Comorbidity profiles in chronic obstructive pulmonary disease: a multicohort study.

Area of Science:

• Pulmonary Medicine
• Data Science
• Clinical Research

Background:

• Comorbidities are prevalent in Chronic Obstructive Pulmonary Disease (COPD), negatively impacting patient health.
• Previous cluster analysis studies in COPD have yielded inconsistent results, necessitating further investigation.
• This research applies uniform machine learning techniques to large COPD cohorts to identify reliable comorbidity clusters.

Purpose of the Study:

• To identify distinct comorbidity clusters in COPD patients using advanced machine learning.
• To analyze phenotypic variability and dominant subgroups within COPD populations.
• To explore the potential of identified comorbidity profiles for improving COPD disease management.

Main Methods:

• Hierarchical clustering on self-organising maps was employed to group 13 comorbidities in COPD patients from the ECLIPSE study.
• The clustering analysis was replicated in a separate cohort, the Groningen Severe COPD Cohort (GSCC), for validation.
• Machine learning techniques were uniformly applied across large COPD cohorts to ensure consistent results.

Main Results:

• Four distinct comorbidity clusters were identified in the ECLIPSE cohort: Musculoskeletal, Mental, Circulatory, and Metabolic.
• Replication in the GSCC cohort confirmed three clusters, with two aligning with ECLIPSE findings and a third combining features.
• One cluster (Musculoskeletal) in the ECLIPSE cohort showed a significant association with increased mortality.

Conclusions:

• The study confirms the existence of distinct comorbidity clusters in COPD patients, associated with different subgroups and symptom burdens.
• While cohort-specific variations in disease outcomes were observed, the findings support the utility of comorbidity profiles in COPD management.
• These identified comorbidity profiles offer potential avenues for refining disease management strategies in COPD.