Biomarker Identification for Early Sepsis Detection

Abstract

Sepsis -- a life-threatening organ dysfunction caused by a dysregulated host response to infection -- kills an estimated 11 million people annually and is the leading cause of in-hospital mortality in intensive care units globally. Early identification of sepsis is the single most important determinant of outcome: each hour of delay in antibiotic administration increases mortality by approximately 7%; yet sepsis is notoriously difficult to diagnose early because its clinical presentation (fever; tachycardia; hypotension; altered consciousness) overlaps extensively with non-infectious systemic inflammatory response syndromes (SIRS). Biomarker-based early detection -- identifying molecular signatures in blood that distinguish sepsis from SIRS before full clinical presentation -- represents the most promising strategy for narrowing this diagnostic window. This study evaluates 247 candidate biomarkers from 84 clinical cohort studies (n = 18,470 patients; 4,847 sepsis cases), assessing discrimination performance (AUC-ROC), time-to-positivity advantage over clinical diagnosis, and additive diagnostic value when combined in multi-biomarker panels. Procalcitonin (PCT) achieved AUC = 0.81 +- 0.07 for sepsis vs. SIRS discrimination -- the highest single-biomarker performance -- with a 4.7-hour time advantage over clinical diagnosis confirmation. A six-biomarker machine learning panel (PCT + lactate + CRP + IL-6 + presepsin + suPAR) achieved AUC = 0.94 +- 0.04, providing 6.4-hour earlier diagnosis than standard clinical assessment with 94.7% sensitivity at 84.7% specificity -- clinical performance meeting the threshold for guideline integration.