Clinical Chemistry 44:7
14431451 (1998)
New immunoseparation-based homogeneous assay
for HDL-cholesterol compared with three
homogeneous and two heterogeneous methods
for HDL-cholesterol
Matthias Nauck,* Winfried Marz, and Heinrich Wieland
We evaluated four new commercial methods for HDL-
cholesterol determination. The three completely homo-
geneous assays were an immunoseparation-based (IS)
method from Wako, a polyethylene glycol-modified
enzyme (PEG) method from Boehringer Mannheim, and
a synthetic polymer-based (SP) method from Genzyme.
The fourth method was a new heterogeneous method in
which lipoproteins are removed using dextran sulfate-
coated magnetic beads and Mg21 (MB, Reference Diag-
nostics). We compared these methods with the conven-
tional phosphotungstic acid/MgCl2 precipitation (PTA)
procedure. The homogeneous assays had good intraas-
say imprecision with total CVs <2.3%, whereas the CVs
of the MB assay were <5.9%. Adding HDL to serum to
achieve HDL-cholesterol (HDL-C) concentrations up to
1000 mg/L revealed nearly complete recoveries in the IS,
PEG, and MB assays, whereas the SP assay showed a
lower recovery (;70%). The SP HDL-C apparently in-
creased at increasing LDL-cholesterol and VLDL-tri-
glyceride concentrations, whereas the IS, PEG, and MB
methods were not influenced by LDL-cholesterol up to
6 000 mg/L (MB, 5 000 mg/L) and VLDL-triglycerides up
to 9 000 mg/L. Free fatty acids above ;2 mmol/L pro-
duced falsely high HDL-C in the IS and SP assays, the
error amounting to as much as 50% in some samples. An
intermethod comparison in 291 fresh serum samples
yielded correlation coefficients of at least r 5 0.95 for all
assays, when compared with the PTA procedure. The
slopes and intercepts of the regression lines were 1.05
and 57 (IS), 1.12 and 9.9 (PEG), 1.00 and 39 (SP), and 1.0
and 38 mg/L (MB), respectively. The new assays are
precise and simplify the determination of HDL-C, but in
University Hospital of Freiburg, 79106 Freiburg, Germany.
*Address correspondence to this author at: Department of Medicine,
Division of Clinical Chemistry, Hugstetter Strasse 55, 79106 Freiburg i. Br.,
Germany. Fax 49-761-270 3444; e-mail manauck@mzl200.ukl.uni-freiburg.de.
Received January 21, 1998; revision accepted March 17, 1998.
1443
Lipids and
Lipoproteins
part they lack specificity or are susceptible to interfer-
ences, resulting in discrepancies when compared with
the established PTA procedure.
The positive association of coronary heart disease risk
with total cholesterol and LDL-cholesterol (LDL-C)1 con-
centrations and its negative association with HDL-choles-
terol (HDL-C) concentrations are well established (1 4).
HDL-C concentrations ,350 mg/L are considered as a
cardiovascular risk factor; HDL-C concentrations exceed-
ing 600 mg/L may be protective (3). According to Friede-
wald et al. (5), together with total cholesterol and triglyc-
erides, the determination of HDL-C allows the calculation
of LDL-C. For these reasons, reliable and easy-to-perform
methods are needed to quantify HDL-C.
Most frequently, HDL-C is measured in the superna-
tant after precipitation of apolipoprotein (apo) B-contain-
ing lipoproteins by phosphotungstic acid/MgCl2 (PTA) or
dextran sulfate (6, 7). All of these methods involve both a
precipitation and a centrifugation step, which prevent full
automation. Previously, new procedures for the determi-
nation of HDL-C were published (8 12). They have in
common that they avoid at least the tedious centrifuga-
tion step.
In 1996, we published the results of an evaluation of a
homogeneous HDL-C assay that included polyethylene
glycol (PEG) to complex the non-HDL lipoproteins (re-
agent 1); antibodies against apo B and apo C-III, which
aggregate the non-HDL lipoproteins (reagent 2); enzymes
for cholesterol determination (reagent 3); and guanidine
hydrochloride to stop all reactions (reagent 4) (9). The
performance characteristics of this method were good, but
1
Nonstandard abbreviations: LDL-C, LDL-cholesterol, HDL-C, HDL-cho-
lesterol; apo, apolipoprotein; PTA, phosphotungstic acid/MgCl2 precipitation;
PEG, polyethylene glycol; IS, immunoseparation; SP, synthetic polymer; MB,
dextran sulfate-coated magnetic bead; FFA, free fatty acid; and NCEP, Na-
tional Cholesterol Education Program.
