Objectives: To determine the effect of glycemic variability, assessed by the standard deviation of each patient's mean glucose level, on mortality in a population of critically ill adult patients.
Design: Retrospective review of a large cohort of prospectively evaluated patients.
Setting: Fourteen-bed medical surgical adult intensive care unit of a university affiliated community hospital.
Patients: Three thousand two hundred fifty-two patients consecutively admitted between October 1999 and October 2007 with at least three venous glucose samples.
Interventions: None.
Measurements and main results: The mean (sd) Acute Physiology and Chronic Health Evaluation II score of the 3252 patients was 20.0 (8.9) and their mortality was 24.4%, ranging from 18.1% among patients with mean glucose level 70 mg/dL to 99 mg/dL to 35.9% among patients with mean glucose level 180+ mg/dL. The relationship between glycemic variability and mortality was strongest in the euglycemic range. For the 410 patients with mean glucose level 70 mg/dL to 99 mg/dL, mortality ranged from 5.9% in the first quartile of glycemic variability to 30.1% in the fourth; for the 1031 patients with mean glucose level 100 mg/dL to 119 mg/dL the corresponding range was 9.7% to 31.0%. Mortality among patients in the entire cohort with the lowest quartile of glycemic variability was 12.1%, increasing to 19.9%, 27.7%, and 37.8% in the second, third, and fourth quartiles. Intensive care unit length of stay was shorter among patients in the first quartile compared with those in the other three (p < .001).
Conclusions: This study demonstrates that increasing glycemic variability conferred a strong independent risk of mortality in this heterogeneous population of critically ill patients. Previously published interventional studies of glycemic control may be reinterpreted using the metric of glycemic variability. Measures to ensure a low degree of glycemic variability may improve outcomes in intensive care unit's implementing glycemic control. Finally, ongoing and future investigations should consider including this new metric in their study design.