Assessment of interlaboratory performance in external proficiency testing programs with a direct HDL-cholesterol assay

¹ Department of Laboratory Medicine, Children's Hospital and Department of Pathology, Harvard Medical School, Boston, MA 02115.

² Department of Medicine, Washington University, St. Louis, MO 63110.

³ Department of Pathology and Laboratory Medicine, University of Wisconsin, Madison, WI 53792-2472.
^a Address correspondence to this author at: 600 Highland Avenue, Madison, WI 53792-2472. Fax 608-263-0910; e-mail da.wiebe{at}hosp.wisc.edu.

	Abstract

Top Abstract Introduction Materials and Methods Results and Discussion References

 
Direct assays for the determination of HDL-cholesterol (HDL-C) have recently become available. The methods are precise, require small sample volume, and appear to be less affected by increased triglycerides than traditional precipitation methods. In this study, we describe the inter- and intralaboratory variability of the Boehringer Mannheim Corporation direct HDL-C assay and its performance in external proficiency testing surveys. A comparison study among three laboratories, using different analyzers and 85 serum specimens, showed a correlation coefficient (r) of 0.99. The direct HDL-C assay also showed good agreement with the ultracentrifugation-dextran sulfate-Mg²⁺ method (r = 0.98) and the Cholesterol Reference Method Laboratory Network-Designated Comparison Method (a = 0.98x + 4.75 mg/L, r = 0.98). Total error at medical decision levels ranged from -0.8% to +11.1%. Furthermore, this assay performed adequately in the College of American Pathologists and the ALERT^® surveys as well as the CDC Lipid Standardization Program and met all performance criteria of regulatory agencies.

	Introduction

Top Abstract Introduction Materials and Methods Results and Discussion References

 
The measurement of HDL-cholesterol (HDL-C)¹ currently is recommended by the National Cholesterol Education Program (NCEP) for all adults at the initial screening stage for hypercholesterolemia (1). HDL-C values <350 mg/L are considered an independent risk factor for coronary heart disease, whereas values 600 mg/L are deemed protective. Historically, HDL-C has been assayed in clinical laboratories by a variety of selective precipitation methods (2). Apolipoprotein B-containing lipoproteins (primarily VLDL and LDL) are removed using precipitation reagents such as phosphotungstate-Mg², dextran sulfate-Mg², heparin-Mn², or polyethylene glycol, followed by centrifugation, and the cholesterol component of the supernatant, which represents HDL-C, is measured enzymatically. These assays are labor-intensive, relatively imprecise (CVs of 3–6%), affected to various degrees by the presence of increased triglycerides, and require large sample volume (100–500 µL). Recently, homogeneous assays for the determination of HDL-C have been introduced (3–7). These methods are performed on-line with improved precision (CVs of 1–2%), appear to be less affected by increased triglycerides, and require only a few microliters of sample.

In the Boehringer Mannheim Corporation (BMC) direct HDL-C assay, which represents this new generation of methods, soluble complexes of non-HDL lipoproteins and -cyclodextrin-Mg² are formed. The cholesterol associated with HDL is then quantitated with polyethylene glycol-modified cholesterol oxidase and esterase, which possess reduced reactivity with the complexed lipoproteins. The analytical performance of this assay has been the subject of several recent reports (3)(6)(7). However, the interlaboratory variation of this assay and its performance in external proficiency testing surveys such as the College of American Pathology (CAP), ALERT^®, and CDC Lipid Standardization Program (LSP) have not been examined. In this study, we describe a multicenter evaluation of this direct HDL-C assay, using fresh patient sera from fasting individuals and three different models of Hitachi analyzers (911, 917, and 747). Furthermore, the ability of the assay to meet current NCEP performance goals, proficiency testing requirements, and LSP criteria is examined (8).

