Evaluation of five methods for determining low-density lipoprotein cholesterol (LDL-C) in hemodialysis patients

Clinical Biochemistry 34 (2001) 593 602 Evaluation of five methods for determining low-density lipoprotein cholesterol (LDL-C) in hemodialysis patients Eleni Bairaktari a, *, Moses Elisaf b, Christos Tzallas a, Sonia Anna Karabina d, Alexandros D. Tselepis d, Kostas C. Siamopoulos c, Orestes Tsolas a a Laboratory of Biochemistry, b Department of Internal Medicine, and c Department of Nephrology, University Hospital, Medical School, University of Ioannina, 450 00, Ioannina, Greece, and d Department of Chemistry, University of Ioannina, 450 00, Ioannina, Greece Received 30 July 2001; received in revised form 15 November 2001; accepted 15 November 2001 Abstract Objectives: Current recommendations for the management of dyslipidemia are largely based on the concentration of LDL-C. Most clinical laboratories estimate the concentration of LDL-C by the recommended routine method, the equation of Friedewald, in specimens from fasting subjects and with TG concentrations 4.52 mmol/l. Because of the limitations of the Friedewald calculation, direct methods for an accurate quantification of LDL-C are needed. Design and Methods: In the present study we evaluated the accuracy of the following 5 different procedures for LDL-C in 98 patients on hemodialysis: the Friedewald equation, where LDL-C is calculated from HDL-C, measured either by the precipitation procedure with dextran sulfate-mg 2 (Method 1), or by a direct HDL-C assay (Method 2), the Direct LDL TM assay (Method 3), the homogeneous N-geneous TM LDL assay (Method 4) and the calculated LDL-C values deriving from the ApoB based equation: 0.41TC - 0.32TG 1.70ApoB - 0.27, (Clin Chem 1997;43:808 815) (Method 5). Results: All five LDL-C methods were found to be in good agreement with ultracentrifugation/dextran sulfate-mg 2 precipitation with the coefficients of correlation of the assays to ranging between 0.93 0.95. However, significant differences in the mean values and biases vs. the reference method were observed. The Friedewald equation and the Direct assay were less affected by high LDL-C levels, and they presented higher sensitivity and higher negative predictive value. The N-geneous assay and the ApoB derived calculation were less affected by high triglyceride levels, and they presented higher specificity and higher positive predictive value. At the diagnostic LDL-C level of 3.37 mmol/l, both Friedewald calculations correctly classified 82/92 patients; Direct assay 86/98; N-geneous assay 88/98; and ApoB derived calculation 88/98. At the diagnostic LDL-C level of 2.98 mmol/l, Friedewald calculations (Method 1 and Method 2) correctly classified 82/92 and 81/92 patients, respectively; Direct assay (LDL-3) 87/98; N-geneous assay (LDL-4) 91/98; and ApoB derived calculation (LDL-5) 91/98. Conclusions: Among hemodialysis patients, who commonly present average LDL-C concentrations and high TG levels, the N-geneous assay and the apob derived calculation seem to yield more acceptable results for the estimation of LDL-C. 2002 The Canadian Society of Clinical Chemists. All rights reserved. Keywords: Method comparison; Triglycerides; Hemodialysis patients 1. Introduction Cardiovascular morbidity and mortality are markedly increased in patients with end-stage renal disease [1 2]. Abbreviations: TC: total cholesterol; TG: triglycerides; LDL-C: LDL cholesterol; HDL-C: HDL cholesterol; LDL-UC: LDL cholesterol measured by the ultracentrifugation/dextran sulfate-mg 2 precipitation assay; pospv: positive predictive value; negpv: negative predictive value; CHD: coronary heart disease; HD: hemodialysis patients. * Corresponding author. Tel.: 30-651-99420; fax: 30-651-99418. E-mail address: ebairakt@cc.uoi.gr (E. Bairaktari). Among other risk factors, dyslipidemia is considered critical for the accelerated atherogenesis in these patients [3 4]. Although the cause of this abnormality is probably multifactorial, decreased removal of triglycerides from the circulation is assumed to be one of the most important factors [5 6]. Current recommendations for the management of dyslipidemia are largely based on the concentration of LDL-C [7 8]. According to the National Cholesterol Education Program-Adult Treatment Panel II (NCEP-ATP II), LDL-C values 3.37 mmol/l are considered desirable and those 4.14 mmol/l are considered high. Patients with docu- 0009-9120/01/$ see front matter 2002 The Canadian Society of Clinical Chemists. All rights reserved. PII: S0009-9120(01)00274-0

