Objectives: To determine prevalence and prognostic significance of hyperglycaemia among critically ill non-diabetic children. To evaluate which patients will benefit from insulin treatment.

Methods: Retrospective study using blood glucose levels (9015 values, 923 patients) in our paediatric intensive care unit (PICU) from January 2003 to December 2005. Patients with diabetic ketoacidosis were excluded. Overall PICU mortality was 3.7%. Hyperglycaemia was defined at 6.1 mmol/l and different cutoff values (6.1, 8.3 and 11.1 mmol/l) were analysed for glycaemia at admission (GLUC). Sustained hyperglycaemia was evaluated with the area under the curve normalised per hour (48 h AUC/h) for the first 48 h. The prevalence of hypoglycaemia (<3 mmol/l), hyperglycemia and PICU death were analysed.

Results: Through the use of different cutoff values (>6.1, >8.3 and >11.1 mmol/l), the prevalence of hyperglycaemia at admission was 62.8%, 28.9% and 11%; mortality was 5.2%, 8.1% and 14.9%, respectively, correlated with cutoff values (r  =  0.95, p<0.05). The prevalence of hypoglycaemia at admission was 0.9%. 48 h AUC was computed in 747 children (30 deaths). The prevalence of hyperglycaemic 48 h AUC values was 70.4%, 21.8% and 4.1%, with a respective mortality of 5.1%, 8.9% and 21.4% (r  =  0.97, p<0.03). For those with high GLUC and high 48 h AUC (>11.1 mmol/l) mortality was high (31.5%), but it decreased to 5.5% when 48 h AUC decreased spontaneously to values <8.3 mmol/l per hour. Finally, when patients with severe neurological lesions where excluded, increased mortality was observed only for GLUC and 48 h AUC higher than 11.1 mmol/l.

Conclusions: Hyperglycaemia at admission and sustained hyperglycaemia are highly correlated with mortality in PICU. Children who will have benefit from insulin therapy represent 3% of our population.