What does this research mean for the field?

Multi-biomarker profiling and machine learning identify three biologically and clinically distinct ARDS clusters that exhibit differential 90-day mortality rates. Novelty: ClaimNovelty.INCREMENTAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

This study aims to refine acute respiratory distress syndrome (ARDS) subtypes using biomarker data to enhance personalized treatment approaches.

May 20, 2026

C24-15 Novel Biomarker-driven Clusters of Acute Respiratory Distress Syndrome Among Patients Enrolled in the Reevaluation of Systemic Early Neuromuscular Blockade (rose) Trial

Key Points

This study aims to refine acute respiratory distress syndrome (ARDS) subtypes using biomarker data to enhance personalized treatment approaches.
Analyzed data from 883 patients in the ROSE trial, measuring six plasma biomarkers upon enrollment.
Applied unsupervised machine learning for cluster identification based on clinical characteristics and biomarker levels.
Conducted statistical analyses, including Kruskal–Wallis tests and logistic regression, to assess treatment effects and cluster comparisons.
Identified three distinct ARDS clusters; Cluster 3 had the highest biomarkers, fluid balance, APACHE score, and 90-day mortality.
Cluster 1 had the lowest biomarkers and mortality rates.
Angiopoietin-2, IL-1β, and eNAMPT were linked to increased mortality in multivariate analyses.

Abstract

Abstract Rationale Distinct biological subtypes of acute respiratory distress syndrome (ARDS) have been described utilizing clinical and limited biomarker data. Refining ARDS subtypes may advance personalized therapeutic strategies. Methods Data and plasma samples from 883 patients in the Reevaluation of Systemic Early Neuromuscular Blockade (ROSE) trial were analyzed. Six biomarkers (Angiopoietin-2, IL-1ra, IL-1β, IL-6, eNAMPT, PSGL-1) were measured on Day 1 of enrollment. Unsupervised machine learning was applied to identify latent clusters. Predictor variables included treatment group (neuromuscular blockade vs. usual care), biomarker levels, and baseline clinical characteristics (24-hour fluid balance, P/F ratio, APACHE comorbidity-age score, barotrauma, driving pressure, sex, and BMI). Missing data were imputed using K-nearest neighbors (k = 5), and biomarkers were log10-transformed. Factor Analysis of Mixed Data (FAMD) was performed on covariates, followed by hierarchical clustering on principal components. Cluster comparisons used Kruskal–Wallis tests for continuous and chi-square or Fisher’s exact tests for categorical variables. Treatment effect heterogeneity was tested via logistic regression including treatment, cluster, and their interaction. Analyses were conducted in R (version 4.5.1) using VIM and FactoMineR. Statistical significance was set at α = 0.05. Results Three latent clusters were identified and cluster 3 had significantly higher levels of all six biomarkers, a significantly higher baseline positive fluid balance, a lower P/F ratio, a higher APACHE score, and the highest 90-day mortality when compared to clusters 2 and 1. Cluster 1 had the lowest biomarker levels and lowest 90-day mortality. Multivariate analysis identified Angiopoietin-2, IL-1β, and eNAMPT as plasma biomarkers associated with increased mortality. Conclusion A biomarker-driven analysis identified three biologically and clinically distinct ARDS clusters with differential 90-day mortality. These findings support multi-biomarker profiling and machine learning as tools for defining ARDS subtypes and guiding future precision-medicine trials. This abstract is funded by: NIH

Bookmark

C24-15 Novel Biomarker-driven Clusters of Acute Respiratory Distress Syndrome Among Patients Enrolled in the Reevaluation of Systemic Early Neuromuscular Blockade (rose) Trial

Key Points

Abstract

Cite This Study

Also Consider

Also Consider