Advertisement

Comparison of lipoprotein (a) serum concentrations measured by six commercially available immunoassays

Open AccessPublished:August 27, 2019DOI:https://doi.org/10.1016/j.atherosclerosis.2019.08.015

      Highlights

      • -
        Lipoprotein (a) is an established risk factor for cardiovascular disease
      • -
        The comparison between commercially available assays revealed substantial differences
      • -
        Further efforts to harmonize assays for lipoprotein(a) are needed

      Abstract

      Background and aims

      Lipoprotein (a) [Lp(a)] is an established causal risk factor for cardiovascular disease (CVD), independently of low-density lipoproteins (LDL) and other risk factors. The recognition of Lp(a) as an atherogenic molecule has raised the demand for reliable quantification methods in the clinical laboratory. The aim of this work is to compare commercial immunochemical assays.

      Methods

      We measured Lp(a) serum concentrations using six different assays, providing Lp(a) in mg/dl (Denka Seiken, Abbott Quantia, Beckman, Diasys 21FS, and Siemens N Latex) or in nmol/l (Roche TinaQuant, Diasys 21 FS) in 144 serum samples covering the clinically relevant range of Lp(a) concentrations. All assays relied on five-point calibrations using calibrators provided by the manufacturers. Apolipoprotein(a) phenotyping was performed by sodium dodecyl sulfate-agarose gel electrophoresis (SDS-agarose) followed by immunoblotting.

      Results

      Most bivariate correlation coefficients were greater than 0.90. Compared to an established IFCC-proposed reference material, the results of the different assays diverged from the target values (43.3 mg/dl or 96.6 nmol/l) by −8% (Siemens N Latex) and +22% (Abbott Quantia). Stratification of the samples into five groups with increasing Lp(a) concentrations and difference plots showed that the differences among assays were concentration-dependent. Some assays overestimated Lp(a) at high concentrations compared to the Denka Seiken assay.

      Conclusions

      Current commercial immunological assays for measuring Lp(a) concentrations are differently calibrated. Their biases differ significantly across the clinically relevant concentration range in a non-linear manner. This is not conclusively explained by apolipoprotein (a) phenotypes. Further international efforts to harmonize assays for Lp(a) are needed.

