| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Lipids and Lipoproteins |
1
Department of Chemical Pathology, Princess Alexandra Hospital, Ipswich Rd., Woolloongabba, Queensland 4102, Australia.
2
Department of Laboratory Medicine, Children's Hospital
and Harvard Medical School, Boston, MA 02115.
3
Institute of Medical Genetics, University of Oslo and
Ullevål University Hospital, N-0315 Oslo, Norway.
4
Service de Biochimie, Tenon Hospital, F-75970 Paris,
France.
5
Scientific Affairs Chemistry, Dade Behring, Inc.,
D-35001 Marburg, Germany.
6
Institute for Medical Biochemistry, University of Graz,
A-8010 Graz, Austria.
7
Department of Clinical Laboratory, Omiya Medical Center,
Jichi Medical School, Saitama 330-0834, Japan.
8
Center for Internal Medicine, University of Marburg,
D-35001 Marburg, Germany.
a Author for correspondence. Fax 61-7-3240-7070; e-mail tatej{at}health.qld.gov.au.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The inadequate standardization of Lp(a) assays has been clearly shown in a number of surveys (7)(8)(9)(10), and the large between-method discrepancies in Lp(a) values have been attributed mainly to the use of different standard materials, which vary in isoform composition. A lack of assay precision and optimization has also contributed to poor performance by laboratories using commercially available test kit assays (7)(9)(10)(11)(12). Additional surveys in which Lp(a) assay systems were calibrated using a common reference material (either lyophilized, liquid-stable, fresh, or frozen sera) have shown a reduction in coefficients of variation within and between methods (8)(13)(14)(15). Despite this improved comparability, the complete harmonization of Lp(a) values is more difficult to achieve. Method- and sample-dependent variations exist that may largely reflect antibody-related differences in immunoreactivity to the various apo(a) isoform sizes, with the difference in isoform composition between sample and calibrator contributing to either the over- or underestimation of Lp(a) values (16)(17)(18).
Improved measurement of Lp(a) will depend on the use of internationally accepted reference materials for Lp(a), which will enhance uniformity of values between clinical laboratories and enable the establishment, possibly worldwide, of valid Lp(a) decision cut-points. Towards these aims, the International Federation of Clinical Chemistry Working Group for the Standardization of Lp(a) Assays [IFCC WG Lp(a)] has initiated a project to select a suitable secondary reference material for Lp(a) that can create closer comparability of Lp(a) values between assay systems, similar to the project for apolipoproteins A-I and B (19). The Lp(a) standardization project will consist of three phases:
(1) assessment of the analytical performance of existing Lp(a) assays and testing of commercial Lp(a) calibrator materials for commutability;
(2) development of a common calibrator for Lp(a) measurement, including stability monitoring, evaluation of assay performance in optimized assays, and choice of a selected Lp(a) method for value assignment; and
(3) testing of a panel of fresh-frozen serum samples for method harmonization and bias estimation after the calibration of test systems with the secondary Lp(a) reference standard.
In Phase 1, 40 test systems were evaluated for analytical performance by testing serum samples for precision, linearity, and parallelism characteristics. In addition, commercial Lp(a) calibrator materials were tested for commutability properties across the range of different methods. We report here the results of Phase 1.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
materials
Samples consisted of freshly-prepared serum, liquid-frozen sera,
and manufactured Lp(a) materials. An in-house serum pool (IHSP) was
freshly prepared by each individual participating laboratory by
combining sera from two or more normotriglyceridemic subjects with
Lp(a) >400 mg/L. Two liquid-frozen Lp(a) serum controls (FSC A and FSC
B) were prepared by Immuno AG (Vienna, Austria), which assigned Lp(a)
concentrations of 101 mg/L and 377 mg/L to FSC A and FSC B by
electroimmunodiffusion (EID), using a secondary standard calibrated
against a purified Lp(a) preparation of known protein and lipid
composition. The eight calibrator materials (proposed reference
materials, PRM 1–8) were provided by seven manufacturers. Four of
these preparations were lyophilized (PRM 1, 2, 4, and 8), three were
liquid (PRM 3, 5, and 6), and one was liquid-frozen (PRM 7); all
consisted of plasma from one or more donors. PRMs were stored and
handled as recommended by the manufacturers. FSC A, FSC B, and PRM 7
were transported on dry ice to laboratories in Europe, the United
States, Japan, and Australia. On arrival, they were stored at
-20 °C until use. Vials of the other lyophilized and liquid Lp(a)
materials were sent on coolant and stored at 4–8 °C. It was
stipulated that the IHSP be stored at 4–8 °C and used within 1 week
of collection. Two laboratories were unable to provide an IHSP.