594 E. Bairaktari et al. / Clinical Biochemistry 34 (2001) 593 602 mented CHD are recommended to maintain their LDL-C concentrations below 2.59 mmol/l. In patients with endstage renal failure, cholesterol concentrations are typically similar to those in the general population, or lower, but this pattern often conceals a highly abnormal lipid subfraction profile with a predominance of atherogenic small dense LDL-particles. Hence, in patients with renal failure reduction of LDL-C concentrations may well lower the risk of cardiovascular disease among those with average (or even below average) LDL-C concentrations [9,10]. The -quantification method, which involves ultracentrifugation and chemical precipitation step, is the reference procedure for measuring LDL-C [11,12]. However, most clinical laboratories estimate the concentration of LDL-C by the recommended routine method, the equation of Friedewald [13], in specimens from fasting subjects and with TG concentrations 4.52 mmol/l. This calculation has contributed greatly to the management of hypercholesterolemia over the past decades, and it has produced acceptable results for the hemodialysis patients as well [14,15]. However, the reliability of the Friedewald equation is considerably decreased even with triglyceride concentrations of 2.26 4.52 mmol/l [16,17]. The validity of calculated LDL-C depends on the accuracy, the precision, and the biologic variation of three other assays, TC, TG and HDL-C, plus a mathematical calculation factor that estimates the amount of cholesterol in the very low density lipoproteins (VLDL). The new direct homogeneous assays for determining HDL-C [18 21] have been proven in recent studies to meet the current performance goal that is a total error 13%, and this improvement is also expected to improve the accuracy of the calculated LDL-C value. Because of the limitations of the Friedewald calculation, direct methods for an accurate quantification of LDL-C are needed [11]. Several new approaches have been published in recent years for the direct measurement of LDL-C or for alternative ways of its calculation [22 31]. A method for measuring LDL cholesterol directly from serum, the Direct LDL TM (Sigma Diagnostics, St. Louis, MO, USA), is based on the immunoseparation of LDL particles from chylomicrons, VLDL, and HDL using antibodies against apolipoproteins A-I and E [22 25]. N-geneous TM LDL assay (Genzyme Diagnostics, Cambridge, MA, USA) is an homogeneous and automated method that uses two detergents for the direct determination of LDL-C in unprocessed serum [26 28]. As each LDL particle contains one molecule of ApoB, the values of total plasma apob concentration have been considered to be a more accurate representation of the atherogenic particles, even for hypertriglyceridemic patients. The formula LDL-C 0.41TC - 0.32TG 1.70ApoB - 0.27 expressed in mmol/l has been proposed for the estimation of LDL-C, using lipid constituents directly measured in total serum, namely total cholesterol, triglycerides and apolipoprotein B [31]. In the present study, the accuracy of the following 5 different procedures for LDL-C has been evaluated in 98 patients on hemodialysis: 1.) the Friedewald equation, where HDL-C is measured by the precipitation procedure with dextran sulfate- Mg 2 [Method-1] 2.) the Friedewald equation, where HDL-C is measured by a direct HDL-C assay [Method-2] 3.) the Direct LDL TM assay (Sigma Diagnostics, St. Louis, MO, USA) [Method 3] 4.) the homogeneous N-geneous TM LDL assay (Genzyme Diagnostics, Cambridge, MA, USA) [Method 4] and 5.) the calculated LDL-C values deriving from the ApoB based equation [Method 5] These five procedures were compared with the ultracentrifugation/dextran sulfate-mg 2 precipitation method (LDL-UC) and their ability to classify HD patients into the risk categories was also examined. In addition, the influence of increased concentrations of LDL-C and TGs on the LDL-C methods were tested by bias plots. This study may contribute to the selection of an accurate and convenient methodology for the estimation of the atherogenic LDL-C in hemodialysis patients who commonly exhibit both quantitative lipoprotein abnormalities, such as hypertriglyceremia and low HDL-C, but also important qualitative lipoprotein changes, which can interfere with the routine laboratory measurements [5 6]. 2. Materials 2.1. Samples The population studied consisted of 98 patients undergoing hemodialysis (HD). Clinical and laboratory parameters of the study population were as follows: Age: 57 (29 78); Sex (F/M): 43/55; duration of dialysis: 21.6 1.9 months; Serum total cholesterol (mmol/l): 4.97 1.42; Triglycerides (mmol/l): 2.26 1.20; HDL-Cholesterol (mmol/l): 0.94 0.30. Blood samples were obtained in the morning after an overnight fast from all subjects before the dialysis session. Serum was isolated within 2 h (centrifugation at 1500 g for 15 min) and was stored at 4 C until analysis (generally within three days). Written informed consent was obtained from all subjects. The study was approved by the Scientific Committee of our hospital. 3. Methods 3.1. Total cholesterol and triglycerides measurements Concentrations of total cholesterol (TC) and triglycerides (TG) were determined enzymatically on the Olym-

