Intensive Insulin Therapy for the Prevention of Infection in the Critically Ill: A Meta-analysis of Prospective, Randomized Trials

Background: Hyperglycemia has been shown to increase rates of infection and mortality in patients with and without diabetes. However, studies examining the role of intensive insulin therapy (IIT) in the critically ill have reached conflicting conclusions. We undertook a meta-analysis to determine the impact of IIT on the incidence of infection.

Objective: To evaluate the impact of IIT on the incidence of infection in critically ill medical and surgical patients.

Methods: Data Sources: we searched 15 international databases including MEDLINE, EMBASE, and CENTRAL from database inception until 5/28/2009 using the terms “intensive insulin therapy” and nine synonyms. Study Selection: prospective randomized controlled trial of critical care or surgical patients were included, where IIT was compared with standard care and at least one type of infection was reported.  IIT was defined as a target blood glucose concentration of 8.3mmol/L or lower. Data Extraction: we extracted general study characteristics, incidence of infection, mortality, and hypoglycemia. Subgroups of medical (<50% surgical patients) and surgical (>50% surgical, perioperative or intraoperative patients) patients were performed for primary endpoints. Infections were analyzed as composite infection, pneumonia, wound, urinary tract, and sepsis. Data Synthesis: 20 studies including one unpublished trial were included in the analysis. Data were analyzed using random-effects model. Heterogeneity was measured with a Cochran Q statistic and I².

Results: Primary endpoints: composite infection was significantly reduced in the IIT group (20 studies, pooled RR=0.80 CI=0.71, 0.90; P=0.0002) with moderate heterogeneity (I²=53.5%),as shown in the figure. Subgroup group analysis of composite infection showed significant reduction in the surgical subgroup (11 studies, pooled RR=0.66, CI=0.57, 0.76; P<0.001) but non-significant difference in the medical subgroup. Sepsis overall was not significantly reduced with IIT. However, subgroup analysis of sepsis showed significant reduction in the surgical subgroup with no heterogeneity (8 studies, pooled RR=0.64, CI=0.48, 0.85; P=0.0024) but non-significant difference in the medical subgroup. Among the subgroups of infections we examined, only pneumonia was significantly reduced (10 studies, RR=0.84, CI=0.72, 0.97, P=0.0212). There was a four fold higher risk of hypoglycemia; however heterogeneity was high (I² =99.4%).

Conclusions: We found that IIT is associated with a significant reduction in overall infection. Subgroup analysis revealed reduction in combined infection and sepsis in the surgical subgroup but not in the medical subgroup. The risk of hypoglycemia must be taken into consideration when employing IIT for reducing infection.