Elevated lipoprotein(a) levels are associated with myocardial infarction (MI) in some but not all studies. Limitations of previous studies include lack of risk estimates for extreme lipoprotein(a) levels, measurements in long-term frozen samples, no correction for regression dilution bias, and lack of absolute risk estimates in the general population. We tested the hypothesis that extreme lipoprotein(a) levels predict MI in the general population, measuring levels shortly after sampling, correcting for regression dilution bias, and calculating hazard ratios and absolute risk estimates.