E. Bairaktari et al. / Clinical Biochemistry 34 (2001) 593 602 595 pus AU560 Clinical Chemistry analyser (Olympus Diagnostica, Hamburg, Germany). Our laboratory is currently participating in the Murex Clinical Chemistry Quality Assessment Program. Our CV values in this program in the past two years (four cycles) have ranged between 0.7 and 1.1% for cholesterol, and between 0.9 and 2.5% for triglycerides. 3.2. Dextran sulfate-mg assay This assay uses dextran sulfate and MgCl 2 to precipitate all lipoproteins except HDL, which remains in the supernatant and is assayed [32]. 3.3. Direct (homogeneous) HDL-C assay The direct Olympus HDL-C assay was performed according to the manufacturer s instructions on the Olympus AU560 Clinical Chemistry analyser [20]. 3.4. Friedewald calculation The LDL-C was calculated using the equation LDL-C TC HDL-C TG/2.2 expressed in mmol/l, excluding samples with TG concentrations 4.52 mmol/l. HDL-C was measured either by the precipitation with Dextran sulfate- Mg 2 assay or by the direct assay mentioned above. 3.5. Direct LDL TM assay (Sigma Diagnostics, St. Louis, MO, USA) LDL were isolated by the direct immunoseparation method according to the manufacturer s instructions. 200 l of LDL-C reagent was pipetted into separation tubes and then 30 l of serum was added. After vortexing immediately, the tubes were incubated for 10 min and centrifuged at 6000 g for 5 min. The cholesterol in the filtrate was measured on a Olympus AU560 Clinical Chemistry analyser using a calibration curve suitable for low cholesterol values. 3.6. Homogeneous N-geneous TM LDL assay (Genzyme Diagnostics, Cambridge, MA, USA) Reagent 1 contains a detergent, which disrupts selectively all non-ldl lipoproteins, and the cholesterol released is removed in a colorless reaction. The second reagent disrupts LDL to release its cholesterol, which is measured using a conventional chromogen system. The test was performed according to the manufacturers instructions using the calibrator and controls included in the test kits. 3.7. ApoB derived calculation The LDL-C was calculated using the equation LDL-C 0.41TC 0.32TG 1.70ApoB - 0.27, expressed in mmol/l. 3.8. Ultracentrifugation/Dextran sulfate-mg assay In this assay VLDL and chylomicrons are removed by ultracentrifugation before LDL is precipitated by dextran sulfate-mgcl 2. HDL-C is measured in the supernatant. Serum samples of 3 ml were ultracentrifuged at d 1.006 Kg/L in a Beckman L7 65 ultracentrifuge at 40 000 rpm, 14 C for 10 h, using type NVT 65 rotor (Beckman Instruments, Fullerton, CA, USA). After ultracentrifugation, the tubes were sliced to remove the floated VLDL and chylomicron particles. The volume of the infranate was then adjusted to 3 ml by adding a solution of 9 g/l NaCl and HDL-C was determined as above after precipitation of the apo B-containing lipoproteins (mainly LDL) by dextran sulfate-mgcl 2. 3.9. Apolipoproteins Apo B was measured with a Dade Behring Nephelometer BN100, and reagents (antibodies and calibrators) from Dade Behring Diagnostics GmbH (Liederbach, Germany). The assay was calibrated according to the IFCC standards. 4. Statistical analysis Values were expressed as mean SD. Linear regression analysis was used for the correlation between parameters. Means were compared by using t-test. Significance levels were set at 0.05 in all cases. Assay bias was calculated as the test method result minus the Ultracentrifugation/Dextran sulfate-mg 2 assay result. Analyses were performed with Statistica Ver. 5.0 (StatSoft Inc. Tulsa, OK, USA). The sensitivity of an LDL-C assay at a specified cut-point was calculated as: [true positives/(true positives false negatives)] x 100, where true positives means the number of diseased patients correctly classified by the LDL-C test and false negatives mean the number of diseased patients misclassified by the test. The specificity of an LDL-C assay at a specified cut-point was calculated as: [true negatives/ (true negatives false positives)] x 100, where true negatives means the number of nondiseased patients correctly classified by the LDL-C test and false positives means the number of nondiseased patients misclassified by the test. The positive predictive value (pospv) and the negative predictive value (negpv) of an LDL-C assay at a specified cut-point were calculated as: [true positives/(true positives false positives)] x 100 and [true negatives/(true negatives false negatives)] x 100, respectively [33,26]. For the predictive value test the Ultracentrifugation/Dextran sulfate-mg 2 procedure used in the method comparison was considered as the reference method.

