E L S E V I E R Clinica Chimica Acta 256 (1996) 125 134

Significant reduction of the bias among commercial immunoassays for lipoprotein(a) after

use of a uniform calibrator

Giuseppe Lippi, Claudia Lo Cascio, Orazio Ruzzenente, Claudio Brentegani, Giancesare Guidi*

Laboratorio di Analisi Chimico-Cliniche e Microbiologiehe, Istituto di Chimica e Microscopia Clinica, Universitgt di Verona, Centro Ospedaliero Clinicizzato, 37067 Valeggio s /M Verona, Italy

Received 26 March 1996; revised 3 July 1996; accepted 3 August 1996

Abstract

Despite the increasing interest in the measurements of lipoprotein(a) (Lp(a)) in serum or plasma, at present there is no effective standardization for Lp(a) assays; the main problems to solve are represented either by the lack of a suitable primary standard or by the absence of a reliable and widely available reference method. A first step is hence the uniformity of calibration of different immunoassays. We calibrated three commercial immunoassays for Lp(a) (enzyme linked immunosorbent assay (ELISA), latex-enhanced immunonephelometric assay (LINA), and immunonephelometric assay (INA)) with either proprietary standards or purified Lp(a) material obtained with a rapid and simple procedure. Final results of purified Lp(a) calibration were reported in terms of protein Lp(a) mass whereas we were able to quantify the exact protein concentration of our purified lipoprotein. The uniformity of the calibration of the different assays led to a significant improvement of regression slopes (from 1.88 to 0.90 ELISA vs. LINA, from 1.45 to 0.95 ELISA vs. INA and from 1.27 to 0.96 INA vs. LINA) and correlation coefficients (from 0.990 to 0.994 ELISA vs. LINA, from 0.987 to 0.990 ELISA vs. INA and from 0.985 to 0.987 INA vs. LINA). Furthermore, the significant differences among Lp(a) values obtained after calibration with proprietary standards were minimized, becoming non-significant in two out of three cases. In conclusion, we demon- strated that a better agreement of Lp(a) values obtained with different commercial assays could be simply reached by uniformity of calibration and by employing standards with values accurately measured. Copyright © 1996 Elsevier Science B.V.

* Corresponding author. Fax: + 39 45 7950188.

0009-8981/96/$15.00 Copyright © 1996 Elsevier Science B.V. All rights reserved Pl l S0009-8981(96)06413-3

126 G. Lippi et al. / Clinica Chimica Acta 256 (1996) 125-134

Keywords: Lipoprotein(a); Immunoassay; Atherosclerosis; Purification

I. Introduction

Several studies indicate lipoprotein(a) (Lp(a)) as an independent risk factor for cardiovascular diseases, even at a young age [1-6]. Lp(a), firstly discovered by K~re Berg in 1963 [7], resembles in structure and properties a low density lipoprotein but with the presence of the unique apolipo- protein(a) (apo(a)) linked by a single disulfide bridge to apolipoprotein B10o [8,9].

Despite the great clinical interest in measuring Lp(a) in serum or plasma, at present there is no effective standardization for lipoprotein(a) assays [9-12], chiefly due to the extreme size-polymorphism of the molecule; more than 34 alleles have been discovered for the apo(a) locus [13] and almost 34 isoforms have been resolved in plasma employing SDS agarose gel elec- trophoresis [14]. Such high polymorphism, due to multiple repeats of a single domain, the kringle IV type 2, gives rise to several problems in Lp(a) immunoassays, in which antibodies directed against epitopes located on the repeated motifs of apolipoprotein(a) usually overestimate samples contain- ing high-molecular weight isoforms and underestimate samples containing low-molecular weight isoforms [15].

An additional source of inter-laboratory variability in Lp(a) measure- ments originates from the lack of a suitable reference material, a primary 'gold standard' that should be applied to transfer values to secondary standards and hence to commercial calibrators. Assuming that by unifor- mity of calibration we would reach a better agreement of Lp(a) values using different commercial immunoassays, we evaluated 33 samples with three different kits, calibrated either with their own proprietary standards or with Lp(a) purified by a rapid and original procedure [16].

