LSC Abstract – Identification of distinct biological phenotypes of ARDS by cluster analysis and association with mortality

Immune-modulatory therapy in ARDS has been unsuccessful, despite promising results in animal studies. This might be due to poor biological phenotyping of the included patients. We hypothesized that ARDS patients can be clustered based on concentrations of 20 plasma and that these unbiased biological phenotypes correspond to clinical outcomes. All ICU-patients were screened for ARDS for three years. Plasma was analyzed for 20 markers of inflammation, coagulation and endothelial activation. Cluster analysis based on biomarker concentrations alone was used to separate the biological phenotypes. A decision tree was used to predict cluster membership based on a limited set of biomarkers. The independent association of clusters with ICU-mortality was studied by logistic regression. 771 patients with ARDS were separated into three clusters: “mild” (N=383), “moderate” (N=224) and “strong responders” (N=164). ICU-mortality increased with response (16%, 26% and 47%, P In conclusion, three biological clusters were identified. “Strong responders” showed a profound cytokine and coagulation response, which was associated with higher mortality and more organ failure. These results may have important implications for inclusion of patients into future clinical trials. Cluster-analysis based on plasma biomarkers This abstract has been presented previously at the European Respiratory Society9s Lung Science Conference in March 2016.