	Materials and Methods

Top Abstract Introduction Materials and Methods Results and Discussion References

 
analyzers
All measurements were performed on three models of BMC/Hitachi analyzers, a 911 at Children's Hospital in Boston, MA, (laboratory 1), a 917 at Washington University in St. Louis, MO, (laboratory 2), and a 747 at the University of Wisconsin in Madison, WI (laboratory 3). Each of these analyzers is the principal analytical system used to perform routine patient testing at these institutions. The instruments were operated and calibrated using the standard protocols recommended by the manufacturer.

reagents
Direct HDL-C reagent and enzymatic cholesterol and triglyceride reagents were provided by BMC. Dextran sulfate (M_r 50 000) was purchased from Genzyme Corporation. The triglyceride assay was corrected for the presence of endogenous glycerol. The imprecision of these assays, which is reflected by CV, was <1.5% for total cholesterol, and the CV of the triglyceride assay was within 2.0% in the three laboratories. All other reagents used were of the highest quality available from local suppliers.

Calibration (Precical®) and quality-control materials (Precinorm® and Precipath®) were provided by BMC to monitor the performance of the analytical systems. These serum-based quality-control materials were prepared and assayed in every analytical run.

samples
Interlaboratory patient sample comparison study.
Blood specimens collected from fasting individuals and processed for routine lipid analyses were used for the comparison study. On the day of analysis, two 0.5-mL serum aliquots were prepared from 85 patient specimens [total cholesterol (mean ± SD), 2178 ± 427 mg/L; triglycerides, 1559 ± 794 mg/L] after laboratory 1 completed their routine work. Aliquots were sent by overnight courier on wet ice to laboratories 2 and 3 for analysis of HDL-C, using the direct assay. Thus, all the fresh specimens were processed within 1 day, without preselection for a desired HDL-C concentration range or other factors. In addition to the measurement of HDL-C by the direct method, laboratory 1 assayed HDL-C using a combination of ultracentrifugation at 1.006 kg/L, dextran sulfate precipitation of the bottom fraction, and subsequent enzymatic cholesterol analysis of the supernate (the modified ß-quantification procedure (UC-DS) (9) . In another experiment, laboratory 2 performed a comparison study of the direct HDL-C assay with the Cholesterol Reference Method Laboratory Network-Designated Comparison Method (DCM) (G. Russell Warnick, Pacific Biometrics, Inc., Seattle, WA, personal communication) (10) for HDL-C, using 82 fresh patient specimens. The DCM method for HDL-C was modified from the procedure of Warnick et al. (11)(12). In brief, 1 part of precipitating reagent [dextran sulfate (M_r 50 000, 100 mg/L) and magnesium chloride (0.35 mol/L)] was mixed with 10 parts of serum and incubated at ambient temperature for 10–30 min. Precipitates of apolipoprotein B-containing lipoproteins were pelleted by centrifugation at 4 °C for 30 min at 1500g, and the resulting supernates were inspected for clarity. Cholesterol concentrations were measured in the clear supernates by the Abell-Kendall reference method (13).

External proficiency testing programs.
Each of the three laboratories participates in multiple external proficiency surveys as part of their standard practice to ensure that the performance meets analytical goals. Two of these programs, CAP and ALERT, are used by all three laboratories. In addition, both laboratory 1 and laboratory 2 participate in the CDC-LSP (14). The direct HDL-C method was evaluated by at least one challenge from each of these programs.

interference studies
Standard addition techniques were used to prepare serum specimens in each laboratory with increasing amounts of bilirubin (conjugated and unconjugated), ascorbic acid, hemoglobin, or triglycerides, while maintaining constant concentrations of HDL-C. For each experiment, a serum pool was divided in half, and a concentrated solution of the potential interfering substance was added to one half (high pool). An equal volume of the appropriate diluent was added to the other half (low pool). Appropriate volumes of each pool were mixed to give 11 specimens of constant HDL-C concentration with concentrations of interfering substances ranging from 0% to 100% of the high pool.

