To evaluate the feasibility and safe operationalization of a pediatric glycemic control protocol in the setting of a general pediatric intensive care unit in a developing country.
Prospective, observational cohort study carried out over 12 months.
Fourteen-bed pediatric intensive care unit in Brazil.
Children requiring mechanical ventilation with at least one organ system dysfunction were included.
Glucose was monitored and insulin used for persistent hyperglycemia (glucose >140 mg/dL [7.8 mmol/L] for at least two observations separated by at least a 1-hr interval), with a target glucose during insulin use of 60-140 mg/dL (3.3-7.8 mmol/L).
Out of 410 admissions, 144 children met the criteria for applying the protocol. One hundred fourteen of 144 (79%) children had at least one peak glucose level that was hyperglycemic, but only 44 (31%) children required insulin. Insulin infusion was most frequently started on day 1 (61%), with a glucose level at the time of 229 ± 79 mg/dL (12.7 ± 4.4 mmol/L). The mean glucose level after 6 hrs of insulin was 172 ± 87 mg/dL (9.6 ± 4.8 mmol/L), and the time to achieve the target glucose range was 9.5 (2-20) hrs (median [interquartile range]). The overall duration of insulin was 24.5 (10-48) hrs, and the average dose required was 0.06 ± 0.03 U/kg/hr. In the whole series, the peak glucose level was 202 ± 93 mg/dL (11.2 ± 5.2 mmol/L), with no difference between survivors and nonsurvivors. There was no difference in mortality when different glucose bands were considered and no association between glucose level and mortality. The overall rate of hypoglycemia (glucose <40 mg/dL [2.2 mmol/L]) was 8.3%, and it was more common in those receiving insulin (20% vs. 3%, p < .05).
Hyperglycemia is frequent in critically ill children managed in a pediatric intensive care unit in a developing country. Using a glycemic control protocol, one-third of these children required insulin, but attendants should be aware of a significant risk of hypoglycemia in this setting. Based on these data, a trial to detect a 20% relative reduction in mortality (power 90%, p = .05) associated with insulin in a similar population would need to screen >10,000 patients.