1444 Nauck et al.: HDL-cholesterol methods
the disadvantage of the procedure was that it relied on
four different reagents, which limited its application to a
small number of automatic analyzers.
The aim of this study was to evaluate three homoge-
neous assays [an immunoseparation-based (IS) assay
from Wako, a PEG-based assay from Boehringer Mann-
heim, and a synthetic polymer-based (SP) assay from
Genzyme] and a dextran sulfate-coated magnetic bead-
based (MB) procedure from Reference Diagnostics for the
determination of HDL-C in parallel with the conventional
PTA assay with respect to precision, linearity, interfer-
ences, and ease of handling.
Materials and Methods
samples
Blood was drawn from 291 in- and outpatients of the
University Hospital of Freiburg, including nonfasting
samples. The blood was allowed to clot at room temper-
ature, and serum was obtained by centrifugation at 1500g
for 15 min. All analyses with the homogeneous and the
precipitation assays were performed within 1 day of
blood collection.
lipid measurements
Cholesterol and triglycerides were determined enzymat-
ically with the CHOD-PAP and GPO-PAP methods, re-
spectively. These reagents were purchased from Boehr-
inger Mannheim. All measurements, including the
HDL-C assays, were performed on a Boehringer Mann-
heim/Hitachi automatic analyzer type 911. Control sera
Precinorm L and Precipath L (both from Boehringer
Mannheim) were included in each analytical run. In
addition, the nonesterified fatty acids were determined
with the NEFA C assay, purchased from Wako, on a
Wako 30R automatic analyzer in a subset of 130 samples.
All tests, including the new homogeneous assays, were
performed according to the manufacturers recommenda-
tions and calibrated using the standards included in the
test kits. Reagents of the homogeneous assays were ly-
ophilized and solubilized before use.
is-based homogeneous hdl-c assay from wako
The reagents for the IS assay were obtained from Wako. In
this assay, anti-human apo B antibodies react with non-
HDL particles, i.e., chylomicrons, VLDL, and LDL (re-
agent 1). These complexes do not react with the enzymes
involved in the subsequent enzymatic cholesterol deter-
mination, whereas the cholesterol content of the HDL
particles is measured (reagent 2).
homogeneous hdl-c assay from boehringer
mannheim based on peg-modified enzymes
The reagents for the PEG assay were obtained from
Boehringer Mannheim. At a neutral pH value (pH 7.0)
and in the presence of MgCl2, sulfated a-cyclodextrin and
dextran sulfate form water-soluble complexes with LDL,
VLDL, and chylomicrons (reagent 1), which are not acces-
sible to PEG-modified enzymes (reagent 2). Because HDL
is not complexed, the cholesterol moiety is readily avail-
able for enzymatic quantification using cholesterol ester-
ase and cholesterol oxidase coupled to PEG (8).
sp-based homogeneous hdl-c assay from genzyme
The reagents for the SP assay were obtained from Greiner
Biochemica, Flacht, Germany. Synthetic polymers and
polyanions adsorb to the surface of LDL, VLDL, and
chylomicrons and convert these lipoproteins into dena-
turation-resistant complexes (reagent 1). In a second step,
a detergent destroys the HDL structure, and the choles-
terol content of the HDL particles is determined enzymat-
ically (reagent 2) (11).
mb-based hdl-c assay
All reagents, cups, and tubes for this assay were obtained
from Reference Diagnostics. In a flat bottom cup, 500 mL
of serum was incubated with 100 mL of reagent containing
250 mmol/L Mg21 and MBs. After the cups were mixed
for a short period on a vortex-type mixer and incubated
for 10 min, they were transferred to a magnet-containing
tube, which was then directly placed in the Hitachi
analyzer. The dextran sulfate binds all apo B-containing
particles, which are pulled down in the magnetic field
within 1 min. In severe hypertriglyceridemic samples, this
separation can take up to 30 min. The cholesterol content
of the HDL particles that remain in the supernate can
easily be measured by the CHOD-PAP method. We
applied a modified version of this assay, in which the
accuracy at low cholesterol concentrations has been
claimed to be improved by using a higher sample volume.
The dilution of the sample was accounted for by multi-
plying the results by 1.2 (10).
hdl-c determination by a pta-based method
We used the PTA procedure as the comparison method.
The reagents were purchased from Boehringer Mann-
heim. Two hundred microliters of serum were mixed with
500 mL of the PTA reagent and incubated for at least 5
min. The apo B-containing lipoproteins were sedimented
by centrifugation (5 min), and the cholesterol content was
measured in the supernate with the CHOD-PAP method.
hdl-c determination by a combined
ultracentrifugation and precipitation method
In 130 samples, a combined ultracentrifugation-precipita-
tion assay was used as the comparison method. Essen-
tially, the protocol of the Lipid Research Clinics Program
was followed, with modifications previously described
(13, 14). In brief, VLDL were floated by ultracentrifuga-
tion (density 5 1.006 kg/L), and the LDL was separated
from the HDL in the infranatant by PTA.