2. Materials and methods

Blood was collected from 33 healthy volunteers after 1 night fasting in 10 ml vacuum tubes with no additive and centrifuged at 3000 g to obtain serum. Haemolytic and lipaemic sera were excluded from the study.

Lipoprotein(a) was measured with three commercial kits, strictly follow- ing the instructions of the manufacturers. 1) Macra Lp(a) (Strategic Diag- nostic, Newark, N J, USA) is an enzyme linked immunosorbent assay (ELISA) that uses monoclonal anti-apo(a) antibody-coated plates for cap- turing the Lp(a) particles and horseradish peroxidase (HRP)-conjugated

G. Lippi et al. / Clinica Chimica Acta 256 (1996) 125 134 127

anti-apo(a) polyclonal antibody for detection. Results are obtained after plotting by polynomial regression the standard curve of mean absorbance vs. respective concentrations of Lp(a) scalar solutions. Each sample was analyzed in duplicate and final results were expressed as total lipoprotein mass. Intra- and inter-assays coefficient of variations (CVs) were < 5% and < 10.4% respectively. 2) The Lp(a) latex-enhanced immunonephelometric assay (LINA) (Behringwerke, AG, Marburg, Germany) was performed on Behring Nephelometric Analyser (BNA) (Behring). The method is based on polystyrene particles coated with three anti-Lp(a) monoclonal antibodies [17]. The intensity of scattered light produced by the agglutination of the particles as consequence of the reaction with Lp(a) is proportional to the concentration of Lp(a) in the sample. Results are evaluated by means of logitlog function of light scattered intensities vs. respective concentrations of scalar dilutions of Lp(a) standard (Behring). Each sample was analysed singly and final results were expressed as total lipoprotein mass. Intra- and inter-assay CVs ranged respectively from 1.5-3.0% and from 1.7-3.2%. 3) LPA immunonephelometric assay (INA) (Beckman Instruments Inc, Gal- way, Ireland) was performed on ARRAY instrument (Beckman). The method measures the rate of increase in light scattered from particles suspended in solution as a result of complexes formed during the reaction between Lp(a) and anti-Lp(a) antibodies. Following calibration, the peak rate signal, which is a function of the concentration of Lp(a) in the sample, is automatically converted to concentration units by the analyzer. Each sample was analyzed singly and final results were reported in terms of total lipoprotein mass. Intra- and inter-assay CVs were respectively < 5 and < 8%. Purified lipoprotein(a) was obtained as previously described [16]. Briefly, 15 g of lysine-Sepharose (Pharmacia, Uppsala, Sweden) was swollen in 50 ml of distilled water, packed into a refrigerated 40 x 2.6 cm I.D. chromatographic column and extensively washed with 500 ml of 50 mmol/1 (NH4)HCO3, pH 7.95. 100 ml of serum obtained from a single healthy volunteer homozygous for a previously identified medium size isoform were filtered and further injected in a single passage into the column. Non-spe- cific proteins were initially washed-out with almost 500 ml of pH 7.95, 50 mmol/1 (NH4)HCO3 and non-Lp(a) kringle-containing proteins exhibiting low affinity for lysine-Sepharose were eluted with 50 ml of 0.5 tool/1 sodium phosphate buffer, pH 7.4. Lipoprotein(a) was finally eluted with 50 ml of 0.5 mol/1 sodium phosphate buffer, pH 4.4.

Total proteins concentration of the purified material was measured with the method of Lowry [18] and final concentration was, therefore, expressed in terms of Lp(a) protein mass instead of total lipoprotein mass.

The calibration of each assay was either performed employing proprietary standards or purified Lp(a), strictly following the manufacturer's proce-

128 G. Lippi et al. / Clinica Chirnica Acta 256 (1996) 125-134

qD

< : Z

I %-. E:Z.

140

120

I O0

80

60

40

20

I I I a

//" I I I

0 I I I I I I

0 20 40 60 80 1 O0 120

Lp(o)-EL[SA (mg /d l )

40

v

< ( Z m J

I %-,

f f

14

12

10

0 I I I 0 6 S 10 12 14

I I J l ~ I

b O0

/ °

0

/

/ @ ~ I I E _1 I _ I

2 4

Lp(a)-EL/SA (mg /d l )

Fig. 1. Comparison between ELISA and LINA values of Lp(a) in 33 serum samples employing: (a) Proprietary standard calibration (total Lp(a) lipoprotein mass): rs = 0.990; equation of the regression: y = 1.88x + 3.06; (b) Purified Lp(a) material calibration (Lp(a) protein mass): rs 0.994; equation of the regression: y = 0.90x + 0.11.

