Toward precision medicine in ARDS: Heterogeneity, subphenotypes, and individualized therapy − A narrative review

Acute respiratory distress syndrome (ARDS) remains a major unmet challenge in critical care, with substantial mortality despite advances in supportive management. A central reason for repeated therapeutic failure is that ARDS is a clinically defined syndrome encompassing multidimensional heterogeneity. Subphenotyping has provided a framework for organizing this heterogeneity into clinically and biologically meaningful states. The value of subphenotyping lies not only in refining ARDS pathobiology, but also in generating hypotheses about differential treatment response. In particular, ARDS subphenotypes may help identify patients more likely to benefit from selected supportive or immunomodulatory interventions, while emerging machine-learning approaches aim to estimate heterogeneous and individualized treatment effects (ITE) beyond conventional average treatment effects. However, translation into practice remains limited by uncertain reproducibility, temporal instability, limited bedside feasibility, and the predominance of retrospective, observational, or post hoc evidence. Evidence from randomized trials directly testing subphenotype-guided treatment allocation in ARDS remains lacking. This narrative review summarizes current evidence on ARDS heterogeneity, subphenotypes, and ITE estimation, with emphasis on their translational limitations and requirements for prospective validation. Progress in precision medicine in ARDS will depend on parsimonious classifiers, prospective validation, and trial designs that explicitly test treatment-effect heterogeneity before implementation in routine care.