Abstract 4141166: Identification of Clinical Phenotypes and Heterogeneous Treatment Effects of Surgical Revascularization in Ischemic Cardiomyopathy: A Machine Learning Consensus Clustering Analysis

Introduction: The Surgical Treatment for Ischemic Heart Failure (STICH) trial confirmed the survival benefits of surgical revascularization compared to medical therapy alone in patients with ischemic cardiomyopathy (ICM). However, the complexities of clinical characteristics in this population may lead to distinct clinical phenotypes with heterogeneity in risk stratification and therapeutic responses to medical therapy or surgical revascularization. Aims: To identify ICM patients with different clinical phenotypes for evaluating their outcomes and heterogeneous treatment effects (HTEs) of coronary artery bypass grafting (CABG). Methods: We applied a machine learning-based consensus, K-Medoids clustering analysis to the STICH trial dataset. A total of 20 clinically available variables were utilized to distinguish the phenotypes. We compared the risk of all-cause mortality and cardiovascular mortality among different phenotypes. Then, the survival benefits of CABG compared with medical therapy alone were assessed in the identified phenotypes for evaluating HTEs. Results: From July 2002 to May 2007, 1212 patients with ICM were enrolled in the STICH trial, with a median follow-up time of 9.7 years. Consensus clustering analysis identified four distinct clinical phenotypes. Specifically, phenotype 1 (n=415) was characterized by higher hemoglobin (14.4 ± 1.4 g/dL) and hyperlipidemia rate (75.2%). Phenotype 2 (n=263) was younger (56.9 ± 8.5 years) and tended to be male (92.4%), with lower LVESVI (69.2 ± 25.2 mL/m 2 ). Phenotype 3 (n=280) was older (62.3 ± 9.5 years), with higher creatinine (1.3 ± 0.7 mg/dL) and lower LVEF (24.0 ± 7.8 %). Phenotype 4 (n=254) had higher LVEF (31.4 ± 8.7 %), and higher rates of diabetes mellitus (62.2%) and hypertension (72.8%). Compared with other phenotypes, phenotype 3 had the highest risk of all-cause mortality (HR 1.69, 95% CI 1.41-2.02; P <0.001) and cardiovascular mortality (HR 1.80, 95% CI 1.45-2.22; P <0.001). Among phenotype 2, CABG was associated with a significantly lower risk of all-cause mortality (HR 0.70, 95% CI 0.50–0.98; P =0.033) and cardiovascular mortality (HR 0.63, 95% CI 0.43–0.92; P =0.015) compared with medical therapy alone. Conclusions: We identified four clinical phenotypes with distinct outcomes and HTEs among ICM patients. Phenotype 3 (older, higher creatinine, lower LVEF) had the poorest clinical outcomes. Phenotype 2 (younger, male, lower LVESVI) was more likely to derive greater survival benefits from CABG.