Analytical performance and clinical utility of a direct LDL-cholesterol assay in a hyperlipidemic pediatric population

Transcription

1 ClinicalC/,emisliy 42: (1996) Analytical performance and clinical utility of a direct LDL-cholesterol assay in a hyperlipidemic pediatric population NEIL HARRIS,1 2 ELLIS J NEUFELD,3 JANE W NEWBIJRGER,4 BARUCH TIcHo,4 ANNETTE BAKER,4 GEOFFREY S GINSBURG,4 5 ERIC RIMM,6 and NADER RIFAII,2* This study compares a new latex immunoseparation method for the direct determination of plasma low-density lipoprotein cholesterol () with the reference procedure for (fl-quantification) in a pediatric hyperlipidemic population The direct assay has a mean bias of -98 mg/l in a fasting group (n = 96) of patients (mean triglycerides 157 ± 72 mg/l) and a bias of +177 mg/l in a nonfasting group (n = 42, mean triglycerides 4854 ± 5457 mg/l) The mean total analytical error calculated from our data is 138% The direct assay and the commonly used Friedewald calculation respectively classified 81% and 84% of fasting patients correctly, according to the cutoffs of 11 and 13 mgil for set by the National Cholesterol Education Program for pediatric patients Of combined fasting and nonfasting patients, 8% were correctly classffied by the direct assay Therefore, despite several analytical shortcomings, the direct assay may be useful in managing hyperlipidenuc children without the need for a fasting specimen INDEXING TERMS: methods comparison #{149} heart disease lipoproteins triglycerides risk factor hypercholesterolemia Atherosclerosis is initiated early in life [1] and progressessilently for decades untilit manifests in adulthood, often as coronary heart diseasealthough screening for lipoproteindisordersin the general pediatricpopulation is not advised,the National Cholesterol Education Program (NCEP) pediatricpanel and Departments of Laboratory Medicine, 2 Pathology, Medicine, and Cardiology, Children s Hospital, and Beth Israel Hospital, HarvardMedicalSchool, Boston, MA 2115 Department of Nutrition, Harvard School of Public Health, Boston, MA 2115 Address correspondence to this author, at: Department of Laboratory Medicine, Farley 7, Children s Hospital, 3 Longwood Ave, Boston, MA 2115 Fax ; rifai@altchharvardedu Received October 25, 1995; accepted February 27, 1996 the American Academy of Pediatrics Committee on Nutrition recommend selectivescreening of individualsbetween ages 2 and 2 years[2] Currently, only those childrenwhose parents or grandparents developed premature atheroscleroticvascular disease (55 years or younger) or children who have one parent whose total cholesterol is 24 mg/l are recommended for screening [218 The major atherogenic fraction of serum cholesterol is low-density lipoprotein cholesterol () [3, 4] The NCEP has classified concentrationsin children and adolescents from families with hypercholesterolemia or premature atherosclerosis into three main categories [2], based on the 75th (11 mgfl) and 9th percentile(13 mgfl) of in American children and adolescents: acceptable <11 mgfl, borderline mg/l, and high 13 mgfl Plasma is currently estimated in most clinical laboratories by the formula of Friedewald et al [5, 6], which is based on the observationthat,in the fasting state, very-low-density lipoprotein (VLDL) cholesterol is -2% of the total triglycerides(tg), measured in units of mg/l Therefore, = total cholesterol minus high-density lipoprotein cholesterol (HDL-C) minus (TG/5) This formula, however, has certain constraints [6-8]: It is invalidat TG concentrations 4 mg/l and it can be utilized only in the fasting state [6] Furthermore, the latestncep recommendations [9] define specificperformance criteriafor assaysthatcannot be met by the Friedewald calculation[6-8] A new direct immunological procedure for detennining (directldl cholesterolor D), LDL Cholesterol#{174}, has been developed by Genzyme Corp, Cambridge, MA, and ismarketed by Sigma Diagnostics, StLouis,MO The assay, which has been assessed in adults [1-12], relies on the immunoseparation of LDL particles from HDL and VLDL by Nonstandard abbreviations: NCEP, National Cholesterol Education Program;, low-density lipoprotein cholesterol; HDL-C, high-density lipoprotein cholesterol; TG, triglycerides; ape, apolipoprotein; and D, direct 8 To convert cholesterol mg/l to mmol/l, divide by 387 9To convert triglyceride mgil to mmovl, divide by