Lp(a) ASSAY SYSTEMS
Among the 40 test systems evaluated in Phase 1, 7 different Lp(a)
methods were used. These included a Lp(a)-cholesterol assay (CHOL),
dissociation-enhanced ligand fluorescence immunoassays (DELFIAs), EID,
ELISAs, immunonephelometric assays (INAs), immunoturbidimetric assays
(ITAs), and RIA (Table 1
). The ITAs and INAs included latex-enhanced assay systems, and
the INAs included both rate and endpoint systems. Four test systems
(DELFIA and ELISA methods) used the a:B assay format, in which the
polyclonal apo B detection antibody identified the apo B-100 component
of Lp(a). Two ELISA systems and RIA used a monoclonal apo(a) detection
antibody, and the 31 other immunochemical assays used polyclonal
antibodies specific for apo(a). Detection antibodies were stated not to
be cross-reactive to plasminogen or LDL-apo B.
|
Lp(a)-cholesterol, Lp(a) protein, and apo(a) values were converted into Lp(a) lipoprotein values, using appropriate conversion factors in the CHOL, EID, and RIA assays. Lp(a)-cholesterol was converted to Lp(a) according to a regression equation based on method comparison to an ELISA a:B Lp(a) assay (correlation coefficient of 0.98). In the RIA, one unit of apo(a) was determined to approximately equal 0.7 mg Lp(a) by calibration of the Apo(a) RIA® diagnostic test kit against a validated commercial Lp(a) preparation containing S1 and S3 isoforms. For the conversion of Lp(a) protein to Lp(a) lipoprotein, an average value of 3 was used. This factor was obtained from the chemical analysis of purified Lp(a) preparations (20). Purified Lp(a), consisting of 20–24 kringle 4 (K4) domains in apo(a), yielded an average 33% of protein, by weight, which contained 15% carbohydrates.
assay protocol
The experiments in Phase 1 were designed to evaluate existing
Lp(a) assay systems for analytical performance and reliability, using
sera and manufacturers' Lp(a) calibrators. Two vials of each PRM and
controls FSC A and FSC B were brought to room temperature if two
analyses were performed in 1 day; otherwise only one vial of each
material was used. Lyophilized PRM 1, 2, 4, and 8 were reconstituted by
the addition of 1 mL of distilled water to each vial, gently swirled
until the contents had dissolved, and left for an additional 30 min.
All materials were equilibrated on a roller-mixer for 5 min. Contents
of each two identical vials were combined and mixed again before making
dilutions.
PRM 1–8, FSC A, FSC B, and the IHSP were diluted to 20%, 40%, 60%, and 80% in phosphate-buffered saline (0.14 mol/L sodium chloride, 0.01 mol/L phosphate buffer, pH 7.4), using either manual or automated, calibrated pipettes. Materials and their dilutions were analyzed in duplicate in four separate analyses performed over either 2 or 4 days. Freshly prepared dilutions of all materials were required when analyzed on separate days.
Lp(a) assay systems were calibrated according to the manufacturers' established procedures, using the manufacturers' recommended calibrators, each system being recalibrated in the four separate analyses. The manufacturers' controls or appropriate in-house controls were assayed at the beginning and end of each run. Samples of PRM 1–8, FSC A, FSC B, and the IHSP were analyzed in descending then ascending order of dilutions.
analysis of sample composition
FSC A, FSC B, and PRM 1–8 were analyzed for lipid and lipoprotein
composition. Cholesterol and triglycerides were measured enzymatically
using commercial test kits (Boehringer Mannheim GmbH). HDL-cholesterol
was measured after the precipitation of VLDL and LDL with polyethylene
glycol, using a commercial HDL-cholesterol reagent kit (Immuno AG);
LDL-cholesterol was calculated according to the Friedewald equation.
Lipoprotein fractionation was performed by agarose gel electrophoresis
and staining with Fat Red 7B (Titan Gel System, Helena); fractions were
compared with a serum control pattern. Lp(a) isoform composition was
determined from both apo(a) phenotype pattern and the number of K4
domain repeats, with the assays performed separately in two IFCC WG
Lp(a) laboratories using sodium dodecyl sulfate-polyacrylamide gel
electrophoresis (SDS-PAGE) and Western blotting. Lp(a) isoform patterns
of PRM 1–8 were compared with commercial markers of known apo(a)
phenotype (Sanwa Chemical Co.), or K4 repeat number (Immuno AG).