 Clinical Chemistry 44, No. 7, 1998 1445

isolation of vldl, ldl, and hdl, and preparation statistical methods

of lipoprotein-deficient serum
VLDL, LDL, and HDL were isolated by sequential ultra-
centrifugation using density ,1.006 kg/L, 1.006 , den-
sity , 1.063 kg/L, and 1.063 , density , 1.21 kg/L as
density limits (15). Lipoprotein-deficient serum was ob-
tained by 2 3 repeated ultracentrifugation at 150 000g for
48 h after adjusting the serum to density 5 1.25 kg/L by
adding solid KBr.
apo a-i
Apo A-I was measured turbidimetrically on the Wako-
30R analyzer with reagents purchased from Greiner Bio-
chemica. The assay was calibrated against the IFCC apo
A-I standard.
linearity
Testing of linearity was performed by adding HDL iso-
lated by ultracentrifugation to serum. To a constant
volume of serum, increasing amounts of HDL were
added. To modify the matrix as minimally as possible, the
samples were brought to equal volumes with lipoprotein-
deficient serum.
interferences
The influences of LDL and VLDL on HDL-C measure-
ments were investigated in experiments in which they
were added to samples. In addition, free fatty acids
(FFAs) and bilirubin were measured in the samples of the
intermethod comparisons. Interference from hemoglobin
was analyzed according to Glick et al. (16).
Precision data were calculated according to recommenda-
tions of the NCCLS/EP5-T protocol (17). Regression anal-
yses were performed using the method of Passing and
Bablok (18). Total error was calculated as the sum of the
systematic error plus random error (19, 20). Systematic
error is calculated from the linear regression equation yc
5 bxc 1 a, where b is the slope of the regression line and
a is the y-axis intercept. At an HDL-C concentration of xc,
systematic error is the absolute value of yc 2 xc. Random
error is 1.963 SD from the run-to-run precision study.
Results
precision
Two commercial control sera with low and medium
HDL-C concentrations and one human serum pool were
used to assess the precision of the HDL-C assays (Table 1).
The within-day imprecision values were determined as
the average of the imprecision obtained each day. The
between-day imprecision values were calculated from the
imprecision of the means obtained each day, which were
then adjusted for the within-day imprecision component.
The total CVs ranged from 1.1% to 2.3% for the three
homogeneous assays, from 3.9% to 5.9% for the MB assay,
and from 3.1% to 3.7% for the PTA assay, respectively.
linearity
The new HDL-C assays were linear up to at least 1000
mg/L, as revealed by increasing amounts of HDL pre-
pared by sequential ultracentrifugation to sera (Fig. 1).
For the SP assay, a slope of ;0.7 was obtained when the

Table 1. Summary of the precision studies in control materials and a human serum pool.
Within-dayb Between-day Totalc
Mean,
Assay Samplea mg/L SD, mg/L CV, % SD, mg/L CV, % SD, mg/L CV, %
IS HSPd 397.8 3.4 0.9 5.5 1.4 6.4 1.6
PNL 493.1 4.2 0.9 5.5 1.1 6.9 1.4
PPL 342.7 2.9 0.8 5.4 1.6 6.1 1.8
PEG HSP 357.9 2.3 0.7 6.0 1.7 6.4 1.8
PNL 468.3 2.4 0.5 4.3 0.9 4.9 1.1
PPL 315.8 1.8 0.6 6.9 2.2 7.1 2.3
SP HSP 388.7 3.5 0.9 5.0 1.3 6.1 1.6
PNL 488.4 4.0 0.8 3.6 0.7 5.4 1.1
PPL 398.4 2.8 0.7 4.3 1.1 5.1 1.3
MB HSP 336.7 7.4 2.2 10.9 3.2 13.2 3.9
PNL 390.2 10.2 2.6 20.5 5.3 22.9 5.9
PPL 251.0 6.1 2.4 12.3 4.9 13.7 5.5
PTA HSP 305.5 5.7 1.9 9.3 3.0 10.9 3.6
PNL 409.2 6.0 1.5 11.3 2.8 12.8 3.1
PPL 261.8 4.3 1.7 8.7 3.3 9.7 3.7
a
Two control sera and one human pool serum were analyzed with each HDL-C assay. Pools were divided into 300-mL aliquots and stored at 4 C until tested. One
aliquot of each pool was assayed in duplicate on each of 21 days.
b
Calculations for within-day and between-day SD were based on 21 samples; for within-day SD, calculations were based on a pair of 21 samples.
c 2
SDTotal 5 SD2W 1 SD2B; where SDW 5 within-day standard deviation, and SDB 5 between-day standard deviation.
d
HSP, human serum pool; PNL, Precinorm L; PPL, Precipath L.
