Identification of metabolomics-based prognostic prediction models for ICU septic patients

Sepsis-related global mortality remains unacceptably high in intensive care units. Identifying the various molecular processes between survival and death in septic patients may assist in better treatment. Accurate prognostic evaluation of sepsis is an essentially unmet need. This study analyzed the metabolite changes in plasma between healthy controls and septic patients, as well as between survival and dead septic patients using liquid chromatography/mass spectrometry. Univariate and multivariate analyses were applied to identify differential metabolites. The differential metabolites and clinical indicators within 24 h after sepsis diagnosis were run through multivariate logistic regression models to determine the 28-day, hospital, and 90-day septic mortality prediction models. The results suggested markedly changed amino acids metabolism in septic patients compared to healthy controls; 10, 4, and 22 primary differential metabolites related to amino acid and fatty acid metabolisms were identified in the survival and death groups at 28-day, hospital, and 90-day, respectively. Further, we found that model 1 (indoleacetic acid, 3-methylene-indolenine, heart rate, respiratory support, and application of pressure drugs), model 2 (lymphocyte count, alkaline phosphatase, SOFA, and L-alpha-amino-1H-pyrrole-1-hexanoic acid), and model 3 (dopamine, delta-12-prostaglandin J2, heart rate, respiratory support, and application of pressure drugs) could predict 28-day, hospital, and 90-day mortality of sepsis with a sensitivity of 75.51%, 73.58%, and 83.33%, specificity of 78.72%, 72.09%, and 78.57%, and the area under the receiver operating characteristic curve of 0.881, 0.830, 0.886, respectively. Thus, this research presents three multiple-biomarker-based prognostic models for 28-day, hospital, and 90-day mortality septic patients and could be used to guide sepsis treatment.