596 E. Bairaktari et al. / Clinical Biochemistry 34 (2001) 593 602 Table 1 Comparison of LDL-C data on hemodialysis patients by different methods 5. Results 5.1. Inter-method comparison LDL-C mean SD (mmol/l) Linear regression** Population: 98(92*) Method-1 2.92 1.11 Method-2 2.86 1.11 Method-3 2.84 1.03 Method-4 2.78 0.92 Method-5 2.60 0.90 LDL-UC 2.70 0.98 Bias, mean SD Method-1-LDL-UC 0.22 0.34 Method-2-LDL-UC 0.16 0.34 Method-3-LDL-UC 0.14 0.37 Method-4-LDL-UC 0.08 0.29 Method-5-LDL-UC 0.10 0.29 Method-1 / LDL-UC 1.08x 0.04 (0.95) Method-2 / LDL-UC 1.08x 0.06 (0.95) Method-3 / LDL-UC 0.99x 0.17 (0.93) Method-4 / LDL-UC 0.90x 0.34 (0.95) Method-5 / LDL-UC 0.88x 0.21 (0.95) LDL-1: Friedewald calculation with Direct HDL-C assay; LDL-2: Friedewald calculation with precipitation HDL-C assay; LDL-3: Direct LDL; LDL-4: N-geneous LDL; LDL-5: ApoB based equation; LDL-UC: Ultracentrifugation/Dextran sulfate-mg procedure. * when Friedewald calculation (LDL-1, LDL-2) is involved (TG 4.52, n 92) ** y bx a (r) A total of 98 serum samples of hemodialysis patients were analyzed for their LDL-C concentrations in parallel, using 5 methods: the Friedewald equation [either with HDL-C precipitation assay (Method 1) or with a direct assay (Method 2)], the Direct LDL assay (Method 3), the N-geneous assay (Method 4), the ApoB derived calculation (Method 5) and the Ultracentrifugation/Dextran sulfate-mg 2 procedure as the reference method (LDL- UC). The LDL-C concentrations for the Friedewald equation were calculated only for those samples with TG concentrations 4.52 mmol/l. In Table 1, the mean values, the biases and the regression lines of LDL-C for all methods used vs. the ultracentrifugation method are represented. Significant differences between the mean values for all methods used compared with the reference method were found (p 0.01 for all comparisons). The LDL-4 assay and the LDL-5 calculation yielded better results than the two Friedewald calculations (LDL-1 and LDL-2) and the LDL-3, since these two methods exhibited lower bias. The direct HDL-C assay has improved slightly the accuracy of the LDL-C calculation by the Friedewald formula. All five methods were correlated with the Reference method with a p value of 0.001 (Table 1 and Fig. 1). The comparison-of-methods plots [LDL-UC (x) vs. test method (y)] showed correlation coefficients 0.95 for the two Friedewald calculations, 0.93 for the Direct assay (LDL-3) and 0.95 for the N-geneous assay (LDL-4) and the ApoB derived calculation (LDL-5). The parameters of the regression lines for the five methods compared (LDL-1 - LDL-5) were slopes 1.08, 1.08, 0.99, 0.90 and 0.88 and intercepts 0.04, 0.06, 0.17, 0.34 and 0.21 mmol/l, respectively. To illustrate better the performance of the methods compared in hypertriglyceridemic patients, samples were divided on the basis of their TG levels into three groups according to serum TG concentrations 2.26, 2.26 4.52, and 4.52 mmol/l. The mean values, the biases and the regression lines for the three groups are shown in Table 2. In samples with TG 2.26 mmol/l, the correlation coefficients for all methods were high (0.97 0.98), and the biases were improved compared to those of the whole group (Table 1) for all methods, except for the ApoB derived calculation (LDL-5), where a slight increase was noticed. In the second group (2.26 TG 4.52 mmol/l), the regression lines worsened for all methods (r: 0.94 0.95) when compared to samples with TG values 2.26 mmol/l. The biases for hypertriglyceridemic patients compared to normotriglyceridemic values showed a significant increase for the Friedewald equations, as it has been previously shown [14]; as well as for the Direct (LDL-3) and the N-geneous (LDL-4) assays, whereas the ApoB calculation was less affected by the increased TG levels. In the third group (TG 4.52 mmol/l) the Friedewald calculations were excluded. The correlation in the LDL-3 fell to 0.92 and the intercept and the bias increased to 0.92 and 0.90 mmol/l, respectively. The LDL-4 and LDL-5 showed a very good correlation without significant alteration of the bias with the LDL-5 giving the better results. 5.2. Bias plots The concentration difference of each of the five methods from the LDL-UC (bias) was plotted as a function of either increased serum TG or LDL-C. A positive correlation between the bias and TG concentration was noticed for all assays, except for the ApoB-derived method, where no significant correlation between assay bias and serum TG was evident (Fig. 2). In the whole group of patients, increased concentrations of LDL-C resulted in a positive bias for the Friedewald equation, a negative bias for the N- geneous assay and ApoB calculation, whereas the negative bias of the Direct assay was not significant. However, when patients with LDL-C concentrations below 3.37 mmol/l (72% of the whole group) were examined separately, the influence of increased LDL-C concentrations was not significant for all five methods (Table 3).