G. Lippi et aL / Clinica Chimica Acta 256 (1996) 125-134 129

6 0

5 0

c - 4O q:3

F < 50 Z

I

Z" c~ 2@

10

I I I I a

O 0 0 0 0

o 10 20 50 40

hp(a) -Eh lSA ( m g / d l )

I I

5 0 67

F . <

Z

I

_J

14

12

10

I I I b

o j

4 ! o

2

0 o 2 4 6

I I I I

j o ~ ' o

O

I

I I I

8 10 12 14

Lp(a)-ELISA ( m g / d l )

Fig. 2. Comparison between ELISA and INA values of Lp(a) in 33 serum samples employing: (a) Proprietary standard calibration (total Lp(a) lipoprotein mass): rs = 0.987; equation of the regression: y = 1.45x + 0.21; (b) Purified Lp(a) material calibration (Lp(a) protein mass): r s = 0.990; equation of the regression: y = 0 . 9 5 x - 0.01.

130 G. Lippi et al. / Clinica Chimica Acta 256 (1996) 125-134

~D

<~ Z

r 'S"

k

140

120

100

8O

60

¢0

20

q I J L / o/ /

/ ©

¢ /

Y o 2o

I r I I

40 60 80 100

hp(a) INA ( rag /d0

120 40

Z

I

14

12

10

F T ]

b / / < o

, /

0 l r 0 2 4 6 8

I I

10 12

-t r

Lp(,a)-INA (mg /d l )

Fig. 3. Comparison between INA and LINA values of Lp(a) in 33 serum samples employing: (a) Proprietary standard calibration (total Lp(a) lipoprotein mass): rs = 0.985; equation of the regression: y = 1.27x + 2.93; (b) Purified Lp(a) material c/dibration (Lp(a) protein mass): rs = 0.987; equation of the regression: y = 0.96x + 0.1.

G. Lippi et al. / Clinica Chimica Acta 256 (1996) 125 134 131

dures. The statistical analysis was performed using Astute (DDU Software, The University of Leeds, UK). As Lp(a) values in Caucasians have been reported to follow a non-Gaussian distribution (e.g. [19]), regressions were performed according to the Passing and Bablok approach [20] and correla- tion coefficients were calculated using the Spearman rank correlation. Statistical significance of the difference in Lp(a) values separately obtained with the three assays was evaluated with the Wilcoxon signed rank test.

3. Results

Results of regression analysis between methods and respective correlation coefficients after proprietary standard calibration are illustrated in Fig. l a, Fig. 2a and Fig. 3a. As shown, slopes and intercepts were unsatisfactory, as previously observed in several other reports [21-24]. In particular, Lp(a) values obtained with the ELISA are consistently lower than those obtained, either with the LINA or with the INA (slopes: 1.88 vs. 1.45 respectively) and Lp(a) values obtained with the INA are lower than those obtained with the LINA (the slope was 1.27). Moreover, a significant difference could be observed in Lp(a) concentrations obtained with the three assays using the Wilcoxon signed rank test (Table 1).

After common calibration employing purified Lp(a) material, a significant improvement of the regressions could be observed (Fig. lb, Fig. 2b and Fig. 3b), indicating that an excellent agreement of results could be achieved and suggesting that values of total Lp(a) protein mass were virtually identical as confirmed by the Wilcoxon signed rank test, with the only exception of the comparison between ELISA and LINA (Table 1). The value of the slopes improved almost constantly and progressed from 1.88 to 0.90 (ELISA vs. LINA), from 1.45 to 0.95 (ELISA vs. INA) and from 1.27 to 0.96 (INA vs. LINA) before and after use of uniform calibration respectively, showing an additional positive effect on the relative values of the intercepts.

