Prognostic biomarkers for sepsis mortality based on the literature and LC-MS-based metabolomics of sepsis patients.

Objectives: The management of sepsis, a potentially lethal overreaction to infection, is limited by the lack of prognostic tools to guide its treatment. Our aim is to identify a novel metabolic biomarker panel for predicting sepsis mortality based on a literature review and liquid chromatography-mass spectrometry (LC-MS)-based metabolomics.

Methods: In the literature, we found metabolomics biomarkers reported to predict sepsis mortality. We determined the classifications, reported frequency, and KEGG pathway enrichment of these markers. Using serum samples from 20 sepsis survivors and 20 non-survivors within 28 days after admission to the intensive care unit (ICU), we performed LC-MS-based metabolomics. Based on the literature review and metabolomics, a prognostic biomarker panel for sepsis was identified and its area under the curve (AUC) values was assessed.

Results: Kynurenate, caffeine, and lysoPC 22:4 were selected as a prognostic biomarker panel for sepsis. The panel had an area under the curve (AUC) of 0.885 (95% CI, 0.694-1) evaluated by linear support vector machine (SVM) and 0.849 (0.699-1) by random forest (RF), which was higher than that of the Sequential Organ Failure Assessment (SOFA). A combination of kynurenate, caffeine, and lysoPC 22:4 and SOFA provided the best discriminating performance, with AUCs of 0.961 (0.878-1) for SVM and 0.916 (0.774-1) for RF.

Conclusions: The prognostic biomarker panel consisting of kynurenate, caffeine, and lysoPC 22:4 may aid in the identification of sepsis patients at a high risk of death, leading to personalized therapy in clinical practice that will improve sepsis survival.