E. Bairaktari et al. / Clinical Biochemistry 34 (2001) 593 602 597 Fig. 1. Least-squares linear regression analysis of LDL-C in hemodialysis patients obtained by comparison of the five methods vs. ultracentrifugation procedure (LDL-1: A, LDL-2: B, LDL-3: C, LDL-4: D and LDL-5: E). 5.3. Classification of patients into the risk categories The ability of the five methods compared to classify correctly subjects into the risk categories was evaluated in the particular population of hemodialysis patients (n 98), using the LDL-UC concentrations as the true values. The two cut-off values selected were 2.98 mmol/l proposed by the European Atherosclerosis Society and 3.37 Table 2 Comparison of LDL-C data by different methods at different TG levels TG in mmol/l 2.26 2.26 4.52 4.52 Sample size 54 38 6 LDL-C, mean SD (mmol/l) Method-1 2.74 1.02 3.16 1.19 Method-2 2.68 1.02 3.10 1.19 Method-3 2.60 0.99 3.04 1.00 3.70 1.07 Method-4 2.62 0.85 2.97 0.99 3.00 1.03 Method-5 2.49 0.85 2.70 0.96 2.94 1.02 Method-UC 2.63 1.00 2.77 0.95 2.80 0.99 Bias, mean SD Method-1-LDL-UC 0.11 0.22 0.39 0.41 Method-2-LDL-UC 0.05 0.22 0.33 0.41 Method-3-LDL-UC 0.04 0.25 0.27 0.30 0.90 0.42 Method-4-LDL-UC 0.01 0.25 0.20 0.30 0.19 0.27 Method-5-LDL-UC 0.14 0.27 0.07 0.32 0.13 0.16 Linear regression** Method-1 / LDL-UC 1.00x 0.12 (0.98) 1.20x 0.15 (0.95) Method-2 / LDL-UC 1.00x 0.06 (0.98) 1.20x 0.22 (0.95) Method-3 / LDL-UC 0.95x 0.10 (0.97) 1.01x 0.24 (0.95) 0.99x 0.92 (0.92) Method-4 / LDL-UC 0.83x 0.45 (0.98) 0.99x 0.21 (0.95) 1.00x 0.20 (0.97) Method-5 / LDL-UC 0.82x 0.33 (0.97) 0.95x 0.06 (0.94) 1.01x 0.10 (0.99) LDL-1: Friedewald calculation with Direct HDL-C assay; LDL-2: Friedewald calculation with precipitation HDL-C assay; LDL-3: Direct LDL; LDL-4: N-geneous LDL; LDL-5: ApoB based equation; LDL-UC: Ultracentrifugation/Dextran sulfate-mg procedure. ** y bx a (r)

598 E. Bairaktari et al. / Clinical Biochemistry 34 (2001) 593 602 Fig. 2. Bias (test method concentration-reference method concentration) plotted as a function of serum TG concentration of the five methods compared. mmol/l proposed by the American National Cholesterol Education Program-Adult Treatment Panel II (NCEP- ATP II) guidelines as the cut-off values in patients with high risk. In the Friedewald calculation, only patients with TG concentrations 4.52 mmol/l (n 92) were included. Overall, 28% of hemodialysis patients exceeded the diagnostic LDL-C level of 3.37 mmol/l. Both Friedewald calculations (LDL-1, LDL-2) correctly classified 82/92 patients; Direct assay (LDL-3) 86/98; N-geneous assay (LDL-4) 88/98; and ApoB derived calculation (LDL-5) 88/ 98. At the diagnostic level of 2.98 mmol/l, 40% of the patients showed pathologic levels. Friedewald calculations LDL-1 and LDL-2 correctly classified 82/92 and 81/92 patients, respectively; Direct assay (LDL-3) 87/98; N-geneous assay (LDL-4) 91/98; and ApoB derived calculation (LDL-5) 91/98. The sensitivity, specificity, positive predictive value (pospv) and negative predictive value (negpv) at the two diagnostic cutoff values (2.98 and 3.37 mmol/l) are shown in Table 4 and Fig. 3. The sensitivity of LDL-C estimated by either of the five methods decreased as LDL-C concentrations increased (Fig. 4a). In both cutoff points, the Friedewald equations showed the higher sensitivity (100% and 97.2% at 2.98 mmol/l; 88% for both at 3.37 mmol/l). The LDL-3 and LDL-4 presented lower sensitivity (92.3% at 2.98 mmol/l and 77.8% at 3.37 mmol/l for both) and LDL-5 the lowest one (84.6% at 2.98 mmol/l and 63% at 3.37 mmol/l). The specificity increased as LDL-C concentrations increased. At 3.37 mmol/l, it ranged between 89.6 97.2% for all methods with the ApoB derived equation showing the highest value. At 2.98 mmol/l, the specificity for the Table 3 Correlation of the bias as a function of serum LDL-C concentration. LDL-1 LDL-2 LDL-3 LDL-4 LDL-5 All samples n 92 92 98 98 98 LDL-UC r 0.236 0.239 0.029 0.320 0.385 p 0.024 0.022 NS 0.001 0.000 LDL-UC 3.37 mmol/l n 67 67 71 71 71 LDL-UC r 0.230 0.232 0.084 0.224 0.190 p NS NS NS NS NS LDL-UC: Ultracentrifugation/Dextran sulfate-mg procedure

E. Bairaktari et al. / Clinical Biochemistry 34 (2001) 593 602 599 Table 4 Classification of HD patients correctly in to the risk categories LDL-C cut-off 2.98 mmol/l test PPV NPV Sensitivity Specificity LDL1 78.3 100.0 100.0 82.1 LDL2 77.8 97.9 97.2 82.1 LDL3 81.8 94.4 92.3 86.4 LDL4 90.0 94.8 92.3 93.2 LDL5 97.1 90.6 84.6 98.3 LDL-C cut-off 3.37 mmol/l test PPV NPV Sensitivity Specificity LDL1 75.9 95.2 88.0 89.6 LDL2 75.9 95.2 88.0 89.6 LDL3 77.8 91.5 77.8 91.5 LDL4 84.0 91.8 77.8 94.4 LDL5 89.5 87.3 63.0 97.2 Friedewald method was 82.6%, the LDL-3 86.4%, the LDL-4 93.2% and the LDL-5 98.3%. The positive predictive value (pospv) was slightly higher at 2.98 mmol/l than at 3.37 mmol/l and increased in order from LDL-1 LDL-5 for both cut-off values (from 77.8% 97.1% at 2.98 mmol/l and from 75.9% 89.5% at 3.37 mmol/l). The negative predictive value (negpv) was above 87.3% for all methods and for both cutoffs. At 2.98 mmol/l, it was slightly higher than at 3.37 mmol/l and showed a slight decrease in order from LDL-1 LDL-5. 