The Association Between Empirical Antibiotic Treatment and Mortality in Severe Infections Caused by Carbapenem-resistant Gram-negative Bacteria: A Prospective Study.

Abstract

Empirical colistin should be avoided. We aimed to evaluate the association between covering empirical antibiotics (EAT) and mortality for infections caused by carbapenem-resistant gram-negative bacteria (CRGNB). This was a secondary analysis of a randomized controlled trial, including adults with bloodstream infections, pneumonia, or urosepsis caused by CRGNB. All patients received EAT followed by covering targeted therapy. The exposure variable was covering EAT in the first 48 hours. The outcome was 28-day mortality. We adjusted the analyses by multivariable regression analysis and propensity score matching. The study included 406 inpatients with severe CRGNB infections, mostly Acinetobacter baumannii (312/406 [77%]). Covering EAT was given to 209 (51.5%) patients, mostly colistin (n = 200). Patients receiving noncovering EAT were older, more frequently unconscious and dependent, carrying catheters, and mechanically ventilated with pneumonia. Mortality was 84 of 197 (42.6%) with noncovering vs 96 of 209 (45.9%) with covering EAT (P = .504). Covering EAT was not associated with survival in the adjusted analysis; rather, there was a weak association with mortality (odds ratio [OR], 1.37; 95% confidence interval [CI], 1.02-1.84). Results were similar for colistin monotherapy and colistin-carbapenem combination EAT. In the propensity score-matched cohort (n = 338) covering antibiotics were not significantly associated with mortality (OR, 1.42; 95% CI, .91-2.22). Similar results were obtained in an analysis of 14-day mortality. Empirical use of colistin before pathogen identification, with or without a carbapenem, was not associated with survival following severe infections caused by CRGNBs, mainly A. baumannii.