Clinical Chemistry 52:9
1722–1727 (2006)
Lipids, Lipoproteins,
and Cardiovascular
Risk Factors
Disparate LDL Phenotypic Classification among 4
Different Methods Assessing LDL Particle
Characteristics
Wayne Ensign,1 Nicole Hill,2 and Christopher B. Heward2*
Background: Our study seeks to clarify the extent of
differences in analytical results, from a clinical perspective, among 4 leading technologies currently used in
clinical reference laboratories for the analysis of LDL
subfractions: gradient gel electrophoresis (GGE), ultracentrifugation–vertical auto profile (VAP), nuclear magnetic resonance (NMR), and tube gel electrophoresis
(TGE).
Methods: We collected 4 simultaneous blood samples
from 40 persons (30 males and 10 females) to determine
LDL subclasses in 4 different clinical reference laboratories using different methods for analysis. LDL subfractions were assessed according to LDL particle size
and the results categorized according to LDL phenotype. We compared results obtained from the different
technologies.
Results: We observed substantial heterogeneity of results and interpretations among the 4 methods. Complete agreement among methods with respect to LDL
subclass phenotyping occurred in only 8% (n ⴝ 3) of the
persons studied. NMR and GGE agreed most frequently
at 70% (n ⴝ 28), whereas VAP matched least often.
Conclusions: As measurement of LDL subclasses becomes increasingly important, standardization of methods is needed. Variation among currently available
methods renders them unreliable and limits their clinical usefulness.
© 2006 American Association for Clinical Chemistry
Cardiovascular disease (CVD)3 is characterized by the
progressive appearance of fibro-fatty plaque (atheroma)
along the vascular endothelium and the coronary arteries
(1 ). Early studies established that increased LDLcholesterol (LDL-C) concentrations accelerated plaque
formation (2 ). However, increased LDL-C concentrations
do not completely predict plaque formation. Chemical
modification of polyunsaturated fats carried in the LDL
particle can greatly enhance its atherogenicity (3–5 ).
Considerable heterogeneity exists in the physical
properties (size and density) of LDL lipoprotein particles among groups of individuals (5 ). Moreover, many
CVD patients do not have increased LDL-C (1, 6 – 8 ).
These ambiguities were partially resolved by the demonstration of a significant association between LDL
subclasses, intermediate-density lipoprotein, increased
plasma triglyceride concentrations, and low plasma
HDL-cholesterol (9 ). Adjustment for triglycerides, however, reduced the relative risk of myocardial infarction
from 3 to 1.5 (95% confidence interval, 0.8 –3.2), indicating
a close association between certain types of LDL particle
and plasma triglyceride concentration. Those investigators (9 ) grouped patients into type A and type B LDL
patterns on the basis of the size/density distribution of
the LDL particle subclasses (10 –15 ).
There are now suggestions that environmental and
metabolic factors can influence atherogenic profiles and
potentially shift a type B LDL pattern toward a type A
LDL pattern (16 –18 ). Thus, clinical use of the LDL profile
as a basis for dietary counseling, lifestyle adjustments,
and the effectiveness of pharmacologic agents in reducing
the risk of CVD could help guide potential treatment
interventions (19 ).
1
Naval Health Research Center, San Diego, CA.
Kronos Science Laboratories, Inc., Phoenix, AZ.
* Address correspondence to this author at: Kronos Science Laboratories,
Inc., 2222 E. Highland Ave., Suite 220, Phoenix, AZ 85016. Fax 602-667-5623;
e-mail [email protected].
Received August 31, 2005; accepted May 4, 2006.
Previously published online at DOI: 10.1373/clinchem.2005.059949
2
3
Nonstandard abbreviations: CVD, cardiovascular disease; LDL-C, LDLcholesterol; GGE, gradient gel electrophoresis; NMR, nuclear magnetic resonance; VAP, ultracentrifugation–vertical auto profile; and TGE, tube gel
electrophoresis.
1722
Clinical Chemistry 52, No. 9, 2006
Analytical ultracentrifugation is considered to be the
benchmark for determining lipoprotein subfractions
(20, 21 ). However, other methodologies have also been
used, including nondenaturing gradient gel electrophoresis (GGE) (20, 22 ), density gradient ultracentrifugation
(23 ), and nuclear magnetic resonance (NMR).
