Subphenotypes in acute respiratory distress syndrome: A scoping review across clinical, biological, computational, imaging, omics, and artificial intelligence approaches.

BACKGROUND

Acute respiratory distress syndrome (ARDS) is a heterogeneous syndrome with high mortality. Subphenotyping may identify more homogeneous groups for prognostic enrichment and precision therapies.

METHODS

We conducted a scoping review (January 2013-December 31, 2025) in PubMed, Embase, and the Cochrane Library, complemented by reference screening. We included original adult studies deriving, applying, or validating ARDS phenotypes/subphenotypes and excluded pediatric, preclinical, review/editorial, and abstract-only reports.

RESULTS

Sixty studies met eligibility criteria. Subphenotypes were reported across clinical, biological, computational, imaging, omics, and artificial intelligence (AI) domains, with uneven evidence maturity; reproducibility and validation were strongest in biological and computational frameworks, whereas imaging and omics evidence was more heterogeneous and less frequently externally validated. The most robust distinction separated hyperinflammatory and hypoinflammatory groups, differing in mortality, ventilator-free days, and organ failure, and showing heterogeneity of treatment effect in secondary analyses of randomized trials (fluid management, statins, corticosteroids, and recruitment manoeuvres). Clinical and computational approaches provided parsimonious classifiers for near real-time assignment, while biological studies implicated inflammatory, epithelial, and endothelial injury markers. AI models integrated multimodal data and reproduced known phenotypes, but external validation and interpretability were inconsistent.

CONCLUSIONS

ARDS subphenotypes-particularly hyperinflammatory and hypoinflammatory classes-are prognostically meaningful and associated with heterogeneity of treatment effect. Parsimonious clinical and computational classifiers appear closest to bedside translation. Future research should prioritise prospective phenotype-stratified/adaptive trials using standardized, transparent algorithms in diverse international cohorts.