      Graphical abstract

      Keywords

      1. Introduction

      Lipoprotein (a) (Lp(a)) is a causal risk factor for the development of cardiovascular disease (CVD), independent of low-density lipoproteins (LDL) and other risk factors [
      • Nordestgaard B.G.
      • Chapman M.J.
      • Ray K.
      • et al.
      Lipoprotein(a) as a cardiovascular risk factor: current status.
      ,
      • Kostner K.M.
      • Marz W.
      • Kostner G.M.
      When should we measure lipoprotein (a)?.
      ,
      • Piepoli M.F.
      • Hoes A.W.
      • Agewall S.
      • et al.
      2016 european guidelines on cardiovascular disease prevention in clinical practice: the sixth joint task force of the european society of cardiology and other societies on cardiovascular disease prevention in clinical practice (constituted by representatives of 10 societies and by invited experts)Developed with the special contribution of the european association for cardiovascular prevention & rehabilitation (EACPR).
      ]. The risk of developing coronary artery diseases (CAD) in Caucasians is more than two times higher at elevated Lp(a) [
      • Holmer S.R.
      • Hengstenberg C.
      • Kraft H.G.
      • et al.
      Association of polymorphisms of the apolipoprotein(a) gene with lipoprotein(a) levels and myocardial infarction.
      ,
      • Emerging Risk Factors Collaboration
      • Erqou S.
      • Kaptoge S.
      • et al.
      Lipoprotein(a) concentration and the risk of coronary heart disease, stroke, and nonvascular mortality.
      ]. Causality of the relationship between Lp(a) plasma concentration and CVD was first reported in 1992 [
      • Sandholzer C.
      • Saha N.
      • Kark J.D.
      • et al.
      Apo(a) isoforms predict risk for coronary heart disease. A study in six populations.
      ] using the algorithm that was denominated in subsequent studies “Mendelian Randomization” [
      • Kamstrup P.R.
      • Tybjaerg-Hansen A.
      • Steffensen R.
      • et al.
      Genetically elevated lipoprotein(a) and increased risk of myocardial infarction.
      ]. Plasma Lp(a) concentrations vary from less than 0.1 mg/dl to more than 300 mg/dl and their frequency distribution is strongly skewed to the right in the white population. A cut-off concentration of 30 mg/dl has originally been defined above which the risk of myocardial infarction increases [
      • Kostner G.M.
      • Avogaro P.
      • Cazzolato G.
      • et al.
      Lipoprotein Lp(a) and the risk for myocardial infarction.
      ]. A meta-analysis of prospective studies enrolling a total of more than 126.000 individuals [
      • Emerging Risk Factors Collaboration
      • Erqou S.
      • Kaptoge S.
      • et al.
      Lipoprotein(a) concentration and the risk of coronary heart disease, stroke, and nonvascular mortality.
      ] found incidence rates of coronary heart disease in the top and bottom tertiles of Lp(a) of 5.6 (95% CI, 5.4–5.9) and 4.4 (95% CI, 4.2–4.6) per 1000 person-years. This meta-analysis also shows that the relationship between Lp(a) and CHD might not be strictly continuous, in that significant increases of risk were only seen when Lp(a) concentrations exceed 30 mg/dl and graded increases in risk may occur above this threshold. A less rigorous cut-off of 50 mg/dl has also been suggested [
      • Nordestgaard B.G.
      • Chapman M.J.
      • Ray K.
      • et al.
      Lipoprotein(a) as a cardiovascular risk factor: current status.
      ].
      Lp(a) is an LDL-like lipoprotein to which apolipoprotein (a) [apo(a)] is linked through a disulfide bridge. Apo(a) consists of repetitive domains, so called kringles (K), which are highly homologous to kringle four (K-IV) of plasminogen [
      • McLean J.W.
      • Tomlinson J.E.
      • Kuang W.-J.
      • et al.
      cDNA sequence of human apolipoprotein(a) is homologous to plasminogen.
      ]. Apo(a) in addition contains one copy of a plasminogen K-V-like domain and an enzymatically inactive protease. In humans, more than 30 genetically determined apo(a) size isoforms (phenotypes) exist; their sizes are inversely correlated with circulating Lp(a) concentrations accounting for approximately 50% of the inter-individual concentration differences [
      • Utermann G.
      • Menzel H.J.
      • Kraft H.G.
      • et al.
      Lp(a) glycoprotein phenotypes. Inheritance and relation to Lp(a)-lipoprotein concentrations in plasma.
      ], yet numerous additional polymorphisms and mutations with varying functionality have been described [
      • Coassin S.
      • Schonherr S.
      • Weissensteiner H.
      • et al.
      A comprehensive map of single-base polymorphisms in the hypervariable LPA kringle IV type 2 copy number variation region.
      ]. The size isoforms differ by varying numbers of identical copies of K-IV type 2; the smallest apo(a) contains 11, the largest one more than 54 ones [
      • Kronenberg F.
      • Utermann G.
      Lipoprotein(a): resurrected by genetics.
      ]. The sequences of the remaining K-IV motifs differ from K-IV type 2 and among themselves but share a high degree of homology. As a consequence of the variations in the number of KIV-type repeats, the molecular weights of the apo(a) phenotypes (which also includes a considerable degree of glycosylation) range from approx. 300 to more than 800 kD [
      • Utermann G.
      The mysteries of lipoprotein(a).
      ].
      The recognition of Lp(a) as a causal atherogenic molecule has raised the demand for reliable methods of quantification. Lp(a) has mostly been measured with immunochemical methods relying on the detection of its apo(a) moiety. One of the problems in the quantification of Lp(a) arises from the size polymorphism of apo(a) [
      • Marcovina S.M.
      • Albers J.J.
      Lipoprotein (a) measurements for clinical application.
      ]. While the epitopes recognized by the majority of antibodies in commercial assays are unknown, they are likely to bind to epitopes residing both on the repetitive and non-repetitive K domains. Because the K-IV domains share a high degree of homology, apo(a) is expected to contain many ‘repetitive’ epitopes, the number of which varies in parallel to the number of repetitive K-IV type 2 domains in a given individual. Therefore, depending on the assay type and antibody specificity, the size polymorphism of apo(a) may impact the results. For instance, the molar concentration of Lp(a) may be underestimated in individuals with small isoforms and high concentrations, while it may be overestimated in individuals with large isoforms and low concentrations when one uses a single reference standard [
      • Marcovina S.M.
      • Albers J.J.
      • Gabel B.
      • et al.
      Effect of the number of apolipoprotein(a) kringle 4 domains on immunochemical measurements of lipoprotein(a).
      ]. Further methodical issues are: (i) Lp(a) particles additionally vary in molar mass and hydrated density, independently of the apo(a) size polymorphism; (ii) purification of Lp(a) for primary standards is difficult and purified Lp(a) is unstable; and (iii) Lp(a) forms mixed aggregates with LDL that are not always fully dissociable (for a comprehensive review covering this topic, see Lippi et al. [
      • Lippi G.
      • Guidi G.
      Standardization and clinical management of lipoprotein(a) measurements.
      ]).
      To deal with these problems, a Lp(a) standardization working group supported by the WHO and the IFCC was established [
      • Tate J.R.
      • Berg K.
      • Couderc R.
      • et al.
      International federation of clinical chemistry and laboratory medicine (IFCC) standardization project for the measurement of lipoprotein(a). Phase 2: selection and properties of a proposed secondary reference material for lipoprotein(a).
      ,
      • Dati F.
      • Tate J.R.
      • Marcovina S.M.
      • et al.
      First WHO/IFCC international reference reagent for lipoprotein(a) for immunoassay--lp(a) SRM 2B.
      ]. This group found a large between-method variation mainly due to the lack of a common reference material for Lp(a). Although most manufacturers mention in their protocol that their assays were standardized with the IFCC-proposed reference material (PRM)-2 [
      • Tate J.R.
      • Berg K.
      • Couderc R.
      • et al.
      International federation of clinical chemistry and laboratory medicine (IFCC) standardization project for the measurement of lipoprotein(a). Phase 2: selection and properties of a proposed secondary reference material for lipoprotein(a).
      ], vast differences of Lp(a) concentrations reported by different laboratories are regularly encountered in clinical practice. We therefore sought to systematically compare currently used assays and methods in clinical laboratories.

      2. Materials and methods

      2.1 Serum samples

      We selectively collected 144 fasting serum samples from patients attending the Medical University of Graz, in which Lp(a) measurements were requested for clinical reasons and which covered the entire, clinically relevant range of Lp(a) concentrations. Serum was recovered by centrifugation, aliquoted and stored at −20° C at maximum for 90 days. According to the Siemens N Latex assay, which is currently used in our routine laboratory, the Lp(a) concentrations in our samples were distributed as follows: 2.5–10 mg/dl, n = 15; 11–20 mg/dl, n = 15; 21–30 mg/dl, n = 15; 31–40 mg/dl, n = 15; 41–50 mg/dl, n = 15; 51–60 mg/dl, n = 17; 61–70 mg/dl, n = 14; 71–90 mg/dl, n = 15; 91–110 mg/dl, n = 11; >110 mg/dl, n = 13. The study was approved by the ethics review committee of the Medial University of Graz (ek 27-305ex 14/15).

