Use of the Friedewald formula to estimate LDL-cholesterol in patients with chronic renal failure on dialysis.

Increased plasma cholesterol, particularly that portion associated with LDL, is an established risk factor for coronary heart disease [1, 2]. As a consequence, it is widely recommended [3] that LDL-cholesterol be determined in individuals with increased total cholesterol. Because isolation of the LDL fraction requires ultracentrifugation, a technique not generally available in service laboratories, the concentration of LDL-cholesterol is usually calculated by the formula of Friedewald et al. [4]. The Friedewald formula provides an adequate estimate of LDL-cholesterol for most fasting specimens but is known to be less reliable as triglyceride concentration increases [5]. Moreover, in patients with chronic renal failure, in whom an accumulation of remnant particles and increased concentrations of IDL-cholesterol have been found [6, 7], the estimate is possibly further compromised [8]. Here we report results on specimens obtained from patients on dialysis with chronic renal failure that show no greater disparity between measured and estimated LDL-cholesterol concentrations than those seen in other patient groups in which the Friedewald formula is frequently used. Additionally we have reexamined the effect of IDL and VLDL composition on the reliability of estimations made with the formula. Two sets of blood specimens taken 6 months apart were obtained from 106 patients receiving hemodialysis or continual ambulatory peritoneal dialysis [9]. Within this 6-month period, some patients received an inhibitor of hydroxymethylglutaryl-CoA reductase, some an inhibitor of angiotensin-converting enzyme, some both treatments, and some placebo only. Blood collected after an overnight fast and separated within 4 h provided serum specimens for analysis. Total cholesterol and triglyceride were assayed enzymically with reagents from Boehringer Mannheim. HDLcholesterol was assayed after precipitation of the apo B-containing lipoproteins with phosphotungstate and magnesium ions [10]. The performance of these assays was regularly certified by reference to samples provided by the Centers for Disease Control and Prevention, Atlanta, GA. Ultracentrifugation was carried out as described by others [11] for 15 h at 120 000g with a TFT 80.4 fixed-angle rotor in a Centrikon T-2070 centrifuge (Kontron Instruments). Cholesterol and triglyceride contents of supernatant fractions obtained after the tubes were sliced [11] were measured; the fraction floating at a relative density (d) of 1.006 was taken to represent VLDL and that floating at d 1.019 obtained by adjustment of 20 volumes of serum with 1 volume of NaBr solution (d 5 1.27, ;3.9 mol/L) was taken to represent VLDL plus IDL. The contents of IDL alone were calculated by subtraction. When in other experiments this technique was extended to permit sequential isolation of LDL and of HDL [11], the mean recovery of total cholesterol was 90% (range 86– 95%) and of triglyceride was 92% (range 88–95%). LDL-cholesterol was calculated as total cholesterol 2 HDL-cholesterol 2 VLDL-cholesterol; VLDL-cholesterol was either measured directly (after ultracentrifugation) or calculated as 0.456 3 total triglyceride concentration expressed in mmol/L (Friedewald). Analysis of 204 specimens with complete data from the 106 patients showed concentrations (mmol/L) of total cholesterol of 3.16–9.25 (median 6.00), triglyceride of 0.48–7.74 (median 2.01), VLDL-cholesterol of 0.09–3.78 (median 0.92), IDL-cholesterol of 0.15–1.58 (median 0.45), and HDL-cholesterol of 0.36–2.28 (median 0.93). LDLcholesterol derived from ultracentrifugation data ranged from 1.13 to 7.27 (median 3.92) mmol/L. Thirty-six specimens (18%) had a LDL-cholesterol calculated from the Friedewald formula that differed from the ultracentrifugation value by .10%. This proportion of specimens with significant error is similar to those proportions (15–19%) found in much larger series of patients of more diverse clinical conditions [12, 13]. The proportion is substantially less than the 42% of 45 individuals with chronic renal failure reported by others [8], despite the lower lipid values in their series. Recognition that the Friedewald formula is unreliable in the presence of chylomicronemia and hypertriglyceridemia has led to the use of various indices for the exclusion of such specimens. Total triglyceride concentration commonly is used [5], although a low total cholesterol:triglyceride ratio has been proposed as a more relevant criterion for rejection [14]. Additionally, others have identified increased cholesterol:triglyceride ratios in VLDL and IDL and increased IDL-cholesterol alone as being associated with increased error in the formula [8]. We plotted the number of errors detected against each index arranged in centiles (Fig. 1). The median (range) for each index was: total triglyceride concentration, 2.01 (0.48–7.74) mmol/L; total cholesterol:triglyceride ratio, 2.92 (0.86–7.86); cholesterol:triglyceride ratio in VLDL, 0.75 (0.31–1.28); IDL-cholesterol 0.45 (0.15–1.58) mmol/L; cholesterol:triglyceride ratio in IDL, 2.69 (1.11–6.80). With total triglyceride concentration and total cholesterol:triglyceride ratio (by inverse ranking so that increased incidence of error was positively associated with the index), error detection was similar with the number of errors initially small but rising steeply from the 60th centile. By contrast, with the ratio of cholesterol:triglyceride in VLDL and in IDL and with IDL-cholesterol concentration, the errors were more evenly distributed among the centiles. Quantitatively, 18 errors (50%) were included when triglyceride concentration was 3.51 mmol/L (86th centile), total cholesterol:triglyceride ratio was 1.91 (81st centile), cholesterol:triglyceride ratio in VLDL was 0.82 (70th centile), IDL-cholesterol was 0.53 mmol/L (65th Technical Briefs