Cholesterol Screening: Comparative Evaluation of On-Site and Laboratory-Based Measurements

Transcription

1 CLIN. CHEM. 36/2, (1990) Cholesterol Screening: Comparative Evaluation of On-Site and Laboratory-Based Measurements Paul S. Bachorlk,1 Robert Rock,2 Teresa Cloey,1 Ester Treciak,2 Diane Becker,3 and WIllIam Sigmund3 We measured cholesterol in capillary blood samples from 9683 volunteers over a four-day on-site community screening program, using the Reflotron desk-top analyzer (Boehnnger-M#{225}nnheimDiagnostics, Indianapolis, IN). We also measured cholesterol in venous blood samples from 3% of those screened (a) with the Reflotron at the screening sites, (b) in a qualified hospital clinical laboratory, and (c) in a Centers for Disease Control standardized lipoprotein research laboratory. The sensitivity (and specificity) of the Refiotron measurements, with use of the lipoprotein laboratory measurements as the point of reference, was 0.95 (0.73) in capillary blood samples and 0.88 (0.93) in venous blood samples, compared with 0.99 (0.87) in the hospital clinical laboratory. The Reflotron measurements correlated less well with the lipoprotein laboratory values in both venous blood (r = 0.91) and capillary blood (r = 0.89) samples than did the clinical laboratory values (r >0.99). Furthermore, the capillary blood measurements averaged 7% higher than venous measurementswhen both kinds of samples were analyzed in the Reflotron. Additional Keyphrases: capillaty-venous differences - reflectance photometry - sensitm4 - specificity - variation, source of The association of high concentrations of blood cholesterol with premature coronary heart disease and the effectiveness of lowering the blood cholesterol concentration in decreasing coronary-disease risk are now well established (1-4). Within the past several years, attention has turned to cholesterol screening as a public health measure to identify at-risk individuals, who are then referred for follow-up evaluation and treatment as appropriate. The National Cholesterol Education Program (NOEP) Expert Panel on the Detection and Treatment of High Blood Cholesterol in Adults (5) recommended the use of uniform cutpoints for identifying individuals who had desirable, borderline-high, and high blood cholesterol concentrations, and developed guidelines for identifying and treating patients who are at risk by virtue of above-normal concentrations of low-density lipoprotein (LDL) cholesterol.4 In an attempt to facilitate the use of these common cutpoints, which are based on studies in which cholesterol was measured with standardized, highly accurate, and precise laboratory methods, the NCEP Laboratory Standardization Panel subsequently evaluated the current state of choles- Departments of Pediatrics and 3Department of Medicine, The Johns Laboratory Medicine, and Hopkins University School of Medicine, Baltimore, MD Depment of Laboratory Medicine, The Johns Hopkins Hospital. abbreviations: NCEP, National Cholesterol Educatiodensity Program; LSP, Lipid Standardization lipoprotein; CDC, Centers for Disease Panel; LDL, low- Control; LL, Lipsprotein Laboratory; and CL, Clinical Laboratory. Received August 18, 1989; accepted November 9, terol measurement in the United States and issued guidelines for interim and ultimate performance standards for accuracy and precision of cholesterol measurements (6). More-or-less coincident with these developments has been the introduction of portable desk-top analyzers that can quickly measure cholesterolin microliterquantitiesof blood. Designed to be operated with only periodic calibration, these instruments can be used in nonlaboratory settings such as the physician s office or field-testing sites. Some of these instruments can provide measurements in whole blood, serum, or plasma, and all of them can use either venous or capillary blood samples. The Greater Baltimore Cholesterol Screening Project was initiated to develop and evaluate screening strategies and evaluate the reliability of cholesterol measurements made with one of the currently used desk-top analyzers. As part of this program, we performed split-sample analysis of a subset of samples drawn from 9683 subjects in the Baltimore area who were being screened at five sites with a whole-blood cholesterol analyzer. Here we describe the accuracy and precision of the cholesterol measurements made in venous whole-blood and capillary whole-blood samples at the screening sites and compare these with measurements made in venous plasma in a