turbidity analysis
PRM 1–8 were examined for turbidity in a nephelometric assay
(Array 360 Protein System, Beckman Instruments Inc., Brea, CA; rabbit
polyclonal Lp(a) antiserum, Dako AS). Each material was tested in the
dilution prescribed for sample analysis, using the recommended diluent
(Apo Diluent, Beckman Instruments), and light scattering was recorded
in the presence and absence of antibody. The measuring ranges of PRM
1–8 were further extended by using a dilution of either 1:2 or 1:3
instead of the usual 1:6 dilution, depending on the Lp(a)
concentration.
data analysis
Data were processed and statistical analysis was performed using
the SAS software (SAS Institute Inc.). Descriptive statistics of the
samples PRM 1–8, FSC A, FSC B, and the IHSP were performed separately
for each assay system and sample at each stage of dilution (20%, 40%,
60%, 80%, and 100%), using reported concentration values. At each
dilution point, the within-assay CV was determined from duplicate
measurements of sample dilutions in four separate analyses (n =
8). Sample imprecision was judged as two or more CVs exceeding a cutoff
value of 15% over the five dilution points. An assay system was
regarded as imprecise if two or more CV values exceeded 15% in the
five stages of dilution of the IHSP.
Testing for linearity by linear regression was performed separately for each assay system and sample. At each stage of dilution the percentage of recovery measured in an assay system was calculated as the mean concentration of eight replicates divided by the theoretical expected value (i.e., the mean concentration at each dilution point within a system was recalculated to the 100% concentration value, and then the five values were averaged). The observed recoveries were regressed against the target dilution values to obtain slope and intercept data. Sample linearity was acceptable if the regression slope was between 0.90 and 1.10 and the intercept was between -10 and 10 (percentage units). An assay system was regarded as optimized for linearity if the recovery of the IHSP was within the slope and intercept regression limits.
In the parallelism study, PRM 1–8 were compared with the frozen serum control FSC B and fresh IHSP within each assay system. The mean Lp(a) concentration at each stage of dilution for each sample and separate assay was entered into a general linear model that tested for parallelism between samples (PROC GLM). Regression estimates of PRM 1–8 were compared with FSC B and the IHSP by one-way ANOVA; contrasts giving a probability of <0.05 were regarded as nonparallel.
The among-assay CV of Lp(a) measurement for each PRM was calculated from the mean Lp(a) concentration measured at 100% dilution in a maximum of 40 assay systems and was a measure of the harmonization effect. Lp(a) values were first corrected for calibrator differences between assay systems, using a conversion factor calculated for each system from the assigned value of FSC B (377 mg/L) divided by the mean Lp(a) concentration of FSC B measured at 100% dilution.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
). FSC A and FSC B had values within the same concentration
ranges as the manufactured Lp(a) materials. Triglyceride concentration
was highest in FSC A, PRM 1, and PRM 7, but a concentration of 1.4 g/L
or less is unlikely to cause interference in INAs and ITAs. In a
separate experiment, turbidity of PRM 1–8 was assessed in a
nephelometric system. A 1:6 predilution of PRM 7, when assayed by
nephelometry in the absence of antibody, gave a reaction scatter equal
to 9% of the scatter measured with antibody. Other PRMs gave a
negligible reaction in the absence of antibody. However, a lesser
dilution (1:2) produced an increased false-positive scatter for PRM 1,
4, and 7 equal to 48%, 21%, and 77%, respectively, of the total
scatter measured by nephelometry in the presence of antibody (data not
shown).
|
Lipoprotein fractionation by electrophoresis and comparison with a
serum control showed serum-like patterns for PRM 1, 2, 3, and 8,
altered mobility of lipoproteins for PRM 4 and 5, and unusual smeared
patterns for PRM 6 and 7. PRM 2, 3, and 8 showed additional bands on
SDS-PAGE and Western blotting (data not shown). The frozen serum
controls FSC A and FSC B did not contain any degradation products.
Lp(a) isoform composition was determined both by apo(a) phenotyping, in
which six major isoforms are identifiable according to the Utermann
classification (21), and by the number of apo(a) K4 domain
repeats, which are known to vary from 12 to 51 (22). The
difference in marker composition between the two systems means that
only an approximate correlation between methods was possible (e.g.,
B 
14 K4 repeats; S2 21) (23). The Lp(a)
materials generally contained more than one Lp(a) isoform, with both
serum controls and calibrators consisting of isoforms of 20–24 K4
repeats (Table 2
). PRM 6 and 7 contained additional isoforms of 17 and
14 repeats, respectively, PRM 2 was a pool of several isoforms, the
predominant one having 21 repeats; and FSC A and FSC B also contained
larger molecular mass isoforms of 42 and 29 K4 repeats, respectively.
evaluation of precision, linearity, and parallelism
Precision, linearity, and parallelism properties of the panel of
sera and commercial Lp(a) calibrators were tested in a maximum of 40
test systems. The analytical performance of each assay system was
assessed from precision and linearity testing of a freshly prepared
IHSP. The manufactured materials were assessed in these same systems to
determine their performance characteristics and commutability
properties.