Unconjugated bilirubin (stock: 20 mg in 1 mL of NaOH, 0.1 mol/L) was obtained from Pfanstiehl. Conjugated bilirubin, as ditaurobilirubin, (stock: 9.0 mg in 10 mL of serum) was obtained from Lee Scientific, Inc. Ascorbic acid (stock: 100 mg in 100 mL of 9 g/L NaCl) was obtained from Sigma Chemical Co. Concentrated hemoglobin was prepared by washing packed red cells six times with 9 g/L NaCl, followed by freezing to rupture the cells and centrifugation to remove cellular debris. Triglyceride-rich lipoprotein concentrate for the lipemia experiment was prepared by ultracentrifugation of sera at 100 000g for 18 h.

data analysis
The means, medians, and SDs were calculated with Microsoft Excel, Ver. 5.0 (Microsoft). Student's paired t-test and least-squares linear regression analysis were performed using the SigmaPlot statistic program (Jandel Scientific). The Student's t-test was considered significant at P <0.05. Biases were calculated as the test procedure result, in this case the direct HDL-C assay result, minus the indicated comparative method.

Total error was calculated as the summation of the systematic and random error. Systematic error, with its two components, constant and proportional error, was derived from the linear regression equation y = bx a, where b was the slope of the linear regression and represented the proportional error, and a was the y-axis intercept and represented the constant error. Systematic error at a specified HDL-C concentration (x_c) was defined as the absolute value of y_c - x_c, where y_c = bx_c a. Random error was calculated as the day-to-day precision of the assay multiplied by the factor, 1.96.

	Results and Discussion

Top Abstract Introduction Materials and Methods Results and Discussion References

 
Historically, the HDL-C concentration was determined as a means to estimate the LDL-cholesterol (LDL-C) with the ß-quantification method and Friedewald calculations to confirm the diagnosis of hypercholesterolemia. However, the NCEP-Adult Treatment Panel II guidelines stipulate that HDL-C should be measured along with total cholesterol in all adults at the initial screening stage. This recommendation presents an interesting challenge to the clinical laboratory because it substantially increases the workload of the labor-intensive HDL-C testing. The introduction of the on-line direct HDL-C assays alleviates the pretreatment processing and provides several analytical advantages over existing methodologies. The direct HDL-C assays have been shown to be apparently less affected by increased triglycerides, have better precision, and require smaller sample volumes than chemical precipitation methods (3)(4)(5)(6)(7). However, the variability of HDL-C measurement by these assays among laboratories and the performance of these methods in proficiency testing programs have not yet been evaluated. In this study, we describe the intra- and interlaboratory variability in HDL-C measurement by the BMC direct assay as well as the ability of this method to meet current NCEP performance goals and proficiency testing programs criteria.

To determine the performance comparability of this homogeneous assay among different laboratories, HDL-C concentrations of 85 freshly collected serum specimens were simultaneously determined using three models of analyzers in three different laboratories. The HDL-C mean reported by laboratory 1 [direct HDL-C (mean ± SD)] was 509.5 ± 162.2 mg/L, by laboratory 2 was 496.6 ± 162.1 mg/L, and by laboratory 3 was 487.5 ± 152.9 mg/L. The findings of this study revealed good comparability and excellent correlation among the three laboratories (laboratory 1 vs laboratory 2, slope = 1.00, intercept = 14.4 mg/L, r = 0.99; laboratory 2 vs laboratory 3, slope = 0.94, intercept = 19.4 mg/L, r = 0.99; laboratory 1 vs laboratory 3, slope = 0.94, intercept = 9.2 mg/L, r = 0.99) (Fig. 1 ). Although statistically significant differences were observed between HDL-C values obtained by the direct assay in laboratories 2 and 3 and the UC-DS method (505.1 ± 174.0 mg/L), these differences were not judged to have clinical significance.

View larger version (20K): [in this window] [in a new window]   Figure 1. Assessment of accuracy of the direct HDL-C method vs the UC-DS method. Serum was collected from 85 fasting patients at laboratory 1. Aliquots of serum were analyzed for HDL-C by the direct method in each of the three laboratories. In addition, laboratory 1 subjected the specimens to UC-DS and measured HDL-C in the bottom fraction by the dextran sulfate method. () laboratory 1; () laboratory 2; (+) laboratory 3.