2 ClinicalChemistry 42, No 8, solid-phase-boundgoat polyclonalantibodiesto apolipoproteins (apos) A-I and E [12] Here, we report our assessment of the D assay in a hyperlipidemic pediatric population (ages 2 years) We compared both the D assay and the Ref- Friedewald estimation with a modified #{176} 13-quantification erence Method [6, 13, 14] for determining Historically, the 13-quantification assay included a heparin and manganese precipitation step, which has largely been replaced by precipitation with dextran sulfate and magnesium; hence, we describe the 13-quantification assay used as a modified procedure Mate,ialsand Methods PLASMA SAMPLES Of the 138 blood samples used in this study, 17 were from patientsattendingthe Children shospital Lipid Clinicbecause of a positivefamily history of premature atherosclerosisor hypercholesterolemia; 96 of these blood samples were obtained afteran overnight fastand the remaining 11 were drawn from nonfasting subjects This study protocol was approved by the Internal Review Board of Children s Hospital, Boston In addition to these Lipid Clinic samples, 31 plasma samples were obtained randomly (without conscious bias)from nonfasting hypertriglyceridemic hospital patients (TG 25 mg/l) We therefore studieda total of 42 nonfasting patients In allcaseswe assayed heparinized plasma that had been stored at 4 #{176}C [15] PROCEDURES Modified (3-quantification [16] We centrifuged23-jil plasma samples at 25 g for 3 h at 4 #{176}C in a Beckman TLIOO rotor with 7 X 2 mm polyallomertubes (Beckman Instruments, Palo Alto, CA) We then sliced the tubes and theircontentsso as to isolate the >16 kg/l infranate Having restored the volume of the infranateto thatof the original sample by adding isotonic saline (9 g/l NaCI), we quantified its HDL-C content after precipitating[17] the LDL particleswith MgCI2 and dextran sulfate (Mr 5 )The was then calculatedas infranatant totalcholesterol- HDL-C, the V having been removed by the ultracentrifugation The proficiency of our (3-quantification technique is checked three times a year as part of the Alert#{174} LDL cholesterolstandardizationprogram (Pacific Biometrics Research Foundation, Seattle, WA) The mean percentage biasfor concentrationsof mg/l in threeconsecutiveproficiencytests,fivesamples per test,was 138% (95% confidence interval, 8-268%) Totalcholesterol, HDL-C, and TG assaysconcentrations of total cholesterol,hdl-c, and TG were determined enzymatically with a Hitachi 911 automated analyzer (Boehringer Mannheim, Indianapolis, IN) TG measurement was corrected for the presence of endogenous glycerol HDL was separated as describedabove and its cholesterolcomponent was measured with the 911 analyzer Our laboratory is currently enrolled in the National Heart, Lung, and Blood Institute and Centers for Disease Control and Prevention Lipid Standardization Program D Sigma Diagnostics donated all the reagents and we performed the D assay at room temperature according to the manufacturer sinstructions Friedewald calculation We measured totalcholesterol,hdl-c, and TG in all plasma samples but applied the Friedewald calculation only to fasting samples with TG <4 mg/l Statistical analysis We calculated means, medians, SDs, and t-testvalueswith Microsoft Excel (Microsoft, Redmond, WA); t-tests were judged significant at P <5 Linear regression analysis by the least-squares method was calculated by the SigmaPlot#{174}statistics program (Jandel Scientific, San Rafael, CA) Biases were calculated as the test method result minus the Reference Method result We calculated total error as the sum of the systematic error plus the random error Systematic error has two components, constant systematic error and