Unfortunately, the estimated number of LDL subfractions is method dependent. Therefore, different methods
may classify a patient’s LDL particle type differently,
making comparisons only tenuous and extrapolation to
patient results difficult. The increasing clinical use of
these newer technologies makes accurate and consistent
patient classification by different laboratories crucial. To
evaluate the performance of these new technologies for
determining LDL subclasses and their effects on patient
classification based on LDL subclass pattern, we conducted this comparison study.
1723
subclasses labeled as LDL1 (most buoyant) through LDL6
(most dense). LDL1 and LDL2 comprise pattern type A
and LDL3 and LDL4 comprise pattern type B (26 ).
method 3: nmr
The laboratory report from Liposcience, Inc., contains 3
sections: a lipoprotein panel, a risk assessment panel, and
a subclass panel. No references are provided for the basis
of the “risk categories”. The LDL size in nanometers and
its classification as type A or B are also reported. The
subclass panel contains the cholesterol content of the
subclasses for intermediate-density lipoprotein and the
triglyceride content for large, intermediate, and small
VLDL (27 ).
method 4: tube gel electrophoresis
Blood samples were collected into BD Vacutainer Tubes
(BD Diagnostics) from 40 apparently healthy persons (30
males) 23 to 61 years of age. Serum and EDTA-plasma
were kept at ambient temperature and separated within
2 h by centrifugation at 1430g for 15 min at 4 °C, aliquoted, and then frozen at ⫺70 °C. Samples from each
person were sent to 4 different laboratories for analysis. A
portion of each serum sample was shipped frozen to
Atherotech (Birmingham, AL), and another portion was
retained frozen at our laboratory (Kronos Science Laboratories, Phoenix, AZ) for LDL subfractionation using the
LipoprintTM system (Quantimetrix). EDTA-plasma samples were shipped frozen for analysis to Liposcience
(Raleigh, NC) and Berkley HeartLab (Alameda, CA). Each
of the laboratories used a different method for LDL
fractionation and LDL particle size determination (9 ).
Quantimetrix Lipoprint LDL System (28 ) was used according to the manufacturer’s specifications at Kronos
Science Laboratories, Inc. In this system, lipoprotein fractions are separated and summed in a weighted manner to
yield an “LDLSF” score. Specific ranges of LDLSF scores
correspond to different LDL phenotype patterns: normal
(type A; LDLSF score ⬍5.5), intermediate (type AB,
LDLSF score 5.5– 8.5), and atherogenic (type B; LDLSF
score ⬎8.5) (28 ).
This technology does not measure the LDL particle size
directly, but estimates LDL particle size by comparing
particle electrophoretic mobility with the electrophoretic
mobilities of particles of known sizes. The total cholesterol
of the sample must be measured independently of the
Lipoprint system. The Lipoprint report contains the patient’s scan and the cholesterol concentration within each
of the fractions based on retardation factor (Rf), which
correspond to each of the LDL subfractions. The scan
contains 7 possible LDL subfractions.
method 1: gge
statistical analysis
To assess LDL subfractions, Berkeley Heartlab uses LDL
Segmented Gradient Gel Electrophoresis (LDL-S3GGETM).
This technique separates LDL particles into 7 LDL subfractions based on particle size and shape (20, 22, 23 ). The
LDL-S3GGE for LDL typing classifications are as follows:
26.35–28.5 nm, large LDL (pattern type A); 25.75–26.34
nm, intermediate (pattern type AB); and 22.0 –25.74 nm,
small LDL (pattern type B). The LDL subfractions are
reported as percentages based on the area under the curve
for each of the 7 subfractions. Small LDL particles correspond to the subfractions LDL IIIa and LDL IIIb. These
subfractions are indicators of the “severity” of the
“atherogenic lipoprotein profile”.
For across-method comparisons, we used a repeatedmeasures ANOVA and pairwise contrasts. Significance
was established with the Bonferroni adjustment. When
the sphericity assumption was in question, the Greenhouse–Geisler adjustment for the degrees of freedom was
used to determine overall statistical significance. We used
the intraclass correlation coefficient to establish the agreement among the within-subject data across laboratory
values. When possible, the distribution of cholesterol
contained in LDL and the estimates for LDL particle sizes
for each laboratory were evaluated by use of boxplots.