Precision.
The serum samples FSC A, FSC B, and IHSP were
tested for analytical precision within assay systems over five stages
of dilution from 20% to 100%. At 100% dilution, the CVs in up to 40
test systems were 6.9%, 4.4%, and 3.0% for FSC A, FSC B, and the
IHSP, respectively, with values approximately doubling at 20% dilution
to 17.2%, 8.0%, and 5.8%. The low concentration serum control, FSC
A, failed the precision criteria in ~50% of the assay systems.
However, only 4 of 38 systems were assessed as having inadequate
precision according to the criterion of two or more CVs >15% for the
IHSP diluted from 20% to 100% (Table 3
). The total analytical variation for 40 Lp(a) measurements,
determined from within- and between-run replicates at five dilutions of
the IHSP, ranged from 2% to 27% for assay systems and was <15% in
85% of the systems. Among the method groups, precision was optimal for
the DELFIA and RIA groups, with an assay CV of 4–5%. In the larger
method groups, the CVs ranged from 7% to 27% for ELISAs, 2% to 26%
for INA systems, and 2% to 16% for ITA systems.
|
Calibrator materials PRM 1–8 were also tested for their analytical
reproducibility when diluted from 20% to 100%. Assay CVs ranged from
3.1% to 5.0% at 100% dilution and were highest at 20% dilution,
ranging from 4.4% (PRM 5) to 9.3% (PRM 7; Fig. 1
). On the basis of the number of systems with two or more CVs
>15%, PRM 8 lacked precision in 18% of the test systems (7 of 40),
compared with acceptable precision for PRM 5 in all systems (Table 3
).
Other PRMs lacked precision in 3–8% of the test systems.
|
Linearity.
The linearity of FSC A, FSC B, the IHSP, and PRM
1–8 was evaluated for each assay system over five stages of dilution
from 20% to 100%. FSC A, the low concentration serum control, gave
the worst assay linearity, with only 13 of 38 assay systems linear
(Table 3
). By comparison, FSC B and the IHSP performed better but were
still nonlinear in more than one-third of systems, 16 of these testing
nonlinear for the IHSP. Nonlinearity presented either as higher
recoveries at 80% and 100% dilutions of the IHSP and a lower recovery
at 20% dilution or vice versa, with values being underestimated at
80% and 100% dilution and higher at 20% dilution. This was indicated
by the large deviation of the regression slope, which ranged from 0.71
to 1.33 between systems (Fig. 2
). Regression intercept values were generally within ±10%
limits. Within the larger method groups, the percentages of systems
that tested linear for the IHSP were only 44% and 35% for ELISA and
ITA assays, respectively, compared with 86% for INAs. Of the four a:B
Lp(a) systems, two were nonlinear for the IHSP.
|
As a group, PRM 1–8 tested linear in 67% of test systems, with PRM 5
linear in more systems than any other calibrator material; only seven
systems tested nonlinear (Table 3
). Participating laboratories
commented that the concentrations of PRM 4 and 7 were too low in
immunonephelometric, immunoturbidimetric, and EID systems, with
dilutions giving values below the detection limit. EID lacked
sensitivity at Lp(a) concentrations <100 mg/L, which resulted in a
loss of linearity at 20% dilution for some tested materials. Among the
other method groups, the DELFIA a:a and a:B assays were linear for all
sera and calibrators. RIA was linear for PRM 1, 3, 4, 5, and 8, but
because of possible assay-matrix interference, grossly underestimated
Lp(a) values for PRM 2, 6, and 7 at 100% dilution when compared with
the consensus mean values.
Parallelism.
To test for parallelism properties between serum
and calibrator materials, PRM 1–8 were compared with the serum samples
FSC B and IHSP by identical dilution and Lp(a) measurement. The
less-than-desired results of FSC A in terms of poor precision and
nonlinearity within many assay systems negated its use in the
parallelism study. The percentage of systems in which calibrators
tested parallel was variable and ranged from 49% for PRM 8 to 77% for
PRM 4 when compared with FSC B, and from 50% for PRM 7 to 79% for PRM
5 when compared with the IHSP. As a group, PRM 1–8 were parallel to
FSC B in 61% of the assays and to the group of IHSPs in 66% of the
assays (Table 3
). Not all systems were able to show parallelism between
the IHSP and FSC B. In addition, despite excellent assay precision and
linearity in some systems, PRM 1–8 did not necessarily test parallel
to the serum samples; in turn, some nonlinear assay systems exhibited
parallelism.