licensed hospital clinical laboratory. We found thal, first, the hospitallaboratory-based measurements were the most reliable; second, the screening measurements correctly identified 90-95% of those screened with cholesterolconcentrations 2000 mg/l; and third, on-site screening measurements in venous whole-blood specimens tended to be more nearly accurate than those in capillary blood samples. MaterIals and Methods Study Design We screened approximately 9683 adults, ages 18 years and older-unselected individualswho volunteered-to identify those with plasma cholesterol concentrations of 2000 mg/l. We conducted the screening program during a four-day period in early April 1989 at five field sitesshopping areas and malls in the Baltimore area. Capillary blood was sampled by fingerstick and the cholesterol in whole blood was measured with the Refiotron desk-top analyzer (Boehringer-Mannheim Diagnostics, Indianapolis, IN). We also obtained a second venous blood sample from -3% of the screenees at each site, drawing blood into an evacuated collection tube (Vacutainer Tubes; Becton Dickinson, Rutherford, NJ) containing disodium EDTA (final concentration, 1.5 gil) as the anticoagulant. We measured cholesterol in both the capillary and the venous whole-blood samples at the screening sites, analyzing each capillary-venous pair in the same Refiotron instrument. The remainder of each venous sample was transported at 4#{176}C to The Johns Hopkins Hospital Clinical Laboratory (CL), where it was centrifuged (1500 x g, 15 mm, 400) to sediment the cells. The plasma was transferred to storage vials, which were then sealed and stored at 4 #{176}C for several days until analysis. CLINICAL CHEMISTRY, Vol. 36, No. 2,

2 We obtained a total of 314 venous-blood samples. We present here the results for 290 samples after excluding results for 21 persons being screened for whom capillary samples were not recorded and for three subjects in whom the screening values were below the measurement range for the Reflotron analyzer (<1000 mg/l). Fifty Reflotron analyzers, each equipped with a surge suppressor, were used at the five screening sites, and the measurements were made by trained operators. Before the instruments were deployed at the field sites, the operators had been instructed by the manufacturer s representatives on how to maintain and use them. Operators also went through a 2-h period of familiarization during which they analyzed specimens and quality-control sera. During the actual screening period, two lyophilized quality-control pools, furnished by the manufacturer, were analyzed each day in each instrument. The total cholesterol concentrations of these pools were approximately 1700 and 2220 mg/l; an aliquot of each was analyzed at the beginning of each screening session, and again after every 20 samples. Two separate lots of each control pool were used, one lot of each pool primarily at two of the screening sites and the other lot at the remaining three sites. In all, these pools were thus analyzed 354 to 425 times with the screening method. One lot of each pool was also analyzed in the CL and in the Johns Hopkins Hospital Lipoprotein Analytical Laboratory (IL), eight to nine times, in duplicate, at each laboratory. For the screening analyses, we applied 30 L of either capillary or venous whole blood to the test strips and placed the strips into the instruments. After 3 mm, the cholesterol concentration of each sample was displayed as a digital readout. The LL analyzed the venous plasma samples by a semiautomated method in which 10 AL of the sample was mixed with 1.0 ml of cholesterol reagent (Chol-PAP cholesterol reagent, cat. no ; Boehringer-Mannheim Diagnostics), incubated for 15 min at 37#{176}C, and the absorbance then measured at 500 nm. Standard curves were constructed with use of pure cholesterol standards (Preciset Standards, cat. no ; Boehringer-Mannheim), and the cholesterol concentrations were calculated from linearregression equations that related the absorbance of the standards to their concentration. During the study, the LL remained standardized for cholesterol measurement according to the criteria of the Centers for Disease Control (CDC)-National Heart, Lung, and Blood Institute Cholesterol Standardization Program (7). Day-to-day quality control in the LL was maintained by using two frozen control pools (Q15, Q18) furnished by the Clinical Chemistry Standardization Section, CDC. The cholesterol concentrations of these pools were established by CDC, by using the