1446 Nauck et al.: HDL-cholesterol methods
Fig. 1. Determination of HDL-C: linearity and analytical recovery.
Increasing amounts of HDL isolated by ultracentrifugation were added to a serum
pool. The measured HDL increased linearly with the amounts of HDL added up to
at least 1000 mg/L. The equations of the regression lines are: IS (f), y 5
0.98x 1 21.3; PEG (), y 5 0.99x 1 19.3; SP (F), y 5 70x 1 14.4; and MB (),
y 5 0.90x 1 23.1.
measured HDL-C was plotted against HDL-C added,
suggesting that the recovery of this assay was ,100%.
specificity
HDL-C concentrations are plotted against added LDL-C
and VLDL-triglycerides, respectively, in Fig. 2.
Cross-reactivity with LDL varied between assays. The
IS assay was not influenced by LDL-C up to 6000 mg/L.
Fig. 2. Specificity of the HDL-C assays.
Cross-reactivity with LDL (A) and cross reactivity with VLDL (B) in the IS (f), PEG (), SP (F), and MB () assays.
The PEG assay showed a minimal increase over the whole
concentration range, but deviations .5% occurred only in
samples with LDL-C above 6000 mg/L. LDL-C increased
the HDL-C values of the SP assay at any concentration.
The MB assay was markedly influenced at LDL-C concen-
trations above 5000 mg/L.
Cross-reactivity with VLDL also varied between as-
says. Triglycerides did not substantially interfere with the
IS and PEG assays up to 9000 mg/L VLDL-triglycerides.
The HDL-C values of the SP assay increased over the
whole concentration range, whereas VLDL-triglycerides
interfered with the MB assay at concentrations of ;5000
mg/L. Both the IS and the MB assays showed an unex-
pected negative bias with increasing VLDL-triglycerides.
intermethod comparisons
As the comparison method, we used the PTA procedure
either in whole serum (n 5 161), or after removal of VLDL
by ultracentrifugation (n 5 130).
A total of 291 fresh sera were analyzed with each of the
new methods and the comparison method. Mean total
cholesterol and total triglycerides were 1690 and 1680
mg/L, respectively. Maximum cholesterol and triglycer-
ide concentrations were 3670 and 6830 mg/L, respec-
tively. HDL-C concentrations ranged between 35 and 906
mg/L, determined with the PTA method.
The correlation coefficients were between 0.949 and
0.987 (Table 2). When we estimated the parameters of the
regression lines according to Passing and Bablok, slopes
relating the homogeneous HDL-C IS and PEG assays
to the PTA assay were significantly higher than 1.00
(P,0.05), whereas the SP and MB methods produced
slopes close to 1.00 (18). All intercepts were significantly
(P,0.05) different from 0.0 mg/L, which produced signif-
 Clinical Chemistry 44, No. 7, 1998 1447

Table 2. Summary of intermethod comparisons.a

Regression line
Correlation New procedure for
Assay coefficient Slope Intercept PTA HDL-C, x# b HDL-C,y
IS 0.980 1.05c 56.8d 336 414e
PEG 0.987 1.12c 9.9d 336 388e
SP 0.949 1.00 39.0d 336 372e
MB 0.987 1.00 37.8d 336 380e
a
Regression analyses were done according to Passing and Bablok (18).
b
Concentrations are given in mg/L.
c
Slopes are significantly different from 1.00.
d
Intercepts are significantly different from 0.00.
e
The means were significantly higher, compared with the PTA method.

icantly (P,0.05) higher HDL-C values for all new meth- as HDL-C concentrations increased (Fig. 3). At HDL-C
ods, compared with the conventional precipitation proce- concentrations ,350 mg/L, some of the IS results showed
dure. The bias plots of the IS and PEG assays revealed a deviation of approximately 2200 mg/L, whereas the
small but systematic deviations, which increased slightly PEG assay produced only one outlier. The bias plot of the

Fig. 3. Bias plots of HDL-C (new assay minus PTA assay) in the IS (A), PEG (B), SP (C), and MB (D) assays.