proportional systematic error, which are calculatedfrom the linearregressionequationy = nix + b, m being the slopeof the regression equation (proportional error) and b the y-axis intercept(constanterror)systematic error at an concentration of x is defined as the absolute valueof jy-x,7,wherey = mx + b Random erroris196sd of the run-to-run precision study [18, 19] The positive predictive value of an assay at the x mg/l cutpoint was calculatedas [true positive/(true positive + false positive)] X 1, where true positivemeans that results of both the Reference Method and the test method are >x mg/l, and falsepositivemeans thatthe testmethod result is >x mg/l when the Reference Method is <x mgfl Results Heparinized plasma is preferred to serum in our laboratory because it can be processed immediately aftercollectionin preliminary experiments we determined that the choice of heparinized plasma or serum had no statistically significant effect(p 37) on the outcome of the modified 13-quantificationor the D assay(table 1)We thereforeused plasma for the remaining experimentsfreezing alsohad no effecton the modified 13-quantification of : Eight plasma samples (mean ± SD 1186 ± 441 mg/l, range mg/l) frozenat -8 #{176}C showed no significantdifference(p = 53) in concentration when thawed (1178 ± 455 mg/l) Table 1 Measurement of concentrations in paired serum and plasma samples (n = 14) by 13-quantification and D assay lj-quant, mg/l D Plasma Serum Plasma Serum Mean SD Mm Max

3 1184 Harris et al: Direct LDL-cholesterol pediatric assay For the D assay, however, frozen plasma samples are unsuitablewhen 29 frozensamples (mean ± SD 1913 ± 633 mgfl) were assayed by both the D assay and the Reference Method, the D assay had a bias (relative to the Reference Method) of +4 to -13 mg/l [mean, -491 mg/l (-25%)] Nevertheless, the D concentrationisstableat 4 #{176}C for at least 3 weeks: Five samples with 13-quantification concentrations of mgfl (mean ± SD 139 ± 381 mg/l) were assayed with the D reagent once a week for 3 weeks, the samples being kept at4 #{176}C between measurements We found no significant difference between the initial (3-quantification concentration and the initial or subsequent D measurements (P 47) All 96 plasma samples from fastingpediatricpatientsof the Lipid Clinic were analyzed with the D assay; analyzed for total cholesterol, HDL-C, and TG concentrations to calculate the Friedewald values; and analyzed for concentrations with the modified 13-quantification procedure Both the Friedewald and the D were compared with the concentration obtained with the Reference Method The data from this study, summarized in Table 2, indicate that the mean biasof the D assayisnegative,whereas thatof the Friedewald calculationis positivethe comparison-ofmethods plot-13-quantification(x)vs testmethod (y)- gave a least-squares linear regression equation of y = 9x + 68 (r = 93) for the D, andy = 96x + 13 (r = 95) forthe Friedewald calculation (Fig IA and B) The resulting D systematicerror calculatedwas 13 mg/l (16%), 117 mg/l (63%), and 132 mg/l (66%) at concentrations of 81, 185, and 2 mg/l, respectively (see Materialsand Methods) The imprecision,defined here as the run-to-runcv (n = 2), was 52%, 45%, and 4% for these three respective concentrationsthe percentage totalerror (random errorplus systematicerror)for these same D concentrationswas 118%, 151%, and 144%, respectively, with a mean total error of 138% The effect of TG concentration on the performance of the D assay for 42 nonfasting samples (mean ± SD TG, a, E Q a a #{149} Beta-Quant (mg/l) Beta-Quant (mg/l) Table 2 CharacteristIcs of the groups used in the studya Patient status Fasting Nontasting Total Total sample no Age, years 117 ± ± ± 47 Range Total cholesterol ± ± ± 1 5 TG 157± ± ±3514 Median Range HDL-C 373 ± ± ± Quant ± ± ± 588 Friedewald 1 77 ± 566 Not calc Not caic Bias 586 ± 172 Not calc Not caic D ± ± ± 629 Bias -982 ± ± ± 323 Concentrations of cholesterol and TG are in mg/l Results are given as mean ± SD, except where noted 35 3 Q 25 -j a S c I Beta-Quant (mg/l) Fig 1 D (A) and the Friedewald (B) results compared with the equivalent value obtained by the modified 13-quantification in fasting pediatric patients (n = 96) by least-squares linear regression analyses (C) Datafrom the combinedfasting and nonfastingpatients (n = 138) I