Among the method groups, 85% of PRMs tested parallel to FSC B in
DELFIA and RIA assays, 65% in EID, 70% in CHOL, and 48% to 60% in
ELISA, INA, and ITA assays. Examples of parallelism and nonparallelism
of calibrator materials compared with the two serum samples FSC B
and IHSP are shown for a typical immunonephelometric Lp(a) assay system
in Fig. 3
. The parallel and nonparallel regression lines reflect either
the similarity or difference in measurement characteristics of these
materials within the described assay system.
|
among-assay harmonization of Lp(a) VALUES
To test whether the manufactured calibrator materials were able to
give comparable Lp(a) values between assay systems and hence reduce the
large method differences observed for Lp(a) measurement, Lp(a)
concentrations were first corrected for calibration differences by
applying a conversion factor that was calculated for each assay from
the assigned and observed values for FSC B. The variation in conversion
factor values between assay systems was large, from 0.6 to 3.1, and
up to threefold in ELISA and immunoturbidimetric assays and twofold for
other method groups. The mean corrected Lp(a) concentrations of PRM
1–8 for all test systems are given in Table 4
, and the differences in mean values between method groups are
shown for each sample. Extreme values of <100 mg/L and <10 mg/L were
obtained for PRM 2 and 7 when assayed by RIA at 100% dilution. Within
each method group, the range of variation in Lp(a) values differed
between calibrator samples. In INA systems for example, PRM 6 gave a CV
of 10% compared with a CV of 34% for PRM 7.
|
In a total of 40 test systems, the among-assay CV of uncorrected Lp(a)
values ranged from 25% for PRM 8 to 77% for PRM 7 (data not shown).
After adjustment for calibration differences between systems, there was
only little or no improvement in CV, with PRM 5 giving the lowest
value of 22% (Table 4
). However, exclusion of those systems that
were nonlinear and/or imprecise for IHSPs, or which gave unexpectedly
low values for the frozen control FSC B, substantially lowered
among-assay CVs for all calibrators. Among-assay CVs within 18
optimized systems ranged from 16% for PRM 3 and 8 to 35% for PRM 7.
Although numbers within method groups were too small for a meaningful
statistical comparison, it was apparent that values measured in
optimized assays were capable of close agreement between methods (Table 5
). PRM 5, for example, gave mean Lp(a) concentrations of
857, 848, and 847 mg/L when measured by INA, ITA, and a group of
systems consisting of DELFIA, EID, ELISA, and RIA.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The IFCC WG Lp(a), working in collaboration with diagnostic companies, has initiated an international standardization study to investigate and, if possible, select a suitable material to serve as the internationally-certified secondary Lp(a) reference material. Such material must be compatible with and stable in different Lp(a) test systems; exhibit precision, linearity, and parallelism characteristics comparable with serum samples; and most importantly, create closer comparability of Lp(a) values between methods. Only those assay systems with optimal analytical performance and antibodies specific for one single epitope on apo(a) or the apo B-100 component of Lp(a) can measure Lp(a) without any isoform-related bias. However, the majority of commercially available Lp(a) test kits use antibodies that are directed against the apo(a) repetitive epitope. Therefore, Phase 1 of the standardization study evaluated the analytical performance and reliability of Lp(a) methods currently in use worldwide, and at the same time tested manufactured Lp(a) calibrators for their suitability as a possible secondary reference material for Lp(a).
The evaluation samples consisted of a panel of sera and commercially prepared calibrators. The manufactured materials were lyophilized, liquid, or liquid-frozen plasma preparations and contained lipid and lipoprotein concentrations similar to the serum controls. By electrophoretic lipoprotein fractionation, some of the calibrators were observed to be serum-like, whereas others had an altered electrophoretic mobility or were possibly degraded. Ideally a good reference material requires similarity to serum samples. However, the suitability of the manufactured materials as potential reference materials cannot be judged only from the electrophoretic pattern, as indicated by PRM 5, which performed best in the quantitative data assessment but had an electrophoretic mobility different from serum. Lp(a) isoform composition was determined by SDS-PAGE and Western blotting, which also detected the presence of apo(a) fragments in some calibrator materials. These fragments may be products of proteolytic modification of Lp(a) by endogenous proteases present in plasma (24). Apart from smaller Lp(a) particles contained in PRM 2, 6, and 7, the calibrator materials contained intermediate-sized Lp(a) isoforms of between 20 and 24 apo(a) K4 repeats. It has been suggested that an intermediate apo(a) size is the clinically relevant isoform (17)(25). Therefore, it is desirable that a secondary reference material for Lp(a) contains isoforms of <25 K4 repeats, as was the case for all tested calibrators.
