Glycaemic variability, infections and mortality in a medical-surgical intensive care unit.

Objective: In critically ill patients, glycaemic variability (GV) was reported as a better predictor of mortality than mean blood glucose level (BGL). We compared the ability of different GV indices and mean BGLs to predict mortality and intensive care unit-acquired infections in a population of ICU patients. Design, setting and participants: Retrospective study on adult ICU patients with >= three BGL measurements. GV was assessed by SD, coefficient of variation (CV) and mean amplitude of glycaemic excursion (MAGE), and by one time-weighted index, the glycaemic lability index (GLI), and compared with mean BGL. We studied 2782 patients admitted to the 12-bed medical-surgical ICU of a teaching hospital from January 2004 until December 2010. Main outcome measures: Logistic regression analyses were performed to assess the association between GV and ICU mortality and ICU-acquired infections. The areas under receiver operating characteristic curves were calculated to compare the discriminatory ability of GV and mean BGL for infections and mortality. Results: Mortality was 16.6%, and 30% of patients had at least one infection. Patients with infections or diabetes or who were treated with insulin had a higher mean BGL and GV than other patients. GLI, SD, CV and MAGE were significantly associated with infections and mortality; mean BGL was not. Quartiles of increasing GLI were independently associated with higher mortality and an increased infection rate. Patients in the upper quartile of mean BGL and GLI had the strongest association with infections (odds ratio, 5.044 [95% Cl, 1.695-15.007]; P=0.004). Conclusion: High GV is associated with higher risk of ICU-acquired infection and mortality.