      2.2 Determination of the Lp(a) concentration

      We measured Lp(a) with five immunoturbidimetric assays and one immunonephelometric assay using the conditions shown in Table 1. All assays were conducted as single values in aliquots frozen and thawed once. Because the Abbott Quantia and the Roche Tina Quant assays were not available at the Clinical Institute of Medical and Chemical Laboratory Diagnostics, Graz, we provided frozen aliquots of the samples to the Department of Laboratory Medicine, Konventhospital Barmherzige Brüder, Linz, for running the Abbott Quantia and the Roche TinaQuant assays. Samples were shipped on dry ice.
      Table 1Immunological assays used for the measurement of Lp(a).
      AssayUnitCalibrators (n)MethodAnalyserSite/laboratoryDetection antibodyReference material
      Denka Seikenmg/dl5
      Five different calibrator pools.
      ImmunturbidimetryOlympus AU640CIMCL, Graz
      Clinical Institute of Medical and Chemical Laboratory Diagnostics, Medical University Graz, Graz, Austria.
      Anti-Lp(a) (rabbit), polyclonalInternal reference material not specified
      Abbott Quantiamg/dl5
      Five different calibrator pools.
      ImmunturbidimetryAbbott Architect c16000Lab Med, Linz
      Department of Laboratory Medicine, Konventhospital Barmherzige Brueder, Linz, Austria.
      Anti-Lp(a) (rabbit), polyclonalInternal reference material not specified
      Beckmanmg/dl5
      Five different calibrator pools.
      ImmunturbidimetryOlympus AU640CIMCL, Graz
      Clinical Institute of Medical and Chemical Laboratory Diagnostics, Medical University Graz, Graz, Austria.
      Anti-Lp(a) (rabbit), polyclonalInternal reference material not specified
      DiaSys 21 FS a
      Diasys 21 FS a and Diasys 21 FS b are independent assays, each using five calibrators provided by the manufacturer, the two assays therefore provide different results.
      mg/dl5
      Five different calibrator pools.
      ImmunturbidimetryOlympus AU640CIMCL, Graz
      Clinical Institute of Medical and Chemical Laboratory Diagnostics, Medical University Graz, Graz, Austria.
      Anti-Lp(a) (rabbit), polyclonalSRM2B
      DiaSys 21 FS b
      Diasys 21 FS a and Diasys 21 FS b are independent assays, each using five calibrators provided by the manufacturer, the two assays therefore provide different results.
      nmol/l5
      Five different calibrator pools.
      ImmunturbidimetryOlympus AU640CIMCL, Graz
      Clinical Institute of Medical and Chemical Laboratory Diagnostics, Medical University Graz, Graz, Austria.
      Anti-Lp(a) (rabbit), polyclonalSRM2B
      Roche TinaQuantnmol/l5
      Five different calibrator pools.
      ImmunturbidimetryRoche Cobas c501Lab Med, Linz
      Department of Laboratory Medicine, Konventhospital Barmherzige Brueder, Linz, Austria.
      Anti-Lp(a) (rabbit), polyclonalSRM2B
      Roche TinaQuant converted
      Results from the Roche TinaQant, which are in nmol/l, were converted into mg/dl using Lp(a) (mg/dl) = Lp(a) (nmol/l) * 0,4167 as suggested by the manufacturer of the assay.
      mg/dl5Calculation (nmol/l * 0.4167)
      Siemens N Latexmg/dl5
      Serial dilution of one calibrator pool.
      NephelometryBN II NephelometerCIMCL, Graz
      Clinical Institute of Medical and Chemical Laboratory Diagnostics, Medical University Graz, Graz, Austria.
      Anti-Lp(a) (rabbit), polyclonalInternal reference material not specified
      a Clinical Institute of Medical and Chemical Laboratory Diagnostics, Medical University Graz, Graz, Austria.
      b Department of Laboratory Medicine, Konventhospital Barmherzige Brueder, Linz, Austria.
      c Diasys 21 FS a and Diasys 21 FS b are independent assays, each using five calibrators provided by the manufacturer, the two assays therefore provide different results.
      d Results from the Roche TinaQant, which are in nmol/l, were converted into mg/dl using Lp(a) (mg/dl) = Lp(a) (nmol/l) * 0,4167 as suggested by the manufacturer of the assay.
      e Five different calibrator pools.
      f Serial dilution of one calibrator pool.
      The immunoturbidimetric assays used five-point calibration curves constructed with five lyophilized standards which were reconstituted separately. The immunonephelometric assay (Siemens N Latex) used a five-point calibration curve generated by serial dilutions of a Lp(a) standard, the dilutions were automatically prepared by the analyser. Diasys 21 FS a and Diasys 21 FS b provide Lp(a) calibrator values in mg/dl and in nmol/l, respectively. Both use independent calibrators provided by the manufacturer, and we therefore evaluated the two assays independently. The Roche TinaQuant assay measures Lp(a) in nmol/l; for comparison purposes, we converted these values into mg/dl by multiplication with the factor 0.4167 recommended by the manufacturer and report the results as Roche TinaQuant converted.
      The analyses were conducted on six consecutive working days in each of the participating laboratories and the samples were allocated randomly to these days: day 1, n = 23; day 2, n = 25, day 3, 25; day 4, n = 23; day 5, n = 24; day 6, n = 24. Thereby, we made sure, that samples across the entire range of Lp(a) concentrations were examined on each day. Into each run, we included two control samples (low and high level) the position of which within the individual batches was blinded to the technicians performing the analyses. We also included as a reference material the PRM-1 originally prepared by the IFCC working group. The reference material was prepared from purified Lp(a), consisting of 24 (predominant band) kringle 4 domains [
      • Tate J.R.
      • Berg K.
      • Couderc R.
      • et al.
      International federation of clinical chemistry and laboratory medicine (IFCC) standardization project for the measurement of lipoprotein(a). Phase 2: selection and properties of a proposed secondary reference material for lipoprotein(a).
      ]. A target value of 96.6 nmol/l or 43.3 mg/dl was adopted for this material. PRM-1 was stored since 1999 in lyophilized form at minus 70 °C.The PRM-1 reference samples were measured at a blinded position on days 1 and 4.

