Aims
Despite intensive research, sepsis remains a life-threatening organ dysfunction with high mortality (25-50%) [1, 2]. Early sepsis diagnosis is complicated by primarily nonspecific symptoms. Up to 40% of patients with severe sepsis are admitted to the emergency department [3]. Early recognition in the emergency department and initiation of therapeutic interventions is necessary to increase and improve patient survival [2,4,5]. The appearance, accumulation, or persistence of metabolites during the infection-related host response could serve as a surrogate for impaired metabolic control and for monitoring disease severity. Thus, the study aims to identify the most appropriate metabolites or combinations of metabolites for early sepsis diagnosis, identification of sepsis-related organ dysfunction and risk stratification of patients with suspected sepsis.
Materials and Methods
A total of 188 metabolites comprising six analyte classes were measured using LC-MS/MS in lithium heparin plasma samples from 400 patients on admission to the emergency department of the university hospital Jena. 160 and 24 patients developed sepsis and septic shock within 96 hours, respectively. Procalcitonin was used as reference laboratory parameter.
Results
The primary study endpoint was to define metabolites that can early identify patients with sepsis or septic shock. 44 and 30 metabolites were altered after correction for multiple testing in patients developing sepsis or septic shock, respectively. Mainly three metabolite classes: amino acids, lysophosphatidylcholines and phosphatidylcholines were associated with sepsis or septic shock and mostly decreased in their concentration during infection. Promising metabolites were selected and combined with a LASSO-regression with a 20-fold cross-validation. ROC-analysis of these models showed a sensitivity of 80.6% or 91.7% and a specificity of 79.6% or 90.2% for early detection of sepsis or septic shock, respectively.
Furthermore, amino acids, biogenic amines and lysophosphatidylcholines showed high potential for providing information of organ dysfunction or poor patients’ outcome. Significant altered metabolites were combined for prognosis of unfavorable outcome of patients or the need of interventions at the intermediate care unit.
Conclusion
The present targeted metabolomics approach allowed to achieve the two study endpoints: a) to indicate metabolite patterns that are of early diagnostic value for sepsis or septic shock; b) to identify some metabolites that can early provide information about risk stratification of patients. These findings will lead to the development of improved novel diagnostic tools for early diagnosis and prognosis of sepsis and septic shock.