CDC modification of the reference method of Abell et al. (8). The Q15 and Q18 pools had reference cholesterol concentrations of 1800 and 2760 mg/l, respectively. Aliquots of two additional pools (VHA-1, VHA-2) were also assayed in the LL. The latter two pools, with cholesterol concentrations of 3120 and 1856 mg/l, respectively, were those used for day-to-day bench control in the CL. Aliquots of the venous plasma samples were also assayed in the CL in a Hitachi 737 analyzer (Boehringer-Mannheim Diagnostics). The analyses were performed according to the manufacturer s instructions, with an enzymatic cholesterol reagent from the same source as that used in the LL. Day-to-day quality control in the CL was maintained by using the VHA-1 and VHA-2 control pools described above. The CL also analyzed the Q15 and Q18 pools described above. The venous blood samples were analyzed in nine analytical runs done during three days in both the IL and CL. Results Accuracy and precision. In this study we used the CDO standardized cholesterol measurements made in the IL as the point of reference. As indicated in Table 1, the CV for Table 1. ScreenIng and Laboratory Analyses In QualIty Control Pools Laboratory analyses IL CL Screening-aft., analyses Chol conan, Chol cones, Chol cones, pa n mg/l CV, % mg/l CV, % n mg/l CV, % XLS-54A (31)b (24) (65) XLS-53A (43) (43) (110) (90) (159) VHA (39) (33) VHA (59) (87) (35) (40) (38) (56) #{149} The first four pools were used to monitor the screening analyses. The VHA pools were used for benchcontrol in the clinical laboratory,andthe 0 pools were used for this purpose in the Lipoprotein Analytical Laboratory.The referencevalues for015 and 018 were 1800 and 2760 mg/i, respectively,and were established bythe COC Clinical Chemistry StandardizationSection, by using the CDC modificationof the method of Abell et al. (. b Mean (and SD). 256 CLINICAL CHEMISTRY, Vol.36, No. 2, 1990

3 cholesterol measurement in the LL was 1.3% to 2.1%, as assessed from the measurements in all six quality-control sera. The analyses in the CL were performed with CVs of 1.5% to 2.8%, similarly estimated. The mean cholesterol concentrations of all six pools differed by 20 to 50 mg/l in the two laboratories, with the CL values tending to be slightly lower than the LL values and slightly closer to reference values as judged from the measurements of the Q15 and Q18 pools. The mean cholesterol concentrations measured in the control pools used at the screening sites (XLS-53A, XLS-54A) were within 1% of the values obtained in the LL. The CVs for both of these pools at the screening sites was <5%. As mentioned earlier, these two lots of pools were used primarily at two of the screening sites. The two lots used at the other three screening sites were not assayed by the LL. The CVs for the pools not assayed at the LL (lots and 33780) were 7.1% and 5.5%, respectively. Venous-capillary comparisons. The group mean cholesterol concentration for the 290 venous blood samples measured in the LL was 2160 (SD 419) mgfl. The values obtained in the CL averaged 2211 (SD 433) mgfl, with mean paired differences 2.4% higher (P <0.001), but the group standard deviations obtained in the two laboratories were similar. The values obtained in the venous wholeblood samples with the Refiotron analyzer, 2140 (SD 465) mg/l, averaged 0.9% lower than the LL values (not statistically significant). The greater group standard deviation of the screening analyses suggested somewhat greater variability of the Reflotron measurement as compared with the LL measurements. The measurements in the capillary blood samples averaged 2289 mg/l, 6% higher than the LL analyses in the venous plasma samples (P <0.001) and 7.0% higher than the Reflotron measurements in venous whole-blood samples (P <0.00 1). In addition, the group standard deviation of the Reflotron measurements in capillary samples was considerably higher (527 mgfl) than that in the LL, suggesting even greater variability of the screening analyses in capillary samples. Table 2 shows the linear-regression parameters relating each of the screening methods to the LL values. There was almost a perfect correlation between the CL and LL values (r = 0.995). In contrast, the coefficient of correlation between the Reflotron and LL measurements in both venous and capillary specimens was Furthermore, the correlation between measurements in capillary and venous whole-blood samples was only 0.88, despite the fact that Table 2. LInear RegressIon Parameters RelatIng the