Shelsey Johnson, MD
Rank
Instructor
Department
Medicine
Pulmonary and Critical Care medicine
Authors
Shelsey W. Johnson*, MD, Carrie Pistenmaa, MD, MS, Bina Choi, MD, MPH, Pietro Nardelli, PhD, Raul San Jose Estepar, PhD, George Washko, MD, Farbod Rahaghi, MD, PhD
Principal Investigator
Farbod Rahaghi, MD, PhD
Twitter / Website
Categories
Background: Pulmonary hypertension (PH) frequently complicates chronic obstructive pulmonary disease (COPD) and is associated with high mortality yet there are no approved treatments. To date, therapeutic investigations are limited by significant disease heterogeneity and advanced phenotyping efforts are needed. Methods: We applied k-means cluster analysis to smokers from the Phase 2 visit of the COPDGene study with spirometric evidence of COPD and a chest computed tomography (CT) pulmonary artery to aorta ratio (PA/Ao) > 1 (a CT surrogate of PH) to identify imaging-based COPD-PH phenotypes. Variables included in clustering were FEV1 and diffusion capacity (% predicted) and CT measurements of the ratio of pulmonary arterial small to total blood vessel volume (aBV5/TBV) and % emphysema. Logistic regression assessed cluster association with all-cause mortality. Results: Three clusters of COPD patients with presumed PH were identified (n=297). Cluster 1 (n=135) had minimal emphysema (4[10]) but loss of small arterial volume, or pruning (aBV5/TBV, 0.45 (0.07)). Cluster 2 (n=62) had a similar degree of pruning (0.44 (0.08)) but substantial emphysema (45 (14)). Cluster 3 (n=100) had minimal emphysema and pruning (57 (7) and 1[3], respectively). When adjusted for age and smoking pack-years, the odds of death for subjects in Cluster 1 was significantly decreased as compared to reference Cluster 2 with the highest mortality risk (OR Cluster1 0.492, 95% CI 0.26-0.93). Conclusion: Cluster analysis deconstructs COPD-PH heterogeneity to identify imaging-based sub-phenotypes with differential pruning and emphysema. COPD-PH sub-phenotyping efforts should focus on the incorporation of non-invasive CT features to inform ongoing efforts to identify effective treatments.