To assess the analytical performance of current Lp(a) systems,
precision and linearity criteria were tested using a fresh IHSP.
Precision was determined by the repeated Lp(a) measurement of IHSP
dilutions and was acceptable in 34 of 38 test systems. PRM 1–8 were
tested for precision within the same assay systems over the same range
of dilutions. Except for PRM 8, the group of calibrator materials gave
acceptable precision within >90% of the test systems when diluted
from 20% to 100%. However, precision generally decreased at the lower
end of the measuring range for both serum and calibrator samples (Fig. 1
), confirming the suboptimal performance of methods at low Lp(a)
concentrations.
The linearity of each assay system was evaluated from the recovery of IHSPs diluted from 20% to 100%. Because no target values were available for the IHSP of each system, the theoretical expected value of each was calculated and observed recoveries regressed against target dilution values. Of the 38 systems tested for linearity, only 22 were acceptable. The majority of nonlinear assays were from ELISA and ITA method groups, with ~60% of these assays deviating from linearity at either the lower or higher end of the Lp(a) measuring range of their system. Therefore, current test systems may over- or underestimate Lp(a) concentrations not only because of an isoform bias but because of nonlinear methods.
The group of calibrators was tested for linearity in the same way as for IHSPs over the same range of dilutions. Similar to the results for IHSPs, PRM 1–8 were nonlinear in approximately one-third of assay systems. However, it was observed that PRM 5 was linear in more systems than the IHSP. The consensus Lp(a) concentration for PRM 5 of 850 mg/L was greater than the Lp(a) value of most IHSPs. Therefore, Lp(a) assays probably do better analytically at higher concentrations than lower ones. Although four other calibrators had similar values to PRM 5, they performed only slightly better or the same as the IHSPs in linearity experiments, suggesting that an additional matrix interference problem was present for these materials.
An additional important characteristic of Lp(a) measurement is the parallelism property of an assay. Calibrator and serum samples should behave in a like (i.e., parallel) manner within an assay system in order that parallelism between the different types of materials is valid. Calibrator materials were tested for parallelism to the freshly prepared IHSP and to the frozen serum control FSC B by the comparison of Lp(a) measurement for identical sample dilutions. PRM 5 was parallel in ~75% of systems compared with PRM 7 and 8, which were parallel in only 50% of systems and showed the least parallelism. Comments from participating laboratories about unstable reactions, high blank signals, and the instability of PRM 7 and 8, particularly in INA and ITA systems, offers a possible explanation for the poorer parallelism characteristics of these calibrator materials.
Whereas nonparallelism between calibrator and both the IHSP and FSC B is suggestive of a possible matrix-reagent incompatibility within a system, the lack of parallelism to either the IHSP or FSC B may indicate an isoform-related bias. Non-equimolar affinity of different sized Lp(a) isoforms for the assay antibody in apo(a)-specific systems may cause either falsely high or low Lp(a) values; the amount of bias is dependent on the isoform difference between samples. Interestingly however, 71% of optimized systems (linear and precise for the IHSP) tested parallel between the serum samples, IHSP and FSC B, despite probable isoform differences for the group of IHSPs and FSC B. Of these systems, only RIA used an apo(a) detection antibody not directed against the K4–2 epitope on apo(a); except for a:B assays, all other tested systems used polyclonal or monoclonal antibodies, which are considered apo(a) size-sensitive. However, the parallelism results for these systems indicate that there was negligible isoform-related concentration bias for samples containing isoforms from 20 to 24 apo(a) K4 repeats in size. Of the group of calibrators, PRM 3 and 5 tested parallel in 77% and 82%, respectively, of optimized systems. Therefore, although some commercial Lp(a) calibrators appear not to be commutable between different assays, others are compatible within the majority of systems and show similar reaction characteristics to serum in addition to specified precision and linearity requirements.