1448 Nauck et al.: HDL-cholesterol methods

SP assay showed substantial scattering over the entire assays using IS and PEG (16). The SP assays yielded lower
range. The results of the MB assay were negatively biased values, whereas in the MB and the PTA method HDL-C
throughout the entire concentration range, with only a increased markedly when hemoglobin was added (Fig. 5).
small scattering.
The correlation coefficients between the apo A-I mea- Discussion
surements and the HDL-C concentrations of the IS, PEG, As shown in many epidemiological and clinical studies, a
SP, and MB assays were 0.92, 0.93, 0.89, and 0.93, respec- low concentration of HDL-C is an important risk factor for
tively. coronary artery disease (1). Thus far, HDL-C has widely
been determined by methods in which apo B-containing
total error lipoproteins are precipitated with polyanions and biva-
The total errors were calculated at the decision points of lent cations. These methods are precise and agree well
350 mg/L and 600 mg/L, respectively. As shown in Table with ultracentrifugation methods but are tedious in re-
3, the systematic deviations contributed much more to the spect to the centrifugation step involved (26, 27). Recently,
total errors than did the random errors. The highest total several new HDL-C assays have been devised which
errors were found for the IS and PEG assays. havewith exceptionssatisfactory performance charac-
teristics (8 12).
free fatty acids Here we present the results of a comparison study that
Conditions with increased FFAs occur mainly in inpa- included three principally different homogeneous HDL-C
tients receiving intravenous heparin therapy. At concen- assays: the first (IS) uses antibodies for the separation of
trations .2 mmol/L, there is substantial binding of FFAs non-HDL lipoproteins coupled to a conventional enzy-
to lipoproteins (21). This increases the electrophoretic matic cholesterol assay; the second (PEG) uses MgCl2,
mobilities of the lipoproteins and may change their inter- sulfated a-cyclodextrin, dextran sulfate, and PEG-coupled
action with precipitating reagents (2224). The results enzymes; the third (SP) uses synthetic polymers, polyan-
obtained in those serum samples of the intermethod ions, and detergents coupled with a conventional enzy-
comparisons that had FFA concentrations .2.0 mmol/L matic cholesterol assay for the quantification of HDL-C.
(n 5 22) indicated that the IS and SP assay were suscep- The fourth method (MB) is based on the absorption of apo
tible to interference from FFAs, whereas the assays from B-containing lipoproteins to dextran sulfate-coated mag-
PEG and MB were not (Fig. 4). Experiments with samples netic beads. All of these assays were compared with the
enriched in FFA, according to the method of Spector and conventional PTA procedure.
Hoak (25), completely confirmed these findings (data not General advantages of the homogeneous methods are
shown). that they obviate pretreatment of the samples and that
only small sample volumes are needed. These features
interferences allow the fully automated determination of HDL-C even
Hemoglobin at concentrations up to 2 g/L did not inter- in the emergency laboratory, where HDL-C may be useful
fere substantially with the new homogeneous HDL-C in the diagnosis of serious infections, such as malaria, or
to monitor patients after liver transplantation (28, 29).
The National Cholesterol Education Program (NCEP)
Table 3. Total error for the new HDL-C assays. performance goals for 1998 demand that the CV of HDL-C
Assay Systematic error, % Random error, % Total error, % determinations be ,4% at concentrations .420 mg/L
IS (30), a criterion that was met by all investigated HDL-C
350 mg/L 21.2 1.1 22.3 assays. The NCEP precision goal for HDL-C concentra-
HDL-C
tions ,420 mg/L is a SD ,17 mg/L (30). The homoge-
600 mg/L 14.5 1.1 15.6
HDL-C neous assays and the PTA procedure fulfilled these crite-
PEG ria. The MB assay produced a SD .17 mg/L in one of
350 mg/L 14.8 1.4 16.2 three pools examined. Thus, the NCEP precision goals
HDL-C will obviously be met by all homogeneous assays consid-
600 mg/L 13.7 0.8 14.5 ered here at both high and low HDL-C concentrations,
HDL-C whereas the MB assay may need further improvement.
SP With regard to the precision, the differences between the
350 mg/L 11.1 0.8 11.9
HDL-C
three homogeneous HDL-C assays are small and appear
600 mg/L 6.5 0.7 7.2
not clinically relevant. These data agree with previous
HDL-C publications (9 12).
