Evaluation of the multimorbidity network and its relationship with clinical phenotypes in chronic obstructive pulmonary disease: The GALAXIA study

Background: Chronic obstructive pulmonary disease (COPD) is a complex and heterogeneous condition, in which taking into consideration clinical phenotypes and multimorbidity is relevant to disease management. Network analysis, a procedure designed to study complex systems, allows to represent connections between the distinct features found in COPD.

Methods: Network analysis was applied to a cohort of patients with COPD in order to explore the degree of connectivity between different diseases, taking into account the presence of two phenotypic traits commonly used to categorize patients in clinical practice: chronic bronchitis (CB⁺ /CB^- ) and the history of previous severe exacerbations (Ex⁺ /Ex^- ). The strength of association between diseases was quantified using the correlation coefficient Phi (ɸ).

Results: A total of 1726 patients were included, and 91 possible links between 14 diseases were established. Although the four phenotypically defined groups presented a similar underlying comorbidity pattern, with special relevance for cardiovascular diseases and/or risk factors, classifying patients according to the presence or absence of CB implied differences between groups in network density (mean ɸ: 0.098 in the CB^- group and 0.050 in the CB⁺ group). In contrast, between-group differences in network density were small and of questionable significance when classifying patients according to prior exacerbation history (mean ɸ: 0.082 among Ex^- subjects and 0.072 in the Ex⁺ group). The degree of connectivity of any given disease with the rest of the network also varied depending on the selected phenotypic trait. The classification of patients according to the CB^- /CB⁺ groups revealed significant differences between groups in the degree of conectivity between comorbidities. On the other side, grouping the patients according to the Ex^- /Ex⁺ trait did not disclose differences in connectivity between network nodes (diseases).

Conclusions: The multimorbidity network of a patient with COPD differs according to the underlying clinical characteristics, suggesting that the connections linking comorbidities between them vary for different phenotypes and that the clinical heterogeneity of COPD could influence the expression of latent multimorbidity. Network analysis has the potential to delve into the interactions between COPD clinical traits and comorbidities and is a promising tool to investigate possible specific biological pathways that modulate multimorbidity patterns.