Recent developments in proteomics suggest opportunities to understand the pathophysiological heterogeneity of sepsis and provide precision medicine tailored to individual patients. This study aims to evaluate the serum proteomic profiles of patients with sepsis using Sequential Window Acquisition of All Theoretical Mass Spectra proteomics and identify novel biomarkers for assessing sepsis severity and predicting patient outcomes in the ICU. This retrospective cohort study included 217 adult patients diagnosed with bacterial sepsis who were admitted to a medical ICU at a single tertiary hospital between January 2011 and January 2020, along with 292 healthy controls. Proteomic analysis was performed to compare patients with sepsis to the 292 healthy controls and analyze differences among sepsis subgroups. The subgroups were classified according to their outcome: early death (within 3 days), late death (after 3 days), and recovery (survived and discharged). Five key proteins—β-actin (ACTB), fibronectin (FINC), metalloproteinase inhibitor 1 (TIMP1), platelet factor 4 (PF4), and C-X-C motif chemokine 7 (CXCL7)—were identified as significant discriminators of sepsis subgroups with AUCs ranging from 0.786 to 0.833. A six-variable model including 5 proteins and the SOFA score showed the highest AUC value of 0.903 for predicting in-hospital mortality. Multivariable Cox proportional hazards analysis revealed that increased ACTB (hazard ratio HR 1.21 1.07–1.36, p = 0.002), decreased FINC (HR 0.88 0.79–0.98, p = 0.024), higher SOFA scores (HR 1.08 1.03–1.13, p = 0.002) and gram-negative sepsis (HR 1.42 1.02–1.97, p = 0.038) were significantly associated with increased in-hospital mortality. A novel set of biomarkers was identified in this study, including ACTB, FINC, CXCL7, TIMP1, and PF4, for assessing sepsis prognosis. The six-variable model incorporating SOFA scores demonstrated a high prognostic value for in-hospital mortality, potentially enabling more accurate risk stratification for patients with sepsis in the ICU.
Lee et al. (Wed,) studied this question.