Comorbidity profiles in chronic obstructive pulmonary disease: a multicohort study.
Abstract
BACKGROUND
Comorbidities are common in COPD, adversely affecting patients’ health. Cluster analysis has been proposed to study phenotypic variability, but inconsistent results have raised concerns. This study uses uniform machine learning techniques across large COPD cohorts to identify comorbidity clusters.
METHODS
Cluster analysis was conducted using data from COPD patients from the ECLIPSE study. 13 comorbidities were grouped using hierarchical clustering on self-organising maps to identify dominant subgroups. The analysis was replicated in COPD patients from the Groningen Severe COPD Cohort (GSCC).
RESULTS
2054 of 2164 ECLIPSE participants (1337 (65%) men; mean±sd age 63.3±7.1 years; 22.0% Global Initiative for Chronic Obstructive Lung Disease (GOLD) stage E) with sufficient data were included in the analysis. Four clusters were identified: Musculoskeletal (C1: more females, lower nutritional status), Mental (C2: younger females, highest comorbidity count), Circulatory (C3: more males, older) and Metabolic (C4: older males, high nutritional status). The replication analysis (776 of 1030 GSCC participants; 266 (34.3%) men; mean±sd age 61.4±6.9 years; 58.9% GOLD stage E) confirmed three clusters. Two clusters (C1, C2) showed similar phenotypic signatures to those in ECLIPSE, while the third combined features of C3 and C4. In ECLIPSE, C1 was associated with increased mortality (HR 2.30, 95% CI 1.26-4.13, p=0.006), which was not replicated in GSSC.
CONCLUSION
Comorbidity clusters exist in COPD and are linked to different patient subgroups with varying symptom burden. While cohort-specific variations in disease outcomes were observed, overall, the study’s findings support the presence of comorbidity profiles with potential to improve disease management.
