Abstract Increased blood levels of omega-3 fatty acids have consistently been shown to be associated with reduced risk of cardiovascular diseases (CVD). However, large scale placebo-controlled trials of omega-3 supplementation have had discordant results: in 2018 REDUCE-IT reported a 25% risk reduction, whereas in 2020 STRENGTH reported no risk reduction. The two trials had both different placebos (mineral vs corn oil) and omega-3 formulations (EPA vs DHA+EPA). There is not yet a consensus on what explains these results. Here we describe the profiling of omega-3 fatty acid subclasses, eicosapentaenoic acid (EPA), docosahexaenoic acid (DHA) with nuclear magnetic resonance (NMR), and test their associations with incident CVD in 500,000 individuals from UK Biobank. We utilized 500 MHz NMR spectrometers to analyze total omega-3 and its subclasses in EDTA plasma samples and employed biomarker-specific quantification algorithms to measure their absolute concentrations. In this study, we developed a novel algorithm to quantify EPA, complementing previously described algorithms for omega-3 and DHA. Correlation (r^2) of this novel NMR-based quantification and gas chromatography-based measurements was 0.77 for EPA. On average, DHA comprised 44.5% of the total omega-3 concentration, EPA accounted for 22.2%, and other subclasses constituted 34.3% in the cohort. We computed age and sex adjusted associations of omega-3, DHA, and EPA with 10-year risk of CVD among individuals who were not taking cholesterol lowering medication. We discovered protective associations: for DHA a hazard ratio of 0.85 (0.86-0.89, 95% confidence interval) per standard deviation versus 0.96 (0.95-0.98) for EPA, and 0.95 (0.93-0.97) for total omega-3 (Figure). Since DHA (used in the STRENGTH trial with a null result) has a significantly stronger epidemiological association than EPA, these results from a large-scale observational study do not support the hypothesis that the differential use of DHA and EPA in the REDUCE-IT and STRENGTH trials explains their divergent results.Figure