      2.3 Determination of apo(a) phenotypes

      Apo(a) phenotyping was performed by sodium dodecyl sulfate-agarose gel electrophoresis (SDS-agarose) of serum under reducing conditions as previously described [
      • Erhart G.
      • Lamina C.
      • Lehtimaki T.
      • et al.
      Genetic factors explain a major fraction of the 50% lower lipoprotein(a) concentrations in Finns.
      ]. All samples were diluted with phosphate buffer to a standardized Lp(a) concentration of approximately 30 ng/μl. Samples with an Lp(a) concentration <15 ng/μl were not diluted, but a higher volume of sample was applied on the gel to obtain an equal apo(a) mass. A mixture of five plasma samples expressing only one apo(a) isoform each (13, 19, 23, 27, and 35 KIV repeats, determined by pulsed-field gel electrophoresis) was applied as apo(a) isoform size standard. SDS-agarose electrophoresis was followed by immunoblotting using the monoclonal anti-apo(a) antibody 1A2 [
      • Dieplinger H.
      • Gruber G.
      • Krasznai K.
      • et al.
      Kringle 4 of human apolipoprotein[a] shares a linear antigenic site with human catalase.
      ] for detection of apo(a) phenotypes. Samples with at least one isoform with less than 23 KIV repeats were considered as low molecular weight (LMW) apo(a); high-molecular weight (HMW) isoforms contained 23 or more KIV repeats.

      2.4 Statistical analysis

      We used least squares regression, Pearson and Spearman correlation coefficients and the non-parametric method of Passing and Bablock [
      • Passing H.
      • Bablok W.
      A new biometrical procedure for testing the equality of measurements from two different analytical methods. Application of linear regression procedures for method comparison studies in clinical chemistry, Part I.
      ] implemented in the tool Analyze-it for Microsoft Excel for bivariate comparisons (https://analyse-it.com). Day to day imprecision was expressed as the coefficient of variation of the two control samples included in each run.

      3. Results

      3.1 Imprecision

      Table 2 shows the means and standard deviations of Lp(a) concentrations obtained for the two control samples analyzed on six consecutive working days. Day-to-day imprecision was in general low, with the exception of the Beckman assay which revealed coefficients of variation of 12 and 13% in the low- and the high-level control sample, respectively.
      Table 2Mean Lp(a) concentrations obtained in two control samples analyzed on six consecutive working days with commercial assays.
      AssayControl 1Control 2
      MeanSDCVDifference (%)
      The entries contain means with standard deviations of two control samples (top rows) analyzed on six consecutive working days and the relative differences of the respective mean value from the mean value of the Denka Seiken assay. For assays providing Lp(a) in nmol/l we did not calculate the differences.
      MeanSDCVDifference (%)
      The entries contain means with standard deviations of two control samples (top rows) analyzed on six consecutive working days and the relative differences of the respective mean value from the mean value of the Denka Seiken assay. For assays providing Lp(a) in nmol/l we did not calculate the differences.
      Denka Seiken (mg/dl)10332.6reference1314.7reference
      Abbott Quantia (mg/dl)12621.6+22170,41.8+31
      Beckman (mg/dl)871113.0−1611112−15
      DiaSys 21 FS a
      Diasys 21 FS a and Diasys 21 FS b are independent assays, each using five calibrators provided by the manufacturer, the two assays therefore provide different results.
      (mg/dl)
      12143.5+17100.11.1−23
      DiaSys 21 FS b
      Diasys 21 FS a and Diasys 21 FS b are independent assays, each using five calibrators provided by the manufacturer, the two assays therefore provide different results.
      (nmol/l)
      237104.22227.2
      Roche TinaQuant (nmol/l)26062.3230.52.2
      Roche TinaQant converted
      Results from the Roche TinaQant, which are in nmol/l, were converted into mg/dl using Lp(a) (mg/dl) = Lp(a) (nmol/l) * 0.4167 as suggested by the manufacturer of the assay.
      (mg/dl)
      1092.52.3+6100.22.2−23
      Siemens N Latex (mg/dl)10754.7+415213+15
      a The entries contain means with standard deviations of two control samples (top rows) analyzed on six consecutive working days and the relative differences of the respective mean value from the mean value of the Denka Seiken assay. For assays providing Lp(a) in nmol/l we did not calculate the differences.
      b Diasys 21 FS a and Diasys 21 FS b are independent assays, each using five calibrators provided by the manufacturer, the two assays therefore provide different results.
      c Results from the Roche TinaQant, which are in nmol/l, were converted into mg/dl using Lp(a) (mg/dl) = Lp(a) (nmol/l) * 0.4167 as suggested by the manufacturer of the assay.