4 ClinicalChemistry 42, No 8, S r=362,n=96,p<o1 A 4 2 a, E 4 2 U) Ca ,s #{149} #{149} S Beta-Quant (mgil) Triglycerides (mgil) 25 r=142, n = 138, NS B a) E U) 3 C) I I I I o i Beta-quant (mgil) Fig 2 determined by modified 13-quantification and by D assay in 96 fasting plasma samples (A) and in 138 samples from combined fasting and nonfasting patients (B) Bias (D minus /3-quantification) is plotted as a function of the modified /3-quantification concentrations Significance of r was determined by the t-test for correlation coefficients The unbroken line represents the best fit as determined by linear regression The horizontal dashed line reflects zero bias NS, not statistically significant Triglycerides (mgil) Fig 3 determined by modified 13-quantification and by D assay in 96 fasting plasma samples (A) and in 138 samples from combined fasting and nonfasting patients (B) Bias (Dminus /3-quantification)is plotted as a function of the plasma TG concentration, displayed on a linear scale (A) or logarithmically (B) Significance of rwas determined by the t-test for correlation coefficients 4854 ± 5457 mg/l) was examined by comparing the resultsof (3-quantification (x) with those of the D assay (y)(table 2) These samples gave the linear regression equation y = 124x (r= 85) The combined 138 samples (fasting plus nonfasting)gave the equationy = 92x + 19 (r = 86) (Fig 1C) Whereas the 96 fasting samples displayed a negative mean (± SD) bias of -982 ± 215 mg/l, the 42 nonfasting samples had a positive mean bias of 177 ± 432 mg/l (Table 2) Figure 2 shows a plot of the bias of the D results in relation to the concentration determined by 13-quantification In the 96 fasting samples (Fig 2A), this bias is negatively correlated with the 13-quantification (r = -254, P <5); in the combined fastingand nonfasting patients(n = 138), however, the correlation is not significant (Fig 2B) Figure 3 displaysthe bias as a function of plasma TG concentration in both the 96 fasting(fig 3A) and the 138 combined fastingand nonfastingpatients(fig 3B) Both anal-

5 1186 Harris et al: Direct LDL-cholesterol pediatric assay yses showed a significant positive correlation (P <1) between the bias and the plasma TG concentration We then assessedthe diagnosticperformance of the assays, using the NCEP cutpoints that divide the pediatric population on the basis of concentrations into acceptable, borderline, and high categories Fig 4 compares these clinical classifications as a grid display of diagnostic sensitivities that shows both correct and incorrect classifications The Friedewald calculation (Fig 4A) correctly classified 81 of the 96 fasting patients(84%), whereas for the same populationthe D (Fig 4B) correctlyclassified78 (81%) For the fastingand nonfasting populations combined (Fig 4C), the D correctly classified 8% (11 of 138) of the population We calculated the positive predictive value of each method at both of the NCEP cutpoints for pediatric patients For the D assay, the positive predictive value at the 13 mg/l and the >11 mg/l concentrations was 85% and 95%, respectively For the Friedewald formula, these were 97% and 98% at the same cutpoints Table 3, in which samples are placed into three groups according to plasma TG concentrations (<1, , and 2 mg/l), displays the fractionof patientscorrectlyclassified (accordingto the NCEP cutpoints) by the D results or the Friedewald calculation in comparison with the reference procedure value Both testmethods yieldedcomparable results Discussion In the routine clinical chemistry laboratory, the Friedewald formula [5] is widely used to calculate but often fails to meet the most recentncep performance goals[9]the current referenceprocedure, known as 13-quantification [13], isolates the population