A high serum lipoprotein(a) [Lp(a)] level, which is genetically determined by apolipoprotein(a) [apo(a)] size polymorphism, is an independent risk factor for coronary atherosclerosis. However, the associations among Lp(a) levels, apo(a) phenotypes, and myocardial infarction (MI) have not been studied. Patients with MI (cases, n = 101, M/F: 86/15, age: 62+/-10y) and control subjects (n = 92, M/F: 53/39, age: 58+/-14y) were classified into quintile groups (Groups I to V) according to Lp(a) levels. Apo(a) isoform phenotyping was performed by a sensitive, high-resolution technique using sodium dodecyl sulfate-agarose/gradient polyacrylamide gel electrophoresis (3-6%), which identified 26 different apo(a) phenotypes, including a null type. Groups with higher Lp(a) levels (Groups II, III, and V) had higher percentages of MI patients than that with the lowest Lp(a) levels (Group I) (54%, 56%, or 75% vs. 32%, p<0.05). Groups with different Lp(a) levels had different frequency distributions of apo(a) isoprotein phenotypes: Groups II, III, IV, and V, which had increasing Lp(a) levels, had increasingly higher percentages of smaller isoforms (A1-A4, A5-A9) and decreasingly lower percentages of large isoforms (A10-A20, A21-A25) compared to Group I. An apparent inverse relationship existed between Lp(a) and the apo(a) phenotype. Subjects with the highest Lp(a) levels (Group V) had significantly (p<0.05) higher serum levels of total cholesterol, apo B, and Lp(a). Patients with MI and the controls had different distributions of apo(a) phenotypes: i.e., more small isoforms and more large size isoforms, respectively (A1-A4/A5-A9/A10-A20/A21-A25: 35.7%/27.7%/20.8%/15.8% and 22.8%/23.9%/29.4%/23.9%, respectively). Lp(a) (parameter estimate +/- standard error: 0.70+/-0.20, Wald chi2 = 12.4, p = 0.0004), apo(a) phenotype (-0.43+/-0.15, Wald chi2 = 8.17, p = 0.004), High-density lipoprotein-cholesterol, apo A-I, and apo B were significantly associated with MI after adjusting for age, gender, and conventional risk factors, as assessed by a univariate logistic regression analysis. The association between Lp(a) and MI was independent of the apo(a) phenotype, but the association between the apo(a) phenotype and MI was not independent of Lp(a), as assessed by a multivariate logistic regression analysis. This association was not influenced by other MI- or Lp(a)-related lipid variables. These results suggest that apo(a) phenotype contributes to, but does not completely explain, the increased Lp(a) levels in MI. A stepwise logistic regression analysis with and without Lp(a) in the model identified Lp(a) and the apo(a) phenotype as significant predictors for MI, respectively.