249. A metabolomic study of patients with A. baumannii bacteremia

Abstract

Background A. baumannii has become an emerging pathogen of healthcare-associated infection with significant mortality. The present study aimed to identify specific biomarkers to predict patient survival of A. baumannii bacteremia by metabolomics.MethodsFrom July 2011 to November 2014, a total of 60 patients with A. baumannii bacteremia and available blood samples within 4 days of the onset (Day 0) of bacteremia were included for analysis. They were categorized into two groups depending on their survival at Day 14. Metabolomic profiles of the blood specimens collected at Day 0–4 of survival and death groups were compared to identify specific biomarkers to predict patient survival at Day 14. The patients were divided in the training (n=40) and validation (n=20) sets, and the logistic regression-based receiver-operation characteristic (ROC) was used to find the potential markers.ResultsThe Day 14 mortality of the included patients was 20.0% (12/60). The partial least square-discriminate analysis (PLS-DA) scores plot separated the survival and death groups (Figure 1). Thirteen metabolites, L-Isoleucine, Ofloxacin, P-Hydroxybenzaldehyde, Hippurate, Indolelactic acid, Kynurenate, N-Acetyl-L-alanine, Sebacic acid, N-Acetylaspartylglutamic acid, Hematoporphyrin IX, and Urocanic acid reached the statistical significance (p < 0.05) and the accuracies of training and validation sets were greater than 0.8 and 0.6, respectively (Figure 2 and Table 1). Moreover, the Wilcoxon rank sum test results of those metabolites reached the statistical significance (Table 1).Future 1: (A) PLS-DA scores plot for death and survival groups. (B) The loading plot of PLS-DA and the distribution of the thirty important metabolites with VIP values of >1.5. Figure 2. Box plots showing significant changes of the 13 potential markers in the level of metabolites among the death and survival groups. Table 2: Identified metabolites, p value, VIP, and the result of logistic regression. ConclusionMetabolomics had the potential to identify metabolites to predict survival in patients with A. baumannii bacteremia.Disclosures All Authors: No reported disclosures