Table 1 Wilcoxon signed rank test of Lp(a) values obtained with three commercial kits following calibration with either proprietary standard or purified Lp(a) material

Proprietary standard calibration Purified Lp(a) calibration

ELISA vs. L I N A P<0.0001 P = 0.016 ELISA vs. I N A P<0.0001 P = 0.1 I N A vs. L I N A P<0.0001 P = 0.43

132 G. Lippi et al. / Clinica Chimica Acta 256 (1996) 125-134

Although the Spearman coefficients of correlation among the three methods with proprietary calibration were always greater than 0.985, a slight improvement could be observed after use of uniform calibration (Figs. 1-3).

4. Discussion

The main problems to solve in the standardization of Lp(a) measure- ments are the lack of a suitable reference material and the lack of reliable commercial certified methods in order to achieve standardization and uniformity of expression of Lp(a) measurements [9-12]. At present, the only known reference method has been presented by Marcovina et al. [15] but it is not yet commercially available. While awaiting a definitive and wide- spread reference method, a first important step is, hence, the identification and characterization of a common standard.

The correct procedure for the standardization of methods involves the transfer of values from a certified reference material to secondary standards. The use of a single company's standard is not practicable at the moment inasmuch as the value declared by the companies is often obtained through uncertified procedures. Therefore, we decided directly to use purified Lp(a) material as a secondary standard since we were able to measure the exact protein concentration, reporting final results of the assays in terms of protein Lp(a) mass. The purified material was obtained from single serum instead of pooled sera. We selected a medium sized isoform in order to minimize the bias due to different immunoreactivity displayed by a mixture of isoforms of different size contained in each proprietary standard towards size-sensitive anti-Lp(a) or anti-apo(a) antibodies [10].

As predicted, the introduction of a common standard with an accurate measurement of the Lp(a) value greatly improved the whole comparability among the assays.

The first achievement was a remarkable improvement of the regressions (Figs. 1-3). The reliability of the results was further confirmed by the Wilcoxon signed rank test (Table 1). Before uniformity of calibration, Lp(a) values showed a significant difference among all the assays (P < 0.0001); after standardization with purified Lp(a) material, the differences among protein Lp(a) values were minimized, becoming non-significant in two out of three cases. In the case of ELISA vs. LINA, the non-significance could not be reached and we suggest that the disparity in immunoreactivity could play a major role in hindering the concordance among data. Further support for this suggestion is shown by the improvement of the Spearman correlation coefficients attained after uniformity of calibration (Figs. 1-3).

G. Lippi et al. / Clinica Chimica Acta 256 (1996) 125-134 133

Although we demonstrated that the correlation between some commercial immunoassays is acceptable, the lack of standardization, as appeared in the present study, can lead to a serious misclassification of subjects at risk for cardiovascular diseases, chiefly depending on the commercial assay em- ployed. We were able to demonstrate that, by uniformity of calibration, a significant reduction of the differences between commercial immunoassays based on different analytical techniques, could be attained. We believe that this observation can be considered a prior but not negligible step in the complex effort that will be needed to achieve uniformity and harmonization of results of Lp(a) measurements.

References

[1] Rhoads GG, Dahlen GH, Berg K, Morton NE, Dannenberg AL. Lp(a) lipoprotein as a risk factor for myocardial infarction. J Am Med Assoc 1986;256:2540-2544.

[2] Sandkamp M, Funke H, Schulte H, K6hler E, Assmann G. Lipoprotein(a) is an independent risk factor for myocardial infarction at a young age. Clin Chem 1990;36/ 1:20-23.

[3] Labeur C, De Bacquer D, De Backer G, Vincke J, Muyldermans L, Vandekerckhove Y et al. Plasma lipoprotein(a) values and severity of coronary artery disease in a large population of patients undergoing coronary angiography. Clin Chem 1992;38:2261- 2266.

[4] Jiirgens G, Taddei-Peters W, K61tringer P, Petek W, Chen Q, Greilberger J, Macomber P, Butman B, Stead A, Ransom J. Lipoprotein(a) serum concentration and apolipo- protein(a) phenotype correlate with severity and presence of ischemic cerebrovascular disease. Stroke 1995;26:1841-1848.

[5] Willeit J, Kiechl S, Santer P, Oberhollenzer F, Egger G, Jarosch E, Mair A. Lipo- protein(a) and asymptomatic carotid artery disease. Stroke 1995;26:1582-1587.

