Low-density lipoprotein-cholesterol (LDL-C) is currently estimated clinically by using the Friedewald formula, when plasma triglycerides are < 4000 mg/L, or as the difference between infranatant and high-density lipoprotein-cholesterol (HDL-C) values after ultracentrifugation of plasma at native density, when plasma triglycerides are > or = 4000 mg/L (beta quantification). HDL-C is measured by selective precipitation of apolipoprotein B-containing lipoproteins from whole plasma or from the density > 1.006 kg/L infranatant. We compared these conventional methods for LDL-C and HDL-C with "high-performance" gel chromatography (HPGC), a method that directly and simultaneously measures both LDL-C and HDL-C in a single, microliter volume of plasma. Not surprisingly, we found that the results by all these methods were highly correlated. However, LDL-C values were significantly higher and HDL-C values significantly lower by the direct HPGC method than by the conventional methods (paired t-test). In addition, both Bland-Altman plots and concordance correlation analyses indicated lack of agreement between the methods' results in the majority of patients' subgroups.