Association Between Inflammatory Biomarkers And Carotid Ultrasound Parameters In The Northern Manhattan Stroke Study

Inflammation contributes to atherosclerosis and vascular remodeling, but the immune pathways associated with distinct carotid ultrasound phenotypes, including carotid intima-media thickness (cIMT), maximal plaque thickness (MPT), and arterial stiffness remain incompletely characterized. We investigated associations between circulating immune biomarkers and carotid ultrasound measures in a stroke-free, multi-ethnic cohort. We analyzed data from 1,134 stroke-free participants (mean age 70 ± 9 years, 59% women, 66% Hispanic) in the Northern Manhattan Study (NOMAS) who underwent high-resolution B-mode carotid ultrasound. cIMT and MPT were measured using automated edge-detection software, while arterial stiffness was assessed via systolic and diastolic diameter measurements, strain, and β-stiffness. Plasma levels of 60 immune biomarkers were quantified using a multiplex immunoassay. Biomarker selection was performed using LASSO, Random Forest, and XGBoost; markers selected by ≥ 2 methods were prioritized. Associations were tested using multivariable linear regression adjusted for age, sex, race/ethnicity, and cardiovascular risk factors (hypertension, hyperlipidemia, diabetes, and smoking). Machine-learning analyses identified partially distinct candidate biomarker profiles for cIMT, MPT, and arterial stiffness. In fully adjusted regression models, greater cIMT was associated with higher IL-4 and leptin levels and lower IL-10, VCAM-1, and SerpinE1 levels. None of the machine-learning-selected biomarkers was independently associated with MPT at the conventional p < 0.05 threshold. Greater arterial stiffness was associated with higher SCF levels, while CXCL10 demonstrated a nonsignificant inverse trend. Models incorporating the selected biomarkers had modestly higher adjusted R² values than models containing clinical covariates alone (cIMT, 0.072 versus 0.109; MPT, 0.134 versus 0.147; arterial stiffness, 0.063 versus 0.075). In a stroke-free, multiethnic cohort, machine-learning analyses identified distinct and overlapping candidate immune biomarker profiles across cIMT, plaque thickness, and arterial stiffness. However, only a subset of selected biomarkers demonstrated independent associations in fully adjusted regression models. These findings support a hypothesis-generating role for inflammatory and immunometabolic pathways in subclinical carotid disease and vascular aging and require validation in longitudinal studies.