MB The slopes of the experiments with added HDL indi-
350 mg/L 10.8 2.4 13.2 cated nearly complete recoveries with the IS and PEG
HDL-C assay. Our data confirm that the specificity of the PEG
600 mg/L 6.3 4.0 10.3 method is good (10, 12, 30, 31). Furthermore, they indicate
HDL-C
that the antibody concentrations of the IS assay are
 Clinical Chemistry 44, No. 7, 1998
Fig. 4. Interferences of FFAs with the determination of HDL-C (ratio, new assay/PTA assay) in the IS (A), PEG (B), SP (C), and MB (D) assays.
sufficient to complex all non-HDL particles. The specific-
ity of the IS assay is thus at least as good as that of the
homogeneous HDL-C from International Reagents Cor-
Fig. 5. Interferences of hemoglobin with the IS (f), PEG (), SP (F), MB
(), and PTA (E) HDL-C assays, tested according to Glick et al. (16).
1449
poration, Kobe, which also contains antibodies to form
stable aggregates of the non-HDL lipoproteins. The latter
method, however, has the disadvantage of including four
different reagents (8).
On the other hand, only ;70% of the added HDL was
recovered with the SP method. The addition of LDL and
VLDL produced falsely high HDL-C concentrations, indi-
cating limited specificity of this method, as described
previously (11, 12).
The MB assay was specific, but the interferences from
LDL-C and VLDL-triglycerides started at lower concen-
trations than in the IS and PEG assays.
Recently, Demacker et al. (27) showed that the PTA
procedure agrees well with the CDC reference method.
Therefore, and because the PTA procedure is very com-
mon in laboratories, it is well suited as a comparison
method.
We obtained excellent correlation coefficients when we
compared the homogeneous HDL-C assays with the PTA
procedure.
The slope of the IS assay was .1.00, which caused the
positive slope of the bias plot (Fig. 3). This was due to a
1450 Nauck et al.: HDL-cholesterol methods
falsely high specification of the calibrator provided by the
manufacturer, so that the results do not exactly agree with
the PTA procedure.
In a previous premarketing evaluation of the PEG
assay, a positive bias of 3% was observed (31). The
optimistic data of this previous multicenter study were
not reproduced in the current study. The PEG assay
showed a slope significantly (P,0.05) exceeding unity.
This deviation was due to a falsely high declaration of the
calibrator, which may have occurred on large-scale pro-
duction. Finally, a very recently performed evaluation of
the PEG method performed in our laboratory showed that
the positive bias had disappeared, indicating that this
assay is now working well (y 5 0.98x 1 11.9 mg/L; r 5
0.997; n 5 100).
The unsatisfactory results of the SP assay have been
reported by other investigators as well (11, 12). The low
specificity of this assay is hampering the use of this
method in the clinical chemical laboratory, in spite of the
good imprecision and a slope of 1.0 in the intermethod
comparison. The scattering of the bias plots, caused by the
nonspecificity of this method, is not acceptable.
The MB assay showed the best agreement with the
comparison method, including an ideal slope of 1.0. It is
worth noting that all results of the MB assay were slightly
higher than the PTA procedure, caused by a small posi-
tive intercept of the regression line.
The NCEP goals for 1998 demand a total error of ,13%
(30). In the current study, the IS and PEG homogeneous
assays have not met these criteria because of high system-
atic error caused by an incorrect calibration (slopes .1
plus a positive y-intercept in the intermethod compari-
son). The SP assay had the lowest total error because of a
very small systematic error. Obviously, the limited spec-
ificity of the SP assay for HDL is balanced by a suitable
calibrator, which accounts for the fact that LDL-C and
VLDL-C are in part determined as HDL-C. The MB assay
meets the NCEP criteria, the total error amounting to
13.2% at 350 mg/L.
A common denominator of the present study and of
many earlier reports is that homogeneous assays tend to
overestimate HDL-C in the range of ;10%. This raises the
suspicion that homogeneous assays may yield an HDL
fraction that differs in some respect from HDL found in
the supernate of precipitation reactions. Such a systematic
difference between homogeneous and conventional meth-
ods would not be entirely surprising. HDL represents a
polydisperse system of particles markedly differing by
physicochemical characteristic, apolipoprotein composi-
tion, and pathobiochemical importance. On the basis of
their apo E content, two populations of HDL particles can
be distinguished, those without apo E and those with apo
E. The latter accounts for ;10% of the total HDL-C. In the
PTA assay, the apo E-containing HDL particles are pre-
cipitated and therefore not included in the HDL-C (32). In
the homogeneous PEG assay, in contrast, these particles
contribute to HDL-C (33), a finding that may at least in
part account for the bias found between the homogeneous
and the precipitation methods. It is currently not known
how different HDL subfractions behave in the other
homogeneous methods.