Cholesterol Measurements In Venous and CapIllary Samples Slop. x y (andse) LL CL (0.006) LL Ven (0.028) LL FS (0.033) Ven FS (0.031) (andse) (4.225) (19.710) (23.723) (24.890) Ven, screening measurements for venous whole blood; ES, screening measurementsfor capillary whole blood. n = 290. both measurements for each capillary-venous pair had been made in samples obtained at the same time, with the same method, and in the same Refiotron instruments. The standard errors for the slopes and intercepts for the Reflotron analyses were four to five times greater than those we observed for the CL analyses, again suggesting considerably greater variation of the Reflotron values than that obtained with the laboratory-based method. Accuracy of classifying subjects. In view of the apparently greater variability of the Reflotron measurements, we examined the sensitivity and specificity of these measurements for detecting hypercholesterolemia (cholesterol 2000 mgfl) in both capillary and venous samples at the screening sites, and for comparison, the venous plasma measurements made in the CL (Table 3). For this purpose, we used the LL measurements as the point of reference. The sensitivity and specificity of the cholesterol analyses in the CL were 0.99 and 0.87, respectively. In contrast, the Reflotron measurements in venous samples had a sensitivity of 0.88 and a specificity of 0.93, and the respective values for capillary samples were 0.95 and Thus, the CL correctly identified 99% of the hypercholesterolemics in this study and misclassified 13% of the normal subjects as being hypercholesterolemic. The Reflotron measurements in capillary blood samples correctly identified 95% of the hypercholesterolemics, but at the expense of misclassifring 27% of those who actually had desirable cholesterol concentrations. Despite the greater accuracy of the Reflotron analyses for venous blood samples, these measurements identified only 88% of the hypercholesterolemics, but misclassified only 7% of those with normal cholesterol concentrations. The greater variability of the Refiotron measurements compared with those in the CL is illustrated in Table 4, with the LL measurements used as the point of reference. Over half of the measurements of capillary samples differed (i.e., were either greater or less than) from the IL measurements by >100 mg/l or about 5%. One-third of the measurements differed by >200 mg/l, and almost 20% of the samples differed by >300 mgfl. The Reflotron measurements of venous blood samples were less variable: 19% Table 3. SensitIvIty and SpecIficIty of the ScreenIng Tests for DetectIng Hypercholesterolemlc Subjects Method SensItivity SpecificIty Reflotron Capillary blood Venous blood CL values Subjectswith cholesterol 2000 mg/l, withthe standardized Lipid Laboratoryvalues considered to be the true values. r Table 4. ProportIon of Samples with Large DIfferences between ScreenIng and Reference Measurements Percentage of values differing from IL values by (mg/i) Method >100 >120 >150 >200 > Reflotron Capillary blood Venousblood CLvalues CLINICAL CHEMISTRY, Vol. 36, No. 2,

4 Screen Capillary blood samples. The range of uncertainty of classification was even greater than that for venous blood samples (Figure 1). Lipid Lob -0-. Screen Venous Lipid Lab Screen Clin Lab Lipid Lab Cholesterol Conc (mg/l) Fig. 1. Ranges of uncertainty of screening analyses calculated from instances in which the reference and screeningvaluesgaveopposite impressions of risk Cutoffused, 2000 mg/l. (0, #{149}) n = 15; (& A) n = 30; (0, ) n 38. Points labeled Lipid Lab indicate the group meansas measuredin the LL for each of the three groups of samples,and theerror bars aroundthese points are the group SD8. The pointslabeled Screen Clin Lab, Screen Capillary, and ScreenVenous each indicatethe group mean values for the respective screening methods; the error bars around these points represent ± 2 SD for the means of the paired differences between the LL and screeningvalues differed from the LL measurements by >100 mgfl, and 6% of the samples differed by >300 mg/l. In contrast, only 11% of the CL measurements in venous plasma differed from the LL values by >100 mg/l, and only 1% of the samples differed by >150 mg/l. Finally, we examined the results for samples in which the screening measurements in venous or capillary blood samples, and the LL measurements in venous plasma samples, gave opposite impressions of risk. The lower portion of Figure 1 shows the results for the 15 venous samples in which opposite