Abstract
Plasma (1)H-NMR metabolomic measures have yielded significant insight into the pathophysiology of cardiometabolic disease, but their inter-related nature complicates causal inference and clinical interpretation. This study aimed to investigate the associations of unrelated (1)H-NMR metabolomic profiles with coronary artery disease (CAD) and ischemic stroke (IS). Principal component (PC) analysis was performed on 168 (1)H-NMR metabolomic measures in 56,712 unrelated European participants from UK Biobank to retrieve uncorrelated PCs, which were used in Cox-proportional hazard models. For each outcome, two-sample Mendelian randomization analyses were then conducted based on three nonoverlapping databases, followed by a meta-analysis. The first six PCs collectively explaining 88% of the total variance were identified. For CAD, results from Cox and Mendelian randomization analyses were generally directionally consistent. The pooled odds ratios (95% CI) for CAD per one-SD increase in genetically influenced PC1 and PC3 (both characterized by distinct apolipoprotein B [ApoB]-associated lipoprotein profiles) were 1.04 (1.03, 1.05) and 0.94 (0.93, 0.96), respectively. Besides, the pooled odds ratio (95% CI) for CAD per one-SD increase in genetically influenced PC4, characterized by simultaneously decreased small HDL and increased large HDL, and independent of ApoB, was 1.05 (1.03, 1.07). For IS, increases of PC3 and PC5 (characterized by increased amino acids) were associated with a lower risk and a higher risk, respectively. This study confirms associations of ApoB-associated lipoprotein profiles with CAD and IS, and highlights the possible existence of an ApoB-independent lipoprotein profile, characterized by a distinctive HDL subparticle distribution, driving CAD.