Cortisol Testing in Septic Shock: An Evaluation of Diagnostic Performance and Predictors of Corticosteroid Use in a Middle Eastern Cohort

TL;DR

This study evaluated cortisol testing for guiding corticosteroid use in septic shock patients in the UAE and found that clinical judgment is more reliable than cortisol levels.

Contribution

The study demonstrates that cortisol testing has poor diagnostic accuracy and no clinical utility in non-Western septic shock populations.

Findings

Baseline total cortisol levels showed poor diagnostic accuracy (AUC 0.61) for predicting corticosteroid use in septic shock.

Clinical factors like high vasopressor requirements and prolonged shock predicted corticosteroid initiation better than cortisol levels.

Cortisol-guided therapy had zero net benefit compared to clinical practice strategies.

Abstract

Background: Critical illness-related corticosteroid insufficiency (CIRCI) diagnosis remains controversial, largely due to the complex pathophysiology of sepsis, which challenges the reliability of conventional biochemical testing. Recent international guidelines have moved away from strict cortisol threshold-based diagnostic criteria for adrenal insufficiency, relying more on clinical evaluation. However, the applicability and diagnostic accuracy of these revised approaches in non-Western populations remain unexplored. Objective: This study aimed to assessthe diagnostic accuracy and clinical utility of baseline total cortisol levels for guiding corticosteroid therapy in a cohort of patients admitted to the intensive care unit (ICU) with septic shock in a tertiary care hospital in the United Arab Emirates. Methods: A ten-year retrospective observational study (2012–2022) of intensive care patients with septic shock was conducted. CIRCI was operationally defined by a documented clinical decision to administer hydrocortisone >24 h. Receiver Operating Characteristic (ROC) analysis assessed diagnostic performance; Decision Curve Analysis (DCA) evaluated clinical utility of performed cortisol levels. Results: Among 43 patients in the ICU with septic shock, 13 (30.2%) received hydrocortisone (CIRCI group). Mean cortisol was found to be paradoxically higher in the CIRCI group (1341.6 ± 1112.5 vs. 976.0 ± 798.7 nmol/L, p = 0.24). ROC analysis demonstrated poor diagnostic performance (AUC 0.61, 95% CI: 0.44–0.78). International guideline cutoff of <276 nmol/L showed 0% sensitivity in identifying CIRCI. Multiple thresholds yielded negative Youden indices, indicating performance of cortisol levels being worse than a random chance. DCA demonstrated zero net benefit for cortisol-guided therapy across all threshold probabilities when compared to clincial practice strategies. Only clinical factors predicted corticosteroid initiation: high vasopressor requirements (OR 3.2, 95% CI: 1.1–9.4, p = 0.03) and persistent shock >48 h (OR 2.8, 95% CI: 1.0–7.9, p = 0.05). Cortisol level had no predictive value (OR 1.0, 95% CI: 0.9–1.0, p = 0.89). Conclusions: In this cohort, baseline total cortisol demonstrated poor diagnostic accuracy and lacked clinical utility for guiding corticosteroid therapy in patients with sepsis. Our findings reinforce the importance of clinical judgement over biochemical testing in identifying patients with septic shock requiring corticosteroid therapy, in line with the recent international guidelines.

Figures