Cardiometabolic and Psychobehavioral Phenotypes Define Cardiovascular Risk Heterogeneity in Rheumatoid Arthritis
TL;DR
Among adults with rheumatoid arthritis (RA), distinct data-driven phenotypes were identified based on combined patterns of metabolic, inflammatory, renal, behavioral, and psychosocial factors, with differing burdens of prevalent cardiovascular disease (CVD).
Abstract
Abstract Cardiovascular disease (CVD) is a major cause of morbidity and premature mortality in people with rheumatoid arthritis (RA). Whether multidimensional cardiometabolic, inflammatory, renal, behavioral, and psychosocial factors define distinct cardiovascular risk phenotypes in RA remains unclear. This study aimed to identify data-driven phenotypes and evaluate their associations with prevalent CVD. We analyzed 3,252 adults aged over 20 years with self-reported RA from the 2005-2018 National Health and Nutrition Examination Survey (NHANES). Unsupervised partitioning around medoids (PAM) clustering using Gower distance identified distinct data-driven cardiovascular-risk profiles. Cluster validity was assessed using silhouette analysis and internal train test validation. Survey-weighted binary logistic regression evaluated associations between phenotype membership and prevalent CVD. Six clinically interpretable data-driven phenotypes were identified with varying CVD prevalence: Severe Metabolic Diabetic (40.3%), Aging Diabetic Hypertensive (34.9%), Non-Diabetic Intermediate (20.2%), Mild Metabolic (14.8%), Cardiometabolic (10.4%), and Smoking-Predominant (19.4%). Compared with the Cardiometabolic phenotype, the Severe Metabolic Diabetic phenotype exhibited the highest odds of prevalent CVD (aOR 4.11, 95%CI:2.56-6.59), followed by the Aging Diabetic-Hypertensive phenotype (aOR 3.38, 95%CI:2.22-5.15). Higher odds of prevalent CVD were also observed in the Smoking-Predominant (aOR 1.99, 95% CI: 1.26-3.13) and Non-Diabetic Intermediate (aOR 1.77, 95% CI: 1.10-2.84) phenotypes. The six-cluster solution demonstrated moderate separation and strong internal correspondence across training and testing sets, while the phenotype-informed model showed moderate discrimination for prevalent CVD (AUC = 0.72). Among adults with rheumatoid arthritis (RA), distinct data-driven phenotypes were identified based on combined patterns of metabolic, inflammatory, renal, behavioral, and psychosocial factors, with differing burdens of prevalent cardiovascular disease (CVD). Phenotype-based approaches may provide a comprehensive framework for characterizing multidimensional cardiovascular risk in RA.