The cholesterol, triglyceride, and major lipoprotein cholesterol values were taken from the patient reports obtained from each method. Because this report primarily
focuses on LDL particle characteristics and their use for
LDL phenotype classification, HDL subfractions were not
included in this analysis. All LDL particle sizes were
converted to nanometers for comparisons across methods.
A value of 5 nm was added to the NMR estimates of LDL
particle size per the laboratory’s statement included in the
Materials and Methods
method 2: ultracentrifugation-vertical auto
profile
The Vertical Auto Profile-II, or VAP-II test (Atherotech), is
derived from density gradient ultracentrifugation
(24, 25 ). The VAP-II generates a series of absorbance
curves from which proprietary software produces 6 LDL
1724
Ensign et al.: Disparate LDL Classification among Different Methods
patient report. All statistical analyses were performed
with the Statistical Package for the Social Sciences, Ver. 8.0
(SPSS, Inc.).
Results
The distribution of LDL phenotypes among the 40 persons for each method is shown in Fig. 1. Results varied
considerably among the methods. The tube gel electrophoresis (TGE) classified 30 of 38 (79%) persons as type A
pattern, whereas VAP classified only 3 of 39 (⬃8%) as
type A. The VAP and NMR classified 21 persons (⬃54%)
as type B, but TGE classified only 2 of the 38 persons
(⬃5%) as having a type B pattern. TGE and GGE classified
6 and 5 persons, respectively (n ⫽ 5) as having an
intermediate pattern (type AB), whereas VAP classified
2.5 times as many persons (n ⫽ 15) as type AB. Although
TGE and GGE classified the same number of persons as
type AB, these methods did not agree on the AB classification for the same individuals. The laboratory using the
NMR methodology does not include an intermediate
(type AB) classification.
Patient classification data for each of the 4 methods are
shown in Table 1. Complete agreement among the laboratory methods occurred in 3 (persons 2, 5, and 11) of 39
persons (⬃8%). NMR matched other methods to a greatest extent (marginal totals) and VAP matched the least.
NMR and GGE produced the same result for 28 of
40 (70%) individuals. When we assumed that all AB
patients reported by Berkeley Heart Laboratory (GGE)
could reasonably be reported as pattern A, the agreement
between NMR and GGE improved to 32 of 40 (80%).
Twenty persons (⬃50%) were similarly classified by 3 of
the 4 laboratories. Eight of these 20 individuals (⬃40%)
were classified as type A by the TGE, NMR, and GGE
methods, whereas 10 of 20 (50%) were classified type B by
NMR, VAP, and GGE. Two of 20 persons (10%) were
classified as type B when TGE was 1 of the 3 methods.
The original basis for distinguishing the LDL phenotype among individuals was LDL particle size (9 ). In 3 of
the laboratory methods (TGE, GGE, and NMR), it was
Fig. 1. Histogram showing the number of persons categorized into
each of the 3 LDL pattern types (A, B, or AB) by each of the 4
measurement methods (GGE, VAP, NMR, and TGE).
possible to determine the sizes of the main LDL particle
fraction, providing the basis for LDL phenotypic classification, and in 1 case (GGE), 2 LDL particle sizes were
given in the laboratory report, corresponding to LDL
peaks 1 (GGE1) and 2 (GGE2). Total LDL-C could also
be determined in 3 of the methods. Laboratory comparisons of LDL-C and particle size estimates are indicated in
Fig. 2.
We observed significant differences among methods
for LDL-C concentration (Fig. 2A). Although the absolute
cholesterol concentrations were significantly different
among methods, the within-subject LDL-C was relatively
consistent (intraclass correlation ⫽ 0.904). These results
indicate a degree of bias within each method, but the
relative LDL-C within individuals was maintained across
methodologies. The distribution of LDL particle sizes
among the methods is shown in Fig. 2B. Differences
among the methods for estimated LDL particle sizes were
also significant. Unlike the LDL-C concentration, the
within-subject particle size estimates were not sustained
across methods (intraclass correlation ⫽ 0.246).
Discussion
Several clinical laboratories, including the 2 largest national clinical reference laboratories, now offer LDL subfraction analysis performed with different technologies.
