Abstract 11924: Phenotypes of Heart Failure Identified via Cluster Analysis Do Not Recapitulate Hemodynamic Profiles: Insights From the ESCAPE Trial

Abstract

Background: Heart failure (HF) classification criteria do not adequately describe disease state. We used cluster analysis to identify distinct phenotypes of patients in the pulmonary artery catheter (PAC) arm of the ESCAPE trial, and examine association of clusters with hemodynamic profiles. Methods: Cluster analysis was performed on baseline clinical variables and PAC measurements in a cohort of 172 subjects from the ESCAPE trial, a randomized study examining PAC versus usual care in advanced systolic HF patients. ANOVA examined association between patient clusters and clinically determined hemodynamic profiles (warm/cold/wet/dry); association with clinical outcomes was assessed using Cox proportional hazards models. Results: Four HF clusters were identified with the following general characteristics: Cluster 1(n=75): Male Caucasians with ischemic cardiomyopathy, multiple comorbidities, lowest BNP levels; Cluster 2 (n=33): Females with non-ischemic cardiomyopathy, few comorbidities, most favorable hemodynamics; Cluster 3 (n=29): Young African American males with non-ischemic cardiomyopathy, most adverse hemodynamics and advanced disease; Cluster 4 (n=25): Older Caucasians with ischemic cardiomyopathy, concomitant renal insufficiency, and highest BNP levels. We noted no association between hemodynamic profiles and clusters (P=0.70). Cluster 4 patients had the highest risk of all adverse clinical outcomes, whereas cluster 2 had the lowest. Compared to patients in cluster 4, patients in other clusters had 45-70% lower risk of all-cause mortality. Conclusions: Using a novel approach that clusters patients according to similar clinical variables and PAC measures, we identified 4 clinically relevant phenotypes of HF patients, with no discernable relationship to hemodynamic profiles. Our analysis suggests that HF classification can be enhanced by simultaneous considerations of etiology, comorbid conditions, and biomarker levels.