The accuracy of the Friedewald formula in estimating low-density lipoprotein (LDL) cholesterol was investigated in 47 alcoholic patients with liver disease (21 minimal-change, 26 cirrhotic) by comparing the results with those obtained by sequential preparative ultracentrifugation. In 14% of subjects with minimal-change disease, the error in the estimated LDL cholesterol was 50% +/- 9% (mean +/- SD; range 40-59%) and was related to the degree of attendant hypertriglyceridemia (r = 0.98; P < 0.001). A similar degree of error was observed in patients with cirrhosis, despite the absence of hypertriglyceridemia; an abnormal VLDL cholesterol: triglyceride ratio was the contributory factor in the discrepancy. We conclude that, as is the case in other clinical pathologies in which abnormalities of lipoprotein composition have been described (e.g., diabetes), the Friedewald formula to estimate LDL cholesterol may be inappropriate in chronic alcoholics, particularly those in whom a degree of hepatic dysfunction may be suspected.