6. Discussion The association between increased concentrations of low-density lipoprotein cholesterol (LDL-C) and increased rate of premature coronary heart disease (CHD) has been clearly demonstrated. Currently, most clinical laboratories use the Friedewald equation to calculate the LDL-C levels, since the reference method, -quantification by ultracentrifugation, is not suitable for routine use. However, the use of the equation has been repeatedly questioned, particularly since it is based on the assumption that the majority of triglycerides reside in the VLDL fraction and that the relationship between triglycerides and cholesterol in this fraction is constant. This statement does not hold in certain conditions associated with hypertriglyceridemia, in type III hyperlipidemia, and certain secondary dyslipidemias, possibly including uremia. In fact, in renal failure, accumulation of partly metabolized triglyceride-rich particles (predominantly VLDL and IDL remnants) is observed, causing hypertriglyceridemia and low HDL-C concentrations. Furthermore, even although LDL-C levels are typically similar to those in the general population (or lower), this pattern often conceals a highly abnormal lipid subfraction profile with a predominance of atherogenic small dense LDL-particles. The purpose of this study was to evaluate the analytical and clinical performance of five methods for the estimation of LDL-C, with the ultracentrifugation/dextran sulfate-mg 2 precipitation as the reference method in 98 hemodialysis patients. As our results show, despite the good correlation be- Fig. 3. Bias (test method concentration-reference method concentration) plotted as a function of serum LDL-C concentration of the five methods compared.

600 E. Bairaktari et al. / Clinical Biochemistry 34 (2001) 593 602 Fig. 4. The sensitivity, specificity, positive predictive value (pospv) and negative predictive value (negpv) for the five LDL-C methods compared at the two diagnostic cut-off values for LDL-C (2.98 and 3.37 mmol/l). tween all five LDL-C methods examined and the ultracentrifugation/dextran sulfate-mg 2 precipitation, there were significant differences in the mean values and biases vs. the reference method. The Friedewald equation showed the higher bias, which was positive, and there was a considerable influence of hypertriglyceridemia on its accuracy. This influence was revealed when patients were divided in normo- and hyper-lipidmic groups, and when the bias from the reference method was plotted as a function of serum TGs. These results are in agreement with those described previously, where hyperlipidemia is considered to be one of the main causes of error in the estimation of LDL-C by the Friedewald equation, in which it is postulated that total triglycerides divided by 2.2 is a reliable estimate of VLDL-C [14,17]. However, this is not always true, as for example is seen in type III hyperlipoproteinemia, where the accumulation of remnant particles results in an increased ratio of cholesterol to triglycerides in the VLDL particles. Similarly, in hemodialysis patients, where the catabolism of the remnant lipoproteins is impaired, possibly due to reduced postheparin lipase activity, an increased ratio of cholesterol to triglycerides may be present. As a consequence, dividing by 2.2 will yield low estimates for VLDL-C and slightly higher calculated values for LDL-C than in the general population, as it has been showed previously in patients on hemodialysis and on continuous ambulatory peritoneal dialysis [14]. The Direct LDL-C assay showed a positive bias although lower than that seen in the Friedewald equation and a considerable influence of increased TGs levels on the accuracy of LDL-C measurement. Although this assay has been proposed for the direct measurement of LDL-C in the nonfasting state, a positive correlation with increased TGs was also previously observed in the general population [22,23] as well as in a group of patients with secondary dyslipidemia, namely the diabetic patients [24]. In another study concerning renal patients the direct assay was compared with the Friedewald equation but not with the ultracentrifugation procedure [25]. Although the Direct assay seems to yield slightly better results than the Friedewald calculation, it does not have appreciable advantages for the routine clinical laboratories because of the requirement of a pretreatment step which is not automated. The N-geneous assay presented the lower bias than all the other methods compared to the reference method, and it was moderately influenced by the increased concentrations of TG. In previous studies, the N-geneous assay and in general homogeneous assays for LDL-C, when examined in the general and pediatric population, were found to provide clinical laboratories with the means to measure LDL-C in