The Laboratory Standardization Panel of the NCEP has recently issued performance goal criteria for HDL-C. An interim goal until 1998 stipulates that HDL-C concentrations should be determined with <10% bias and <6% imprecision for HDL-C values 420 mg/L, or SD ± 25 mg/L for HDL-C values <420 mg/L, (total allowable error of 22%). A more stringent criteria was established for 1998. HDL-C concentrations should be measured with a bias of <5% and an imprecision of <4% CV for HDL-C values 420 mg/L, or SD ± 17 mg/L for HDL-C values <420 mg/L, (total allowable error of 13%). In this study, the day-to-day reproducibility of the assay in three different laboratories at two concentrations (335 and 473 mg/L) ranged between 1.87% and 4.37% (Table 1 ). The lowest imprecision at both concentrations was seen using the Hitachi 747 analyzer, whereas the highest was seen using the Hitachi 911 system. For the low HDL-C concentration control, the SD for laboratories 1, 2, and 3 were 14.8, 12.9, and 7.8 mg/L, respectively. These data indicate that the direct HDL-C assay currently can meet the rigid requirements for precision established for 1998, for both the HDL-C cutoff of 420 mg/L and <420 mg/L on the three analyzers. The bias for the direct HDL-C assay was determined by the linear regression analysis comparison with two systems, the UC-DS assay, and the DCM. When determined by the former, the absolute bias calculated at the two HDL-C concentrations of 335 and 473 mg/L ranged from -15 to 12 mg/L (-3.2% to 3.7%). At HDL-C concentrations of 335 and 473 mg/L, the absolute bias determined by laboratory 2 using DCM was 5 and 3 mg/L (1.4% and 0.7%), respectively (Fig. 2 ). On the basis of these findings, the total allowable error ranged from 0.5% to 11.3% (Table 1 ). Therefore, this direct HDL-C assay meets the analytical performance requirements set forth by the NCEP for 1998 and beyond, using three analyzers.

View larger version (18K): [in this window] [in a new window]   Figure 2. Assessment of accuracy of the direct HDL-C method vs DCM. Serum was collected from 82 fasting subjects over a period of 17 weeks. Within 24 h, HDL-C was measured in each specimen by both the direct method and the DCM. All specimens had triglyceride concentrations <600 mg/L.

The reducing agents bilirubin and ascorbic acid are known to interfere with the peroxidase-dependent cholesterol measurement. In addition, hemoglobin and triglycerides are common interferents with laboratory testing. To determine the effects of these interferents on the determination of HDL-C by the homogeneous assay, known amounts of bilirubin, ascorbic acid, hemoglobin, or triglycerides (VLDL and chylomicrons) were added to pooled sera. The addition of up to 178 mg/L bilirubin, 150 mg/L ascorbate, or 1030 mg/L hemoglobin did not significantly affect the determination of HDL-C in any of the three laboratories (data not shown). This finding is consistent with a previously publish report (3). However, the effect of the triglycerides on the determination HDL-C by the three laboratories was not consistent (Fig. 3 ). A positive bias of ~8%, 5%, and 3% was seen in laboratory 1, laboratory 2, and laboratory 3, respectively, as triglyceride concentrations approached 10 000 mg/L. Although the exact cause of this triglyceride-related bias is unknown, it may be either analyzer-related or dependent on the nature and/or composition of the triglyceride-rich lipoprotein used in the interference studies. Laboratory 1 used nonfasting patient sera to prepare their triglyceride concentrate, which was rich in both chylomicrons and VLDL, whereas laboratories 2 and 3 used fasting sera, which consisted of predominately VLDL. In addition, laboratory 2 measured the apparent HDL-C concentration using the direct method in the original sample and ultracentrifugal bottom fractions of sera from subjects with various triglyceride concentrations. A positive bias was observed in whole sera at triglycerides above 5000 mg/L, which increased with increasing triglyceride concentrations (data not shown), in the same pattern that laboratory 1 observed in the initial lipemia study (see Fig. 3 ).

View larger version (15K): [in this window] [in a new window]   Figure 3. Effect of lipemia on the direct HDL-C method. A concentrated mixture of triglyceride-rich lipoproteins was prepared in each laboratory by ultracentrifugation of nonfasting (laboratory 1) or fasting (laboratories 2 and 3) sera. Aliquots of the lipoproteins were added to serum with a triglyceride concentration <2000 mg/L, and the HDL-C was measured by the direct method.

