In order to evaluate cardiovascular risk, we analyzed the lipid composition of HDL and the presence of lipoprotein(a) [Lp(a)] by both agarose gel electrophoresis and enzyme-linked immunoassay (ELISA). In 681 plasmas we found a close correspondence between the existence of a visible sinking pre-~ lipoprotein band and a concentration of Lp(a) higher than 300 mg/L. In the sinking pre-~(+) samples, the HDL-cholesterol level obtained by differential ultracentrifugation was significantly higher than that obtained by precipitation with the MgCl 2-phosphotungstic acid reagent; and the difference between these HDL-cholesterol values was linearly correlated with plasma Lp(a) concentration. Moreover, the other HDL lipid components and the lipid mass ratios of HDL isolated by ultracentrifugation were significantly different from those of HDL isolated by precipitation, and these changes were also correlated with plasma Lp(a). These differences are attributed to Lp(a) because it was detected in the 1 .063-1.21 kg/L plasma fractions, whereas it was absent in the plasma supernates after precipitation with MgCl2- phosphotungstic acid. Although to a lesser extent, Lp(a) was also present in the LDL and VLDL density ranges and it directly depended on both the Lp(a) and the triglyceride plasma concentrations. The proportion of Lp(a) in HDL as related to that in LDL density fractions decreased as Lp(a) plasma levels increased, reflecting an interindividual variation of Lp(a) density species. Since 90% of our study population had detectable Lp(a) in plasma, the results reinforce the concept that the ultracentrifugation method is not equivalent to precipitation in most samples, and the contaminant effect of Lp(a) cannot be predicted because of Lp(a) partition into the different lipoprotein fractions.