Lipoprotein data from 9477 subjects, covering a wide range of total plasma cholesterol levels, were used to examine the validity of the Friedewald formula for estimating plasma concentrations of low-density lipoprotein cholesterol (LDL-C) using high-density lipoprotein cholesterol (HDL-C) and triglyceride (TG) concentrations. Values of LDL-C obtained from the Friedewald formula were compared with values of LDL-C derived from preparative ultracentrifugation used as a reference method. We found that the bias associated with the Friedewald formula was not related to plasma LDL-C levels and was smaller than -4.0% even for plasma LDL-C values <3.0 mmol/l. Moreover, in the subgroup of individuals with plasma TG levels < or =4.5 mmol/l, the Friedewald formula underestimated LDL-C levels with a bias between -3.1% and -1.9% according to TG quartiles. Interestingly, the Friedewald formula showed no significant bias in patients with plasma TG levels between 4.51 and 8.82 mmol/l, suggesting that the calculated LDL-C are reliable and could be clinically useful in patients with plasma TG levels higher than 4.5 mmol/l which is the reference cut-point value used by most clinical laboratories. Finally, multiple regression analyses showed that the very low-density lipoprotein cholesterol (VLDL-C)/TG ratio represented nearly 63% (P < 0.0001) of the variance of the bias associated with the Friedewald formula. We concluded that the Friedewald formula may be reliable at low LDL-C levels and at TG levels up to 9 mmol/l but may be used with caution when the VLDL-C/TG ratio is high as observed in patients with type III dysbetalipoproteinemia.