Association between early plasma cortisol levels and all-cause in-hospital mortality in patients with sepsis: a multicenter retrospective study.

Abstract

Clinical heterogeneity complicates early accurate prognosis for sepsis. Cortisol is robustly elevated by hypothalamic-pituitary-adrenal (HPA) axis activation during septic stress, yet large multicenter data clarifying its independent association with in-hospital mortality are scarce. This study investigated the prognostic value of early plasma cortisol and its incremental predictive value over classic critical illness severity scores. We retrospectively enrolled 2,364 adult patients meeting Sepsis-3.0 criteria from MIMIC-IV and eICU-CRD databases, with cortisol tested within 24 h after intensive care unit (ICU) admission. Individuals with pituitary/adrenal dysfunction or prior corticosteroid administration before cortisol measurement were excluded. The primary endpoint was all-cause in-hospital mortality. Multivariable Cox regression, restricted cubic splines and threshold models were adopted to characterize cortisol-mortality relationships; receiver operating characteristic (ROC) curves, integrated discrimination improvement (IDI) and net reclassification improvement (NRI) were calculated to evaluate predictive performance. After full adjustment for confounders, each 1 µg/dL rise in cortisol independently increased mortality risk (HR = 1.004, 95 %CI 1.001-1.006, P = 0.007). A nonlinear threshold of 43.8 µg/dL was detected: below this cutoff, mortality risk rose 1.7 % per µg/dL (HR = 1.017, P < 0.001), whereas no significant correlation existed above the threshold (HR = 0.999, P = 0.696). Cortisol alone showed moderate discrimination (AUC = 0.668). When added to Sequential Organ Failure Assessment (SOFA) score, Acute Physiology Score III (APSIII), or Oxford Acute Severity of Illness Score (OASIS), it significantly improved discrimination (IDI 1.73 %-2.17 %) and risk reclassification (NRI 36.6 %-39.8 %). Early plasma cortisol independently predicts sepsis in-hospital mortality with a distinct nonlinear threshold effect. Although its single-biomarker predictive power is limited, cortisol substantially enhances the risk-stratification capacity of established severity scores, supporting its clinical application as an adjunctive prognostic biomarker.