Background: Extreme glucose values in critically ill patients have been studied and shown to be predictors of mortality. In addition, glucose fluctuation is also proposed to have been a determinant of poor outcome. In our study, we attempt to understand whether there is a relation between the fluctuations of glucose in critically ill patients to their mortality.

Methods: Data was obtained from "Medical Information Mart for Intensive Care III" database between 2001 and 2012. Fluctuation was measured as coefficient of variance of glucose (CVG). Patients were split into four groups based on the CVG of glucose levels during their ICU stay, namely Group 1: <25%, Group 2: 25%-50%, Group 3: 50%-100% and Group 4: >100%. Outcomes were measured at 3-day mortality interval and the relative risk was compared across the four groups. Multinomial logistic regression with length of stay, prednisone administration and Simplified Acute Physiology Score (SAPS) II as variables were run to assess the relative risk ratios (RRR) of increased CVG with mortality.

Results: A total of 45,302 adult ICU admissions were considered, 82.1% (n=37,192) had no diabetes, 17.6% (n=7,952) were type 2 diabetic (T2DM) and 0.04% (n=168) were type 1 diabetic (T1DM). The RRR for 3-day mortality, among the nondiabetics compared to the base outcome (Group 1) was 1.51 (p<0.0001) in Group 2, 3.76 (p<0.0001) in Group 3 and 2.17 (p=0.10) in Group 4. Among T2DM, the RRR for 3-day mortality was 1.04 (p=0.7) in Group 2, 2.97 (p<0.0001) in Group 3, 14.03 (p=0.025) in Group 4. Due to the small number of observations for T1DM, they were excluded.

Conclusion: In nondiabetics there was higher mortality at a lower CVG as compared to the diabetics. In diabetics the increased mortality was observed at slightly higher CVG. In patients admitted to critical care units, increased fluctuation in glucose during their ICU admission is an independent predictor of mortality and additional caution should be taken to avoid wide fluctuations.


N. Kanaparthy: None. S. Elsa Mathews: None. K. Kallur: None.

Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at