of particles that have a hydrated density between 16 and 163 kg/l Besides LDL particles, however, this heterogeneous population includes intermediate-density lipoproteins ( kg/l) and lipoprotein(a) (15-18 kg/l) For any new technique to be clinically useful and widely accepted, it should correlate with the 13-quantification method, which is the basisfor the database derived from the Lipid Research Clinics However, given that 13-quantification actuallymeasures a heterogeneous population of particles, the potential task of developing a precise and accuratedirectassayfordeterminingthe cholesterolcomponent of this wide-density LDL fraction appears all the more difficult [13, 2-22] Here, we describe the performance of a newly introduced assay (D) and its ability to classify hyperlipidemic children correctly according to the NCEP cutpoints The D assay requires a single measurement, whereas measurements of total cholesterol, HDL-C, and TG are needed for the Friedewald calculation The D assay is unaffected by substituting heparinized plasma for serum Frozen plasma samples,however, are unsuitableand show, on average, 25% less measured by D when compared with the Reference Method This finding is disappointing and may preclude use of the assayin long-term clinicaltrials,in which samples are usually stored frozen until analysis However, we did A Friedewaid Beta-quant < B Beta-quant < C 13 Beta-quant < < /13 54% 3/13 23% 3/13 23% % 4/11 364% 5/li 454% /72 % 2/72 28% 7/12 972% D < /13 846% 1/13 77% 1/13 77% 2/lI 181% 6/11 546% 3/il 273% 2/72 28% 9/72 125% 61/12 847% D < /29 793% 3/29 13% 3/29 13% 4/19 211% 1/19 526% 5/19 283% 2/9 22% 11/9 122% 77/9 855% Fig 4 Diagnostic performance of the Friedewald calculation (A) and the D assay (B and C) assessed accordingto NCEP guidelines All values are in mg/l Left-hand column, classification of the sample population by the /3-quantification Reference Method: top row, classification according to either the Friedewald calculation (A) or the Dassay (B and C) The blocks with the bold borders display the number of patients correctly classified by these assays (both absolute values and percentages) The remaining blocks are discrepant classifications The data in A and B are from the fasting patients (n = 96); those in C are from the combined fasting and nonfasting group (n = 138)

6 Clinical Chemistry 42, No 8, Table 3 Effect oftg concentration on classification into correct pediatric NCEP group TO, mg/l < By D 4/55 (73)8 27/3 (9) 11/11 (1) Results aregiven as no of subjectscorrectly (and % correctlyclassified) By F#{241}edewald 43/55 (78) 28/3 (93) 1/11 (91) classified/total no of subjects find that D-assayed results are stableat4 #{176}C forat least 3 weeks The random error of an assay is established by performing replicate measurements of a sample The run-to-run imprecision of the D assay has a concentration-dependent CV range of 4-52%, and therefore does not always meet the CV recommended by the NCEP (4%) Initially, the estimation of the run-to-run precision of the D assay presented a problem Our experience with the manufacturer s lyophilized controls has been that, once reconstituted and stored at 4 #{176}C, they show a gradual upward drift of measured concentrations over a 2-3 week period, resulting in an increase of -6-1% When the reconstituted control material was aliquoted and stored at 4#{176}C in 15-mL Microfuge tubes, the variation was less Freezing the control material does not appear to be a viable alternative because frozen samples show a continuous time-dependent decrease in D concentration [1] Table 2 and Fig 1 demonstrate that in the fasting pediatric subjects the D assay had a negative mean bias (±SD) of -98 (±2 15) mg/l vs the reference procedure In the same population the bias of the Friedewald calculation was + 59 (± 172) mgfl Another indicator of analytical error, the systematic error of an assay, may be further quantified by least-squares linear regression [18, 19] The NCEP performance goals require that an assay have an analytical bias of ± 4% [9] We