impressions of risk were obtained by the LL and CL, the group means (and SD) for these samples being 2019(30) mgfl in the CL and 1959 (26) mgfl in the LL. The average paired difference between the CL and LL measurements was 60 mgfl, and the standard deviation of the paired difference was 30 mg/l. Because the precision of the analyses in both laboratories was similar, misclassification resulted primarily from the bias between the measurements, and occurred most often in the concentration range of 1900 to 2000 mg/l (measured in the LL). Thirty of the venous blood samples were misclassified on the basis of the Refiotron measurements, the mean (and SD) cholesterol concentration of these samples being 2033 (93) mgfl. For these samples, the Refiotron values averaged 1971 mgfl, or 63 mgil lower than the LL values. The standard deviation of the paired difference between the Reflotron and LL values, however, was 214 mg/l, or about fivefold that for the CL analyses, and the range of uncertainty of classification was correspondingly larger (Figure Thirty-eight individuals were misclassified on the basis of the results for the capillary blood samples. The mean (and SD) cholesterolconcentrationin these samples was 1977 (140) mg/l as measured in the corresponding venous plasma samples in the LL. The Refiotron measurements for capillary blood samples averaged 88 mgfl higher, and the standard deviation of the paired differences was 423 mg/l, or >10-fold higher than the CL measurements for venous DIscussion In this study we evaluated the reliability of essentially three different kinds of cholesterol measurements: those in capillary or venous blood samples assayed with the Reflotron desk-top analyzer, and those in venous plasma samples assayed in a licensed, hospital clinical laboratory. We used the measurements made for venous plasma samples in a CDC-standardized lipoprotein research laboratory as the reference point for these comparisons. Reliability of clinical laboratory measurements. The most nearly accurate and precise measurements were made in the licensed clinical laboratory. Assuming that our findings in the subset of subjects we examined can be applied to the entire 9683 subjects in the population screened, we would expect to have misclassified 1% of the hypercholesterolemice, or 61 of the 6100 subjects with cholesterol concentrations of 2000 mg/l. Furthermore, we would expect to misclassify 466 (13%) of the subjects with cholesterol concentrations <2000 mgfl. This represents an overall misclassification rate (false positives plus false negatives) of 5.4%, or 527 of the 9683 individuals screened. Considering the accuracy and precision of the analyses in the CL, this degree of misclassification probably represents about the best that can be reasonably achieved by a qualified clinical laboratory making the measurements in venous samples. It is evident, but might be mentioned nevertheless, that our estimates of misclassification are based on the assumption that the LL measurements were without bias. As can be seen from the results for the two Q-pools (Table 1), there was a slight positive bias in the LL, and the estimates of misclassification will have been slightly over- or understated. Reliability of screening measurements at the field testing sites. Desk-top analyzers, because of their portabilityand the speed with which results are obtained, lend themselves to large-scale screening programs. They are, however, generally operated in nonlaboratory settings by nurses, phlebotomists, or others with limited formal laboratory training. The results obtained under these conditions are known to be less reliable than laboratory-based measurements, or measurements made at nonlaboratory screening sites by qualified laboratory technologists (9). In the present study, sites were established as state-licensed laboratories to conform with Maryland state regulations. The five nonlaboratory-based screening measurements were made at these sites by nonlaboratory personnel under the supervision of qualified laboratory technologists, who were required by state regulation to be present at each of the screening sites. Even under these conditions, 732 hypercholesterolemic subjects and 251 normoeholesterolemic subjects, or 10.2% of the screenees, would have been misclassified on the basis of the LL measurements in venous plasma samples. Because there was no significant bias, on average, between the Reflotron and LL measurements in venous samples, this misclassification can be attributed primarily to the greater variability of cholesterol analysis with the Reflotron analyzer. On the basis of Reflotron measurements in capillary blood samples, 305 hypercholesterolemic subjects and 967 normocholesterolemic subjects would have been misclassified, for a total misclassification rate of about 13% of the 9683 subjects screened. The 258 CLINICAL CHEMISTRY, Vol. 36, No. 2, 1990

