TY  - JOUR
T1  - Subphenotypes of Acute Respiratory Distress Syndrome: From Biological Heterogeneity to Precision Medicine 
AU  - Li, Hao
AU  - Meng, Xianglin
JF  - Journal of Translational Critical Care Medicine
VL  - 8
IS  - 3
SN  - 2590-3438
SP  - e00002
EP  - e00002
Y1  - 2026-09-30
DO  - 10.14218/JTCCM.2026.00002
UR  - https://www.xiahepublishing.com/2590-3438/JTCCM-2026-00002
AB  - Despite the 2012 Berlin definition standardizing acute respiratory distress syndrome (ARDS) diagnosis, pharmacological therapies have consistently failed in phase 3 trials, partly reflecting the syndrome’s profound biological heterogeneity. We synthesize evidence on ARDS subphenotyping to evaluate prospects for precision medicine. This narrative review was informed by targeted searches of PubMed/MEDLINE, Embase, and Web of Science from database inception through February 2026, supplemented by an updated PubMed/MEDLINE search through September 2026. Representative clinical trials, multicenter cohorts, mechanistic studies, and methodological reports were selected to support a critical synthesis of ARDS subphenotyping, its therapeutic implications, and barriers to clinical translation. Hyperinflammatory and hypoinflammatory subphenotypes—originally identified using plasma biomarkers (IL-6, IL-8, and soluble tumor necrosis factor receptor 1) in the ARMA and ALVEOLI cohorts—demonstrate reproducible differences in mortality, with worse outcomes in the hyperinflammatory subphenotype, and have shown potential treatment-effect heterogeneity in secondary analyses. Approximately 30%–35% of patients exhibit the hyperinflammatory profile, characterized by greater vasopressor requirements and organ failure. Radiological phenotypes (focal versus non-focal) on computed tomography imaging have been associated with differences in lung recruitability and physiological responses to ventilatory interventions, although the LIVE trial did not demonstrate an overall survival benefit from morphology-guided ventilation. Recent multi-omics studies (transcriptomic, proteomic, and metabolomic) reveal distinct endotypes involving dysregulated innate immunity, endothelial activation, and aberrant coagulation. Machine learning algorithms, particularly unsupervised clustering and deep learning-based computed tomography radiomics, have been investigated for more rapid phenotyping, but reported performance varies by task, cohort, and validation design. However, prospective, phenotype-guided randomized trials remain limited. Subphenotype stratification offers a mechanistically grounded framework for re-evaluating therapies that were ineffective in unselected populations and for investigating phenotype-informed ventilatory strategies. Clinical implementation requires standardized, bedside-compatible biomarker panels and definitive trials demonstrating that phenotype-specific interventions improve patient-centered outcomes.