The accuracy of a routine laboratory test is usually monitored and determined via participation in proficiency testing programs. The CAP Survey, the most prominent proficiency testing program in North America, uses lyophilized human serum pools to assess the performance of lipid testing. In addition, laboratories that are heavily involved in lipid research may also participate in the more rigid LSP, which is administered by the CDC and which uses frozen serum pools. Clinical laboratories that are not involved in lipid research but wish to have an additional means to assess the accuracy of their lipid testing may participate in the ALERT survey program, which uses fresh serum pools. The performance of the homogeneous HDL-C assay in the above-mentioned proficiency testing programs was evaluated by the three participating laboratories.

All three laboratories performed very well in four of the five pools provided as part of the CAP Survey (Table 2 ). The reason for the discrepancy between the HDL-C values obtained for LP-07 by the direct HDL-C assay and the confirmatory test is not clear at present. Perhaps the lyophilization process and the nature of the lipoproteins in this particular serum pool might be responsible for this discrepancy. The direct assay is sensitive to sample matrix and is very specific to human lipoproteins. For example, great differences were seen when rodent HDL-C concentrations were determined by this assay and by preparative ultracentrifugation (data not shown). Therefore, non- human-based sera may not be used in proficiency testing materials, and the use of this assay in veterinary medicine must be carefully evaluated in the species of interest. The direct HDL-C assay performed extremely well in both the CDC-LSP and the ALERT survey (Table 2 ). Our data indicate that this assay may be used by clinical laboratories participating in proficiency testing programs that use fresh, frozen, or lyophilized materials. Furthermore, this assay allows laboratories to meet the criteria established by federal regulatory agencies and the CDC-LSP.

As indicated earlier, the ß-quantification method for LDL-C, involves ultracentrifugation combined with a chemical precipitation step. Historically, with the reference method, the latter step was accomplished using heparin-Mn² precipitation reagents. However, most lipid reference laboratories in the US currently use dextran sulfate-Mg² reagent to determine HDL-C concentration in the triglyceride-rich lipoprotein-free fraction. Because both clinical and lipid reference laboratories are moving toward changing their HDL-C methodologies from precipitation to direct methods, we examined the impact of measuring HDL-C by the direct assay on the determination of LDL-C concentrations by the UC-DS method.

After the removal of the triglyceride-rich lipoprotein fraction, the HDL-C concentration was determined simultaneously by the dextran sulfate-Mg² precipitation method and the direct assay in 181 samples (total cholesterol, 2140 ± 420 mg/L; triglycerides, 1878 ± 1586 mg/L). No statistically significant difference was seen in HDL-C concentrations determined by the two methods (dextran sulfate-Mg², 481 ± 172 mg/L; direct, 494 ± 177 mg/L). In addition, no significant difference was seen in the derived LDL-C concentrations, using either the precipitation (1280 ± 383 mg/L) or the homogeneous (1289 ± 376 mg/L) method for HDL-C estimation (y = 0.99x 18.8 mg/L; r = 1.00; Fig. 4 ). Our data indicate that the direct HDL-C assay is an adequate alternative to precipitation methods in the determination of LDL-C concentrations by this approach.

View larger version (18K): [in this window] [in a new window]   Figure 4. Determination of LDL-C, using the direct HDL-C method. LDL-C was measured in 181 fresh patient specimens in laboratory 1 by ß-quantification, using both the dextran sulfate-Mg2+ and direct methods for the determination of HDL-C in the bottom fraction. There was no statistically significant difference between the two methods by the paired Student's t-test.

In conclusion, the BMC direct HDL-C assay demonstrated acceptable intralaboratory variability and fulfilled the NCEP performance goal criteria established for 1998 on different Hitachi systems. Furthermore, the assay was shown to perform adequately in proficiency testing programs and to meet the requirements of federal regulatory agencies and the CDC-LSP.

	Acknowledgments

 
We thank the staff at BMC for their support and donation of reagents and supplies for this study. Specifically, we thank Deborah Bruton for kind assistance.

