Large-scale proteomic profiling identifies distinct inflammatory phenotypes in Acute Respiratory Distress Syndrome (ARDS): A multi-center, prospective cohort study.

Summary

In a multicenter prospective cohort of 1048 ARDS patients, latent class analysis of early serum proteomics defined three inflammatory phenotypes with distinct clinical, radiographic, and molecular features. The high-risk C1 phenotype had the worst 90-day outcomes and different responses to steroids and ventilation, supporting biomarker-guided precision therapies.

Key Findings

• Three proteomic inflammatory phenotypes (C1–C3) were identified and externally validated among 1048 ARDS patients.
• C1 had greater poorly/non-inflated lung on CT, highest 90-day mortality and shock, and fewest ventilator-free days.
• Phenotypes demonstrated heterogeneous treatment effects to glucocorticoids and ventilation; an XGBoost classifier enabled phenotype prediction.

Clinical Implications

Phenotype assignment at diagnosis could stratify risk and inform steroid use and ventilation strategies. Implementing a parsimonious classifier may enable bedside biomarker-guided care.

Limitations

• Non-randomized treatment exposure may confound heterogeneous treatment effect estimates
• Serum proteomics may not fully reflect lung compartment biology

Future Directions

Prospective stratified or adaptive trials should test phenotype-guided therapies and validate parsimonious biomarker panels for bedside use.