Metabolite Profiles in Sepsis: Developing Prognostic Tools Based on the Type of Infection

Objectives: Currently used biomarkers insufficiently discriminate between patients with systemic inflammatory response syndrome of non-infectious origin and sepsis. The aim of this study was to identify surrogate markers that distinguish between systemic inflammatory response syndrome and sepsis as well as the underlying type of infection by targeted metabolomics. Design: Retrospective analysis. Settings: Six sites of the Hellenic Sepsis Study Group and at Jena University Hospital. Patients: A total of 406 patients were analyzed: 66 fulfilling criteria for diagnosis of systemic inflammatory response syndrome, 100 for community-acquired pneumonia, 112 for urinary tract infection, 83 for intra-abdominal infection and 45 for bloodstream infection. Patients were divided into test cohort (n = 268) and confirmation cohort (n = 138). Interventions: A total of 186 metabolites were determined by liquid chromatography tandem mass spectrometry. Measurements and Main Results: Serum concentrations of most acylcarnitines, glycerophospholipids and sphingolipids were altered in sepsis compared to systemic inflammatory response syndrome. A regression model combining the sphingolipid SM C22:3 and the glycerophospholipid lysoPCaC24:0 was discovered for sepsis diagnosis with a sensitivity of 84.1% and specificity of 85.7%. Furthermore, specific metabolites could be used for the discrimination of different types of infection. The glycerophospholipid lysoPCaC26:1 identified patients with community-acquired pneumonia in sepsis or severe sepsis/septic shock. Within severe sepsis/septic shock, patients with bloodstream infection could be discriminated by a decrease of acetylornithine. Changes of metabolites between sepsis and severe sepsis/septic shock also varied according to the underlying type of infection, showing that putrescine, lysoPCaC18:0 and SM C16:1 are associated with unfavorable outcome in community-acquired pneumonia, intra-abdominal infections and bloodstream infections, respectively. Conclusions: Using a metabolomics approach, single metabolites are identified that allow a good, albeit at about 14% false positive rate of sepsis diagnosis. Additionally, metabolites might be also useful for differentiation and prognosis according to the type of underlying infection. However, confirmation of the findings in ongoing studies is mandatory before they can be applied in the development of novel diagnostic tools for the management of sepsis.