The consistency of patient classification among laboratories is unknown because there is no standardization
among the different methods. To gain insight into the
consistency of LDL phenotype classification, we sent a
sample from each of 40 patients to 4 different laboratories
that use different methods for LDL subfraction analysis.
In this comparison study, we focused on the diagnostic
classification of patient LDL pattern type. Typically, individuals with large LDL particles are classified as type A,
whereas individuals with small, dense LDL particles are
classified as type B and are at a greater risk for CVD
(9, 15 ). Our results indicated only 3 of 39 patients (⬃8%)
were classified as having the same LDL phenotype by
all 4 methods. The best agreement between 2 methods
was between GGE and NMR (70%); this agreement may
actually be ⬎70%, as noted in the Results section. Removal of the 5 samples classified as AB, the 3 samples
classified as not available (NA), and the 1 sample classified as indeterminate (IND) from the GGE report
(Table 1), leaves 31 samples for comparison. The betweenlaboratory agreement for these 31 samples for the GGE
and NMR methods would have been 90% (28 of 31).
The inability to consistently classify patients did not
seem to be dependent on the LDL particle per se, because
the inconsistency in classification was more or less evenly
distributed among the methods used, suggesting that the
classification is method specific. The absolute LDL-C
concentration was significantly different among the compared methods, but the within-subject reproducibility
(intraclass correlation ⫽ 0.904) was consistent across the
methods. The LDL particle sizes were also significantly
1725
Clinical Chemistry 52, No. 9, 2006
Table 1. Shows the patient classification data for each of the 4 laboratories along with results for total cholesterol,
HDL-C, LDL-C, and triglycerides.
Pattern typea
Patient
TGE
NMR
VAP
GGE
TC,b mg/L
HDL-C, mg/L
LDL-C, mg/L
TG, mg/L
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
NA
A
A
A
A
AB
A
A
A
AB
A
AB
A
A
A
A
AB
A
NA
A
A
A
A
A
A
A
A
A
AB
A
B
A
B
A
A
A
A
AB
A
A
B
A
NA
A
A
B
B
A
B
B
A
B
A
B
A
B
B
A
B
B
A
B
B
A
A
A
B
A
B
B
B
A
B
A
A
B
B
A
B
A
AB
A
NA
B
A
B
B
B
AB
B
A
B
AB
AB
AB
AB
B
AB
B
B
AB
B
B
AB
AB
B
B
B
B
AB
B
AB
AB
B
B
B
AB
AB
B
B
B
A
NA
AB
A
B
B
AB
B
B
A
B
A
B
NA
B
B
A
B
B
A
A
A
A
A
NA
IND
A
B
B
A
AB
B
A
A
B
AB
A
B
AB
NA
2090
1720
2230
2000
1860
2290
2110
2260
1890
2170
2090
2150
2320
1980
2530
2070
2250
NA
2610
1640
1870
1250
1900
1170
2330
1360
1740
2460
2000
1240
1970
1820
2000
2040
1770
1950
1810
2830
1900
NA
450
340
450
440
370
340
490
350
320
660
450
500
460
350
340
300
380
NA
430
380
350
200
360
330
630
270
340
460
380
460
380
330
420
340
320
380
460
450
490
NA
1260
1220
1610
1290
1110
1240
1380
1500
1170
1310
1280
1360
1530
1450
1920
1460
1590
NA
1730
1080
1200
870
1320
630
1380
8200
1190
1530
1190
580
1230
710
1280
1390
1100
1220
1170
1800
1200
NA
1400
510
1110
650
1770
3150
990
1370
2590
760
1620
960
1680
960
1410
2250
920
NA
2150
910
1930
890
890
830
1850
840
1090
1510
1430
460
1330
4790
1140
1320
1450
1400
370
2620
490
a
b
Shaded results do not match any other result for that patient.
TC, total cholesterol; HDL-C, HDL-cholesterol; TG, triglycerides; NA, test results not available; IND, indeterminate.
different among the methods, but unlike LDL-C, there
was a poor correlation within individuals across methods.
If the basis for patient classification is LDL particle size
and if the particle size estimates for a given individual are
different depending on the method used, then it is not
surprising that patient classification across laboratories is
not consistent.