      3.2 Accuracy

      In the high-level control sample (Table 2), mean Lp(a) concentrations decreased in the following order (assays with mg/dl): Abbott Quantia > Diasys 21 FS a > Roche TinaQuant converted > Siemens N Latex > Denka Seiken > Beckman. The mean concentration obtained with Roche TinaQuant (nmol/l) was higher than that of the DiaSys 21 FS b (nmol/l). For the low-level control the ranking was different: Abbott Quantia > Siemens N Latex > Denka Seiken, > Beckman > Diasys 21 FS a > Roche Tina Quant converted.
      We also measured Lp(a) in the PRM-1 reference material (Table 3). The results suggested that the assays were differently calibrated. The deviations from the target values (96.6 nmol/l or 43.3 mg/dl) ranged from −8% (Siemens N Latex) to +22% (Abbott Quantia). The Denka Seiken assay is only slightly affected by the apo(a) size variation [
      • Marcovina S.M.
      • Albers J.J.
      Lipoprotein (a) measurements for clinical application.
      ]. For this reason, we compared all further analyses using this assay as provisional reference.
      Table 3Mean Lp(a) concentrations of the IFCC-PRM-1 reference material with an assumed target value of 43.3 mg/dl or 96.6 nmol/l, observed on two different working days.
      AssayMean of 2 measurements
      The PRM-1 reference material was included in single measurement on days 1 and 4 of our study.
      Target valueDifference (%)
      Denka Seiken (mg/dl)48.543.3+12
      Abbott Quantia (mg/dl)52.743.3+22
      Beckman (mg/dl)51.543.3+19
      DiaSys 21 FS a
      Diasys 21 FS a and Diasys 21 FS b are independent assays, each using five calibrators provided by the manufacturer, the two assays therefore provide different results.
      (mg/dl)
      46.043.3+6
      DiaSys 21 FS b
      Diasys 21 FS a and Diasys 21 FS b are independent assays, each using five calibrators provided by the manufacturer, the two assays therefore provide different results.
      (nmol/l)
      104.796.6+8
      Roche TinaQuant (nmol/l)109.496.6+13
      Roche TinaQant converted
      Results from the Roche TinaQant, which are in nmol/l, were converted into mg/dl using Lp(a) (mg/dl) = Lp(a) (nmol/l) * 0.4167 as suggested by the manufacturer of the assay.
      (mg/dl)
      45.643.3+5
      Siemens N Latex (mg/dl)39.743.3−8
      a The PRM-1 reference material was included in single measurement on days 1 and 4 of our study.
      b Diasys 21 FS a and Diasys 21 FS b are independent assays, each using five calibrators provided by the manufacturer, the two assays therefore provide different results.
      c Results from the Roche TinaQant, which are in nmol/l, were converted into mg/dl using Lp(a) (mg/dl) = Lp(a) (nmol/l) * 0.4167 as suggested by the manufacturer of the assay.

      3.3 Assay comparison

      The bivariate correlation coefficients for pairwise comparisons were greater than 0.90 throughout (Table 4, upper diagonal), with the exception of the correlation between the Siemens N Latex and the Beckman assay (r according to Spearman = 0.854, Table 4). The slopes of the regression lines of the Denka Seiken assay ranged between 0.93 and 1.31 by considering only assays providing Lp(a) in mg/dl (Table 4, lower diagonal, left column); the slopes of the regression lines compared to the Denka Seiken assay decreased in the following order: Abbott Quantia > Siemens N Latex > DiaSys 21 FS a > Roche TinaQuant converted > Beckman. The same ranking was found when we considered the slopes of least squares regression lines (Table 4).
      Table 4Regression according to Passing and Bablock and correlation coefficients according to Spearman and Pearson of immunological assays for measuring Lp(a).
      Table thumbnail fx2
      a The fields above the diagonal contain non-parametric Spearman correlation coefficients (top) and correlation coefficients according to Pearson (bottom). The fields below the diagonal contain the slopes (top) and the intercepts (bottom) of the regression lines, all calculated according to Passing and Bablock [
      • Passing H.
      • Bablok W.
      A new biometrical procedure for testing the equality of measurements from two different analytical methods. Application of linear regression procedures for method comparison studies in clinical chemistry, Part I.
      ]. Regression analyses were calculated using assays in columns as independent variable (X) and the methods in the rows as dependent variable (Y). The dark grey fields contain assays producing results in different units. In these cases, the slopes necessarily depart from unity.
      b Diasys 21 FS a and Diasys 21 FS b are independent assays, each using five calibrators provided by the manufacturer, the two assays therefore provide differenr results.
      c Results from the Roche TinaQant, which are in nmol/l, were converted into mg/dl using Lp(a) (mg/dl) = Lp(a) (nmol/l) * 0.4167 as suggested by the manufacturer of the assay.
      Comparison of the mean and median values in our 144 samples also reflected the different calibrations of the assays. Overall the mean values were in the following order: Abbott Quantia > Siemens N Latex > Diasys 21 FS a > Denka Seiken > Roche TinaQuant converted > Beckman (Table 5).
      Table 5Mean ± standard deviations, medians and interquartile ranges of Lp(a) concentrations obtained with commercial assays of 144 samples.
      The entries contain means with standard deviations (top rows), medians with interquartile ranges (medium rows) and the relative differences of the median and the mean values in percent compared to the median and mean values of the Denka Seiken assay in the complete set of samples and in groups of samples with different Lp(a) concentrations. The stratification of samples according to Lp(a) was conducted using the results of the Siemens N Latex assay at the Clinical Institute of Medical and Chemical Laboratory Diagnostics, Medical University Graz. Relative differences between Roche TinaQuant (nmol/l) and DiaSys 21 FS b (nmol/l) and Denka Seiken (mg/dl) were not calculated.
      Assay/concentration range (mg/dl)0–175≤19.920–39.940–59.960–79.980–175
      N = 1442930322132
      Denka Seiken (mg/dl)43 ± 298 ± 422 ± 846 ± 1456 ± 1282 ± 19
      40 (16–63)8 (4–11)20 (16–26)43 (37–52)55 (49–60)82 (66–101)
      Abbott Quantia (mg/dl)57 ± 4112 ± 529 ± 756 ± 1574 ± 16116 ± 31
      48 (23–80)12 (7–16)27 (24–32)51 (44–66)73 (63–80)110 (101–130)
      +33/+20+50/+49+29/+37+21/+73+31/+32+41/+35
      Beckman (mg/dl)38 ± 298 ± 317 ± 741 ± 1552 ± 1472 ± 28
      34 (14–53)8 (5–10)16 (13–19)37 (29–51)56 (42–63)70 (50–92)
      −12/−15+1/−1−24/−21−12/−16−8/+1−12/−14
      DiaSys 21 FS a
      Diasys 21 FS a and Diasys 21 FS b are independent assays, each using five calibrators provided by the manufacturer, the two assays therefore provide different results.
      (mg/dl)
      44 ± 366 ± 317 ± 943 ± 1556 ± 1292 ± 31
      38 (12–62)6 (4–8)15 (12–19)42 (35–51)56 (48–60)86 (69–110)
      +2/−5−40/−23−40/−26−28/−4−23/+2−14/+6
      DiaSys 21 FS b
      Diasys 21 FS a and Diasys 21 FS b are independent assays, each using five calibrators provided by the manufacturer, the two assays therefore provide different results.
      (nmol/l)
      89 ± 7312 ± 735 ± 1786 ± 34112 ± 28184 ± 66
      78 (26–131)11 (6–17)31 (26–37)79 (65–104)111 (90–118)176 (136–222)
      Roche TinaQuant (nmol/l)100 ± 8313 ± 641 ± 17101 ± 32129 ± 27212 ± 80
      87 (29–146)13 (7–18)36 (29–48)93 (78–124)131 (113–143)192 (165–230)
      Roche TinaQant converted
      Results from the Roche TinaQant, which are in nmol/l, were converted into mg/dl using Lp(a) (mg/dl) = Lp(a) (nmol/l) * 0.4167 as suggested by the manufacturer of the assay.
      (mg/dl)
      42 ± 356 ± 317 ± 742 ± 1354 ± 1188 ± 33
      36 (16–57)6 (3–8)15 (12–20)39 (32–52)55 (47–59)80 (69–96)
      −3/−10−28/−29−22/−25−9/−11−4/−1+8/−2
      Siemens N Latex (mg/dl)54 ± 3810 ± 530 ± 650 ± 668 ± 6111 ± 26
      49 (25–75)11 (5–13)30 (25–36)51 (45–55)68 (62–71)104 (91–128)
      +26/+23+29/+42+35/+49+9/+17+21/+23+36/+27
      a The entries contain means with standard deviations (top rows), medians with interquartile ranges (medium rows) and the relative differences of the median and the mean values in percent compared to the median and mean values of the Denka Seiken assay in the complete set of samples and in groups of samples with different Lp(a) concentrations. The stratification of samples according to Lp(a) was conducted using the results of the Siemens N Latex assay at the Clinical Institute of Medical and Chemical Laboratory Diagnostics, Medical University Graz. Relative differences between Roche TinaQuant (nmol/l) and DiaSys 21 FS b (nmol/l) and Denka Seiken (mg/dl) were not calculated.
      b Diasys 21 FS a and Diasys 21 FS b are independent assays, each using five calibrators provided by the manufacturer, the two assays therefore provide different results.
      c Results from the Roche TinaQant, which are in nmol/l, were converted into mg/dl using Lp(a) (mg/dl) = Lp(a) (nmol/l) * 0.4167 as suggested by the manufacturer of the assay.
      Stratification of the results into five groups with increasing Lp(a) concentrations (obtained originally with the Siemens N Latex assay) indicated that the differences between the assays were concentration-dependent (Table 5). We therefore generated difference plots in comparison to the Denka Seiken assay (Fig. 1A–E). The Abbott Quantia assay positively deviated from Denka Seiken assay, less so at low, but markedly at high concentrations. Compared to the Denka Seiken assay, the Beckman assay slightly underestimated Lp(a) concentrations, whereby the differences scattered randomly and widely at concentrations above 30 mg/dl. The DiaSys 21 FS assay coincided well with the Denka Seiken assay at concentrations up to 80 mg/dl; beyond this value it clearly overestimated Lp(a). The same applied to the Roche TinaQuant converted. The Siemens N Latex assay underestimated Lp(a) below 20 mg/dl, but overestimated Lp(a) above this value in comparison to the Denka Seiken assay. The DiaSys 21 FS b (nmol/l) assay underestimated Lp(a) in relation to the Roche TinaQuant assay (nmol/l) which became particularly apparent at concentrations above 50 nmol/l (Fig. 1F).
      Fig. 1
      Fig. 1Bland-Altman difference plots of Lp(a) assays considering the apo(a) isoform size.
      The samples were stratified into groups with high molecular weight (blue circles) and low molecular weight (red circles) apo(a) isoforms as described in the Materials and Methods section. To obtain Roche TinaQuant converted, the results from the Roche TinaQant assay (in nmol/l) were converted into mg/dl using a factor of 0.4167 as suggested by the manufacturer. HMW, high molecular weight; LMW, low molecular weight.

