The respiratory physiome: Clustering based on a comprehensive lung function assessment in patients with COPD

Abstract

Background: While spirometry and particularly airflow limitation is still considered as an important tool in therapeutic decision making, it poorly reflects the heterogeneity of respiratory impairment in chronic obstructive pulmonary disease (COPD). The aims of this study were to identify pathophysiological clusters in COPD based on an integrated set of standard lung function attributes and to investigate whether these clusters can predict patient-related outcomes and differ in clinical characteristics.

Methods: Clinically stable COPD patients referred for pulmonary rehabilitation underwent an integrated assessment including clinical characteristics, dyspnea score, exercise performance, mood and health status, and lung function measurements (post-bronchodilator spirometry, body plethysmography, diffusing capacity, mouth pressures and arterial blood gases). Self-organizing maps were used to generate lung function based clusters.

Results: Clustering of lung function attributes of 518 patients with mild to very severe COPD identified seven different lung function clusters. Cluster 1 includes patients with better lung function attributes compared to the other clusters. Airflow limitation is attenuated in clusters 1 to 4 but more pronounced in clusters 5 to 7. Static hyperinflation is more dominant in clusters 5 to 7. A different pattern occurs for carbon monoxide diffusing capacity, mouth pressures and for arterial blood gases. Related to the different lung function profiles, clusters 1 and 4 demonstrate the best functional performance and health status while this is worst for clusters 6 and 7. All clusters show differences in dyspnea score, proportion of men/women, age, number of exacerbations and hospitalizations, proportion of patients using long-term oxygen and number of comorbidities.

Conclusion: Based on an integrated assessment of lung function variables, seven pathophysiological clusters can be identified in COPD patients. These clusters poorly predict functional performance and health status.

Fig 1. Heterogeneity of lung function impairment in COPD.

Fig 2. The lung function clusters and the related functional and health status characteristics.

The seven lung function clusters in chronic obstructive pulmonary disease (COPD) and the related functional and health status characteristics. When looking at the different lung function, functional characteristics and health status, subjects “raise a red flag” if the attribute is relatively high within this sample, present “a green flag” if the clinical attribute is moderate, and present “a blue flag” when the clinical attribute is relatively low within this sample. In this way the maps can be interpreted. The Viscovery program placed all subjects on a specific position on the map based on their profile of a comprehensive lung function assessment. The more subjects resemble in terms of their lung function the closer they are on the map. Contrarily, the more they differ the further they are away from each other. By drawing lines on the map, the Viscovery program could identify seven different clusters of patients with COPD with a significant different respiratory physiome (95% confidence interval).