The results of the analytical assessment of Lp(a) assays have shown that ~40% of existing test systems, including both manufacturers' test kits and research assays, are unable to produce a linear dose–response curve for a pooled serum of medium-to-high Lp(a) concentration, and ~50% of the assays lack adequate precision and/or linearity. In particular, the failure of some systems to measure Lp(a) at low concentrations (<100 to 150 mg/L) indicates the need for Lp(a) test kits to include a statement of the assay detection limit and the precision at this value. In addition, assay reagent formulations should be optimized to enable better linearity of both standard and serum samples and at the same time produce parallelism between these materials. Of the Lp(a) assays tested, it appears that there is no one particular method type that is better or worse than another; acceptable analytical precision, linearity, and parallelism were shown by individual test systems from all method groups. In addition, none of the manufacturers' test systems performed identically in the hands of all users. The decision as to which of the current assays are most accurate and least biased towards Lp(a) isoform size will be determined in Phase 3 of the standardization project, when fresh-frozen serum samples with Lp(a) values assigned against the selected method are analyzed.
From the results of Phase 1, it seems highly likely that the suboptimal analytical performance of a large number of existing Lp(a) assay systems has contributed to the poor comparability of values between methods to a greater extent than recognized in previous Lp(a) surveys. Consequently, it is not surprising that, even after correction of Lp(a) values for calibrator differences between systems, there was not a closer comparison of Lp(a) values for the 40 test systems. In some assay systems, unexpectedly low values for the frozen serum control FSC B, against which all values were referenced, have partly contributed to poorer harmonization results. However, when nonoptimized systems were excluded from the analysis for harmonization, Lp(a) values compared more favorably after uniform calibration, and among-assay variation was reduced to 16%. This closer comparability of values further reinforces the requirement for Lp(a) assays to meet specified precision and linearity performance standards. The final, unequivocal confirmation of successful Lp(a) harmonization will take place in Phase 3 of the standardization project, when a series of samples of varying Lp(a) concentration and isoform composition are analyzed in assay systems that have been optimized for analytical performance and uniformly calibrated against the selected secondary reference material for Lp(a).
The evaluation of manufactured calibrator materials in existing Lp(a) test systems has shown some calibrators to be comparable with serum samples in terms of precision, linearity, and parallelism assay characteristics. Results from optimized assays indicated that Lp(a) values agreed closely between methods for some calibrators but deviated widely for others, indicating their lack of commutability across all Lp(a) assay systems. Therefore, any future selection of a manufactured material as a reference material for Lp(a) must consider its analytical performance, harmonization effect, commutability, and stability properties. Most importantly, monitoring of the long-term stability of such a material is essential because sample degradation leading to changes in recognition of the apo(a) epitope by the antibody may affect Lp(a) measurement in some assay systems. In combination with these requirements is the prerequisite that proposed reference materials be quantitatively tested in optimized assay systems. The development of a common calibrator and choice of a selected Lp(a) method for value assignment are currently in progress in Phase 2 of the standardization project.
In summary, an evaluation of 40 Lp(a) test systems was undertaken to determine the analytical performance of existing Lp(a) assays in use worldwide. From the Phase 1 results, it is apparent that a significant number of the assays were not optimized (i.e., did not meet the stated criteria of precision, linearity, and parallelism) and will not be helped or improved by any standardization effort. More work on the part of the manufacturers is required to address these concerns. The IFCC WG Lp(a) recommends that a similar evaluation consisting of method assessment and comparison of samples be done when selecting an Lp(a) method to be used in large clinical trials or epidemiological studies. Finally, the desired harmonization of Lp(a) assays will require the use of a universal standard, and results presented here indicate the possibility of commercial Lp(a) materials being a satisfactory source of secondary reference material for Lp(a).
|
Acknowledgments |
|---|
Participating laboratories:
WG Lp(a): Institute of Medical Genetics, University of Oslo, Norway; Service de Biochimie, Tenon Hospital, Paris, France; Medical Biochemistry, University of Graz, Austria; Laboratory Medicine, Children's Hospital, Boston, USA; Clinical Laboratory, Omiya Medical Center, Jichi Medical School, Saitama, Japan; and Chemical Pathology, Princess Alexandra Hospital, Brisbane, Australia.