      3.4 Apo(a) phenotypes

      We examined whether the relative differences between the assays were related to the smallest of the apo(a) isoform(s) expressed in each sample and grouped them in LMW (less or equal than 22 K-IV repeats) or HMW (more than 22 K-IV repeats) apo(a) phenotypes (Table 6, Fig. 1A–F). Table 6 shows that in some cases the relative differences seen in samples with HMW apo(a) were identical to the relative difference in samples with LMW apo(a), in other cases biases were different (e.g. Diasys 21FS or RocheTinaQant converted versus Denka Seiken).
      As expected, the difference plots showed an overlap of samples with HMW apo (a) and LMW apo(a) in the concentration range between 40 and 70 mg/dl (Fig. 1). We could, however, not find discernible shifts of the differences within this concentration range (which would have been anticipated if the differences were related to apo(a) size). An exception was the difference plot of the Siemens N Latex versus the Senka Daiken assay (Fig. 1E) in which the HMW apo(a) samples had a stronger positive bias than the LMW apo(a) samples.
      We also generated difference plots with relative rather than absolute differences again accounting for HMW and LMW apo(a) (Supplementary Fig. 1). Only at very low Lp(a) concentrations did the Abbott Quantia (Supplementary Fig. 1A), the Beckman (Supplementary Fig. 1B), the Diaysys 21 FS a (Supplementary Fig. 1C), and the Roche TinaQant converted (Supplementary Fig. 1D) show positive proportional bias in comparison to higher concentrations. This may, however, be of limited practical significance, because it applies to Lp(a) concentrations far below clinical decision limits. Finally and deviating from the other methods, the Siemens N Latex assay in tendency showed a positive bias across all HMW apo(a) samples (Supplementary Table 1E).