What are the consequences of these differences in
practice? The large epidemiological studies, which estab-
lished the role of low HDL-C concentrations as a risk
factor of coronary artery disease, were performed with
precipitation methods adjusted to the CDC reference
method. These studies will probably not be repeated in
the future because they are expensive and the results
would be biased through the widespread use of lipid-
lowering therapies, even in primary prevention. It will
probably not be feasible to change the cutoff point for
HDL-C from 350 (without apo E-containing HDL-C) to
380 mg/L (including apo E-containing HDL-C). More
pragmatic, however, will be adjusting the new methods to
the CDC reference method (31, 34), thereby deliberately
disregarding the fact that the homogeneous assays yield
an HDL fraction that may physicochemically be distinct
from HDL produced by precipitation reactions.
The reagents for the homogeneous HDL-C assay are
approximately threefold more expensive than conven-
tional precipitation reagents. With the exception of the SP
assay, the homogeneous assays evaluated here are conve-
nient to use in routine laboratories. The MB assay, which
also avoids centrifugation, is substantially cheaper than
the homogeneous assays and therefore represents an
alternative for laboratories with a workload of ,50 sam-
ples per day. However, in most laboratories, the higher
reagent costs for the new HDL-C assays will probably be
offset by economizing the laboratory work.
In summary, the homogeneous IS and PEG assays pro-
duce precise and accurate determinations of HDL-C. Even
hypertriglyceridemic samples up to at least 9000 mg/L
revealed unbiased results. The MB assay showed the best
agreement with the PTA procedure, except in of hyper-
triglyceridemic and hemolytic samples. This assay is
cheaper than the homogeneous assays but is not fully
automated. All new HDL-C assays, with the exception of
the SP assay, represent a major improvement in our
methodology to quantify HDL-C and may facilitate the
identification of individuals at increased risk of athero-
sclerosis. Although it was the only homogeneous assay to
meet the NCEP performance goals, the SP assay proved
nonspecific and is therefore not recommended for use in
the clinical laboratory.
We are grateful for the excellent technical assistance of
Gabi Herr, Brigitte Kreisel, and Sibylle Rall. We also thank
Wako, Boehringer Mannheim, Greiner, and Reference
Diagnostics for providing HDL-C assay kits free of
charge.
 Clinical Chemistry 44, No. 7, 1998
References
1. Law MR, Wald NJ, Thompson SG. By how much and how quickly
does reduction in serum cholesterol concentration lower risk of
ischaemic heart disease? Br Med J 1994;308:36772.
2. Wilson PW, Abbott RD, Castelli WP. High density lipoprotein
cholesterol and mortality. The Framingham Heart Study. Arterio-
sclerosis 1988;8:737 41.
3. National Cholesterol Education Program. Second report of the
expert panel on detection, evaluation, and treatment of high blood
cholesterol in adults (adult treatment panel II). Circulation 1994;
89:1330 445.
4. International Task Force for Prevention of Coronary Heart Disease.
Prevention of coronary heart disease: scientific background and
new clinical guidelines. Recommendations of the European Ath-
erosclerosis Society. Nutr Metab Cardiovasc Dis 1992;2:11356.
5. Friedewald WT, Leva RJ, Fredrickson DS. Estimation of the con-
centration of low density lipoprotein cholesterol in plasma, without
use of preparative ultracentrifuge. Clin Chem 1972;18:499 509.
6. Warnick GR, Cheung MC, Albers JJ. Comparison of current meth-
ods for high density lipoprotein cholesterol quantification. Clin
Chem 1979;25:596 604.
7. Demacker PNM, Vos-Janssen HE, Hijmans AGM, vant Laar A,
Janssen AP. Measurement of high-density lipoprotein cholesterol
in serum: comparison of six isolation methods with enzymatic
cholesterol analysis. Clin Chem 1980;26:1780 6.
8. Sugiuchi H, Yoshinori U, Okabe H, Irie T, Uekama K, Kayahara N,
Miyauchi K. Direct measurement of high-density lipoprotein cho-
lesterol in serum with polyethylene glycol-modified enzymes and
a-cyclodextrin. Clin Chem 1995;41:71723.
9. Nauck M, Marz W, Haas B, Wieland H. Homogeneous assay for
direct determination of high-density lipoprotein cholesterol evalu-
ated. Clin Chem 1996;42:424 9.
10. Harris N, Galpachian V, Rifai N. Three routine methods for
measuring high-density lipoprotein cholesterol compared with the
reference method. Clin Chem 1996;42:738 43.
11. Harris N, Galpachian V, Thomas J, Iannotti E, Law T, Rifai N. Three
generations of high-density lipoprotein cholesterol assays com-
pared with ultracentrifugation/dextran sulfate-Mg21 method. Clin
Chem 1997;43:816 23.
12. Huang YC, Kao JT, Tsai KS. Evaluation of two homogeneous
methods for measuring high-density lipoprotein cholesterol. Clin
Chem 1997;43:1048 55.
