Elevated levels of lipoprotein(a) [Lp(a)] and the presence of small isoforms of apolipoprotein(a) [apo(a)] have been associated with coronary artery disease (CAD) in whites but not in African Americans. Because of marked race/ethnicity differences in the distribution of Lp(a) levels across apo(a) sizes, we tested the hypothesis that apo(a) isoform size determines the association between Lp(a) and CAD. We related Lp(a) levels, apo(a) isoforms, and the levels of Lp(a) associated with different apo(a) isoforms to the presence of CAD (>/=50% stenosis) in 576 white and African American men and women. Only in white men were Lp(a) levels significantly higher among patients with CAD than in those without CAD (28.4 versus 16.5 mg/dL, respectively; P:=0.004), and only in this group was the presence of small apo(a) isoforms (<22 kringle 4 repeats) associated with CAD (P:=0.043). Elevated Lp(a) levels (>/=30 mg/dL) were found in 26% of whites and 68% of African Americans, and of those, 80% of whites but only 26% of African Americans had a small apo(a) isoform. Elevated Lp(a) levels with small apo(a) isoforms were significantly associated with CAD (P:<0.01) in African American and white men but not in women. This association remained significant after adjusting for age, diabetes mellitus, smoking, hypertension, HDL cholesterol, LDL cholesterol, and triglycerides. We conclude that elevated levels of Lp(a) with small apo(a) isoforms independently predict risk for CAD in African American and white men. Our study, by determining the predictive power of Lp(a) levels combined with apo(a) isoform size, provides an explanation for the apparent lack of association of either measure alone with CAD in African Americans. Furthermore, our results suggest that small apo(a) size confers atherogenicity to Lp(a).