GV are each associated with mortality in distinct ways. Further, we hypothesize that a single method of describing glucose variability is superior and that identifying this measure of GV may predict mortality in critically ill trauma patients. Methods: This was a single-center retrospective review of trauma patients age 18 years and older admitted to the Surgical Intensive Care Unit at a Level I Trauma Center between January 2005 through December 2007 who were managed with a TGC protocol. Subjects were identified by trauma registry query and blood glucose values were obtained through the hospital database. Patients with an ISS<9 and ICU LOS<3 days were excluded. Data collected included demographics, diagnoses, mechanism of injury, medical history including diabetes, length of ICU stay, clinical course and disposition. Glucose variability was described as the following: standard deviation and coefficient of variability of all recorded glucose values, glycemic lability index (defined as the summation of glucose changes squared divided by hour and week), triangular index (the mode divided by total number of glucose measurements), successive changes in glucose levels (greatest increase, greatest decrease, greatest absolute change), and mean amplitude of glycemic excursion (the mean of absolute values of any change between consecutive glucose values that are greater than 1 SD). The primary endpoint was in-hospital mortality. Multivariate logistic regression analysis was used, controlling for GV indices, ISS, TRISS, and presence or absence of traumatic brain injury. Results: Five-hundred and forty five subjects were included. Most were male (69%) with mean ISS of 25±11, TRISS 0.7±0.3. Average ICU length of stay was 13 days (median 8). Mean ventilator days were 8 (median 4). A total of 98,864 glucose measurements were obtained, with 14 mean (mode 7) daily glucose measurements per subject. High standard deviation (p<0.006, OR 1.106), and coefficient of variability (p<0.002, OR 1.029) were associated with mortality. Low triangular index (p<0.04, OR 0.857) was associated with mortality. Glycemic lability index, successive changes in glucose levels, and mean amplitude of glycemic excursion were not significantly associated with mortality. Conclusions: High glucose variability is associated with mortality in critically ill trauma patients. Coefficient of variability was superior to other methods of describing glucose variability and was most highly associated with mortality. It is unclear from these data whether glucose variability is associated with underlying patient disease physiology or a tight glycemic control protocol.