Objective: To explore the relationship between the levels and variability of blood glucose and the prognosis of critical patients., Methods: A prospective study was conducted. Blood glucose monitoring and prognosis observation were performed for the adult nondiabetic patients admitted in intensive care unit (ICU) from June 2011 to January 2012. Blood glucose monitoring terminal was 72 hours after admitting in ICU, prognosis was observed for 28 days after the end of turning into ICU. Acute physiology and chronic health evaluation II(APACHE II) scores when transferred into ICU and blood glucose variability [standard deviation (SD) of blood glucose, mean absolute blood glucose fluctuation amplitude (MAGE) and glycemic instability index (GLI)] were calculated. Patients were divided into death group and survival group according to the outcome, and the APACHE II score, mean blood glucose and blood glucose variability were compared between the two groups. Patients were divided into different groups based on the blood glucose, and the APACHE II score, blood glucose variability and 28-day mortality were compared among groups., Results: Total 85 cases were enrolled. Compared with survivor group (n=58), in death group (n=27), APACHE II score (28.9±6.6 vs. 23.8±5.9), mean blood glucose (11.9±2.9 mmol/L vs. 9.4±1.8 mmol/L), SD of blood glucose (3.7±1.6 mmol/L vs. 2.4±1.0 mmol/L), MAGE (0.86±0.46 mmol/L vs. 0.54±0.25 mmol/L) and GLI (255.9±232.7 vs. 111.7±110.9) were increased (all P<0.05). SD of blood glucose (4.3±1.4 mmol/L), MAGE (1.1±0.4 mmol/L), GLI (345.3±210.3) and 28-day mortality (63.6%) in blood glucose >11.1 mmol/L group (n=22) were higher than those in ≤11.1 mmol/L group (n=63, 2.3±0.9 mmol/L, 0.5±0.2 mmol/L, 91.9±91.2, 20.6%, respectively, all P<0.05) and 7.8-11.1 mmol/L group (n=52, 2.3±0.9 mmol/L, 0.5±0.2 mmol/L, 85.2±66.4, 25.0%, respectively, all P<0.05). There were no significant differences between 7.8-11.1 mmol/L group and <7.8 mmol/L group (n=11) in SD of blood glucose (2.0±0.9 mmol/L), MAGE (0.5±0.3 mmol/L), GLI (123.8±166.7) and 28-day mortality (0, all P>0.05)., Conclusion: Blood glucose variability is associated with critical patient's 28-day mortality, and may predict mortality as good as APACHE II score.