Diagnostic Companies: Beckman Instruments Inc., Brea, CA, USA; Behringwerke AG, Marburg, Germany; bioMérieux, L'Etoile, France; Biopool AB, Umea, Sweden; Boehringer Mannheim GmbH, Tutzing, Germany; Cosmo Research Institute, Tokyo, Japan; Daiichi Pure Chemicals Co. Ltd., Tokyo, Japan; Dako A/S, Glostrup, Denmark; Denka Seiken Co. Ltd., Tokyo, Japan; DiaSys, Holzheim, Germany; Genzyme Corp., Cambridge, MA, USA; F.Hoffmann-La Roche AG, Basel, Switzerland; Immuno AG, Vienna, Austria; Immuno GmbH, Heidelberg, Germany; Incstar Corp., Stillwater, OK, USA (represented by the ROGOSIN Institute); Innogenetics N.V., Zwijnaarde, Belgium; International Enzymes Inc., Fallbrook, CA, USA; Kaketsuken, Kumamoto, Japan; Medical & Biological Laboratories, Co. Ltd., Ina City, Japan; Mercodia AB, Uppsala, Sweden; Nitto Boseki Co. Ltd., Tokyo, Japan; Organon Teknika (PerImmune Inc.), Rockville, MD, USA; Orion Diagnostica, Espoo, Finland; Sanwa Kagaku Kenkyusho Co. Ltd., Nagoya, Japan; Shima Co. Ltd., Tokyo, Japan; Terumo Corp., Kanagawa, Japan
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  The Emerging Risk Factors Collaboration Lipoprotein(a) Concentration and the Risk of Coronary Heart Disease, Stroke, and Nonvascular Mortality JAMA, July 22, 2009; 302(4): 412 - 423. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. R. Kamstrup, A. Tybjaerg-Hansen, R. Steffensen, and B. G. Nordestgaard Genetically Elevated Lipoprotein(a) and Increased Risk of Myocardial Infarction JAMA, June 10, 2009; 301(22): 2331 - 2339. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Bennet, E. Di Angelantonio, S. Erqou, G. Eiriksdottir, G. Sigurdsson, M. Woodward, A. Rumley, G. D. O. Lowe, J. Danesh, and V. Gudnason Lipoprotein(a) Levels and Risk of Future Coronary Heart Disease: Large-Scale Prospective Data Arch Intern Med, March 24, 2008; 168(6): 598 - 608. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. R. Kamstrup, M. Benn, A. Tybjaerg-Hansen, and B. G. Nordestgaard Extreme Lipoprotein(a) Levels and Risk of Myocardial Infarction in the General Population: The Copenhagen City Heart Study Circulation, January 15, 2008; 117(2): 176 - 184. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N. Rifai, J. Ma, F. M. Sacks, P. M. Ridker, W. J. L. Hernandez, M. J. Stampfer, and S. M. Marcovina Apolipoprotein(a) Size and Lipoprotein(a) Concentration and Future Risk of Angina Pectoris with Evidence of Severe Coronary Atherosclerosis in Men: The Physicians' Health Study Clin. Chem., August 1, 2004; 50(8): 1364 - 1371. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. M. Kimberly, H. W. Vesper, S. P. Caudill, G. R. Cooper, N. Rifai, F. Dati, and G. L. Myers Standardization of Immunoassays for Measurement of High-Sensitivity C-reactive Protein. Phase I: Evaluation of Secondary Reference Materials Clin. Chem., April 1, 2003; 49(4): 611 - 616. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 U. Nowak-Gottl, R. Schobess, K. Kurnik, D. Schwabe, G. Fleischhack, R. Junker ;, W. Korte, J. Greiner, A. Feldges, and W. Riesen Elevated lipoprotein(a) concentration is an independent risk factor of venous thromboembolism Blood, May 1, 2002; 99(9): 3476 - 3478. [Full Text] [PDF]
|
|
|
|
|
|
S. M. Marcovina, J. J. Albers, A. M. Scanu, H. Kennedy, F. Giaculli, K. Berg, R. Couderc, F. Dati, N. Rifai, I. Sakurabayashi, et al. Use of a Reference Material Proposed by the International Federation of Clinical Chemistry and Laboratory Medicine to Evaluate Analytical Methods for the Determination of Plasma Lipoprotein(a) Clin. Chem., December 1, 2000; 46(12): 1956 - 1967. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. M. Kostner, A. Ibovnik, H. Holzer, and H. Grillhofer Preparation of a stable fresh frozen primary lipoprotein[a] (Lp[a]) standard J. Lipid Res., December 1, 1999; 40(12): 2255 - 2263. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
) was determined from the duplicate
measurement of 100% and 20% sample dilutions in four separate
analyses. The distribution of CVs for each PRM is indicated by the 50th
percentile (
) and 90th percentile (—) demarcations.
), INA
(*), ITA (), RIA (
), and CHOL (#). At each dilution
point in a test system, the percentage of Lp(a) recovery was calculated
as the mean concentration of eight replicate measurements divided by
the theoretical expected value (calculated for a system as the average
of the Lp(a) concentration at each dilution point recalculated to
100%).