As new methods for measuring LDL subclasses are
introduced, standardization becomes increasingly important, particularly when these methods are based on dif-
ferent principles. If LDL particle size is to be meaningfully
incorporated into a patient’s heart disease risk profile,
then a standardization program should be considered. At
the very least, a proficiency testing program for routine
interlaboratory comparison of test results should be implemented. Such a program could help identify and
eliminate inconsistencies among methods. In the absence
of such a program, the clinical utility of current methods
for determining LDL subclasses phenotype is open to
question.
1726
Ensign et al.: Disparate LDL Classification among Different Methods
Fig. 2. Distribution of LDL-C concentration (A) and LDL particle size (B).
(A), boxplots showing the distribution of LDL-C estimates for 3 methods in which LDL-C content was estimated. E indicate outliers. (B), boxplots indicating the
distribution of LDL particle sizes for the main LDL fraction used for patient classification. GGE1 and GGE2 refer to LDL peaks 1 and 2 as indicated in the patient report
for the Berkeley Heart Laboratory.
Although our study demonstrated disparities among
various methods for differentiating LDL particle size,
certain potential limitations of our study cannot be ignored. Our samples were a combination of plasma and
serum (provided as requested by each laboratory), and
there may be a 3% to 7% difference in the concentration of
lipids between plasma and serum (29, 30 ). Additionally,
one cannot completely ignore the physical effects of
freezing and thawing on LDL particle subclasses, but
several studies have demonstrated that freezing and
thawing do not alter LDL phenotypes when GGE or
ultracentrifugation is used for analysis (31–33 ). Whether
this lack of an effect holds true for the other methods must
be determined by further investigation.
References
1. Griffin BA. Lipoprotein atherogenicity: an overview of current
mechanisms. Proc Nutr Soc 1999;58:163–9.
2. Brown MS, Faust JR, Goldstein JL. Role of the low density
lipoprotein receptor in regulating the content of free and esterified
cholesterol in human fibroblasts. J Clin Invest 1975;55:783–93.
3. Goldstein JL, Ho YK, Basu SK, Brown MS. Binding site on
macrophages that mediates uptake and degradation of acetylated
low density lipoprotein, producing massive cholesterol deposition.
Proc Natl Acad Sci U S A 1979;76:333–7.
4. Morel DW, DiCorleto PE, Chisolm GM. Endothelial and smooth
muscle cells alter low density lipoprotein in vitro by free radical
oxidation. Arteriosclerosis 1984;4:357– 64.
5. Henriksen T, Mahoney EM, Steinberg D. Enhanced macrophage
degradation of low density lipoprotein previously incubated with
cultured endothelial cells: recognition by receptors for acetylated
low density lipoproteins. Proc Natl Acad Sci U S A 1981;78:
6499 –503.
6. Musliner TA, Krauss RM. Lipoprotein subspecies and risk of
coronary disease [Review]. Clin Chem 1988;34:B78 – 83.
7. Fruchart JC, Packard CJ. Is cholesterol the major lipoprotein risk
factor in coronary heart disease? A Franco-Scottish overview. Curr
Med Res Opin 1997;13:603–16.
8. Castelli WP. Epidemiology of coronary heart disease: the Framingham study. Am J Med 1984;76:4 –12.
9. Austin MA, Breslow JL, Hennekens CH, Buring JE, Willett WC,
Krauss RM. Low-density lipoprotein subclass patterns and risk of
myocardial infarction. JAMA 1988;260:1917–21.
10. Campos H, Genest JJ Jr, Blijlevens E, McNamara JR, Jenner JL,
Ordovas JM, et al. Low density lipoprotein particle size and
coronary artery disease. Arterioscler Thromb 1992;12:187–95.
11. Austin MA, Hokanson JE, Brunzell JD. Characterization of lowdensity lipoprotein subclasses: methodologic approaches and
clinical relevance. Curr Opin Lipidol 1994;5:395– 403.
12. Frohlich JJ, Seccombe DW, Pritchard PH. The role of apoproteins
in disorders of lipoprotein metabolism. Clin Biochem 1989;22:
51– 6.
13. Slyper AH. Low-density lipoprotein density and atherosclerosis:
unraveling the connection. JAMA 1994;272:305– 8.
14. Krauss RM. Atherogenic lipoprotein phenotype and diet-gene
interactions. J Nutr 2001;131:340S–3S.