E. Bairaktari et al. / Clinical Biochemistry 34 (2001) 593 602 601 hypertriglyceridemic samples, even although they were slightly affected by increased TGs [27,28]. However, the assay was not tested previously in patients with secondary dyslipidemias including patients with impaired renal function. The ApoB derived calculation presented a low and negative bias and interestingly was the only one assay found to be independent of increased TG levels until 4.52 mmol/l, either when the bias was examined as a mean value, or when it was examined as a plot function. In the small group of patients (n 6) with TG 4.52 mmol/l, all biases were positive but ApoB derived calculation had presented the smaller bias. Similar results were also found in previous studies in the general population and in hyperlipidemic patients [31,34], while the equation is tested for the first time in patients with secondary dyslipidemias. The reason the ApoB derived calculation is independent from increased TG concentrations may be that the ApoB levels are representative of the number of apob containing particles, even with alteration of their lipid content, as it occurs in patients with abnormal abundance of small LDL particles. LDL IDL account for 90% of atherogenic apob containing particles even in hypertriglyceridemic patients and IDL are the minor components of these particles. The analytical differentiation between the assays examined also reflected on the classification of the patients into the selected risk categories. All assays classified correctly a high percentage ( 88%) of the patients in both cut-point values (2.98 and 3.37 mmol/l). However, the sensitivity, specificity, positive predictive value (pospv) and negative predictive value (negpv) at the two diagnostic cut-off points (2.98 and 3.37 mmol/l) were found to vary between these methods. As it is shown in Table 4 and Fig. 3, the sensitivity of LDL-C estimated by either of the five methods decreased as LDL-C concentrations increased. The Friedewald formula showed the highest values (100 and 88.0% in the two cut-off points, respectively), while ApoB the lowest ones (84.6 and 63.0%, respectively). The low sensitivity presented by the ApoB derived calculation, especially at the cut-off value of 3.37 mmol/l, is probably due to the negative bias of the assay in the high LDL-C concentrations, thus resulting in an increased number of false negative results. However, as the majority of the samples of the hemodialysis patients studied present low LDL-C (in 72% of the patients LDL-C 3.37 mmol/l), the negative predictive value of the assay was less affected by this negative bias. The specificity for all assays was greater than 82% in both cut-off points. The Friedewald calculation showed the lowest values (82.1 and 89.6% in the two cut-off points, respectively), while ApoB the highest ones (98.3 and 97.2%, respectively). The low specificity presented by the Friedewald calculation is probably due to the positive bias of the assay in high TG levels, thus resulting in an increased number of false positive results. Although the specificity of the Friedewald is not very low, it has a considerable influence on the positive predictive value since 45% of the patients presented TG levels 2.26 mmol/l. In conclusion, the new methodologies and particularly the N-geneous and the ApoB derived calculation, offer an improved approach to LDL-C analysis compared to the Friedewald equation in hemodialysis, enabling a correct classification of these patients. The ApoB derived calculation that was found to be independent of increased TG levels and had correctly classified the highest percentage of patients may represent a good alternative to the Friedewald calculation in patients with increased TG, but low to moderate LDL-C levels. References [1] Lazarus JM, Lowry EG, Hampers CL, Merrill P. Cardiovascular disease in uremic patients on hemodialysis. Kidney Int 1975;7(Suppl 2):167 75. [2] Linder AL, Charra B, Sherrard DJ, Scribner BN. Accelerated atherosclerosis in prolonged maintenance hemodialysis. N Engl J Med 1974;290:697 701. [3] Attman P, Samuelsson O, Alaupovic P. Lipoprotein metabolism and renal failure. Am J Kidney Dis 1993;21:573 92. [4] Chan MK, Varghese Z, Moorhead JF. Lipid abnormalities in uremia, dialysis, and transplantation. Kidney Int 1981;19:625 37. [5] Cattran DC, Steiner G, Wilson DR, Fenton SSA. Defective triglyceride removal in lipidemia associated with peritoneal dialysis and hemodialysis. Ann Intern Med 1976;85:29 33. [6] Chan MK, Persuad JW, Varghese Z, Moorhead JF. Pathogenic roles of postheparin lipases in lipid abnormalities in hemodialysis patients. Kidney Int 1984;25:812 16. [7] Expert panel on detection, evaluation, and treatment of high blood cholesterol in adults. Summary of the second report of the National Cholesterol Education Program (NCEP) expert panel on detection, evaluation, and treatment of high blood cholesterol in adults (Adult Treatment Panel II). JAMA 1993;269:3015 23. [8] International Task Force for Prevention of Coronary Heart Disease. Prevention of coronary heart disease: scientific background and new clinical guidelines. Recommendations for the European Atherosclerosis Society. Nutr Metab Cardiovasc Dis 1992;2:113 56. [9] Baigent C, Burbury K, Wheeler D. Premature cardiovascular disease in chronic renal failure. Lancet 2000;356[Review]:147 52. [10] Baigent C, Wheeler D. Should we reduce blood cholesterol to prevent cardiovascular disease among patients with chronic renal failure? Nephrol Dial Transplant 2000;15[Editorial Comments]:1118 9. [11] Bachorik PS, Ross JW, for the National Cholesterol Education Program Working Group on