Reappraisal with meta-analysis of bacteremia, endotoxemia, and mortality in gram-negative sepsis.

Abstract

Among patients with suspected sepsis, endotoxemia is variably present in association with gram-negative bacteremia. A total of 738 patients with suspected sepsis from 11 studies could be classified into four groups: 131 (18%) patients had both endotoxemia and gram-negative bacteremia (group 1), 87 (12%) had only gram-negative bacteremia (group 2), and 143 (19%) had only endotoxemia (group 3); in 377 (51%) patients neither could be detected (group 4). By the statistical techniques of meta-analysis, the fatality risk for patients with either endotoxemia or gram-negative bacteremia or both was estimated with group-specific case fatality rates from these studies and expressed as an odds ratio (OR; 95% confidence interval [95% CI]) versus patients with these factors absent. This risk was increased marginally when endotoxemia was detected without gram-negative bacteremia (OR, 2.3; 95% CI, 1.3 to 4.0) or the converse (OR, 2.0; 95% CI, 1.0 to 3.8), in contrast to the striking increase when both endotoxemia and gram-negative bacteremia were detected (OR, 6.3; 95% CI, 4.0 to 10.0). Alone, neither endotoxemia nor gram-negative bacteremia is a strong predictor of outcome in patients with suspected sepsis. However, in combination, the two identify a subpopulation with a substantially increased risk of mortality.