15. Krauss RM, Siri PW. Metabolic abnormalities: triglyceride and
low-density lipoprotein. Endocrinol Metab Clin North Am 2004;33:
405–15.
16. Ensign WY. Association of regional and total body composition,
plasma lipids, lipoproteins, apolipoproteins and the effects of
long-term resistance training in premenopausal females [PhD
dissertation], AZ: University of Arizona, 1993.
17. Lamon-Fava S, McNamara JR, Farber HW, Hill NS, Schaefer EJ.
Acute changes in lipid, lipoprotein, apolipoprotein, and low-density
lipoprotein particle size after an endurance triathlon. Metabolism
1989;38:921–5.
18. Reaven GM, Chen YD, Jeppesen J, Maheux P, Krauss RM. Insulin
resistance and hyperinsulinemia in individuals with small, dense
low density lipoprotein particles. J Clin Invest 1993;92:141– 6.
19. Shimabukuro M, Higa N, Asahi T, Oshiro Y, Takasu N. Fluvastatin
improves endothelial dysfunction in overweight postmenopausal
women through small dense low-density lipoprotein reduction.
Metabolism 2004;53:733–9.
20. Krauss RM, Blanche PJ. Detection and quantitation of LDL subfractions. Curr Opin Lipidol 1992;3:377– 83.
21. Krauss RM, Burke DJ. Identification of multiple subclasses of
Clinical Chemistry 52, No. 9, 2006
22.
23.
24.
25.
26.
27.
plasma low density lipoproteins in normal humans. J Lipid Res
1982;23:97–104.
Nichols AV, Krauss RM, Musliner TA. Nondenaturing polyacrylamide gradient gel electrophoresis. Methods Enzymol 1986;128:
417–31.
Lee DM, Downs D. A quick and large-scale density gradient
subfractionation method for low density lipoproteins. J Lipid Res
1982;23:14 –27.
Kulkarni KR, Garber DW, Jones MK, Segrest JP. Identification and
cholesterol quantification of low density lipoprotein subclasses in
young adults by VAP-II methodology. J Lipid Res 1995;36:2291–
302.
Austin MA, King MC, Vranizan KM, Krauss RM. Atherogenic
lipoprotein phenotype: a proposed genetic marker for coronary
heart disease risk. Circulation 1990;82:495–506.
Chung BH, Segrest JP, Ray MJ, Brunzell JD, Hokanson JE, Krauss
RM, et al. Single vertical spin density gradient ultracentrifugation.
Methods Enzymol 1986;128:181–209.
Otvos JD, Jeyarajah EJ, Bennett DW, Krauss RM. Development of
a proton nuclear magnetic resonance spectroscopic method for
determining plasma lipoprotein concentrations and subspecies
distributions from a single, rapid measurement. Clin Chem 1992;
38:1632– 8.
1727
28. Hoefner DM, Hodel SD, O’Brien JF, Branum EL, Sun D, Meissner
I, et al. Development of a rapid, quantitative method for LDL
subfractionation with use of the Quantimetrix Lipoprint LDL System. Clin Chem 2001;47:266 –74.
29. Cooper GR, Myers GL, Smith SJ, Sampson EJ. Standardization of
lipid, lipoprotein, and apolipoprotein measurements [Review]. Clin
Chem 1988;34:B95–105.
30. Cloey T, Bachorik PS, Becker D, Finney C, Lowry D, Sigmund W.
Reevaluation of serum-plasma differences in total cholesterol
concentration. JAMA 1990;263:2788 –9.
31. Dormans TP, Swinkels DW, de Graaf J, Hendriks JC, Stalenhoef
AF, Demacker PN. Single-spin density-gradient ultracentrifugation
vs gradient gel electrophoresis: two methods for detecting lowdensity-lipoprotein heterogeneity compared. Clin Chem 1991;37:
853– 8.
32. McNamara JR, Campos H, Ordovas JM, Wilson PW, Schaefer EJ.
Gradientgel electrophoresis analysis of low density lipoprotein.
Am Biotechnol Lab 1988;6:28 –33.
33. Hokanson JE, Austin MA, Zambon A, Brunzell JD. Plasma triglyceride and LDL heterogeneity in familial combined hyperlipidemia.
Arterioscler Thromb 1993;13:427–34.