note, however, that substituting a range of values (6-35 mgfl) into our D regression equation for fasting patients (y = 9x + 68) yielded calculated systematic errors of + 14% to -8% (data not shown) We therefore found a greater negative bias than did several other investigators who studied the performance of the D assay in adults [1-12] The explanation for the overall negative bias of the D is unclear Theoretically, the D assay determines the cholesterol that is present only in LDL and lipoprotein(a) particles-and therefore not in intermediate-density lipoprotein [13], unlike the 13-quantification procedure Perhaps this is the source of the D negative bias Alternatively, some LDL-containing plasma may be trapped by the latex beads in the D assay and thereby be excluded from the filtrate [1] Combining systematic error and random error to obtain the total analytical error [18, 19], we obtained a mean total analytical error in the 96 fasting patients at three D concentrations of 138% (the NCEP recommends a total analytical error of l2%) In contrast, for nonfasting samples (Table 2), the D assay had a mean positive bias of +177 (±432) mg/l Further analysis showed a positive correlation between plasma TG concentration and D bias (D - (3-quantification ) in both the fasting patients and the combined fasting and nonfasting patients (Fig 3) Two possible explanations for the D positive bias are that (a) the anti-apo A-I and anti-apo E on the latex beads become saturated in hypertriglyceridemic plasma, thus allowing the excess TG-rich lipoproteins to escape into the filtrate and increase the measured cholesterol, or (b) the larger TG-rich particles present in hypertriglyceridemia interfere with antibody binding through steric hindrance [1] Awareness of the triglyceride-dependent bias of the D assay is important in pediatrics because mean TG concentrations in this population are low relative to adults [1], implying that this negative bias will predominate We also noted an inverse correlation between the D bias and the measured by 13-quantification (Fig 2), the negative bias becoming more pronounced at higher concentrations Despite the above-mentioned negative systematic error of the D assay in fasting pediatric patients, the diagnostic performance of this assay was at least comparable with that of the Friedewald calculation as assessed by its ability to correctly classify such patients into the three NCEP groups However, the positive predictive value of the Friedewald calculation was better than that of the D assay at the 13 mg/l concentration When nonfasting patients were included in the analysis, the diagnostic performance of the D assay was largelyunaffected(fig4c) TG concentrations >4 mg/l (for which use of the Friedewald calculation is precluded) are unusual in pediatric practice In a fasting pediatric population, one would therefore commonly be able to use either the Friedewald calculation or the D assay Our results indicate that the latter would give little gain in terms of analytical or diagnostic performance We do, however, envision an application for the D assay in nonfasting pediatric patients, the diagnostic performance of this assay being comparable in both the fasting and nonfasting groups The D assay need not be used in the initial assessment of new pediatric hyperlipidemic patients Because fasting values for TG and HDL-C are needed in such patients, can be adequately derived from the Friedewald calculation In conclusion, therefore, despite its borderline-acceptable analytical performance with respect to the NCEP performance goals for total error, we recognize that the D assay is a potentially useful tool in the follow-up nonfasting management of children with hypercholesterolemia This population is particularly relevant for use with such a method because patient compliance for drawing a morning fasting blood sample is often achieved with some difficulty We are grateful to Sigma Diagnostics (St Louis, MO) for the generous donation of the D kits Vartouhi Galpchian and Mary Ellen Chmil provided excellent technical assistance

