Background: Apolipoprotein B (apoB) levels, reflecting the total number of atherogenic lipoproteins, is increasingly recognized as the optimal lipid risk predictor. Whether the lipoprotein type, cholesterol content, or size of apoB particles adds predictive value for atherosclerotic risk beyond total apoB is unclear. We explored the relative importance of different apoB particle classes and size in the risk of developing coronary artery disease (CAD). Methods: This was a prospective analysis of UK Biobank data involving 207,368 participants with available lipoprotein profiling and without prevalent coronary, peripheral artery disease, stroke, diabetes or taking lipid-lowering medication. We examined associations of nuclear magnetic resonance (NMR)-measured non-Lp(a) apoB particle count, concentrations of individual lipoprotein classes (very low [VLDL], intermediate [IDL], low-density lipoprotein [LDL]) and size subclasses (analyzed per standard deviation [1-SD], per-particle [equimolar] or as % of apoB increase), average particle diameter, and immunoassay-measured Lp(a) with incident CAD. Results: NMR non-Lp(a) apoB was associated with a 33% higher risk of developing CAD per 1-SD (equal to 370 nmol/L) increase (adjusted hazard ratio [aHR] 1.33, 95%CI: 1.30-1.37). A higher proportion of VLDL relative to other apoB-containing particles (% VLDL) was associated with an increased risk of CAD (aHR per 1% absolute increase in %VLDL 1.03, 1.01-1.05). However, the addition of % VLDL or other VLDL or LDL parameters to a model with clinical covariates and NMR non-Lp(a) apoB did not improve prediction accuracy. In contrast to detailed phenotyping of VLDL and IDL/LDL, the inclusion of Lp(a) added independent prognostic value for CAD (aHR 1.18, 1.16-1.20). Conclusions: Lipid-related atherosclerotic risk is most accurately reflected by the total count of apoB and is largely unaffected by the major particle type (VLDL, IDL/LDL) or size. Elevated levels of Lp(a) adds additional risk, and thus adequate assessment of atherogenic risk from dyslipidemia is best accomplished by consideration of both apoB and Lp(a) levels.