Cardiovascular disease (CVD) remains the leading cause of morbidity and mortality globally, posing a great burden to human health and public health resources. Recent advances in comprehensive high-throughput metabolomics techniques have opened opportunities for the large-scale discovery of novel biomarkers and improvement of the understanding of disease aetiology. This thesis aimed to investigate the associations of common CVD related factors (CVRFs) with metabolite profiling and the associations of metabolite profiling with CVD outcomes.

I first conducted a comprehensive review of the published literature on the topic to provide background. Next, I conducted cross-sectional analyses assessing the associations of ten common CVRFs with the human metabolic profile across 246 measures quantified by nuclear magnetic resonance, using data from 131,976 participants without prior vascular disease from the UK Biobank (UKB) and INTERVAL studies. Overall, age, sex, ethnicity, history of diabetes, smoking, alcohol consumption, lipid-lowering medication, physical activity, body mass index, and systolic blood pressure collectively explained 13.7% of the variability in the metabolic profile, with 6.9% explained by modifiable factors. Results also showed differential independent patterns of metabolic associations of ten CVRFs that may potentially explain their associations with CVD risk. Subsequently, I investigated associations of these circulating metabolites with first-onset coronary heart disease (CHD) (2,231 cases) and stroke (1,787 cases), as well as all-cause mortality (6,562 deaths) among 148,126 UKB and INTERVAL participants. I also secondarily assessed and compared the metabolic associations across initial manifestations of 14 cardiovascular conditions.

In further analyses aiming to improve understanding of lipoprotein characteristics and CVD, I assessed detailed associations of particle concentrations and sizes of lipoproteins with incident cardiovascular events among 135,035 participants (~ 3,800 cases) from UKB and INTERVAL studies, as well as meta-analyses with all prior evidence, involving a total of > 312,000 individuals and > 19,000 events. Concentrations of high-density lipoprotein (HDL) particles (HDL-P) were always associated with lower risks of cardiovascular events independent of HDL cholesterol (HDL-C), whereas HDL-P tended to attenuate associations of HDL-C with events towards null. Similar patterns were noted for low-density lipoprotein (LDL) particles. Despite the potential opposite pattern for very-low-density lipoprotein (VLDL) and CHD, the smallest-sized VLDL particle (VLDL-P) subclass, which comprised the largest proportion of cholesterol content, was strongly associated with higher risks of cardiovascular events independent of HDL, VLDL, LDL, or triglycerides.

Finally, I assessed detailed associations of circulating fatty acid (FA) subtypes, either individually (n= 28) or in combination (n= 18), with incident cardiovascular events among 172,891 individuals (~ 20,000 cases) from the European Prospective Investigation into Cancer and Nutrition-CVD (EPIC-CVD), UKB, and INTERVAL studies. I conducted a systematic literature search and updated meta-analysis to contextualise the current study findings and provide a quantitative synthesis of all available evidence thus far. Final results included up to 22,802 CHD and 14,221 stroke cases and provided the strongest evidence yet on the topic. Key findings were that cardiovascular risk associations differed importantly by saturated FA (SFA) subtypes (positive associations for even-chain SFA and negative associations with odd-chain SFA) and supported a favourable role of higher long-chain n-3 polyunsaturated FAs and higher linoleic acids on cardiovascular health – a finding generally consistent with some available trial evidence.

Overall, this PhD thesis provides an up-to-date summary of the current evidence on high-dimensional metabolite profiling and the epidemiology of CVD. Using the most powerful data sources available to date, it illustrates and quantifies the metabolic patterns associated with common CVRFs and reveals key metabolic signatures of multiple cardiovascular conditions. It further highlights the importance of lipoprotein particles per se beyond the traditional cholesterol biomarkers and challenges the current broad dietary recommendations focused solely on lowering total SFA intake. The findings from this thesis contribute to the understanding of cardiovascular health from the perspective of human metabolic profiles, which may inform future interventions and could have important implications for the improvement of personalized disease prevention and management.