[6] Cambillau M, Simon A, Amar J, Giral PH, Atger V, Segond P, Levenson J, Merli I, Megnien JL, Plainfosse MC, Moatti N and PCVMETRA Group. Serum Lp(a) as a discriminant marker of early atherosclerotic plaque at three extracoronary sites in hypercholesterolemic men. Arterioscler Thromb 1992; 11:1346-1352.

[7] Berg K. A new serum type system in man: the Lp system. Acta Pathol Microbiol Scand 1963;59:369-382.

[8] Albers J J, Kennedy H, Mareovina SM. Evidence that Lp(a) contains one molecule of apo(a) and one molecule of apoB: evaluation of amino acid analysis data. J Lipid Res 1996;37:192 196.

[9] Marcovina S, Levine DM, Lippi G. Lipoprotein(a): structure, measurement and clinical significance. In: Rifai N, Warnick R, eds. Laboratory measurement of lipids, lipo- proteins and apolipoproteins. A A C C Press 1994:235-263.

[10] Albers JJ, Marcovina SM, Lodge MS. The unique lipoprotein(a): properties and immunochemical measurement. Clin Chim Acta 1990;36:2019-2026.

[11] Kostner GM. Standardization of Lp(a) assays. Clin Chem 1992;21 l: 191 195. [12] Guidi GC, Lippi G. Standardization of lipoprotein(a) measurements: an overview.

Biochim Clin 1995;19:703 709.

134 G. Lippi et al. / Clinica Chimica Acta 256 (1996) 125-134

[13] Lackner C, Cohen J, Hobbs H. Molecular definition of the extreme size polymorphism in apolipoprotein(a). Hum Molec Genet 1993;2:933-940.

[14] Marcovina S, Zhang Z, Gaur V, Albers J. Identification of 34 apolipoprotein(a) isoforms: differential expression of apolipoprotein(a) alleles between American blacks and whites. Biochem Biophys Res Commun 1993;191:1192-1196.

[15] Marcovina SM, Albers JJ, Gabel B, Koschinsky L, Gaur VP. Effect of the number of apolipoprotein(a) kringle 4 domains on immunochemical measurements of lipo- protein(a). Clin Chem 1995;41:246-255.

[16] Lippi G, Lo Cascio C, Ruzzenente O, Poli G, Brentegani C, Guidi GC. Simple and rapid procedure for the purification of lipoprotein(a). J Chromatogr B (in press).

[17] Lippi G, Ruzzenente O, Facchinetti R, Poll G, Schiesari F, Guidi GC. Latex im- munonephelometry for lipopoprotein(a) assay. Eur J Lab Med 1994;2:23-28.

[18] Lowry OH, Rosebrough NJ, Farr AL, Randall RJ. Protein measurement with folin phenol reagent. J Biol Chem 1951;193:265-275.

[19] Marcovina SM, Albers JJ, Jacobs DR, Perkins LL, Lewis CE, Howard BV, Savage P. Lipoprotein(a) concentrations and apolipoprotein(a) phenotypes in Caucasians and African Americans. Arterioscler Thromb 1993; 13:1037-1045.

[20] Passing H, Bablok W. Comparison of several regression procedures for method com- parison studies and determination of sample size. Part II. J Clin Chem Clin Biochem 1984;22:431-435.

[21] Mfirz W, Siekmeier R, Grog, E, Gr Grog, W. Determination of lipoprotein(a): enzyme immunoassay and immunoradiometric assay compared. Clin Chim Acta 1993;214:153- 163.

[22] M/irz W, Siekmeier R, Grog E, Kostner GM. Determination of lipoprotein(a): evalua- tion of three methods. Eur J Clin Chem Clin Biochem 1993;31:295-301.

[23] Leus F, Leerink C, Prins J and van Rijn H. Influence of apolipoprotein(a) phenotype on lipoprotein(a) quantification: evaluation of three methods. Clin Biochem 1994;27:449- 455.

[24] Sundvall J, Sulonen GB, Hiltunen O, Kiuru J, Pursiainen M, Jauhiainen M. Compari- son of a new immunoturbidimetric assay of human serum lipoprotein(a) to the ELISA and the IRMA methods. Scan J Clin Lab Invest 1995;55:171-180.