Objectives: Although 1H-nuclear magnetic resonance (NMR)-based lipid/metabolomics has been used to detect atherosclerosis, data regarding lipid/metabolomic signature in rheumatoid arthritis (RA)-related atherosclerosis are scarce. We aimed to identify the distinct lipid/metabolomic profiling and develop a prediction score model for RA patients with subclinical atherosclerosis (SA).
Methods: Serum levels of lipid metabolites were determined using 1H-NMR-based lipid/metabolomics in 65 RA patients and 12 healthy controls (HCs). The occurrence of SA was defined as the presence of carotid plaques revealed in ultrasound images.
Results: Compared with HC, RA patients had significantly higher levels of phenylalanine and glycoprotein acetyls (GlycA) and lower levels of leucine and isoleucine. RA patients with SA had significantly higher levels of phenylalanine, creatinine, and glycolysis_total and lower levels of total lipid in HDL(HDL_L) than RA patients without SA. The Lasso logistic regression analysis revealed that age, creatinine, HDL_L, and glycolysis_total were significant predictors for the presence of SA. The prediction scoring algorithm was built as ( -0.657 + 0.011*Age + 0.004*Creatinine -0.120*HDL_L + 0.056*glycolysis-related measures), with AUC 0.90, sensitivity 83.3%, and specificity 87.2%. Serum phenylalanine levels were significantly decreased, and the levels of HDL_L and HDL_Particle were significantly increased in 20 RA patients, paralleling the decrease in disease activity score for 28-joints.
Conclusions: With 1H-NMR-based lipid/metabolomics, distinct profiling of lipid metabolites was identified between RA patients and HC or between RA patients with and without SA. We further developed a scoring model based on lipid/metabolomics profiling for predicting RA-associated SA.