      4. Discussion

      This systematic comparison reveals substantial differences between contemporary commercial assays to measure Lp(a). While the imprecision of all assays was within an acceptable range, the actually obtained concentrations due to calibration differed most likely due to biases. The assays applied in this study produced remarkably different values obtained for the PRM-1 reference standard, which has assigned concentrations of 43.3 mg/dl or 96.6 nmol/l, respectively. In addition, by examining differences plots in detail, we noticed that the differences between the assays were not proportional, but rather varied across the concentration ranges.
      We conclude from these analyses that the apo(a) size variation may not conclusively explain the key finding of our study, namely a very consistent positive bias at high Lp(a) concentrations of several assays (Abbott Quantia, DiaSys 21 FS a, Roche TinaQuant converted, Siemens N Latex) in relation to the Denka Seiken. We therefore speculate that the major “overestimation” of Lp(a) at high Lp(a) results from the use of the five independent reference standards across the concentration range of Lp(a). Evidently, these materials may have been assigned different Lp(a) concentrations by the manufacturers. According to Marcovina and Albers [
      • Marcovina S.M.
      • Albers J.J.
      Lipoprotein (a) measurements for clinical application.
      ], the turbidimetric Denka Seiken assay relies on five independent reference standards, each containing an apo(a) isoform distribution representative for the respective concentration range, i.e. the high-level calibrators contain small isoforms and the low-level calibrators containing large ones. Marcovina and Albers claimed from the results of a comparative study using their isoform-independent sandwich enzyme linked immunoassay as the reference that the turbidimetric Denka Seiken assay was largely independent of the apo (a) size polymorphism, despite the fact that the polyclonal antibody used in the latter assay recognizes a repetitive (K4 type (2) epitope [
      • Marcovina S.M.
      • Albers J.J.
      Lipoprotein (a) measurements for clinical application.
      ]. If this was true, then the turbidimetric signal per molecule of Lp(a) would be lower at high concentrations of Lp(a) and the high-level reference standards should already have been “up-calibrated” to compensate for the isoform dependence of the polyclonal antibody. In the case of those assays yielding a positive bias at higher apparent Lp(a) (Abbott Quantia, DiaSys 21 FS a, Roche TinaQuant converted, Siemens N Latex) such a compensation may have even been overdone in the high-level calibrators. Alternatively, “up-calibration” of the Denka Seiken high level standard may have been insufficient. Because the protocols for the assignment of the calibrators in different assays are not public, this explanation of course needs to remain speculative and, importantly, we believe that none of the methods under examinations should be coined as a reference method.
      Also, beyond the methods examined here, there is until today no widely accepted commercial gold-standard method for the measurement of Lp(a) nor has a consistent standardization procedure been established. Therefore, Our study was not designed to decide which commercial Lp(a) assayis preferable. Our observations much more demonstrate that further intense international efforts to harmonize the of Lp(a) assays amongst providers of reagents are needed. In the nearer future, the measurement of the apo(a) molar concentrations by liquid chromatography – mass spectrometry using targeting appropriate signature peptides might provide another option that could be employed as a reference method to resolve the methodical issues uncovered by the current study. Efforts along this line are currently underway [
      • International Federation of Clinical Chemistry and Laboratory Medicine
      Apolipoproteins by Mass Spectrometry.
      ].
      It has been suggested that substantial reductions of Lp(a) plasma concentrations starting from very high concentrations at baseline would be needed to demonstrate a cardiovascular benefit [
      • Burgess S.
      • Ference B.A.
      • Staley J.R.
      • et al.
      Association of variants with risk of coronary disease and the implications for lipoprotein(a)-lowering therapies: a mendelian randomization analysis.
      ,
      • Lamina C.
      • Kronenberg F.
      • Lp(a)-GWAS-Consortium
      Estimation of the required lipoprotein(a)-lowering therapeutic effect size for reduction in coronary heart disease outcomes: a mendelian randomization analysis.
      ]. For this purpose it appears crucial to measure Lp(a) levels reliably in particular at high levels .
      Table 6Relative differences (in %) of Lp(a) concentrations in 144 samples obtained with commercial assays according to molecular weight of apo(a).
      Table thumbnail fx3
      a The entries contain means with standard deviations (in %), high molecular weight apo(a) (HMW, top row) and low molecular weight apo(a) (LMW, bottom row). The differences were calculated according to the following formula: Difference (%) = [B - A] * 100/[(A + B)/2].
      b Diasys 21 FS a and Diasys 21 FS b are independent assays, each using five calibrators provided by the manufacturer, the two assays therefore provide differenr results.
      c Results from the Roche TinaQant, which are in nmol/l, were converted into mg/dl using Lp(a) (mg/dl) = Lp(a) (nmol/l) * 0.4167 as suggested by the manufacturer of the assay.

      Conflicts of interest

      The authors declared they do not have anything to disclose regarding conflict of interest with respect to this manuscript.

      Author contributions

      HS, GK, WM, TG designed the study; HS, TS, BD, GE, WM collected data; HS, HD, GK, WM, TG analysed and interpreted data; HS, WM performed statistical analysis; HS, TG, WM wrote the manuscript; TS, BD, HD, GK, MH validated, reviewed and edited the manuscript.

      Acknowledgements

      The authors thank Lusik Balayan, Anton Ibovnik, Sabine Paulitsch, and Elisabeth Pozar for technical assistance. We thank the manufacturers for providing the Lp(a) assays for this project without charge.