5 measurements in capillary blood were less precise than those in venous blood samples and exhibited a significant positive bias with respect to the LL values. The positive bias led to a somewhat greater sensitivity for detecting hypercholesterolemia than that observed with venous blood samples, but reduced the specificity of the analyses by fourfold with respect to that in the venous blood samples. The accuracy and precision of the measurements in both the CL and LL satisfied the ultimate performance criteria for cholesterol measurement specified by the NCEP Laboratory Standardization Panel (6). These recommendations specify that cholesterol measurements should be accurate to within 3% of reference values and should be operated with a CV of <3%. Based on the LL measurements in venous plasma, the cholesterol measurements made with the Reflotron analyzer were also within 3% of reference values, and in two of the five screening sites, the CV for cholesterol measurement was <5%. In the other three sites, however, the CVs exceeded 5%, and for one of the control pools the CV was almost 8%. Because those pools were not analyzed in the LL, we cannot state with certainty whether the lower precision was due to analytical error at the three screening sites or to differences in the poois themselves. Probably, however, the lower precision was due to analytical error. Several of our findings are consonant with those we obtained in a previous study, in which we examined the accuracy and precision of cholesterol measured in capillary samples with the Reflotron analyzer at four centers of the Lipid Research Clinics Program (10). In that study, the sensitivity of the measurements, with 2000 mg/l as the cutoff, was also -=95%. The specificity, however, was about 80% to 90%, compared with 73% in the present study. The lower specificity in the present study probably results from the slightly positive bias of the measurements in the capillary samples used in the present study, whereas in the earlier study the bias was slightly negative (10). In addition, in our previous study, in some samples the difference between the screening and laboratory measurements exceeded 300 mg/l, a difference not accounted for by the method-based bias or by random error; this occurred in 5-15% of the samples (10) vs 18% in the present study. Comparison of measurements in venous and capillary blood samples. Finally, the measurements in capillary blood samples were about 7% higher than those in venous blood samples, when both kinds of sample were analyzed with the same Reflotron. The venous measurements were made on blood that had been collected with EDTA as the anticoagulant, and would be expected to be about 3% lower than they would have been for serum (11). Heparin exerts no appreciable osmotic effect, however, and the measurements in the capillary samples therefore reflect values for capillary blood-serum. Thus, the actual positive bias in capillary blood samples would have been about 4%, due primarily to the physiological difference between venous and capillary blood. Other reasons can be discounted as follows: first, each venous-capillary pair was measured at the same time with the same Refiotron instrument. Second, the measurements for both venous and capillary samples were made on whole blood, and the differences therefore could not be attributed to the presence or absence of blood cells, per se, in the sample. Third, on average, the measurements made in the venous blood samples with the Reflotron were not significantly different from those made in venous plasma in the LL. Thus, the positive bias of the measurements in the capillary samples was not a consequence of inaccuracy in the Reflotron method itself. Various, and sometimes contradictory, estimates of venous-capillary specimen differences have been reported, from as little as 4% or less (12) to 8-12% (13, 14). Indeed, in our previous study (10), screening measurements in capillary blood agreed within about 4% with the measurements made in venous blood-plasma samples with the reference method. In the present study, the larger group standard deviation for cholesterol