Temporal stability of phenotypes of acute respiratory distress syndrome: clinical implications for early corticosteroid therapy and mortality.

Summary

Using IPD from six multicenter RCTs and a large retrospective cohort (total n≈9,536), the authors built an open-source clinical classifier to track ARDS inflammatory phenotypes over 30 days. Hyperinflammatory ARDS had lower mortality with corticosteroids (HR 0.81), while hypoinflammatory ARDS had higher mortality (HR 1.26); benefit persisted at day 3 only if patients remained hyperinflammatory.

Key Findings

• Clinical AI classifier identified 39% hyperinflammatory and 61% hypoinflammatory ARDS.
• 30-day mortality was 49% in hyperinflammatory vs 24% in hypoinflammatory ARDS (p<0.001).
• Phenotypes were dynamic: 49% of hyperinflammatory transitioned to hypoinflammatory; 7% of hypoinflammatory transitioned to hyperinflammatory (p<0.001).
• Corticosteroids reduced mortality in hyperinflammatory ARDS (IPW-weighted HR 0.81 [0.67–0.98], p=0.033).
• Corticosteroids increased mortality in hypoinflammatory ARDS (IPW-weighted HR 1.26 [1.06–1.50], p=0.009).
• At day 3, benefit from corticosteroids persisted only among patients remaining hyperinflammatory (adjusted OR 0.51, 95% CI 0.32–0.80, p=0.004).

Clinical Implications

Consider early phenotype assessment and re-assessment (e.g., by day 3) to guide corticosteroid use: favor corticosteroids in hyperinflammatory ARDS and avoid or de-escalate them in hypoinflammatory ARDS pending prospective validation.

Limitations

• Observational analyses may have residual confounding and treatment indication bias despite IPW adjustment
• Phenotyping relies on clinical surrogates rather than direct biomarker panels; heterogeneity in corticosteroid regimens

Future Directions

Prospective, randomized trials to test phenotype-guided corticosteroid strategies with real-time reassessment; integration of biomarker panels to refine classification.