	Footnotes

 
Presented in part at the 49th AACC Annual Meeting, Atlanta, GA, July 20–24, 1997.

¹ Nonstandard abbreviations: HDL-C, HDL-cholesterol; NCEP, National Cholesterol Education Program; BMC, Boehringer Mannheim Corporation; CAP, College of American Pathologists; LSP, Lipid Standardization Program; UC-DS, ultracentrifugation-dextran sulfate; DCM, Designated Comparison Method; and LDL-C, LDL-cholesterol.

	References

Top Abstract Introduction Materials and Methods Results and Discussion References

 

1. . The Expert Panel. 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;269:3015-3023. [ISI][Medline] [Order article via Infotrieve]
2. Wiebe DA, Warnick GR. Measurement of high-density lipoprotein cholesterol. Rifai N Warnick GR Dominiczak MH eds. Handbook of lipoprotein testing 1997:127-144 AACC Press Washington. .
3. Nauck M, Marz W, Jarausch J, Cobbaert C, Sagers A, Bernard D, et al. Multicenter evaluation of a homogeneous assay for HDL-cholesterol without sample pretreatment. Clin Chem 1997;43:1622-1629. [Abstract/Free Full Text]
4. Sugiuchi H, Yoshinori U, Okabe H, Irie T, Uekama K, Kayahara N, Miyauchi K. Direct measurement of high-density lipoprotein cholesterol in serum with polyethylene glycol-modified enzymes and -cyclodextrin. Clin Chem 1995;41:717-723. [Abstract/Free Full Text]
5. Harris N, Galpachian V, Thomas J, Iannotti E, Law T, Rifai N. Three generations of high-density lipoprotein cholesterol assays compared with the ultracentrifugation/dextran sulfate-Mg²⁺ method. Clin Chem 1997;43:816-823. [Abstract/Free Full Text]
6. Huang YC, Kao JT, Tsai KS. Evaluation of two homogeneous methods for measuring high-density lipoprotein cholesterol evaluated. Clin Chem 1997;43:1048-1055. [Abstract/Free Full Text]
7. Harris N, Galpachian V, Rifai N. Three routine methods for measuring high-density lipoprotein cholesterol compared with the reference method. Clin Chem 1996;42:738-743. [Abstract/Free Full Text]
8. . for the National Cholesterol Education Program Working Group on Lipoprotein MeasurementWarnick GR, Wood PD. National Cholesterol Education Program recommendations for measurement of high-density lipoprotein cholesterol: executive summary. Clin Chem 1995;41:1427-1433. [Free Full Text]
9. Harris N, Neufeld EJ, Newburger JW, Ticho B, Baker A, Ginsburg G, et al. Analytical performance and clinical utility of a direct LDL-cholesterol assay in a hyperlipidemic pediatric population. Clin Chem 1996;42:1182-1188. [Abstract/Free Full Text]
10. Kimberly MM, Waymack PP, Smith SJ. Evaluation of frozen vs fresh serum samples using the Designated Comparison Method for HDL cholesterol in the Cholesterol Reference Method Laboratory Network [Abstract]. Clin Chem 1995;41:S136.
11. Warnick GR, Benderson J, Albers JJ. Dextran sulfate-Mg²⁺ precipitation procedure for quantitation of high density lipoprotein cholesterol. Cooper GR eds. Selected methods of clinical chemistry 1983;Vol 10:91-99 American Association for Clinical Chemistry Washington. .
12. Warnick GR, Benderson J, Albers JJ. Dextran sulfate-Mg²⁺ precipitation for quantitation of high-density lipoprotein cholesterol. Clin Chem 1982;28:1379-1388. [Free Full Text]
13. Abell LL, Levy BB, Kendall FE. Simplified methods for the estimation of total cholesterol in serum and demonstration of its specificity. J Biol Chem 1951;195:357-366.
14. Myers GL, Cooper GR, Henderson O, Hassemer DJ, Kimberly MM. Standardization of lipid and lipoprotein measurements. Rifai N Warnick GR Dominiczak MH eds. Handbook of lipoprotein testing 1997:223-250 AACC Press Washington. .