13. Manual of laboratory operation. Lipid Research Clinics Program.
DHEW No. (NIH) 75 628. Bethesda, MD: National Heart and Lung
Institute, 1974:174.
14. Wanner C, Horl WH, Luley C, Wieland H. Effects of HMG-CoA
reductase inhibitors in hypercholesterolemic patients on hemodi-
alysis. Kidney Int 1991;39:754 9.
15. Havel RJ, Eder HA, Bragdon JH. The distribution and chemical
composition of ultracentrifugally separated lipoproteins in human
plasma. J Clin Invest 1955;34:134553.
16. Glick MR, Ryder KW, Jackson SA. Graphical comparisons of
interferences in clinical chemistry instrumentation. Clin Chem
1986;32:470 5.
17. NCCLS. User evaluation of precision performance of clinical
chemistry devices. Tentative Guideline EP5-T. Villanova, PA: Na-
tional Committee for Clinical Laboratory Standards, 1984.
1451
18. Passing H, Bablok W. Comparison of several regression proce-
dures for method comparison studies and determination of sam-
ple size. Part II. J Clin Chem Biochem 1984;22:431 45.
19. Carey RN, Garber CC. Evaluation of methods. In: Kaplan LA, Pesce
AJ, eds. Clinical chemistry: theory, analysis and correlation, 3rd
ed. St. Louis: CV Mosby, 1996:40223.
20. Westgard JO, Carey RN, Wold S. Criteria for judging precision and
accuracy in method development and evaluation. Clin Chem
1974;20:82533.
21. Nauck M, Marz W, Wieland H. Adding albumin normalizes electro-
phoretic mobility of lipoproteins in sera with high concentrations
of free fatty acids. Clin Chem 1996;42:12835.
22. Armstrong V, Seidel D. Evaluation of a commercial kit for the
determination of LDL cholesterol in serum based on precipitation
of LDL with dextran sulfate. Arztl Lab 1985;31:32533.
23. Nauck M, Friedrich I, Wieland H. Measurement of LDL and VLDL
cholesterol with precipitation techniques. A comparison with the
ultracentrifugation method. Klin Lab 1994;40:16776.
24. Siekmeier R, Marz W, Gross W. Insufficient accuracy and speci-
ficity of polyanion precipitation methods for quantifying low-den-
sity lipoproteins. Clin Chem 1990;36:2109 13.
25. Spector AA, Hoak JC. An improved method for the addition of
long-chain free fatty acid to protein solutions. Anal Biochem
1969;32:297302.
26. Marz W, Gross W. Analysis of plasma lipoproteins by ultracentrif-
ugation in a new fixed angle rotor: evaluation of a phosphotungstic
acid/MgCl2 precipitation and a quantitative lipoprotein electro-
phoresis assay. Clin Chim Acta 1986;160:118.
27. Demacker PNM, Hessels M, Toenhake-Dijkstra H, Baadenhuijsen
H. Precipitation methods for high-density lipoprotein cholesterol
measurement compared, and final evaluation under routine oper-
ating conditions of a method with a low-to-reagent ratio. Clin Chem
1997;43:663 8.
28. Nilsson-Ehle I, Nilsson-Ehle P. Changes in plasma lipoproteins in
acute malaria. J Intern Med 1990;227:1515.
29. Nikkila K. Liver transplantation restores low serum levels of very
low density and high density lipoproteins in end-stage primary
cirrhosis. Ann Med 1992;24:129 36.
30. Warnick GR, Wood PD for the National Cholesterol Education
Program Working Group on Lipoprotein Measurement. National
Cholesterol Education Program recommendations for measure-
ment of high-density lipoprotein cholesterol: executive summary.
Clin Chem 1995;41:142733.
31. Nauck M, Marz W, Jarausch J, Cobbaert C, Sagers A, Bernard D, et al.
Multicenter evaluation of a homogeneous assay for HDL cholesterol
without sample pre-treatment. Clin Chem 1997;43:16229.
32. Okamoto Y, Tsujii H, Haga Y, Tanaka S, Nakano H. Determination
of apolipoprotein E in high-density lipoprotein fraction by immuno-
fixation method and turbidimetric immunoassay after precipita-
tion. J Atheroscler Thromb 1994;1:239.
33. Okamoto Y, Tanaka S, Nakano H. Direct measurement of HDL
cholesterol preferable to precipitation method. Clin Chem 1995;
41:1784.
34. Warnick GR, Wood PD. National Cholesterol Education Program
recommendations for measurement of high-density lipoprotein
cholesterol: executive summary. Clin Chem 1995;41:142733.