CLINICAL AND BIOLOGICAL MARKERS FOR PREDICTING ARDS AND OUTCOMES IN SEPTIC PATIENTS, COMBINING EMERGENCY MEDICINE, CRITICAL CARE, AND SEPSIS, MAKING IT HIGHLY RELEVANT AND BROADLY APPLICABLE

Abstract

Background: Sepsis is still a leading cause of death worldwide, and acute respiratory distress syndrome (ARDS) is one of its most severe consequences. Early diagnosis of septic individuals with a high risk of developing ARDS is critical for timely intervention and better clinical outcomes. This review summarizes current information regarding the predictive validity of clinical signs, biological biomarkers, and machine learning-based models for ARDS development and prognosis in septic patients.

Methods: A narrative review was undertaken using literature from PubMed/MEDLINE, Scopus, Web of Science, Embase, Google Scholar, and the Cochrane Library. Studies published between January 2015 and March 2025 that assessed clinical predictors, circulating biomarkers, molecular signatures, or artificial intelligence-based prediction models in adult patients with sepsis or sepsis-associated ARDS were considered. Eighty-five relevant trials with 41,782 individuals were analyzed.

Results: Clinical markers such as the SOFA score, APACHE II score, PaO₂/FiO₂ ratio, serum lactate concentration, and mechanical ventilation demand accurately predict disease severity and death. RAGE and SP-D, two epithelial damage indicators, were found to have high correlations with early ARDS and bad outcomes. Angiopoietin-2 had the best diagnostic performance among endothelial biomarkers (AUC = 0.91); CRP, IL-6, procalcitonin, D-dimer, and plasminogen activator inhibitor-1 were the most commonly validated inflammatory and coagulation biomarkers. Emerging molecular markers, including as microRNAs, genomic and transcriptomic signatures, extracellular vesicles, and cell-free DNA, have shown great promise in precision risk stratification. Individual biomarkers were outperformed by combined clinical-biomarker prediction models (AUC = 0.91; sensitivity 88%; specificity 86%), whereas deep learning and XGBoost algorithms had the highest predictive performance among machine learning approaches (AUC up to 0.94).

Conclusions: Evidence suggests that combining clinical severity scores with multimarker panels and artificial intelligence significantly improves early detection of ARDS and bad outcomes in septic patients. Future precision medicine solutions that combine clinical, molecular, and computational data have the potential to revolutionize risk assessment and personalized care of sepsis-associated ARDS.