Admission Cytokine Profiling for ICU Mortality Prediction in Heterogeneous Acute Respiratory Failure: An Exploratory...

Background/Objectives: Acute respiratory failure (ARF) encompasses heterogeneous etiologies, and early bedside prognostication remains challenging. Cytokines and chemokines may capture underlying biological severity and identify high-risk patients. We evaluated whether admission cytokine/chemokine profiles add incremental prognostic value over clinical risk factors in unselected ARF patients. Methods: This prospective, single-center cohort study enrolled adult patients admitted to medical ICUs with ARF requiring high-intensity respiratory support. Plasma samples were collected within 24 h of ARF diagnosis, and 19 cytokines/chemokines were measured using multiplex immunoassays. The primary outcome was ICU mortality. Univariate and multivariable logistic regression models assessed associations between biomarkers and mortality, with discrimination evaluated by the area under the receiver operating characteristic curve (AUC). Results: Among 41 patients, 15 (37%) died in the ICU. Non-survivors had higher rates of immunosuppression (80% vs. 38%, p = 0.010) and hematologic malignancy (67% vs. 31%, p = 0.026). CXCL10 (IP-10), IL-18, and CCL2 (MCP-1) were significantly higher in non-survivors, and IL-1Ra showed a marked numerical increase with a significant univariable association with ICU mortality, despite comparable severity scores and oxygenation indices at admission. A clinical core model (SOFA, immunosuppression, hematologic malignancy) achieved an AUC of 0.74 (95% CI 0.58-0.90); adding cytokines improved discrimination modestly (AUC 0.76-0.80). In highest-quartile survival analyses, IL-1Ra (p = 0.002), CXCL10 (p = 0.005), and CCL2 (p = 0.009) demonstrated significant survival separation. Conclusions: At ICU admission, CXCL10 (IP-10), IL-18, CCL2 (MCP-1), and IL-1Ra showed exploratory associations with ICU mortality and were prioritized as candidate inflammatory biomarkers. These findings require validation in larger multicenter cohorts.