7 1188 Harris et al: Direct LDL-cholesterol pediatric assay We also appreciate the helpful statistical insights provided by Kimberlee Gauvreau of the Department of Cardiology, Children s Hospital References 1 Rifai N, Kwiterovich PC Jr Disordersof lipid and lipoprotein metabolism in children and adolescents In: Soldin Si, Rifal N, Hicks JMB, eds Biochemical basis of pediatric disease, 2nd ed Washington, DC: AACCPress, 1995: National Cholesterol Education Program (NCEP) Highlights of the report of the Expert Panel on Blood Cholesterol Levels in Children and Adolescents Pediatrics 1992;89: Schaefer EJ, Levi RI The pathogenesis and management of lipoprotein disorders N Engl J Med 1985;312: Schaefer El, McNamara JR Overview of the diagnosis and treatment of lipid disorders In: Rifal N, Warnick GR, eds Laboratory measurement of lipids, lipoproteins and apolipoproteins Washington, DC: AACC Press, 1994: Friedewald WI, Levy RI, Fredrickson OS Estimation of the concentration of low-density lipoprotein cholesterol in plasma without use of the ultracentrifuge Clin Chem 1972;18: Belcher JD, McNamara JR, Grinstead GF, Rifai N, Warnick GR, Bachorik P Measurement of low-density lipoprotein cholesterol concentration In: Rifai N, Wamick GR, eds Laboratory measurement of lipids, lipoproteins and apolipoproteins Washington, DC: AACC Press, 1994: Wamick GR, Knopp RH, Fitzpatrick V, Branson L Estimating low-density lipoprotein cholesterol by the Friedewald equation is adequate for classifying patients on the basis of nationally recommended cutpoints Clin Chem 199;36: McNamara JR, Cohn is, Wilson PWF, Schaefer El Calculated values for low-density lipoprotein cholesterol in the assessment of lipid abnormalities and coronary disease risk Clin Chem 199; 36: Bachorik PS, Ross JW National Cholesterol Education Program recommendations for measurement of low-density lipoprotein cholesterol Clin Chem 1995;41: McNamara JR, Cole TG, Contois JH, Ferguson CA, Ordovas JM, Schaefer El Immunoseparation method for measuring low-density lipoprotein cholesterol directly from serum evaluated Clin Chem 1995;41: Jialal I, Hirany SV, Devaraj S, Sherwood TA Comparison of an immunoprecipitation method for direct measurement of LDLcholesterol with /3-quantification (ultracentrifugation) Am J Clin Pathol 1995;1O4: Pisani I, Gebski CP, Leary El,Wamick GR, Ollington ifaccurate direct determination of low-density lipoprotein cholesterol using an immunoseparation reagent and enzymatic cholesterol assay Arch Pathol Lab Med 1995;119: RifaiN, WarnickGR, McNamara JR, BelcherJD, GrinsteadGF, FrantzIDJrMeasurement oflow-densitylipoprotein cholesterol in serum:a statusreportclinchem 1992:38: Hainline A Jr, Karon J, Lippel K, eds Manual of laboratory operations:lipidand lipoprotein analysis, 2nd ed HEW PubI No (NIH) (rev) US Govt Printing Office PubI No Bethesda, MD: National Heart, Lung and Blood institute, Lipid Research Clinics Program, 1982: Lum G, Gambino R A comparison of serum versus heparinized plasma forroutine chemistry tests Am J ClinPathol1974;61: Wu LL, Warnick GR, Wu JT,WilliamsRR, LalouelJM A rapid micro-scale procedure for determination of the total lipid profile Clin Chem 1989:35: Warnick GR, Benderson J, Albersii Dextran sulfate-mg2 precipitation for quantitation of high-density lipoprotein cholesterol Clin Chem 1982:28: Carey RN, GarberCC Evaluation of methods In: Kaplan LA, Pesce AJ, eds Clinical chemistry: theory, analysis and calculation St Louis, MO: CV Mosby, 1984: Westgard JO, Carey RN, Wold S Criteria for judging precision and accuracyin method development and evaluation Clin Chem 1974;2: Weiland H, Seidel D A simple specific method for precipitation of low density lipoproteins i Lipid Res 1983;24: KersherL,Schiefer5, DraegerB, MalerJ,Ziegenhom J Precipitation methods for the determination of LDL cholesterol Clin Biochem 1985:18: Demacker PN, Hijmans AG, Brenninkmeijer BJ, Jansen AP, van t Laar A Five methods for determining low-density lipoprotein compared Clin Chem 1984;3:1797-8