measurements in capillary blood samples compared with those in venous blood samples measured with the Reflotron is also consonant with greater variability in measurements made in capillary samples. The source(s) of the bias of the capillary blood measurements is not entirely clear. It could be related to the technique used for sample collection or to actual differences in the cholesterol concentration of venous and capillary blood samples, or both. According to our current findings, the most reliable screening measurements were obtained when the analyses were performed in venous plasma samples by a qualified clinical laboratory. Although the overall accuracy of measurements made in venous blood samples with the Reflotron analyzer compared favorably with that of the venous plasma measurements in the LL, the former results were more variable, and the degree of misclassification based on such measurements was attributable to variability rather than to inaccuracy of the measurements. The most-variable measurements were obtained with the capillary samples, and these measurements seemed to be most prone to misclassification overall. Most of the error, however, was misclassification of individuals whose cholesterol values were actually <2000 mg/l. Such an error should be detected in follow-up analyses, so it would therefore presumably be of less consequence than the failure to identify true hypercholesterolemics. Thisstudywas supported inpart by thejohns Hopkins PreventiveCardiology Programs, the Johns Hopkins Lipid Research- Atherosclerosis Unit, and The Johns Hopkins Hospital Clinical Laboratory. Forty-five of the 50 Refiotron analyzers used inthis study were lent to us by Boehringer-Mannheim Diagnostics, Indianapolis, IN. Television station WMAR, Baltimore, MD, provided the publicity for the four-day screening program. We thank Carol McGeeney for secretarial assistance. References 1. Lipid Research ClinicsProgram.The LipidResearch Clinics Coronary Primary Prevention Trial Results. I. Reduction in incidence of coronary heart disease. J Am Med Assoc 1984;251: Lipid Research Clinics Program. The Lipid Research Clinics Coronary Primary Prevention Trial Results. II. The relationship of reduction in incidence of coronary heart disease to cholesterol lowering. J Am Med Assoc 1984;251: Frick MH, Elo 0, Haapa K, et al. Helsinki Heart Study: primary prevention trial with gemfibrozil in middle-aged men with dyslipidemia. N Engl J Med 1987;317: Brensike JF, Levy RI, Kelsey SF, et al. Effects of therapy with cholestyramine on progression of coronary arteriosclerosis: results of the NHLBI Type H Coronary Intervention Study. Circulation 1984;69: The Expert Panel. Report of the National Cholesterol Education Program Expert Panel on Detection, Evaluation and Treatment of High Blood Cholesterol in Adults. Arch Intern Med 1988;148: Laboratory Standardization Panel of the National Cholesterol Education Program. Current status of blood cholesterol measurement in clinical laboratories in the United States. NIH Publ. No Bethesda, MD: Natl. Institutes of Health. January CLINICAL CHEMISTRY, Vol. 36, No. 2,

6 7. Myers GL, Cooper GR, Winn CL, Smith SJ. The Centers for Disease Control-National Heart, Lung, and Blood Institute Lipid Standardization Program. Clin Lab Med 1989;9: Abell LL, Levy BB, Broclie BB, Kendall FE. A simplified method for the estimation of total cholesterol in serum and demonstration of its specificity. J Biol Chem 1952;195: Belsey R, Vandernbark M, Goitein RK, Baer DM. Evaluation of a laboratory system intended for use in physician s offices. H. Reliability of results produced by health care workers without formal or professional laboratory training. J Am Med Assoc 1987;258: Bachorik PS, Bradford RH, Cole T, et al. Accuracy and precision of analyses for total cholesterol as measured with the Refiotron cholesterol method. Clin Chem 1989;35: Laboratory Methods Program of the National Committee of the Lipid Research Clinics Heart, Lung, and Blood Institute. Cholesterol and triglyceride concentrations in serum/plasma pairs. Clin Chem 1977;23: Koch TR, Mehta U, Lee H, et al. Bias and precision of cholesterol analysis by physician s office analyzers. Clin Chem 1987;33: Kupke IR, Zeugner S, Gottschalk A, Kather B. Difference in lipid and lipoprotein concentrations of capillary and venous blood samples. Clin Chim Acts 1979;97: El-Dering S, Ng RH, Staatland BE. Evaluation of the Kodak Ektachem DT-60 Analyzer [Tech Brief]. Clin Chem 1986;32: CLINICAL CHEMISTRY, Vol. 36, No. 2, 1990