Abstract
COPD is an heterogeneous disease better characterized by multidimensional phenotyping. Clustering is a technique used to identify discrete subgroups with similar combinations of traits. Except for pulmonary function variables, many other characteristics (6MWT, BMI, QOL) are not unique to COPD.
Aim: To compare how discrete clusters form in a mixed cohort of 120 individuals with COPD and controls.
Methods: Hundred and twenty patients matched by age and gender were selected, 90 with COPD of which 60 died at 3 years of follow-up. Hierarchical clustering was applied using pulmonary function, functional, anthropometric and QOL variables. Clusters were compared against a selected reference.
Results: Four clusters were identified and their composition is shown in A. Cluster 4 composed with 84% controls was used as reference and the comparison and descriptions are shown in B.
Conclusion: Clustering is a useful tool to discriminate clinical meaningful phenotypes and by including “controls” we showed that >10% were assigned in COPD predominant clusters. This could be important when designing exploratory studies.
- Copyright ©the authors 2016
