Abstract
Background: Identifying clinically meaningful groups of COPD patients is a crucial goal to explore COPD heterogeneity. We attempt to define groups using unsupervised clustering methods.
Methods: Data from the 2164 COPD patients in the Evaluation of COPD Longitudinally to Identify Predictive Surrogate Endpoints (ECLIPSE) study were assessed. Using forty-one baseline variables describing demographic, clinical, quality of life, laboratory and biomarker values, twelve factors were identified via factor analysis that accounted for 61% of the variance in the data set. The variables with the highest loadings for those factors were used to define five patient groups using unsupervised clustering, and relationships to longitudinal outcomes were assessed.
Results: Demographic profiles are shown in table 1. Over three years, higher mortality was seen in Cluster 2 (characterized by higher comorbidity and BMI, despite FEV1 values that were not substantially lower than other groups) and Cluster 5, characterized by more airflow limitation.
Conclusion: Unsupervised cluster analysis identified 5 groups of COPD patients in ECLIPSE that differ in their baseline demographics and outcomes over 3 years. These may represent subtypes of COPD.
Funded by GlaxoSmithKline. (SCO104960, NCT00292552)
- © 2011 ERS