Glycemic control in critically ill patients with or without diabetes

BACKGROUND: Early randomized controlled trials have demonstrated the benefits of tight glucose control. Subsequent NICE-SUGAR study found that tight glucose control increased mortality. The optimal glucose target in diabetic and nondiabetic patients remains unclear. This study aimed to evaluate the relationship between blood glucose levels and outcomes in critically ill patients with or without diabetes. METHODS: This was a retrospective analysis of the eICU database. Repeat ICU stays, ICU stays of less than 2 days, patients transferred from other ICUs, those with less than 2 blood glucose measurements, and those with missing data on hospital mortality were excluded. The primary outcome was hospital mortality. Generalised additive models were used to model relationship between glycemic control and mortality. Models were adjusted for age, APACHE IV scores, body mass index, admission diagnosis, mechanical ventilation, and use of vasopressor or inotropic agents. RESULTS: There were 52,107 patients in the analysis. Nondiabetes patients exhibited a J-shaped association between time-weighted average glucose and hospital mortality, while this association in diabetes patients was right-shifted and flattened. Using a TWA glucose of 100 mg/dL as the reference value, the adjusted odds ratio (OR) of TWA glucose of 140 mg/dL was 3.05 (95% confidence interval (CI) 3.03-3.08) in nondiabetes and 1.14 (95% CI 1.08-1.20) in diabetes patients. The adjusted OR of TWA glucose of 180 mg/dL were 4.20 (95% CI 4.07-4.33) and 1.49 (1.41-1.57) in patients with no diabetes and patients with diabetes, respectively. The adjusted ORs of TWA glucose of 80 mg/dL compared with 100 mg/dL were 1.74 (95% CI 1.57-1.92) in nondiabetes and 1.36 (95% CI 1.12-1.66) in patients with diabetes. The glucose ranges associated with a below-average risk of mortality were 80-120 mg/dL and 90-150 mg/dL for nondiabetes and diabetes patients, respectively. Hypoglycemia was associated with increased hospital mortality in both groups but to a lesser extent in diabetic patients. Glucose variability was positively associated with hospital mortality in nondiabetics. CONCLUSIONS: Time-weighted average glucose, hypoglycemia, and glucose variability had different impacts on clinical outcomes in patients with and without diabetes. Compared with nondiabetic patients, diabetic patients showed a more blunted response to hypo- and hyperglycemia and glucose variability. Glycemic control strategies should be reconsidered to avoid both hypoglycemia and hyperglycemia.