Lipoprotein Measurement. Recommendations for measurement of low-density lipoprotein cholesterol: Executive Summary. Clin Chem 1995;41:1414 20. [12] Bachorik PS. Measurement of low-density lipoprotein cholesterol, In: Rifai N, Warnick GR, Dominiczak MH editors: Handbook of lipoprotein testing, chap 8. Washington, U.S.A.: AACC press, 1997. p. 145 60. [13] Friedewald WT, Levy RI, Fredrickson DS. Estimation of the concentration of low-density lipoprotein cholesterol in plasma without use of a preparative ultracentrifuge. Clin Chem 1972;18:499 502. [14] Nauck M, Krämer-Guth A, Bartens W, März W, Wieland H, Wanner C. Is the determination of LDL cholesterol according to Friedewald accurate in CAPD and HD patients? Clin Nephrol 1996;46:319 25. [15] Johnson R, McNutt P, MacMahon S, Robson R. Use of the Friedewald formula to estimate LDL-cholesterol in patients with chronic

602 E. Bairaktari et al. / Clinical Biochemistry 34 (2001) 593 602 renal failure on dialysis. [Technical Brief]. Clin Chem 1997;43:2183 84. [16] Warnick GR, Knopp RH, Fitzpatrick V, Branson L. Estimating lowdensity lipoprotein cholesterol by the Friedewald equation is adequate for classifying patients on the basis of nationally recommended cutpoints. Clin Chem 1990;36:15 9. [17] McNamara JR, Cohn JS, Wilson PWF, Schaefer EJ. Calculated values for low-density lipoprotein cholesterol in the assessment of lipid abnormalities and coronary disease risk. Clin Chem 1990;36:36 42. [18] Nauck M, März W, Jarausch J, Cobbaert C, Sägers A, Bernard D, Delanghe J, Honauer G, Lehmann P, Oestrich E, Eckardstein A, Walch S, Wieland H, Assmann G. Multicenter evaluation of a homogeneous assay for HDL-cholesterol without sample pretreatment. Clin Chem 1997;43:1622 29. [19] Nauck M, Marz W, Wieland H. New immunoseparation-based homogeneous assay for HDL-cholesterol compared with three homogeneous and two heterogeneous methods for HDL-cholesterol. Clin Chem 1998;44:1443 51. [20] Bairaktari E, Elisaf M, Katsaraki A, Tsimihodimos V, Tselepis AD, Siamopoulos KC, Tsolas O. Homogeneous HDL-cholesterol assay versus ultracentrifugation/dextran sulfate-mg 2 precipitation, and dextran sulfate-mg 2 precipitation in healthy population, and in hemodialysis patients. Clin Biochem 1999;32:339 46. [21] Rifai N, Cole TG, Iannotti E, Law T, Macke M, Miller R, et al. Assessment of interlaboratory performance in external proficiency testing programs with a direct HDL-cholesterol assay. Clin Chem 1998;44:1452 58. [22] McNamara JR, Cole TG, Contois JH, Ferguson CA, Otvoras JM, Schaefer EJ. Immunoseparation method for measuring low-density lipoprotein cholesterol directly from serum evaluated. Clin Chem 1995;41:232 40. [23] Harris N, Neufeld EJ, Newburger JW, Ticho B, Baker A, Ginsburg GS, Rimm E, Rifai N. Analytical performance and clinical utility of a direct LDL-cholesterol assay in a hyperlipidemic pediatric population. Clin Chem 1996;42:1182 8. [24] Hirary S, Li D, Jialal I. A more valid measurement of low-density lipoprotein cholesterol in diabetic patients. Am J Med 1997;102:48 53. [25] Akanji AO. Direct method for the measurement of low-density lipoprotein cholesterol levels in patients with chronic renal disease: a comparative assessment. Nephron 1988;79:154 61. [26] Rifai N, Iannotti E, DeAngelis K, Law T. Analytical and clinical performance of a homogeneous enzymatic LDL-cholesterol assay compared with the ultracentrifugation-dextran sulfate-mg 2 method. Clin Chem 1998;44:1242 50. [27] Nauck M, Rifai N. Analytical performance and clinical efficacy of three routine procedures for LDL cholesterol measurement compared with the ultracentrifugation-dextran sulfate-mg 2 method. Clin Chim Acta 2000;294:77 92. [28] Yu HH, Markowitz R, De Ferranti SD, Neufeld EJ, Farrow G, Bernstein HH, Rifai N. Direct measurement of LDL-C in children: performance of two surfactant-based methods in a general pediatric population. Clin Biochem 2000;33:89 95. [29] Sugiuchi H, Irie T, Uji Y, Ueno T, Chaen T, Uekama K, Okabe H. Homogeneous assay for measuring low-density lipoprotein cholesterol in serum with triblock copolymer and alpha-cyclodextrin sulfate. Clin Chem 1998;44:522 31. [30] Nauck M, Graziani MS, Bruton D, Cobbaert C, Cole TG, Lefevre F, Riesen W, Bachorik PS, Rifai N. Analytical and clinical performance of a detergent-based homogeneous LDL-cholesterol assay: a multicenter evaluation. Clin Chem 2000;46:506 14. [31] Planetta T, Cortes M, Martinez-Bru C, Gonzalez-Sastre F, Ordonez- Llanos J. Calculation of LDL-cholesterol by using apolipoprotein B for classification of nonchylomicronemic dyslipemia. Clin Chem 1997;43:808 15. [32] Warnick GR, Benderson J, Albers JJ. Dextran sulfate-mg 2 precipitation procedure for quantitation of high-density-lipoprotein cholesterol. Clin Chem 1982;28:1379 88. [33] Galen RS, Peters T. Analytical goals, and clinical relevance of laboratory procedures, In: Tietz NW, editor. Fundamentals of clinical chemistry, 3 rd ed., chap 6. Washington, Philadelphia, U.S.A.: WB Saunders Company, 1987. p. 213 224. [34] Bairaktari E, Hadzidimou K, Tzallas C, Vini M, Katsaraki A, Tselepis A, Elisaf M, Tsolas O. Estimation of LDL Cholesterol based on Friedewald formula and on Apo B levels. Clin Biochem 2000;33: 549 55.