Integrative plasma metabolomic profiling identifies distinct vascular and inflammatory signatures in hospital emergency department patients

Background: Early risk stratification of vascular and inflammatory conditions in the hospital emergency department (ED) is crucial as vascular conditions are the leading global cause of morbidity and mortality. Blood plasma lipidomics and metabolomics, the high-throughput analysis of small molecules, may aid in the identification of potential biomarkers for risk stratification of clinical outcomes in patients with these conditions as it provides an additional physiological readout. This may help to address limitations of existing methods, thereby improving prognosis and management of these conditions. Methods: A comprehensive survey of 338 lipids and 102 polar metabolites was conducted on a cohort of 500 patients with vascular and inflammatory conditions who visited the ED at the National University Hospital, Singapore. Lipids were analysed using targeted lipidomics with normalization against class specific internal standards, while polar metabolites were analysed using untargeted and semi-targeted metabolomics with relative quantification against the median peak areas of study samples. These patients were monitored prospectively in the hospital for clinical outcomes of interest. We identified analytes that were significantly correlated with disposition, transfer from general ward (GW) to intensive care unit (ICU), hospital length of stay (LOS), 30-day reattendance to ED and 30-day readmission to ward in vascular and inflammatory subgroups. Prediction models for these outcomes were derived using multivariate logistic regression with elastic net penalty. Results: In patients with vascular conditions, asymmetric dimethyl arginine (ADMA), citrulline and hypotaurine were the strongest predictors for disposition; thromboxane B3 (TXB3) for transfer from GW to ICU; pregnenolone for 30-day ED reattendances; and phosphatidylcholines (PCs) and estrone 3-sulfate for 30-day ward readmissions. In patients with inflammatory conditions, monounsaturated ether lysophosphatidylcholines, PC 32:1 and L-tryptophan were the strongest predictors for disposition; d18:1 ceramides for transfer from GW to ICU; triacylglycerols for hospital LOS; and 3-ureidopropionic acid and phosphatidylinositol 32:0 for 30-day hospital readmissions. Interestingly, most lipids decreased in concentrations across all clinical outcomes. The predictive models for disposition showed a good predictive performance, giving areas under the curves of 0.870 and 0.882 for vascular and inflammatory subgroups, respectively. Conclusions: Novel and distinct metabolomic signatures significantly associated with clinical outcomes were identified in patients with vascular and inflammatory conditions. The proposed biomarkers in this study may be viable tools for future development of strategies that may aid clinicians with timely treatment, monitoring and risk stratification of ED patients using a precision medicine-based approach.