      Appendix A. Supplementary data

      The following is the Supplementary data to this article:

      References

        • Nordestgaard B.G.
        • Chapman M.J.
        • Ray K.
        • et al.
        Lipoprotein(a) as a cardiovascular risk factor: current status.
        Eur. Heart J. 2010; 31: 2844-2853
        • Kostner K.M.
        • Marz W.
        • Kostner G.M.
        When should we measure lipoprotein (a)?.
        Eur. Heart J. 2013; 34: 3268-3276
        • Piepoli M.F.
        • Hoes A.W.
        • Agewall S.
        • et al.
        2016 european guidelines on cardiovascular disease prevention in clinical practice: the sixth joint task force of the european society of cardiology and other societies on cardiovascular disease prevention in clinical practice (constituted by representatives of 10 societies and by invited experts)Developed with the special contribution of the european association for cardiovascular prevention & rehabilitation (EACPR).
        Eur. Heart J. 2016; 37: 2315-2381
        • Holmer S.R.
        • Hengstenberg C.
        • Kraft H.G.
        • et al.
        Association of polymorphisms of the apolipoprotein(a) gene with lipoprotein(a) levels and myocardial infarction.
        Circulation. 2003; 107: 696-701
        • Emerging Risk Factors Collaboration
        • Erqou S.
        • Kaptoge S.
        • et al.
        Lipoprotein(a) concentration and the risk of coronary heart disease, stroke, and nonvascular mortality.
        J. Am. Med. Assoc. 2009; 302: 412-423
        • Sandholzer C.
        • Saha N.
        • Kark J.D.
        • et al.
        Apo(a) isoforms predict risk for coronary heart disease. A study in six populations.
        Arterioscler. Thromb. 1992; 12: 1214-1226
        • Kamstrup P.R.
        • Tybjaerg-Hansen A.
        • Steffensen R.
        • et al.
        Genetically elevated lipoprotein(a) and increased risk of myocardial infarction.
        J. Am. Med. Assoc. 2009; 301: 2331-2339
        • Kostner G.M.
        • Avogaro P.
        • Cazzolato G.
        • et al.
        Lipoprotein Lp(a) and the risk for myocardial infarction.
        Atherosclerosis. 1981; 38: 51-61
        • McLean J.W.
        • Tomlinson J.E.
        • Kuang W.-J.
        • et al.
        cDNA sequence of human apolipoprotein(a) is homologous to plasminogen.
        Nature. 1987; 330: 132-137
        • Utermann G.
        • Menzel H.J.
        • Kraft H.G.
        • et al.
        Lp(a) glycoprotein phenotypes. Inheritance and relation to Lp(a)-lipoprotein concentrations in plasma.
        J. Clin. Investig. 1987; 80: 458-465
        • Coassin S.
        • Schonherr S.
        • Weissensteiner H.
        • et al.
        A comprehensive map of single-base polymorphisms in the hypervariable LPA kringle IV type 2 copy number variation region.
        J. Lipid Res. 2019; 60: 186-199
        • Kronenberg F.
        • Utermann G.
        Lipoprotein(a): resurrected by genetics.
        J. Intern. Med. 2013; 273: 6-30
        • Utermann G.
        The mysteries of lipoprotein(a).
        Science. 1989; 246: 904-910
        • Marcovina S.M.
        • Albers J.J.
        Lipoprotein (a) measurements for clinical application.
        J. Lipid Res. 2016; 57: 526-537
        • Marcovina S.M.
        • Albers J.J.
        • Gabel B.
        • et al.
        Effect of the number of apolipoprotein(a) kringle 4 domains on immunochemical measurements of lipoprotein(a).
        Clin. Chem. 1995; 41: 246-255
        • Lippi G.
        • Guidi G.
        Standardization and clinical management of lipoprotein(a) measurements.
        Clin. Chem. Lab. Med. 1998; 36: 5-16
        • Tate J.R.
        • Berg K.
        • Couderc R.
        • et al.
        International federation of clinical chemistry and laboratory medicine (IFCC) standardization project for the measurement of lipoprotein(a). Phase 2: selection and properties of a proposed secondary reference material for lipoprotein(a).
        Clin. Chem. Lab. Med. 1999; 37: 949-958
        • Dati F.
        • Tate J.R.
        • Marcovina S.M.
        • et al.
        First WHO/IFCC international reference reagent for lipoprotein(a) for immunoassay--lp(a) SRM 2B.
        Clin. Chem. Lab. Med. 2004; 42: 670-676
        • Erhart G.
        • Lamina C.
        • Lehtimaki T.
        • et al.
        Genetic factors explain a major fraction of the 50% lower lipoprotein(a) concentrations in Finns.
        Arterioscler. Thromb. Vasc. Biol. 2018; 38: 1230-1241
        • Dieplinger H.
        • Gruber G.
        • Krasznai K.
        • et al.
        Kringle 4 of human apolipoprotein[a] shares a linear antigenic site with human catalase.
        J. Lipid Res. 1995; 36: 813-822
        • Passing H.
        • Bablok W.
        A new biometrical procedure for testing the equality of measurements from two different analytical methods. Application of linear regression procedures for method comparison studies in clinical chemistry, Part I.
        J. Clin. Chem. Clin. Biochem. 1983; 21: 709-720
        • International Federation of Clinical Chemistry and Laboratory Medicine
        Apolipoproteins by Mass Spectrometry.
        WG-APO MS, 2019
        Date accessed: June 18, 2019
        • Burgess S.
        • Ference B.A.
        • Staley J.R.
        • et al.
        Association of variants with risk of coronary disease and the implications for lipoprotein(a)-lowering therapies: a mendelian randomization analysis.
        JAMA Cardiol. 2018; 3: 619-627
        • Lamina C.
        • Kronenberg F.
        • Lp(a)-GWAS-Consortium
        Estimation of the required lipoprotein(a)-lowering therapeutic effect size for reduction in coronary heart disease outcomes: a mendelian randomization analysis.
        JAMA Cardiol. 2019; 4 (Apr 24. [Epub ahead of print]): 575-579

      Linked Article

      • The challenges of measuring Lp(a): A fight against Hydra?
        AtherosclerosisVol. 289
        • Preview
          The Hydra of Lerna is a serpentine water monster in Greek mythology. The Hydra possessed many heads that had an enormous regenerative capacity: whenever a head was cut off, the Hydra would regrow two heads. Thus, it was not easy for the ancient hero Herakles to win against this constantly regenerating problem.
        • Full-Text
        • PDF